diff options
| author | Linus Torvalds <torvalds@linux-foundation.org> | 2022-10-11 15:02:25 -0700 |
|---|---|---|
| committer | Linus Torvalds <torvalds@linux-foundation.org> | 2022-10-11 15:02:25 -0700 |
| commit | d465bff130bf4ca17b6980abe51164ace1e0cba4 (patch) | |
| tree | cf63ff5514126f777e8c1f727bb832b33d07505e /tools/perf | |
| parent | 041bc24d867a2a577a06534d6d25e500b24a01ef (diff) | |
| parent | d79310700590b8b40d8c867012d6c899ea6fd505 (diff) | |
| download | linux-rpi-d465bff130bf4ca17b6980abe51164ace1e0cba4.tar.gz linux-rpi-d465bff130bf4ca17b6980abe51164ace1e0cba4.tar.bz2 linux-rpi-d465bff130bf4ca17b6980abe51164ace1e0cba4.zip | |
Merge tag 'perf-tools-for-v6.1-1-2022-10-07' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux
Pull perf tools updates from Arnaldo Carvalho de Melo:
- Add support for AMD on 'perf mem' and 'perf c2c', the kernel
enablement patches went via tip.
Example:
$ sudo perf mem record -- -c 10000
^C[ perf record: Woken up 227 times to write data ]
[ perf record: Captured and wrote 58.760 MB perf.data (836978 samples) ]
$ sudo perf mem report -F mem,sample,snoop
Samples: 836K of event 'ibs_op//', Event count (approx.): 8418762
Memory access Samples Snoop
N/A 700620 N/A
L1 hit 126675 N/A
L2 hit 424 N/A
L3 hit 664 HitM
L3 hit 10 N/A
Local RAM hit 2 N/A
Remote RAM (1 hop) hit 8558 N/A
Remote Cache (1 hop) hit 3 N/A
Remote Cache (1 hop) hit 2 HitM
Remote Cache (2 hops) hit 10 HitM
Remote Cache (2 hops) hit 6 N/A
Uncached hit 4 N/A
$
- "perf lock" improvements:
- Add -E/--entries option to limit the number of entries to
display, say to ask for just the top 5 contended locks.
- Add -q/--quiet option to suppress header and debug messages.
- Add a 'perf test' kernel lock contention entry to test 'perf
lock'.
- "perf lock contention" improvements:
- Ask BPF's bpf_get_stackid() to skip some callchain entries.
The ones closer to the tooling are bpf related and not that
interesting, the ones calling the locking function are the ones
we're interested in, example of a full, unskipped callstack:
- Allow changing the callstack depth and number of entries to skip.
1 10.74 us 10.74 us 10.74 us spinlock __bpf_trace_contention_begin+0xb
0xffffffffc03b5c47 bpf_prog_bf07ae9e2cbd02c5_contention_begin+0x117
0xffffffffc03b5c47 bpf_prog_bf07ae9e2cbd02c5_contention_begin+0x117
0xffffffffbb8b8e75 bpf_trace_run2+0x35
0xffffffffbb7eab9b __bpf_trace_contention_begin+0xb
0xffffffffbb7ebe75 queued_spin_lock_slowpath+0x1f5
0xffffffffbc1c26ff _raw_spin_lock+0x1f
0xffffffffbb841015 tick_do_update_jiffies64+0x25
0xffffffffbb8409ee tick_irq_enter+0x9e
- Show full callstack in verbose mode (-v option), sometimes this
is desirable instead of showing just one callstack entry.
- Allow multiple time ranges in 'perf record --delay' to help in
reducing the amount of data collected from hardware tracing (Intel
PT, etc) when there is a rough idea of periods of time where events
of interest take time.
- Add Intel PT to record only decoder debug messages when error
happens.
- Improve layout of Intel PT man page.
- Add new branch types: alignment, data and inst faults and arch
specific ones, such as fiq, debug_halt, debug_exit, debug_inst and
debug_data on arm64.
Kernel enablement went thru the tip tree.
- Fix 'perf probe' error log check in 'perf test' when no debuginfo is
available.
- Fix 'perf stat' aggregation mode logic, it should be looking at the
CPU not at the core number.
- Fix flags parsing in 'perf trace' filters.
- Introduce compact encoding of CPU range encoding on perf.data, to
avoid having a bitmap with all the CPUs.
- Improvements to the 'perf stat' metrics, including adding
"core_wide", and computing "smt" from the CPU topology.
- Add support to the new PERF_FORMAT_LOST perf_event_attr.read_format,
that allows tooling to ask for the precise number of lost samples for
a given event.
- Add 'addr' sort key to see just the address of sampled instructions:
$ perf record -o- true | perf report -i- -s addr
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.000 MB - ]
# Samples: 12 of event 'cycles:u'
# Event count (approx.): 252512
#
# Overhead Address
# ........ ..................
42.96% 0x7f96f08443d7
29.55% 0x7f96f0859b50
14.76% 0x7f96f0852e02
8.30% 0x7f96f0855028
4.43% 0xffffffff8de01087
perf annotate: Toggle full address <-> offset display
- Add 'f' hotkey to the 'perf annotate' TUI interface when in
'disassembler output' mode ('o' hotkey) to toggle showing full
virtual address or just the offset.
- Cache DSO build-ids when synthesizing PERF_RECORD_MMAP records for
pre-existing threads, at the start of a 'perf record' session,
speeding up that record startup phase.
- Add a command line option to specify build ids in 'perf inject'.
- Update JSON event files for the Intel alderlake, broadwell,
broadwellde, broadwellx, cascadelakex, haswell, haswellx, icelake,
icelakex, ivybridge, ivytown, jaketown, sandybridge, sapphirerapids,
skylake, skylakex, and tigerlake processors.
- Update vendor JSON event files for the ARM Neoverse V1 and E1
platforms.
- Add a 'perf test' entry for 'perf mem' where a struct has false
sharing and this gets detected in the 'perf mem' output, tested with
Intel, AMD and ARM64 systems.
- Add a 'perf test' entry to test the resolution of java symbols, where
an output like this is expected:
8.18% jshell jitted-50116-29.so [.] Interpreter
0.75% Thread-1 jitted-83602-1670.so [.] jdk.internal.jimage.BasicImageReader.getString(int)
- Add tests for the ARM64 CoreSight hardware tracing feature, with
specially crafted pureloop, memcpy, thread loop and unroll tread that
then gets traced and the output compared with expected output.
Documentation explaining it is also included.
- Add per thread Intel PT 'perf test' entry to check that
PERF_RECORD_TEXT_POKE events are recorded per CPU, resulting in a
mixture of per thread and per CPU events and mmaps, verify that this
gets all recorded correctly.
- Introduce pthread mutex wrappers to allow for building with clang's
-Wthread-safety, i.e. using the "guarded_by" "pt_guarded_by"
"lockable", "exclusive_lock_function", "exclusive_trylock_function",
"exclusive_locks_required", and "no_thread_safety_analysis" compiler
function attributes.
- Fix empty version number when building outside of a git repo.
- Improve feature detection display when multiple versions of a feature
are present, such as for binutils libbfd, that has a mix of possible
ways to detect according to the Linux distribution.
Previously in some cases we had:
Auto-detecting system features
<SNIP>
... libbfd: [ on ]
... libbfd-liberty: [ on ]
... libbfd-liberty-z: [ on ]
<SNIP>
Now for this case we show just the main feature:
Auto-detecting system features
<SNIP>
... libbfd: [ on ]
<SNIP>
- Remove some unused structs, variables, macros, function prototypes
and includes from various places.
* tag 'perf-tools-for-v6.1-1-2022-10-07' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux: (169 commits)
perf script: Add missing fields in usage hint
perf mem: Print "LFB/MAB" for PERF_MEM_LVLNUM_LFB
perf mem/c2c: Avoid printing empty lines for unsupported events
perf mem/c2c: Add load store event mappings for AMD
perf mem/c2c: Set PERF_SAMPLE_WEIGHT for LOAD_STORE events
perf mem: Add support for printing PERF_MEM_LVLNUM_{CXL|IO}
perf amd ibs: Sync arch/x86/include/asm/amd-ibs.h header with the kernel
tools headers UAPI: Sync include/uapi/linux/perf_event.h header with the kernel
perf stat: Fix cpu check to use id.cpu.cpu in aggr_printout()
perf test coresight: Add relevant documentation about ARM64 CoreSight testing
perf test: Add git ignore for tmp and output files of ARM CoreSight tests
perf test coresight: Add unroll thread test shell script
perf test coresight: Add unroll thread test tool
perf test coresight: Add thread loop test shell scripts
perf test coresight: Add thread loop test tool
perf test coresight: Add memcpy thread test shell script
perf test coresight: Add memcpy thread test tool
perf test: Add git ignore for perf data generated by the ARM CoreSight tests
perf test: Add arm64 asm pureloop test shell script
perf test: Add asm pureloop test tool
...
Diffstat (limited to 'tools/perf')
225 files changed, 15513 insertions, 5245 deletions
diff --git a/tools/perf/.gitignore b/tools/perf/.gitignore index 4b9c71faa01a..a653311d9693 100644 --- a/tools/perf/.gitignore +++ b/tools/perf/.gitignore @@ -15,13 +15,14 @@ perf*.1 perf*.xml perf*.html common-cmds.h -perf.data -perf.data.old +perf*.data +perf*.data.old output.svg perf-archive perf-iostat tags TAGS +stats-*.csv cscope* config.mak config.mak.autogen @@ -29,6 +30,7 @@ config.mak.autogen *-flex.* *.pyc *.pyo +*.stdout .config-detected util/intel-pt-decoder/inat-tables.c arch/*/include/generated/ diff --git a/tools/perf/Documentation/itrace.txt b/tools/perf/Documentation/itrace.txt index 6b189669c450..0916bbfe64cb 100644 --- a/tools/perf/Documentation/itrace.txt +++ b/tools/perf/Documentation/itrace.txt @@ -64,6 +64,7 @@ debug messages will or will not be logged. Each flag must be preceded by either '+' or '-'. The flags are: a all perf events + e output only on errors (size configurable - see linkperf:perf-config[1]) o output to stdout If supported, the 'q' option may be repeated to increase the effect. diff --git a/tools/perf/Documentation/perf-arm-coresight.txt b/tools/perf/Documentation/perf-arm-coresight.txt new file mode 100644 index 000000000000..c117fc50a2a9 --- /dev/null +++ b/tools/perf/Documentation/perf-arm-coresight.txt @@ -0,0 +1,5 @@ +Arm CoreSight Support +===================== + +For full documentation, see Documentation/trace/coresight/coresight-perf.rst +in the kernel tree. diff --git a/tools/perf/Documentation/perf-c2c.txt b/tools/perf/Documentation/perf-c2c.txt index f1f7ae6b08d1..5c5eb2def83e 100644 --- a/tools/perf/Documentation/perf-c2c.txt +++ b/tools/perf/Documentation/perf-c2c.txt @@ -19,9 +19,10 @@ C2C stands for Cache To Cache. The perf c2c tool provides means for Shared Data C2C/HITM analysis. It allows you to track down the cacheline contentions. -On x86, the tool is based on load latency and precise store facility events +On Intel, the tool is based on load latency and precise store facility events provided by Intel CPUs. On PowerPC, the tool uses random instruction sampling -with thresholding feature. +with thresholding feature. On AMD, the tool uses IBS op pmu (due to hardware +limitations, perf c2c is not supported on Zen3 cpus). These events provide: - memory address of the access @@ -49,7 +50,8 @@ RECORD OPTIONS -l:: --ldlat:: - Configure mem-loads latency. (x86 only) + Configure mem-loads latency. Supported on Intel and Arm64 processors + only. Ignored on other archs. -k:: --all-kernel:: @@ -135,11 +137,15 @@ Following perf record options are configured by default: -W,-d,--phys-data,--sample-cpu Unless specified otherwise with '-e' option, following events are monitored by -default on x86: +default on Intel: cpu/mem-loads,ldlat=30/P cpu/mem-stores/P +following on AMD: + + ibs_op// + and following on PowerPC: cpu/mem-loads/ diff --git a/tools/perf/Documentation/perf-config.txt b/tools/perf/Documentation/perf-config.txt index 0420e71698ee..39c890ead2dc 100644 --- a/tools/perf/Documentation/perf-config.txt +++ b/tools/perf/Documentation/perf-config.txt @@ -729,6 +729,13 @@ auxtrace.*:: If the directory does not exist or has the wrong file type, the current directory is used. +itrace.*:: + + debug-log-buffer-size:: + Log size in bytes to output when using the option --itrace=d+e + Refer 'itrace' option of linkperf:perf-script[1] or + linkperf:perf-report[1]. The default is 16384. + daemon.*:: daemon.base:: diff --git a/tools/perf/Documentation/perf-inject.txt b/tools/perf/Documentation/perf-inject.txt index ffc293fdf61d..c972032f4ca0 100644 --- a/tools/perf/Documentation/perf-inject.txt +++ b/tools/perf/Documentation/perf-inject.txt @@ -25,10 +25,17 @@ OPTIONS ------- -b:: --build-ids:: - Inject build-ids into the output stream + Inject build-ids of DSOs hit by samples into the output stream. + This means it needs to process all SAMPLE records to find the DSOs. ---buildid-all: - Inject build-ids of all DSOs into the output stream +--buildid-all:: + Inject build-ids of all DSOs into the output stream regardless of hits + and skip SAMPLE processing. + +--known-build-ids=:: + Override build-ids to inject using these comma-separated pairs of + build-id and path. Understands file://filename to read these pairs + from a file, which can be generated with perf buildid-list. -v:: --verbose:: diff --git a/tools/perf/Documentation/perf-intel-pt.txt b/tools/perf/Documentation/perf-intel-pt.txt index 3dc3f0ccbd51..92464a5d7eaf 100644 --- a/tools/perf/Documentation/perf-intel-pt.txt +++ b/tools/perf/Documentation/perf-intel-pt.txt @@ -943,12 +943,15 @@ event packets are recorded only if the "pwr_evt" config term was used. Refer to the config terms section above. The power events record information about C-state changes, whereas CBR is indicative of CPU frequency. perf script "event,synth" fields display information like this: + cbr: cbr: 22 freq: 2189 MHz (200%) mwait: hints: 0x60 extensions: 0x1 pwre: hw: 0 cstate: 2 sub-cstate: 0 exstop: ip: 1 pwrx: deepest cstate: 2 last cstate: 2 wake reason: 0x4 + Where: + "cbr" includes the frequency and the percentage of maximum non-turbo "mwait" shows mwait hints and extensions "pwre" shows C-state transitions (to a C-state deeper than C0) and @@ -956,6 +959,7 @@ Where: "exstop" indicates execution stopped and whether the IP was recorded exactly, "pwrx" indicates return to C0 + For more details refer to the Intel 64 and IA-32 Architectures Software Developer Manuals. @@ -969,8 +973,10 @@ are quite important. Users must know if what they are seeing is a complete picture or not. The "e" option may be followed by flags which affect what errors will or will not be reported. Each flag must be preceded by either '+' or '-'. The flags supported by Intel PT are: + -o Suppress overflow errors -l Suppress trace data lost errors + For example, for errors but not overflow or data lost errors: --itrace=e-o-l @@ -980,11 +986,16 @@ decoded packets and instructions. Note that this option slows down the decoder and that the resulting file may be very large. The "d" option may be followed by flags which affect what debug messages will or will not be logged. Each flag must be preceded by either '+' or '-'. The flags support by Intel PT are: + -a Suppress logging of perf events +a Log all perf events + +e Output only on decoding errors (size configurable) +o Output to stdout instead of "intel_pt.log" + By default, logged perf events are filtered by any specified time ranges, but -flag +a overrides that. +flag +a overrides that. The +e flag can be useful for analyzing errors. By +default, the log size in that case is 16384 bytes, but can be altered by +linkperf:perf-config[1] e.g. perf config itrace.debug-log-buffer-size=30000 In addition, the period of the "instructions" event can be specified. e.g. diff --git a/tools/perf/Documentation/perf-lock.txt b/tools/perf/Documentation/perf-lock.txt index 193c5d8b8db9..3b1e16563b79 100644 --- a/tools/perf/Documentation/perf-lock.txt +++ b/tools/perf/Documentation/perf-lock.txt @@ -40,6 +40,10 @@ COMMON OPTIONS --verbose:: Be more verbose (show symbol address, etc). +-q:: +--quiet:: + Do not show any message. (Suppress -v) + -D:: --dump-raw-trace:: Dump raw trace in ASCII. @@ -94,6 +98,11 @@ REPORT OPTIONS EventManager_De 1845 1 636 futex-default-S 1609 0 0 +-E:: +--entries=<value>:: + Display this many entries. + + INFO OPTIONS ------------ @@ -105,6 +114,7 @@ INFO OPTIONS --map:: dump map of lock instances (address:name table) + CONTENTION OPTIONS -------------- @@ -148,6 +158,16 @@ CONTENTION OPTIONS --map-nr-entries:: Maximum number of BPF map entries (default: 10240). +--max-stack:: + Maximum stack depth when collecting lock contention (default: 8). + +--stack-skip + Number of stack depth to skip when finding a lock caller (default: 3). + +-E:: +--entries=<value>:: + Display this many entries. + SEE ALSO -------- diff --git a/tools/perf/Documentation/perf-mem.txt b/tools/perf/Documentation/perf-mem.txt index 66177511c5c4..005c95580b1e 100644 --- a/tools/perf/Documentation/perf-mem.txt +++ b/tools/perf/Documentation/perf-mem.txt @@ -85,7 +85,8 @@ RECORD OPTIONS Be more verbose (show counter open errors, etc) --ldlat <n>:: - Specify desired latency for loads event. (x86 only) + Specify desired latency for loads event. Supported on Intel and Arm64 + processors only. Ignored on other archs. In addition, for report all perf report options are valid, and for record all perf record options. diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt index 0228efc96686..e41ae950fdc3 100644 --- a/tools/perf/Documentation/perf-record.txt +++ b/tools/perf/Documentation/perf-record.txt @@ -400,6 +400,7 @@ following filters are defined: For the platforms with Intel Arch LBR support (12th-Gen+ client or 4th-Gen Xeon+ server), the save branch type is unconditionally enabled when the taken branch stack sampling is enabled. + - priv: save privilege state during sampling in case binary is not available later + The option requires at least one branch type among any, any_call, any_ret, ind_call, cond. @@ -410,6 +411,7 @@ is enabled for all the sampling events. The sampled branch type is the same for The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k Note that this feature may not be available on all processors. +-W:: --weight:: Enable weightened sampling. An additional weight is recorded per sample and can be displayed with the weight and local_weight sort keys. This currently works for TSX @@ -433,8 +435,10 @@ if combined with -a or -C options. -D:: --delay=:: After starting the program, wait msecs before measuring (-1: start with events -disabled). This is useful to filter out the startup phase of the program, which -is often very different. +disabled), or enable events only for specified ranges of msecs (e.g. +-D 10-20,30-40 means wait 10 msecs, enable for 10 msecs, wait 10 msecs, enable +for 10 msecs, then stop). Note, delaying enabling of events is useful to filter +out the startup phase of the program, which is often very different. -I:: --intr-regs:: diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt index 24efc0583c93..4533db2ee56b 100644 --- a/tools/perf/Documentation/perf-report.txt +++ b/tools/perf/Documentation/perf-report.txt @@ -73,7 +73,7 @@ OPTIONS Sort histogram entries by given key(s) - multiple keys can be specified in CSV format. Following sort keys are available: pid, comm, dso, symbol, parent, cpu, socket, srcline, weight, - local_weight, cgroup_id. + local_weight, cgroup_id, addr. Each key has following meaning: @@ -114,6 +114,7 @@ OPTIONS - local_ins_lat: Local instruction latency version - p_stage_cyc: On powerpc, this presents the number of cycles spent in a pipeline stage. And currently supported only on powerpc. + - addr: (Full) virtual address of the sampled instruction By default, comm, dso and symbol keys are used. (i.e. --sort comm,dso,symbol) diff --git a/tools/perf/Makefile.config b/tools/perf/Makefile.config index 2171f02daf59..6fd4b1384b97 100644 --- a/tools/perf/Makefile.config +++ b/tools/perf/Makefile.config @@ -19,6 +19,11 @@ detected_var = $(shell echo "$(1)=$($(1))" >> $(OUTPUT).config-detected) CFLAGS := $(EXTRA_CFLAGS) $(filter-out -Wnested-externs,$(EXTRA_WARNINGS)) HOSTCFLAGS := $(filter-out -Wnested-externs,$(EXTRA_WARNINGS)) +# Enabled Wthread-safety analysis for clang builds. +ifeq ($(CC_NO_CLANG), 0) + CFLAGS += -Wthread-safety +endif + include $(srctree)/tools/scripts/Makefile.arch $(call detected_var,SRCARCH) @@ -1291,6 +1296,8 @@ perf_examples_instdir_SQ = $(subst ','\'',$(perf_examples_instdir)) STRACE_GROUPS_INSTDIR_SQ = $(subst ','\'',$(STRACE_GROUPS_INSTDIR)) tip_instdir_SQ = $(subst ','\'',$(tip_instdir)) +export perfexec_instdir_SQ + # If we install to $(HOME) we keep the traceevent default: # $(HOME)/.traceevent/plugins # Otherwise we install plugins into the global $(libdir). @@ -1301,14 +1308,18 @@ endif print_var = $(eval $(print_var_code)) $(info $(MSG)) define print_var_code - MSG = $(shell printf '...%30s: %s' $(1) $($(1))) + MSG = $(shell printf '...%40s: %s' $(1) $($(1))) endef +ifeq ($(feature_display),1) + $(call feature_display_entries) +endif + ifeq ($(VF),1) # Display EXTRA features which are detected manualy # from here with feature_check call and thus cannot # be partof global state output. - $(foreach feat,$(FEATURE_TESTS_EXTRA),$(call feature_print_status,$(feat),)) + $(foreach feat,$(FEATURE_TESTS_EXTRA),$(call feature_print_status,$(feat),) $(info $(MSG))) $(call print_var,prefix) $(call print_var,bindir) $(call print_var,libdir) @@ -1318,11 +1329,12 @@ ifeq ($(VF),1) $(call print_var,JDIR) ifeq ($(dwarf-post-unwind),1) - $(call feature_print_text,"DWARF post unwind library", $(dwarf-post-unwind-text)) + $(call feature_print_text,"DWARF post unwind library", $(dwarf-post-unwind-text)) $(info $(MSG)) endif - $(info ) endif +$(info ) + $(call detected_var,bindir_SQ) $(call detected_var,PYTHON_WORD) ifneq ($(OUTPUT),) @@ -1352,7 +1364,3 @@ endif # tests. $(shell rm -f $(FEATURE_DUMP_FILENAME)) $(foreach feat,$(FEATURE_TESTS),$(shell echo "$(call feature_assign,$(feat))" >> $(FEATURE_DUMP_FILENAME))) - -ifeq ($(feature_display),1) - $(call feature_display_entries) -endif diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf index bd947885a639..a432e59afc42 100644 --- a/tools/perf/Makefile.perf +++ b/tools/perf/Makefile.perf @@ -629,7 +629,16 @@ sync_file_range_tbls := $(srctree)/tools/perf/trace/beauty/sync_file_range.sh $(sync_file_range_arrays): $(linux_uapi_dir)/fs.h $(sync_file_range_tbls) $(Q)$(SHELL) '$(sync_file_range_tbls)' $(linux_uapi_dir) > $@ -all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS) +TESTS_CORESIGHT_DIR := $(srctree)/tools/perf/tests/shell/coresight + +tests-coresight-targets: FORCE + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR) + +tests-coresight-targets-clean: + $(call QUIET_CLEAN, coresight) + $(Q)$(MAKE) -C $(TESTS_CORESIGHT_DIR) O=$(OUTPUT) clean >/dev/null + +all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS) tests-coresight-targets # Create python binding output directory if not already present _dummy := $(shell [ -d '$(OUTPUT)python' ] || mkdir -p '$(OUTPUT)python') @@ -1006,7 +1015,10 @@ install-tests: all install-gtk $(INSTALL) tests/shell/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell'; \ $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'; \ $(INSTALL) tests/shell/lib/*.sh -m 644 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'; \ - $(INSTALL) tests/shell/lib/*.py -m 644 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib' + $(INSTALL) tests/shell/lib/*.py -m 644 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/lib'; \ + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/coresight' ; \ + $(INSTALL) tests/shell/coresight/*.sh '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/tests/shell/coresight' + $(Q)$(MAKE) -C tests/shell/coresight install-tests install-bin: install-tools install-tests install-traceevent-plugins @@ -1077,7 +1089,7 @@ endif # BUILD_BPF_SKEL bpf-skel-clean: $(call QUIET_CLEAN, bpf-skel) $(RM) -r $(SKEL_TMP_OUT) $(SKELETONS) -clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean +clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean tests-coresight-targets-clean $(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-iostat $(LANG_BINDINGS) $(Q)find $(or $(OUTPUT),.) -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete $(Q)$(RM) $(OUTPUT).config-detected diff --git a/tools/perf/arch/x86/util/intel-pt.c b/tools/perf/arch/x86/util/intel-pt.c index 13933020a79e..793b35f2221a 100644 --- a/tools/perf/arch/x86/util/intel-pt.c +++ b/tools/perf/arch/x86/util/intel-pt.c @@ -11,6 +11,7 @@ #include <linux/bitops.h> #include <linux/log2.h> #include <linux/zalloc.h> +#include <linux/err.h> #include <cpuid.h> #include "../../../util/session.h" @@ -426,20 +427,14 @@ static int intel_pt_track_switches(struct evlist *evlist) if (!evlist__can_select_event(evlist, sched_switch)) return -EPERM; - err = parse_event(evlist, sched_switch); - if (err) { - pr_debug2("%s: failed to parse %s, error %d\n", + evsel = evlist__add_sched_switch(evlist, true); + if (IS_ERR(evsel)) { + err = PTR_ERR(evsel); + pr_debug2("%s: failed to create %s, error = %d\n", __func__, sched_switch, err); return err; } - evsel = evlist__last(evlist); - - evsel__set_sample_bit(evsel, CPU); - evsel__set_sample_bit(evsel, TIME); - - evsel->core.system_wide = true; - evsel->no_aux_samples = true; evsel->immediate = true; return 0; diff --git a/tools/perf/arch/x86/util/mem-events.c b/tools/perf/arch/x86/util/mem-events.c index 5214370ca4e4..f683ac702247 100644 --- a/tools/perf/arch/x86/util/mem-events.c +++ b/tools/perf/arch/x86/util/mem-events.c @@ -1,7 +1,9 @@ // SPDX-License-Identifier: GPL-2.0 #include "util/pmu.h" +#include "util/env.h" #include "map_symbol.h" #include "mem-events.h" +#include "linux/string.h" static char mem_loads_name[100]; static bool mem_loads_name__init; @@ -12,18 +14,43 @@ static char mem_stores_name[100]; #define E(t, n, s) { .tag = t, .name = n, .sysfs_name = s } -static struct perf_mem_event perf_mem_events[PERF_MEM_EVENTS__MAX] = { +static struct perf_mem_event perf_mem_events_intel[PERF_MEM_EVENTS__MAX] = { E("ldlat-loads", "%s/mem-loads,ldlat=%u/P", "%s/events/mem-loads"), E("ldlat-stores", "%s/mem-stores/P", "%s/events/mem-stores"), E(NULL, NULL, NULL), }; +static struct perf_mem_event perf_mem_events_amd[PERF_MEM_EVENTS__MAX] = { + E(NULL, NULL, NULL), + E(NULL, NULL, NULL), + E("mem-ldst", "ibs_op//", "ibs_op"), +}; + +static int perf_mem_is_amd_cpu(void) +{ + struct perf_env env = { .total_mem = 0, }; + + perf_env__cpuid(&env); + if (env.cpuid && strstarts(env.cpuid, "AuthenticAMD")) + return 1; + return -1; +} + struct perf_mem_event *perf_mem_events__ptr(int i) { + /* 0: Uninitialized, 1: Yes, -1: No */ + static int is_amd; + if (i >= PERF_MEM_EVENTS__MAX) return NULL; - return &perf_mem_events[i]; + if (!is_amd) + is_amd = perf_mem_is_amd_cpu(); + + if (is_amd == 1) + return &perf_mem_events_amd[i]; + + return &perf_mem_events_intel[i]; } bool is_mem_loads_aux_event(struct evsel *leader) diff --git a/tools/perf/bench/epoll-ctl.c b/tools/perf/bench/epoll-ctl.c index 4256dc5d6236..521d1ff97b06 100644 --- a/tools/perf/bench/epoll-ctl.c +++ b/tools/perf/bench/epoll-ctl.c @@ -23,6 +23,7 @@ #include <sys/eventfd.h> #include <perf/cpumap.h> +#include "../util/mutex.h" #include "../util/stat.h" #include <subcmd/parse-options.h> #include "bench.h" @@ -58,10 +59,10 @@ static unsigned int nested = 0; /* amount of fds to monitor, per thread */ static unsigned int nfds = 64; -static pthread_mutex_t thread_lock; +static struct mutex thread_lock; static unsigned int threads_starting; static struct stats all_stats[EPOLL_NR_OPS]; -static pthread_cond_t thread_parent, thread_worker; +static struct cond thread_parent, thread_worker; struct worker { int tid; @@ -174,12 +175,12 @@ static void *workerfn(void *arg) struct timespec ts = { .tv_sec = 0, .tv_nsec = 250 }; - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); threads_starting--; if (!threads_starting) - pthread_cond_signal(&thread_parent); - pthread_cond_wait(&thread_worker, &thread_lock); - pthread_mutex_unlock(&thread_lock); + cond_signal(&thread_parent); + cond_wait(&thread_worker, &thread_lock); + mutex_unlock(&thread_lock); /* Let 'em loose */ do { @@ -367,9 +368,9 @@ int bench_epoll_ctl(int argc, const char **argv) for (i = 0; i < EPOLL_NR_OPS; i++) init_stats(&all_stats[i]); - pthread_mutex_init(&thread_lock, NULL); - pthread_cond_init(&thread_parent, NULL); - pthread_cond_init(&thread_worker, NULL); + mutex_init(&thread_lock); + cond_init(&thread_parent); + cond_init(&thread_worker); threads_starting = nthreads; @@ -377,11 +378,11 @@ int bench_epoll_ctl(int argc, const char **argv) do_threads(worker, cpu); - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); while (threads_starting) - pthread_cond_wait(&thread_parent, &thread_lock); - pthread_cond_broadcast(&thread_worker); - pthread_mutex_unlock(&thread_lock); + cond_wait(&thread_parent, &thread_lock); + cond_broadcast(&thread_worker); + mutex_unlock(&thread_lock); sleep(nsecs); toggle_done(0, NULL, NULL); @@ -394,9 +395,9 @@ int bench_epoll_ctl(int argc, const char **argv) } /* cleanup & report results */ - pthread_cond_destroy(&thread_parent); - pthread_cond_destroy(&thread_worker); - pthread_mutex_destroy(&thread_lock); + cond_destroy(&thread_parent); + cond_destroy(&thread_worker); + mutex_destroy(&thread_lock); for (i = 0; i < nthreads; i++) { unsigned long t[EPOLL_NR_OPS]; diff --git a/tools/perf/bench/epoll-wait.c b/tools/perf/bench/epoll-wait.c index 2728b0140853..c1cdf03c075d 100644 --- a/tools/perf/bench/epoll-wait.c +++ b/tools/perf/bench/epoll-wait.c @@ -79,6 +79,7 @@ #include <perf/cpumap.h> #include "../util/stat.h" +#include "../util/mutex.h" #include <subcmd/parse-options.h> #include "bench.h" @@ -109,10 +110,10 @@ static bool multiq; /* use an epoll instance per thread */ /* amount of fds to monitor, per thread */ static unsigned int nfds = 64; -static pthread_mutex_t thread_lock; +static struct mutex thread_lock; static unsigned int threads_starting; static struct stats throughput_stats; -static pthread_cond_t thread_parent, thread_worker; +static struct cond thread_parent, thread_worker; struct worker { int tid; @@ -189,12 +190,12 @@ static void *workerfn(void *arg) int to = nonblocking? 0 : -1; int efd = multiq ? w->epollfd : epollfd; - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); threads_starting--; if (!threads_starting) - pthread_cond_signal(&thread_parent); - pthread_cond_wait(&thread_worker, &thread_lock); - pthread_mutex_unlock(&thread_lock); + cond_signal(&thread_parent); + cond_wait(&thread_worker, &thread_lock); + mutex_unlock(&thread_lock); do { /* @@ -485,9 +486,9 @@ int bench_epoll_wait(int argc, const char **argv) getpid(), nthreads, oneshot ? " (EPOLLONESHOT semantics)": "", nfds, nsecs); init_stats(&throughput_stats); - pthread_mutex_init(&thread_lock, NULL); - pthread_cond_init(&thread_parent, NULL); - pthread_cond_init(&thread_worker, NULL); + mutex_init(&thread_lock); + cond_init(&thread_parent); + cond_init(&thread_worker); threads_starting = nthreads; @@ -495,11 +496,11 @@ int bench_epoll_wait(int argc, const char **argv) do_threads(worker, cpu); - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); while (threads_starting) - pthread_cond_wait(&thread_parent, &thread_lock); - pthread_cond_broadcast(&thread_worker); - pthread_mutex_unlock(&thread_lock); + cond_wait(&thread_parent, &thread_lock); + cond_broadcast(&thread_worker); + mutex_unlock(&thread_lock); /* * At this point the workers should be blocked waiting for read events @@ -522,9 +523,9 @@ int bench_epoll_wait(int argc, const char **argv) err(EXIT_FAILURE, "pthread_join"); /* cleanup & report results */ - pthread_cond_destroy(&thread_parent); - pthread_cond_destroy(&thread_worker); - pthread_mutex_destroy(&thread_lock); + cond_destroy(&thread_parent); + cond_destroy(&thread_worker); + mutex_destroy(&thread_lock); /* sort the array back before reporting */ if (randomize) diff --git a/tools/perf/bench/futex-hash.c b/tools/perf/bench/futex-hash.c index f05db4cf983d..2005a3fa3026 100644 --- a/tools/perf/bench/futex-hash.c +++ b/tools/perf/bench/futex-hash.c @@ -23,6 +23,7 @@ #include <sys/mman.h> #include <perf/cpumap.h> +#include "../util/mutex.h" #include "../util/stat.h" #include <subcmd/parse-options.h> #include "bench.h" @@ -34,10 +35,10 @@ static bool done = false; static int futex_flag = 0; struct timeval bench__start, bench__end, bench__runtime; -static pthread_mutex_t thread_lock; +static struct mutex thread_lock; static unsigned int threads_starting; static struct stats throughput_stats; -static pthread_cond_t thread_parent, thread_worker; +static struct cond thread_parent, thread_worker; struct worker { int tid; @@ -73,12 +74,12 @@ static void *workerfn(void *arg) unsigned int i; unsigned long ops = w->ops; /* avoid cacheline bouncing */ - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); threads_starting--; if (!threads_starting) - pthread_cond_signal(&thread_parent); - pthread_cond_wait(&thread_worker, &thread_lock); - pthread_mutex_unlock(&thread_lock); + cond_signal(&thread_parent); + cond_wait(&thread_worker, &thread_lock); + mutex_unlock(&thread_lock); do { for (i = 0; i < params.nfutexes; i++, ops++) { @@ -165,9 +166,9 @@ int bench_futex_hash(int argc, const char **argv) getpid(), params.nthreads, params.nfutexes, params.fshared ? "shared":"private", params.runtime); init_stats(&throughput_stats); - pthread_mutex_init(&thread_lock, NULL); - pthread_cond_init(&thread_parent, NULL); - pthread_cond_init(&thread_worker, NULL); + mutex_init(&thread_lock); + cond_init(&thread_parent); + cond_init(&thread_worker); threads_starting = params.nthreads; pthread_attr_init(&thread_attr); @@ -203,11 +204,11 @@ int bench_futex_hash(int argc, const char **argv) CPU_FREE(cpuset); pthread_attr_destroy(&thread_attr); - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); while (threads_starting) - pthread_cond_wait(&thread_parent, &thread_lock); - pthread_cond_broadcast(&thread_worker); - pthread_mutex_unlock(&thread_lock); + cond_wait(&thread_parent, &thread_lock); + cond_broadcast(&thread_worker); + mutex_unlock(&thread_lock); sleep(params.runtime); toggle_done(0, NULL, NULL); @@ -219,9 +220,9 @@ int bench_futex_hash(int argc, const char **argv) } /* cleanup & report results */ - pthread_cond_destroy(&thread_parent); - pthread_cond_destroy(&thread_worker); - pthread_mutex_destroy(&thread_lock); + cond_destroy(&thread_parent); + cond_destroy(&thread_worker); + mutex_destroy(&thread_lock); for (i = 0; i < params.nthreads; i++) { unsigned long t = bench__runtime.tv_sec > 0 ? diff --git a/tools/perf/bench/futex-lock-pi.c b/tools/perf/bench/futex-lock-pi.c index 0abb3f7ee24f..2d0417949727 100644 --- a/tools/perf/bench/futex-lock-pi.c +++ b/tools/perf/bench/futex-lock-pi.c @@ -8,6 +8,7 @@ #include <pthread.h> #include <signal.h> +#include "../util/mutex.h" #include "../util/stat.h" #include <subcmd/parse-options.h> #include <linux/compiler.h> @@ -34,10 +35,10 @@ static u_int32_t global_futex = 0; static struct worker *worker; static bool done = false; static int futex_flag = 0; -static pthread_mutex_t thread_lock; +static struct mutex thread_lock; static unsigned int threads_starting; static struct stats throughput_stats; -static pthread_cond_t thread_parent, thread_worker; +static struct cond thread_parent, thread_worker; static struct bench_futex_parameters params = { .runtime = 10, @@ -83,12 +84,12 @@ static void *workerfn(void *arg) struct worker *w = (struct worker *) arg; unsigned long ops = w->ops; - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); threads_starting--; if (!threads_starting) - pthread_cond_signal(&thread_parent); - pthread_cond_wait(&thread_worker, &thread_lock); - pthread_mutex_unlock(&thread_lock); + cond_signal(&thread_parent); + cond_wait(&thread_worker, &thread_lock); + mutex_unlock(&thread_lock); do { int ret; @@ -197,9 +198,9 @@ int bench_futex_lock_pi(int argc, const char **argv) getpid(), params.nthreads, params.runtime); init_stats(&throughput_stats); - pthread_mutex_init(&thread_lock, NULL); - pthread_cond_init(&thread_parent, NULL); - pthread_cond_init(&thread_worker, NULL); + mutex_init(&thread_lock); + cond_init(&thread_parent); + cond_init(&thread_worker); threads_starting = params.nthreads; pthread_attr_init(&thread_attr); @@ -208,11 +209,11 @@ int bench_futex_lock_pi(int argc, const char **argv) create_threads(worker, thread_attr, cpu); pthread_attr_destroy(&thread_attr); - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); while (threads_starting) - pthread_cond_wait(&thread_parent, &thread_lock); - pthread_cond_broadcast(&thread_worker); - pthread_mutex_unlock(&thread_lock); + cond_wait(&thread_parent, &thread_lock); + cond_broadcast(&thread_worker); + mutex_unlock(&thread_lock); sleep(params.runtime); toggle_done(0, NULL, NULL); @@ -224,9 +225,9 @@ int bench_futex_lock_pi(int argc, const char **argv) } /* cleanup & report results */ - pthread_cond_destroy(&thread_parent); - pthread_cond_destroy(&thread_worker); - pthread_mutex_destroy(&thread_lock); + cond_destroy(&thread_parent); + cond_destroy(&thread_worker); + mutex_destroy(&thread_lock); for (i = 0; i < params.nthreads; i++) { unsigned long t = bench__runtime.tv_sec > 0 ? diff --git a/tools/perf/bench/futex-requeue.c b/tools/perf/bench/futex-requeue.c index b6faabfafb8e..69ad896f556c 100644 --- a/tools/perf/bench/futex-requeue.c +++ b/tools/perf/bench/futex-requeue.c @@ -15,6 +15,7 @@ #include <pthread.h> #include <signal.h> +#include "../util/mutex.h" #include "../util/stat.h" #include <subcmd/parse-options.h> #include <linux/compiler.h> @@ -34,8 +35,8 @@ static u_int32_t futex1 = 0, futex2 = 0; static pthread_t *worker; static bool done = false; -static pthread_mutex_t thread_lock; -static pthread_cond_t thread_parent, thread_worker; +static struct mutex thread_lock; +static struct cond thread_parent, thread_worker; static struct stats requeuetime_stats, requeued_stats; static unsigned int threads_starting; static int futex_flag = 0; @@ -82,12 +83,12 @@ static void *workerfn(void *arg __maybe_unused) { int ret; - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); threads_starting--; if (!threads_starting) - pthread_cond_signal(&thread_parent); - pthread_cond_wait(&thread_worker, &thread_lock); - pthread_mutex_unlock(&thread_lock); + cond_signal(&thread_parent); + cond_wait(&thread_worker, &thread_lock); + mutex_unlock(&thread_lock); while (1) { if (!params.pi) { @@ -209,9 +210,9 @@ int bench_futex_requeue(int argc, const char **argv) init_stats(&requeued_stats); init_stats(&requeuetime_stats); pthread_attr_init(&thread_attr); - pthread_mutex_init(&thread_lock, NULL); - pthread_cond_init(&thread_parent, NULL); - pthread_cond_init(&thread_worker, NULL); + mutex_init(&thread_lock); + cond_init(&thread_parent); + cond_init(&thread_worker); for (j = 0; j < bench_repeat && !done; j++) { unsigned int nrequeued = 0, wakeups = 0; @@ -221,11 +222,11 @@ int bench_futex_requeue(int argc, const char **argv) block_threads(worker, thread_attr, cpu); /* make sure all threads are already blocked */ - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); while (threads_starting) - pthread_cond_wait(&thread_parent, &thread_lock); - pthread_cond_broadcast(&thread_worker); - pthread_mutex_unlock(&thread_lock); + cond_wait(&thread_parent, &thread_lock); + cond_broadcast(&thread_worker); + mutex_unlock(&thread_lock); usleep(100000); @@ -297,9 +298,9 @@ int bench_futex_requeue(int argc, const char **argv) } /* cleanup & report results */ - pthread_cond_destroy(&thread_parent); - pthread_cond_destroy(&thread_worker); - pthread_mutex_destroy(&thread_lock); + cond_destroy(&thread_parent); + cond_destroy(&thread_worker); + mutex_destroy(&thread_lock); pthread_attr_destroy(&thread_attr); print_summary(); diff --git a/tools/perf/bench/futex-wake-parallel.c b/tools/perf/bench/futex-wake-parallel.c index e47f46a3a47e..6682e49d0ee0 100644 --- a/tools/perf/bench/futex-wake-parallel.c +++ b/tools/perf/bench/futex-wake-parallel.c @@ -10,6 +10,7 @@ #include "bench.h" #include <linux/compiler.h> #include "../util/debug.h" +#include "../util/mutex.h" #ifndef HAVE_PTHREAD_BARRIER int bench_futex_wake_parallel(int argc __maybe_unused, const char **argv __maybe_unused) @@ -49,8 +50,8 @@ static u_int32_t futex = 0; static pthread_t *blocked_worker; static bool done = false; -static pthread_mutex_t thread_lock; -static pthread_cond_t thread_parent, thread_worker; +static struct mutex thread_lock; +static struct cond thread_parent, thread_worker; static pthread_barrier_t barrier; static struct stats waketime_stats, wakeup_stats; static unsigned int threads_starting; @@ -125,12 +126,12 @@ static void wakeup_threads(struct thread_data *td, pthread_attr_t thread_attr) static void *blocked_workerfn(void *arg __maybe_unused) { - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); threads_starting--; if (!threads_starting) - pthread_cond_signal(&thread_parent); - pthread_cond_wait(&thread_worker, &thread_lock); - pthread_mutex_unlock(&thread_lock); + cond_signal(&thread_parent); + cond_wait(&thread_worker, &thread_lock); + mutex_unlock(&thread_lock); while (1) { /* handle spurious wakeups */ if (futex_wait(&futex, 0, NULL, futex_flag) != EINTR) @@ -294,9 +295,9 @@ int bench_futex_wake_parallel(int argc, const char **argv) init_stats(&waketime_stats); pthread_attr_init(&thread_attr); - pthread_mutex_init(&thread_lock, NULL); - pthread_cond_init(&thread_parent, NULL); - pthread_cond_init(&thread_worker, NULL); + mutex_init(&thread_lock); + cond_init(&thread_parent); + cond_init(&thread_worker); for (j = 0; j < bench_repeat && !done; j++) { waking_worker = calloc(params.nwakes, sizeof(*waking_worker)); @@ -307,11 +308,11 @@ int bench_futex_wake_parallel(int argc, const char **argv) block_threads(blocked_worker, thread_attr, cpu); /* make sure all threads are already blocked */ - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); while (threads_starting) - pthread_cond_wait(&thread_parent, &thread_lock); - pthread_cond_broadcast(&thread_worker); - pthread_mutex_unlock(&thread_lock); + cond_wait(&thread_parent, &thread_lock); + cond_broadcast(&thread_worker); + mutex_unlock(&thread_lock); usleep(100000); @@ -332,9 +333,9 @@ int bench_futex_wake_parallel(int argc, const char **argv) } /* cleanup & report results */ - pthread_cond_destroy(&thread_parent); - pthread_cond_destroy(&thread_worker); - pthread_mutex_destroy(&thread_lock); + cond_destroy(&thread_parent); + cond_destroy(&thread_worker); + mutex_destroy(&thread_lock); pthread_attr_destroy(&thread_attr); print_summary(); diff --git a/tools/perf/bench/futex-wake.c b/tools/perf/bench/futex-wake.c index 201a3555f09a..9ecab6620a87 100644 --- a/tools/perf/bench/futex-wake.c +++ b/tools/perf/bench/futex-wake.c @@ -14,6 +14,7 @@ #include <pthread.h> #include <signal.h> +#include "../util/mutex.h" #include "../util/stat.h" #include <subcmd/parse-options.h> #include <linux/compiler.h> @@ -34,8 +35,8 @@ static u_int32_t futex1 = 0; static pthread_t *worker; static bool done = false; -static pthread_mutex_t thread_lock; -static pthread_cond_t thread_parent, thread_worker; +static struct mutex thread_lock; +static struct cond thread_parent, thread_worker; static struct stats waketime_stats, wakeup_stats; static unsigned int threads_starting; static int futex_flag = 0; @@ -65,12 +66,12 @@ static const char * const bench_futex_wake_usage[] = { static void *workerfn(void *arg __maybe_unused) { - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); threads_starting--; if (!threads_starting) - pthread_cond_signal(&thread_parent); - pthread_cond_wait(&thread_worker, &thread_lock); - pthread_mutex_unlock(&thread_lock); + cond_signal(&thread_parent); + cond_wait(&thread_worker, &thread_lock); + mutex_unlock(&thread_lock); while (1) { if (futex_wait(&futex1, 0, NULL, futex_flag) != EINTR) @@ -178,9 +179,9 @@ int bench_futex_wake(int argc, const char **argv) init_stats(&wakeup_stats); init_stats(&waketime_stats); pthread_attr_init(&thread_attr); - pthread_mutex_init(&thread_lock, NULL); - pthread_cond_init(&thread_parent, NULL); - pthread_cond_init(&thread_worker, NULL); + mutex_init(&thread_lock); + cond_init(&thread_parent); + cond_init(&thread_worker); for (j = 0; j < bench_repeat && !done; j++) { unsigned int nwoken = 0; @@ -190,11 +191,11 @@ int bench_futex_wake(int argc, const char **argv) block_threads(worker, thread_attr, cpu); /* make sure all threads are already blocked */ - pthread_mutex_lock(&thread_lock); + mutex_lock(&thread_lock); while (threads_starting) - pthread_cond_wait(&thread_parent, &thread_lock); - pthread_cond_broadcast(&thread_worker); - pthread_mutex_unlock(&thread_lock); + cond_wait(&thread_parent, &thread_lock); + cond_broadcast(&thread_worker); + mutex_unlock(&thread_lock); usleep(100000); @@ -224,9 +225,9 @@ int bench_futex_wake(int argc, const char **argv) } /* cleanup & report results */ - pthread_cond_destroy(&thread_parent); - pthread_cond_destroy(&thread_worker); - pthread_mutex_destroy(&thread_lock); + cond_destroy(&thread_parent); + cond_destroy(&thread_worker); + mutex_destroy(&thread_lock); pthread_attr_destroy(&thread_attr); print_summary(); diff --git a/tools/perf/bench/numa.c b/tools/perf/bench/numa.c index 20eed1e53f80..e78dedf9e682 100644 --- a/tools/perf/bench/numa.c +++ b/tools/perf/bench/numa.c @@ -6,8 +6,6 @@ */ #include <inttypes.h> -/* For the CLR_() macros */ -#include <pthread.h> #include <subcmd/parse-options.h> #include "../util/cloexec.h" @@ -35,6 +33,7 @@ #include <linux/zalloc.h> #include "../util/header.h" +#include "../util/mutex.h" #include <numa.h> #include <numaif.h> @@ -67,7 +66,7 @@ struct thread_data { u64 system_time_ns; u64 user_time_ns; double speed_gbs; - pthread_mutex_t *process_lock; + struct mutex *process_lock; }; /* Parameters set by options: */ @@ -137,16 +136,16 @@ struct params { struct global_info { u8 *data; - pthread_mutex_t startup_mutex; - pthread_cond_t startup_cond; + struct mutex startup_mutex; + struct cond startup_cond; int nr_tasks_started; - pthread_mutex_t start_work_mutex; - pthread_cond_t start_work_cond; + struct mutex start_work_mutex; + struct cond start_work_cond; int nr_tasks_working; bool start_work; - pthread_mutex_t stop_work_mutex; + struct mutex stop_work_mutex; u64 bytes_done; struct thread_data *threads; @@ -524,30 +523,6 @@ static void * setup_private_data(ssize_t bytes) return alloc_data(bytes, MAP_PRIVATE, 0, g->p.init_cpu0, g->p.thp, g->p.init_random); } -/* - * Return a process-shared (global) mutex: - */ -static void init_global_mutex(pthread_mutex_t *mutex) -{ - pthread_mutexattr_t attr; - - pthread_mutexattr_init(&attr); - pthread_mutexattr_setpshared(&attr, PTHREAD_PROCESS_SHARED); - pthread_mutex_init(mutex, &attr); -} - -/* - * Return a process-shared (global) condition variable: - */ -static void init_global_cond(pthread_cond_t *cond) -{ - pthread_condattr_t attr; - - pthread_condattr_init(&attr); - pthread_condattr_setpshared(&attr, PTHREAD_PROCESS_SHARED); - pthread_cond_init(cond, &attr); -} - static int parse_cpu_list(const char *arg) { p0.cpu_list_str = strdup(arg); @@ -1220,22 +1195,22 @@ static void *worker_thread(void *__tdata) } if (g->p.serialize_startup) { - pthread_mutex_lock(&g->startup_mutex); + mutex_lock(&g->startup_mutex); g->nr_tasks_started++; /* The last thread wakes the main process. */ if (g->nr_tasks_started == g->p.nr_tasks) - pthread_cond_signal(&g->startup_cond); + cond_signal(&g->startup_cond); - pthread_mutex_unlock(&g->startup_mutex); + mutex_unlock(&g->startup_mutex); /* Here we will wait for the main process to start us all at once: */ - pthread_mutex_lock(&g->start_work_mutex); + mutex_lock(&g->start_work_mutex); g->start_work = false; g->nr_tasks_working++; while (!g->start_work) - pthread_cond_wait(&g->start_work_cond, &g->start_work_mutex); + cond_wait(&g->start_work_cond, &g->start_work_mutex); - pthread_mutex_unlock(&g->start_work_mutex); + mutex_unlock(&g->start_work_mutex); } gettimeofday(&start0, NULL); @@ -1254,17 +1229,17 @@ static void *worker_thread(void *__tdata) val += do_work(thread_data, g->p.bytes_thread, 0, 1, l, val); if (g->p.sleep_usecs) { - pthread_mutex_lock(td->process_lock); + mutex_lock(td->process_lock); usleep(g->p.sleep_usecs); - pthread_mutex_unlock(td->process_lock); + mutex_unlock(td->process_lock); } /* * Amount of work to be done under a process-global lock: */ if (g->p.bytes_process_locked) { - pthread_mutex_lock(td->process_lock); + mutex_lock(td->process_lock); val += do_work(process_data, g->p.bytes_process_locked, thread_nr, g->p.nr_threads, l, val); - pthread_mutex_unlock(td->process_lock); + mutex_unlock(td->process_lock); } work_done = g->p.bytes_global + g->p.bytes_process + @@ -1361,9 +1336,9 @@ static void *worker_thread(void *__tdata) free_data(thread_data, g->p.bytes_thread); - pthread_mutex_lock(&g->stop_work_mutex); + mutex_lock(&g->stop_work_mutex); g->bytes_done += bytes_done; - pthread_mutex_unlock(&g->stop_work_mutex); + mutex_unlock(&g->stop_work_mutex); return NULL; } @@ -1373,7 +1348,7 @@ static void *worker_thread(void *__tdata) */ static void worker_process(int process_nr) { - pthread_mutex_t process_lock; + struct mutex process_lock; struct thread_data *td; pthread_t *pthreads; u8 *process_data; @@ -1381,7 +1356,7 @@ static void worker_process(int process_nr) int ret; int t; - pthread_mutex_init(&process_lock, NULL); + mutex_init(&process_lock); set_taskname("process %d", process_nr); /* @@ -1540,11 +1515,11 @@ static int init(void) g->data = setup_shared_data(g->p.bytes_global); /* Startup serialization: */ - init_global_mutex(&g->start_work_mutex); - init_global_cond(&g->start_work_cond); - init_global_mutex(&g->startup_mutex); - init_global_cond(&g->startup_cond); - init_global_mutex(&g->stop_work_mutex); + mutex_init_pshared(&g->start_work_mutex); + cond_init_pshared(&g->start_work_cond); + mutex_init_pshared(&g->startup_mutex); + cond_init_pshared(&g->startup_cond); + mutex_init_pshared(&g->stop_work_mutex); init_thread_data(); @@ -1633,17 +1608,17 @@ static int __bench_numa(const char *name) * Wait for all the threads to start up. The last thread will * signal this process. */ - pthread_mutex_lock(&g->startup_mutex); + mutex_lock(&g->startup_mutex); while (g->nr_tasks_started != g->p.nr_tasks) - pthread_cond_wait(&g->startup_cond, &g->startup_mutex); + cond_wait(&g->startup_cond, &g->startup_mutex); - pthread_mutex_unlock(&g->startup_mutex); + mutex_unlock(&g->startup_mutex); /* Wait for all threads to be at the start_work_cond. */ while (!threads_ready) { - pthread_mutex_lock(&g->start_work_mutex); + mutex_lock(&g->start_work_mutex); threads_ready = (g->nr_tasks_working == g->p.nr_tasks); - pthread_mutex_unlock(&g->start_work_mutex); + mutex_unlock(&g->start_work_mutex); if (!threads_ready) usleep(1); } @@ -1661,10 +1636,10 @@ static int __bench_numa(const char *name) start = stop; /* Start all threads running. */ - pthread_mutex_lock(&g->start_work_mutex); + mutex_lock(&g->start_work_mutex); g->start_work = true; - pthread_mutex_unlock(&g->start_work_mutex); - pthread_cond_broadcast(&g->start_work_cond); + mutex_unlock(&g->start_work_mutex); + cond_broadcast(&g->start_work_cond); } else { gettimeofday(&start, NULL); } diff --git a/tools/perf/builtin-c2c.c b/tools/perf/builtin-c2c.c index 438fc222e213..a9190458d2d5 100644 --- a/tools/perf/builtin-c2c.c +++ b/tools/perf/builtin-c2c.c @@ -679,28 +679,35 @@ STAT_FN(ld_l2hit) STAT_FN(ld_llchit) STAT_FN(rmt_hit) -static uint64_t total_records(struct c2c_stats *stats) +static uint64_t get_load_llc_misses(struct c2c_stats *stats) { - uint64_t lclmiss, ldcnt, total; - - lclmiss = stats->lcl_dram + - stats->rmt_dram + - stats->rmt_hitm + - stats->rmt_hit; + return stats->lcl_dram + + stats->rmt_dram + + stats->rmt_hitm + + stats->rmt_hit; +} - ldcnt = lclmiss + - stats->ld_fbhit + - stats->ld_l1hit + - stats->ld_l2hit + - stats->ld_llchit + - stats->lcl_hitm; +static uint64_t get_load_cache_hits(struct c2c_stats *stats) +{ + return stats->ld_fbhit + + stats->ld_l1hit + + stats->ld_l2hit + + stats->ld_llchit + + stats->lcl_hitm; +} - total = ldcnt + - stats->st_l1hit + - stats->st_l1miss + - stats->st_na; +static uint64_t get_stores(struct c2c_stats *stats) +{ + return stats->st_l1hit + + stats->st_l1miss + + stats->st_na; +} - return total; +static uint64_t total_records(struct c2c_stats *stats) +{ + return get_load_llc_misses(stats) + + get_load_cache_hits(stats) + + get_stores(stats); } static int @@ -737,21 +744,8 @@ tot_recs_cmp(struct perf_hpp_fmt *fmt __maybe_unused, static uint64_t total_loads(struct c2c_stats *stats) { - uint64_t lclmiss, ldcnt; - - lclmiss = stats->lcl_dram + - stats->rmt_dram + - stats->rmt_hitm + - stats->rmt_hit; - - ldcnt = lclmiss + - stats->ld_fbhit + - stats->ld_l1hit + - stats->ld_l2hit + - stats->ld_llchit + - stats->lcl_hitm; - - return ldcnt; + return get_load_llc_misses(stats) + + get_load_cache_hits(stats); } static int @@ -2376,10 +2370,7 @@ static void print_c2c__display_stats(FILE *out) int llc_misses; struct c2c_stats *stats = &c2c.hists.stats; - llc_misses = stats->lcl_dram + - stats->rmt_dram + - stats->rmt_hit + - stats->rmt_hitm; + llc_misses = get_load_llc_misses(stats); fprintf(out, "=================================================\n"); fprintf(out, " Trace Event Information \n"); @@ -3290,6 +3281,7 @@ static int perf_c2c__record(int argc, const char **argv) */ if (e->tag) { e->record = true; + rec_argv[i++] = "-W"; } else { e = perf_mem_events__ptr(PERF_MEM_EVENTS__LOAD); e->record = true; diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c index 2a0f992ca0be..e254f18986f7 100644 --- a/tools/perf/builtin-inject.c +++ b/tools/perf/builtin-inject.c @@ -21,6 +21,7 @@ #include "util/data.h" #include "util/auxtrace.h" #include "util/jit.h" +#include "util/string2.h" #include "util/symbol.h" #include "util/synthetic-events.h" #include "util/thread.h" @@ -38,6 +39,7 @@ #include <linux/string.h> #include <linux/zalloc.h> #include <linux/hash.h> +#include <ctype.h> #include <errno.h> #include <signal.h> #include <inttypes.h> @@ -123,6 +125,7 @@ struct perf_inject { char event_copy[PERF_SAMPLE_MAX_SIZE]; struct perf_file_section secs[HEADER_FEAT_BITS]; struct guest_session guest_session; + struct strlist *known_build_ids; }; struct event_entry { @@ -433,8 +436,10 @@ static struct dso *findnew_dso(int pid, int tid, const char *filename, } if (dso) { + mutex_lock(&dso->lock); nsinfo__put(dso->nsinfo); dso->nsinfo = nsi; + mutex_unlock(&dso->lock); } else nsinfo__put(nsi); @@ -617,6 +622,7 @@ static int dso__read_build_id(struct dso *dso) if (dso->has_build_id) return 0; + mutex_lock(&dso->lock); nsinfo__mountns_enter(dso->nsinfo, &nsc); if (filename__read_build_id(dso->long_name, &dso->bid) > 0) dso->has_build_id = true; @@ -630,13 +636,78 @@ static int dso__read_build_id(struct dso *dso) free(new_name); } nsinfo__mountns_exit(&nsc); + mutex_unlock(&dso->lock); return dso->has_build_id ? 0 : -1; } +static struct strlist *perf_inject__parse_known_build_ids( + const char *known_build_ids_string) +{ + struct str_node *pos, *tmp; + struct strlist *known_build_ids; + int bid_len; + + known_build_ids = strlist__new(known_build_ids_string, NULL); + if (known_build_ids == NULL) + return NULL; + strlist__for_each_entry_safe(pos, tmp, known_build_ids) { + const char *build_id, *dso_name; + + build_id = skip_spaces(pos->s); + dso_name = strchr(build_id, ' '); + if (dso_name == NULL) { + strlist__remove(known_build_ids, pos); + continue; + } + bid_len = dso_name - pos->s; + dso_name = skip_spaces(dso_name); + if (bid_len % 2 != 0 || bid_len >= SBUILD_ID_SIZE) { + strlist__remove(known_build_ids, pos); + continue; + } + for (int ix = 0; 2 * ix + 1 < bid_len; ++ix) { + if (!isxdigit(build_id[2 * ix]) || + !isxdigit(build_id[2 * ix + 1])) { + strlist__remove(known_build_ids, pos); + break; + } + } + } + return known_build_ids; +} + +static bool perf_inject__lookup_known_build_id(struct perf_inject *inject, + struct dso *dso) +{ + struct str_node *pos; + int bid_len; + + strlist__for_each_entry(pos, inject->known_build_ids) { + const char *build_id, *dso_name; + + build_id = skip_spaces(pos->s); + dso_name = strchr(build_id, ' '); + bid_len = dso_name - pos->s; + dso_name = skip_spaces(dso_name); + if (strcmp(dso->long_name, dso_name)) + continue; + for (int ix = 0; 2 * ix + 1 < bid_len; ++ix) { + dso->bid.data[ix] = (hex(build_id[2 * ix]) << 4 | + hex(build_id[2 * ix + 1])); + } + dso->bid.size = bid_len / 2; + dso->has_build_id = 1; + return true; + } + return false; +} + static int dso__inject_build_id(struct dso *dso, struct perf_tool *tool, struct machine *machine, u8 cpumode, u32 flags) { + struct perf_inject *inject = container_of(tool, struct perf_inject, + tool); int err; if (is_anon_memory(dso->long_name) || flags & MAP_HUGETLB) @@ -644,6 +715,10 @@ static int dso__inject_build_id(struct dso *dso, struct perf_tool *tool, if (is_no_dso_memory(dso->long_name)) return 0; + if (inject->known_build_ids != NULL && + perf_inject__lookup_known_build_id(inject, dso)) + return 1; + if (dso__read_build_id(dso) < 0) { pr_debug("no build_id found for %s\n", dso->long_name); return -1; @@ -2112,12 +2187,16 @@ int cmd_inject(int argc, const char **argv) }; int ret; bool repipe = true; + const char *known_build_ids = NULL; struct option options[] = { OPT_BOOLEAN('b', "build-ids", &inject.build_ids, "Inject build-ids into the output stream"), OPT_BOOLEAN(0, "buildid-all", &inject.build_id_all, "Inject build-ids of all DSOs into the output stream"), + OPT_STRING(0, "known-build-ids", &known_build_ids, + "buildid path [,buildid path...]", + "build-ids to use for given paths"), OPT_STRING('i', "input", &inject.input_name, "file", "input file name"), OPT_STRING('o', "output", &inject.output.path, "file", @@ -2257,6 +2336,15 @@ int cmd_inject(int argc, const char **argv) */ inject.tool.ordered_events = true; inject.tool.ordering_requires_timestamps = true; + if (known_build_ids != NULL) { + inject.known_build_ids = + perf_inject__parse_known_build_ids(known_build_ids); + + if (inject.known_build_ids == NULL) { + pr_err("Couldn't parse known build ids.\n"); + goto out_delete; + } + } } if (inject.sched_stat) { @@ -2285,6 +2373,7 @@ int cmd_inject(int argc, const char **argv) guest_session__exit(&inject.guest_session); out_delete: + strlist__delete(inject.known_build_ids); zstd_fini(&(inject.session->zstd_data)); perf_session__delete(inject.session); out_close_output: diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c index ea40ae52cd2c..9722d4ab2e55 100644 --- a/tools/perf/builtin-lock.c +++ b/tools/perf/builtin-lock.c @@ -28,7 +28,6 @@ #include <sys/types.h> #include <sys/prctl.h> #include <semaphore.h> -#include <pthread.h> #include <math.h> #include <limits.h> @@ -57,6 +56,9 @@ static bool combine_locks; static bool show_thread_stats; static bool use_bpf; static unsigned long bpf_map_entries = 10240; +static int max_stack_depth = CONTENTION_STACK_DEPTH; +static int stack_skip = CONTENTION_STACK_SKIP; +static int print_nr_entries = INT_MAX / 2; static enum { LOCK_AGGR_ADDR, @@ -561,29 +563,50 @@ enum acquire_flags { READ_LOCK = 2, }; -static int report_lock_acquire_event(struct evsel *evsel, - struct perf_sample *sample) +static int get_key_by_aggr_mode_simple(u64 *key, u64 addr, u32 tid) { - struct lock_stat *ls; - struct thread_stat *ts; - struct lock_seq_stat *seq; - const char *name = evsel__strval(evsel, sample, "name"); - u64 addr = evsel__intval(evsel, sample, "lockdep_addr"); - int flag = evsel__intval(evsel, sample, "flags"); - u64 key; - switch (aggr_mode) { case LOCK_AGGR_ADDR: - key = addr; + *key = addr; break; case LOCK_AGGR_TASK: - key = sample->tid; + *key = tid; break; case LOCK_AGGR_CALLER: default: pr_err("Invalid aggregation mode: %d\n", aggr_mode); return -EINVAL; } + return 0; +} + +static u64 callchain_id(struct evsel *evsel, struct perf_sample *sample); + +static int get_key_by_aggr_mode(u64 *key, u64 addr, struct evsel *evsel, + struct perf_sample *sample) +{ + if (aggr_mode == LOCK_AGGR_CALLER) { + *key = callchain_id(evsel, sample); + return 0; + } + return get_key_by_aggr_mode_simple(key, addr, sample->tid); +} + +static int report_lock_acquire_event(struct evsel *evsel, + struct perf_sample *sample) +{ + struct lock_stat *ls; + struct thread_stat *ts; + struct lock_seq_stat *seq; + const char *name = evsel__strval(evsel, sample, "name"); + u64 addr = evsel__intval(evsel, sample, "lockdep_addr"); + int flag = evsel__intval(evsel, sample, "flags"); + u64 key; + int ret; + + ret = get_key_by_aggr_mode_simple(&key, addr, sample->tid); + if (ret < 0) + return ret; ls = lock_stat_findnew(key, name, 0); if (!ls) @@ -654,19 +677,11 @@ static int report_lock_acquired_event(struct evsel *evsel, const char *name = evsel__strval(evsel, sample, "name"); u64 addr = evsel__intval(evsel, sample, "lockdep_addr"); u64 key; + int ret; - switch (aggr_mode) { - case LOCK_AGGR_ADDR: - key = addr; - break; - case LOCK_AGGR_TASK: - key = sample->tid; - break; - case LOCK_AGGR_CALLER: - default: - pr_err("Invalid aggregation mode: %d\n", aggr_mode); - return -EINVAL; - } + ret = get_key_by_aggr_mode_simple(&key, addr, sample->tid); + if (ret < 0) + return ret; ls = lock_stat_findnew(key, name, 0); if (!ls) @@ -727,19 +742,11 @@ static int report_lock_contended_event(struct evsel *evsel, const char *name = evsel__strval(evsel, sample, "name"); u64 addr = evsel__intval(evsel, sample, "lockdep_addr"); u64 key; + int ret; - switch (aggr_mode) { - case LOCK_AGGR_ADDR: - key = addr; - break; - case LOCK_AGGR_TASK: - key = sample->tid; - break; - case LOCK_AGGR_CALLER: - default: - pr_err("Invalid aggregation mode: %d\n", aggr_mode); - return -EINVAL; - } + ret = get_key_by_aggr_mode_simple(&key, addr, sample->tid); + if (ret < 0) + return ret; ls = lock_stat_findnew(key, name, 0); if (!ls) @@ -793,19 +800,11 @@ static int report_lock_release_event(struct evsel *evsel, const char *name = evsel__strval(evsel, sample, "name"); u64 addr = evsel__intval(evsel, sample, "lockdep_addr"); u64 key; + int ret; - switch (aggr_mode) { - case LOCK_AGGR_ADDR: - key = addr; - break; - case LOCK_AGGR_TASK: - key = sample->tid; - break; - case LOCK_AGGR_CALLER: - default: - pr_err("Invalid aggregation mode: %d\n", aggr_mode); - return -EINVAL; - } + ret = get_key_by_aggr_mode_simple(&key, addr, sample->tid); + if (ret < 0) + return ret; ls = lock_stat_findnew(key, name, 0); if (!ls) @@ -903,6 +902,23 @@ bool is_lock_function(struct machine *machine, u64 addr) return false; } +static int get_symbol_name_offset(struct map *map, struct symbol *sym, u64 ip, + char *buf, int size) +{ + u64 offset; + + if (map == NULL || sym == NULL) { + buf[0] = '\0'; + return 0; + } + + offset = map->map_ip(map, ip) - sym->start; + + if (offset) + return scnprintf(buf, size, "%s+%#lx", sym->name, offset); + else + return strlcpy(buf, sym->name, size); +} static int lock_contention_caller(struct evsel *evsel, struct perf_sample *sample, char *buf, int size) { @@ -923,7 +939,7 @@ static int lock_contention_caller(struct evsel *evsel, struct perf_sample *sampl /* use caller function name from the callchain */ ret = thread__resolve_callchain(thread, cursor, evsel, sample, - NULL, NULL, CONTENTION_STACK_DEPTH); + NULL, NULL, max_stack_depth); if (ret != 0) { thread__put(thread); return -1; @@ -940,20 +956,13 @@ static int lock_contention_caller(struct evsel *evsel, struct perf_sample *sampl break; /* skip first few entries - for lock functions */ - if (++skip <= CONTENTION_STACK_SKIP) + if (++skip <= stack_skip) goto next; sym = node->ms.sym; if (sym && !is_lock_function(machine, node->ip)) { - struct map *map = node->ms.map; - u64 offset; - - offset = map->map_ip(map, node->ip) - sym->start; - - if (offset) - scnprintf(buf, size, "%s+%#lx", sym->name, offset); - else - strlcpy(buf, sym->name, size); + get_symbol_name_offset(node->ms.map, sym, node->ip, + buf, size); return 0; } @@ -978,7 +987,7 @@ static u64 callchain_id(struct evsel *evsel, struct perf_sample *sample) /* use caller function name from the callchain */ ret = thread__resolve_callchain(thread, cursor, evsel, sample, - NULL, NULL, CONTENTION_STACK_DEPTH); + NULL, NULL, max_stack_depth); thread__put(thread); if (ret != 0) @@ -994,7 +1003,7 @@ static u64 callchain_id(struct evsel *evsel, struct perf_sample *sample) break; /* skip first few entries - for lock functions */ - if (++skip <= CONTENTION_STACK_SKIP) + if (++skip <= stack_skip) goto next; if (node->ms.sym && is_lock_function(machine, node->ip)) @@ -1008,6 +1017,27 @@ next: return hash; } +static u64 *get_callstack(struct perf_sample *sample, int max_stack) +{ + u64 *callstack; + u64 i; + int c; + + callstack = calloc(max_stack, sizeof(*callstack)); + if (callstack == NULL) + return NULL; + + for (i = 0, c = 0; i < sample->callchain->nr && c < max_stack; i++) { + u64 ip = sample->callchain->ips[i]; + + if (ip >= PERF_CONTEXT_MAX) + continue; + + callstack[c++] = ip; + } + return callstack; +} + static int report_lock_contention_begin_event(struct evsel *evsel, struct perf_sample *sample) { @@ -1016,21 +1046,11 @@ static int report_lock_contention_begin_event(struct evsel *evsel, struct lock_seq_stat *seq; u64 addr = evsel__intval(evsel, sample, "lock_addr"); u64 key; + int ret; - switch (aggr_mode) { - case LOCK_AGGR_ADDR: - key = addr; - break; - case LOCK_AGGR_TASK: - key = sample->tid; - break; - case LOCK_AGGR_CALLER: - key = callchain_id(evsel, sample); - break; - default: - pr_err("Invalid aggregation mode: %d\n", aggr_mode); - return -EINVAL; - } + ret = get_key_by_aggr_mode(&key, addr, evsel, sample); + if (ret < 0) + return ret; ls = lock_stat_find(key); if (!ls) { @@ -1044,6 +1064,12 @@ static int report_lock_contention_begin_event(struct evsel *evsel, ls = lock_stat_findnew(key, caller, flags); if (!ls) return -ENOMEM; + + if (aggr_mode == LOCK_AGGR_CALLER && verbose) { + ls->callstack = get_callstack(sample, max_stack_depth); + if (ls->callstack == NULL) + return -ENOMEM; + } } ts = thread_stat_findnew(sample->tid); @@ -1099,21 +1125,11 @@ static int report_lock_contention_end_event(struct evsel *evsel, u64 contended_term; u64 addr = evsel__intval(evsel, sample, "lock_addr"); u64 key; + int ret; - switch (aggr_mode) { - case LOCK_AGGR_ADDR: - key = addr; - break; - case LOCK_AGGR_TASK: - key = sample->tid; - break; - case LOCK_AGGR_CALLER: - key = callchain_id(evsel, sample); - break; - default: - pr_err("Invalid aggregation mode: %d\n", aggr_mode); - return -EINVAL; - } + ret = get_key_by_aggr_mode(&key, addr, evsel, sample); + if (ret < 0) + return ret; ls = lock_stat_find(key); if (!ls) @@ -1234,7 +1250,7 @@ static void print_bad_events(int bad, int total) for (i = 0; i < BROKEN_MAX; i++) broken += bad_hist[i]; - if (broken == 0 && !verbose) + if (quiet || (broken == 0 && !verbose)) return; pr_info("\n=== output for debug===\n\n"); @@ -1251,14 +1267,16 @@ static void print_result(void) struct lock_stat *st; struct lock_key *key; char cut_name[20]; - int bad, total; + int bad, total, printed; - pr_info("%20s ", "Name"); - list_for_each_entry(key, &lock_keys, list) - pr_info("%*s ", key->len, key->header); - pr_info("\n\n"); + if (!quiet) { + pr_info("%20s ", "Name"); + list_for_each_entry(key, &lock_keys, list) + pr_info("%*s ", key->len, key->header); + pr_info("\n\n"); + } - bad = total = 0; + bad = total = printed = 0; while ((st = pop_from_result())) { total++; if (st->broken) @@ -1296,6 +1314,9 @@ static void print_result(void) pr_info(" "); } pr_info("\n"); + + if (++printed >= print_nr_entries) + break; } print_bad_events(bad, total); @@ -1457,21 +1478,23 @@ static void sort_contention_result(void) sort_result(); } -static void print_contention_result(void) +static void print_contention_result(struct lock_contention *con) { struct lock_stat *st; struct lock_key *key; - int bad, total; + int bad, total, printed; - list_for_each_entry(key, &lock_keys, list) - pr_info("%*s ", key->len, key->header); + if (!quiet) { + list_for_each_entry(key, &lock_keys, list) + pr_info("%*s ", key->len, key->header); - if (show_thread_stats) - pr_info(" %10s %s\n\n", "pid", "comm"); - else - pr_info(" %10s %s\n\n", "type", "caller"); + if (show_thread_stats) + pr_info(" %10s %s\n\n", "pid", "comm"); + else + pr_info(" %10s %s\n\n", "type", "caller"); + } - bad = total = 0; + bad = total = printed = 0; if (use_bpf) bad = bad_hist[BROKEN_CONTENDED]; @@ -1492,10 +1515,30 @@ static void print_contention_result(void) /* st->addr contains tid of thread */ t = perf_session__findnew(session, pid); pr_info(" %10d %s\n", pid, thread__comm_str(t)); - continue; + goto next; } pr_info(" %10s %s\n", get_type_str(st), st->name); + if (verbose) { + struct map *kmap; + struct symbol *sym; + char buf[128]; + u64 ip; + + for (int i = 0; i < max_stack_depth; i++) { + if (!st->callstack || !st->callstack[i]) + break; + + ip = st->callstack[i]; + sym = machine__find_kernel_symbol(con->machine, ip, &kmap); + get_symbol_name_offset(kmap, sym, ip, buf, sizeof(buf)); + pr_info("\t\t\t%#lx %s\n", (unsigned long)ip, buf); + } + } + +next: + if (++printed >= print_nr_entries) + break; } print_bad_events(bad, total); @@ -1603,6 +1646,8 @@ static int __cmd_contention(int argc, const char **argv) .target = &target, .result = &lockhash_table[0], .map_nr_entries = bpf_map_entries, + .max_stack = max_stack_depth, + .stack_skip = stack_skip, }; session = perf_session__new(use_bpf ? NULL : &data, &eops); @@ -1611,6 +1656,8 @@ static int __cmd_contention(int argc, const char **argv) return PTR_ERR(session); } + con.machine = &session->machines.host; + /* for lock function check */ symbol_conf.sort_by_name = true; symbol__init(&session->header.env); @@ -1629,8 +1676,6 @@ static int __cmd_contention(int argc, const char **argv) signal(SIGCHLD, sighandler); signal(SIGTERM, sighandler); - con.machine = &session->machines.host; - con.evlist = evlist__new(); if (con.evlist == NULL) { err = -ENOMEM; @@ -1702,7 +1747,7 @@ static int __cmd_contention(int argc, const char **argv) setup_pager(); sort_contention_result(); - print_contention_result(); + print_contention_result(&con); out_delete: evlist__delete(con.evlist); @@ -1824,6 +1869,7 @@ int cmd_lock(int argc, const char **argv) "file", "vmlinux pathname"), OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name, "file", "kallsyms pathname"), + OPT_BOOLEAN('q', "quiet", &quiet, "Do not show any message"), OPT_END() }; @@ -1845,6 +1891,7 @@ int cmd_lock(int argc, const char **argv) "combine locks in the same class"), OPT_BOOLEAN('t', "threads", &show_thread_stats, "show per-thread lock stats"), + OPT_INTEGER('E', "entries", &print_nr_entries, "display this many functions"), OPT_PARENT(lock_options) }; @@ -1866,6 +1913,13 @@ int cmd_lock(int argc, const char **argv) "Trace on existing thread id (exclusive to --pid)"), OPT_CALLBACK(0, "map-nr-entries", &bpf_map_entries, "num", "Max number of BPF map entries", parse_map_entry), + OPT_INTEGER(0, "max-stack", &max_stack_depth, + "Set the maximum stack depth when collecting lock contention, " + "Default: " __stringify(CONTENTION_STACK_DEPTH)), + OPT_INTEGER(0, "stack-skip", &stack_skip, + "Set the number of stack depth to skip when finding a lock caller, " + "Default: " __stringify(CONTENTION_STACK_SKIP)), + OPT_INTEGER('E', "entries", &print_nr_entries, "display this many functions"), OPT_PARENT(lock_options) }; diff --git a/tools/perf/builtin-mem.c b/tools/perf/builtin-mem.c index 9e435fd23503..f7dd8216de72 100644 --- a/tools/perf/builtin-mem.c +++ b/tools/perf/builtin-mem.c @@ -122,6 +122,7 @@ static int __cmd_record(int argc, const char **argv, struct perf_mem *mem) (mem->operation & MEM_OPERATION_LOAD) && (mem->operation & MEM_OPERATION_STORE)) { e->record = true; + rec_argv[i++] = "-W"; } else { if (mem->operation & MEM_OPERATION_LOAD) { e = perf_mem_events__ptr(PERF_MEM_EVENTS__LOAD); diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 0f711f88894c..52d254b1530c 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -10,6 +10,7 @@ #include "util/build-id.h" #include <subcmd/parse-options.h> +#include <internal/xyarray.h> #include "util/parse-events.h" #include "util/config.h" @@ -21,6 +22,7 @@ #include "util/evsel.h" #include "util/debug.h" #include "util/mmap.h" +#include "util/mutex.h" #include "util/target.h" #include "util/session.h" #include "util/tool.h" @@ -143,6 +145,11 @@ static const char *thread_spec_tags[THREAD_SPEC__MAX] = { "undefined", "cpu", "core", "package", "numa", "user" }; +struct pollfd_index_map { + int evlist_pollfd_index; + int thread_pollfd_index; +}; + struct record { struct perf_tool tool; struct record_opts opts; @@ -171,6 +178,9 @@ struct record { int nr_threads; struct thread_mask *thread_masks; struct record_thread *thread_data; + struct pollfd_index_map *index_map; + size_t index_map_sz; + size_t index_map_cnt; }; static volatile int done; @@ -608,17 +618,18 @@ static int process_synthesized_event(struct perf_tool *tool, return record__write(rec, NULL, event, event->header.size); } +static struct mutex synth_lock; + static int process_locked_synthesized_event(struct perf_tool *tool, union perf_event *event, struct perf_sample *sample __maybe_unused, struct machine *machine __maybe_unused) { - static pthread_mutex_t synth_lock = PTHREAD_MUTEX_INITIALIZER; int ret; - pthread_mutex_lock(&synth_lock); + mutex_lock(&synth_lock); ret = process_synthesized_event(tool, event, sample, machine); - pthread_mutex_unlock(&synth_lock); + mutex_unlock(&synth_lock); return ret; } @@ -1074,6 +1085,70 @@ static void record__free_thread_data(struct record *rec) zfree(&rec->thread_data); } +static int record__map_thread_evlist_pollfd_indexes(struct record *rec, + int evlist_pollfd_index, + int thread_pollfd_index) +{ + size_t x = rec->index_map_cnt; + + if (realloc_array_as_needed(rec->index_map, rec->index_map_sz, x, NULL)) + return -ENOMEM; + rec->index_map[x].evlist_pollfd_index = evlist_pollfd_index; + rec->index_map[x].thread_pollfd_index = thread_pollfd_index; + rec->index_map_cnt += 1; + return 0; +} + +static int record__update_evlist_pollfd_from_thread(struct record *rec, + struct evlist *evlist, + struct record_thread *thread_data) +{ + struct pollfd *e_entries = evlist->core.pollfd.entries; + struct pollfd *t_entries = thread_data->pollfd.entries; + int err = 0; + size_t i; + + for (i = 0; i < rec->index_map_cnt; i++) { + int e_pos = rec->index_map[i].evlist_pollfd_index; + int t_pos = rec->index_map[i].thread_pollfd_index; + + if (e_entries[e_pos].fd != t_entries[t_pos].fd || + e_entries[e_pos].events != t_entries[t_pos].events) { + pr_err("Thread and evlist pollfd index mismatch\n"); + err = -EINVAL; + continue; + } + e_entries[e_pos].revents = t_entries[t_pos].revents; + } + return err; +} + +static int record__dup_non_perf_events(struct record *rec, + struct evlist *evlist, + struct record_thread *thread_data) +{ + struct fdarray *fda = &evlist->core.pollfd; + int i, ret; + + for (i = 0; i < fda->nr; i++) { + if (!(fda->priv[i].flags & fdarray_flag__non_perf_event)) + continue; + ret = fdarray__dup_entry_from(&thread_data->pollfd, i, fda); + if (ret < 0) { + pr_err("Failed to duplicate descriptor in main thread pollfd\n"); + return ret; + } + pr_debug2("thread_data[%p]: pollfd[%d] <- non_perf_event fd=%d\n", + thread_data, ret, fda->entries[i].fd); + ret = record__map_thread_evlist_pollfd_indexes(rec, i, ret); + if (ret < 0) { + pr_err("Failed to map thread and evlist pollfd indexes\n"); + return ret; + } + } + return 0; +} + static int record__alloc_thread_data(struct record *rec, struct evlist *evlist) { int t, ret; @@ -1121,18 +1196,12 @@ static int record__alloc_thread_data(struct record *rec, struct evlist *evlist) thread_data[t].pipes.msg[0]); } else { thread_data[t].tid = gettid(); - if (evlist->ctl_fd.pos == -1) - continue; - ret = fdarray__dup_entry_from(&thread_data[t].pollfd, evlist->ctl_fd.pos, - &evlist->core.pollfd); - if (ret < 0) { - pr_err("Failed to duplicate descriptor in main thread pollfd\n"); + + ret = record__dup_non_perf_events(rec, evlist, &thread_data[t]); + if (ret < 0) goto out_free; - } - thread_data[t].ctlfd_pos = ret; - pr_debug2("thread_data[%p]: pollfd[%d] <- ctl_fd=%d\n", - thread_data, thread_data[t].ctlfd_pos, - evlist->core.pollfd.entries[evlist->ctl_fd.pos].fd); + + thread_data[t].ctlfd_pos = -1; /* Not used */ } } @@ -1784,6 +1853,74 @@ record__switch_output(struct record *rec, bool at_exit) return fd; } +static void __record__read_lost_samples(struct record *rec, struct evsel *evsel, + struct perf_record_lost_samples *lost, + int cpu_idx, int thread_idx) +{ + struct perf_counts_values count; + struct perf_sample_id *sid; + struct perf_sample sample = {}; + int id_hdr_size; + + if (perf_evsel__read(&evsel->core, cpu_idx, thread_idx, &count) < 0) { + pr_err("read LOST count failed\n"); + return; + } + + if (count.lost == 0) + return; + + lost->lost = count.lost; + if (evsel->core.ids) { + sid = xyarray__entry(evsel->core.sample_id, cpu_idx, thread_idx); + sample.id = sid->id; + } + + id_hdr_size = perf_event__synthesize_id_sample((void *)(lost + 1), + evsel->core.attr.sample_type, &sample); + lost->header.size = sizeof(*lost) + id_hdr_size; + record__write(rec, NULL, lost, lost->header.size); +} + +static void record__read_lost_samples(struct record *rec) +{ + struct perf_session *session = rec->session; + struct perf_record_lost_samples *lost; + struct evsel *evsel; + + /* there was an error during record__open */ + if (session->evlist == NULL) + return; + + lost = zalloc(PERF_SAMPLE_MAX_SIZE); + if (lost == NULL) { + pr_debug("Memory allocation failed\n"); + return; + } + + lost->header.type = PERF_RECORD_LOST_SAMPLES; + + evlist__for_each_entry(session->evlist, evsel) { + struct xyarray *xy = evsel->core.sample_id; + + if (xy == NULL || evsel->core.fd == NULL) + continue; + if (xyarray__max_x(evsel->core.fd) != xyarray__max_x(xy) || + xyarray__max_y(evsel->core.fd) != xyarray__max_y(xy)) { + pr_debug("Unmatched FD vs. sample ID: skip reading LOST count\n"); + continue; + } + + for (int x = 0; x < xyarray__max_x(xy); x++) { + for (int y = 0; y < xyarray__max_y(xy); y++) { + __record__read_lost_samples(rec, evsel, lost, x, y); + } + } + } + free(lost); + +} + static volatile int workload_exec_errno; /* @@ -1921,6 +2058,7 @@ static int record__synthesize(struct record *rec, bool tail) } if (rec->opts.nr_threads_synthesize > 1) { + mutex_init(&synth_lock); perf_set_multithreaded(); f = process_locked_synthesized_event; } @@ -1934,8 +2072,10 @@ static int record__synthesize(struct record *rec, bool tail) rec->opts.nr_threads_synthesize); } - if (rec->opts.nr_threads_synthesize > 1) + if (rec->opts.nr_threads_synthesize > 1) { perf_set_singlethreaded(); + mutex_destroy(&synth_lock); + } out: return err; @@ -2294,10 +2434,14 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) record__uniquify_name(rec); + /* Debug message used by test scripts */ + pr_debug3("perf record opening and mmapping events\n"); if (record__open(rec) != 0) { err = -1; goto out_free_threads; } + /* Debug message used by test scripts */ + pr_debug3("perf record done opening and mmapping events\n"); session->header.env.comp_mmap_len = session->evlist->core.mmap_len; if (rec->opts.kcore) { @@ -2436,6 +2580,14 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) } } + err = event_enable_timer__start(rec->evlist->eet); + if (err) + goto out_child; + + /* Debug message used by test scripts */ + pr_debug3("perf record has started\n"); + fflush(stderr); + trigger_ready(&auxtrace_snapshot_trigger); trigger_ready(&switch_output_trigger); perf_hooks__invoke_record_start(); @@ -2534,8 +2686,9 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) record__thread_munmap_filtered, NULL) == 0) draining = true; - evlist__ctlfd_update(rec->evlist, - &thread->pollfd.entries[thread->ctlfd_pos]); + err = record__update_evlist_pollfd_from_thread(rec, rec->evlist, thread); + if (err) + goto out_child; } if (evlist__ctlfd_process(rec->evlist, &cmd) > 0) { @@ -2558,6 +2711,14 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) } } + err = event_enable_timer__process(rec->evlist->eet); + if (err < 0) + goto out_child; + if (err) { + err = 0; + done = 1; + } + /* * When perf is starting the traced process, at the end events * die with the process and we wait for that. Thus no need to @@ -2630,6 +2791,7 @@ out_free_threads: if (rec->off_cpu) rec->bytes_written += off_cpu_write(rec->session); + record__read_lost_samples(rec); record__synthesize(rec, true); /* this will be recalculated during process_buildids() */ rec->samples = 0; @@ -2779,6 +2941,12 @@ static int perf_record_config(const char *var, const char *value, void *cb) return 0; } +static int record__parse_event_enable_time(const struct option *opt, const char *str, int unset) +{ + struct record *rec = (struct record *)opt->value; + + return evlist__parse_event_enable_time(rec->evlist, &rec->opts, str, unset); +} static int record__parse_affinity(const struct option *opt, const char *str, int unset) { @@ -3240,8 +3408,10 @@ static struct option __record_options[] = { OPT_CALLBACK('G', "cgroup", &record.evlist, "name", "monitor event in cgroup name only", parse_cgroups), - OPT_INTEGER('D', "delay", &record.opts.initial_delay, - "ms to wait before starting measurement after program start (-1: start with events disabled)"), + OPT_CALLBACK('D', "delay", &record, "ms", + "ms to wait before starting measurement after program start (-1: start with events disabled), " + "or ranges of time to enable events e.g. '-D 10-20,30-40'", + record__parse_event_enable_time), OPT_BOOLEAN(0, "kcore", &record.opts.kcore, "copy /proc/kcore"), OPT_STRING('u', "uid", &record.opts.target.uid_str, "user", "user to profile"), diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 91ed41cc7d88..8361890176c2 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -752,6 +752,22 @@ static int count_sample_event(struct perf_tool *tool __maybe_unused, return 0; } +static int count_lost_samples_event(struct perf_tool *tool, + union perf_event *event, + struct perf_sample *sample, + struct machine *machine __maybe_unused) +{ + struct report *rep = container_of(tool, struct report, tool); + struct evsel *evsel; + + evsel = evlist__id2evsel(rep->session->evlist, sample->id); + if (evsel) { + hists__inc_nr_lost_samples(evsel__hists(evsel), + event->lost_samples.lost); + } + return 0; +} + static int process_attr(struct perf_tool *tool __maybe_unused, union perf_event *event, struct evlist **pevlist); @@ -761,6 +777,7 @@ static void stats_setup(struct report *rep) memset(&rep->tool, 0, sizeof(rep->tool)); rep->tool.attr = process_attr; rep->tool.sample = count_sample_event; + rep->tool.lost_samples = count_lost_samples_event; rep->tool.no_warn = true; } diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index a5cf243c337f..f93737eef07b 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -7,6 +7,7 @@ #include "util/evlist.h" #include "util/evsel.h" #include "util/evsel_fprintf.h" +#include "util/mutex.h" #include "util/symbol.h" #include "util/thread.h" #include "util/header.h" @@ -184,8 +185,8 @@ struct perf_sched { struct task_desc **pid_to_task; struct task_desc **tasks; const struct trace_sched_handler *tp_handler; - pthread_mutex_t start_work_mutex; - pthread_mutex_t work_done_wait_mutex; + struct mutex start_work_mutex; + struct mutex work_done_wait_mutex; int profile_cpu; /* * Track the current task - that way we can know whether there's any @@ -245,6 +246,7 @@ struct perf_sched { const char *time_str; struct perf_time_interval ptime; struct perf_time_interval hist_time; + volatile bool thread_funcs_exit; }; /* per thread run time data */ @@ -632,35 +634,34 @@ static void *thread_func(void *ctx) prctl(PR_SET_NAME, comm2); if (fd < 0) return NULL; -again: - ret = sem_post(&this_task->ready_for_work); - BUG_ON(ret); - ret = pthread_mutex_lock(&sched->start_work_mutex); - BUG_ON(ret); - ret = pthread_mutex_unlock(&sched->start_work_mutex); - BUG_ON(ret); - cpu_usage_0 = get_cpu_usage_nsec_self(fd); + while (!sched->thread_funcs_exit) { + ret = sem_post(&this_task->ready_for_work); + BUG_ON(ret); + mutex_lock(&sched->start_work_mutex); + mutex_unlock(&sched->start_work_mutex); - for (i = 0; i < this_task->nr_events; i++) { - this_task->curr_event = i; - perf_sched__process_event(sched, this_task->atoms[i]); - } + cpu_usage_0 = get_cpu_usage_nsec_self(fd); - cpu_usage_1 = get_cpu_usage_nsec_self(fd); - this_task->cpu_usage = cpu_usage_1 - cpu_usage_0; - ret = sem_post(&this_task->work_done_sem); - BUG_ON(ret); + for (i = 0; i < this_task->nr_events; i++) { + this_task->curr_event = i; + perf_sched__process_event(sched, this_task->atoms[i]); + } - ret = pthread_mutex_lock(&sched->work_done_wait_mutex); - BUG_ON(ret); - ret = pthread_mutex_unlock(&sched->work_done_wait_mutex); - BUG_ON(ret); + cpu_usage_1 = get_cpu_usage_nsec_self(fd); + this_task->cpu_usage = cpu_usage_1 - cpu_usage_0; + ret = sem_post(&this_task->work_done_sem); + BUG_ON(ret); - goto again; + mutex_lock(&sched->work_done_wait_mutex); + mutex_unlock(&sched->work_done_wait_mutex); + } + return NULL; } static void create_tasks(struct perf_sched *sched) + EXCLUSIVE_LOCK_FUNCTION(sched->start_work_mutex) + EXCLUSIVE_LOCK_FUNCTION(sched->work_done_wait_mutex) { struct task_desc *task; pthread_attr_t attr; @@ -672,10 +673,8 @@ static void create_tasks(struct perf_sched *sched) err = pthread_attr_setstacksize(&attr, (size_t) max(16 * 1024, (int)PTHREAD_STACK_MIN)); BUG_ON(err); - err = pthread_mutex_lock(&sched->start_work_mutex); - BUG_ON(err); - err = pthread_mutex_lock(&sched->work_done_wait_mutex); - BUG_ON(err); + mutex_lock(&sched->start_work_mutex); + mutex_lock(&sched->work_done_wait_mutex); for (i = 0; i < sched->nr_tasks; i++) { struct sched_thread_parms *parms = malloc(sizeof(*parms)); BUG_ON(parms == NULL); @@ -691,7 +690,30 @@ static void create_tasks(struct perf_sched *sched) } } +static void destroy_tasks(struct perf_sched *sched) + UNLOCK_FUNCTION(sched->start_work_mutex) + UNLOCK_FUNCTION(sched->work_done_wait_mutex) +{ + struct task_desc *task; + unsigned long i; + int err; + + mutex_unlock(&sched->start_work_mutex); + mutex_unlock(&sched->work_done_wait_mutex); + /* Get rid of threads so they won't be upset by mutex destrunction */ + for (i = 0; i < sched->nr_tasks; i++) { + task = sched->tasks[i]; + err = pthread_join(task->thread, NULL); + BUG_ON(err); + sem_destroy(&task->sleep_sem); + sem_destroy(&task->ready_for_work); + sem_destroy(&task->work_done_sem); + } +} + static void wait_for_tasks(struct perf_sched *sched) + EXCLUSIVE_LOCKS_REQUIRED(sched->work_done_wait_mutex) + EXCLUSIVE_LOCKS_REQUIRED(sched->start_work_mutex) { u64 cpu_usage_0, cpu_usage_1; struct task_desc *task; @@ -699,7 +721,7 @@ static void wait_for_tasks(struct perf_sched *sched) sched->start_time = get_nsecs(); sched->cpu_usage = 0; - pthread_mutex_unlock(&sched->work_done_wait_mutex); + mutex_unlock(&sched->work_done_wait_mutex); for (i = 0; i < sched->nr_tasks; i++) { task = sched->tasks[i]; @@ -707,12 +729,11 @@ static void wait_for_tasks(struct perf_sched *sched) BUG_ON(ret); sem_init(&task->ready_for_work, 0, 0); } - ret = pthread_mutex_lock(&sched->work_done_wait_mutex); - BUG_ON(ret); + mutex_lock(&sched->work_done_wait_mutex); cpu_usage_0 = get_cpu_usage_nsec_parent(); - pthread_mutex_unlock(&sched->start_work_mutex); + mutex_unlock(&sched->start_work_mutex); for (i = 0; i < sched->nr_tasks; i++) { task = sched->tasks[i]; @@ -734,8 +755,7 @@ static void wait_for_tasks(struct perf_sched *sched) sched->runavg_parent_cpu_usage = (sched->runavg_parent_cpu_usage * (sched->replay_repeat - 1) + sched->parent_cpu_usage)/sched->replay_repeat; - ret = pthread_mutex_lock(&sched->start_work_mutex); - BUG_ON(ret); + mutex_lock(&sched->start_work_mutex); for (i = 0; i < sched->nr_tasks; i++) { task = sched->tasks[i]; @@ -745,6 +765,8 @@ static void wait_for_tasks(struct perf_sched *sched) } static void run_one_test(struct perf_sched *sched) + EXCLUSIVE_LOCKS_REQUIRED(sched->work_done_wait_mutex) + EXCLUSIVE_LOCKS_REQUIRED(sched->start_work_mutex) { u64 T0, T1, delta, avg_delta, fluct; @@ -3316,11 +3338,14 @@ static int perf_sched__replay(struct perf_sched *sched) print_task_traces(sched); add_cross_task_wakeups(sched); + sched->thread_funcs_exit = false; create_tasks(sched); printf("------------------------------------------------------------\n"); for (i = 0; i < sched->replay_repeat; i++) run_one_test(sched); + sched->thread_funcs_exit = true; + destroy_tasks(sched); return 0; } @@ -3444,8 +3469,6 @@ int cmd_sched(int argc, const char **argv) }, .cmp_pid = LIST_HEAD_INIT(sched.cmp_pid), .sort_list = LIST_HEAD_INIT(sched.sort_list), - .start_work_mutex = PTHREAD_MUTEX_INITIALIZER, - .work_done_wait_mutex = PTHREAD_MUTEX_INITIALIZER, .sort_order = default_sort_order, .replay_repeat = 10, .profile_cpu = -1, @@ -3559,8 +3582,10 @@ int cmd_sched(int argc, const char **argv) .fork_event = replay_fork_event, }; unsigned int i; - int ret; + int ret = 0; + mutex_init(&sched.start_work_mutex); + mutex_init(&sched.work_done_wait_mutex); for (i = 0; i < ARRAY_SIZE(sched.curr_pid); i++) sched.curr_pid[i] = -1; @@ -3572,11 +3597,10 @@ int cmd_sched(int argc, const char **argv) /* * Aliased to 'perf script' for now: */ - if (!strcmp(argv[0], "script")) - return cmd_script(argc, argv); - - if (strlen(argv[0]) > 2 && strstarts("record", argv[0])) { - return __cmd_record(argc, argv); + if (!strcmp(argv[0], "script")) { + ret = cmd_script(argc, argv); + } else if (strlen(argv[0]) > 2 && strstarts("record", argv[0])) { + ret = __cmd_record(argc, argv); } else if (strlen(argv[0]) > 2 && strstarts("latency", argv[0])) { sched.tp_handler = &lat_ops; if (argc > 1) { @@ -3585,7 +3609,7 @@ int cmd_sched(int argc, const char **argv) usage_with_options(latency_usage, latency_options); } setup_sorting(&sched, latency_options, latency_usage); - return perf_sched__lat(&sched); + ret = perf_sched__lat(&sched); } else if (!strcmp(argv[0], "map")) { if (argc) { argc = parse_options(argc, argv, map_options, map_usage, 0); @@ -3594,7 +3618,7 @@ int cmd_sched(int argc, const char **argv) } sched.tp_handler = &map_ops; setup_sorting(&sched, latency_options, latency_usage); - return perf_sched__map(&sched); + ret = perf_sched__map(&sched); } else if (strlen(argv[0]) > 2 && strstarts("replay", argv[0])) { sched.tp_handler = &replay_ops; if (argc) { @@ -3602,7 +3626,7 @@ int cmd_sched(int argc, const char **argv) if (argc) usage_with_options(replay_usage, replay_options); } - return perf_sched__replay(&sched); + ret = perf_sched__replay(&sched); } else if (!strcmp(argv[0], "timehist")) { if (argc) { argc = parse_options(argc, argv, timehist_options, @@ -3618,16 +3642,21 @@ int cmd_sched(int argc, const char **argv) parse_options_usage(NULL, timehist_options, "w", true); if (sched.show_next) parse_options_usage(NULL, timehist_options, "n", true); - return -EINVAL; + ret = -EINVAL; + goto out; } ret = symbol__validate_sym_arguments(); if (ret) - return ret; + goto out; - return perf_sched__timehist(&sched); + ret = perf_sched__timehist(&sched); } else { usage_with_options(sched_usage, sched_options); } - return 0; +out: + mutex_destroy(&sched.start_work_mutex); + mutex_destroy(&sched.work_done_wait_mutex); + + return ret; } diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index 029b4330e59b..7ca238277d83 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -882,7 +882,7 @@ static int print_bstack_flags(FILE *fp, struct branch_entry *br) br->flags.in_tx ? 'X' : '-', br->flags.abort ? 'A' : '-', br->flags.cycles, - br->flags.type ? branch_type_name(br->flags.type) : "-"); + get_branch_type(br)); } static int perf_sample__fprintf_brstack(struct perf_sample *sample, @@ -2243,9 +2243,6 @@ static void __process_stat(struct evsel *counter, u64 tstamp) struct perf_cpu cpu; static int header_printed; - if (counter->core.system_wide) - nthreads = 1; - if (!header_printed) { printf("%3s %8s %15s %15s %15s %15s %s\n", "CPU", "THREAD", "VAL", "ENA", "RUN", "TIME", "EVENT"); @@ -3849,9 +3846,10 @@ int cmd_script(int argc, const char **argv) "Valid types: hw,sw,trace,raw,synth. " "Fields: comm,tid,pid,time,cpu,event,trace,ip,sym,dso," "addr,symoff,srcline,period,iregs,uregs,brstack," - "brstacksym,flags,bpf-output,brstackinsn,brstackinsnlen,brstackoff," - "callindent,insn,insnlen,synth,phys_addr,metric,misc,ipc,tod," - "data_page_size,code_page_size,ins_lat", + "brstacksym,flags,data_src,weight,bpf-output,brstackinsn," + "brstackinsnlen,brstackoff,callindent,insn,insnlen,synth," + "phys_addr,metric,misc,srccode,ipc,tod,data_page_size," + "code_page_size,ins_lat", parse_output_fields), OPT_BOOLEAN('a', "all-cpus", &system_wide, "system-wide collection from all CPUs"), diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index 0b4a62e4ff67..265b05157972 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -191,6 +191,7 @@ static bool append_file; static bool interval_count; static const char *output_name; static int output_fd; +static char *metrics; struct perf_stat { bool record; @@ -291,13 +292,8 @@ static inline void diff_timespec(struct timespec *r, struct timespec *a, static void perf_stat__reset_stats(void) { - int i; - evlist__reset_stats(evsel_list); perf_stat__reset_shadow_stats(); - - for (i = 0; i < stat_config.stats_num; i++) - perf_stat__reset_shadow_per_stat(&stat_config.stats[i]); } static int process_synthesized_event(struct perf_tool *tool __maybe_unused, @@ -488,46 +484,6 @@ static void read_counters(struct timespec *rs) } } -static int runtime_stat_new(struct perf_stat_config *config, int nthreads) -{ - int i; - - config->stats = calloc(nthreads, sizeof(struct runtime_stat)); - if (!config->stats) - return -1; - - config->stats_num = nthreads; - - for (i = 0; i < nthreads; i++) - runtime_stat__init(&config->stats[i]); - - return 0; -} - -static void runtime_stat_delete(struct perf_stat_config *config) -{ - int i; - - if (!config->stats) - return; - - for (i = 0; i < config->stats_num; i++) - runtime_stat__exit(&config->stats[i]); - - zfree(&config->stats); -} - -static void runtime_stat_reset(struct perf_stat_config *config) -{ - int i; - - if (!config->stats) - return; - - for (i = 0; i < config->stats_num; i++) - perf_stat__reset_shadow_per_stat(&config->stats[i]); -} - static void process_interval(void) { struct timespec ts, rs; @@ -536,7 +492,6 @@ static void process_interval(void) diff_timespec(&rs, &ts, &ref_time); perf_stat__reset_shadow_per_stat(&rt_stat); - runtime_stat_reset(&stat_config); read_counters(&rs); if (STAT_RECORD) { @@ -661,9 +616,7 @@ static void process_evlist(struct evlist *evlist, unsigned int interval) if (evlist__ctlfd_process(evlist, &cmd) > 0) { switch (cmd) { case EVLIST_CTL_CMD_ENABLE: - if (interval) - process_interval(); - break; + __fallthrough; case EVLIST_CTL_CMD_DISABLE: if (interval) process_interval(); @@ -901,8 +854,6 @@ try_again: evlist__for_each_cpu(evlist_cpu_itr, evsel_list, affinity) { counter = evlist_cpu_itr.evsel; - if (!counter->reset_group && !counter->errored) - continue; if (!counter->reset_group) continue; try_again_reset: @@ -1017,7 +968,6 @@ try_again_reset: evlist__copy_prev_raw_counts(evsel_list); evlist__reset_prev_raw_counts(evsel_list); - runtime_stat_reset(&stat_config); perf_stat__reset_shadow_per_stat(&rt_stat); } else { update_stats(&walltime_nsecs_stats, t1 - t0); @@ -1148,14 +1098,23 @@ static int enable_metric_only(const struct option *opt __maybe_unused, return 0; } -static int parse_metric_groups(const struct option *opt, +static int append_metric_groups(const struct option *opt __maybe_unused, const char *str, int unset __maybe_unused) { - return metricgroup__parse_groups(opt, str, - stat_config.metric_no_group, - stat_config.metric_no_merge, - &stat_config.metric_events); + if (metrics) { + char *tmp; + + if (asprintf(&tmp, "%s,%s", metrics, str) < 0) + return -ENOMEM; + free(metrics); + metrics = tmp; + } else { + metrics = strdup(str); + if (!metrics) + return -ENOMEM; + } + return 0; } static int parse_control_option(const struct option *opt, @@ -1299,7 +1258,7 @@ static struct option stat_options[] = { "measure SMI cost"), OPT_CALLBACK('M', "metrics", &evsel_list, "metric/metric group list", "monitor specified metrics or metric groups (separated by ,)", - parse_metric_groups), + append_metric_groups), OPT_BOOLEAN_FLAG(0, "all-kernel", &stat_config.all_kernel, "Configure all used events to run in kernel space.", PARSE_OPT_EXCLUSIVE), @@ -1792,11 +1751,11 @@ static int add_default_attributes(void) * on an architecture test for such a metric name. */ if (metricgroup__has_metric("transaction")) { - struct option opt = { .value = &evsel_list }; - - return metricgroup__parse_groups(&opt, "transaction", + return metricgroup__parse_groups(evsel_list, "transaction", stat_config.metric_no_group, - stat_config.metric_no_merge, + stat_config.metric_no_merge, + stat_config.user_requested_cpu_list, + stat_config.system_wide, &stat_config.metric_events); } @@ -2183,6 +2142,8 @@ static int __cmd_report(int argc, const char **argv) input_name = "perf.data"; } + perf_stat__init_shadow_stats(); + perf_stat.data.path = input_name; perf_stat.data.mode = PERF_DATA_MODE_READ; @@ -2262,8 +2223,6 @@ int cmd_stat(int argc, const char **argv) argc = parse_options_subcommand(argc, argv, stat_options, stat_subcommands, (const char **) stat_usage, PARSE_OPT_STOP_AT_NON_OPTION); - perf_stat__collect_metric_expr(evsel_list); - perf_stat__init_shadow_stats(); if (stat_config.csv_sep) { stat_config.csv_output = true; @@ -2430,6 +2389,34 @@ int cmd_stat(int argc, const char **argv) target.system_wide = true; } + if ((stat_config.aggr_mode == AGGR_THREAD) && (target.system_wide)) + target.per_thread = true; + + stat_config.system_wide = target.system_wide; + if (target.cpu_list) { + stat_config.user_requested_cpu_list = strdup(target.cpu_list); + if (!stat_config.user_requested_cpu_list) { + status = -ENOMEM; + goto out; + } + } + + /* + * Metric parsing needs to be delayed as metrics may optimize events + * knowing the target is system-wide. + */ + if (metrics) { + metricgroup__parse_groups(evsel_list, metrics, + stat_config.metric_no_group, + stat_config.metric_no_merge, + stat_config.user_requested_cpu_list, + stat_config.system_wide, + &stat_config.metric_events); + zfree(&metrics); + } + perf_stat__collect_metric_expr(evsel_list); + perf_stat__init_shadow_stats(); + if (add_default_attributes()) goto out; @@ -2449,9 +2436,6 @@ int cmd_stat(int argc, const char **argv) } } - if ((stat_config.aggr_mode == AGGR_THREAD) && (target.system_wide)) - target.per_thread = true; - if (evlist__fix_hybrid_cpus(evsel_list, target.cpu_list)) { pr_err("failed to use cpu list %s\n", target.cpu_list); goto out; @@ -2479,12 +2463,6 @@ int cmd_stat(int argc, const char **argv) */ if (stat_config.aggr_mode == AGGR_THREAD) { thread_map__read_comms(evsel_list->core.threads); - if (target.system_wide) { - if (runtime_stat_new(&stat_config, - perf_thread_map__nr(evsel_list->core.threads))) { - goto out; - } - } } if (stat_config.aggr_mode == AGGR_NODE) @@ -2617,6 +2595,7 @@ out: iostat_release(evsel_list); zfree(&stat_config.walltime_run); + zfree(&stat_config.user_requested_cpu_list); if (smi_cost && smi_reset) sysfs__write_int(FREEZE_ON_SMI_PATH, 0); @@ -2624,7 +2603,6 @@ out: evlist__delete(evsel_list); metricgroup__rblist_exit(&stat_config.metric_events); - runtime_stat_delete(&stat_config); evlist__close_control(stat_config.ctl_fd, stat_config.ctl_fd_ack, &stat_config.ctl_fd_close); return status; diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index e2e9ad929baf..c36296bb7637 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -215,6 +215,19 @@ static struct per_pid *find_create_pid(struct timechart *tchart, int pid) return cursor; } +static struct per_pidcomm *create_pidcomm(struct per_pid *p) +{ + struct per_pidcomm *c; + + c = zalloc(sizeof(*c)); + if (!c) + return NULL; + p->current = c; + c->next = p->all; + p->all = c; + return c; +} + static void pid_set_comm(struct timechart *tchart, int pid, char *comm) { struct per_pid *p; @@ -233,12 +246,9 @@ static void pid_set_comm(struct timechart *tchart, int pid, char *comm) } c = c->next; } - c = zalloc(sizeof(*c)); + c = create_pidcomm(p); assert(c != NULL); c->comm = strdup(comm); - p->current = c; - c->next = p->all; - p->all = c; } static void pid_fork(struct timechart *tchart, int pid, int ppid, u64 timestamp) @@ -277,11 +287,8 @@ static void pid_put_sample(struct timechart *tchart, int pid, int type, p = find_create_pid(tchart, pid); c = p->current; if (!c) { - c = zalloc(sizeof(*c)); + c = create_pidcomm(p); assert(c != NULL); - p->current = c; - c->next = p->all; - p->all = c; } sample = zalloc(sizeof(*sample)); @@ -369,16 +376,13 @@ static void c_state_end(struct timechart *tchart, int cpu, u64 timestamp) tchart->power_events = pwr; } -static void p_state_change(struct timechart *tchart, int cpu, u64 timestamp, u64 new_freq) +static struct power_event *p_state_end(struct timechart *tchart, int cpu, + u64 timestamp) { - struct power_event *pwr; - - if (new_freq > 8000000) /* detect invalid data */ - return; + struct power_event *pwr = zalloc(sizeof(*pwr)); - pwr = zalloc(sizeof(*pwr)); if (!pwr) - return; + return NULL; pwr->state = cpus_pstate_state[cpu]; pwr->start_time = cpus_pstate_start_times[cpu]; @@ -386,11 +390,23 @@ static void p_state_change(struct timechart *tchart, int cpu, u64 timestamp, u64 pwr->cpu = cpu; pwr->type = PSTATE; pwr->next = tchart->power_events; - if (!pwr->start_time) pwr->start_time = tchart->first_time; tchart->power_events = pwr; + return pwr; +} + +static void p_state_change(struct timechart *tchart, int cpu, u64 timestamp, u64 new_freq) +{ + struct power_event *pwr; + + if (new_freq > 8000000) /* detect invalid data */ + return; + + pwr = p_state_end(tchart, cpu, timestamp); + if (!pwr) + return; cpus_pstate_state[cpu] = new_freq; cpus_pstate_start_times[cpu] = timestamp; @@ -698,22 +714,12 @@ static void end_sample_processing(struct timechart *tchart) #endif /* P state */ - pwr = zalloc(sizeof(*pwr)); + pwr = p_state_end(tchart, cpu, tchart->last_time); if (!pwr) return; - pwr->state = cpus_pstate_state[cpu]; - pwr->start_time = cpus_pstate_start_times[cpu]; - pwr->end_time = tchart->last_time; - pwr->cpu = cpu; - pwr->type = PSTATE; - pwr->next = tchart->power_events; - - if (!pwr->start_time) - pwr->start_time = tchart->first_time; if (!pwr->state) pwr->state = tchart->min_freq; - tchart->power_events = pwr; } } @@ -726,12 +732,9 @@ static int pid_begin_io_sample(struct timechart *tchart, int pid, int type, struct io_sample *prev; if (!c) { - c = zalloc(sizeof(*c)); + c = create_pidcomm(p); if (!c) return -ENOMEM; - p->current = c; - c->next = p->all; - p->all = c; } prev = c->io_samples; diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index fd8fd913c533..4b3ff7687236 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -136,10 +136,10 @@ static int perf_top__parse_source(struct perf_top *top, struct hist_entry *he) } notes = symbol__annotation(sym); - pthread_mutex_lock(¬es->lock); + mutex_lock(¬es->lock); if (!symbol__hists(sym, top->evlist->core.nr_entries)) { - pthread_mutex_unlock(¬es->lock); + mutex_unlock(¬es->lock); pr_err("Not enough memory for annotating '%s' symbol!\n", sym->name); sleep(1); @@ -155,7 +155,7 @@ static int perf_top__parse_source(struct perf_top *top, struct hist_entry *he) pr_err("Couldn't annotate %s: %s\n", sym->name, msg); } - pthread_mutex_unlock(¬es->lock); + mutex_unlock(¬es->lock); return err; } @@ -196,6 +196,7 @@ static void perf_top__record_precise_ip(struct perf_top *top, struct hist_entry *he, struct perf_sample *sample, struct evsel *evsel, u64 ip) + EXCLUSIVE_LOCKS_REQUIRED(he->hists->lock) { struct annotation *notes; struct symbol *sym = he->ms.sym; @@ -208,19 +209,19 @@ static void perf_top__record_precise_ip(struct perf_top *top, notes = symbol__annotation(sym); - if (pthread_mutex_trylock(¬es->lock)) + if (!mutex_trylock(¬es->lock)) return; err = hist_entry__inc_addr_samples(he, sample, evsel, ip); - pthread_mutex_unlock(¬es->lock); + mutex_unlock(¬es->lock); if (unlikely(err)) { /* * This function is now called with he->hists->lock held. * Release it before going to sleep. */ - pthread_mutex_unlock(&he->hists->lock); + mutex_unlock(&he->hists->lock); if (err == -ERANGE && !he->ms.map->erange_warned) ui__warn_map_erange(he->ms.map, sym, ip); @@ -230,7 +231,7 @@ static void perf_top__record_precise_ip(struct perf_top *top, sleep(1); } - pthread_mutex_lock(&he->hists->lock); + mutex_lock(&he->hists->lock); } } @@ -250,7 +251,7 @@ static void perf_top__show_details(struct perf_top *top) symbol = he->ms.sym; notes = symbol__annotation(symbol); - pthread_mutex_lock(¬es->lock); + mutex_lock(¬es->lock); symbol__calc_percent(symbol, evsel); @@ -271,7 +272,7 @@ static void perf_top__show_details(struct perf_top *top) if (more != 0) printf("%d lines not displayed, maybe increase display entries [e]\n", more); out_unlock: - pthread_mutex_unlock(¬es->lock); + mutex_unlock(¬es->lock); } static void perf_top__resort_hists(struct perf_top *t) @@ -724,13 +725,13 @@ repeat: static int hist_iter__top_callback(struct hist_entry_iter *iter, struct addr_location *al, bool single, void *arg) + EXCLUSIVE_LOCKS_REQUIRED(iter->he->hists->lock) { struct perf_top *top = arg; - struct hist_entry *he = iter->he; struct evsel *evsel = iter->evsel; if (perf_hpp_list.sym && single) - perf_top__record_precise_ip(top, he, iter->sample, evsel, al->addr); + perf_top__record_precise_ip(top, iter->he, iter->sample, evsel, al->addr); hist__account_cycles(iter->sample->branch_stack, al, iter->sample, !(top->record_opts.branch_stack & PERF_SAMPLE_BRANCH_ANY), @@ -836,12 +837,12 @@ static void perf_event__process_sample(struct perf_tool *tool, else iter.ops = &hist_iter_normal; - pthread_mutex_lock(&hists->lock); + mutex_lock(&hists->lock); if (hist_entry_iter__add(&iter, &al, top->max_stack, top) < 0) pr_err("Problem incrementing symbol period, skipping event\n"); - pthread_mutex_unlock(&hists->lock); + mutex_unlock(&hists->lock); } addr_location__put(&al); @@ -893,10 +894,10 @@ static void perf_top__mmap_read_idx(struct perf_top *top, int idx) perf_mmap__consume(&md->core); if (top->qe.rotate) { - pthread_mutex_lock(&top->qe.mutex); + mutex_lock(&top->qe.mutex); top->qe.rotate = false; - pthread_cond_signal(&top->qe.cond); - pthread_mutex_unlock(&top->qe.mutex); + cond_signal(&top->qe.cond); + mutex_unlock(&top->qe.mutex); } } @@ -1100,10 +1101,10 @@ static void *process_thread(void *arg) out = rotate_queues(top); - pthread_mutex_lock(&top->qe.mutex); + mutex_lock(&top->qe.mutex); top->qe.rotate = true; - pthread_cond_wait(&top->qe.cond, &top->qe.mutex); - pthread_mutex_unlock(&top->qe.mutex); + cond_wait(&top->qe.cond, &top->qe.mutex); + mutex_unlock(&top->qe.mutex); if (ordered_events__flush(out, OE_FLUSH__TOP)) pr_err("failed to process events\n"); @@ -1217,8 +1218,8 @@ static void init_process_thread(struct perf_top *top) ordered_events__set_copy_on_queue(&top->qe.data[0], true); ordered_events__set_copy_on_queue(&top->qe.data[1], true); top->qe.in = &top->qe.data[0]; - pthread_mutex_init(&top->qe.mutex, NULL); - pthread_cond_init(&top->qe.cond, NULL); + mutex_init(&top->qe.mutex); + cond_init(&top->qe.cond); } static int __cmd_top(struct perf_top *top) @@ -1349,7 +1350,7 @@ static int __cmd_top(struct perf_top *top) out_join: pthread_join(thread, NULL); out_join_thread: - pthread_cond_signal(&top->qe.cond); + cond_signal(&top->qe.cond); pthread_join(thread_process, NULL); return ret; } @@ -1706,6 +1707,7 @@ int cmd_top(int argc, const char **argv) if (evlist__create_maps(top.evlist, target) < 0) { ui__error("Couldn't create thread/CPU maps: %s\n", errno == ENOENT ? "No such process" : str_error_r(errno, errbuf, sizeof(errbuf))); + status = -errno; goto out_delete_evlist; } @@ -1758,11 +1760,13 @@ int cmd_top(int argc, const char **argv) if (top.sb_evlist == NULL) { pr_err("Couldn't create side band evlist.\n."); + status = -EINVAL; goto out_delete_evlist; } if (evlist__add_bpf_sb_event(top.sb_evlist, &perf_env)) { pr_err("Couldn't ask for PERF_RECORD_BPF_EVENT side band events.\n."); + status = -EINVAL; goto out_delete_evlist; } } diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 0bd9d01c0df9..d3c757769b96 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -615,11 +615,8 @@ bool strarray__strtoul_flags(struct strarray *sa, char *bf, size_t size, u64 *re if (isalpha(*tok) || *tok == '_') { if (!strarray__strtoul(sa, tok, toklen, &val)) return false; - } else { - bool is_hexa = tok[0] == 0 && (tok[1] = 'x' || tok[1] == 'X'); - - val = strtoul(tok, NULL, is_hexa ? 16 : 0); - } + } else + val = strtoul(tok, NULL, 0); *ret |= (1 << (val - 1)); @@ -2173,13 +2170,10 @@ static void thread__update_stats(struct thread *thread, struct thread_trace *ttr stats = inode->priv; if (stats == NULL) { - stats = malloc(sizeof(*stats)); + stats = zalloc(sizeof(*stats)); if (stats == NULL) return; - stats->nr_failures = 0; - stats->max_errno = 0; - stats->errnos = NULL; init_stats(&stats->stats); inode->priv = stats; } @@ -2762,11 +2756,7 @@ static size_t trace__fprintf_tp_fields(struct trace *trace, struct evsel *evsel, printed += scnprintf(bf + printed, size - printed, "%s", printed ? ", " : ""); - /* - * XXX Perhaps we should have a show_tp_arg_names, - * leaving show_arg_names just for syscalls? - */ - if (1 || trace->show_arg_names) + if (trace->show_arg_names) printed += scnprintf(bf + printed, size - printed, "%s: ", field->name); printed += syscall_arg_fmt__scnprintf_val(arg, bf + printed, size - printed, &syscall_arg, val); diff --git a/tools/perf/perf.c b/tools/perf/perf.c index c21b3973641a..7af135dea1cd 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c @@ -99,10 +99,16 @@ struct pager_config { int val; }; +static bool same_cmd_with_prefix(const char *var, struct pager_config *c, + const char *header) +{ + return (strstarts(var, header) && !strcmp(var + strlen(header), c->cmd)); +} + static int pager_command_config(const char *var, const char *value, void *data) { struct pager_config *c = data; - if (strstarts(var, "pager.") && !strcmp(var + 6, c->cmd)) + if (same_cmd_with_prefix(var, c, "pager.")) c->val = perf_config_bool(var, value); return 0; } @@ -121,9 +127,9 @@ static int check_pager_config(const char *cmd) static int browser_command_config(const char *var, const char *value, void *data) { struct pager_config *c = data; - if (strstarts(var, "tui.") && !strcmp(var + 4, c->cmd)) + if (same_cmd_with_prefix(var, c, "tui.")) c->val = perf_config_bool(var, value); - if (strstarts(var, "gtk.") && !strcmp(var + 4, c->cmd)) + if (same_cmd_with_prefix(var, c, "gtk.")) c->val = perf_config_bool(var, value) ? 2 : 0; return 0; } diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/branch.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/branch.json index 2f2d137f5f55..2f2d137f5f55 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/branch.json +++ b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/branch.json diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/bus.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/bus.json index 75d850b781ac..75d850b781ac 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/bus.json +++ b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/bus.json diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/cache.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/cache.json index 118c5cb0674b..118c5cb0674b 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/cache.json +++ b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/cache.json diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/dpu.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/dpu.json index b8e402a91bdd..b8e402a91bdd 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/dpu.json +++ b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/dpu.json diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/exception.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/exception.json index 27c3fe9c831a..27c3fe9c831a 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/exception.json +++ b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/exception.json diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/ifu.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/ifu.json index 13178c5dca14..13178c5dca14 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/ifu.json +++ b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/ifu.json diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/instruction.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/instruction.json index 2e0d60779dce..2e0d60779dce 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/instruction.json +++ b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/instruction.json diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/memory.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/memory.json index 18d527f7fad4..18d527f7fad4 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/memory.json +++ b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/memory.json diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/pipeline.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/pipeline.json index eeac798d403a..eeac798d403a 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a65/pipeline.json +++ b/tools/perf/pmu-events/arch/arm64/arm/cortex-a65-e1/pipeline.json diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a76-n1/memory.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a76-n1/memory.json index 20a929e7728d..5bed2514b245 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a76-n1/memory.json +++ b/tools/perf/pmu-events/arch/arm64/arm/cortex-a76-n1/memory.json @@ -4,6 +4,9 @@ "ArchStdEvent": "MEM_ACCESS" }, { + "ArchStdEvent": "REMOTE_ACCESS" + }, + { "ArchStdEvent": "MEM_ACCESS_RD" }, { diff --git a/tools/perf/pmu-events/arch/arm64/arm/cortex-a76-n1/other.json b/tools/perf/pmu-events/arch/arm64/arm/cortex-a76-n1/other.json deleted file mode 100644 index 20d8365756c5..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/cortex-a76-n1/other.json +++ /dev/null @@ -1,5 +0,0 @@ -[ - { - "ArchStdEvent": "REMOTE_ACCESS" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/branch.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/branch.json deleted file mode 100644 index 2f2d137f5f55..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/branch.json +++ /dev/null @@ -1,17 +0,0 @@ -[ - { - "ArchStdEvent": "BR_MIS_PRED" - }, - { - "ArchStdEvent": "BR_PRED" - }, - { - "ArchStdEvent": "BR_IMMED_SPEC" - }, - { - "ArchStdEvent": "BR_RETURN_SPEC" - }, - { - "ArchStdEvent": "BR_INDIRECT_SPEC" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/bus.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/bus.json deleted file mode 100644 index 75d850b781ac..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/bus.json +++ /dev/null @@ -1,17 +0,0 @@ -[ - { - "ArchStdEvent": "CPU_CYCLES" - }, - { - "ArchStdEvent": "BUS_ACCESS" - }, - { - "ArchStdEvent": "BUS_CYCLES" - }, - { - "ArchStdEvent": "BUS_ACCESS_RD" - }, - { - "ArchStdEvent": "BUS_ACCESS_WR" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/cache.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/cache.json deleted file mode 100644 index 3ad15e3a93a9..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/cache.json +++ /dev/null @@ -1,107 +0,0 @@ -[ - { - "ArchStdEvent": "L1I_CACHE_REFILL" - }, - { - "ArchStdEvent": "L1I_TLB_REFILL" - }, - { - "ArchStdEvent": "L1D_CACHE_REFILL" - }, - { - "ArchStdEvent": "L1D_CACHE" - }, - { - "ArchStdEvent": "L1D_TLB_REFILL" - }, - { - "ArchStdEvent": "L1I_CACHE" - }, - { - "ArchStdEvent": "L1D_CACHE_WB" - }, - { - "ArchStdEvent": "L2D_CACHE" - }, - { - "ArchStdEvent": "L2D_CACHE_REFILL" - }, - { - "ArchStdEvent": "L2D_CACHE_WB" - }, - { - "ArchStdEvent": "L1D_CACHE_ALLOCATE" - }, - { - "ArchStdEvent": "L2D_CACHE_ALLOCATE" - }, - { - "ArchStdEvent": "L1D_TLB" - }, - { - "ArchStdEvent": "L1I_TLB" - }, - { - "ArchStdEvent": "L3D_CACHE_ALLOCATE" - }, - { - "ArchStdEvent": "L3D_CACHE_REFILL" - }, - { - "ArchStdEvent": "L3D_CACHE" - }, - { - "ArchStdEvent": "L2D_TLB_REFILL" - }, - { - "ArchStdEvent": "L2D_TLB" - }, - { - "ArchStdEvent": "DTLB_WALK" - }, - { - "ArchStdEvent": "ITLB_WALK" - }, - { - "ArchStdEvent": "LL_CACHE_RD" - }, - { - "ArchStdEvent": "LL_CACHE_MISS_RD" - }, - { - "ArchStdEvent": "L1D_CACHE_RD" - }, - { - "ArchStdEvent": "L1D_CACHE_WR" - }, - { - "ArchStdEvent": "L1D_CACHE_REFILL_RD" - }, - { - "ArchStdEvent": "L1D_CACHE_REFILL_WR" - }, - { - "ArchStdEvent": "L1D_CACHE_REFILL_INNER" - }, - { - "ArchStdEvent": "L1D_CACHE_REFILL_OUTER" - }, - { - "ArchStdEvent": "L2D_CACHE_RD" - }, - { - "ArchStdEvent": "L2D_CACHE_WR" - }, - { - "ArchStdEvent": "L2D_CACHE_REFILL_RD" - }, - { - "ArchStdEvent": "L2D_CACHE_REFILL_WR" - }, - { - "ArchStdEvent": "L3D_CACHE_RD" - }, - { - "ArchStdEvent": "L3D_CACHE_REFILL_RD" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/exception.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/exception.json deleted file mode 100644 index 27c3fe9c831a..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/exception.json +++ /dev/null @@ -1,14 +0,0 @@ -[ - { - "ArchStdEvent": "EXC_TAKEN" - }, - { - "ArchStdEvent": "MEMORY_ERROR" - }, - { - "ArchStdEvent": "EXC_IRQ" - }, - { - "ArchStdEvent": "EXC_FIQ" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/instruction.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/instruction.json deleted file mode 100644 index 6c3b8f772e7f..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/instruction.json +++ /dev/null @@ -1,65 +0,0 @@ -[ - { - "ArchStdEvent": "SW_INCR" - }, - { - "ArchStdEvent": "LD_RETIRED" - }, - { - "ArchStdEvent": "ST_RETIRED" - }, - { - "ArchStdEvent": "INST_RETIRED" - }, - { - "ArchStdEvent": "EXC_RETURN" - }, - { - "ArchStdEvent": "CID_WRITE_RETIRED" - }, - { - "ArchStdEvent": "PC_WRITE_RETIRED" - }, - { - "ArchStdEvent": "BR_IMMED_RETIRED" - }, - { - "ArchStdEvent": "BR_RETURN_RETIRED" - }, - { - "ArchStdEvent": "INST_SPEC" - }, - { - "ArchStdEvent": "TTBR_WRITE_RETIRED" - }, - { - "ArchStdEvent": "BR_RETIRED" - }, - { - "ArchStdEvent": "BR_MIS_PRED_RETIRED" - }, - { - "ArchStdEvent": "LD_SPEC" - }, - { - "ArchStdEvent": "ST_SPEC" - }, - { - "ArchStdEvent": "LDST_SPEC" - }, - { - "ArchStdEvent": "DP_SPEC" - }, - { - "ArchStdEvent": "ASE_SPEC" - }, - { - "ArchStdEvent": "VFP_SPEC" - }, - { - "ArchStdEvent": "CRYPTO_SPEC" - }, - { - "ArchStdEvent": "ISB_SPEC" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/memory.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/memory.json deleted file mode 100644 index 78ed6dfcedc1..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/memory.json +++ /dev/null @@ -1,23 +0,0 @@ -[ - { - "ArchStdEvent": "MEM_ACCESS" - }, - { - "ArchStdEvent": "REMOTE_ACCESS_RD" - }, - { - "ArchStdEvent": "MEM_ACCESS_RD" - }, - { - "ArchStdEvent": "MEM_ACCESS_WR" - }, - { - "ArchStdEvent": "UNALIGNED_LD_SPEC" - }, - { - "ArchStdEvent": "UNALIGNED_ST_SPEC" - }, - { - "ArchStdEvent": "UNALIGNED_LDST_SPEC" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/pipeline.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/pipeline.json deleted file mode 100644 index eeac798d403a..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/pipeline.json +++ /dev/null @@ -1,8 +0,0 @@ -[ - { - "ArchStdEvent": "STALL_FRONTEND" - }, - { - "ArchStdEvent": "STALL_BACKEND" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/spe.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/spe.json deleted file mode 100644 index 20f2165c85fe..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-e1/spe.json +++ /dev/null @@ -1,14 +0,0 @@ -[ - { - "ArchStdEvent": "SAMPLE_POP" - }, - { - "ArchStdEvent": "SAMPLE_FEED" - }, - { - "ArchStdEvent": "SAMPLE_FILTRATE" - }, - { - "ArchStdEvent": "SAMPLE_COLLISION" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-n2/memory.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-n2/memory.json index e522113aeb96..7b2b21ac150f 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-n2/memory.json +++ b/tools/perf/pmu-events/arch/arm64/arm/neoverse-n2/memory.json @@ -3,6 +3,9 @@ "ArchStdEvent": "MEM_ACCESS" }, { + "ArchStdEvent": "REMOTE_ACCESS" + }, + { "ArchStdEvent": "MEM_ACCESS_RD" }, { diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-n2/other.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-n2/other.json deleted file mode 100644 index 20d8365756c5..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-n2/other.json +++ /dev/null @@ -1,5 +0,0 @@ -[ - { - "ArchStdEvent": "REMOTE_ACCESS" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/instruction.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/instruction.json index 25825e14c535..e29b88fb7f24 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/instruction.json +++ b/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/instruction.json @@ -85,5 +85,35 @@ }, { "ArchStdEvent": "RC_ST_SPEC" + }, + { + "ArchStdEvent": "ASE_INST_SPEC" + }, + { + "ArchStdEvent": "SVE_INST_SPEC" + }, + { + "ArchStdEvent": "SVE_PRED_SPEC" + }, + { + "ArchStdEvent": "SVE_PRED_EMPTY_SPEC" + }, + { + "ArchStdEvent": "SVE_PRED_FULL_SPEC" + }, + { + "ArchStdEvent": "SVE_PRED_PARTIAL_SPEC" + }, + { + "ArchStdEvent": "SVE_LDFF_SPEC" + }, + { + "ArchStdEvent": "SVE_LDFF_FAULT_SPEC" + }, + { + "ArchStdEvent": "FP_SCALE_OPS_SPEC" + }, + { + "ArchStdEvent": "FP_FIXED_OPS_SPEC" } ] diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/memory.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/memory.json index e3d08f1f7c92..5aff6e93c1ad 100644 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/memory.json +++ b/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/memory.json @@ -3,6 +3,9 @@ "ArchStdEvent": "MEM_ACCESS" }, { + "ArchStdEvent": "REMOTE_ACCESS" + }, + { "ArchStdEvent": "MEM_ACCESS_RD" }, { diff --git a/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/other.json b/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/other.json deleted file mode 100644 index 20d8365756c5..000000000000 --- a/tools/perf/pmu-events/arch/arm64/arm/neoverse-v1/other.json +++ /dev/null @@ -1,5 +0,0 @@ -[ - { - "ArchStdEvent": "REMOTE_ACCESS" - } -] diff --git a/tools/perf/pmu-events/arch/arm64/mapfile.csv b/tools/perf/pmu-events/arch/arm64/mapfile.csv index 406f6edd4e12..ad502d00f460 100644 --- a/tools/perf/pmu-events/arch/arm64/mapfile.csv +++ b/tools/perf/pmu-events/arch/arm64/mapfile.csv @@ -17,7 +17,8 @@ 0x00000000420f1000,v1,arm/cortex-a53,core 0x00000000410fd040,v1,arm/cortex-a35,core 0x00000000410fd050,v1,arm/cortex-a55,core -0x00000000410fd060,v1,arm/cortex-a65,core +0x00000000410fd060,v1,arm/cortex-a65-e1,core +0x00000000410fd4a0,v1,arm/cortex-a65-e1,core 0x00000000410fd070,v1,arm/cortex-a57-a72,core 0x00000000410fd080,v1,arm/cortex-a57-a72,core 0x00000000410fd090,v1,arm/cortex-a73,core @@ -34,7 +35,6 @@ 0x00000000410fd470,v1,arm/cortex-a710,core 0x00000000410fd480,v1,arm/cortex-x2,core 0x00000000410fd490,v1,arm/neoverse-n2,core -0x00000000410fd4a0,v1,arm/neoverse-e1,core 0x00000000420f5160,v1,cavium/thunderx2,core 0x00000000430f0af0,v1,cavium/thunderx2,core 0x00000000460f0010,v1,fujitsu/a64fx,core diff --git a/tools/perf/pmu-events/arch/test/test_soc/cpu/metrics.json b/tools/perf/pmu-events/arch/test/test_soc/cpu/metrics.json index 42d9b5242fd7..70ec8caaaf6f 100644 --- a/tools/perf/pmu-events/arch/test/test_soc/cpu/metrics.json +++ b/tools/perf/pmu-events/arch/test/test_soc/cpu/metrics.json @@ -34,15 +34,15 @@ "MetricName": "DCache_L2_All_Miss" }, { - "MetricExpr": "dcache_l2_all_hits + dcache_l2_all_miss", + "MetricExpr": "DCache_L2_All_Hits + DCache_L2_All_Miss", "MetricName": "DCache_L2_All" }, { - "MetricExpr": "d_ratio(dcache_l2_all_hits, dcache_l2_all)", + "MetricExpr": "d_ratio(DCache_L2_All_Hits, DCache_L2_All)", "MetricName": "DCache_L2_Hits" }, { - "MetricExpr": "d_ratio(dcache_l2_all_miss, dcache_l2_all)", + "MetricExpr": "d_ratio(DCache_L2_All_Miss, DCache_L2_All)", "MetricName": "DCache_L2_Misses" }, { diff --git a/tools/perf/pmu-events/arch/x86/alderlake/adl-metrics.json b/tools/perf/pmu-events/arch/x86/alderlake/adl-metrics.json index 095dd8c7f161..e06d26ad5138 100644 --- a/tools/perf/pmu-events/arch/x86/alderlake/adl-metrics.json +++ b/tools/perf/pmu-events/arch/x86/alderlake/adl-metrics.json @@ -1,22 +1,852 @@ [ { + "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", + "MetricExpr": "topdown\\-fe\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) - INT_MISC.UOP_DROPPING / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. Sample with: FRONTEND_RETIRED.LATENCY_GE_4_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "(topdown\\-fetch\\-lat / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) - INT_MISC.UOP_DROPPING / SLOTS)", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: FRONTEND_RETIRED.LATENCY_GE_16_PS;FRONTEND_RETIRED.LATENCY_GE_8_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses", + "MetricExpr": "ICACHE_DATA.STALLS / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses. Sample with: FRONTEND_RETIRED.L2_MISS_PS;FRONTEND_RETIRED.L1I_MISS_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "ICACHE_TAG.STALLS / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: FRONTEND_RETIRED.STLB_MISS_PS;FRONTEND_RETIRED.ITLB_MISS_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "INT_MISC.CLEAR_RESTEER_CYCLES / CLKS + tma_unknown_branches", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage", + "MetricExpr": "(tma_branch_mispredicts / tma_bad_speculation) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears", + "MetricExpr": "(1 - (tma_branch_mispredicts / tma_bad_speculation)) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "INT_MISC.UNKNOWN_BRANCH_CYCLES / CLKS", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: FRONTEND_RETIRED.UNKNOWN_BRANCH", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty. Sample with: FRONTEND_RETIRED.DSB_MISS_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "DECODE.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "3 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: FRONTEND_RETIRED.MS_FLOWS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "max(0, tma_frontend_bound - tma_fetch_latency)", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend. Sample with: FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_2_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.MITE_CYCLES_ANY - IDQ.MITE_CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck. Sample with: FRONTEND_RETIRED.ANY_DSB_MISS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles where decoder-0 was the only active decoder", + "MetricExpr": "(cpu_core@INST_DECODED.DECODERS\\,cmask\\=1@ - cpu_core@INST_DECODED.DECODERS\\,cmask\\=2@) / CORE_CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_decoder0_alone", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.DSB_CYCLES_ANY - IDQ.DSB_CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to LSD (Loop Stream Detector) unit", + "MetricExpr": "(LSD.CYCLES_ACTIVE - LSD.CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "FetchBW;LSD;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_lsd", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to LSD (Loop Stream Detector) unit. LSD typically does well sustaining Uop supply. However; in some rare cases; optimal uop-delivery could not be reached for small loops whose size (in terms of number of uops) does not suit well the LSD structure.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", + "MetricExpr": "max(1 - (tma_frontend_bound + tma_backend_bound + tma_retiring), 0)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "topdown\\-br\\-mispredict / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: TOPDOWN.BR_MISPREDICT_SLOTS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "max(0, tma_bad_speculation - tma_branch_mispredicts)", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", + "MetricExpr": "topdown\\-be\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. Sample with: TOPDOWN.BACKEND_BOUND_SLOTS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "topdown\\-mem\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((EXE_ACTIVITY.BOUND_ON_LOADS - MEMORY_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_RETIRED.L1_HIT_PS;MEM_LOAD_RETIRED.FB_HIT_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "min(7 * cpu_core@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE, max(CYCLE_ACTIVITY.CYCLES_MEM_ANY - MEMORY_ACTIVITY.CYCLES_L1D_MISS, 0)) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_INST_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_load - tma_load_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_hit", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk", + "MetricExpr": "DTLB_LOAD_MISSES.WALK_ACTIVE / CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_miss", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(16 * max(0, MEM_INST_RETIRED.LOCK_LOADS - L2_RQSTS.ALL_RFO) + (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES) * (10 * L2_RQSTS.RFO_HIT + min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO))) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_INST_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_INST_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "L1D_PEND_MISS.FB_FULL / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "(MEMORY_ACTIVITY.STALLS_L1D_MISS - MEMORY_ACTIVITY.STALLS_L2_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEMORY_ACTIVITY.STALLS_L2_MISS - MEMORY_ACTIVITY.STALLS_L3_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "((25 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD * (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM / (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM + OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD))) + (24 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "(24 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD + MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD * (1 - (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM / (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM + OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD)))) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "(9 * Average_Frequency) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "(XQ.FULL_CYCLES + L1D_PEND_MISS.L2_STALLS) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(MEMORY_ACTIVITY.STALLS_L3_MISS / CLKS)", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu_core@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "EXE_ACTIVITY.BOUND_ON_STORES / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_INST_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((MEM_STORE_RETIRED.L2_HIT * 10 * (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES))) + (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "(28 * Average_Frequency) * OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "MEM_INST_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_INST_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores", + "MetricExpr": "9 * OCR.STREAMING_WR.ANY_RESPONSE / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_streaming_stores", + "PublicDescription": "This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should Streaming stores be a bottleneck. Sample with: OCR.STREAMING_WR.ANY_RESPONSE", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(7 * cpu_core@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE) / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_INST_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_store - tma_store_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_hit", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk", + "MetricExpr": "DTLB_STORE_MISSES.WALK_ACTIVE / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_miss", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "max(0, tma_backend_bound - tma_memory_bound)", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.DIVIDER_ACTIVE / CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_ACTIVE", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "(cpu_core@EXE_ACTIVITY.3_PORTS_UTIL\\,umask\\=0x80@ + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * cpu_core@EXE_ACTIVITY.2_PORTS_UTIL\\,umask\\=0xc@)) / CLKS if (ARITH.DIVIDER_ACTIVE < (CYCLE_ACTIVITY.STALLS_TOTAL - EXE_ACTIVITY.BOUND_ON_LOADS)) else (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * cpu_core@EXE_ACTIVITY.2_PORTS_UTIL\\,umask\\=0xc@) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "cpu_core@EXE_ACTIVITY.3_PORTS_UTIL\\,umask\\=0x80@ / CLKS + tma_serializing_operation * (CYCLE_ACTIVITY.STALLS_TOTAL - EXE_ACTIVITY.BOUND_ON_LOADS) / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations", + "MetricExpr": "RESOURCE_STALLS.SCOREBOARD / CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_serializing_operation", + "PublicDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out-of-order execution which may limit performance. Sample with: RESOURCE_STALLS.SCOREBOARD", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions", + "MetricExpr": "CPU_CLK_UNHALTED.PAUSE / CLKS", + "MetricGroup": "TopdownL6;tma_serializing_operation_group", + "MetricName": "tma_slow_pause", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions. Sample with: CPU_CLK_UNHALTED.PAUSE_INST", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to LFENCE Instructions.", + "MetricExpr": "13 * MISC2_RETIRED.LFENCE / CLKS", + "MetricGroup": "TopdownL6;tma_serializing_operation_group", + "MetricName": "tma_memory_fence", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued", + "MetricExpr": "160 * ASSISTS.SSE_AVX_MIX / CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_mixing_vectors", + "PublicDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "EXE_ACTIVITY.1_PORTS_UTIL / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful. Sample with: EXE_ACTIVITY.1_PORTS_UTIL", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "EXE_ACTIVITY.2_PORTS_UTIL / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop. Sample with: EXE_ACTIVITY.2_PORTS_UTIL", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "UOPS_EXECUTED.CYCLES_GE_3 / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Sample with: UOPS_EXECUTED.CYCLES_GE_3", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED.PORT_0 + UOPS_DISPATCHED.PORT_1 + UOPS_DISPATCHED.PORT_5_11 + UOPS_DISPATCHED.PORT_6) / (5 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED.PORT_0", + "MetricExpr": "UOPS_DISPATCHED.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED.PORT_1", + "MetricExpr": "UOPS_DISPATCHED.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED.PORT_6", + "MetricExpr": "UOPS_DISPATCHED.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3_10", + "MetricExpr": "UOPS_DISPATCHED.PORT_2_3_10 / (3 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations Sample with: UOPS_DISPATCHED.PORT_7_8", + "MetricExpr": "(UOPS_DISPATCHED.PORT_4_9 + UOPS_DISPATCHED.PORT_7_8) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", + "MetricExpr": "topdown\\-retiring / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.SLOTS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "max(0, tma_retiring - tma_heavy_operations)", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "tma_retiring * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents overall Integer (Int) select operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_int_vector_128b + tma_int_vector_256b + tma_shuffles", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_int_operations", + "PublicDescription": "This metric represents overall Integer (Int) select operations fraction the CPU has executed (retired). Vector/Matrix Int operations and shuffles are counted. Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents 128-bit vector Integer ADD/SUB/SAD or VNNI (Vector Neural Network Instructions) uops fraction the CPU has retired.", + "MetricExpr": "(INT_VEC_RETIRED.ADD_128 + INT_VEC_RETIRED.VNNI_128) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;IntVector;Pipeline;TopdownL4;tma_int_operations_group", + "MetricName": "tma_int_vector_128b", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents 256-bit vector Integer ADD/SUB/SAD or VNNI (Vector Neural Network Instructions) uops fraction the CPU has retired.", + "MetricExpr": "(INT_VEC_RETIRED.ADD_256 + INT_VEC_RETIRED.MUL_256 + INT_VEC_RETIRED.VNNI_256) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;IntVector;Pipeline;TopdownL4;tma_int_operations_group", + "MetricName": "tma_int_vector_256b", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents Shuffle (cross \"vector lane\" data transfers) uops fraction the CPU has retired.", + "MetricExpr": "INT_VEC_RETIRED.SHUFFLES / (tma_retiring * SLOTS)", + "MetricGroup": "HPC;Pipeline;TopdownL4;tma_int_operations_group", + "MetricName": "tma_shuffles", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", + "MetricExpr": "tma_light_operations * MEM_UOP_RETIRED.ANY / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_memory_operations", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions", + "MetricExpr": "tma_light_operations * INST_RETIRED.MACRO_FUSED / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fused_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions. The instruction pairs of CMP+JCC or DEC+JCC are commonly used examples.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused", + "MetricExpr": "tma_light_operations * (BR_INST_RETIRED.ALL_BRANCHES - INST_RETIRED.MACRO_FUSED) / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_non_fused_branches", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused. Non-conditional branches like direct JMP or CALL would count here. Can be used to examine fusible conditional jumps that were not fused.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions", + "MetricExpr": "tma_light_operations * INST_RETIRED.NOP / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_nop_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body. Sample with: INST_RETIRED.NOP", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", + "MetricExpr": "max(0, tma_light_operations - (tma_fp_arith + tma_int_operations + tma_memory_operations + tma_fused_instructions + tma_non_fused_branches + tma_nop_instructions))", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_other_light_ops", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "topdown\\-heavy\\-ops / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences. Sample with: UOPS_RETIRED.HEAVY", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops", + "MetricExpr": "tma_heavy_operations - tma_microcode_sequencer", + "MetricGroup": "TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_few_uops_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "UOPS_RETIRED.MS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: UOPS_RETIRED.MS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * cpu_core@ASSISTS.ANY\\,umask\\=0x1B@ / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: ASSISTS.ANY", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Page Faults", + "MetricExpr": "99 * ASSISTS.PAGE_FAULT / SLOTS", + "MetricGroup": "TopdownL5;tma_assists_group", + "MetricName": "tma_page_faults", + "PublicDescription": "This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Page Faults. A Page Fault may apply on first application access to a memory page. Note operating system handling of page faults accounts for the majority of its cost.", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Floating Point (FP) Assists", + "MetricExpr": "30 * ASSISTS.FP / SLOTS", + "MetricGroup": "HPC;TopdownL5;tma_assists_group", + "MetricName": "tma_fp_assists", + "PublicDescription": "This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Floating Point (FP) Assists. FP Assist may apply when working with very small floating point values (so-called denormals).", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops as a result of handing SSE to AVX* or AVX* to SSE transition Assists. ", + "MetricExpr": "63 * ASSISTS.SSE_AVX_MIX / SLOTS", + "MetricGroup": "HPC;TopdownL5;tma_assists_group", + "MetricName": "tma_avx_assists", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources. Sample with: FRONTEND_RETIRED.MS_FLOWS", + "ScaleUnit": "100%", + "Unit": "cpu_core" + }, + { + "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", + "MetricExpr": "100 * (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches))", + "MetricGroup": "Bad;BadSpec;BrMispredicts", + "MetricName": "Mispredictions", + "Unit": "cpu_core" + }, + { + "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_mem_bandwidth / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_sq_full / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full))) + (tma_l1_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_fb_full / (tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) ", + "MetricGroup": "Mem;MemoryBW;Offcore", + "MetricName": "Memory_Bandwidth", + "Unit": "cpu_core" + }, + { + "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_mem_latency / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_l3_hit_latency / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full)) + (tma_l2_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)))", + "MetricGroup": "Mem;MemoryLat;Offcore", + "MetricName": "Memory_Latency", + "Unit": "cpu_core" + }, + { + "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", + "MetricExpr": "100 * tma_memory_bound * ((tma_l1_bound / max(tma_memory_bound, tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_dtlb_load / max(tma_l1_bound, tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) + (tma_store_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_dtlb_store / (tma_dtlb_store + tma_false_sharing + tma_split_stores + tma_store_latency + tma_streaming_stores))) ", + "MetricGroup": "Mem;MemoryTLB;Offcore", + "MetricName": "Memory_Data_TLBs", + "Unit": "cpu_core" + }, + { "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.COND + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL) ) / TOPDOWN.SLOTS)", + "MetricExpr": "100 * ((BR_INST_RETIRED.COND + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL)) / SLOTS)", "MetricGroup": "Ret", "MetricName": "Branching_Overhead", "Unit": "cpu_core" }, { + "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", + "MetricExpr": "100 * tma_fetch_latency * (tma_itlb_misses + tma_icache_misses + tma_unknown_branches) / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)", + "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB", + "MetricName": "Big_Code", + "Unit": "cpu_core" + }, + { + "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", + "MetricExpr": "100 * (tma_frontend_bound - tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) - Big_Code", + "MetricGroup": "Fed;FetchBW;Frontend", + "MetricName": "Instruction_Fetch_BW", + "Unit": "cpu_core" + }, + { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC", "Unit": "cpu_core" }, { + "BriefDescription": "Uops Per Instruction", + "MetricExpr": "(tma_retiring * SLOTS) / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;Ret;Retire", + "MetricName": "UPI", + "Unit": "cpu_core" + }, + { + "BriefDescription": "Instruction per taken branch", + "MetricExpr": "(tma_retiring * SLOTS) / BR_INST_RETIRED.NEAR_TAKEN", + "MetricGroup": "Branches;Fed;FetchBW", + "MetricName": "UpTB", + "Unit": "cpu_core" + }, + { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI", "Unit": "cpu_core" }, @@ -30,14 +860,14 @@ { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", "MetricExpr": "TOPDOWN.SLOTS", - "MetricGroup": "TmaL1", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS", "Unit": "cpu_core" }, { "BriefDescription": "Fraction of Physical Core issue-slots utilized by this Logical Processor", - "MetricExpr": "TOPDOWN.SLOTS / ( TOPDOWN.SLOTS / 2 ) if #SMT_on else 1", - "MetricGroup": "SMT;TmaL1", + "MetricExpr": "SLOTS / (TOPDOWN.SLOTS / 2) if #SMT_on else 1", + "MetricGroup": "SMT;tma_L1_group", "MetricName": "Slots_Utilization", "Unit": "cpu_core" }, @@ -51,21 +881,21 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.DISTRIBUTED", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC", "Unit": "cpu_core" }, { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / CPU_CLK_UNHALTED.DISTRIBUTED", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc", "Unit": "cpu_core" }, { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( FP_ARITH_DISPATCHED.PORT_0 + FP_ARITH_DISPATCHED.PORT_1 + FP_ARITH_DISPATCHED.PORT_5 ) / ( 2 * CPU_CLK_UNHALTED.DISTRIBUTED )", + "MetricExpr": "(FP_ARITH_DISPATCHED.PORT_0 + FP_ARITH_DISPATCHED.PORT_1 + FP_ARITH_DISPATCHED.PORT_5) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common).", @@ -73,12 +903,19 @@ }, { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2 ) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP", "Unit": "cpu_core" }, { + "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", + "MetricExpr": "(1 - tma_core_bound / tma_ports_utilization if tma_core_bound < tma_ports_utilization else 1) if SMT_2T_Utilization > 0.5 else 0", + "MetricGroup": "Cor;SMT", + "MetricName": "Core_Bound_Likely", + "Unit": "cpu_core" + }, + { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", "MetricExpr": "CPU_CLK_UNHALTED.DISTRIBUTED", "MetricGroup": "SMT", @@ -129,14 +966,14 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP", "Unit": "cpu_core" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW.", @@ -160,7 +997,7 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting.", @@ -168,7 +1005,7 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting.", @@ -182,13 +1019,20 @@ "Unit": "cpu_core" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions", "Unit": "cpu_core" }, { + "BriefDescription": "Average number of Uops retired in cycles where at least one uop has retired.", + "MetricExpr": "(tma_retiring * SLOTS) / cpu_core@UOPS_RETIRED.SLOTS\\,cmask\\=1@", + "MetricGroup": "Pipeline;Ret", + "MetricName": "Retire", + "Unit": "cpu_core" + }, + { "BriefDescription": "Estimated fraction of retirement-cycles dealing with repeat instructions", "MetricExpr": "INST_RETIRED.REP_ITERATION / cpu_core@UOPS_RETIRED.SLOTS\\,cmask\\=1@", "MetricGroup": "Pipeline;Ret", @@ -238,6 +1082,13 @@ "Unit": "cpu_core" }, { + "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", + "MetricExpr": "100 * (tma_fetch_latency * tma_dsb_switches / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches) + tma_fetch_bandwidth * tma_mite / (tma_dsb + tma_lsd + tma_mite))", + "MetricGroup": "DSBmiss;Fed", + "MetricName": "DSB_Misses", + "Unit": "cpu_core" + }, + { "BriefDescription": "Number of Instructions per non-speculative DSB miss (lower number means higher occurrence rate)", "MetricExpr": "INST_RETIRED.ANY / FRONTEND_RETIRED.ANY_DSB_MISS", "MetricGroup": "DSBmiss;Fed", @@ -252,6 +1103,13 @@ "Unit": "cpu_core" }, { + "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricGroup": "Bad;BrMispredicts", + "MetricName": "Branch_Misprediction_Cost", + "Unit": "cpu_core" + }, + { "BriefDescription": "Fraction of branches that are non-taken conditionals", "MetricExpr": "BR_INST_RETIRED.COND_NTAKEN / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", @@ -267,7 +1125,7 @@ }, { "BriefDescription": "Fraction of branches that are CALL or RET", - "MetricExpr": "( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "CallRet", "Unit": "cpu_core" @@ -281,7 +1139,7 @@ }, { "BriefDescription": "Fraction of branches of other types (not individually covered by other metrics in Info.Branches group)", - "MetricExpr": "1 - ( (BR_INST_RETIRED.COND_NTAKEN / BR_INST_RETIRED.ALL_BRANCHES) + (BR_INST_RETIRED.COND_TAKEN / BR_INST_RETIRED.ALL_BRANCHES) + (( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES) + ((BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES) )", + "MetricExpr": "1 - (Cond_NT + Cond_TK + CallRet + Jump)", "MetricGroup": "Bad;Branches", "MetricName": "Other_Branches", "Unit": "cpu_core" @@ -296,77 +1154,77 @@ { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP", "Unit": "cpu_core" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI", "Unit": "cpu_core" }, { "BriefDescription": "L1 cache true misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.ALL_DEMAND_DATA_RD / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI_Load", "Unit": "cpu_core" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI", "Unit": "cpu_core" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All", "Unit": "cpu_core" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load", "Unit": "cpu_core" }, { "BriefDescription": "L2 cache hits per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( L2_RQSTS.REFERENCES - L2_RQSTS.MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricExpr": "1000 * (L2_RQSTS.REFERENCES - L2_RQSTS.MISS) / INST_RETIRED.ANY", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_All", "Unit": "cpu_core" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load", "Unit": "cpu_core" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI", "Unit": "cpu_core" }, { "BriefDescription": "Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries)", "MetricExpr": "1000 * MEM_LOAD_RETIRED.FB_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "FB_HPKI", "Unit": "cpu_core" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING ) / ( 4 * CPU_CLK_UNHALTED.DISTRIBUTED )", + "MetricExpr": "(ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING) / (4 * CORE_CLKS)", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization", "Unit": "cpu_core" @@ -401,28 +1259,28 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T", "Unit": "cpu_core" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T", "Unit": "cpu_core" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T", "Unit": "cpu_core" }, { "BriefDescription": "Average per-thread data access bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * OFFCORE_REQUESTS.ALL_REQUESTS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Access_BW", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "L3_Cache_Access_BW_1T", "Unit": "cpu_core" @@ -436,14 +1294,14 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency", "Unit": "cpu_core" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine.", @@ -451,7 +1309,7 @@ }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization", "Unit": "cpu_core" @@ -479,7 +1337,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * ( arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@ ) / 1000000 / duration_time / 1000", + "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1000000 / duration_time / 1000", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use", "Unit": "cpu_core" @@ -500,41 +1358,408 @@ }, { "BriefDescription": "Counts the number of issue slots that were not consumed by the backend due to frontend stalls.", - "MetricExpr": "TOPDOWN_FE_BOUND.ALL / (5 * CPU_CLK_UNHALTED.CORE)", + "MetricExpr": "TOPDOWN_FE_BOUND.ALL / SLOTS", "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", + "MetricName": "tma_frontend_bound", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not delivered by the frontend due to frontend bandwidth restrictions due to decode, predecode, cisc, and other limitations.", + "MetricExpr": "TOPDOWN_FE_BOUND.FRONTEND_LATENCY / SLOTS", + "MetricGroup": "TopdownL2;tma_frontend_bound_group", + "MetricName": "tma_frontend_latency", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not delivered by the frontend due to instruction cache misses.", + "MetricExpr": "TOPDOWN_FE_BOUND.ICACHE / SLOTS", + "MetricGroup": "TopdownL3;tma_frontend_latency_group", + "MetricName": "tma_icache", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not delivered by the frontend due to Instruction Table Lookaside Buffer (ITLB) misses.", + "MetricExpr": "TOPDOWN_FE_BOUND.ITLB / SLOTS", + "MetricGroup": "TopdownL3;tma_frontend_latency_group", + "MetricName": "tma_itlb", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not delivered by the frontend due to BACLEARS, which occurs when the Branch Target Buffer (BTB) prediction or lack thereof, was corrected by a later branch predictor in the frontend", + "MetricExpr": "TOPDOWN_FE_BOUND.BRANCH_DETECT / SLOTS", + "MetricGroup": "TopdownL3;tma_frontend_latency_group", + "MetricName": "tma_branch_detect", + "PublicDescription": "Counts the number of issue slots that were not delivered by the frontend due to BACLEARS, which occurs when the Branch Target Buffer (BTB) prediction or lack thereof, was corrected by a later branch predictor in the frontend. Includes BACLEARS due to all branch types including conditional and unconditional jumps, returns, and indirect branches.", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not delivered by the frontend due to BTCLEARS, which occurs when the Branch Target Buffer (BTB) predicts a taken branch.", + "MetricExpr": "TOPDOWN_FE_BOUND.BRANCH_RESTEER / SLOTS", + "MetricGroup": "TopdownL3;tma_frontend_latency_group", + "MetricName": "tma_branch_resteer", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not delivered by the frontend due to frontend bandwidth restrictions due to decode, predecode, cisc, and other limitations.", + "MetricExpr": "TOPDOWN_FE_BOUND.FRONTEND_BANDWIDTH / SLOTS", + "MetricGroup": "TopdownL2;tma_frontend_bound_group", + "MetricName": "tma_frontend_bandwidth", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not delivered by the frontend due to the microcode sequencer (MS).", + "MetricExpr": "TOPDOWN_FE_BOUND.CISC / SLOTS", + "MetricGroup": "TopdownL3;tma_frontend_bandwidth_group", + "MetricName": "tma_cisc", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not delivered by the frontend due to decode stalls.", + "MetricExpr": "TOPDOWN_FE_BOUND.DECODE / SLOTS", + "MetricGroup": "TopdownL3;tma_frontend_bandwidth_group", + "MetricName": "tma_decode", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not delivered by the frontend due to wrong predecodes.", + "MetricExpr": "TOPDOWN_FE_BOUND.PREDECODE / SLOTS", + "MetricGroup": "TopdownL3;tma_frontend_bandwidth_group", + "MetricName": "tma_predecode", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not delivered by the frontend due to other common frontend stalls not categorized.", + "MetricExpr": "TOPDOWN_FE_BOUND.OTHER / SLOTS", + "MetricGroup": "TopdownL3;tma_frontend_bandwidth_group", + "MetricName": "tma_other_fb", + "ScaleUnit": "100%", "Unit": "cpu_atom" }, { "BriefDescription": "Counts the total number of issue slots that were not consumed by the backend because allocation is stalled due to a mispredicted jump or a machine clear", - "MetricExpr": "TOPDOWN_BAD_SPECULATION.ALL / (5 * CPU_CLK_UNHALTED.CORE)", + "MetricExpr": "(SLOTS - (TOPDOWN_FE_BOUND.ALL + TOPDOWN_BE_BOUND.ALL + TOPDOWN_RETIRING.ALL)) / SLOTS", "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", + "MetricName": "tma_bad_speculation", "PublicDescription": "Counts the total number of issue slots that were not consumed by the backend because allocation is stalled due to a mispredicted jump or a machine clear. Only issue slots wasted due to fast nukes such as memory ordering nukes are counted. Other nukes are not accounted for. Counts all issue slots blocked during this recovery window including relevant microcode flows and while uops are not yet available in the instruction queue (IQ). Also includes the issue slots that were consumed by the backend but were thrown away because they were younger than the mispredict or machine clear.", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not consumed by the backend due to branch mispredicts.", + "MetricExpr": "TOPDOWN_BAD_SPECULATION.MISPREDICT / SLOTS", + "MetricGroup": "TopdownL2;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the total number of issue slots that were not consumed by the backend because allocation is stalled due to a machine clear (nuke) of any kind including memory ordering and memory disambiguation.", + "MetricExpr": "TOPDOWN_BAD_SPECULATION.MACHINE_CLEARS / SLOTS", + "MetricGroup": "TopdownL2;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not consumed by the backend due to a machine clear (slow nuke).", + "MetricExpr": "TOPDOWN_BAD_SPECULATION.NUKE / SLOTS", + "MetricGroup": "TopdownL3;tma_machine_clears_group", + "MetricName": "tma_nuke", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of machine clears relative to the number of nuke slots due to SMC. ", + "MetricExpr": "tma_nuke * (MACHINE_CLEARS.SMC / MACHINE_CLEARS.SLOW)", + "MetricGroup": "TopdownL4;tma_nuke_group", + "MetricName": "tma_smc", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of machine clears relative to the number of nuke slots due to memory ordering. ", + "MetricExpr": "tma_nuke * (MACHINE_CLEARS.MEMORY_ORDERING / MACHINE_CLEARS.SLOW)", + "MetricGroup": "TopdownL4;tma_nuke_group", + "MetricName": "tma_memory_ordering", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of machine clears relative to the number of nuke slots due to FP assists. ", + "MetricExpr": "tma_nuke * (MACHINE_CLEARS.FP_ASSIST / MACHINE_CLEARS.SLOW)", + "MetricGroup": "TopdownL4;tma_nuke_group", + "MetricName": "tma_fp_assist", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of machine clears relative to the number of nuke slots due to memory disambiguation. ", + "MetricExpr": "tma_nuke * (MACHINE_CLEARS.DISAMBIGUATION / MACHINE_CLEARS.SLOW)", + "MetricGroup": "TopdownL4;tma_nuke_group", + "MetricName": "tma_disambiguation", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of machine clears relative to the number of nuke slots due to page faults. ", + "MetricExpr": "tma_nuke * (MACHINE_CLEARS.PAGE_FAULT / MACHINE_CLEARS.SLOW)", + "MetricGroup": "TopdownL4;tma_nuke_group", + "MetricName": "tma_page_fault", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not consumed by the backend due to a machine clear classified as a fast nuke due to memory ordering, memory disambiguation and memory renaming.", + "MetricExpr": "TOPDOWN_BAD_SPECULATION.FASTNUKE / SLOTS", + "MetricGroup": "TopdownL3;tma_machine_clears_group", + "MetricName": "tma_fast_nuke", + "ScaleUnit": "100%", "Unit": "cpu_atom" }, { "BriefDescription": "Counts the total number of issue slots that were not consumed by the backend due to backend stalls", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "TOPDOWN_BE_BOUND.ALL / (5 * CPU_CLK_UNHALTED.CORE)", + "MetricExpr": "TOPDOWN_BE_BOUND.ALL / SLOTS", "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", + "MetricName": "tma_backend_bound", "PublicDescription": "Counts the total number of issue slots that were not consumed by the backend due to backend stalls. Note that uops must be available for consumption in order for this event to count. If a uop is not available (IQ is empty), this event will not count. The rest of these subevents count backend stalls, in cycles, due to an outstanding request which is memory bound vs core bound. The subevents are not slot based events and therefore can not be precisely added or subtracted from the Backend_Bound_Aux subevents which are slot based.", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles due to backend bound stalls that are core execution bound and not attributed to outstanding demand load or store stalls. ", + "MetricExpr": "max(0, tma_backend_bound - tma_load_store_bound)", + "MetricGroup": "TopdownL2;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles the core is stalled due to stores or loads. ", + "MetricExpr": "min((TOPDOWN_BE_BOUND.ALL / SLOTS), (LD_HEAD.ANY_AT_RET / CLKS) + tma_store_bound)", + "MetricGroup": "TopdownL2;tma_backend_bound_group", + "MetricName": "tma_load_store_bound", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles the core is stalled due to store buffer full.", + "MetricExpr": "tma_mem_scheduler * (MEM_SCHEDULER_BLOCK.ST_BUF / MEM_SCHEDULER_BLOCK.ALL)", + "MetricGroup": "TopdownL3;tma_load_store_bound_group", + "MetricName": "tma_store_bound", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles that the oldest load of the load buffer is stalled at retirement due to a load block.", + "MetricExpr": "LD_HEAD.L1_BOUND_AT_RET / CLKS", + "MetricGroup": "TopdownL3;tma_load_store_bound_group", + "MetricName": "tma_l1_bound", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles that the oldest load of the load buffer is stalled at retirement due to a store forward block.", + "MetricExpr": "LD_HEAD.ST_ADDR_AT_RET / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles that the oldest load of the load buffer is stalled at retirement due to a first level TLB miss.", + "MetricExpr": "LD_HEAD.DTLB_MISS_AT_RET / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_stlb_hit", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles that the oldest load of the load buffer is stalled at retirement due to a second level TLB miss requiring a page walk.", + "MetricExpr": "LD_HEAD.PGWALK_AT_RET / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_stlb_miss", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles that the oldest load of the load buffer is stalled at retirement due to a number of other load blocks.", + "MetricExpr": "LD_HEAD.OTHER_AT_RET / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_other_l1", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles a core is stalled due to a demand load which hit in the L2 Cache.", + "MetricExpr": "(MEM_BOUND_STALLS.LOAD_L2_HIT / CLKS) - (MEM_BOUND_STALLS_AT_RET_CORRECTION * MEM_BOUND_STALLS.LOAD_L2_HIT / MEM_BOUND_STALLS.LOAD)", + "MetricGroup": "TopdownL3;tma_load_store_bound_group", + "MetricName": "tma_l2_bound", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles a core is stalled due to a demand load which hit in the Last Level Cache (LLC) or other core with HITE/F/M.", + "MetricExpr": "(MEM_BOUND_STALLS.LOAD_LLC_HIT / CLKS) - (MEM_BOUND_STALLS_AT_RET_CORRECTION * MEM_BOUND_STALLS.LOAD_LLC_HIT / MEM_BOUND_STALLS.LOAD)", + "MetricGroup": "TopdownL3;tma_load_store_bound_group", + "MetricName": "tma_l3_bound", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles the core is stalled due to a demand load miss which hit in DRAM or MMIO (Non-DRAM).", + "MetricExpr": "(MEM_BOUND_STALLS.LOAD_DRAM_HIT / CLKS) - (MEM_BOUND_STALLS_AT_RET_CORRECTION * MEM_BOUND_STALLS.LOAD_DRAM_HIT / MEM_BOUND_STALLS.LOAD)", + "MetricGroup": "TopdownL3;tma_load_store_bound_group", + "MetricName": "tma_dram_bound", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles the core is stalled due to a demand load miss which hits in the L2, LLC, DRAM or MMIO (Non-DRAM) but could not be correctly attributed or cycles in which the load miss is waiting on a request buffer.", + "MetricExpr": "max(0, tma_load_store_bound - (tma_store_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_dram_bound))", + "MetricGroup": "TopdownL3;tma_load_store_bound_group", + "MetricName": "tma_other_load_store", + "ScaleUnit": "100%", "Unit": "cpu_atom" }, { "BriefDescription": "Counts the total number of issue slots that were not consumed by the backend due to backend stalls", - "MetricExpr": "(TOPDOWN_BE_BOUND.ALL / (5 * CPU_CLK_UNHALTED.CORE))", + "MetricExpr": "tma_backend_bound", "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound_Aux", + "MetricName": "tma_backend_bound_aux", "PublicDescription": "Counts the total number of issue slots that were not consumed by the backend due to backend stalls. Note that UOPS must be available for consumption in order for this event to count. If a uop is not available (IQ is empty), this event will not count. All of these subevents count backend stalls, in slots, due to a resource limitation. These are not cycle based events and therefore can not be precisely added or subtracted from the Backend_Bound subevents which are cycle based. These subevents are supplementary to Backend_Bound and can be used to analyze results from a resource perspective at allocation. ", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the total number of issue slots that were not consumed by the backend due to backend stalls", + "MetricExpr": "tma_backend_bound", + "MetricGroup": "TopdownL2;tma_backend_bound_aux_group", + "MetricName": "tma_resource_bound", + "PublicDescription": "Counts the total number of issue slots that were not consumed by the backend due to backend stalls. Note that uops must be available for consumption in order for this event to count. If a uop is not available (IQ is empty), this event will not count. ", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not consumed by the backend due to memory reservation stalls in which a scheduler is not able to accept uops.", + "MetricExpr": "TOPDOWN_BE_BOUND.MEM_SCHEDULER / SLOTS", + "MetricGroup": "TopdownL3;tma_resource_bound_group", + "MetricName": "tma_mem_scheduler", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles, relative to the number of mem_scheduler slots, in which uops are blocked due to store buffer full", + "MetricExpr": "tma_mem_scheduler * (MEM_SCHEDULER_BLOCK.ST_BUF / MEM_SCHEDULER_BLOCK.ALL)", + "MetricGroup": "TopdownL4;tma_mem_scheduler_group", + "MetricName": "tma_st_buffer", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles, relative to the number of mem_scheduler slots, in which uops are blocked due to load buffer full", + "MetricExpr": "tma_mem_scheduler * MEM_SCHEDULER_BLOCK.LD_BUF / MEM_SCHEDULER_BLOCK.ALL", + "MetricGroup": "TopdownL4;tma_mem_scheduler_group", + "MetricName": "tma_ld_buffer", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cycles, relative to the number of mem_scheduler slots, in which uops are blocked due to RSV full relative ", + "MetricExpr": "tma_mem_scheduler * MEM_SCHEDULER_BLOCK.RSV / MEM_SCHEDULER_BLOCK.ALL", + "MetricGroup": "TopdownL4;tma_mem_scheduler_group", + "MetricName": "tma_rsv", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not consumed by the backend due to IEC or FPC RAT stalls, which can be due to FIQ or IEC reservation stalls in which the integer, floating point or SIMD scheduler is not able to accept uops.", + "MetricExpr": "TOPDOWN_BE_BOUND.NON_MEM_SCHEDULER / SLOTS", + "MetricGroup": "TopdownL3;tma_resource_bound_group", + "MetricName": "tma_non_mem_scheduler", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not consumed by the backend due to the physical register file unable to accept an entry (marble stalls).", + "MetricExpr": "TOPDOWN_BE_BOUND.REGISTER / SLOTS", + "MetricGroup": "TopdownL3;tma_resource_bound_group", + "MetricName": "tma_register", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not consumed by the backend due to the reorder buffer being full (ROB stalls).", + "MetricExpr": "TOPDOWN_BE_BOUND.REORDER_BUFFER / SLOTS", + "MetricGroup": "TopdownL3;tma_resource_bound_group", + "MetricName": "tma_reorder_buffer", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not consumed by the backend due to certain allocation restrictions.", + "MetricExpr": "TOPDOWN_BE_BOUND.ALLOC_RESTRICTIONS / SLOTS", + "MetricGroup": "TopdownL3;tma_resource_bound_group", + "MetricName": "tma_alloc_restriction", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of issue slots that were not consumed by the backend due to scoreboards from the instruction queue (IQ), jump execution unit (JEU), or microcode sequencer (MS).", + "MetricExpr": "TOPDOWN_BE_BOUND.SERIALIZATION / SLOTS", + "MetricGroup": "TopdownL3;tma_resource_bound_group", + "MetricName": "tma_serialization", + "ScaleUnit": "100%", "Unit": "cpu_atom" }, { "BriefDescription": "Counts the numer of issue slots that result in retirement slots. ", - "MetricExpr": "TOPDOWN_RETIRING.ALL / (5 * CPU_CLK_UNHALTED.CORE)", + "MetricExpr": "TOPDOWN_RETIRING.ALL / SLOTS", "MetricGroup": "TopdownL1", - "MetricName": "Retiring", + "MetricName": "tma_retiring", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of uops that are not from the microsequencer. ", + "MetricExpr": "(TOPDOWN_RETIRING.ALL - UOPS_RETIRED.MS) / SLOTS", + "MetricGroup": "TopdownL2;tma_retiring_group", + "MetricName": "tma_base", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of floating point operations per uop with all default weighting.", + "MetricExpr": "UOPS_RETIRED.FPDIV / SLOTS", + "MetricGroup": "TopdownL3;tma_base_group", + "MetricName": "tma_fp_uops", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of uops retired excluding ms and fp div uops.", + "MetricExpr": "(TOPDOWN_RETIRING.ALL - UOPS_RETIRED.MS - UOPS_RETIRED.FPDIV) / SLOTS", + "MetricGroup": "TopdownL3;tma_base_group", + "MetricName": "tma_other_ret", + "ScaleUnit": "100%", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of uops that are from the complex flows issued by the micro-sequencer (MS)", + "MetricExpr": "UOPS_RETIRED.MS / SLOTS", + "MetricGroup": "TopdownL2;tma_retiring_group", + "MetricName": "tma_ms_uops", + "PublicDescription": "Counts the number of uops that are from the complex flows issued by the micro-sequencer (MS). This includes uops from flows due to complex instructions, faults, assists, and inserted flows.", + "ScaleUnit": "100%", "Unit": "cpu_atom" }, { @@ -551,19 +1776,19 @@ }, { "BriefDescription": "", - "MetricExpr": "5 * CPU_CLK_UNHALTED.CORE", + "MetricExpr": "5 * CLKS", "MetricName": "SLOTS", "Unit": "cpu_atom" }, { "BriefDescription": "Instructions Per Cycle", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.CORE", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricName": "IPC", "Unit": "cpu_atom" }, { "BriefDescription": "Cycles Per Instruction", - "MetricExpr": "CPU_CLK_UNHALTED.CORE / INST_RETIRED.ANY", + "MetricExpr": "CLKS / INST_RETIRED.ANY", "MetricName": "CPI", "Unit": "cpu_atom" }, @@ -623,7 +1848,7 @@ }, { "BriefDescription": "Instructions per Far Branch", - "MetricExpr": "INST_RETIRED.ANY / ( BR_INST_RETIRED.FAR_BRANCH / 2 )", + "MetricExpr": "INST_RETIRED.ANY / (BR_INST_RETIRED.FAR_BRANCH / 2)", "MetricName": "IpFarBranch", "Unit": "cpu_atom" }, @@ -665,7 +1890,7 @@ }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.CORE / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricName": "Turbo_Utilization", "Unit": "cpu_atom" }, @@ -682,12 +1907,6 @@ "Unit": "cpu_atom" }, { - "BriefDescription": "Estimated Pause cost. In percent", - "MetricExpr": "100 * SERIALIZATION.NON_C01_MS_SCB / (5 * CPU_CLK_UNHALTED.CORE)", - "MetricName": "Estimated_Pause_Cost", - "Unit": "cpu_atom" - }, - { "BriefDescription": "Cycle cost per L2 hit", "MetricExpr": "MEM_BOUND_STALLS.LOAD_L2_HIT / MEM_LOAD_UOPS_RETIRED.L2_HIT", "MetricName": "Cycles_per_Demand_Load_L2_Hit", @@ -707,19 +1926,19 @@ }, { "BriefDescription": "Percent of instruction miss cost that hit in the L2", - "MetricExpr": "100 * MEM_BOUND_STALLS.IFETCH_L2_HIT / ( MEM_BOUND_STALLS.IFETCH )", + "MetricExpr": "100 * MEM_BOUND_STALLS.IFETCH_L2_HIT / (MEM_BOUND_STALLS.IFETCH)", "MetricName": "Inst_Miss_Cost_L2Hit_Percent", "Unit": "cpu_atom" }, { "BriefDescription": "Percent of instruction miss cost that hit in the L3", - "MetricExpr": "100 * MEM_BOUND_STALLS.IFETCH_LLC_HIT / ( MEM_BOUND_STALLS.IFETCH )", + "MetricExpr": "100 * MEM_BOUND_STALLS.IFETCH_LLC_HIT / (MEM_BOUND_STALLS.IFETCH)", "MetricName": "Inst_Miss_Cost_L3Hit_Percent", "Unit": "cpu_atom" }, { "BriefDescription": "Percent of instruction miss cost that hit in DRAM", - "MetricExpr": "100 * MEM_BOUND_STALLS.IFETCH_DRAM_HIT / ( MEM_BOUND_STALLS.IFETCH )", + "MetricExpr": "100 * MEM_BOUND_STALLS.IFETCH_DRAM_HIT / (MEM_BOUND_STALLS.IFETCH)", "MetricName": "Inst_Miss_Cost_DRAMHit_Percent", "Unit": "cpu_atom" }, diff --git a/tools/perf/pmu-events/arch/x86/alderlake/cache.json b/tools/perf/pmu-events/arch/x86/alderlake/cache.json index 887dce4dfeba..2cc62d2779d2 100644 --- a/tools/perf/pmu-events/arch/x86/alderlake/cache.json +++ b/tools/perf/pmu-events/arch/x86/alderlake/cache.json @@ -1,5 +1,29 @@ [ { + "BriefDescription": "Counts the number of cacheable memory requests that miss in the LLC. Counts on a per core basis.", + "CollectPEBSRecord": "2", + "Counter": "0,1,2,3,4,5", + "EventCode": "0x2e", + "EventName": "LONGEST_LAT_CACHE.MISS", + "PEBScounters": "0,1,2,3,4,5", + "SampleAfterValue": "200003", + "Speculative": "1", + "UMask": "0x41", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts the number of cacheable memory requests that access the LLC. Counts on a per core basis.", + "CollectPEBSRecord": "2", + "Counter": "0,1,2,3,4,5", + "EventCode": "0x2e", + "EventName": "LONGEST_LAT_CACHE.REFERENCE", + "PEBScounters": "0,1,2,3,4,5", + "SampleAfterValue": "200003", + "Speculative": "1", + "UMask": "0x4f", + "Unit": "cpu_atom" + }, + { "BriefDescription": "Counts the number of cycles the core is stalled due to an instruction cache or TLB miss which hit in the L2, LLC, DRAM or MMIO (Non-DRAM).", "CollectPEBSRecord": "2", "Counter": "0,1,2,3,4,5", @@ -210,8 +234,8 @@ }, { "BriefDescription": "Counts the number of tagged loads with an instruction latency that exceeds or equals the threshold of 128 cycles as defined in MEC_CR_PEBS_LD_LAT_THRESHOLD (3F6H). Only counts with PEBS enabled.", - "CollectPEBSRecord": "3", - "Counter": "0,1,2,3,4,5", + "CollectPEBSRecord": "2", + "Counter": "0,1", "Data_LA": "1", "EventCode": "0xd0", "EventName": "MEM_UOPS_RETIRED.LOAD_LATENCY_GT_128", @@ -219,7 +243,7 @@ "MSRIndex": "0x3F6", "MSRValue": "0x80", "PEBS": "2", - "PEBScounters": "0,1,2,3,4,5", + "PEBScounters": "0,1", "SampleAfterValue": "1000003", "TakenAlone": "1", "UMask": "0x5", @@ -227,8 +251,8 @@ }, { "BriefDescription": "Counts the number of tagged loads with an instruction latency that exceeds or equals the threshold of 16 cycles as defined in MEC_CR_PEBS_LD_LAT_THRESHOLD (3F6H). Only counts with PEBS enabled.", - "CollectPEBSRecord": "3", - "Counter": "0,1,2,3,4,5", + "CollectPEBSRecord": "2", + "Counter": "0,1", "Data_LA": "1", "EventCode": "0xd0", "EventName": "MEM_UOPS_RETIRED.LOAD_LATENCY_GT_16", @@ -236,7 +260,7 @@ "MSRIndex": "0x3F6", "MSRValue": "0x10", "PEBS": "2", - "PEBScounters": "0,1,2,3,4,5", + "PEBScounters": "0,1", "SampleAfterValue": "1000003", "TakenAlone": "1", "UMask": "0x5", @@ -244,8 +268,8 @@ }, { "BriefDescription": "Counts the number of tagged loads with an instruction latency that exceeds or equals the threshold of 256 cycles as defined in MEC_CR_PEBS_LD_LAT_THRESHOLD (3F6H). Only counts with PEBS enabled.", - "CollectPEBSRecord": "3", - "Counter": "0,1,2,3,4,5", + "CollectPEBSRecord": "2", + "Counter": "0,1", "Data_LA": "1", "EventCode": "0xd0", "EventName": "MEM_UOPS_RETIRED.LOAD_LATENCY_GT_256", @@ -253,7 +277,7 @@ "MSRIndex": "0x3F6", "MSRValue": "0x100", "PEBS": "2", - "PEBScounters": "0,1,2,3,4,5", + "PEBScounters": "0,1", "SampleAfterValue": "1000003", "TakenAlone": "1", "UMask": "0x5", @@ -261,8 +285,8 @@ }, { "BriefDescription": "Counts the number of tagged loads with an instruction latency that exceeds or equals the threshold of 32 cycles as defined in MEC_CR_PEBS_LD_LAT_THRESHOLD (3F6H). Only counts with PEBS enabled.", - "CollectPEBSRecord": "3", - "Counter": "0,1,2,3,4,5", + "CollectPEBSRecord": "2", + "Counter": "0,1", "Data_LA": "1", "EventCode": "0xd0", "EventName": "MEM_UOPS_RETIRED.LOAD_LATENCY_GT_32", @@ -270,7 +294,7 @@ "MSRIndex": "0x3F6", "MSRValue": "0x20", "PEBS": "2", - "PEBScounters": "0,1,2,3,4,5", + "PEBScounters": "0,1", "SampleAfterValue": "1000003", "TakenAlone": "1", "UMask": "0x5", @@ -278,8 +302,8 @@ }, { "BriefDescription": "Counts the number of tagged loads with an instruction latency that exceeds or equals the threshold of 4 cycles as defined in MEC_CR_PEBS_LD_LAT_THRESHOLD (3F6H). Only counts with PEBS enabled.", - "CollectPEBSRecord": "3", - "Counter": "0,1,2,3,4,5", + "CollectPEBSRecord": "2", + "Counter": "0,1", "Data_LA": "1", "EventCode": "0xd0", "EventName": "MEM_UOPS_RETIRED.LOAD_LATENCY_GT_4", @@ -287,7 +311,7 @@ "MSRIndex": "0x3F6", "MSRValue": "0x4", "PEBS": "2", - "PEBScounters": "0,1,2,3,4,5", + "PEBScounters": "0,1", "SampleAfterValue": "1000003", "TakenAlone": "1", "UMask": "0x5", @@ -295,8 +319,8 @@ }, { "BriefDescription": "Counts the number of tagged loads with an instruction latency that exceeds or equals the threshold of 512 cycles as defined in MEC_CR_PEBS_LD_LAT_THRESHOLD (3F6H). Only counts with PEBS enabled.", - "CollectPEBSRecord": "3", - "Counter": "0,1,2,3,4,5", + "CollectPEBSRecord": "2", + "Counter": "0,1", "Data_LA": "1", "EventCode": "0xd0", "EventName": "MEM_UOPS_RETIRED.LOAD_LATENCY_GT_512", @@ -304,7 +328,7 @@ "MSRIndex": "0x3F6", "MSRValue": "0x200", "PEBS": "2", - "PEBScounters": "0,1,2,3,4,5", + "PEBScounters": "0,1", "SampleAfterValue": "1000003", "TakenAlone": "1", "UMask": "0x5", @@ -312,8 +336,8 @@ }, { "BriefDescription": "Counts the number of tagged loads with an instruction latency that exceeds or equals the threshold of 64 cycles as defined in MEC_CR_PEBS_LD_LAT_THRESHOLD (3F6H). Only counts with PEBS enabled.", - "CollectPEBSRecord": "3", - "Counter": "0,1,2,3,4,5", + "CollectPEBSRecord": "2", + "Counter": "0,1", "Data_LA": "1", "EventCode": "0xd0", "EventName": "MEM_UOPS_RETIRED.LOAD_LATENCY_GT_64", @@ -321,7 +345,7 @@ "MSRIndex": "0x3F6", "MSRValue": "0x40", "PEBS": "2", - "PEBScounters": "0,1,2,3,4,5", + "PEBScounters": "0,1", "SampleAfterValue": "1000003", "TakenAlone": "1", "UMask": "0x5", @@ -329,8 +353,8 @@ }, { "BriefDescription": "Counts the number of tagged loads with an instruction latency that exceeds or equals the threshold of 8 cycles as defined in MEC_CR_PEBS_LD_LAT_THRESHOLD (3F6H). Only counts with PEBS enabled.", - "CollectPEBSRecord": "3", - "Counter": "0,1,2,3,4,5", + "CollectPEBSRecord": "2", + "Counter": "0,1", "Data_LA": "1", "EventCode": "0xd0", "EventName": "MEM_UOPS_RETIRED.LOAD_LATENCY_GT_8", @@ -338,7 +362,7 @@ "MSRIndex": "0x3F6", "MSRValue": "0x8", "PEBS": "2", - "PEBScounters": "0,1,2,3,4,5", + "PEBScounters": "0,1", "SampleAfterValue": "1000003", "TakenAlone": "1", "UMask": "0x5", @@ -359,7 +383,7 @@ }, { "BriefDescription": "Counts the number of stores uops retired. Counts with or without PEBS enabled.", - "CollectPEBSRecord": "3", + "CollectPEBSRecord": "2", "Counter": "0,1,2,3,4,5", "Data_LA": "1", "EventCode": "0xd0", @@ -372,6 +396,61 @@ "Unit": "cpu_atom" }, { + "BriefDescription": "Counts demand data reads that were supplied by the L3 cache.", + "Counter": "0,1,2,3,4,5", + "EventCode": "0xB7", + "EventName": "OCR.DEMAND_DATA_RD.L3_HIT", + "MSRIndex": "0x1a6,0x1a7", + "MSRValue": "0x3F803C0001", + "SampleAfterValue": "100003", + "UMask": "0x1", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts demand data reads that were supplied by the L3 cache where a snoop was sent, the snoop hit, and modified data was forwarded.", + "Counter": "0,1,2,3,4,5", + "EventCode": "0xB7", + "EventName": "OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM", + "MSRIndex": "0x1a6,0x1a7", + "MSRValue": "0x10003C0001", + "SampleAfterValue": "100003", + "UMask": "0x1", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts demand data reads that were supplied by the L3 cache where a snoop was sent, the snoop hit, but no data was forwarded.", + "Counter": "0,1,2,3,4,5", + "EventCode": "0xB7", + "EventName": "OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_NO_FWD", + "MSRIndex": "0x1a6,0x1a7", + "MSRValue": "0x4003C0001", + "SampleAfterValue": "100003", + "UMask": "0x1", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts demand data reads that were supplied by the L3 cache where a snoop was sent, the snoop hit, and non-modified data was forwarded.", + "Counter": "0,1,2,3,4,5", + "EventCode": "0xB7", + "EventName": "OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD", + "MSRIndex": "0x1a6,0x1a7", + "MSRValue": "0x8003C0001", + "SampleAfterValue": "100003", + "UMask": "0x1", + "Unit": "cpu_atom" + }, + { + "BriefDescription": "Counts demand reads for ownership (RFO) and software prefetches for exclusive ownership (PREFETCHW) that were supplied by the L3 cache.", + "Counter": "0,1,2,3,4,5", + "EventCode": "0xB7", + "EventName": "OCR.DEMAND_RFO.L3_HIT", + "MSRIndex": "0x1a6,0x1a7", + "MSRValue": "0x3F803C0002", + "SampleAfterValue": "100003", + "UMask": "0x1", + "Unit": "cpu_atom" + }, + { "BriefDescription": "Counts demand reads for ownership (RFO) and software prefetches for exclusive ownership (PREFETCHW) that were supplied by the L3 cache where a snoop was sent, the snoop hit, and modified data was forwarded.", "Counter": "0,1,2,3,4,5", "EventCode": "0xB7", diff --git a/tools/perf/pmu-events/arch/x86/alderlake/frontend.json b/tools/perf/pmu-events/arch/x86/alderlake/frontend.json index 2cfa70b2d5e1..da1a7ba0e568 100644 --- a/tools/perf/pmu-events/arch/x86/alderlake/frontend.json +++ b/tools/perf/pmu-events/arch/x86/alderlake/frontend.json @@ -48,6 +48,18 @@ "Unit": "cpu_core" }, { + "BriefDescription": "Cycles the Microcode Sequencer is busy.", + "CollectPEBSRecord": "2", + "Counter": "0,1,2,3", + "EventCode": "0x87", + "EventName": "DECODE.MS_BUSY", + "PEBScounters": "0,1,2,3", + "SampleAfterValue": "500009", + "Speculative": "1", + "UMask": "0x2", + "Unit": "cpu_core" + }, + { "BriefDescription": "DSB-to-MITE switch true penalty cycles.", "CollectPEBSRecord": "2", "Counter": "0,1,2,3", diff --git a/tools/perf/pmu-events/arch/x86/alderlake/memory.json b/tools/perf/pmu-events/arch/x86/alderlake/memory.json index 586fb961e46d..f894e4a0212b 100644 --- a/tools/perf/pmu-events/arch/x86/alderlake/memory.json +++ b/tools/perf/pmu-events/arch/x86/alderlake/memory.json @@ -83,6 +83,17 @@ "Unit": "cpu_atom" }, { + "BriefDescription": "Counts demand data reads that were not supplied by the L3 cache.", + "Counter": "0,1,2,3,4,5", + "EventCode": "0xB7", + "EventName": "OCR.DEMAND_DATA_RD.L3_MISS_LOCAL", + "MSRIndex": "0x1a6,0x1a7", + "MSRValue": "0x3F84400001", + "SampleAfterValue": "100003", + "UMask": "0x1", + "Unit": "cpu_atom" + }, + { "BriefDescription": "Counts demand reads for ownership (RFO) and software prefetches for exclusive ownership (PREFETCHW) that were not supplied by the L3 cache.", "Counter": "0,1,2,3,4,5", "EventCode": "0xB7", @@ -94,6 +105,17 @@ "Unit": "cpu_atom" }, { + "BriefDescription": "Counts demand reads for ownership (RFO) and software prefetches for exclusive ownership (PREFETCHW) that were not supplied by the L3 cache.", + "Counter": "0,1,2,3,4,5", + "EventCode": "0xB7", + "EventName": "OCR.DEMAND_RFO.L3_MISS_LOCAL", + "MSRIndex": "0x1a6,0x1a7", + "MSRValue": "0x3F84400002", + "SampleAfterValue": "100003", + "UMask": "0x1", + "Unit": "cpu_atom" + }, + { "BriefDescription": "Execution stalls while L3 cache miss demand load is outstanding.", "CollectPEBSRecord": "2", "Counter": "0,1,2,3", diff --git a/tools/perf/pmu-events/arch/x86/alderlake/other.json b/tools/perf/pmu-events/arch/x86/alderlake/other.json index 67a9c13cc71d..c49d8ce27310 100644 --- a/tools/perf/pmu-events/arch/x86/alderlake/other.json +++ b/tools/perf/pmu-events/arch/x86/alderlake/other.json @@ -1,5 +1,16 @@ [ { + "BriefDescription": "Counts modified writebacks from L1 cache and L2 cache that have any type of response.", + "Counter": "0,1,2,3,4,5", + "EventCode": "0xB7", + "EventName": "OCR.COREWB_M.ANY_RESPONSE", + "MSRIndex": "0x1a6,0x1a7", + "MSRValue": "0x10008", + "SampleAfterValue": "100003", + "UMask": "0x1", + "Unit": "cpu_atom" + }, + { "BriefDescription": "Counts demand data reads that have any type of response.", "Counter": "0,1,2,3,4,5", "EventCode": "0xB7", @@ -104,6 +115,17 @@ "Unit": "cpu_core" }, { + "BriefDescription": "Counts demand data reads that were supplied by DRAM.", + "Counter": "0,1,2,3,4,5,6,7", + "EventCode": "0x2A,0x2B", + "EventName": "OCR.DEMAND_DATA_RD.DRAM", + "MSRIndex": "0x1a6,0x1a7", + "MSRValue": "0x184000001", + "SampleAfterValue": "100003", + "UMask": "0x1", + "Unit": "cpu_core" + }, + { "BriefDescription": "Counts demand read for ownership (RFO) requests and software prefetches for exclusive ownership (PREFETCHW) that have any type of response.", "Counter": "0,1,2,3,4,5,6,7", "EventCode": "0x2A,0x2B", diff --git a/tools/perf/pmu-events/arch/x86/alderlake/pipeline.json b/tools/perf/pmu-events/arch/x86/alderlake/pipeline.json index d02e078a90c9..1a137f7f8b7e 100644 --- a/tools/perf/pmu-events/arch/x86/alderlake/pipeline.json +++ b/tools/perf/pmu-events/arch/x86/alderlake/pipeline.json @@ -331,6 +331,18 @@ "Unit": "cpu_atom" }, { + "BriefDescription": "Counts the number of unhalted reference clock cycles at TSC frequency.", + "CollectPEBSRecord": "2", + "Counter": "0,1,2,3,4,5", + "EventCode": "0x3c", + "EventName": "CPU_CLK_UNHALTED.REF_TSC_P", + "PEBScounters": "0,1,2,3,4,5", + "SampleAfterValue": "2000003", + "Speculative": "1", + "UMask": "0x1", + "Unit": "cpu_atom" + }, + { "BriefDescription": "Counts the number of unhalted core clock cycles. (Fixed event)", "CollectPEBSRecord": "2", "Counter": "Fixed counter 1", @@ -874,7 +886,7 @@ "PEBScounters": "0,1,2,3,4,5,6,7", "SampleAfterValue": "100003", "Speculative": "1", - "UMask": "0x1f", + "UMask": "0x1b", "Unit": "cpu_core" }, { diff --git a/tools/perf/pmu-events/arch/x86/broadwell/bdw-metrics.json b/tools/perf/pmu-events/arch/x86/broadwell/bdw-metrics.json index d65afe3d0b06..c220b1cf1740 100644 --- a/tools/perf/pmu-events/arch/x86/broadwell/bdw-metrics.json +++ b/tools/perf/pmu-events/arch/x86/broadwell/bdw-metrics.json @@ -1,64 +1,552 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: RS_EVENTS.EMPTY_END", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", + "MetricExpr": "ICACHE.IFDATA_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "(14 * ITLB_MISSES.STLB_HIT + cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * ITLB_MISSES.WALK_COMPLETED) / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: ITLB_MISSES.WALK_COMPLETED", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "12 * (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY) / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. ", + "MetricExpr": "BR_MISP_RETIRED.ALL_BRANCHES * tma_branch_resteers / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY)", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. ", + "MetricExpr": "MACHINE_CLEARS.COUNT * tma_branch_resteers / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY)", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "tma_branch_resteers - tma_mispredicts_resteers - tma_clears_resteers", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: BACLEARS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "2 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - (tma_frontend_bound + tma_bad_speculation + tma_retiring)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB) / (CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if (IPC > 1.8) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_UOPS_RETIRED.L1_HIT_PS;MEM_LOAD_UOPS_RETIRED.HIT_LFB_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "(8 * DTLB_LOAD_MISSES.STLB_HIT + cpu@DTLB_LOAD_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * DTLB_LOAD_MISSES.WALK_COMPLETED) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_UOPS_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_UOPS_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_UOPS_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "Load_Miss_Real_Latency * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEM_LOAD_UOPS_RETIRED.L3_HIT / (MEM_LOAD_UOPS_RETIRED.L3_HIT + 7 * MEM_LOAD_UOPS_RETIRED.L3_MISS)) * CYCLE_ACTIVITY.STALLS_L2_MISS / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "(60 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS))) + 43 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS)))) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "43 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS))) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "29 * (MEM_LOAD_UOPS_RETIRED.L3_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS))) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "((OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2) if #SMT_on else OFFCORE_REQUESTS_BUFFER.SQ_FULL) / CORE_CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(1 - (MEM_LOAD_UOPS_RETIRED.L3_HIT / (MEM_LOAD_UOPS_RETIRED.L3_HIT + 7 * MEM_LOAD_UOPS_RETIRED.L3_MISS))) * CYCLE_ACTIVITY.STALLS_L2_MISS / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "RESOURCE_STALLS.SB / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_UOPS_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 9 * (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES))) + (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "60 * OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "2 * MEM_UOPS_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_UOPS_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(8 * DTLB_STORE_MISSES.STLB_HIT + cpu@DTLB_STORE_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * DTLB_STORE_MISSES.WALK_COMPLETED) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_UOPS_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.FPU_DIV_ACTIVE / CORE_CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if (IPC > 1.8) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) - RESOURCE_STALLS.SB - CYCLE_ACTIVITY.STALLS_MEM_ANY) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,inv\\,cmask\\=1@) / 2 if #SMT_on else (CYCLE_ACTIVITY.STALLS_TOTAL - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else 0) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 if #SMT_on else (UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / 2 if #SMT_on else (UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).", + "MetricExpr": "((cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ / 2) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_0 + UOPS_DISPATCHED_PORT.PORT_1 + UOPS_DISPATCHED_PORT.PORT_5 + UOPS_DISPATCHED_PORT.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED_PORT.PORT_0", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_1", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_6", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_2 + UOPS_DISPATCHED_PORT.PORT_3 + UOPS_DISPATCHED_PORT.PORT_7 - UOPS_DISPATCHED_PORT.PORT_4) / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_2", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_2 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_2", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_3", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_3 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data) Sample with: UOPS_DISPATCHED_PORT.PORT_4", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_4", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 7 ([HSW+]simple Store-address) Sample with: UOPS_DISPATCHED_PORT.PORT_7", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_7 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_7", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "INST_RETIRED.X87 * UPI / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * OTHER_ASSISTS.ANY_WB_ASSIST / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: OTHER_ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" }, { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -76,8 +564,8 @@ }, { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI" }, { @@ -88,17 +576,11 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "The ratio of Executed- by Issued-Uops", "MetricExpr": "UOPS_EXECUTED.THREAD / UOPS_ISSUED.ANY", "MetricGroup": "Cor;Pipeline", @@ -107,51 +589,32 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { - "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;Flops_SMT", - "MetricName": "FLOPc_SMT" - }, - { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.THREAD )", + "MetricExpr": "((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)." }, { - "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )", - "MetricGroup": "Cor;Flops;HPC_SMT", - "MetricName": "FP_Arith_Utilization_SMT", - "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common). SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, @@ -193,13 +656,13 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." @@ -220,22 +683,22 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -252,7 +715,7 @@ }, { "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)", - "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )", + "MetricExpr": "IDQ.DSB_UOPS / ((IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS))", "MetricGroup": "DSB;Fed;FetchBW", "MetricName": "DSB_Coverage" }, @@ -264,84 +727,72 @@ }, { "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;BrMispredicts", "MetricName": "Branch_Misprediction_Cost" }, { - "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES", - "MetricGroup": "Bad;BrMispredicts_SMT", - "MetricName": "Branch_Misprediction_Cost_SMT" - }, - { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load" }, { "BriefDescription": "L2 cache hits per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( L2_RQSTS.REFERENCES - L2_RQSTS.MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricExpr": "1000 * (L2_RQSTS.REFERENCES - L2_RQSTS.MISS) / INST_RETIRED.ANY", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_All" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=1@ + cpu@DTLB_LOAD_MISSES.WALK_DURATION\\,cmask\\=1@ + cpu@DTLB_STORE_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * ( DTLB_STORE_MISSES.WALK_COMPLETED + DTLB_LOAD_MISSES.WALK_COMPLETED + ITLB_MISSES.WALK_COMPLETED ) ) / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "(cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=1@ + cpu@DTLB_LOAD_MISSES.WALK_DURATION\\,cmask\\=1@ + cpu@DTLB_STORE_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * (DTLB_STORE_MISSES.WALK_COMPLETED + DTLB_LOAD_MISSES.WALK_COMPLETED + ITLB_MISSES.WALK_COMPLETED)) / CORE_CLKS", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, { - "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", - "MetricExpr": "( cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=1@ + cpu@DTLB_LOAD_MISSES.WALK_DURATION\\,cmask\\=1@ + cpu@DTLB_STORE_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * ( DTLB_STORE_MISSES.WALK_COMPLETED + DTLB_LOAD_MISSES.WALK_COMPLETED + ITLB_MISSES.WALK_COMPLETED ) ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Mem;MemoryTLB_SMT", - "MetricName": "Page_Walks_Utilization_SMT" - }, - { "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]", "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time", "MetricGroup": "Mem;MemoryBW", @@ -361,19 +812,19 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, @@ -391,26 +842,26 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -428,7 +879,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * ( arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@ ) / 1000000 / duration_time / 1000", + "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1000000 / duration_time / 1000", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, diff --git a/tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json b/tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json index b6fdf5ba2c9a..5a074cf7c77d 100644 --- a/tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json +++ b/tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json @@ -1,64 +1,556 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. Sample with: FRONTEND_RETIRED.LATENCY_GE_4_PS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: FRONTEND_RETIRED.LATENCY_GE_16_PS;FRONTEND_RETIRED.LATENCY_GE_8_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses", + "MetricExpr": "ICACHE.IFDATA_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses. Sample with: FRONTEND_RETIRED.L2_MISS_PS;FRONTEND_RETIRED.L1I_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "(14 * ITLB_MISSES.STLB_HIT + cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * ITLB_MISSES.WALK_COMPLETED) / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: FRONTEND_RETIRED.STLB_MISS_PS;FRONTEND_RETIRED.ITLB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "12 * (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY) / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage", + "MetricExpr": "BR_MISP_RETIRED.ALL_BRANCHES * tma_branch_resteers / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY)", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears", + "MetricExpr": "MACHINE_CLEARS.COUNT * tma_branch_resteers / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY)", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "tma_branch_resteers - tma_mispredicts_resteers - tma_clears_resteers", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: FRONTEND_RETIRED.UNKNOWN_BRANCH", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty. Sample with: FRONTEND_RETIRED.DSB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "2 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend. Sample with: FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_2_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck. Sample with: FRONTEND_RETIRED.ANY_DSB_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: TOPDOWN.BR_MISPREDICT_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - (tma_frontend_bound + tma_bad_speculation + tma_retiring)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. Sample with: TOPDOWN.BACKEND_BOUND_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB) / (CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if (IPC > 1.8) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_RETIRED.L1_HIT_PS;MEM_LOAD_RETIRED.FB_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "(8 * DTLB_LOAD_MISSES.STLB_HIT + cpu@DTLB_LOAD_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * DTLB_LOAD_MISSES.WALK_COMPLETED) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_INST_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_INST_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_INST_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "Load_Miss_Real_Latency * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEM_LOAD_UOPS_RETIRED.L3_HIT / (MEM_LOAD_UOPS_RETIRED.L3_HIT + 7 * MEM_LOAD_UOPS_RETIRED.L3_MISS)) * CYCLE_ACTIVITY.STALLS_L2_MISS / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "(60 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS))) + 43 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS)))) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "43 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS))) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "29 * (MEM_LOAD_UOPS_RETIRED.L3_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS))) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "((OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2) if #SMT_on else OFFCORE_REQUESTS_BUFFER.SQ_FULL) / CORE_CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(1 - (MEM_LOAD_UOPS_RETIRED.L3_HIT / (MEM_LOAD_UOPS_RETIRED.L3_HIT + 7 * MEM_LOAD_UOPS_RETIRED.L3_MISS))) * CYCLE_ACTIVITY.STALLS_L2_MISS / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "RESOURCE_STALLS.SB / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_INST_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 9 * (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES))) + (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "60 * OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "2 * MEM_UOPS_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_INST_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(8 * DTLB_STORE_MISSES.STLB_HIT + cpu@DTLB_STORE_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * DTLB_STORE_MISSES.WALK_COMPLETED) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_INST_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.FPU_DIV_ACTIVE / CORE_CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_ACTIVE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if (IPC > 1.8) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) - RESOURCE_STALLS.SB - CYCLE_ACTIVITY.STALLS_MEM_ANY) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,inv\\,cmask\\=1@) / 2 if #SMT_on else (CYCLE_ACTIVITY.STALLS_TOTAL - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else 0) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 if #SMT_on else (UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful. Sample with: EXE_ACTIVITY.1_PORTS_UTIL", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / 2 if #SMT_on else (UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop. Sample with: EXE_ACTIVITY.2_PORTS_UTIL", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "((cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ / 2) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Sample with: UOPS_EXECUTED.CYCLES_GE_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_0 + UOPS_DISPATCHED_PORT.PORT_1 + UOPS_DISPATCHED_PORT.PORT_5 + UOPS_DISPATCHED_PORT.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED.PORT_0", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED.PORT_1", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU)", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED.PORT_6", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3_10", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_2 + UOPS_DISPATCHED_PORT.PORT_3 + UOPS_DISPATCHED_PORT.PORT_7 - UOPS_DISPATCHED_PORT.PORT_4) / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 ([SNB+]Loads and Store-address; [ICL+] Loads)", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_2 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_2", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 ([SNB+]Loads and Store-address; [ICL+] Loads)", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_3 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations Sample with: UOPS_DISPATCHED.PORT_7_8", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data)", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_4", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 7 ([HSW+]simple Store-address)", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_7 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_7", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.SLOTS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "INST_RETIRED.X87 * UPI / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: UOPS_RETIRED.MS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * OTHER_ASSISTS.ANY_WB_ASSIST / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" }, { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -76,8 +568,8 @@ }, { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI" }, { @@ -88,17 +580,11 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "The ratio of Executed- by Issued-Uops", "MetricExpr": "UOPS_EXECUTED.THREAD / UOPS_ISSUED.ANY", "MetricGroup": "Cor;Pipeline", @@ -107,51 +593,32 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { - "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;Flops_SMT", - "MetricName": "FLOPc_SMT" - }, - { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.THREAD )", + "MetricExpr": "((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)." }, { - "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )", - "MetricGroup": "Cor;Flops;HPC_SMT", - "MetricName": "FP_Arith_Utilization_SMT", - "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common). SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, @@ -193,13 +660,13 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." @@ -220,22 +687,22 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -252,7 +719,7 @@ }, { "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)", - "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )", + "MetricExpr": "IDQ.DSB_UOPS / ((IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS))", "MetricGroup": "DSB;Fed;FetchBW", "MetricName": "DSB_Coverage" }, @@ -264,84 +731,72 @@ }, { "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;BrMispredicts", "MetricName": "Branch_Misprediction_Cost" }, { - "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES", - "MetricGroup": "Bad;BrMispredicts_SMT", - "MetricName": "Branch_Misprediction_Cost_SMT" - }, - { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load" }, { "BriefDescription": "L2 cache hits per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( L2_RQSTS.REFERENCES - L2_RQSTS.MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricExpr": "1000 * (L2_RQSTS.REFERENCES - L2_RQSTS.MISS) / INST_RETIRED.ANY", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_All" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=1@ + cpu@DTLB_LOAD_MISSES.WALK_DURATION\\,cmask\\=1@ + cpu@DTLB_STORE_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * ( DTLB_STORE_MISSES.WALK_COMPLETED + DTLB_LOAD_MISSES.WALK_COMPLETED + ITLB_MISSES.WALK_COMPLETED ) ) / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION + 7 * ( DTLB_STORE_MISSES.WALK_COMPLETED + DTLB_LOAD_MISSES.WALK_COMPLETED + ITLB_MISSES.WALK_COMPLETED ) ) / ( 2 * (( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) if #core_wide < 1 else ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else CPU_CLK_UNHALTED.THREAD) )", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, { - "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", - "MetricExpr": "( cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=1@ + cpu@DTLB_LOAD_MISSES.WALK_DURATION\\,cmask\\=1@ + cpu@DTLB_STORE_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * ( DTLB_STORE_MISSES.WALK_COMPLETED + DTLB_LOAD_MISSES.WALK_COMPLETED + ITLB_MISSES.WALK_COMPLETED ) ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Mem;MemoryTLB_SMT", - "MetricName": "Page_Walks_Utilization_SMT" - }, - { "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]", "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time", "MetricGroup": "Mem;MemoryBW", @@ -361,19 +816,19 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, @@ -391,26 +846,26 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -428,33 +883,21 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "( 64 * ( uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@ ) / 1000000000 ) / duration_time", + "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1000000 / duration_time / 1000", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, { - "BriefDescription": "Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches", - "MetricExpr": "1000000000 * ( cbox@event\\=0x36\\,umask\\=0x3\\,filter_opc\\=0x182@ / cbox@event\\=0x35\\,umask\\=0x3\\,filter_opc\\=0x182@ ) / ( cbox_0@event\\=0x0@ / duration_time )", - "MetricGroup": "Mem;MemoryLat;SoC", - "MetricName": "MEM_Read_Latency" - }, - { - "BriefDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches", - "MetricExpr": "cbox@event\\=0x36\\,umask\\=0x3\\,filter_opc\\=0x182@ / cbox@event\\=0x36\\,umask\\=0x3\\,filter_opc\\=0x182\\,thresh\\=1@", - "MetricGroup": "Mem;MemoryBW;SoC", - "MetricName": "MEM_Parallel_Reads" - }, - { - "BriefDescription": "Socket actual clocks when any core is active on that socket", - "MetricExpr": "cbox_0@event\\=0x0@", - "MetricGroup": "SoC", - "MetricName": "Socket_CLKS" + "BriefDescription": "Average latency of all requests to external memory (in Uncore cycles)", + "MetricExpr": "UNC_ARB_TRK_OCCUPANCY.ALL / arb@event\\=0x81\\,umask\\=0x1@", + "MetricGroup": "Mem;SoC", + "MetricName": "MEM_Request_Latency" }, { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "cbox_0@event\\=0x0@ / #num_dies / duration_time / 1000000000", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" + "BriefDescription": "Average number of parallel requests to external memory. Accounts for all requests", + "MetricExpr": "UNC_ARB_TRK_OCCUPANCY.ALL / arb@event\\=0x81\\,umask\\=0x1@", + "MetricGroup": "Mem;SoC", + "MetricName": "MEM_Parallel_Requests" }, { "BriefDescription": "Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]", diff --git a/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json b/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json index a3a15ee52841..e89fa536ca03 100644 --- a/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json +++ b/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json @@ -1,64 +1,576 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: RS_EVENTS.EMPTY_END", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", + "MetricExpr": "ICACHE.IFDATA_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "(14 * ITLB_MISSES.STLB_HIT + cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * ITLB_MISSES.WALK_COMPLETED) / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: ITLB_MISSES.WALK_COMPLETED", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "12 * (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY) / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. ", + "MetricExpr": "BR_MISP_RETIRED.ALL_BRANCHES * tma_branch_resteers / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY)", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. ", + "MetricExpr": "MACHINE_CLEARS.COUNT * tma_branch_resteers / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY)", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "tma_branch_resteers - tma_mispredicts_resteers - tma_clears_resteers", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: BACLEARS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "2 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - (tma_frontend_bound + tma_bad_speculation + tma_retiring)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB) / (CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if (IPC > 1.8) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_UOPS_RETIRED.L1_HIT_PS;MEM_LOAD_UOPS_RETIRED.HIT_LFB_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "(8 * DTLB_LOAD_MISSES.STLB_HIT + cpu@DTLB_LOAD_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * DTLB_LOAD_MISSES.WALK_COMPLETED) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_UOPS_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_UOPS_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_UOPS_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "Load_Miss_Real_Latency * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEM_LOAD_UOPS_RETIRED.L3_HIT / (MEM_LOAD_UOPS_RETIRED.L3_HIT + 7 * MEM_LOAD_UOPS_RETIRED.L3_MISS)) * CYCLE_ACTIVITY.STALLS_L2_MISS / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "(60 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) + 43 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD)))) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "43 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "41 * (MEM_LOAD_UOPS_RETIRED.L3_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "((OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2) if #SMT_on else OFFCORE_REQUESTS_BUFFER.SQ_FULL) / CORE_CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(1 - (MEM_LOAD_UOPS_RETIRED.L3_HIT / (MEM_LOAD_UOPS_RETIRED.L3_HIT + 7 * MEM_LOAD_UOPS_RETIRED.L3_MISS))) * CYCLE_ACTIVITY.STALLS_L2_MISS / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory", + "MetricExpr": "200 * (MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "Server;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_local_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory. Caching will improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM_PS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory", + "MetricExpr": "310 * (MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues", + "MetricExpr": "(200 * (MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) + 180 * (MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD)))) / CLKS", + "MetricGroup": "Offcore;Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_cache", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM_PS;MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "RESOURCE_STALLS.SB / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_UOPS_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 9 * (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES))) + (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "(200 * OFFCORE_RESPONSE.DEMAND_RFO.LLC_MISS.REMOTE_HITM + 60 * OFFCORE_RESPONSE.DEMAND_RFO.LLC_HIT.HITM_OTHER_CORE) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "2 * MEM_UOPS_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_UOPS_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(8 * DTLB_STORE_MISSES.STLB_HIT + cpu@DTLB_STORE_MISSES.WALK_DURATION\\,cmask\\=1@ + 7 * DTLB_STORE_MISSES.WALK_COMPLETED) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_UOPS_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.FPU_DIV_ACTIVE / CORE_CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if (IPC > 1.8) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) - RESOURCE_STALLS.SB - CYCLE_ACTIVITY.STALLS_MEM_ANY) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,inv\\,cmask\\=1@) / 2 if #SMT_on else (CYCLE_ACTIVITY.STALLS_TOTAL - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else 0) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 if #SMT_on else (UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / 2 if #SMT_on else (UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).", + "MetricExpr": "((cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ / 2) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_0 + UOPS_DISPATCHED_PORT.PORT_1 + UOPS_DISPATCHED_PORT.PORT_5 + UOPS_DISPATCHED_PORT.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED_PORT.PORT_0", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_1", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_6", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_2 + UOPS_DISPATCHED_PORT.PORT_3 + UOPS_DISPATCHED_PORT.PORT_7 - UOPS_DISPATCHED_PORT.PORT_4) / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_2", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_2 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_2", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_3", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_3 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data) Sample with: UOPS_DISPATCHED_PORT.PORT_4", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_4", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 7 ([HSW+]simple Store-address) Sample with: UOPS_DISPATCHED_PORT.PORT_7", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_7 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_7", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "INST_RETIRED.X87 * UPI / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * OTHER_ASSISTS.ANY_WB_ASSIST / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: OTHER_ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" }, { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -75,6 +587,12 @@ "MetricName": "UpTB" }, { + "BriefDescription": "Cycles Per Instruction (per Logical Processor)", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", + "MetricName": "CPI" + }, + { "BriefDescription": "Per-Logical Processor actual clocks when the Logical Processor is active.", "MetricExpr": "CPU_CLK_UNHALTED.THREAD", "MetricGroup": "Pipeline", @@ -82,17 +600,11 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "The ratio of Executed- by Issued-Uops", "MetricExpr": "UOPS_EXECUTED.THREAD / UOPS_ISSUED.ANY", "MetricGroup": "Cor;Pipeline", @@ -101,51 +613,32 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { - "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;Flops_SMT", - "MetricName": "FLOPc_SMT" - }, - { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.THREAD )", + "MetricExpr": "((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)." }, { - "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )", - "MetricGroup": "Cor;Flops;HPC_SMT", - "MetricName": "FP_Arith_Utilization_SMT", - "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common). SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, @@ -187,13 +680,13 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." @@ -214,22 +707,22 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -246,7 +739,7 @@ }, { "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)", - "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )", + "MetricExpr": "IDQ.DSB_UOPS / ((IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS))", "MetricGroup": "DSB;Fed;FetchBW", "MetricName": "DSB_Coverage" }, @@ -258,84 +751,72 @@ }, { "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;BrMispredicts", "MetricName": "Branch_Misprediction_Cost" }, { - "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES", - "MetricGroup": "Bad;BrMispredicts_SMT", - "MetricName": "Branch_Misprediction_Cost_SMT" - }, - { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load" }, { "BriefDescription": "L2 cache hits per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( L2_RQSTS.REFERENCES - L2_RQSTS.MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricExpr": "1000 * (L2_RQSTS.REFERENCES - L2_RQSTS.MISS) / INST_RETIRED.ANY", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_All" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION + 7 * ( DTLB_STORE_MISSES.WALK_COMPLETED + DTLB_LOAD_MISSES.WALK_COMPLETED + ITLB_MISSES.WALK_COMPLETED ) ) / ( 2 * CPU_CLK_UNHALTED.THREAD )", + "MetricExpr": "(ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION + 7 * (DTLB_STORE_MISSES.WALK_COMPLETED + DTLB_LOAD_MISSES.WALK_COMPLETED + ITLB_MISSES.WALK_COMPLETED)) / (2 * CORE_CLKS)", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, { - "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", - "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION + 7 * ( DTLB_STORE_MISSES.WALK_COMPLETED + DTLB_LOAD_MISSES.WALK_COMPLETED + ITLB_MISSES.WALK_COMPLETED ) ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )", - "MetricGroup": "Mem;MemoryTLB_SMT", - "MetricName": "Page_Walks_Utilization_SMT" - }, - { "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]", "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time", "MetricGroup": "Mem;MemoryBW", @@ -355,19 +836,19 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, @@ -385,26 +866,26 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -422,13 +903,13 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "( 64 * ( uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@ ) / 1000000000 ) / duration_time", + "MetricExpr": "(64 * (uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@) / 1000000000) / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, { "BriefDescription": "Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches", - "MetricExpr": "1000000000 * ( cbox@event\\=0x36\\,umask\\=0x3\\,filter_opc\\=0x182@ / cbox@event\\=0x35\\,umask\\=0x3\\,filter_opc\\=0x182@ ) / ( cbox_0@event\\=0x0@ / duration_time )", + "MetricExpr": "1000000000 * (cbox@event\\=0x36\\,umask\\=0x3\\,filter_opc\\=0x182@ / cbox@event\\=0x35\\,umask\\=0x3\\,filter_opc\\=0x182@) / (Socket_CLKS / duration_time)", "MetricGroup": "Mem;MemoryLat;SoC", "MetricName": "MEM_Read_Latency" }, @@ -445,12 +926,6 @@ "MetricName": "Socket_CLKS" }, { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "cbox_0@event\\=0x0@ / #num_dies / duration_time / 1000000000", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, - { "BriefDescription": "Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]", "MetricExpr": "INST_RETIRED.ANY / BR_INST_RETIRED.FAR_BRANCH:u", "MetricGroup": "Branches;OS", @@ -499,20 +974,19 @@ "MetricName": "C7_Pkg_Residency" }, { + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "Socket_CLKS / #num_dies / duration_time / 1000000000", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ" + }, + { "BriefDescription": "CPU operating frequency (in GHz)", - "MetricExpr": "( CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000", + "MetricExpr": "(( CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000) / duration_time", "MetricGroup": "", "MetricName": "cpu_operating_frequency", "ScaleUnit": "1GHz" }, { - "BriefDescription": "Cycles per instruction retired; indicating how much time each executed instruction took; in units of cycles.", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY", - "MetricGroup": "", - "MetricName": "cpi", - "ScaleUnit": "1per_instr" - }, - { "BriefDescription": "The ratio of number of completed memory load instructions to the total number completed instructions", "MetricExpr": "MEM_UOPS_RETIRED.ALL_LOADS / INST_RETIRED.ANY", "MetricGroup": "", @@ -530,7 +1004,7 @@ "BriefDescription": "Ratio of number of requests missing L1 data cache (includes data+rfo w/ prefetches) to the total number of completed instructions", "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l1d_mpi_includes_data_plus_rfo_with_prefetches", + "MetricName": "l1d_mpi", "ScaleUnit": "1per_instr" }, { @@ -558,7 +1032,7 @@ "BriefDescription": "Ratio of number of requests missing L2 cache (includes code+data+rfo w/ prefetches) to the total number of completed instructions", "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches", + "MetricName": "l2_mpi", "ScaleUnit": "1per_instr" }, { @@ -591,21 +1065,21 @@ }, { "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) in nano seconds", - "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time", + "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to local memory in nano seconds", - "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time", + "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_local_requests", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to remote memory in nano seconds", - "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time", + "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_remote_requests", "ScaleUnit": "1ns" @@ -640,21 +1114,21 @@ }, { "BriefDescription": "Memory read that miss the last level cache (LLC) addressed to local DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", - "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )", + "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_local_dram", + "MetricName": "numa_reads_addressed_to_local_dram", "ScaleUnit": "1%" }, { "BriefDescription": "Memory reads that miss the last level cache (LLC) addressed to remote DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", - "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )", + "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_remote_dram", + "MetricName": "numa_reads_addressed_to_remote_dram", "ScaleUnit": "1%" }, { "BriefDescription": "Uncore operating frequency in GHz", - "MetricExpr": "UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) / 1000000000", + "MetricExpr": "( UNC_C_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) / 1000000000) / duration_time", "MetricGroup": "", "MetricName": "uncore_frequency", "ScaleUnit": "1GHz" @@ -663,7 +1137,7 @@ "BriefDescription": "Intel(R) Quick Path Interconnect (QPI) data transmit bandwidth (MB/sec)", "MetricExpr": "( UNC_Q_TxL_FLITS_G0.DATA * 8 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "qpi_data_transmit_bw_only_data", + "MetricName": "qpi_data_transmit_bw", "ScaleUnit": "1MB/s" }, { @@ -691,245 +1165,42 @@ "BriefDescription": "Bandwidth of IO reads that are initiated by end device controllers that are requesting memory from the CPU.", "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x19e@ * 64 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_read", + "MetricName": "io_bandwidth_disk_or_network_writes", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Bandwidth of IO writes that are initiated by end device controllers that are writing memory to the CPU.", "MetricExpr": "(( cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x1c8\\,filter_tid\\=0x3e@ + cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x180\\,filter_tid\\=0x3e@ ) * 64 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_write", + "MetricName": "io_bandwidth_disk_or_network_reads", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Uops delivered from decoded instruction cache (decoded stream buffer or DSB) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.DSB_UOPS / UOPS_ISSUED.ANY )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_decoded_icache_dsb", + "MetricName": "percent_uops_delivered_from_decoded_icache", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from legacy decode pipeline (Micro-instruction Translation Engine or MITE) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.MITE_UOPS / UOPS_ISSUED.ANY )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline_mite", + "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from microcode sequencer (MS) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.MS_UOPS / UOPS_ISSUED.ANY )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_microcode_sequencer_ms", + "MetricName": "percent_uops_delivered_from_microcode_sequencer", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from loop stream detector(LSD) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( LSD.UOPS / UOPS_ISSUED.ANY )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_loop_stream_detector_lsd", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", - "MetricExpr": "100 * ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1;PGO", - "MetricName": "tma_frontend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.", - "MetricExpr": "100 * ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "Frontend;TmaL2;m_tma_frontend_bound_percent", - "MetricName": "tma_fetch_latency_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", - "MetricExpr": "100 * ( ICACHE.IFDATA_STALL / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_icache_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.", - "MetricExpr": "100 * ( ( 14 * ITLB_MISSES.STLB_HIT + cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=0x1@ + 7 * ITLB_MISSES.WALK_COMPLETED ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_itlb_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.", - "MetricExpr": "100 * ( ( 12 ) * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_branch_resteers_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", - "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_dsb_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", - "MetricExpr": "100 * ( ILD_STALL.LCP / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_lcp_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.", - "MetricExpr": "100 * ( ( 2 ) * IDQ.MS_SWITCHES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_ms_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", - "MetricExpr": "100 * ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent", - "MetricName": "tma_fetch_bandwidth_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", - "MetricExpr": "100 * ( ( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )", - "MetricGroup": "DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent", - "MetricName": "tma_mite_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", - "MetricExpr": "100 * ( ( IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )", - "MetricGroup": "DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent", - "MetricName": "tma_dsb_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", - "MetricExpr": "100 * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_bad_speculation_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.", - "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent", - "MetricName": "tma_branch_mispredicts_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.", - "MetricExpr": "100 * ( ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )", - "MetricGroup": "BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent", - "MetricName": "tma_machine_clears_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", - "MetricExpr": "100 * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_backend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", - "MetricExpr": "100 * ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) ) * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) )", - "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent", - "MetricName": "tma_memory_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.", - "MetricExpr": "100 * ( max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) , 0 ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l1_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l2_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l3_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.", - "MetricExpr": "100 * ( min( ( ( 1 - ( MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) ) , ( 1 ) ) )", - "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_dram_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.", - "MetricExpr": "100 * ( RESOURCE_STALLS.SB / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_store_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", - "MetricExpr": "100 * ( ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) - ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) ) * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) ) )", - "MetricGroup": "Backend;TmaL2;Compute;m_tma_backend_bound_percent", - "MetricName": "tma_core_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.", - "MetricExpr": "100 * ( ARITH.FPU_DIV_ACTIVE / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) )", - "MetricGroup": "TmaL3;m_tma_core_bound_percent", - "MetricName": "tma_divider_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", - "MetricExpr": "100 * ( ( ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) - RESOURCE_STALLS.SB - CYCLE_ACTIVITY.STALLS_MEM_ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "PortsUtil;TmaL3;m_tma_core_bound_percent", - "MetricName": "tma_ports_utilization_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. ", - "MetricExpr": "100 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_retiring_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved.", - "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )", - "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", - "MetricName": "tma_light_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", - "MetricExpr": "100 * ( ( INST_RETIRED.X87 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / INST_RETIRED.ANY ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) )", - "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_fp_arith_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", - "MetricName": "tma_heavy_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided.", - "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "MicroSeq;TmaL3;m_tma_heavy_operations_percent", - "MetricName": "tma_microcode_sequencer_percent", + "MetricName": "percent_uops_delivered_from_loop_stream_detector", "ScaleUnit": "1%" } ] diff --git a/tools/perf/pmu-events/arch/x86/broadwellx/uncore-cache.json b/tools/perf/pmu-events/arch/x86/broadwellx/uncore-cache.json index abee6f773c1f..449fa723d0aa 100644 --- a/tools/perf/pmu-events/arch/x86/broadwellx/uncore-cache.json +++ b/tools/perf/pmu-events/arch/x86/broadwellx/uncore-cache.json @@ -947,21 +947,19 @@ "Unit": "CBO" }, { - "BriefDescription": "LLC misses - demand and prefetch data reads - excludes LLC prefetches. Derived from unc_c_tor_inserts.miss_opcode", + "BriefDescription": "TOR Inserts; Miss Opcode Match", "Counter": "0,1,2,3", "EventCode": "0x35", - "EventName": "LLC_MISSES.DATA_READ", - "Filter": "filter_opc=0x182", + "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE", "PerPkg": "1", - "ScaleUnit": "64Bytes", "UMask": "0x3", "Unit": "CBO" }, { - "BriefDescription": "LLC misses - demand and prefetch data reads - excludes LLC prefetches", + "BriefDescription": "LLC misses - demand and prefetch data reads - excludes LLC prefetches. Derived from unc_c_tor_inserts.miss_opcode", "Counter": "0,1,2,3", "EventCode": "0x35", - "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE", + "EventName": "LLC_MISSES.DATA_READ", "Filter": "filter_opc=0x182", "PerPkg": "1", "ScaleUnit": "64Bytes", diff --git a/tools/perf/pmu-events/arch/x86/broadwellx/uncore-interconnect.json b/tools/perf/pmu-events/arch/x86/broadwellx/uncore-interconnect.json index 071ce45620d2..cb1916f52607 100644 --- a/tools/perf/pmu-events/arch/x86/broadwellx/uncore-interconnect.json +++ b/tools/perf/pmu-events/arch/x86/broadwellx/uncore-interconnect.json @@ -685,36 +685,34 @@ "Unit": "QPI LL" }, { - "BriefDescription": "Number of data flits transmitted . Derived from unc_q_txl_flits_g0.data", + "BriefDescription": "Flits Transferred - Group 0; Data Tx Flits", "Counter": "0,1,2,3", - "EventName": "QPI_DATA_BANDWIDTH_TX", + "EventName": "UNC_Q_TxL_FLITS_G0.DATA", "PerPkg": "1", - "ScaleUnit": "8Bytes", "UMask": "0x2", "Unit": "QPI LL" }, { - "BriefDescription": "Number of data flits transmitted ", + "BriefDescription": "Number of data flits transmitted . Derived from unc_q_txl_flits_g0.data", "Counter": "0,1,2,3", - "EventName": "UNC_Q_TxL_FLITS_G0.DATA", + "EventName": "QPI_DATA_BANDWIDTH_TX", "PerPkg": "1", "ScaleUnit": "8Bytes", "UMask": "0x2", "Unit": "QPI LL" }, { - "BriefDescription": "Number of non data (control) flits transmitted . Derived from unc_q_txl_flits_g0.non_data", + "BriefDescription": "Flits Transferred - Group 0; Non-Data protocol Tx Flits", "Counter": "0,1,2,3", - "EventName": "QPI_CTL_BANDWIDTH_TX", + "EventName": "UNC_Q_TxL_FLITS_G0.NON_DATA", "PerPkg": "1", - "ScaleUnit": "8Bytes", "UMask": "0x4", "Unit": "QPI LL" }, { - "BriefDescription": "Number of non data (control) flits transmitted ", + "BriefDescription": "Number of non data (control) flits transmitted . Derived from unc_q_txl_flits_g0.non_data", "Counter": "0,1,2,3", - "EventName": "UNC_Q_TxL_FLITS_G0.NON_DATA", + "EventName": "QPI_CTL_BANDWIDTH_TX", "PerPkg": "1", "ScaleUnit": "8Bytes", "UMask": "0x4", diff --git a/tools/perf/pmu-events/arch/x86/broadwellx/uncore-memory.json b/tools/perf/pmu-events/arch/x86/broadwellx/uncore-memory.json index 302e956a82ed..05fab7d2723e 100644 --- a/tools/perf/pmu-events/arch/x86/broadwellx/uncore-memory.json +++ b/tools/perf/pmu-events/arch/x86/broadwellx/uncore-memory.json @@ -72,20 +72,19 @@ "Unit": "iMC" }, { - "BriefDescription": "read requests to memory controller. Derived from unc_m_cas_count.rd", + "BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; All DRAM Reads (RD_CAS + Underfills)", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "LLC_MISSES.MEM_READ", + "EventName": "UNC_M_CAS_COUNT.RD", "PerPkg": "1", - "ScaleUnit": "64Bytes", "UMask": "0x3", "Unit": "iMC" }, { - "BriefDescription": "read requests to memory controller", + "BriefDescription": "read requests to memory controller. Derived from unc_m_cas_count.rd", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "UNC_M_CAS_COUNT.RD", + "EventName": "LLC_MISSES.MEM_READ", "PerPkg": "1", "ScaleUnit": "64Bytes", "UMask": "0x3", @@ -110,20 +109,19 @@ "Unit": "iMC" }, { - "BriefDescription": "write requests to memory controller. Derived from unc_m_cas_count.wr", + "BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; All DRAM WR_CAS (both Modes)", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "LLC_MISSES.MEM_WRITE", + "EventName": "UNC_M_CAS_COUNT.WR", "PerPkg": "1", - "ScaleUnit": "64Bytes", "UMask": "0xC", "Unit": "iMC" }, { - "BriefDescription": "write requests to memory controller", + "BriefDescription": "write requests to memory controller. Derived from unc_m_cas_count.wr", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "UNC_M_CAS_COUNT.WR", + "EventName": "LLC_MISSES.MEM_WRITE", "PerPkg": "1", "ScaleUnit": "64Bytes", "UMask": "0xC", diff --git a/tools/perf/pmu-events/arch/x86/cascadelakex/clx-metrics.json b/tools/perf/pmu-events/arch/x86/cascadelakex/clx-metrics.json index 46613504b816..81de1149297d 100644 --- a/tools/perf/pmu-events/arch/x86/cascadelakex/clx-metrics.json +++ b/tools/perf/pmu-events/arch/x86/cascadelakex/clx-metrics.json @@ -1,148 +1,742 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. Sample with: FRONTEND_RETIRED.LATENCY_GE_4_PS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: FRONTEND_RETIRED.LATENCY_GE_16_PS;FRONTEND_RETIRED.LATENCY_GE_8_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses", + "MetricExpr": "(ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@) / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses. Sample with: FRONTEND_RETIRED.L2_MISS_PS;FRONTEND_RETIRED.L1I_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "ICACHE_64B.IFTAG_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: FRONTEND_RETIRED.STLB_MISS_PS;FRONTEND_RETIRED.ITLB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "INT_MISC.CLEAR_RESTEER_CYCLES / CLKS + tma_unknown_branches", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears", + "MetricExpr": "(1 - (BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT))) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "9 * BACLEARS.ANY / CLKS", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: BACLEARS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty. Sample with: FRONTEND_RETIRED.DSB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "2 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend. Sample with: FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_2_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck. Sample with: FRONTEND_RETIRED.ANY_DSB_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where decoder-0 was the only active decoder", + "MetricExpr": "(cpu@INST_DECODED.DECODERS\\,cmask\\=1@ - cpu@INST_DECODED.DECODERS\\,cmask\\=2@) / CORE_CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_decoder0_alone", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - tma_frontend_bound - (UOPS_ISSUED.ANY + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_RETIRED.L1_HIT_PS;MEM_LOAD_RETIRED.FB_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "min(9 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE, max(CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS, 0)) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_INST_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_load - tma_load_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk", + "MetricExpr": "DTLB_LOAD_MISSES.WALK_ACTIVE / CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(12 * max(0, MEM_INST_RETIRED.LOCK_LOADS - L2_RQSTS.ALL_RFO) + (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES) * (11 * L2_RQSTS.RFO_HIT + min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO))) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_INST_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_INST_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "Load_Miss_Real_Latency * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) / ((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@)) * ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "((44 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM * (OCR.DEMAND_DATA_RD.L3_HIT.HITM_OTHER_CORE / (OCR.DEMAND_DATA_RD.L3_HIT.HITM_OTHER_CORE + OCR.DEMAND_DATA_RD.L3_HIT.HIT_OTHER_CORE_FWD))) + (44 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "(44 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM * (1 - (OCR.DEMAND_DATA_RD.L3_HIT.HITM_OTHER_CORE / (OCR.DEMAND_DATA_RD.L3_HIT.HITM_OTHER_CORE + OCR.DEMAND_DATA_RD.L3_HIT.HIT_OTHER_CORE_FWD)))) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "(17 * Average_Frequency) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "((OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2) if #SMT_on else OFFCORE_REQUESTS_BUFFER.SQ_FULL) / CORE_CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_L3_MISS / CLKS + ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS) - tma_l2_bound) - tma_pmm_bound)", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory", + "MetricExpr": "(59.5 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Server;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_local_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory. Caching will improve the latency and increase performance. Sample with: MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory", + "MetricExpr": "(127 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues", + "MetricExpr": "((89.5 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM + (89.5 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_cache", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM_PS;MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates (based on idle latencies) how often the CPU was stalled on accesses to external 3D-Xpoint (Crystal Ridge, a.k.a", + "MetricExpr": "(((1 - ((19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + 10 * ((MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))))) / ((19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + 10 * ((MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))))) + (25 * (MEM_LOAD_RETIRED.LOCAL_PMM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + 33 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))))))) * (CYCLE_ACTIVITY.STALLS_L3_MISS / CLKS + ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS) - tma_l2_bound)) if (1000000 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM) > MEM_LOAD_RETIRED.L1_MISS) else 0)", + "MetricGroup": "MemoryBound;Server;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_pmm_bound", + "PublicDescription": "This metric roughly estimates (based on idle latencies) how often the CPU was stalled on accesses to external 3D-Xpoint (Crystal Ridge, a.k.a. IXP) memory by loads, PMM stands for Persistent Memory Module. ", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "EXE_ACTIVITY.BOUND_ON_STORES / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_INST_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 11 * (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES))) + (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "((110 * Average_Frequency) * (OCR.DEMAND_RFO.L3_MISS.REMOTE_HITM + OCR.PF_L2_RFO.L3_MISS.REMOTE_HITM) + (47.5 * Average_Frequency) * (OCR.DEMAND_RFO.L3_HIT.HITM_OTHER_CORE + OCR.PF_L2_RFO.L3_HIT.HITM_OTHER_CORE)) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "MEM_INST_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_INST_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(9 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE) / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_INST_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_store - tma_store_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk", + "MetricExpr": "DTLB_STORE_MISSES.WALK_ACTIVE / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.DIVIDER_ACTIVE / CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_ACTIVE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "(EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL)) / CLKS if (ARITH.DIVIDER_ACTIVE < (CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY)) else (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(UOPS_EXECUTED.CORE_CYCLES_NONE / 2 if #SMT_on else CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations", + "MetricExpr": "PARTIAL_RAT_STALLS.SCOREBOARD / CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_serializing_operation", + "PublicDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out-of-order execution which may limit performance. Sample with: PARTIAL_RAT_STALLS.SCOREBOARD", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions", + "MetricExpr": "40 * ROB_MISC_EVENTS.PAUSE_INST / CLKS", + "MetricGroup": "TopdownL6;tma_serializing_operation_group", + "MetricName": "tma_slow_pause", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions. Sample with: MISC_RETIRED.PAUSE_INST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued", + "MetricExpr": "CLKS * UOPS_ISSUED.VECTOR_WIDTH_MISMATCH / UOPS_ISSUED.ANY", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_mixing_vectors", + "PublicDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "((UOPS_EXECUTED.CORE_CYCLES_GE_1 - UOPS_EXECUTED.CORE_CYCLES_GE_2) / 2 if #SMT_on else EXE_ACTIVITY.1_PORTS_UTIL) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "((UOPS_EXECUTED.CORE_CYCLES_GE_2 - UOPS_EXECUTED.CORE_CYCLES_GE_3) / 2 if #SMT_on else EXE_ACTIVITY.2_PORTS_UTIL) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).", + "MetricExpr": "(UOPS_EXECUTED.CORE_CYCLES_GE_3 / 2 if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_3) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_0 + UOPS_DISPATCHED_PORT.PORT_1 + UOPS_DISPATCHED_PORT.PORT_5 + UOPS_DISPATCHED_PORT.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED_PORT.PORT_0", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_1", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_6", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_2 + UOPS_DISPATCHED_PORT.PORT_3 + UOPS_DISPATCHED_PORT.PORT_7 - UOPS_DISPATCHED_PORT.PORT_4) / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_2", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_2 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_2", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_3", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_3 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data) Sample with: UOPS_DISPATCHED_PORT.PORT_4", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_4", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 7 ([HSW+]simple Store-address) Sample with: UOPS_DISPATCHED_PORT.PORT_7", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_7 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_7", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" }, { - "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", - "MetricExpr": "100 * ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "Bad;BadSpec;BrMispredicts", - "MetricName": "Mispredictions" + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "tma_retiring * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_512b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", + "MetricExpr": "tma_light_operations * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_memory_operations", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions", + "MetricExpr": "tma_light_operations * UOPS_RETIRED.MACRO_FUSED / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fused_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions. The instruction pairs of CMP+JCC or DEC+JCC are commonly used examples.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused", + "MetricExpr": "tma_light_operations * (BR_INST_RETIRED.ALL_BRANCHES - UOPS_RETIRED.MACRO_FUSED) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_non_fused_branches", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused. Non-conditional branches like direct JMP or CALL would count here. Can be used to examine fusible conditional jumps that were not fused.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions", + "MetricExpr": "tma_light_operations * INST_RETIRED.NOP / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_nop_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body. Sample with: INST_RETIRED.NOP", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", + "MetricExpr": "max(0, tma_light_operations - (tma_fp_arith + tma_memory_operations + tma_fused_instructions + tma_non_fused_branches + tma_nop_instructions))", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_other_light_ops", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY) / SLOTS", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops", + "MetricExpr": "tma_heavy_operations - tma_microcode_sequencer", + "MetricGroup": "TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_few_uops_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * (FP_ASSIST.ANY + OTHER_ASSISTS.ANY) / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: OTHER_ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" }, { "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", - "MetricExpr": "100 * ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "Bad;BadSpec;BrMispredicts_SMT", - "MetricName": "Mispredictions_SMT" + "MetricExpr": "100 * (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches))", + "MetricGroup": "Bad;BadSpec;BrMispredicts", + "MetricName": "Mispredictions" }, { "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) * ( ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) - ( ( ( 1 - ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) / ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) + ( 25 * ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) + 33 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) ) ) ) ) * (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) if ( 1000000 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD) / #( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) - ( ( ( 1 - ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) / ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) + ( 25 * ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) + 33 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) ) ) ) ) * (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) if ( 1000000 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (OFFCORE_REQUESTS_BUFFER.SQ_FULL / CPU_CLK_UNHALTED.THREAD) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) ) + ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( ((L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )) * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CPU_CLK_UNHALTED.THREAD) / #(max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) ", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_mem_bandwidth / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_sq_full / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full))) + (tma_l1_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_fb_full / (tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) ", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "Memory_Bandwidth" }, { - "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * ( ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) - ( ( ( 1 - ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) / ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) + ( 25 * ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) + 33 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) ) ) ) ) * (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) if ( 1000000 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD) / #( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) - ( ( ( 1 - ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) / ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) + ( 25 * ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) + 33 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) ) ) ) ) * (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) if ( 1000000 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (( OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2 ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) ) + ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( ((L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )) * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CPU_CLK_UNHALTED.THREAD) / #(max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) ", - "MetricGroup": "Mem;MemoryBW;Offcore_SMT", - "MetricName": "Memory_Bandwidth_SMT" - }, - { "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) * ( ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) - ( ( ( 1 - ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) / ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) + ( 25 * ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) + 33 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) ) ) ) ) * (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) if ( 1000000 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD ) / CPU_CLK_UNHALTED.THREAD - (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD)) / #( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) - ( ( ( 1 - ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) / ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) + ( 25 * ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) + 33 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) ) ) ) ) * (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) if ( 1000000 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (( (20.5 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) - (3.5 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) ) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CPU_CLK_UNHALTED.THREAD) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) + ( (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD)) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) )", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_mem_latency / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_l3_hit_latency / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full)) + (tma_l2_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)))", "MetricGroup": "Mem;MemoryLat;Offcore", "MetricName": "Memory_Latency" }, { - "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * ( ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) - ( ( ( 1 - ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) / ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) + ( 25 * ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) + 33 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) ) ) ) ) * (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) if ( 1000000 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD ) / CPU_CLK_UNHALTED.THREAD - (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD)) / #( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) - ( ( ( 1 - ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) / ( ( 19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + 10 * ( (MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) ) ) + ( 25 * ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) + 33 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) ) ) ) ) ) ) * (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) if ( 1000000 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (( (20.5 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) - (3.5 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) ) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CPU_CLK_UNHALTED.THREAD) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) + ( (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD)) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) )", - "MetricGroup": "Mem;MemoryLat;Offcore_SMT", - "MetricName": "Memory_Latency_SMT" - }, - { "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) * ( ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (min( 9 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE , max( CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS , 0 ) ) / CPU_CLK_UNHALTED.THREAD) / (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) + ( (EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (( 9 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE ) / CPU_CLK_UNHALTED.THREAD) / #(EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) ) ) ", + "MetricExpr": "100 * tma_memory_bound * ((tma_l1_bound / max(tma_memory_bound, tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_dtlb_load / max(tma_l1_bound, tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) + (tma_store_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_dtlb_store / (tma_dtlb_store + tma_false_sharing + tma_split_stores + tma_store_latency))) ", "MetricGroup": "Mem;MemoryTLB;Offcore", "MetricName": "Memory_Data_TLBs" }, { - "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * ( ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (min( 9 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE , max( CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS , 0 ) ) / CPU_CLK_UNHALTED.THREAD) / (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) + ( (EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (( 9 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / #(EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) ) ) ", - "MetricGroup": "Mem;MemoryTLB;Offcore_SMT", - "MetricName": "Memory_Data_TLBs_SMT" - }, - { "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.CONDITIONAL + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - ( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) - 2 * BR_INST_RETIRED.NEAR_CALL) ) / (4 * CPU_CLK_UNHALTED.THREAD))", + "MetricExpr": "100 * ((BR_INST_RETIRED.CONDITIONAL + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - (BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN) - 2 * BR_INST_RETIRED.NEAR_CALL)) / SLOTS)", "MetricGroup": "Ret", "MetricName": "Branching_Overhead" }, { - "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.CONDITIONAL + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - ( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) - 2 * BR_INST_RETIRED.NEAR_CALL) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))", - "MetricGroup": "Ret_SMT", - "MetricName": "Branching_Overhead_SMT" - }, - { "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", - "MetricExpr": "100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD))", + "MetricExpr": "100 * tma_fetch_latency * (tma_itlb_misses + tma_icache_misses + tma_unknown_branches) / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)", "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB", "MetricName": "Big_Code" }, { - "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", - "MetricExpr": "100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))", - "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB_SMT", - "MetricName": "Big_Code_SMT" - }, - { "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", - "MetricExpr": "100 * ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) - (100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)))", + "MetricExpr": "100 * (tma_frontend_bound - tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) - Big_Code", "MetricGroup": "Fed;FetchBW;Frontend", "MetricName": "Instruction_Fetch_BW" }, { - "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", - "MetricExpr": "100 * ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) - (100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))))", - "MetricGroup": "Fed;FetchBW;Frontend_SMT", - "MetricName": "Instruction_Fetch_BW_SMT" - }, - { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -159,6 +753,12 @@ "MetricName": "UpTB" }, { + "BriefDescription": "Cycles Per Instruction (per Logical Processor)", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", + "MetricName": "CPI" + }, + { "BriefDescription": "Per-Logical Processor actual clocks when the Logical Processor is active.", "MetricExpr": "CPU_CLK_UNHALTED.THREAD", "MetricGroup": "Pipeline", @@ -166,17 +766,11 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "The ratio of Executed- by Issued-Uops", "MetricExpr": "UOPS_EXECUTED.THREAD / UOPS_ISSUED.ANY", "MetricGroup": "Cor;Pipeline", @@ -185,63 +779,38 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { - "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;Flops_SMT", - "MetricName": "FLOPc_SMT" - }, - { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.THREAD )", + "MetricExpr": "((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)." }, { - "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )", - "MetricGroup": "Cor;Flops;HPC_SMT", - "MetricName": "FP_Arith_Utilization_SMT", - "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common). SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2 ) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", - "MetricExpr": "( 1 - ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)))) / ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) if ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)))) < ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) else 1 ) if 0 > 0.5 else 0", + "MetricExpr": "(1 - tma_core_bound / tma_ports_utilization if tma_core_bound < tma_ports_utilization else 1) if SMT_2T_Utilization > 0.5 else 0", "MetricGroup": "Cor;SMT", "MetricName": "Core_Bound_Likely" }, { - "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", - "MetricExpr": "( 1 - ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))))) / ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) if ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))))) < ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) else 1 ) if (1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 )) > 0.5 else 0", - "MetricGroup": "Cor;SMT_SMT", - "MetricName": "Core_Bound_Likely_SMT" - }, - { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, @@ -283,13 +852,13 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." @@ -310,21 +879,21 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX512", "PublicDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." @@ -336,9 +905,9 @@ "MetricName": "IpSWPF" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -373,17 +942,11 @@ }, { "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", - "MetricExpr": "100 * ( (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (DSB2MITE_SWITCHES.PENALTY_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + ((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD))) * (( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / CPU_CLK_UNHALTED.THREAD / 2) / #((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD))) )", + "MetricExpr": "100 * (tma_fetch_latency * tma_dsb_switches / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches) + tma_fetch_bandwidth * tma_mite / (tma_dsb + tma_mite))", "MetricGroup": "DSBmiss;Fed", "MetricName": "DSB_Misses" }, { - "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", - "MetricExpr": "100 * ( (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * (DSB2MITE_SWITCHES.PENALTY_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + ((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * (( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) / 2) / #((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) )", - "MetricGroup": "DSBmiss;Fed_SMT", - "MetricName": "DSB_Misses_SMT" - }, - { "BriefDescription": "Number of Instructions per non-speculative DSB miss (lower number means higher occurrence rate)", "MetricExpr": "INST_RETIRED.ANY / FRONTEND_RETIRED.ANY_DSB_MISS", "MetricGroup": "DSBmiss;Fed", @@ -397,17 +960,11 @@ }, { "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;BrMispredicts", "MetricName": "Branch_Misprediction_Cost" }, { - "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES", - "MetricGroup": "Bad;BrMispredicts_SMT", - "MetricName": "Branch_Misprediction_Cost_SMT" - }, - { "BriefDescription": "Fraction of branches that are non-taken conditionals", "MetricExpr": "BR_INST_RETIRED.NOT_TAKEN / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", @@ -415,102 +972,96 @@ }, { "BriefDescription": "Fraction of branches that are taken conditionals", - "MetricExpr": "( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", "MetricName": "Cond_TK" }, { "BriefDescription": "Fraction of branches that are CALL or RET", - "MetricExpr": "( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "CallRet" }, { "BriefDescription": "Fraction of branches that are unconditional (direct or indirect) jumps", - "MetricExpr": "(BR_INST_RETIRED.NEAR_TAKEN - ( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_TAKEN - (BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN) - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "Jump" }, { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L1 cache true misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.ALL_DEMAND_DATA_RD / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI_Load" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load" }, { "BriefDescription": "L2 cache hits per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( L2_RQSTS.REFERENCES - L2_RQSTS.MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricExpr": "1000 * (L2_RQSTS.REFERENCES - L2_RQSTS.MISS) / INST_RETIRED.ANY", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_All" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries)", "MetricExpr": "1000 * MEM_LOAD_RETIRED.FB_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "FB_HPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING + EPT.WALK_PENDING ) / ( 2 * CPU_CLK_UNHALTED.THREAD )", + "MetricExpr": "(ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING + EPT.WALK_PENDING) / (2 * CORE_CLKS)", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, { - "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING + EPT.WALK_PENDING ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )", - "MetricGroup": "Mem;MemoryTLB_SMT", - "MetricName": "Page_Walks_Utilization_SMT" - }, - { "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]", "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time", "MetricGroup": "Mem;MemoryBW", @@ -536,37 +1087,37 @@ }, { "BriefDescription": "Rate of silent evictions from the L2 cache per Kilo instruction where the evicted lines are dropped (no writeback to L3 or memory)", - "MetricExpr": "1000 * L2_LINES_OUT.SILENT / INST_RETIRED.ANY", + "MetricExpr": "1000 * L2_LINES_OUT.SILENT / Instructions", "MetricGroup": "L2Evicts;Mem;Server", "MetricName": "L2_Evictions_Silent_PKI" }, { "BriefDescription": "Rate of non silent evictions from the L2 cache per Kilo instruction", - "MetricExpr": "1000 * L2_LINES_OUT.NON_SILENT / INST_RETIRED.ANY", + "MetricExpr": "1000 * L2_LINES_OUT.NON_SILENT / Instructions", "MetricGroup": "L2Evicts;Mem;Server", "MetricName": "L2_Evictions_NonSilent_PKI" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data access bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * OFFCORE_REQUESTS.ALL_REQUESTS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Access_BW", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "L3_Cache_Access_BW_1T" }, @@ -578,68 +1129,47 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0", - "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / 2 / CORE_CLKS if #SMT_on else CORE_POWER.LVL0_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License0_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. This includes non-AVX codes, SSE, AVX 128-bit, and low-current AVX 256-bit codes." }, { - "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / 2 / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Power_SMT", - "MetricName": "Power_License0_Utilization_SMT", - "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. This includes non-AVX codes, SSE, AVX 128-bit, and low-current AVX 256-bit codes. SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1", - "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / 2 / CORE_CLKS if #SMT_on else CORE_POWER.LVL1_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License1_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1. This includes high current AVX 256-bit instructions as well as low current AVX 512-bit instructions." }, { - "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / 2 / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Power_SMT", - "MetricName": "Power_License1_Utilization_SMT", - "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1. This includes high current AVX 256-bit instructions as well as low current AVX 512-bit instructions. SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX)", - "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / 2 / CORE_CLKS if #SMT_on else CORE_POWER.LVL2_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License2_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). This includes high current AVX 512-bit instructions." }, { - "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / 2 / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Power_SMT", - "MetricName": "Power_License2_Utilization_SMT", - "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). This includes high current AVX 512-bit instructions. SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -657,13 +1187,13 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "( 64 * ( uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@ ) / 1000000000 ) / duration_time", + "MetricExpr": "(64 * (uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@) / 1000000000) / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, { "BriefDescription": "Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches", - "MetricExpr": "1000000000 * ( cha@event\\=0x36\\,umask\\=0x21\\,config\\=0x40433@ / cha@event\\=0x35\\,umask\\=0x21\\,config\\=0x40433@ ) / ( cha_0@event\\=0x0@ / duration_time )", + "MetricExpr": "1000000000 * (cha@event\\=0x36\\,umask\\=0x21\\,config\\=0x40433@ / cha@event\\=0x35\\,umask\\=0x21\\,config\\=0x40433@) / (Socket_CLKS / duration_time)", "MetricGroup": "Mem;MemoryLat;SoC", "MetricName": "MEM_Read_Latency" }, @@ -675,38 +1205,38 @@ }, { "BriefDescription": "Average latency of data read request to external 3D X-Point memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches", - "MetricExpr": "( 1000000000 * ( imc@event\\=0xe0\\,umask\\=0x1@ / imc@event\\=0xe3@ ) / imc_0@event\\=0x0@ )", - "MetricGroup": "Mem;MemoryLat;SoC;Server", + "MetricExpr": "(1000000000 * (imc@event\\=0xe0\\,umask\\=0x1@ / imc@event\\=0xe3@) / imc_0@event\\=0x0@)", + "MetricGroup": "Mem;MemoryLat;Server;SoC", "MetricName": "MEM_PMM_Read_Latency" }, { "BriefDescription": "Average latency of data read request to external DRAM memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches", - "MetricExpr": "1000000000 * ( UNC_M_RPQ_OCCUPANCY / UNC_M_RPQ_INSERTS ) / imc_0@event\\=0x0@", - "MetricGroup": "Mem;MemoryLat;SoC;Server", + "MetricExpr": "1000000000 * (UNC_M_RPQ_OCCUPANCY / UNC_M_RPQ_INSERTS) / imc_0@event\\=0x0@", + "MetricGroup": "Mem;MemoryLat;Server;SoC", "MetricName": "MEM_DRAM_Read_Latency" }, { "BriefDescription": "Average 3DXP Memory Bandwidth Use for reads [GB / sec]", - "MetricExpr": "( ( 64 * imc@event\\=0xe3@ / 1000000000 ) / duration_time )", - "MetricGroup": "Mem;MemoryBW;SoC;Server", + "MetricExpr": "((64 * imc@event\\=0xe3@ / 1000000000) / duration_time)", + "MetricGroup": "Mem;MemoryBW;Server;SoC", "MetricName": "PMM_Read_BW" }, { "BriefDescription": "Average 3DXP Memory Bandwidth Use for Writes [GB / sec]", - "MetricExpr": "( ( 64 * imc@event\\=0xe7@ / 1000000000 ) / duration_time )", - "MetricGroup": "Mem;MemoryBW;SoC;Server", + "MetricExpr": "((64 * imc@event\\=0xe7@ / 1000000000) / duration_time)", + "MetricGroup": "Mem;MemoryBW;Server;SoC", "MetricName": "PMM_Write_BW" }, { "BriefDescription": "Average IO (network or disk) Bandwidth Use for Writes [GB / sec]", - "MetricExpr": "( UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART0 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART1 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART2 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART3 ) * 4 / 1000000000 / duration_time", - "MetricGroup": "IoBW;Mem;SoC;Server", + "MetricExpr": "(UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART0 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART1 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART2 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART3) * 4 / 1000000000 / duration_time", + "MetricGroup": "IoBW;Mem;Server;SoC", "MetricName": "IO_Write_BW" }, { "BriefDescription": "Average IO (network or disk) Bandwidth Use for Reads [GB / sec]", - "MetricExpr": "( UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART0 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART1 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART2 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART3 ) * 4 / 1000000000 / duration_time", - "MetricGroup": "IoBW;Mem;SoC;Server", + "MetricExpr": "(UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART0 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART1 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART2 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART3) * 4 / 1000000000 / duration_time", + "MetricGroup": "IoBW;Mem;Server;SoC", "MetricName": "IO_Read_BW" }, { @@ -716,12 +1246,6 @@ "MetricName": "Socket_CLKS" }, { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "cha_0@event\\=0x0@ / #num_dies / duration_time / 1000000000", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, - { "BriefDescription": "Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]", "MetricExpr": "INST_RETIRED.ANY / BR_INST_RETIRED.FAR_BRANCH:u", "MetricGroup": "Branches;OS", @@ -770,27 +1294,19 @@ "MetricName": "C7_Pkg_Residency" }, { - "BriefDescription": "Percentage of time spent in the active CPU power state C0", - "MetricExpr": "100 * CPU_CLK_UNHALTED.REF_TSC / TSC", - "MetricGroup": "", - "MetricName": "cpu_utilization_percent", - "ScaleUnit": "1%" + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "Socket_CLKS / #num_dies / duration_time / 1000000000", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ" }, { "BriefDescription": "CPU operating frequency (in GHz)", - "MetricExpr": "( CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000", + "MetricExpr": "(( CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000) / duration_time", "MetricGroup": "", "MetricName": "cpu_operating_frequency", "ScaleUnit": "1GHz" }, { - "BriefDescription": "Cycles per instruction retired; indicating how much time each executed instruction took; in units of cycles.", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY", - "MetricGroup": "", - "MetricName": "cpi", - "ScaleUnit": "1per_instr" - }, - { "BriefDescription": "The ratio of number of completed memory load instructions to the total number completed instructions", "MetricExpr": "MEM_INST_RETIRED.ALL_LOADS / INST_RETIRED.ANY", "MetricGroup": "", @@ -808,7 +1324,7 @@ "BriefDescription": "Ratio of number of requests missing L1 data cache (includes data+rfo w/ prefetches) to the total number of completed instructions", "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l1d_mpi_includes_data_plus_rfo_with_prefetches", + "MetricName": "l1d_mpi", "ScaleUnit": "1per_instr" }, { @@ -836,7 +1352,7 @@ "BriefDescription": "Ratio of number of requests missing L2 cache (includes code+data+rfo w/ prefetches) to the total number of completed instructions", "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches", + "MetricName": "l2_mpi", "ScaleUnit": "1per_instr" }, { @@ -869,21 +1385,21 @@ }, { "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) in nano seconds", - "MetricExpr": "( ( 1000000000 * ( cha@unc_cha_tor_occupancy.ia_miss\\,config1\\=0x4043300000000@ / cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043300000000@ ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_CLOCKTICKS) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( cha@unc_cha_tor_occupancy.ia_miss\\,config1\\=0x4043300000000@ / cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043300000000@ ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_CLOCKTICKS) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to local memory in nano seconds", - "MetricExpr": "( ( 1000000000 * ( cha@unc_cha_tor_occupancy.ia_miss\\,config1\\=0x4043200000000@ / cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043200000000@ ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_CLOCKTICKS) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( cha@unc_cha_tor_occupancy.ia_miss\\,config1\\=0x4043200000000@ / cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043200000000@ ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_CLOCKTICKS) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_local_requests", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to remote memory in nano seconds", - "MetricExpr": "( ( 1000000000 * ( cha@unc_cha_tor_occupancy.ia_miss\\,config1\\=0x4043100000000@ / cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043100000000@ ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_CLOCKTICKS) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( cha@unc_cha_tor_occupancy.ia_miss\\,config1\\=0x4043100000000@ / cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043100000000@ ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_CLOCKTICKS) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_remote_requests", "ScaleUnit": "1ns" @@ -892,54 +1408,54 @@ "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the ITLB (Instruction TLB) and further levels of TLB.", "MetricExpr": "ITLB_MISSES.WALK_COMPLETED / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "itlb_2nd_level_mpi", + "MetricName": "itlb_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for 2 megabyte and 4 megabyte page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the Instruction Translation Lookaside Buffer (ITLB) and further levels of TLB.", "MetricExpr": "ITLB_MISSES.WALK_COMPLETED_2M_4M / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "itlb_2nd_level_large_page_mpi", + "MetricName": "itlb_large_page_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by demand data loads to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.", "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "dtlb_2nd_level_load_mpi", + "MetricName": "dtlb_load_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for 2 megabyte page sizes) caused by demand data loads to the total number of completed instructions. This implies it missed in the Data Translation Lookaside Buffer (DTLB) and further levels of TLB.", "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED_2M_4M / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "dtlb_2nd_level_2mb_large_page_load_mpi", + "MetricName": "dtlb_2mb_large_page_load_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by demand data stores to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.", "MetricExpr": "DTLB_STORE_MISSES.WALK_COMPLETED / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "dtlb_2nd_level_store_mpi", + "MetricName": "dtlb_store_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Memory read that miss the last level cache (LLC) addressed to local DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", "MetricExpr": "100 * cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043200000000@ / ( cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043200000000@ + cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043100000000@ )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_local_dram", + "MetricName": "numa_reads_addressed_to_local_dram", "ScaleUnit": "1%" }, { "BriefDescription": "Memory reads that miss the last level cache (LLC) addressed to remote DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", "MetricExpr": "100 * cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043100000000@ / ( cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043200000000@ + cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043100000000@ )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_remote_dram", + "MetricName": "numa_reads_addressed_to_remote_dram", "ScaleUnit": "1%" }, { "BriefDescription": "Uncore operating frequency in GHz", - "MetricExpr": "UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_CLOCKTICKS) * #num_packages ) / 1000000000", + "MetricExpr": "( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_CLOCKTICKS) * #num_packages ) / 1000000000) / duration_time", "MetricGroup": "", "MetricName": "uncore_frequency", "ScaleUnit": "1GHz" @@ -948,7 +1464,7 @@ "BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data transmit bandwidth (MB/sec)", "MetricExpr": "( UNC_UPI_TxL_FLITS.ALL_DATA * (64 / 9.0) / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "upi_data_transmit_bw_only_data", + "MetricName": "upi_data_transmit_bw", "ScaleUnit": "1MB/s" }, { @@ -997,35 +1513,35 @@ "BriefDescription": "Bandwidth of IO reads that are initiated by end device controllers that are requesting memory from the CPU.", "MetricExpr": "(( UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART0 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART1 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART2 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART3 ) * 4 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_read", + "MetricName": "io_bandwidth_disk_or_network_writes", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Bandwidth of IO writes that are initiated by end device controllers that are writing memory to the CPU.", "MetricExpr": "(( UNC_IIO_PAYLOAD_BYTES_IN.MEM_WRITE.PART0 + UNC_IIO_PAYLOAD_BYTES_IN.MEM_WRITE.PART1 + UNC_IIO_PAYLOAD_BYTES_IN.MEM_WRITE.PART2 + UNC_IIO_PAYLOAD_BYTES_IN.MEM_WRITE.PART3 ) * 4 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_write", + "MetricName": "io_bandwidth_disk_or_network_reads", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Uops delivered from decoded instruction cache (decoded stream buffer or DSB) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.DSB_UOPS / ( IDQ.DSB_UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS + LSD.UOPS ) )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_decoded_icache_dsb", + "MetricName": "percent_uops_delivered_from_decoded_icache", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from legacy decode pipeline (Micro-instruction Translation Engine or MITE) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.MITE_UOPS / ( IDQ.DSB_UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS + LSD.UOPS ) )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline_mite", + "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from microcode sequencer (MS) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.MS_UOPS / ( IDQ.DSB_UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS + LSD.UOPS ) )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_microcode_sequencer_ms", + "MetricName": "percent_uops_delivered_from_microcode_sequencer", "ScaleUnit": "1%" }, { @@ -1050,255 +1566,10 @@ "ScaleUnit": "1MB/s" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", - "MetricExpr": "100 * ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1;PGO", - "MetricName": "tma_frontend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.", - "MetricExpr": "100 * ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "Frontend;TmaL2;m_tma_frontend_bound_percent", - "MetricName": "tma_fetch_latency_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", - "MetricExpr": "100 * ( ( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=0x1\\,edge\\=0x1@ ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_icache_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.", - "MetricExpr": "100 * ( ICACHE_64B.IFTAG_STALL / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_itlb_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.", - "MetricExpr": "100 * ( INT_MISC.CLEAR_RESTEER_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) + ( ( 9 ) * BACLEARS.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) )", - "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_branch_resteers_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", - "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_dsb_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", - "MetricExpr": "100 * ( ILD_STALL.LCP / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_lcp_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.", - "MetricExpr": "100 * ( ( 2 ) * IDQ.MS_SWITCHES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_ms_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", - "MetricExpr": "100 * ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent", - "MetricName": "tma_fetch_bandwidth_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", - "MetricExpr": "100 * ( ( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )", - "MetricGroup": "DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent", - "MetricName": "tma_mite_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", - "MetricExpr": "100 * ( ( IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )", - "MetricGroup": "DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent", - "MetricName": "tma_dsb_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", - "MetricExpr": "100 * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_bad_speculation_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.", - "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent", - "MetricName": "tma_branch_mispredicts_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.", - "MetricExpr": "100 * ( ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )", - "MetricGroup": "BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent", - "MetricName": "tma_machine_clears_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", - "MetricExpr": "100 * ( 1 - ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( UOPS_ISSUED.ANY + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_backend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", - "MetricExpr": "100 * ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / ( CYCLE_ACTIVITY.STALLS_TOTAL + ( EXE_ACTIVITY.1_PORTS_UTIL + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * EXE_ACTIVITY.2_PORTS_UTIL ) + EXE_ACTIVITY.BOUND_ON_STORES ) ) * ( 1 - ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( UOPS_ISSUED.ANY + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent", - "MetricName": "tma_memory_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.", - "MetricExpr": "100 * ( max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) , 0 ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l1_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) / ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=0x1@ ) ) * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l2_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l3_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.", - "MetricExpr": "100 * ( min( ( ( ( CYCLE_ACTIVITY.STALLS_L3_MISS / ( CPU_CLK_UNHALTED.THREAD ) + ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) - ( ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) / ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=0x1@ ) ) * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( min( ( ( ( ( 1 - ( ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) / ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) + ( 25 * ( ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) + 33 * ( ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) ) ) ) ) * ( CYCLE_ACTIVITY.STALLS_L3_MISS / ( CPU_CLK_UNHALTED.THREAD ) + ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) - ( ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) / ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=0x1@ ) ) * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) if ( ( 1000000 ) * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) , ( 1 ) ) ) ) ) , ( 1 ) ) )", - "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_dram_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric roughly estimates (based on idle latencies) how often the CPU was stalled on accesses to external 3D-Xpoint (Crystal Ridge, a.k.a. IXP) memory by loads, PMM stands for Persistent Memory Module. ", - "MetricExpr": "100 * ( min( ( ( ( ( 1 - ( ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) / ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) + ( 25 * ( ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) + 33 * ( ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) ) ) ) ) * ( CYCLE_ACTIVITY.STALLS_L3_MISS / ( CPU_CLK_UNHALTED.THREAD ) + ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) - ( ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) / ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=0x1@ ) ) * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) if ( ( 1000000 ) * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) , ( 1 ) ) )", - "MetricGroup": "MemoryBound;Server;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_pmm_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.", - "MetricExpr": "100 * ( EXE_ACTIVITY.BOUND_ON_STORES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_store_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", - "MetricExpr": "100 * ( ( 1 - ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( UOPS_ISSUED.ANY + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / ( CYCLE_ACTIVITY.STALLS_TOTAL + ( EXE_ACTIVITY.1_PORTS_UTIL + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * EXE_ACTIVITY.2_PORTS_UTIL ) + EXE_ACTIVITY.BOUND_ON_STORES ) ) * ( 1 - ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( UOPS_ISSUED.ANY + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )", - "MetricGroup": "Backend;TmaL2;Compute;m_tma_backend_bound_percent", - "MetricName": "tma_core_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.", - "MetricExpr": "100 * ( ARITH.DIVIDER_ACTIVE / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "TmaL3;m_tma_core_bound_percent", - "MetricName": "tma_divider_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", - "MetricExpr": "100 * ( ( EXE_ACTIVITY.EXE_BOUND_0_PORTS + ( EXE_ACTIVITY.1_PORTS_UTIL + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * EXE_ACTIVITY.2_PORTS_UTIL ) ) / ( CPU_CLK_UNHALTED.THREAD ) if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else ( EXE_ACTIVITY.1_PORTS_UTIL + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * EXE_ACTIVITY.2_PORTS_UTIL ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "PortsUtil;TmaL3;m_tma_core_bound_percent", - "MetricName": "tma_ports_utilization_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. ", - "MetricExpr": "100 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_retiring_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved.", - "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", - "MetricName": "tma_light_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) )", - "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_fp_arith_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_memory_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions. The instruction pairs of CMP+JCC or DEC+JCC are commonly used examples.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * UOPS_RETIRED.MACRO_FUSED / ( UOPS_RETIRED.RETIRE_SLOTS ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_fused_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused. Non-conditional branches like direct JMP or CALL would count here. Can be used to examine fusible conditional jumps that were not fused.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * ( BR_INST_RETIRED.ALL_BRANCHES - UOPS_RETIRED.MACRO_FUSED ) / ( UOPS_RETIRED.RETIRE_SLOTS ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_non_fused_branches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * INST_RETIRED.NOP / ( UOPS_RETIRED.RETIRE_SLOTS ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_nop_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", - "MetricExpr": "100 * ( max( 0 , ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) - ( ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) ) + ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY ) + ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * UOPS_RETIRED.MACRO_FUSED / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * ( BR_INST_RETIRED.ALL_BRANCHES - UOPS_RETIRED.MACRO_FUSED ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * INST_RETIRED.NOP / ( UOPS_RETIRED.RETIRE_SLOTS ) ) ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_other_light_ops_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", - "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", - "MetricName": "tma_heavy_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "TmaL3;m_tma_heavy_operations_percent", - "MetricName": "tma_few_uops_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided.", - "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "MicroSeq;TmaL3;m_tma_heavy_operations_percent", - "MetricName": "tma_microcode_sequencer_percent", + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to LSD (Loop Stream Detector) unit. LSD typically does well sustaining Uop supply. However; in some rare cases; optimal uop-delivery could not be reached for small loops whose size (in terms of number of uops) does not suit well the LSD structure.", + "MetricExpr": "100 * ( ( LSD.CYCLES_ACTIVE - LSD.CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )", + "MetricGroup": "FetchBW;LSD;TopdownL3;tma_L3_group;tma_fetch_bandwidth_group", + "MetricName": "tma_lsd", "ScaleUnit": "1%" } ] diff --git a/tools/perf/pmu-events/arch/x86/cascadelakex/uncore-memory.json b/tools/perf/pmu-events/arch/x86/cascadelakex/uncore-memory.json index 6facfb244cd3..326b674045c6 100644 --- a/tools/perf/pmu-events/arch/x86/cascadelakex/uncore-memory.json +++ b/tools/perf/pmu-events/arch/x86/cascadelakex/uncore-memory.json @@ -27,20 +27,19 @@ "Unit": "iMC" }, { - "BriefDescription": "read requests to memory controller. Derived from unc_m_cas_count.rd", + "BriefDescription": "All DRAM Read CAS Commands issued (including underfills)", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "LLC_MISSES.MEM_READ", + "EventName": "UNC_M_CAS_COUNT.RD", "PerPkg": "1", - "ScaleUnit": "64Bytes", "UMask": "0x3", "Unit": "iMC" }, { - "BriefDescription": "read requests to memory controller", + "BriefDescription": "read requests to memory controller. Derived from unc_m_cas_count.rd", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "UNC_M_CAS_COUNT.RD", + "EventName": "LLC_MISSES.MEM_READ", "PerPkg": "1", "ScaleUnit": "64Bytes", "UMask": "0x3", @@ -56,20 +55,19 @@ "Unit": "iMC" }, { - "BriefDescription": "write requests to memory controller. Derived from unc_m_cas_count.wr", + "BriefDescription": "All DRAM Write CAS commands issued", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "LLC_MISSES.MEM_WRITE", + "EventName": "UNC_M_CAS_COUNT.WR", "PerPkg": "1", - "ScaleUnit": "64Bytes", "UMask": "0xC", "Unit": "iMC" }, { - "BriefDescription": "write requests to memory controller", + "BriefDescription": "write requests to memory controller. Derived from unc_m_cas_count.wr", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "UNC_M_CAS_COUNT.WR", + "EventName": "LLC_MISSES.MEM_WRITE", "PerPkg": "1", "ScaleUnit": "64Bytes", "UMask": "0xC", diff --git a/tools/perf/pmu-events/arch/x86/cascadelakex/uncore-other.json b/tools/perf/pmu-events/arch/x86/cascadelakex/uncore-other.json index a29bba230f49..e10530c21ef8 100644 --- a/tools/perf/pmu-events/arch/x86/cascadelakex/uncore-other.json +++ b/tools/perf/pmu-events/arch/x86/cascadelakex/uncore-other.json @@ -1477,7 +1477,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x01", - "ScaleUnit": "4Bytes", "UMask": "0x01", "Unit": "IIO" }, @@ -1489,7 +1488,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x02", - "ScaleUnit": "4Bytes", "UMask": "0x01", "Unit": "IIO" }, @@ -1501,7 +1499,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x04", - "ScaleUnit": "4Bytes", "UMask": "0x01", "Unit": "IIO" }, @@ -1513,7 +1510,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x08", - "ScaleUnit": "4Bytes", "UMask": "0x01", "Unit": "IIO" }, @@ -1584,7 +1580,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x01", - "ScaleUnit": "4Bytes", "UMask": "0x04", "Unit": "IIO" }, @@ -1596,7 +1591,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x02", - "ScaleUnit": "4Bytes", "UMask": "0x04", "Unit": "IIO" }, @@ -1608,7 +1602,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x04", - "ScaleUnit": "4Bytes", "UMask": "0x04", "Unit": "IIO" }, @@ -1620,7 +1613,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x08", - "ScaleUnit": "4Bytes", "UMask": "0x04", "Unit": "IIO" }, @@ -2254,7 +2246,7 @@ "Unit": "UPI LL" }, { - "BriefDescription": "FLITs received which bypassed the Slot0 Receive Buffer", + "BriefDescription": "FLITs received which bypassed the Slot0 Recieve Buffer", "Counter": "0,1,2,3", "EventCode": "0x31", "EventName": "UNC_UPI_RxL_BYPASSED.SLOT2", diff --git a/tools/perf/pmu-events/arch/x86/haswell/cache.json b/tools/perf/pmu-events/arch/x86/haswell/cache.json index 3b0f3a264246..719b8e622f59 100644 --- a/tools/perf/pmu-events/arch/x86/haswell/cache.json +++ b/tools/perf/pmu-events/arch/x86/haswell/cache.json @@ -20,7 +20,7 @@ "UMask": "0x2" }, { - "BriefDescription": "L1D miss oustandings duration in cycles", + "BriefDescription": "L1D miss outstanding duration in cycles", "Counter": "2", "CounterHTOff": "2", "EventCode": "0x48", @@ -655,7 +655,7 @@ "UMask": "0x8" }, { - "BriefDescription": "Cacheable and noncachaeble code read requests", + "BriefDescription": "Cacheable and noncacheable code read requests", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0xB0", diff --git a/tools/perf/pmu-events/arch/x86/haswell/frontend.json b/tools/perf/pmu-events/arch/x86/haswell/frontend.json index c45a09abe5d3..18a993297108 100644 --- a/tools/perf/pmu-events/arch/x86/haswell/frontend.json +++ b/tools/perf/pmu-events/arch/x86/haswell/frontend.json @@ -161,7 +161,7 @@ "UMask": "0x4" }, { - "BriefDescription": "Cycles when uops are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Cycles when uops are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "CounterMask": "1", @@ -172,7 +172,7 @@ "UMask": "0x30" }, { - "BriefDescription": "Cycles when uops initiated by Decode Stream Buffer (DSB) are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy.", + "BriefDescription": "Cycles when uops initiated by Decode Stream Buffer (DSB) are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy.", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "CounterMask": "1", @@ -182,7 +182,7 @@ "UMask": "0x10" }, { - "BriefDescription": "Deliveries to Instruction Decode Queue (IDQ) initiated by Decode Stream Buffer (DSB) while Microcode Sequenser (MS) is busy.", + "BriefDescription": "Deliveries to Instruction Decode Queue (IDQ) initiated by Decode Stream Buffer (DSB) while Microcode Sequencer (MS) is busy.", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "CounterMask": "1", @@ -193,7 +193,7 @@ "UMask": "0x10" }, { - "BriefDescription": "Uops initiated by Decode Stream Buffer (DSB) that are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Uops initiated by Decode Stream Buffer (DSB) that are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0x79", @@ -203,7 +203,7 @@ "UMask": "0x10" }, { - "BriefDescription": "Uops initiated by MITE and delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Uops initiated by MITE and delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0x79", @@ -224,7 +224,7 @@ "UMask": "0x30" }, { - "BriefDescription": "Uops delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Uops delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0x79", diff --git a/tools/perf/pmu-events/arch/x86/haswell/hsw-metrics.json b/tools/perf/pmu-events/arch/x86/haswell/hsw-metrics.json index 75dc6dd9a7bc..6cb6603efbd8 100644 --- a/tools/perf/pmu-events/arch/x86/haswell/hsw-metrics.json +++ b/tools/perf/pmu-events/arch/x86/haswell/hsw-metrics.json @@ -1,64 +1,490 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * min(CPU_CLK_UNHALTED.THREAD, IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE) / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: RS_EVENTS.EMPTY_END", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", + "MetricExpr": "ICACHE.IFDATA_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "(14 * ITLB_MISSES.STLB_HIT + ITLB_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: ITLB_MISSES.WALK_COMPLETED", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "12 * (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY) / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "2 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - (tma_frontend_bound + tma_bad_speculation + tma_retiring)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING) + RESOURCE_STALLS.SB) / (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + (cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) if #SMT_on else (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING) - CYCLE_ACTIVITY.STALLS_L1D_PENDING) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_UOPS_RETIRED.L1_HIT_PS;MEM_LOAD_UOPS_RETIRED.HIT_LFB_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "(8 * DTLB_LOAD_MISSES.STLB_HIT + DTLB_LOAD_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_UOPS_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_UOPS_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_UOPS_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "Load_Miss_Real_Latency * cpu@L1D_PEND_MISS.REQUEST_FB_FULL\\,cmask\\=1@ / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L1D_PENDING - CYCLE_ACTIVITY.STALLS_L2_PENDING) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEM_LOAD_UOPS_RETIRED.L3_HIT / (MEM_LOAD_UOPS_RETIRED.L3_HIT + 7 * MEM_LOAD_UOPS_RETIRED.L3_MISS)) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "(60 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS))) + 43 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS)))) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "43 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS))) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "29 * (MEM_LOAD_UOPS_RETIRED.L3_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.L3_MISS))) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "((OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2) if #SMT_on else OFFCORE_REQUESTS_BUFFER.SQ_FULL) / CORE_CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(1 - (MEM_LOAD_UOPS_RETIRED.L3_HIT / (MEM_LOAD_UOPS_RETIRED.L3_HIT + 7 * MEM_LOAD_UOPS_RETIRED.L3_MISS))) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=6@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "RESOURCE_STALLS.SB / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_UOPS_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 9 * (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES))) + (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "60 * OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.HITM_OTHER_CORE / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "2 * MEM_UOPS_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_UOPS_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(8 * DTLB_STORE_MISSES.STLB_HIT + DTLB_STORE_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_UOPS_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "10 * ARITH.DIVIDER_UOPS / CORE_CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + (cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) if #SMT_on else (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) - RESOURCE_STALLS.SB - min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING)) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,inv\\,cmask\\=1@) / 2 if #SMT_on else (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else 0) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 if #SMT_on else (cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / 2 if #SMT_on else (cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).", + "MetricExpr": "((cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ / 2) if #SMT_on else cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_0 + UOPS_DISPATCHED_PORT.PORT_1 + UOPS_DISPATCHED_PORT.PORT_5 + UOPS_DISPATCHED_PORT.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED_PORT.PORT_0", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_1", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_6", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_2 + UOPS_DISPATCHED_PORT.PORT_3 + UOPS_DISPATCHED_PORT.PORT_7 - UOPS_DISPATCHED_PORT.PORT_4) / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_2", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_2 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_2", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_3", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_3 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data) Sample with: UOPS_DISPATCHED_PORT.PORT_4", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_4", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 7 ([HSW+]simple Store-address) Sample with: UOPS_DISPATCHED_PORT.PORT_7", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_7 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_7", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "INST_RETIRED.X87 * UPI / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * OTHER_ASSISTS.ANY_WB_ASSIST / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: OTHER_ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" }, { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -76,8 +502,8 @@ }, { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI" }, { @@ -88,37 +514,25 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "( UOPS_EXECUTED.CORE / 2 / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@) ) if #SMT_on else UOPS_EXECUTED.CORE / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@)", + "MetricExpr": "(UOPS_EXECUTED.CORE / 2 / ((cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2) if #SMT_on else cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@)) if #SMT_on else UOPS_EXECUTED.CORE / ((cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2) if #SMT_on else cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, @@ -159,9 +573,9 @@ "MetricName": "BpTkBranch" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -172,7 +586,7 @@ }, { "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)", - "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )", + "MetricExpr": "IDQ.DSB_UOPS / ((IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS))", "MetricGroup": "DSB;Fed;FetchBW", "MetricName": "DSB_Coverage" }, @@ -184,48 +598,42 @@ }, { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION ) / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "(ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION) / CORE_CLKS", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, { - "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", - "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Mem;MemoryTLB_SMT", - "MetricName": "Page_Walks_Utilization_SMT" - }, - { "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]", "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time", "MetricGroup": "Mem;MemoryBW", @@ -245,19 +653,19 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, @@ -275,19 +683,19 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -305,7 +713,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * ( arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@ ) / 1000000 / duration_time / 1000", + "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1000000 / duration_time / 1000", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, diff --git a/tools/perf/pmu-events/arch/x86/haswellx/cache.json b/tools/perf/pmu-events/arch/x86/haswellx/cache.json index 7557a203a1b6..427c949bed6e 100644 --- a/tools/perf/pmu-events/arch/x86/haswellx/cache.json +++ b/tools/perf/pmu-events/arch/x86/haswellx/cache.json @@ -691,7 +691,7 @@ "UMask": "0x8" }, { - "BriefDescription": "Cacheable and noncachaeble code read requests", + "BriefDescription": "Cacheable and noncacheable code read requests", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0xB0", diff --git a/tools/perf/pmu-events/arch/x86/haswellx/frontend.json b/tools/perf/pmu-events/arch/x86/haswellx/frontend.json index c45a09abe5d3..18a993297108 100644 --- a/tools/perf/pmu-events/arch/x86/haswellx/frontend.json +++ b/tools/perf/pmu-events/arch/x86/haswellx/frontend.json @@ -161,7 +161,7 @@ "UMask": "0x4" }, { - "BriefDescription": "Cycles when uops are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Cycles when uops are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "CounterMask": "1", @@ -172,7 +172,7 @@ "UMask": "0x30" }, { - "BriefDescription": "Cycles when uops initiated by Decode Stream Buffer (DSB) are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy.", + "BriefDescription": "Cycles when uops initiated by Decode Stream Buffer (DSB) are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy.", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "CounterMask": "1", @@ -182,7 +182,7 @@ "UMask": "0x10" }, { - "BriefDescription": "Deliveries to Instruction Decode Queue (IDQ) initiated by Decode Stream Buffer (DSB) while Microcode Sequenser (MS) is busy.", + "BriefDescription": "Deliveries to Instruction Decode Queue (IDQ) initiated by Decode Stream Buffer (DSB) while Microcode Sequencer (MS) is busy.", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "CounterMask": "1", @@ -193,7 +193,7 @@ "UMask": "0x10" }, { - "BriefDescription": "Uops initiated by Decode Stream Buffer (DSB) that are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Uops initiated by Decode Stream Buffer (DSB) that are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0x79", @@ -203,7 +203,7 @@ "UMask": "0x10" }, { - "BriefDescription": "Uops initiated by MITE and delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Uops initiated by MITE and delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0x79", @@ -224,7 +224,7 @@ "UMask": "0x30" }, { - "BriefDescription": "Uops delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Uops delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0x79", diff --git a/tools/perf/pmu-events/arch/x86/haswellx/hsx-metrics.json b/tools/perf/pmu-events/arch/x86/haswellx/hsx-metrics.json index d31d76db9d84..2cd86750986a 100644 --- a/tools/perf/pmu-events/arch/x86/haswellx/hsx-metrics.json +++ b/tools/perf/pmu-events/arch/x86/haswellx/hsx-metrics.json @@ -1,64 +1,514 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * min(CPU_CLK_UNHALTED.THREAD, IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE) / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: RS_EVENTS.EMPTY_END", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", + "MetricExpr": "ICACHE.IFDATA_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "(14 * ITLB_MISSES.STLB_HIT + ITLB_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: ITLB_MISSES.WALK_COMPLETED", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "12 * (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY) / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "2 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - (tma_frontend_bound + tma_bad_speculation + tma_retiring)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING) + RESOURCE_STALLS.SB) / (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + (cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) if #SMT_on else (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING) - CYCLE_ACTIVITY.STALLS_L1D_PENDING) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_UOPS_RETIRED.L1_HIT_PS;MEM_LOAD_UOPS_RETIRED.HIT_LFB_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "(8 * DTLB_LOAD_MISSES.STLB_HIT + DTLB_LOAD_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_UOPS_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_UOPS_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_UOPS_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "Load_Miss_Real_Latency * cpu@L1D_PEND_MISS.REQUEST_FB_FULL\\,cmask\\=1@ / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L1D_PENDING - CYCLE_ACTIVITY.STALLS_L2_PENDING) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEM_LOAD_UOPS_RETIRED.L3_HIT / (MEM_LOAD_UOPS_RETIRED.L3_HIT + 7 * MEM_LOAD_UOPS_RETIRED.L3_MISS)) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "(60 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) + 43 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD)))) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "43 * (MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "41 * (MEM_LOAD_UOPS_RETIRED.L3_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "((OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2) if #SMT_on else OFFCORE_REQUESTS_BUFFER.SQ_FULL) / CORE_CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(1 - (MEM_LOAD_UOPS_RETIRED.L3_HIT / (MEM_LOAD_UOPS_RETIRED.L3_HIT + 7 * MEM_LOAD_UOPS_RETIRED.L3_MISS))) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=6@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory", + "MetricExpr": "200 * (MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "Server;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_local_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory. Caching will improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory", + "MetricExpr": "310 * (MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM_PS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues", + "MetricExpr": "(200 * (MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD))) + 180 * (MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.L3_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD)))) / CLKS", + "MetricGroup": "Offcore;Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_cache", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM_PS;MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "RESOURCE_STALLS.SB / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_UOPS_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 9 * (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES))) + (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "(200 * OFFCORE_RESPONSE.DEMAND_RFO.LLC_MISS.REMOTE_HITM + 60 * OFFCORE_RESPONSE.DEMAND_RFO.LLC_HIT.HITM_OTHER_CORE) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "2 * MEM_UOPS_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_UOPS_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(8 * DTLB_STORE_MISSES.STLB_HIT + DTLB_STORE_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_UOPS_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "10 * ARITH.DIVIDER_UOPS / CORE_CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + (cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) if #SMT_on else (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) - RESOURCE_STALLS.SB - min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING)) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,inv\\,cmask\\=1@) / 2 if #SMT_on else (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else 0) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 if #SMT_on else (cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / 2 if #SMT_on else (cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).", + "MetricExpr": "((cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ / 2) if #SMT_on else cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_0 + UOPS_DISPATCHED_PORT.PORT_1 + UOPS_DISPATCHED_PORT.PORT_5 + UOPS_DISPATCHED_PORT.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED_PORT.PORT_0", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_1", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_6", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_2 + UOPS_DISPATCHED_PORT.PORT_3 + UOPS_DISPATCHED_PORT.PORT_7 - UOPS_DISPATCHED_PORT.PORT_4) / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_2", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_2 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_2", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_3", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_3 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data) Sample with: UOPS_DISPATCHED_PORT.PORT_4", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_4", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 7 ([HSW+]simple Store-address) Sample with: UOPS_DISPATCHED_PORT.PORT_7", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_7 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_7", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "INST_RETIRED.X87 * UPI / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * OTHER_ASSISTS.ANY_WB_ASSIST / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: OTHER_ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" }, { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -75,6 +525,12 @@ "MetricName": "UpTB" }, { + "BriefDescription": "Cycles Per Instruction (per Logical Processor)", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", + "MetricName": "CPI" + }, + { "BriefDescription": "Per-Logical Processor actual clocks when the Logical Processor is active.", "MetricExpr": "CPU_CLK_UNHALTED.THREAD", "MetricGroup": "Pipeline", @@ -82,37 +538,25 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "( UOPS_EXECUTED.CORE / 2 / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@) ) if #SMT_on else UOPS_EXECUTED.CORE / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@)", + "MetricExpr": "(UOPS_EXECUTED.CORE / 2 / ((cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2) if #SMT_on else cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@)) if #SMT_on else UOPS_EXECUTED.CORE / ((cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2) if #SMT_on else cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, @@ -153,9 +597,9 @@ "MetricName": "BpTkBranch" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -166,7 +610,7 @@ }, { "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)", - "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )", + "MetricExpr": "IDQ.DSB_UOPS / ((IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS))", "MetricGroup": "DSB;Fed;FetchBW", "MetricName": "DSB_Coverage" }, @@ -178,48 +622,42 @@ }, { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION ) / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "(ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION) / CORE_CLKS", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, { - "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", - "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Mem;MemoryTLB_SMT", - "MetricName": "Page_Walks_Utilization_SMT" - }, - { "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]", "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time", "MetricGroup": "Mem;MemoryBW", @@ -239,19 +677,19 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, @@ -269,19 +707,19 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -299,13 +737,13 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "( 64 * ( uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@ ) / 1000000000 ) / duration_time", + "MetricExpr": "(64 * (uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@) / 1000000000) / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, { "BriefDescription": "Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches", - "MetricExpr": "1000000000 * ( cbox@event\\=0x36\\,umask\\=0x3\\,filter_opc\\=0x182@ / cbox@event\\=0x35\\,umask\\=0x3\\,filter_opc\\=0x182@ ) / ( cbox_0@event\\=0x0@ / duration_time )", + "MetricExpr": "1000000000 * (cbox@event\\=0x36\\,umask\\=0x3\\,filter_opc\\=0x182@ / cbox@event\\=0x35\\,umask\\=0x3\\,filter_opc\\=0x182@) / (Socket_CLKS / duration_time)", "MetricGroup": "Mem;MemoryLat;SoC", "MetricName": "MEM_Read_Latency" }, @@ -322,12 +760,6 @@ "MetricName": "Socket_CLKS" }, { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "cbox_0@event\\=0x0@ / #num_dies / duration_time / 1000000000", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, - { "BriefDescription": "Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]", "MetricExpr": "INST_RETIRED.ANY / BR_INST_RETIRED.FAR_BRANCH:u", "MetricGroup": "Branches;OS", @@ -376,402 +808,233 @@ "MetricName": "C7_Pkg_Residency" }, { + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "Socket_CLKS / #num_dies / duration_time / 1000000000", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ" + }, + { "BriefDescription": "CPU operating frequency (in GHz)", - "MetricExpr": "( CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000", + "MetricExpr": "(( CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000) / duration_time", "MetricGroup": "", "MetricName": "cpu_operating_frequency", "ScaleUnit": "1GHz" }, { - "BriefDescription": "Cycles per instruction retired; indicating how much time each executed instruction took; in units of cycles.", - "MetricExpr": " CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY ", - "MetricGroup": "", - "MetricName": "cpi", - "ScaleUnit": "1per_instr" - }, - { "BriefDescription": "The ratio of number of completed memory load instructions to the total number completed instructions", - "MetricExpr": " MEM_UOPS_RETIRED.ALL_LOADS / INST_RETIRED.ANY ", + "MetricExpr": "MEM_UOPS_RETIRED.ALL_LOADS / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "loads_per_instr", "ScaleUnit": "1per_instr" }, { "BriefDescription": "The ratio of number of completed memory store instructions to the total number completed instructions", - "MetricExpr": " MEM_UOPS_RETIRED.ALL_STORES / INST_RETIRED.ANY ", + "MetricExpr": "MEM_UOPS_RETIRED.ALL_STORES / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "stores_per_instr", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of requests missing L1 data cache (includes data+rfo w/ prefetches) to the total number of completed instructions", - "MetricExpr": " L1D.REPLACEMENT / INST_RETIRED.ANY ", + "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l1d_mpi_includes_data_plus_rfo_with_prefetches", + "MetricName": "l1d_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of demand load requests hitting in L1 data cache to the total number of completed instructions", - "MetricExpr": " MEM_LOAD_UOPS_RETIRED.L1_HIT / INST_RETIRED.ANY ", + "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L1_HIT / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "l1d_demand_data_read_hits_per_instr", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of code read requests missing in L1 instruction cache (includes prefetches) to the total number of completed instructions", - "MetricExpr": " L2_RQSTS.ALL_CODE_RD / INST_RETIRED.ANY ", + "MetricExpr": "L2_RQSTS.ALL_CODE_RD / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "l1_i_code_read_misses_with_prefetches_per_instr", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed demand load requests hitting in L2 cache to the total number of completed instructions", - "MetricExpr": " MEM_LOAD_UOPS_RETIRED.L2_HIT / INST_RETIRED.ANY ", + "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_HIT / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "l2_demand_data_read_hits_per_instr", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of requests missing L2 cache (includes code+data+rfo w/ prefetches) to the total number of completed instructions", - "MetricExpr": " L2_LINES_IN.ALL / INST_RETIRED.ANY ", + "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches", + "MetricName": "l2_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed data read request missing L2 cache to the total number of completed instructions", - "MetricExpr": " MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY ", + "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "l2_demand_data_read_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of code read request missing L2 cache to the total number of completed instructions", - "MetricExpr": " L2_RQSTS.CODE_RD_MISS / INST_RETIRED.ANY ", + "MetricExpr": "L2_RQSTS.CODE_RD_MISS / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "l2_demand_code_mpi", "ScaleUnit": "1per_instr" }, { + "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) in nano seconds", + "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) ) ) * duration_time", + "MetricGroup": "", + "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency", + "ScaleUnit": "1ns" + }, + { + "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to local memory in nano seconds", + "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) ) ) * duration_time", + "MetricGroup": "", + "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_local_requests", + "ScaleUnit": "1ns" + }, + { + "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to remote memory in nano seconds", + "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) ) ) * duration_time", + "MetricGroup": "", + "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_remote_requests", + "ScaleUnit": "1ns" + }, + { "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the ITLB (Instruction TLB) and further levels of TLB.", - "MetricExpr": " ITLB_MISSES.WALK_COMPLETED / INST_RETIRED.ANY ", + "MetricExpr": "ITLB_MISSES.WALK_COMPLETED / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "itlb_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for 2 megabyte and 4 megabyte page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the Instruction Translation Lookaside Buffer (ITLB) and further levels of TLB.", - "MetricExpr": " ITLB_MISSES.WALK_COMPLETED_2M_4M / INST_RETIRED.ANY ", + "MetricExpr": "ITLB_MISSES.WALK_COMPLETED_2M_4M / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "itlb_large_page_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by demand data loads to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.", - "MetricExpr": " DTLB_LOAD_MISSES.WALK_COMPLETED / INST_RETIRED.ANY ", + "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "dtlb_load_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by demand data stores to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.", - "MetricExpr": " DTLB_STORE_MISSES.WALK_COMPLETED / INST_RETIRED.ANY ", + "MetricExpr": "DTLB_STORE_MISSES.WALK_COMPLETED / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "dtlb_store_mpi", "ScaleUnit": "1per_instr" }, { + "BriefDescription": "Uncore operating frequency in GHz", + "MetricExpr": "( UNC_C_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) / 1000000000) / duration_time", + "MetricGroup": "", + "MetricName": "uncore_frequency", + "ScaleUnit": "1GHz" + }, + { "BriefDescription": "Intel(R) Quick Path Interconnect (QPI) data transmit bandwidth (MB/sec)", - "MetricExpr": "( UNC_Q_TxL_FLITS_G0.DATA * 8 / 1000000) / duration_time", + "MetricExpr": "( UNC_Q_TxL_FLITS_G0.DATA * 8 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "qpi_data_transmit_bw_only_data", + "MetricName": "qpi_data_transmit_bw", "ScaleUnit": "1MB/s" }, { "BriefDescription": "DDR memory read bandwidth (MB/sec)", - "MetricExpr": "( UNC_M_CAS_COUNT.RD * 64 / 1000000) / duration_time", + "MetricExpr": "( UNC_M_CAS_COUNT.RD * 64 / 1000000) / duration_time", "MetricGroup": "", "MetricName": "memory_bandwidth_read", "ScaleUnit": "1MB/s" }, { "BriefDescription": "DDR memory write bandwidth (MB/sec)", - "MetricExpr": "( UNC_M_CAS_COUNT.WR * 64 / 1000000) / duration_time", + "MetricExpr": "( UNC_M_CAS_COUNT.WR * 64 / 1000000) / duration_time", "MetricGroup": "", "MetricName": "memory_bandwidth_write", "ScaleUnit": "1MB/s" }, { "BriefDescription": "DDR memory bandwidth (MB/sec)", - "MetricExpr": "(( UNC_M_CAS_COUNT.RD + UNC_M_CAS_COUNT.WR ) * 64 / 1000000) / duration_time", + "MetricExpr": "(( UNC_M_CAS_COUNT.RD + UNC_M_CAS_COUNT.WR ) * 64 / 1000000) / duration_time", "MetricGroup": "", "MetricName": "memory_bandwidth_total", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Bandwidth of IO reads that are initiated by end device controllers that are requesting memory from the CPU.", - "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x19e@ * 64 / 1000000) / duration_time", + "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x19e@ * 64 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_read", + "MetricName": "io_bandwidth_disk_or_network_writes", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Bandwidth of IO writes that are initiated by end device controllers that are writing memory to the CPU.", - "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x1c8\\,filter_tid\\=0x3e@ * 64 / 1000000) / duration_time", + "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x1c8\\,filter_tid\\=0x3e@ * 64 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_write", + "MetricName": "io_bandwidth_disk_or_network_reads", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Uops delivered from decoded instruction cache (decoded stream buffer or DSB) as a percent of total uops delivered to Instruction Decode Queue", - "MetricExpr": "100 * ( IDQ.DSB_UOPS / UOPS_ISSUED.ANY )", + "MetricExpr": "100 * ( IDQ.DSB_UOPS / UOPS_ISSUED.ANY )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_frodecoded_icache_dsb", + "MetricName": "percent_uops_delivered_from_decoded_icache", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from legacy decode pipeline (Micro-instruction Translation Engine or MITE) as a percent of total uops delivered to Instruction Decode Queue", - "MetricExpr": "100 * ( IDQ.MITE_UOPS / UOPS_ISSUED.ANY )", + "MetricExpr": "100 * ( IDQ.MITE_UOPS / UOPS_ISSUED.ANY )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_frolegacy_decode_pipeline_mite", + "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from microcode sequencer (MS) as a percent of total uops delivered to Instruction Decode Queue", - "MetricExpr": "100 * ( IDQ.MS_UOPS / UOPS_ISSUED.ANY )", + "MetricExpr": "100 * ( IDQ.MS_UOPS / UOPS_ISSUED.ANY )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_fromicrocode_sequencer_ms", + "MetricName": "percent_uops_delivered_from_microcode_sequencer", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from loop stream detector(LSD) as a percent of total uops delivered to Instruction Decode Queue", - "MetricExpr": "100 * ( UOPS_ISSUED.ANY - IDQ.MITE_UOPS - IDQ.MS_UOPS - IDQ.DSB_UOPS ) / UOPS_ISSUED.ANY ", + "MetricExpr": "100 * ( UOPS_ISSUED.ANY - IDQ.MITE_UOPS - IDQ.MS_UOPS - IDQ.DSB_UOPS ) / UOPS_ISSUED.ANY", "MetricGroup": "", - "MetricName": "percent_uops_delivered_froloop_streadetector_lsd", + "MetricName": "percent_uops_delivered_from_loop_stream_detector", "ScaleUnit": "1%" }, { "BriefDescription": "Ratio of number of data read requests missing last level core cache (includes demand w/ prefetches) to the total number of completed instructions", - "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x192@ ) / INST_RETIRED.ANY ", + "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x192@ ) / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "llc_data_read_mpi_demand_plus_prefetch", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of code read requests missing last level core cache (includes demand w/ prefetches) to the total number of completed instructions", - "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x181@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x191@ ) / INST_RETIRED.ANY ", + "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x181@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x191@ ) / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "llc_code_read_mpi_demand_plus_prefetch", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Memory read that miss the last level cache (LLC) addressed to local DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", - "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )", + "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_local_dram", + "MetricName": "numa_reads_addressed_to_local_dram", "ScaleUnit": "1%" }, { "BriefDescription": "Memory reads that miss the last level cache (LLC) addressed to remote DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", - "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )", - "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_remote_dram", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", - "MetricExpr": "100 * ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1, PGO", - "MetricName": "tma_frontend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.", - "MetricExpr": "100 * ( ( 4 ) * ( min( CPU_CLK_UNHALTED.THREAD , IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "Frontend, TmaL2", - "MetricName": "tma_fetch_latency_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", - "MetricExpr": "100 * ( ICACHE.IFDATA_STALL / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot, FetchLat, IcMiss", - "MetricName": "tma_icache_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.", - "MetricExpr": "100 * ( ( 14 * ITLB_MISSES.STLB_HIT + ITLB_MISSES.WALK_DURATION ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot, FetchLat, MemoryTLB", - "MetricName": "tma_itlb_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.", - "MetricExpr": "100 * ( ( 12 ) * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat", - "MetricName": "tma_branch_resteers_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", - "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "DSBmiss, FetchLat", - "MetricName": "tma_dsb_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", - "MetricExpr": "100 * ( ILD_STALL.LCP / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat", - "MetricName": "tma_lcp_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.", - "MetricExpr": "100 * ( ( 2 ) * IDQ.MS_SWITCHES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat, MicroSeq", - "MetricName": "tma_ms_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", - "MetricExpr": "100 * ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( 4 ) * ( min( CPU_CLK_UNHALTED.THREAD , IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "FetchBW, Frontend, TmaL2", - "MetricName": "tma_fetch_bandwidth_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", - "MetricExpr": "100 * ( ( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )", - "MetricGroup": "DSBmiss, FetchBW", - "MetricName": "tma_mite_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", - "MetricExpr": "100 * ( ( IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )", - "MetricGroup": "DSB, FetchBW", - "MetricName": "tma_dsb_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", - "MetricExpr": "100 * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_bad_speculation_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.", - "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "BadSpec, BrMispredicts, TmaL2", - "MetricName": "tma_branch_mispredicts_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.", - "MetricExpr": "100 * ( ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )", - "MetricGroup": "BadSpec, MachineClears, TmaL2", - "MetricName": "tma_machine_clears_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", - "MetricExpr": "100 * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_backend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", - "MetricExpr": "100 * ( ( ( ( min( CPU_CLK_UNHALTED.THREAD , CYCLE_ACTIVITY.STALLS_LDM_PENDING ) ) + RESOURCE_STALLS.SB ) / ( ( ( min( CPU_CLK_UNHALTED.THREAD , CYCLE_ACTIVITY.CYCLES_NO_EXECUTE ) ) + ( cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x1@ - ( cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x3@ if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x2@ ) ) / 2 - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * ( min( CPU_CLK_UNHALTED.THREAD , IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) if #SMT_on else ( ( min( CPU_CLK_UNHALTED.THREAD , CYCLE_ACTIVITY.CYCLES_NO_EXECUTE ) ) + cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x1@ - ( cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x3@ if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x2@ ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * ( min( CPU_CLK_UNHALTED.THREAD , IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) ) * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) )", - "MetricGroup": "Backend, TmaL2", - "MetricName": "tma_memory_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.", - "MetricExpr": "100 * ( max( ( ( min( CPU_CLK_UNHALTED.THREAD , CYCLE_ACTIVITY.STALLS_LDM_PENDING ) ) - CYCLE_ACTIVITY.STALLS_L1D_PENDING ) / ( CPU_CLK_UNHALTED.THREAD ) , 0 ) )", - "MetricGroup": "CacheMisses, MemoryBound, TmaL3mem", - "MetricName": "tma_l1_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L1D_PENDING - CYCLE_ACTIVITY.STALLS_L2_PENDING ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "CacheMisses, MemoryBound, TmaL3mem", - "MetricName": "tma_l2_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) * CYCLE_ACTIVITY.STALLS_L2_PENDING / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "CacheMisses, MemoryBound, TmaL3mem", - "MetricName": "tma_l3_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.", - "MetricExpr": "100 * ( min( ( ( 1 - ( MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) ) * CYCLE_ACTIVITY.STALLS_L2_PENDING / ( CPU_CLK_UNHALTED.THREAD ) ) , ( 1 ) ) )", - "MetricGroup": "MemoryBound, TmaL3mem", - "MetricName": "tma_drabound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.", - "MetricExpr": "100 * ( RESOURCE_STALLS.SB / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "MemoryBound, TmaL3mem", - "MetricName": "tma_store_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", - "MetricExpr": "100 * ( ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) - ( ( ( ( min( CPU_CLK_UNHALTED.THREAD , CYCLE_ACTIVITY.STALLS_LDM_PENDING ) ) + RESOURCE_STALLS.SB ) / ( ( ( min( CPU_CLK_UNHALTED.THREAD , CYCLE_ACTIVITY.CYCLES_NO_EXECUTE ) ) + ( cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x1@ - ( cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x3@ if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x2@ ) ) / 2 - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * ( min( CPU_CLK_UNHALTED.THREAD , IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) if #SMT_on else ( ( min( CPU_CLK_UNHALTED.THREAD , CYCLE_ACTIVITY.CYCLES_NO_EXECUTE ) ) + cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x1@ - ( cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x3@ if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x2@ ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * ( min( CPU_CLK_UNHALTED.THREAD , IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) ) * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) ) )", - "MetricGroup": "Backend, TmaL2, Compute", - "MetricName": "tma_core_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.", - "MetricExpr": "100 * ( 10 * ARITH.DIVIDER_UOPS / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) )", + "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ )", "MetricGroup": "", - "MetricName": "tma_divider_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", - "MetricExpr": "100 * ( ( ( ( ( min( CPU_CLK_UNHALTED.THREAD , CYCLE_ACTIVITY.CYCLES_NO_EXECUTE ) ) + ( cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x1@ - ( cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x3@ if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x2@ ) ) / 2 - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * ( min( CPU_CLK_UNHALTED.THREAD , IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) if #SMT_on else ( ( min( CPU_CLK_UNHALTED.THREAD , CYCLE_ACTIVITY.CYCLES_NO_EXECUTE ) ) + cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x1@ - ( cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x3@ if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else cpu@UOPS_EXECUTED.CORE\\,cmask\\=0x2@ ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * ( min( CPU_CLK_UNHALTED.THREAD , IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) - RESOURCE_STALLS.SB - ( min( CPU_CLK_UNHALTED.THREAD , CYCLE_ACTIVITY.STALLS_LDM_PENDING ) ) ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "PortsUtil", - "MetricName": "tma_ports_utilization_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. ", - "MetricExpr": "100 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_retiring_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved.", - "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )", - "MetricGroup": "Retire, TmaL2", - "MetricName": "tma_light_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "Retire, TmaL2", - "MetricName": "tma_heavy_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided.", - "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "MicroSeq", - "MetricName": "tma_microcode_sequencer_percent", + "MetricName": "numa_reads_addressed_to_remote_dram", "ScaleUnit": "1%" } ] diff --git a/tools/perf/pmu-events/arch/x86/haswellx/uncore-interconnect.json b/tools/perf/pmu-events/arch/x86/haswellx/uncore-interconnect.json index 3e48ff3516b0..eb0a05fbb704 100644 --- a/tools/perf/pmu-events/arch/x86/haswellx/uncore-interconnect.json +++ b/tools/perf/pmu-events/arch/x86/haswellx/uncore-interconnect.json @@ -981,36 +981,34 @@ "Unit": "QPI LL" }, { - "BriefDescription": "Number of data flits transmitted . Derived from unc_q_txl_flits_g0.data", + "BriefDescription": "Flits Transferred - Group 0; Data Tx Flits", "Counter": "0,1,2,3", - "EventName": "QPI_DATA_BANDWIDTH_TX", + "EventName": "UNC_Q_TxL_FLITS_G0.DATA", "PerPkg": "1", - "ScaleUnit": "8Bytes", "UMask": "0x2", "Unit": "QPI LL" }, { - "BriefDescription": "Number of data flits transmitted ", + "BriefDescription": "Number of data flits transmitted . Derived from unc_q_txl_flits_g0.data", "Counter": "0,1,2,3", - "EventName": "UNC_Q_TxL_FLITS_G0.DATA", + "EventName": "QPI_DATA_BANDWIDTH_TX", "PerPkg": "1", "ScaleUnit": "8Bytes", "UMask": "0x2", "Unit": "QPI LL" }, { - "BriefDescription": "Number of non data (control) flits transmitted . Derived from unc_q_txl_flits_g0.non_data", + "BriefDescription": "Flits Transferred - Group 0; Non-Data protocol Tx Flits", "Counter": "0,1,2,3", - "EventName": "QPI_CTL_BANDWIDTH_TX", + "EventName": "UNC_Q_TxL_FLITS_G0.NON_DATA", "PerPkg": "1", - "ScaleUnit": "8Bytes", "UMask": "0x4", "Unit": "QPI LL" }, { - "BriefDescription": "Number of non data (control) flits transmitted ", + "BriefDescription": "Number of non data (control) flits transmitted . Derived from unc_q_txl_flits_g0.non_data", "Counter": "0,1,2,3", - "EventName": "UNC_Q_TxL_FLITS_G0.NON_DATA", + "EventName": "QPI_CTL_BANDWIDTH_TX", "PerPkg": "1", "ScaleUnit": "8Bytes", "UMask": "0x4", diff --git a/tools/perf/pmu-events/arch/x86/haswellx/uncore-memory.json b/tools/perf/pmu-events/arch/x86/haswellx/uncore-memory.json index db3418db312e..c003daa9ed8c 100644 --- a/tools/perf/pmu-events/arch/x86/haswellx/uncore-memory.json +++ b/tools/perf/pmu-events/arch/x86/haswellx/uncore-memory.json @@ -72,20 +72,19 @@ "Unit": "iMC" }, { - "BriefDescription": "read requests to memory controller. Derived from unc_m_cas_count.rd", + "BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; All DRAM Reads (RD_CAS + Underfills)", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "LLC_MISSES.MEM_READ", + "EventName": "UNC_M_CAS_COUNT.RD", "PerPkg": "1", - "ScaleUnit": "64Bytes", "UMask": "0x3", "Unit": "iMC" }, { - "BriefDescription": "read requests to memory controller", + "BriefDescription": "read requests to memory controller. Derived from unc_m_cas_count.rd", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "UNC_M_CAS_COUNT.RD", + "EventName": "LLC_MISSES.MEM_READ", "PerPkg": "1", "ScaleUnit": "64Bytes", "UMask": "0x3", @@ -110,20 +109,19 @@ "Unit": "iMC" }, { - "BriefDescription": "write requests to memory controller. Derived from unc_m_cas_count.wr", + "BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; All DRAM WR_CAS (both Modes)", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "LLC_MISSES.MEM_WRITE", + "EventName": "UNC_M_CAS_COUNT.WR", "PerPkg": "1", - "ScaleUnit": "64Bytes", "UMask": "0xC", "Unit": "iMC" }, { - "BriefDescription": "write requests to memory controller", + "BriefDescription": "write requests to memory controller. Derived from unc_m_cas_count.wr", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "UNC_M_CAS_COUNT.WR", + "EventName": "LLC_MISSES.MEM_WRITE", "PerPkg": "1", "ScaleUnit": "64Bytes", "UMask": "0xC", diff --git a/tools/perf/pmu-events/arch/x86/icelake/cache.json b/tools/perf/pmu-events/arch/x86/icelake/cache.json index b4f28f24ee63..0f6b918484d5 100644 --- a/tools/perf/pmu-events/arch/x86/icelake/cache.json +++ b/tools/perf/pmu-events/arch/x86/icelake/cache.json @@ -18,13 +18,13 @@ "EventCode": "0x48", "EventName": "L1D_PEND_MISS.FB_FULL", "PEBScounters": "0,1,2,3", - "PublicDescription": "Counts number of cycles a demand request has waited due to L1D Fill Buffer (FB) unavailablability. Demand requests include cacheable/uncacheable demand load, store, lock or SW prefetch accesses.", + "PublicDescription": "Counts number of cycles a demand request has waited due to L1D Fill Buffer (FB) unavailability. Demand requests include cacheable/uncacheable demand load, store, lock or SW prefetch accesses.", "SampleAfterValue": "1000003", "Speculative": "1", "UMask": "0x2" }, { - "BriefDescription": "Number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailablability.", + "BriefDescription": "Number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailability.", "CollectPEBSRecord": "2", "Counter": "0,1,2,3", "CounterMask": "1", @@ -32,7 +32,7 @@ "EventCode": "0x48", "EventName": "L1D_PEND_MISS.FB_FULL_PERIODS", "PEBScounters": "0,1,2,3", - "PublicDescription": "Counts number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailablability. Demand requests include cacheable/uncacheable demand load, store, lock or SW prefetch accesses.", + "PublicDescription": "Counts number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailability. Demand requests include cacheable/uncacheable demand load, store, lock or SW prefetch accesses.", "SampleAfterValue": "1000003", "Speculative": "1", "UMask": "0x2" diff --git a/tools/perf/pmu-events/arch/x86/icelake/icl-metrics.json b/tools/perf/pmu-events/arch/x86/icelake/icl-metrics.json index f0356d66a927..3b5ef09eb8ef 100644 --- a/tools/perf/pmu-events/arch/x86/icelake/icl-metrics.json +++ b/tools/perf/pmu-events/arch/x86/icelake/icl-metrics.json @@ -1,26 +1,716 @@ [ { + "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", + "MetricExpr": "topdown\\-fe\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) - INT_MISC.UOP_DROPPING / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. Sample with: FRONTEND_RETIRED.LATENCY_GE_4_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "(5 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING) / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: FRONTEND_RETIRED.LATENCY_GE_16_PS;FRONTEND_RETIRED.LATENCY_GE_8_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses", + "MetricExpr": "ICACHE_16B.IFDATA_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses. Sample with: FRONTEND_RETIRED.L2_MISS_PS;FRONTEND_RETIRED.L1I_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "ICACHE_64B.IFTAG_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: FRONTEND_RETIRED.STLB_MISS_PS;FRONTEND_RETIRED.ITLB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "INT_MISC.CLEAR_RESTEER_CYCLES / CLKS + tma_unknown_branches", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears", + "MetricExpr": "(1 - (BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT))) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "10 * BACLEARS.ANY / CLKS", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: BACLEARS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty. Sample with: FRONTEND_RETIRED.DSB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "3 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "max(0, tma_frontend_bound - tma_fetch_latency)", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend. Sample with: FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_2_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.MITE_CYCLES_ANY - IDQ.MITE_CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck. Sample with: FRONTEND_RETIRED.ANY_DSB_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where decoder-0 was the only active decoder", + "MetricExpr": "(cpu@INST_DECODED.DECODERS\\,cmask\\=1@ - cpu@INST_DECODED.DECODERS\\,cmask\\=2@) / CORE_CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_decoder0_alone", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where (only) 4 uops were delivered by the MITE pipeline", + "MetricExpr": "(cpu@IDQ.MITE_UOPS\\,cmask\\=4@ - cpu@IDQ.MITE_UOPS\\,cmask\\=5@) / CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_mite_4wide", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.DSB_CYCLES_ANY - IDQ.DSB_CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to LSD (Loop Stream Detector) unit", + "MetricExpr": "(LSD.CYCLES_ACTIVE - LSD.CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "FetchBW;LSD;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_lsd", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to LSD (Loop Stream Detector) unit. LSD typically does well sustaining Uop supply. However; in some rare cases; optimal uop-delivery could not be reached for small loops whose size (in terms of number of uops) does not suit well the LSD structure.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", + "MetricExpr": "max(1 - (tma_frontend_bound + tma_backend_bound + tma_retiring), 0)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "max(0, tma_bad_speculation - tma_branch_mispredicts)", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", + "MetricExpr": "topdown\\-be\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + (5 * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=1\\,edge@) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. Sample with: TOPDOWN.BACKEND_BOUND_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_RETIRED.L1_HIT_PS;MEM_LOAD_RETIRED.FB_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "min(7 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE, max(CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS, 0)) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_INST_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_load - tma_load_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk", + "MetricExpr": "DTLB_LOAD_MISSES.WALK_ACTIVE / CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(16 * max(0, MEM_INST_RETIRED.LOCK_LOADS - L2_RQSTS.ALL_RFO) + (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES) * (10 * L2_RQSTS.RFO_HIT + min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO))) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_INST_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_INST_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "L1D_PEND_MISS.FB_FULL / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) / ((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + L1D_PEND_MISS.FB_FULL_PERIODS)) * ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "((29 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM + (23.5 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "(23.5 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "(9 * Average_Frequency) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "L1D_PEND_MISS.L2_STALL / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L3_MISS / CLKS + ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS) - tma_l2_bound)", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "EXE_ACTIVITY.BOUND_ON_STORES / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_INST_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 10 * (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES))) + (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "(32.5 * Average_Frequency) * OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "MEM_INST_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_INST_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores", + "MetricExpr": "9 * OCR.STREAMING_WR.ANY_RESPONSE / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_streaming_stores", + "PublicDescription": "This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should Streaming stores be a bottleneck. Sample with: OCR.STREAMING_WR.ANY_RESPONSE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(7 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE) / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_INST_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_store - tma_store_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk", + "MetricExpr": "DTLB_STORE_MISSES.WALK_ACTIVE / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "max(0, tma_backend_bound - tma_memory_bound)", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.DIVIDER_ACTIVE / CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_ACTIVE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "(cpu@EXE_ACTIVITY.3_PORTS_UTIL\\,umask\\=0x80@ + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL)) / CLKS if (ARITH.DIVIDER_ACTIVE < (CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY)) else (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "cpu@EXE_ACTIVITY.3_PORTS_UTIL\\,umask\\=0x80@ / CLKS + tma_serializing_operation * (CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY) / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations", + "MetricExpr": "RESOURCE_STALLS.SCOREBOARD / CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_serializing_operation", + "PublicDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out-of-order execution which may limit performance. Sample with: RESOURCE_STALLS.SCOREBOARD", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions", + "MetricExpr": "140 * MISC_RETIRED.PAUSE_INST / CLKS", + "MetricGroup": "TopdownL6;tma_serializing_operation_group", + "MetricName": "tma_slow_pause", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions. Sample with: MISC_RETIRED.PAUSE_INST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued", + "MetricExpr": "CLKS * UOPS_ISSUED.VECTOR_WIDTH_MISMATCH / UOPS_ISSUED.ANY", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_mixing_vectors", + "PublicDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "EXE_ACTIVITY.1_PORTS_UTIL / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful. Sample with: EXE_ACTIVITY.1_PORTS_UTIL", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "EXE_ACTIVITY.2_PORTS_UTIL / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop. Sample with: EXE_ACTIVITY.2_PORTS_UTIL", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "UOPS_EXECUTED.CYCLES_GE_3 / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Sample with: UOPS_EXECUTED.CYCLES_GE_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED.PORT_0 + UOPS_DISPATCHED.PORT_1 + UOPS_DISPATCHED.PORT_5 + UOPS_DISPATCHED.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED.PORT_0", + "MetricExpr": "UOPS_DISPATCHED.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED.PORT_1", + "MetricExpr": "UOPS_DISPATCHED.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED.PORT_6", + "MetricExpr": "UOPS_DISPATCHED.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "UOPS_DISPATCHED.PORT_2_3 / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations Sample with: UOPS_DISPATCHED.PORT_7_8", + "MetricExpr": "(UOPS_DISPATCHED.PORT_4_9 + UOPS_DISPATCHED.PORT_7_8) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", + "MetricExpr": "topdown\\-retiring / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "max(0, tma_retiring - tma_heavy_operations)", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "tma_retiring * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_512b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", + "MetricExpr": "tma_light_operations * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_memory_operations", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions.", + "MetricExpr": "tma_light_operations * BR_INST_RETIRED.ALL_BRANCHES / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_branch_instructions", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions", + "MetricExpr": "tma_light_operations * INST_RETIRED.NOP / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_nop_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body. Sample with: INST_RETIRED.NOP", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", + "MetricExpr": "max(0, tma_light_operations - (tma_fp_arith + tma_memory_operations + tma_branch_instructions + tma_nop_instructions))", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_other_light_ops", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer + tma_retiring * (UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=1@) / IDQ.MITE_UOPS", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops", + "MetricExpr": "tma_heavy_operations - tma_microcode_sequencer", + "MetricGroup": "TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_few_uops_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "((tma_retiring * SLOTS) / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * ASSISTS.ANY / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", + "MetricExpr": "100 * (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches))", + "MetricGroup": "Bad;BadSpec;BrMispredicts", + "MetricName": "Mispredictions" + }, + { + "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_mem_bandwidth / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_sq_full / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full))) + (tma_l1_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_fb_full / (tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) ", + "MetricGroup": "Mem;MemoryBW;Offcore", + "MetricName": "Memory_Bandwidth" + }, + { + "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_mem_latency / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_l3_hit_latency / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full)) + (tma_l2_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)))", + "MetricGroup": "Mem;MemoryLat;Offcore", + "MetricName": "Memory_Latency" + }, + { + "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", + "MetricExpr": "100 * tma_memory_bound * ((tma_l1_bound / max(tma_memory_bound, tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_dtlb_load / max(tma_l1_bound, tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) + (tma_store_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_dtlb_store / (tma_dtlb_store + tma_false_sharing + tma_split_stores + tma_store_latency + tma_streaming_stores))) ", + "MetricGroup": "Mem;MemoryTLB;Offcore", + "MetricName": "Memory_Data_TLBs" + }, + { "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.COND + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL) ) / TOPDOWN.SLOTS)", + "MetricExpr": "100 * ((BR_INST_RETIRED.COND + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL)) / SLOTS)", "MetricGroup": "Ret", "MetricName": "Branching_Overhead" }, { "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", - "MetricExpr": "100 * (( 5 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING ) / TOPDOWN.SLOTS) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (ICACHE_16B.IFDATA_STALL / CPU_CLK_UNHALTED.THREAD) + (10 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(( 5 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING ) / TOPDOWN.SLOTS)", + "MetricExpr": "100 * tma_fetch_latency * (tma_itlb_misses + tma_icache_misses + tma_unknown_branches) / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)", "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB", "MetricName": "Big_Code" }, { + "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", + "MetricExpr": "100 * (tma_frontend_bound - tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) - Big_Code", + "MetricGroup": "Fed;FetchBW;Frontend", + "MetricName": "Instruction_Fetch_BW" + }, + { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, { + "BriefDescription": "Uops Per Instruction", + "MetricExpr": "(tma_retiring * SLOTS) / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;Ret;Retire", + "MetricName": "UPI" + }, + { + "BriefDescription": "Instruction per taken branch", + "MetricExpr": "(tma_retiring * SLOTS) / BR_INST_RETIRED.NEAR_TAKEN", + "MetricGroup": "Branches;Fed;FetchBW", + "MetricName": "UpTB" + }, + { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI" }, { @@ -32,13 +722,13 @@ { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", "MetricExpr": "TOPDOWN.SLOTS", - "MetricGroup": "TmaL1", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { "BriefDescription": "Fraction of Physical Core issue-slots utilized by this Logical Processor", - "MetricExpr": "TOPDOWN.SLOTS / ( TOPDOWN.SLOTS / 2 ) if #SMT_on else 1", - "MetricGroup": "SMT;TmaL1", + "MetricExpr": "SLOTS / (TOPDOWN.SLOTS / 2) if #SMT_on else 1", + "MetricGroup": "SMT;tma_L1_group", "MetricName": "Slots_Utilization" }, { @@ -50,30 +740,36 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.DISTRIBUTED", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / CPU_CLK_UNHALTED.DISTRIBUTED", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.DISTRIBUTED )", + "MetricExpr": "((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)." }, { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2 ) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { + "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", + "MetricExpr": "(1 - tma_core_bound / tma_ports_utilization if tma_core_bound < tma_ports_utilization else 1) if SMT_2T_Utilization > 0.5 else 0", + "MetricGroup": "Cor;SMT", + "MetricName": "Core_Bound_Likely" + }, + { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", "MetricExpr": "CPU_CLK_UNHALTED.DISTRIBUTED", "MetricGroup": "SMT", @@ -117,13 +813,13 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." @@ -144,21 +840,21 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX512", "PublicDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." @@ -170,12 +866,18 @@ "MetricName": "IpSWPF" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { + "BriefDescription": "Average number of Uops retired in cycles where at least one uop has retired.", + "MetricExpr": "(tma_retiring * SLOTS) / cpu@UOPS_RETIRED.SLOTS\\,cmask\\=1@", + "MetricGroup": "Pipeline;Ret", + "MetricName": "Retire" + }, + { "BriefDescription": "", "MetricExpr": "UOPS_EXECUTED.THREAD / cpu@UOPS_EXECUTED.THREAD\\,cmask\\=1@", "MetricGroup": "Cor;Pipeline;PortsUtil;SMT", @@ -206,6 +908,12 @@ "MetricName": "DSB_Switch_Cost" }, { + "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", + "MetricExpr": "100 * (tma_fetch_latency * tma_dsb_switches / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches) + tma_fetch_bandwidth * tma_mite / (tma_dsb + tma_lsd + tma_mite))", + "MetricGroup": "DSBmiss;Fed", + "MetricName": "DSB_Misses" + }, + { "BriefDescription": "Number of Instructions per non-speculative DSB miss (lower number means higher occurrence rate)", "MetricExpr": "INST_RETIRED.ANY / FRONTEND_RETIRED.ANY_DSB_MISS", "MetricGroup": "DSBmiss;Fed", @@ -218,6 +926,12 @@ "MetricName": "IpMispredict" }, { + "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricGroup": "Bad;BrMispredicts", + "MetricName": "Branch_Misprediction_Cost" + }, + { "BriefDescription": "Fraction of branches that are non-taken conditionals", "MetricExpr": "BR_INST_RETIRED.COND_NTAKEN / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", @@ -231,7 +945,7 @@ }, { "BriefDescription": "Fraction of branches that are CALL or RET", - "MetricExpr": "( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "CallRet" }, @@ -243,74 +957,74 @@ }, { "BriefDescription": "Fraction of branches of other types (not individually covered by other metrics in Info.Branches group)", - "MetricExpr": "1 - ( (BR_INST_RETIRED.COND_NTAKEN / BR_INST_RETIRED.ALL_BRANCHES) + (BR_INST_RETIRED.COND_TAKEN / BR_INST_RETIRED.ALL_BRANCHES) + (( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES) + ((BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES) )", + "MetricExpr": "1 - (Cond_NT + Cond_TK + CallRet + Jump)", "MetricGroup": "Bad;Branches", "MetricName": "Other_Branches" }, { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L1 cache true misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.ALL_DEMAND_DATA_RD / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI_Load" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( ( OFFCORE_REQUESTS.ALL_DATA_RD - OFFCORE_REQUESTS.DEMAND_DATA_RD ) + L2_RQSTS.ALL_DEMAND_MISS + L2_RQSTS.SWPF_MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricExpr": "1000 * ((OFFCORE_REQUESTS.ALL_DATA_RD - OFFCORE_REQUESTS.DEMAND_DATA_RD) + L2_RQSTS.ALL_DEMAND_MISS + L2_RQSTS.SWPF_MISS) / Instructions", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries)", "MetricExpr": "1000 * MEM_LOAD_RETIRED.FB_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "FB_HPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING ) / ( 2 * CPU_CLK_UNHALTED.DISTRIBUTED )", + "MetricExpr": "(ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING) / (2 * CORE_CLKS)", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, @@ -340,25 +1054,25 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data access bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * OFFCORE_REQUESTS.ALL_REQUESTS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Access_BW", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "L3_Cache_Access_BW_1T" }, @@ -370,40 +1084,40 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0", - "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / CPU_CLK_UNHALTED.DISTRIBUTED", + "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License0_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. This includes non-AVX codes, SSE, AVX 128-bit, and low-current AVX 256-bit codes." }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1", - "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / CPU_CLK_UNHALTED.DISTRIBUTED", + "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License1_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1. This includes high current AVX 256-bit instructions as well as low current AVX 512-bit instructions." }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX)", - "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / CPU_CLK_UNHALTED.DISTRIBUTED", + "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License2_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). This includes high current AVX 512-bit instructions." @@ -428,7 +1142,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * ( arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@ ) / 1000000 / duration_time / 1000", + "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1000000 / duration_time / 1000", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, diff --git a/tools/perf/pmu-events/arch/x86/icelake/pipeline.json b/tools/perf/pmu-events/arch/x86/icelake/pipeline.json index a017a4727050..c74a7369cff3 100644 --- a/tools/perf/pmu-events/arch/x86/icelake/pipeline.json +++ b/tools/perf/pmu-events/arch/x86/icelake/pipeline.json @@ -167,7 +167,7 @@ "UMask": "0x10" }, { - "BriefDescription": "number of branch instructions retired that were mispredicted and taken. Non PEBS", + "BriefDescription": "number of branch instructions retired that were mispredicted and taken.", "CollectPEBSRecord": "2", "Counter": "0,1,2,3,4,5,6,7", "EventCode": "0xc5", diff --git a/tools/perf/pmu-events/arch/x86/icelakex/cache.json b/tools/perf/pmu-events/arch/x86/icelakex/cache.json index 775190bdd063..e4035b3e55ca 100644 --- a/tools/perf/pmu-events/arch/x86/icelakex/cache.json +++ b/tools/perf/pmu-events/arch/x86/icelakex/cache.json @@ -18,13 +18,13 @@ "EventCode": "0x48", "EventName": "L1D_PEND_MISS.FB_FULL", "PEBScounters": "0,1,2,3", - "PublicDescription": "Counts number of cycles a demand request has waited due to L1D Fill Buffer (FB) unavailablability. Demand requests include cacheable/uncacheable demand load, store, lock or SW prefetch accesses.", + "PublicDescription": "Counts number of cycles a demand request has waited due to L1D Fill Buffer (FB) unavailability. Demand requests include cacheable/uncacheable demand load, store, lock or SW prefetch accesses.", "SampleAfterValue": "1000003", "Speculative": "1", "UMask": "0x2" }, { - "BriefDescription": "Number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailablability.", + "BriefDescription": "Number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailability.", "CollectPEBSRecord": "2", "Counter": "0,1,2,3", "CounterMask": "1", @@ -32,7 +32,7 @@ "EventCode": "0x48", "EventName": "L1D_PEND_MISS.FB_FULL_PERIODS", "PEBScounters": "0,1,2,3", - "PublicDescription": "Counts number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailablability. Demand requests include cacheable/uncacheable demand load, store, lock or SW prefetch accesses.", + "PublicDescription": "Counts number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailability. Demand requests include cacheable/uncacheable demand load, store, lock or SW prefetch accesses.", "SampleAfterValue": "1000003", "Speculative": "1", "UMask": "0x2" diff --git a/tools/perf/pmu-events/arch/x86/icelakex/icx-metrics.json b/tools/perf/pmu-events/arch/x86/icelakex/icx-metrics.json index e905458b34b8..b52afc34a169 100644 --- a/tools/perf/pmu-events/arch/x86/icelakex/icx-metrics.json +++ b/tools/perf/pmu-events/arch/x86/icelakex/icx-metrics.json @@ -1,23 +1,743 @@ [ { + "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", + "MetricExpr": "topdown\\-fe\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) - INT_MISC.UOP_DROPPING / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. Sample with: FRONTEND_RETIRED.LATENCY_GE_4_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "(5 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING) / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: FRONTEND_RETIRED.LATENCY_GE_16_PS;FRONTEND_RETIRED.LATENCY_GE_8_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses", + "MetricExpr": "ICACHE_16B.IFDATA_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses. Sample with: FRONTEND_RETIRED.L2_MISS_PS;FRONTEND_RETIRED.L1I_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "ICACHE_64B.IFTAG_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: FRONTEND_RETIRED.STLB_MISS_PS;FRONTEND_RETIRED.ITLB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "INT_MISC.CLEAR_RESTEER_CYCLES / CLKS + tma_unknown_branches", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears", + "MetricExpr": "(1 - (BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT))) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "10 * BACLEARS.ANY / CLKS", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: BACLEARS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty. Sample with: FRONTEND_RETIRED.DSB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "3 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "max(0, tma_frontend_bound - tma_fetch_latency)", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend. Sample with: FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_2_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.MITE_CYCLES_ANY - IDQ.MITE_CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck. Sample with: FRONTEND_RETIRED.ANY_DSB_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where decoder-0 was the only active decoder", + "MetricExpr": "(cpu@INST_DECODED.DECODERS\\,cmask\\=1@ - cpu@INST_DECODED.DECODERS\\,cmask\\=2@) / CORE_CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_decoder0_alone", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where (only) 4 uops were delivered by the MITE pipeline", + "MetricExpr": "(cpu@IDQ.MITE_UOPS\\,cmask\\=4@ - cpu@IDQ.MITE_UOPS\\,cmask\\=5@) / CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_mite_4wide", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.DSB_CYCLES_ANY - IDQ.DSB_CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", + "MetricExpr": "max(1 - (tma_frontend_bound + tma_backend_bound + tma_retiring), 0)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "max(0, tma_bad_speculation - tma_branch_mispredicts)", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", + "MetricExpr": "topdown\\-be\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + (5 * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=1\\,edge@) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. Sample with: TOPDOWN.BACKEND_BOUND_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_RETIRED.L1_HIT_PS;MEM_LOAD_RETIRED.FB_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "min(7 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE, max(CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS, 0)) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_INST_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_load - tma_load_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk", + "MetricExpr": "DTLB_LOAD_MISSES.WALK_ACTIVE / CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(16 * max(0, MEM_INST_RETIRED.LOCK_LOADS - L2_RQSTS.ALL_RFO) + (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES) * (10 * L2_RQSTS.RFO_HIT + min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO))) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_INST_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_INST_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "L1D_PEND_MISS.FB_FULL / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) / ((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + L1D_PEND_MISS.FB_FULL_PERIODS)) * ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "((44 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM * (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM / (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM + OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD))) + (43.5 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "(43.5 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM * (1 - (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM / (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM + OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD)))) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "(19 * Average_Frequency) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "L1D_PEND_MISS.L2_STALL / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_L3_MISS / CLKS + ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS) - tma_l2_bound) - tma_pmm_bound)", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory", + "MetricExpr": "(43.5 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Server;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_local_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory. Caching will improve the latency and increase performance. Sample with: MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory", + "MetricExpr": "(108 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues", + "MetricExpr": "((97 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM + (97 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_cache", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM_PS;MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates (based on idle latencies) how often the CPU was stalled on accesses to external 3D-Xpoint (Crystal Ridge, a.k.a", + "MetricExpr": "(((1 - ((19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + 10 * ((MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))))) / ((19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + 10 * ((MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))))) + (25 * (MEM_LOAD_RETIRED.LOCAL_PMM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + 33 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))))))) * (CYCLE_ACTIVITY.STALLS_L3_MISS / CLKS + ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS) - tma_l2_bound)) if (1000000 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM) > MEM_LOAD_RETIRED.L1_MISS) else 0)", + "MetricGroup": "MemoryBound;Server;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_pmm_bound", + "PublicDescription": "This metric roughly estimates (based on idle latencies) how often the CPU was stalled on accesses to external 3D-Xpoint (Crystal Ridge, a.k.a. IXP) memory by loads, PMM stands for Persistent Memory Module. ", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "EXE_ACTIVITY.BOUND_ON_STORES / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_INST_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 10 * (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES))) + (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "(48 * Average_Frequency) * OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "MEM_INST_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_INST_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores", + "MetricExpr": "9 * OCR.STREAMING_WR.ANY_RESPONSE / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_streaming_stores", + "PublicDescription": "This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should Streaming stores be a bottleneck. Sample with: OCR.STREAMING_WR.ANY_RESPONSE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(7 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE) / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_INST_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_store - tma_store_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk", + "MetricExpr": "DTLB_STORE_MISSES.WALK_ACTIVE / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "max(0, tma_backend_bound - tma_memory_bound)", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.DIVIDER_ACTIVE / CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_ACTIVE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "(cpu@EXE_ACTIVITY.3_PORTS_UTIL\\,umask\\=0x80@ + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL)) / CLKS if (ARITH.DIVIDER_ACTIVE < (CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY)) else (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "cpu@EXE_ACTIVITY.3_PORTS_UTIL\\,umask\\=0x80@ / CLKS + tma_serializing_operation * (CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY) / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations", + "MetricExpr": "RESOURCE_STALLS.SCOREBOARD / CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_serializing_operation", + "PublicDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out-of-order execution which may limit performance. Sample with: RESOURCE_STALLS.SCOREBOARD", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions", + "MetricExpr": "37 * MISC_RETIRED.PAUSE_INST / CLKS", + "MetricGroup": "TopdownL6;tma_serializing_operation_group", + "MetricName": "tma_slow_pause", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions. Sample with: MISC_RETIRED.PAUSE_INST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued", + "MetricExpr": "CLKS * UOPS_ISSUED.VECTOR_WIDTH_MISMATCH / UOPS_ISSUED.ANY", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_mixing_vectors", + "PublicDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "EXE_ACTIVITY.1_PORTS_UTIL / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful. Sample with: EXE_ACTIVITY.1_PORTS_UTIL", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "EXE_ACTIVITY.2_PORTS_UTIL / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop. Sample with: EXE_ACTIVITY.2_PORTS_UTIL", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "UOPS_EXECUTED.CYCLES_GE_3 / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Sample with: UOPS_EXECUTED.CYCLES_GE_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED.PORT_0 + UOPS_DISPATCHED.PORT_1 + UOPS_DISPATCHED.PORT_5 + UOPS_DISPATCHED.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED.PORT_0", + "MetricExpr": "UOPS_DISPATCHED.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED.PORT_1", + "MetricExpr": "UOPS_DISPATCHED.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED.PORT_6", + "MetricExpr": "UOPS_DISPATCHED.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "UOPS_DISPATCHED.PORT_2_3 / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations Sample with: UOPS_DISPATCHED.PORT_7_8", + "MetricExpr": "(UOPS_DISPATCHED.PORT_4_9 + UOPS_DISPATCHED.PORT_7_8) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", + "MetricExpr": "topdown\\-retiring / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "max(0, tma_retiring - tma_heavy_operations)", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "tma_retiring * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_512b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", + "MetricExpr": "tma_light_operations * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_memory_operations", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions.", + "MetricExpr": "tma_light_operations * BR_INST_RETIRED.ALL_BRANCHES / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_branch_instructions", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions", + "MetricExpr": "tma_light_operations * INST_RETIRED.NOP / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_nop_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body. Sample with: INST_RETIRED.NOP", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", + "MetricExpr": "max(0, tma_light_operations - (tma_fp_arith + tma_memory_operations + tma_branch_instructions + tma_nop_instructions))", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_other_light_ops", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer + tma_retiring * (UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=1@) / IDQ.MITE_UOPS", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops", + "MetricExpr": "tma_heavy_operations - tma_microcode_sequencer", + "MetricGroup": "TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_few_uops_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "((tma_retiring * SLOTS) / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * ASSISTS.ANY / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", + "MetricExpr": "100 * (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches))", + "MetricGroup": "Bad;BadSpec;BrMispredicts", + "MetricName": "Mispredictions" + }, + { + "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_mem_bandwidth / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_sq_full / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full))) + (tma_l1_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_fb_full / (tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) ", + "MetricGroup": "Mem;MemoryBW;Offcore", + "MetricName": "Memory_Bandwidth" + }, + { + "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_mem_latency / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_l3_hit_latency / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full)) + (tma_l2_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)))", + "MetricGroup": "Mem;MemoryLat;Offcore", + "MetricName": "Memory_Latency" + }, + { + "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", + "MetricExpr": "100 * tma_memory_bound * ((tma_l1_bound / max(tma_memory_bound, tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_dtlb_load / max(tma_l1_bound, tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) + (tma_store_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_dtlb_store / (tma_dtlb_store + tma_false_sharing + tma_split_stores + tma_store_latency + tma_streaming_stores))) ", + "MetricGroup": "Mem;MemoryTLB;Offcore", + "MetricName": "Memory_Data_TLBs" + }, + { "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.COND + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL) ) / TOPDOWN.SLOTS)", + "MetricExpr": "100 * ((BR_INST_RETIRED.COND + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL)) / SLOTS)", "MetricGroup": "Ret", "MetricName": "Branching_Overhead" }, { "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", - "MetricExpr": "100 * (( 5 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING ) / TOPDOWN.SLOTS) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (ICACHE_16B.IFDATA_STALL / CPU_CLK_UNHALTED.THREAD) + (10 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(( 5 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING ) / TOPDOWN.SLOTS)", + "MetricExpr": "100 * tma_fetch_latency * (tma_itlb_misses + tma_icache_misses + tma_unknown_branches) / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)", "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB", "MetricName": "Big_Code" }, { + "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", + "MetricExpr": "100 * (tma_frontend_bound - tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) - Big_Code", + "MetricGroup": "Fed;FetchBW;Frontend", + "MetricName": "Instruction_Fetch_BW" + }, + { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, { + "BriefDescription": "Uops Per Instruction", + "MetricExpr": "(tma_retiring * SLOTS) / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;Ret;Retire", + "MetricName": "UPI" + }, + { + "BriefDescription": "Instruction per taken branch", + "MetricExpr": "(tma_retiring * SLOTS) / BR_INST_RETIRED.NEAR_TAKEN", + "MetricGroup": "Branches;Fed;FetchBW", + "MetricName": "UpTB" + }, + { + "BriefDescription": "Cycles Per Instruction (per Logical Processor)", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", + "MetricName": "CPI" + }, + { "BriefDescription": "Per-Logical Processor actual clocks when the Logical Processor is active.", "MetricExpr": "CPU_CLK_UNHALTED.THREAD", "MetricGroup": "Pipeline", @@ -26,13 +746,13 @@ { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", "MetricExpr": "TOPDOWN.SLOTS", - "MetricGroup": "TmaL1", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { "BriefDescription": "Fraction of Physical Core issue-slots utilized by this Logical Processor", - "MetricExpr": "TOPDOWN.SLOTS / ( TOPDOWN.SLOTS / 2 ) if #SMT_on else 1", - "MetricGroup": "SMT;TmaL1", + "MetricExpr": "SLOTS / (TOPDOWN.SLOTS / 2) if #SMT_on else 1", + "MetricGroup": "SMT;tma_L1_group", "MetricName": "Slots_Utilization" }, { @@ -44,30 +764,36 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.DISTRIBUTED", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / CPU_CLK_UNHALTED.DISTRIBUTED", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.DISTRIBUTED )", + "MetricExpr": "((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)." }, { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2 ) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { + "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", + "MetricExpr": "(1 - tma_core_bound / tma_ports_utilization if tma_core_bound < tma_ports_utilization else 1) if SMT_2T_Utilization > 0.5 else 0", + "MetricGroup": "Cor;SMT", + "MetricName": "Core_Bound_Likely" + }, + { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", "MetricExpr": "CPU_CLK_UNHALTED.DISTRIBUTED", "MetricGroup": "SMT", @@ -111,13 +837,13 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." @@ -138,21 +864,21 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX512", "PublicDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." @@ -164,12 +890,18 @@ "MetricName": "IpSWPF" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { + "BriefDescription": "Average number of Uops retired in cycles where at least one uop has retired.", + "MetricExpr": "(tma_retiring * SLOTS) / cpu@UOPS_RETIRED.SLOTS\\,cmask\\=1@", + "MetricGroup": "Pipeline;Ret", + "MetricName": "Retire" + }, + { "BriefDescription": "", "MetricExpr": "UOPS_EXECUTED.THREAD / cpu@UOPS_EXECUTED.THREAD\\,cmask\\=1@", "MetricGroup": "Cor;Pipeline;PortsUtil;SMT", @@ -194,6 +926,12 @@ "MetricName": "DSB_Switch_Cost" }, { + "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", + "MetricExpr": "100 * (tma_fetch_latency * tma_dsb_switches / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches) + tma_fetch_bandwidth * tma_mite / (tma_dsb + tma_mite))", + "MetricGroup": "DSBmiss;Fed", + "MetricName": "DSB_Misses" + }, + { "BriefDescription": "Number of Instructions per non-speculative DSB miss (lower number means higher occurrence rate)", "MetricExpr": "INST_RETIRED.ANY / FRONTEND_RETIRED.ANY_DSB_MISS", "MetricGroup": "DSBmiss;Fed", @@ -206,6 +944,12 @@ "MetricName": "IpMispredict" }, { + "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricGroup": "Bad;BrMispredicts", + "MetricName": "Branch_Misprediction_Cost" + }, + { "BriefDescription": "Fraction of branches that are non-taken conditionals", "MetricExpr": "BR_INST_RETIRED.COND_NTAKEN / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", @@ -219,7 +963,7 @@ }, { "BriefDescription": "Fraction of branches that are CALL or RET", - "MetricExpr": "( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "CallRet" }, @@ -231,74 +975,74 @@ }, { "BriefDescription": "Fraction of branches of other types (not individually covered by other metrics in Info.Branches group)", - "MetricExpr": "1 - ( (BR_INST_RETIRED.COND_NTAKEN / BR_INST_RETIRED.ALL_BRANCHES) + (BR_INST_RETIRED.COND_TAKEN / BR_INST_RETIRED.ALL_BRANCHES) + (( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES) + ((BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES) )", + "MetricExpr": "1 - (Cond_NT + Cond_TK + CallRet + Jump)", "MetricGroup": "Bad;Branches", "MetricName": "Other_Branches" }, { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L1 cache true misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.ALL_DEMAND_DATA_RD / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI_Load" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( ( OFFCORE_REQUESTS.ALL_DATA_RD - OFFCORE_REQUESTS.DEMAND_DATA_RD ) + L2_RQSTS.ALL_DEMAND_MISS + L2_RQSTS.SWPF_MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricExpr": "1000 * ((OFFCORE_REQUESTS.ALL_DATA_RD - OFFCORE_REQUESTS.DEMAND_DATA_RD) + L2_RQSTS.ALL_DEMAND_MISS + L2_RQSTS.SWPF_MISS) / Instructions", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries)", "MetricExpr": "1000 * MEM_LOAD_RETIRED.FB_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "FB_HPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING ) / ( 2 * CPU_CLK_UNHALTED.DISTRIBUTED )", + "MetricExpr": "(ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING) / (2 * CORE_CLKS)", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, @@ -328,37 +1072,37 @@ }, { "BriefDescription": "Rate of silent evictions from the L2 cache per Kilo instruction where the evicted lines are dropped (no writeback to L3 or memory)", - "MetricExpr": "1000 * L2_LINES_OUT.SILENT / INST_RETIRED.ANY", + "MetricExpr": "1000 * L2_LINES_OUT.SILENT / Instructions", "MetricGroup": "L2Evicts;Mem;Server", "MetricName": "L2_Evictions_Silent_PKI" }, { "BriefDescription": "Rate of non silent evictions from the L2 cache per Kilo instruction", - "MetricExpr": "1000 * L2_LINES_OUT.NON_SILENT / INST_RETIRED.ANY", + "MetricExpr": "1000 * L2_LINES_OUT.NON_SILENT / Instructions", "MetricGroup": "L2Evicts;Mem;Server", "MetricName": "L2_Evictions_NonSilent_PKI" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data access bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * OFFCORE_REQUESTS.ALL_REQUESTS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Access_BW", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "L3_Cache_Access_BW_1T" }, @@ -370,40 +1114,40 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0", - "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / CPU_CLK_UNHALTED.DISTRIBUTED", + "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License0_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. This includes non-AVX codes, SSE, AVX 128-bit, and low-current AVX 256-bit codes." }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1", - "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / CPU_CLK_UNHALTED.DISTRIBUTED", + "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License1_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1. This includes high current AVX 256-bit instructions as well as low current AVX 512-bit instructions." }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX)", - "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / CPU_CLK_UNHALTED.DISTRIBUTED", + "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License2_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). This includes high current AVX 512-bit instructions." @@ -428,13 +1172,13 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "( 64 * ( uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@ ) / 1000000000 ) / duration_time", + "MetricExpr": "(64 * (uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@) / 1000000000) / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, { "BriefDescription": "Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches", - "MetricExpr": "1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD / UNC_CHA_TOR_INSERTS.IA_MISS_DRD ) / ( cha_0@event\\=0x0@ / duration_time )", + "MetricExpr": "1000000000 * (UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD / UNC_CHA_TOR_INSERTS.IA_MISS_DRD) / (Socket_CLKS / duration_time)", "MetricGroup": "Mem;MemoryLat;SoC", "MetricName": "MEM_Read_Latency" }, @@ -446,38 +1190,38 @@ }, { "BriefDescription": "Average latency of data read request to external 3D X-Point memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches", - "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PMM ) / cha_0@event\\=0x0@ )", - "MetricGroup": "Mem;MemoryLat;SoC;Server", + "MetricExpr": "(1000000000 * (UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PMM) / cha_0@event\\=0x0@)", + "MetricGroup": "Mem;MemoryLat;Server;SoC", "MetricName": "MEM_PMM_Read_Latency" }, { "BriefDescription": "Average latency of data read request to external DRAM memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches", - "MetricExpr": " 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_DDR ) / cha_0@event\\=0x0@", - "MetricGroup": "Mem;MemoryLat;SoC;Server", + "MetricExpr": " 1000000000 * (UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_DDR) / cha_0@event\\=0x0@", + "MetricGroup": "Mem;MemoryLat;Server;SoC", "MetricName": "MEM_DRAM_Read_Latency" }, { "BriefDescription": "Average 3DXP Memory Bandwidth Use for reads [GB / sec]", - "MetricExpr": "( ( 64 * imc@event\\=0xe3@ / 1000000000 ) / duration_time )", - "MetricGroup": "Mem;MemoryBW;SoC;Server", + "MetricExpr": "((64 * imc@event\\=0xe3@ / 1000000000) / duration_time)", + "MetricGroup": "Mem;MemoryBW;Server;SoC", "MetricName": "PMM_Read_BW" }, { "BriefDescription": "Average 3DXP Memory Bandwidth Use for Writes [GB / sec]", - "MetricExpr": "( ( 64 * imc@event\\=0xe7@ / 1000000000 ) / duration_time )", - "MetricGroup": "Mem;MemoryBW;SoC;Server", + "MetricExpr": "((64 * imc@event\\=0xe7@ / 1000000000) / duration_time)", + "MetricGroup": "Mem;MemoryBW;Server;SoC", "MetricName": "PMM_Write_BW" }, { "BriefDescription": "Average IO (network or disk) Bandwidth Use for Writes [GB / sec]", "MetricExpr": "UNC_CHA_TOR_INSERTS.IO_PCIRDCUR * 64 / 1000000000 / duration_time", - "MetricGroup": "IoBW;Mem;SoC;Server", + "MetricGroup": "IoBW;Mem;Server;SoC", "MetricName": "IO_Write_BW" }, { "BriefDescription": "Average IO (network or disk) Bandwidth Use for Reads [GB / sec]", - "MetricExpr": "( UNC_CHA_TOR_INSERTS.IO_HIT_ITOM + UNC_CHA_TOR_INSERTS.IO_MISS_ITOM + UNC_CHA_TOR_INSERTS.IO_HIT_ITOMCACHENEAR + UNC_CHA_TOR_INSERTS.IO_MISS_ITOMCACHENEAR ) * 64 / 1000000000 / duration_time", - "MetricGroup": "IoBW;Mem;SoC;Server", + "MetricExpr": "(UNC_CHA_TOR_INSERTS.IO_HIT_ITOM + UNC_CHA_TOR_INSERTS.IO_MISS_ITOM + UNC_CHA_TOR_INSERTS.IO_HIT_ITOMCACHENEAR + UNC_CHA_TOR_INSERTS.IO_MISS_ITOMCACHENEAR) * 64 / 1000000000 / duration_time", + "MetricGroup": "IoBW;Mem;Server;SoC", "MetricName": "IO_Read_BW" }, { @@ -487,12 +1231,6 @@ "MetricName": "Socket_CLKS" }, { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "cha_0@event\\=0x0@ / #num_dies / duration_time / 1000000000", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, - { "BriefDescription": "Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]", "MetricExpr": "INST_RETIRED.ANY / BR_INST_RETIRED.FAR_BRANCH:u", "MetricGroup": "Branches;OS", @@ -523,11 +1261,10 @@ "MetricName": "C6_Pkg_Residency" }, { - "BriefDescription": "Percentage of time spent in the active CPU power state C0", - "MetricExpr": "100 * CPU_CLK_UNHALTED.REF_TSC / TSC", - "MetricGroup": "", - "MetricName": "cpu_utilization_percent", - "ScaleUnit": "1%" + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "Socket_CLKS / #num_dies / duration_time / 1000000000", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ" }, { "BriefDescription": "CPU operating frequency (in GHz)", @@ -537,13 +1274,6 @@ "ScaleUnit": "1GHz" }, { - "BriefDescription": "Cycles per instruction retired; indicating how much time each executed instruction took; in units of cycles.", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY", - "MetricGroup": "", - "MetricName": "cpi", - "ScaleUnit": "1per_instr" - }, - { "BriefDescription": "The ratio of number of completed memory load instructions to the total number completed instructions", "MetricExpr": "MEM_INST_RETIRED.ALL_LOADS / INST_RETIRED.ANY", "MetricGroup": "", @@ -561,7 +1291,7 @@ "BriefDescription": "Ratio of number of requests missing L1 data cache (includes data+rfo w/ prefetches) to the total number of completed instructions", "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l1d_mpi_includes_data_plus_rfo_with_prefetches", + "MetricName": "l1d_mpi", "ScaleUnit": "1per_instr" }, { @@ -589,7 +1319,7 @@ "BriefDescription": "Ratio of number of requests missing L2 cache (includes code+data+rfo w/ prefetches) to the total number of completed instructions", "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches", + "MetricName": "l2_mpi", "ScaleUnit": "1per_instr" }, { @@ -615,42 +1345,42 @@ }, { "BriefDescription": "Ratio of number of code read requests missing last level core cache (includes demand w/ prefetches) to the total number of completed instructions", - "MetricExpr": "( UNC_CHA_TOR_INSERTS.IA_MISS_CRD ) / INST_RETIRED.ANY", + "MetricExpr": "( UNC_CHA_TOR_INSERTS.IA_MISS_CRD + UNC_CHA_TOR_INSERTS.IA_MISS_CRD_PREF ) / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "llc_code_read_mpi_demand_plus_prefetch", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand data read miss (read memory access) in nano seconds", - "MetricExpr": "( ( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD / UNC_CHA_TOR_INSERTS.IA_MISS_DRD ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD / UNC_CHA_TOR_INSERTS.IA_MISS_DRD ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_demand_data_read_miss_latency", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand data read miss (read memory access) addressed to local memory in nano seconds", - "MetricExpr": "( ( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_LOCAL / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_LOCAL ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_LOCAL) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_LOCAL / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_LOCAL ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_LOCAL) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_demand_data_read_miss_latency_for_local_requests", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand data read miss (read memory access) addressed to remote memory in nano seconds", - "MetricExpr": "( ( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_REMOTE / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_REMOTE ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_REMOTE) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_REMOTE / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_REMOTE ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_REMOTE) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_demand_data_read_miss_latency_for_remote_requests", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand data read miss (read memory access) addressed to Intel(R) Optane(TM) Persistent Memory(PMEM) in nano seconds", - "MetricExpr": "( ( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PMM ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PMM ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_demand_data_read_miss_to_pmem_latency", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand data read miss (read memory access) addressed to DRAM in nano seconds", - "MetricExpr": "( ( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_DDR ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_DDR ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_demand_data_read_miss_to_dram_latency", "ScaleUnit": "1ns" @@ -694,14 +1424,14 @@ "BriefDescription": "Memory read that miss the last level cache (LLC) addressed to local DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", "MetricExpr": "100 * ( UNC_CHA_TOR_INSERTS.IA_MISS_DRD_LOCAL + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_LOCAL ) / ( UNC_CHA_TOR_INSERTS.IA_MISS_DRD_LOCAL + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_LOCAL + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_REMOTE + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_REMOTE )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_local_dram", + "MetricName": "numa_reads_addressed_to_local_dram", "ScaleUnit": "1%" }, { "BriefDescription": "Memory reads that miss the last level cache (LLC) addressed to remote DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", "MetricExpr": "100 * ( UNC_CHA_TOR_INSERTS.IA_MISS_DRD_REMOTE + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_REMOTE ) / ( UNC_CHA_TOR_INSERTS.IA_MISS_DRD_LOCAL + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_LOCAL + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_REMOTE + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_REMOTE )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_remote_dram", + "MetricName": "numa_reads_addressed_to_remote_dram", "ScaleUnit": "1%" }, { @@ -715,7 +1445,7 @@ "BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data transmit bandwidth (MB/sec)", "MetricExpr": "( UNC_UPI_TxL_FLITS.ALL_DATA * (64 / 9.0) / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "upi_data_transmit_bw_only_data", + "MetricName": "upi_data_transmit_bw", "ScaleUnit": "1MB/s" }, { @@ -764,35 +1494,35 @@ "BriefDescription": "Bandwidth of IO reads that are initiated by end device controllers that are requesting memory from the CPU.", "MetricExpr": "(( UNC_CHA_TOR_INSERTS.IO_HIT_PCIRDCUR + UNC_CHA_TOR_INSERTS.IO_MISS_PCIRDCUR ) * 64 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_read", + "MetricName": "io_bandwidth_disk_or_network_writes", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Bandwidth of IO writes that are initiated by end device controllers that are writing memory to the CPU.", "MetricExpr": "(( UNC_CHA_TOR_INSERTS.IO_HIT_ITOM + UNC_CHA_TOR_INSERTS.IO_MISS_ITOM + UNC_CHA_TOR_INSERTS.IO_HIT_ITOMCACHENEAR + UNC_CHA_TOR_INSERTS.IO_MISS_ITOMCACHENEAR ) * 64 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_write", + "MetricName": "io_bandwidth_disk_or_network_reads", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Uops delivered from decoded instruction cache (decoded stream buffer or DSB) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.DSB_UOPS / ( IDQ.DSB_UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS + LSD.UOPS ) )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_decoded_icache_dsb", + "MetricName": "percent_uops_delivered_from_decoded_icache", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from legacy decode pipeline (Micro-instruction Translation Engine or MITE) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.MITE_UOPS / ( IDQ.DSB_UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS + LSD.UOPS ) )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline_mite", + "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from microcode sequencer (MS) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.MS_UOPS / ( IDQ.DSB_UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS + LSD.UOPS ) )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_microcode_sequencer_ms", + "MetricName": "percent_uops_delivered_from_microcode_sequencer", "ScaleUnit": "1%" }, { @@ -824,241 +1554,10 @@ "ScaleUnit": "1MB/s" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", - "MetricExpr": "100 * ( topdown\\-fe\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) )", - "MetricGroup": "TmaL1;PGO", - "MetricName": "tma_frontend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.", - "MetricExpr": "100 * ( ( ( 5 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING ) / ( slots ) )", - "MetricGroup": "Frontend;TmaL2;m_tma_frontend_bound_percent", - "MetricName": "tma_fetch_latency_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", - "MetricExpr": "100 * ( ICACHE_16B.IFDATA_STALL / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_icache_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.", - "MetricExpr": "100 * ( ICACHE_64B.IFTAG_STALL / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_itlb_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.", - "MetricExpr": "100 * ( INT_MISC.CLEAR_RESTEER_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) + ( ( 10 ) * BACLEARS.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) )", - "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_branch_resteers_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", - "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_dsb_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", - "MetricExpr": "100 * ( ILD_STALL.LCP / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_lcp_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.", - "MetricExpr": "100 * ( ( 3 ) * IDQ.MS_SWITCHES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_ms_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", - "MetricExpr": "100 * ( max( 0 , ( topdown\\-fe\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) ) - ( ( ( 5 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING ) / ( slots ) ) ) )", - "MetricGroup": "FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent", - "MetricName": "tma_fetch_bandwidth_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", - "MetricExpr": "100 * ( ( IDQ.MITE_CYCLES_ANY - IDQ.MITE_CYCLES_OK ) / ( CPU_CLK_UNHALTED.DISTRIBUTED ) / 2 )", - "MetricGroup": "DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent", - "MetricName": "tma_mite_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", - "MetricExpr": "100 * ( ( IDQ.DSB_CYCLES_ANY - IDQ.DSB_CYCLES_OK ) / ( CPU_CLK_UNHALTED.DISTRIBUTED ) / 2 )", - "MetricGroup": "DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent", - "MetricName": "tma_dsb_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", - "MetricExpr": "100 * ( max( 1 - ( ( topdown\\-fe\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) ) + ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) + ( ( 5 ) * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=0x1\\,edge\\=0x1@ ) / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) , 0 ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_bad_speculation_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.", - "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( max( 1 - ( ( topdown\\-fe\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) ) + ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) + ( ( 5 ) * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=0x1\\,edge\\=0x1@ ) / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) , 0 ) ) )", - "MetricGroup": "BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent", - "MetricName": "tma_branch_mispredicts_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.", - "MetricExpr": "100 * ( max( 0 , ( max( 1 - ( ( topdown\\-fe\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) ) + ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) + ( ( 5 ) * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=0x1\\,edge\\=0x1@ ) / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) , 0 ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( max( 1 - ( ( topdown\\-fe\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) ) + ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) + ( ( 5 ) * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=0x1\\,edge\\=0x1@ ) / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) , 0 ) ) ) ) )", - "MetricGroup": "BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent", - "MetricName": "tma_machine_clears_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", - "MetricExpr": "100 * ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) + ( ( 5 ) * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=0x1\\,edge\\=0x1@ ) / ( slots ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_backend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", - "MetricExpr": "100 * ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / ( CYCLE_ACTIVITY.STALLS_TOTAL + ( EXE_ACTIVITY.1_PORTS_UTIL + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * EXE_ACTIVITY.2_PORTS_UTIL ) + EXE_ACTIVITY.BOUND_ON_STORES ) ) * ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) + ( ( 5 ) * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=0x1\\,edge\\=0x1@ ) / ( slots ) ) )", - "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent", - "MetricName": "tma_memory_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.", - "MetricExpr": "100 * ( max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) , 0 ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l1_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) / ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + L1D_PEND_MISS.FB_FULL_PERIODS ) ) * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l2_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l3_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.", - "MetricExpr": "100 * ( min( ( ( ( CYCLE_ACTIVITY.STALLS_L3_MISS / ( CPU_CLK_UNHALTED.THREAD ) + ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) - ( ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) / ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + L1D_PEND_MISS.FB_FULL_PERIODS ) ) * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( min( ( ( ( ( 1 - ( ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) / ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) + ( 25 * ( ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) + 33 * ( ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) ) ) ) ) * ( CYCLE_ACTIVITY.STALLS_L3_MISS / ( CPU_CLK_UNHALTED.THREAD ) + ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) - ( ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) / ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + L1D_PEND_MISS.FB_FULL_PERIODS ) ) * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) if ( ( 1000000 ) * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) , ( 1 ) ) ) ) ) , ( 1 ) ) )", - "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_dram_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric roughly estimates (based on idle latencies) how often the CPU was stalled on accesses to external 3D-Xpoint (Crystal Ridge, a.k.a. IXP) memory by loads, PMM stands for Persistent Memory Module. ", - "MetricExpr": "100 * ( min( ( ( ( ( 1 - ( ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) / ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) + ( 25 * ( ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) + 33 * ( ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) ) ) ) ) * ( CYCLE_ACTIVITY.STALLS_L3_MISS / ( CPU_CLK_UNHALTED.THREAD ) + ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) - ( ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) / ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + L1D_PEND_MISS.FB_FULL_PERIODS ) ) * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) if ( ( 1000000 ) * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) , ( 1 ) ) )", - "MetricGroup": "MemoryBound;Server;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_pmm_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.", - "MetricExpr": "100 * ( EXE_ACTIVITY.BOUND_ON_STORES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_store_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", - "MetricExpr": "100 * ( max( 0 , ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) + ( ( 5 ) * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=0x1\\,edge\\=0x1@ ) / ( slots ) ) - ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / ( CYCLE_ACTIVITY.STALLS_TOTAL + ( EXE_ACTIVITY.1_PORTS_UTIL + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * EXE_ACTIVITY.2_PORTS_UTIL ) + EXE_ACTIVITY.BOUND_ON_STORES ) ) * ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) + ( ( 5 ) * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=0x1\\,edge\\=0x1@ ) / ( slots ) ) ) ) )", - "MetricGroup": "Backend;TmaL2;Compute;m_tma_backend_bound_percent", - "MetricName": "tma_core_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.", - "MetricExpr": "100 * ( ARITH.DIVIDER_ACTIVE / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "TmaL3;m_tma_core_bound_percent", - "MetricName": "tma_divider_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. ", - "MetricExpr": "( 100 * ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) + ( 0 * slots )", - "MetricGroup": "TmaL1", - "MetricName": "tma_retiring_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved.", - "MetricExpr": "100 * ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=0x1@ ) / IDQ.MITE_UOPS ) ) )", - "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", - "MetricName": "tma_light_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", - "MetricExpr": "100 * ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) , ( 1 ) ) ) )", - "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_fp_arith_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", - "MetricExpr": "100 * ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=0x1@ ) / IDQ.MITE_UOPS ) ) ) * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_memory_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions.", - "MetricExpr": "100 * ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=0x1@ ) / IDQ.MITE_UOPS ) ) ) * BR_INST_RETIRED.ALL_BRANCHES / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_branch_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body.", - "MetricExpr": "100 * ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=0x1@ ) / IDQ.MITE_UOPS ) ) ) * INST_RETIRED.NOP / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_nop_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", - "MetricExpr": "100 * ( max( 0 , ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=0x1@ ) / IDQ.MITE_UOPS ) ) ) - ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) , ( 1 ) ) ) ) + ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=0x1@ ) / IDQ.MITE_UOPS ) ) ) * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY ) + ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=0x1@ ) / IDQ.MITE_UOPS ) ) ) * BR_INST_RETIRED.ALL_BRANCHES / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=0x1@ ) / IDQ.MITE_UOPS ) ) ) * INST_RETIRED.NOP / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_other_light_ops_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", - "MetricExpr": "100 * ( ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=0x1@ ) / IDQ.MITE_UOPS )", - "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", - "MetricName": "tma_heavy_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", - "MetricExpr": "100 * ( ( ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=0x1@ ) / IDQ.MITE_UOPS ) - ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) ) )", - "MetricGroup": "TmaL3;m_tma_heavy_operations_percent", - "MetricName": "tma_few_uops_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided.", - "MetricExpr": "100 * ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( slots ) )", - "MetricGroup": "MicroSeq;TmaL3;m_tma_heavy_operations_percent", - "MetricName": "tma_microcode_sequencer_percent", + "BriefDescription": "%", + "MetricExpr": "100 * ( ( LSD.CYCLES_ACTIVE - LSD.CYCLES_OK ) / ( CPU_CLK_UNHALTED.DISTRIBUTED ) / 2 )", + "MetricGroup": "FetchBW;LSD;TopdownL3;tma_L3_group;tma_fetch_bandwidth_group", + "MetricName": "tma_lsd", "ScaleUnit": "1%" } ] diff --git a/tools/perf/pmu-events/arch/x86/icelakex/pipeline.json b/tools/perf/pmu-events/arch/x86/icelakex/pipeline.json index 396868f70004..52fba238bf1f 100644 --- a/tools/perf/pmu-events/arch/x86/icelakex/pipeline.json +++ b/tools/perf/pmu-events/arch/x86/icelakex/pipeline.json @@ -167,7 +167,7 @@ "UMask": "0x10" }, { - "BriefDescription": "number of branch instructions retired that were mispredicted and taken. Non PEBS", + "BriefDescription": "number of branch instructions retired that were mispredicted and taken.", "CollectPEBSRecord": "2", "Counter": "0,1,2,3,4,5,6,7", "EventCode": "0xc5", diff --git a/tools/perf/pmu-events/arch/x86/icelakex/uncore-other.json b/tools/perf/pmu-events/arch/x86/icelakex/uncore-other.json index 7783aa2ef5d1..03e99b8aed93 100644 --- a/tools/perf/pmu-events/arch/x86/icelakex/uncore-other.json +++ b/tools/perf/pmu-events/arch/x86/icelakex/uncore-other.json @@ -11779,7 +11779,7 @@ "Unit": "M3UPI" }, { - "BriefDescription": "Flit Gen - Header 1 : Acumullate", + "BriefDescription": "Flit Gen - Header 1 : Accumulate", "Counter": "0,1,2,3", "CounterType": "PGMABLE", "EventCode": "0x51", diff --git a/tools/perf/pmu-events/arch/x86/ivybridge/ivb-metrics.json b/tools/perf/pmu-events/arch/x86/ivybridge/ivb-metrics.json index 3f48e75f8a86..63db3397af0f 100644 --- a/tools/perf/pmu-events/arch/x86/ivybridge/ivb-metrics.json +++ b/tools/perf/pmu-events/arch/x86/ivybridge/ivb-metrics.json @@ -1,64 +1,500 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * min(CPU_CLK_UNHALTED.THREAD, IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE) / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: RS_EVENTS.EMPTY_END", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", + "MetricExpr": "ICACHE.IFETCH_STALL / CLKS - tma_itlb_misses", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "(12 * ITLB_MISSES.STLB_HIT + ITLB_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: ITLB_MISSES.WALK_COMPLETED", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "12 * (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY) / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "3 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - (tma_frontend_bound + tma_bad_speculation + tma_retiring)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING) + RESOURCE_STALLS.SB) / (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if (IPC > 1.8) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING) - CYCLE_ACTIVITY.STALLS_L1D_PENDING) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_UOPS_RETIRED.L1_HIT_PS;MEM_LOAD_UOPS_RETIRED.HIT_LFB_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "(7 * DTLB_LOAD_MISSES.STLB_HIT + DTLB_LOAD_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_UOPS_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_UOPS_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "13 * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_UOPS_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "Load_Miss_Real_Latency * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L1D_PENDING - CYCLE_ACTIVITY.STALLS_L2_PENDING) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEM_LOAD_UOPS_RETIRED.LLC_HIT / (MEM_LOAD_UOPS_RETIRED.LLC_HIT + 7 * MEM_LOAD_UOPS_RETIRED.LLC_MISS)) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "(60 * (MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.LLC_MISS))) + 43 * (MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.LLC_MISS)))) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "43 * (MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.LLC_MISS))) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "29 * (MEM_LOAD_UOPS_RETIRED.LLC_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_RETIRED.LLC_MISS))) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "((OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2) if #SMT_on else OFFCORE_REQUESTS_BUFFER.SQ_FULL) / CORE_CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(1 - (MEM_LOAD_UOPS_RETIRED.LLC_HIT / (MEM_LOAD_UOPS_RETIRED.LLC_HIT + 7 * MEM_LOAD_UOPS_RETIRED.LLC_MISS))) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=6@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "RESOURCE_STALLS.SB / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_UOPS_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 9 * (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES))) + (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "60 * OFFCORE_RESPONSE.DEMAND_RFO.LLC_HIT.HITM_OTHER_CORE / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "2 * MEM_UOPS_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_UOPS_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(7 * DTLB_STORE_MISSES.STLB_HIT + DTLB_STORE_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_UOPS_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.FPU_DIV_ACTIVE / CORE_CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if (IPC > 1.8) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) - RESOURCE_STALLS.SB - min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING)) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,inv\\,cmask\\=1@) / 2 if #SMT_on else (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else 0) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 if #SMT_on else (UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / 2 if #SMT_on else (UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).", + "MetricExpr": "((cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ / 2) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_0 + UOPS_DISPATCHED_PORT.PORT_1 + UOPS_DISPATCHED_PORT.PORT_5) / (3 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED_PORT.PORT_0", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_1", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_2 + UOPS_DISPATCHED_PORT.PORT_3 - UOPS_DISPATCHED_PORT.PORT_4) / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_2", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_2 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_2", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_3", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_3 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data) Sample with: UOPS_DISPATCHED_PORT.PORT_4", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_4", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS * FP_COMP_OPS_EXE.X87 / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * OTHER_ASSISTS.ANY_WB_ASSIST / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: OTHER_ASSISTS.ANY", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" }, { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -76,8 +512,8 @@ }, { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI" }, { @@ -88,17 +524,11 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "The ratio of Executed- by Issued-Uops", "MetricExpr": "UOPS_EXECUTED.THREAD / UOPS_ISSUED.ANY", "MetricGroup": "Cor;Pipeline", @@ -107,37 +537,25 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * (FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) + 8 * SIMD_FP_256.PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { - "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;Flops_SMT", - "MetricName": "FLOPc_SMT" - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, @@ -179,15 +597,15 @@ }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "1 / ( ((FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) / UOPS_EXECUTED.THREAD) + ((FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) / UOPS_EXECUTED.THREAD) )", + "MetricExpr": "1 / (tma_fp_scalar + tma_fp_vector)", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -204,7 +622,7 @@ }, { "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)", - "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )", + "MetricExpr": "IDQ.DSB_UOPS / ((IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS))", "MetricGroup": "DSB;Fed;FetchBW", "MetricName": "DSB_Coverage" }, @@ -216,48 +634,42 @@ }, { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.LLC_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION ) / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "(ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION) / CORE_CLKS", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, { - "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", - "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Mem;MemoryTLB_SMT", - "MetricName": "Page_Walks_Utilization_SMT" - }, - { "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]", "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time", "MetricGroup": "Mem;MemoryBW", @@ -277,19 +689,19 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, @@ -307,26 +719,26 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * (FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) + 8 * SIMD_FP_256.PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -344,7 +756,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * ( arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@ ) / 1000000 / duration_time / 1000", + "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1000000 / duration_time / 1000", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, diff --git a/tools/perf/pmu-events/arch/x86/ivytown/cache.json b/tools/perf/pmu-events/arch/x86/ivytown/cache.json index 27576d53b347..d95b98c83914 100644 --- a/tools/perf/pmu-events/arch/x86/ivytown/cache.json +++ b/tools/perf/pmu-events/arch/x86/ivytown/cache.json @@ -21,7 +21,7 @@ "UMask": "0x2" }, { - "BriefDescription": "L1D miss oustandings duration in cycles", + "BriefDescription": "L1D miss outstanding duration in cycles", "Counter": "2", "CounterHTOff": "2", "EventCode": "0x48", @@ -658,7 +658,7 @@ "UMask": "0x8" }, { - "BriefDescription": "Cacheable and noncachaeble code read requests", + "BriefDescription": "Cacheable and noncacheable code read requests", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0xB0", diff --git a/tools/perf/pmu-events/arch/x86/ivytown/floating-point.json b/tools/perf/pmu-events/arch/x86/ivytown/floating-point.json index 4c2ac010cf55..88891cba54ec 100644 --- a/tools/perf/pmu-events/arch/x86/ivytown/floating-point.json +++ b/tools/perf/pmu-events/arch/x86/ivytown/floating-point.json @@ -91,7 +91,7 @@ "UMask": "0x20" }, { - "BriefDescription": "Number of FP Computational Uops Executed this cycle. The number of FADD, FSUB, FCOM, FMULs, integer MULsand IMULs, FDIVs, FPREMs, FSQRTS, integer DIVs, and IDIVs. This event does not distinguish an FADD used in the middle of a transcendental flow from a s", + "BriefDescription": "Number of FP Computational Uops Executed this cycle. The number of FADD, FSUB, FCOM, FMULs, integer MULs and IMULs, FDIVs, FPREMs, FSQRTS, integer DIVs, and IDIVs. This event does not distinguish an FADD used in the middle of a transcendental flow from a s", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0x10", diff --git a/tools/perf/pmu-events/arch/x86/ivytown/frontend.json b/tools/perf/pmu-events/arch/x86/ivytown/frontend.json index 2b1a82dd86ab..0a295c4e093d 100644 --- a/tools/perf/pmu-events/arch/x86/ivytown/frontend.json +++ b/tools/perf/pmu-events/arch/x86/ivytown/frontend.json @@ -176,41 +176,41 @@ "UMask": "0x4" }, { - "BriefDescription": "Cycles when uops are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Cycles when uops are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "CounterMask": "1", "EventCode": "0x79", "EventName": "IDQ.MS_CYCLES", - "PublicDescription": "Cycles when uops are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy.", + "PublicDescription": "Cycles when uops are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy.", "SampleAfterValue": "2000003", "UMask": "0x30" }, { - "BriefDescription": "Cycles when uops initiated by Decode Stream Buffer (DSB) are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Cycles when uops initiated by Decode Stream Buffer (DSB) are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "CounterMask": "1", "EventCode": "0x79", "EventName": "IDQ.MS_DSB_CYCLES", - "PublicDescription": "Cycles when uops initiated by Decode Stream Buffer (DSB) are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy.", + "PublicDescription": "Cycles when uops initiated by Decode Stream Buffer (DSB) are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy.", "SampleAfterValue": "2000003", "UMask": "0x10" }, { - "BriefDescription": "Deliveries to Instruction Decode Queue (IDQ) initiated by Decode Stream Buffer (DSB) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Deliveries to Instruction Decode Queue (IDQ) initiated by Decode Stream Buffer (DSB) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "CounterMask": "1", "EdgeDetect": "1", "EventCode": "0x79", "EventName": "IDQ.MS_DSB_OCCUR", - "PublicDescription": "Deliveries to Instruction Decode Queue (IDQ) initiated by Decode Stream Buffer (DSB) while Microcode Sequenser (MS) is busy.", + "PublicDescription": "Deliveries to Instruction Decode Queue (IDQ) initiated by Decode Stream Buffer (DSB) while Microcode Sequencer (MS) is busy.", "SampleAfterValue": "2000003", "UMask": "0x10" }, { - "BriefDescription": "Uops initiated by Decode Stream Buffer (DSB) that are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Uops initiated by Decode Stream Buffer (DSB) that are being delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0x79", @@ -220,7 +220,7 @@ "UMask": "0x10" }, { - "BriefDescription": "Uops initiated by MITE and delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Uops initiated by MITE and delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0x79", @@ -242,7 +242,7 @@ "UMask": "0x30" }, { - "BriefDescription": "Uops delivered to Instruction Decode Queue (IDQ) while Microcode Sequenser (MS) is busy", + "BriefDescription": "Uops delivered to Instruction Decode Queue (IDQ) while Microcode Sequencer (MS) is busy", "Counter": "0,1,2,3", "CounterHTOff": "0,1,2,3,4,5,6,7", "EventCode": "0x79", diff --git a/tools/perf/pmu-events/arch/x86/ivytown/ivt-metrics.json b/tools/perf/pmu-events/arch/x86/ivytown/ivt-metrics.json index 19c7f3b41102..99a45c8d8cee 100644 --- a/tools/perf/pmu-events/arch/x86/ivytown/ivt-metrics.json +++ b/tools/perf/pmu-events/arch/x86/ivytown/ivt-metrics.json @@ -1,64 +1,524 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * min(CPU_CLK_UNHALTED.THREAD, IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE) / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: RS_EVENTS.EMPTY_END", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", + "MetricExpr": "ICACHE.IFETCH_STALL / CLKS - tma_itlb_misses", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "(12 * ITLB_MISSES.STLB_HIT + ITLB_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: ITLB_MISSES.WALK_COMPLETED", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "12 * (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY) / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "3 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - (tma_frontend_bound + tma_bad_speculation + tma_retiring)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING) + RESOURCE_STALLS.SB) / (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if (IPC > 1.8) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING) - CYCLE_ACTIVITY.STALLS_L1D_PENDING) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_UOPS_RETIRED.L1_HIT_PS;MEM_LOAD_UOPS_RETIRED.HIT_LFB_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "(7 * DTLB_LOAD_MISSES.STLB_HIT + DTLB_LOAD_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_UOPS_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_UOPS_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "13 * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_UOPS_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "Load_Miss_Real_Latency * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L1D_PENDING - CYCLE_ACTIVITY.STALLS_L2_PENDING) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEM_LOAD_UOPS_RETIRED.LLC_HIT / (MEM_LOAD_UOPS_RETIRED.LLC_HIT + 7 * MEM_LOAD_UOPS_RETIRED.LLC_MISS)) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "(60 * (MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD))) + 43 * (MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD)))) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "43 * (MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "41 * (MEM_LOAD_UOPS_RETIRED.LLC_HIT * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "((OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2) if #SMT_on else OFFCORE_REQUESTS_BUFFER.SQ_FULL) / CORE_CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(1 - (MEM_LOAD_UOPS_RETIRED.LLC_HIT / (MEM_LOAD_UOPS_RETIRED.LLC_HIT + 7 * MEM_LOAD_UOPS_RETIRED.LLC_MISS))) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=6@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory", + "MetricExpr": "200 * (MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "Server;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_local_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory. Caching will improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory", + "MetricExpr": "310 * (MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD))) / CLKS", + "MetricGroup": "Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues", + "MetricExpr": "(200 * (MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD))) + 180 * (MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD * (1 + mem_load_uops_retired.hit_lfb / ((MEM_LOAD_UOPS_RETIRED.L2_HIT + MEM_LOAD_UOPS_RETIRED.LLC_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM + MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS) + MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM + MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD)))) / CLKS", + "MetricGroup": "Offcore;Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_cache", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM_PS;MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "RESOURCE_STALLS.SB / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_UOPS_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 9 * (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES))) + (1 - (MEM_UOPS_RETIRED.LOCK_LOADS / MEM_UOPS_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "(200 * OFFCORE_RESPONSE.DEMAND_RFO.LLC_MISS.REMOTE_HITM + 60 * OFFCORE_RESPONSE.DEMAND_RFO.LLC_HIT.HITM_OTHER_CORE) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "2 * MEM_UOPS_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_UOPS_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(7 * DTLB_STORE_MISSES.STLB_HIT + DTLB_STORE_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_UOPS_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.FPU_DIV_ACTIVE / CORE_CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if (IPC > 1.8) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) - RESOURCE_STALLS.SB - min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_LDM_PENDING)) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,inv\\,cmask\\=1@) / 2 if #SMT_on else (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_EXECUTE) - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else 0) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@) / 2 if #SMT_on else (UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(cpu@UOPS_EXECUTED.CORE\\,cmask\\=2@ - cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@) / 2 if #SMT_on else (UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC - UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).", + "MetricExpr": "((cpu@UOPS_EXECUTED.CORE\\,cmask\\=3@ / 2) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_0 + UOPS_DISPATCHED_PORT.PORT_1 + UOPS_DISPATCHED_PORT.PORT_5) / (3 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED_PORT.PORT_0", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_1", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_2 + UOPS_DISPATCHED_PORT.PORT_3 - UOPS_DISPATCHED_PORT.PORT_4) / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_2", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_2 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_2", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_3", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_3 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data) Sample with: UOPS_DISPATCHED_PORT.PORT_4", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_4", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS * FP_COMP_OPS_EXE.X87 / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * OTHER_ASSISTS.ANY_WB_ASSIST / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: OTHER_ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" }, { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -76,8 +536,8 @@ }, { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI" }, { @@ -88,17 +548,11 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "The ratio of Executed- by Issued-Uops", "MetricExpr": "UOPS_EXECUTED.THREAD / UOPS_ISSUED.ANY", "MetricGroup": "Cor;Pipeline", @@ -107,37 +561,25 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * (FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) + 8 * SIMD_FP_256.PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { - "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;Flops_SMT", - "MetricName": "FLOPc_SMT" - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, @@ -179,15 +621,15 @@ }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "1 / ( ((FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) / UOPS_EXECUTED.THREAD) + ((FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) / UOPS_EXECUTED.THREAD) )", + "MetricExpr": "1 / (tma_fp_scalar + tma_fp_vector)", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -204,7 +646,7 @@ }, { "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)", - "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )", + "MetricExpr": "IDQ.DSB_UOPS / ((IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS))", "MetricGroup": "DSB;Fed;FetchBW", "MetricName": "DSB_Coverage" }, @@ -216,48 +658,42 @@ }, { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_UOPS_RETIRED.LLC_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION ) / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "(ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION) / CORE_CLKS", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, { - "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", - "MetricExpr": "( ITLB_MISSES.WALK_DURATION + DTLB_LOAD_MISSES.WALK_DURATION + DTLB_STORE_MISSES.WALK_DURATION ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Mem;MemoryTLB_SMT", - "MetricName": "Page_Walks_Utilization_SMT" - }, - { "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]", "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time", "MetricGroup": "Mem;MemoryBW", @@ -277,19 +713,19 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, @@ -307,26 +743,26 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * (FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) + 8 * SIMD_FP_256.PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -344,7 +780,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "( 64 * ( uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@ ) / 1000000000 ) / duration_time", + "MetricExpr": "(64 * (uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@) / 1000000000) / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, @@ -355,12 +791,6 @@ "MetricName": "Socket_CLKS" }, { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "cbox_0@event\\=0x0@ / #num_dies / duration_time / 1000000000", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, - { "BriefDescription": "Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]", "MetricExpr": "INST_RETIRED.ANY / BR_INST_RETIRED.FAR_BRANCH:u", "MetricGroup": "Branches;OS", @@ -407,5 +837,11 @@ "MetricExpr": "(cstate_pkg@c7\\-residency@ / msr@tsc@) * 100", "MetricGroup": "Power", "MetricName": "C7_Pkg_Residency" + }, + { + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "Socket_CLKS / #num_dies / duration_time / 1000000000", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ" } ] diff --git a/tools/perf/pmu-events/arch/x86/ivytown/uncore-cache.json b/tools/perf/pmu-events/arch/x86/ivytown/uncore-cache.json index 93e07385eeec..c118ff54c30e 100644 --- a/tools/perf/pmu-events/arch/x86/ivytown/uncore-cache.json +++ b/tools/perf/pmu-events/arch/x86/ivytown/uncore-cache.json @@ -61,7 +61,7 @@ "EventCode": "0x34", "EventName": "UNC_C_LLC_LOOKUP.WRITE", "PerPkg": "1", - "PublicDescription": "Counts the number of times the LLC was accessed - this includes code, data, prefetches and hints coming from L2. This has numerous filters available. Note the non-standard filtering equation. This event will count requests that lookup the cache multiple times with multiple increments. One must ALWAYS set filter mask bit 0 and select a state or states to match. Otherwise, the event will count nothing. CBoGlCtrl[22:17] bits correspond to [M'FMESI] state.; Writeback transactions from L2 to the LLC This includes all write transactions -- both Cachable and UC.", + "PublicDescription": "Counts the number of times the LLC was accessed - this includes code, data, prefetches and hints coming from L2. This has numerous filters available. Note the non-standard filtering equation. This event will count requests that lookup the cache multiple times with multiple increments. One must ALWAYS set filter mask bit 0 and select a state or states to match. Otherwise, the event will count nothing. CBoGlCtrl[22:17] bits correspond to [M'FMESI] state.; Writeback transactions from L2 to the LLC This includes all write transactions -- both Cacheable and UC.", "UMask": "0x5", "Unit": "CBO" }, @@ -999,7 +999,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.ALL", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All transactions inserted into the TOR. This includes requests that reside in the TOR for a short time, such as LLC Hits that do not need to snoop cores or requests that get rejected and have to be retried through one of the ingress queues. The TOR is more commonly a bottleneck in skews with smaller core counts, where the ratio of RTIDs to TOR entries is larger. Note that there are reserved TOR entries for various request types, so it is possible that a given request type be blocked with an occupancy that is less than 20. Also note that generally requests will not be able to arbitrate into the TOR pipeline if there are no available TOR slots.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All transactions inserted into the TOR. This includes requests that reside in the TOR for a short time, such as LLC Hits that do not need to snoop cores or requests that get rejected and have to be retried through one of the ingress queues. The TOR is more commonly a bottleneck in skews with smaller core counts, where the ratio of RTIDs to TOR entries is larger. Note that there are reserved TOR entries for various request types, so it is possible that a given request type be blocked with an occupancy that is less than 20. Also note that generally requests will not be able to arbitrate into the TOR pipeline if there are no available TOR slots.", "UMask": "0x8", "Unit": "CBO" }, @@ -1009,7 +1009,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.EVICTION", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Eviction transactions inserted into the TOR. Evictions can be quick, such as when the line is in the F, S, or E states and no core valid bits are set. They can also be longer if either CV bits are set (so the cores need to be snooped) and/or if there is a HitM (in which case it is necessary to write the request out to memory).", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Eviction transactions inserted into the TOR. Evictions can be quick, such as when the line is in the F, S, or E states and no core valid bits are set. They can also be longer if either CV bits are set (so the cores need to be snooped) and/or if there is a HitM (in which case it is necessary to write the request out to memory).", "UMask": "0x4", "Unit": "CBO" }, @@ -1019,7 +1019,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.LOCAL", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All transactions inserted into the TOR that are satisifed by locally HOMed memory.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All transactions inserted into the TOR that are satisfied by locally HOMed memory.", "UMask": "0x28", "Unit": "CBO" }, @@ -1029,7 +1029,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.LOCAL_OPCODE", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All transactions, satisifed by an opcode, inserted into the TOR that are satisifed by locally HOMed memory.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All transactions, satisfied by an opcode, inserted into the TOR that are satisfied by locally HOMed memory.", "UMask": "0x21", "Unit": "CBO" }, @@ -1039,7 +1039,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.MISS_LOCAL", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions inserted into the TOR that are satisifed by locally HOMed memory.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions inserted into the TOR that are satisfied by locally HOMed memory.", "UMask": "0x2A", "Unit": "CBO" }, @@ -1049,7 +1049,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.MISS_LOCAL_OPCODE", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions, satisifed by an opcode, inserted into the TOR that are satisifed by locally HOMed memory.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions, satisfied by an opcode, inserted into the TOR that are satisfied by locally HOMed memory.", "UMask": "0x23", "Unit": "CBO" }, @@ -1059,7 +1059,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions inserted into the TOR that match an opcode.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions inserted into the TOR that match an opcode.", "UMask": "0x3", "Unit": "CBO" }, @@ -1069,7 +1069,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.MISS_REMOTE", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions inserted into the TOR that are satisifed by remote caches or remote memory.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions inserted into the TOR that are satisfied by remote caches or remote memory.", "UMask": "0x8A", "Unit": "CBO" }, @@ -1079,7 +1079,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.MISS_REMOTE_OPCODE", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions, satisifed by an opcode, inserted into the TOR that are satisifed by remote caches or remote memory.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions, satisfied by an opcode, inserted into the TOR that are satisfied by remote caches or remote memory.", "UMask": "0x83", "Unit": "CBO" }, @@ -1089,7 +1089,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.NID_ALL", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All NID matched (matches an RTID destination) transactions inserted into the TOR. The NID is programmed in Cn_MSR_PMON_BOX_FILTER.nid. In conjunction with STATE = I, it is possible to monitor misses to specific NIDs in the system.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All NID matched (matches an RTID destination) transactions inserted into the TOR. The NID is programmed in Cn_MSR_PMON_BOX_FILTER.nid. In conjunction with STATE = I, it is possible to monitor misses to specific NIDs in the system.", "UMask": "0x48", "Unit": "CBO" }, @@ -1099,7 +1099,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.NID_EVICTION", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; NID matched eviction transactions inserted into the TOR.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; NID matched eviction transactions inserted into the TOR.", "UMask": "0x44", "Unit": "CBO" }, @@ -1109,7 +1109,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.NID_MISS_ALL", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All NID matched miss requests that were inserted into the TOR.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All NID matched miss requests that were inserted into the TOR.", "UMask": "0x4A", "Unit": "CBO" }, @@ -1119,7 +1119,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.NID_MISS_OPCODE", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions inserted into the TOR that match a NID and an opcode.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Miss transactions inserted into the TOR that match a NID and an opcode.", "UMask": "0x43", "Unit": "CBO" }, @@ -1129,7 +1129,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.NID_OPCODE", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Transactions inserted into the TOR that match a NID and an opcode.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Transactions inserted into the TOR that match a NID and an opcode.", "UMask": "0x41", "Unit": "CBO" }, @@ -1139,7 +1139,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.NID_WB", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; NID matched write transactions inserted into the TOR.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; NID matched write transactions inserted into the TOR.", "UMask": "0x50", "Unit": "CBO" }, @@ -1149,7 +1149,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.OPCODE", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Transactions inserted into the TOR that match an opcode (matched by Cn_MSR_PMON_BOX_FILTER.opc)", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Transactions inserted into the TOR that match an opcode (matched by Cn_MSR_PMON_BOX_FILTER.opc)", "UMask": "0x1", "Unit": "CBO" }, @@ -1159,7 +1159,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.REMOTE", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All transactions inserted into the TOR that are satisifed by remote caches or remote memory.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All transactions inserted into the TOR that are satisfied by remote caches or remote memory.", "UMask": "0x88", "Unit": "CBO" }, @@ -1169,7 +1169,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.REMOTE_OPCODE", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All transactions, satisifed by an opcode, inserted into the TOR that are satisifed by remote caches or remote memory.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; All transactions, satisfied by an opcode, inserted into the TOR that are satisfied by remote caches or remote memory.", "UMask": "0x81", "Unit": "CBO" }, @@ -1179,7 +1179,7 @@ "EventCode": "0x35", "EventName": "UNC_C_TOR_INSERTS.WB", "PerPkg": "1", - "PublicDescription": "Counts the number of entries successfuly inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Write transactions inserted into the TOR. This does not include RFO, but actual operations that contain data being sent from the core.", + "PublicDescription": "Counts the number of entries successfully inserted into the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182).; Write transactions inserted into the TOR. This does not include RFO, but actual operations that contain data being sent from the core.", "UMask": "0x10", "Unit": "CBO" }, @@ -1215,7 +1215,7 @@ "EventCode": "0x36", "EventName": "UNC_C_TOR_OCCUPANCY.LOCAL_OPCODE", "PerPkg": "1", - "PublicDescription": "For each cycle, this event accumulates the number of valid entries in the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182); Number of outstanding transactions, satisifed by an opcode, in the TOR that are satisifed by locally HOMed memory.", + "PublicDescription": "For each cycle, this event accumulates the number of valid entries in the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182); Number of outstanding transactions, satisfied by an opcode, in the TOR that are satisfied by locally HOMed memory.", "UMask": "0x21", "Unit": "CBO" }, @@ -1242,7 +1242,7 @@ "EventCode": "0x36", "EventName": "UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE", "PerPkg": "1", - "PublicDescription": "For each cycle, this event accumulates the number of valid entries in the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182); Number of outstanding Miss transactions, satisifed by an opcode, in the TOR that are satisifed by locally HOMed memory.", + "PublicDescription": "For each cycle, this event accumulates the number of valid entries in the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182); Number of outstanding Miss transactions, satisfied by an opcode, in the TOR that are satisfied by locally HOMed memory.", "UMask": "0x23", "Unit": "CBO" }, @@ -1269,7 +1269,7 @@ "EventCode": "0x36", "EventName": "UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE", "PerPkg": "1", - "PublicDescription": "For each cycle, this event accumulates the number of valid entries in the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182); Number of outstanding Miss transactions, satisifed by an opcode, in the TOR that are satisifed by remote caches or remote memory.", + "PublicDescription": "For each cycle, this event accumulates the number of valid entries in the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182); Number of outstanding Miss transactions, satisfied by an opcode, in the TOR that are satisfied by remote caches or remote memory.", "UMask": "0x83", "Unit": "CBO" }, @@ -1350,7 +1350,7 @@ "EventCode": "0x36", "EventName": "UNC_C_TOR_OCCUPANCY.REMOTE_OPCODE", "PerPkg": "1", - "PublicDescription": "For each cycle, this event accumulates the number of valid entries in the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182); Number of outstanding transactions, satisifed by an opcode, in the TOR that are satisifed by remote caches or remote memory.", + "PublicDescription": "For each cycle, this event accumulates the number of valid entries in the TOR that match qualifications specified by the subevent. There are a number of subevent 'filters' but only a subset of the subevent combinations are valid. Subevents that require an opcode or NID match require the Cn_MSR_PMON_BOX_FILTER.{opc, nid} field to be set. If, for example, one wanted to count DRD Local Misses, one should select MISS_OPC_MATCH and set Cn_MSR_PMON_BOX_FILTER.opc to DRD (0x182); Number of outstanding transactions, satisfied by an opcode, in the TOR that are satisfied by remote caches or remote memory.", "UMask": "0x81", "Unit": "CBO" }, @@ -1446,7 +1446,7 @@ "EventCode": "0x2", "EventName": "UNC_C_TxR_INSERTS.BL_CORE", "PerPkg": "1", - "PublicDescription": "Number of allocations into the Cbo Egress. The Egress is used to queue up requests destined for the ring.; Ring transactions from the Corebo destined for the BL ring. This is commonly used for transfering writeback data to the cache.", + "PublicDescription": "Number of allocations into the Cbo Egress. The Egress is used to queue up requests destined for the ring.; Ring transactions from the Corebo destined for the BL ring. This is commonly used for transferring writeback data to the cache.", "UMask": "0x40", "Unit": "CBO" }, @@ -1692,7 +1692,7 @@ "EventCode": "0xb", "EventName": "UNC_H_CONFLICT_CYCLES.LAST", "PerPkg": "1", - "PublicDescription": "Count every last conflictor in conflict chain. Can be used to compute the average conflict chain length as (#Ackcnflts/#LastConflictor)+1. This can be used to give a feel for the conflict chain lenghts while analyzing lock kernels.", + "PublicDescription": "Count every last conflictor in conflict chain. Can be used to compute the average conflict chain length as (#Ackcnflts/#LastConflictor)+1. This can be used to give a feel for the conflict chain lengths while analyzing lock kernels.", "UMask": "0x4", "Unit": "HA" }, @@ -1729,7 +1729,7 @@ "EventCode": "0x41", "EventName": "UNC_H_DIRECTORY_LAT_OPT", "PerPkg": "1", - "PublicDescription": "Directory Latency Optimization Data Return Path Taken. When directory mode is enabled and the directory retuned for a read is Dir=I, then data can be returned using a faster path if certain conditions are met (credits, free pipeline, etc).", + "PublicDescription": "Directory Latency Optimization Data Return Path Taken. When directory mode is enabled and the directory returned for a read is Dir=I, then data can be returned using a faster path if certain conditions are met (credits, free pipeline, etc).", "Unit": "HA" }, { @@ -2686,7 +2686,7 @@ "EventCode": "0x21", "EventName": "UNC_H_SNOOP_RESP.RSPSFWD", "PerPkg": "1", - "PublicDescription": "Counts the total number of RspI snoop responses received. Whenever a snoops are issued, one or more snoop responses will be returned depending on the topology of the system. In systems larger than 2s, when multiple snoops are returned this will count all the snoops that are received. For example, if 3 snoops were issued and returned RspI, RspS, and RspSFwd; then each of these sub-events would increment by 1.; Filters for a snoop response of RspSFwd. This is returned when a remote caching agent forwards data but holds on to its currentl copy. This is common for data and code reads that hit in a remote socket in E or F state.", + "PublicDescription": "Counts the total number of RspI snoop responses received. Whenever a snoops are issued, one or more snoop responses will be returned depending on the topology of the system. In systems larger than 2s, when multiple snoops are returned this will count all the snoops that are received. For example, if 3 snoops were issued and returned RspI, RspS, and RspSFwd; then each of these sub-events would increment by 1.; Filters for a snoop response of RspSFwd. This is returned when a remote caching agent forwards data but holds on to its currently copy. This is common for data and code reads that hit in a remote socket in E or F state.", "UMask": "0x8", "Unit": "HA" }, @@ -2766,7 +2766,7 @@ "EventCode": "0x60", "EventName": "UNC_H_SNP_RESP_RECV_LOCAL.RSPSFWD", "PerPkg": "1", - "PublicDescription": "Number of snoop responses received for a Local request; Filters for a snoop response of RspSFwd. This is returned when a remote caching agent forwards data but holds on to its currentl copy. This is common for data and code reads that hit in a remote socket in E or F state.", + "PublicDescription": "Number of snoop responses received for a Local request; Filters for a snoop response of RspSFwd. This is returned when a remote caching agent forwards data but holds on to its currently copy. This is common for data and code reads that hit in a remote socket in E or F state.", "UMask": "0x8", "Unit": "HA" }, diff --git a/tools/perf/pmu-events/arch/x86/ivytown/uncore-interconnect.json b/tools/perf/pmu-events/arch/x86/ivytown/uncore-interconnect.json index b3b1a08d4acf..10ea4afeffc1 100644 --- a/tools/perf/pmu-events/arch/x86/ivytown/uncore-interconnect.json +++ b/tools/perf/pmu-events/arch/x86/ivytown/uncore-interconnect.json @@ -24,7 +24,7 @@ "EventCode": "0x13", "EventName": "UNC_Q_DIRECT2CORE.FAILURE_CREDITS", "PerPkg": "1", - "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because there were not enough Egress credits. Had there been enough credits, the spawn would have worked as the RBT bit was set and the RBT tag matched.", + "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exclusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because there were not enough Egress credits. Had there been enough credits, the spawn would have worked as the RBT bit was set and the RBT tag matched.", "UMask": "0x2", "Unit": "QPI LL" }, @@ -34,7 +34,7 @@ "EventCode": "0x13", "EventName": "UNC_Q_DIRECT2CORE.FAILURE_CREDITS_MISS", "PerPkg": "1", - "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because the RBT tag did not match and there weren't enough Egress credits. The valid bit was set.", + "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exclusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because the RBT tag did not match and there weren't enough Egress credits. The valid bit was set.", "UMask": "0x20", "Unit": "QPI LL" }, @@ -44,7 +44,7 @@ "EventCode": "0x13", "EventName": "UNC_Q_DIRECT2CORE.FAILURE_CREDITS_RBT", "PerPkg": "1", - "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because there were not enough Egress credits AND the RBT bit was not set, but the RBT tag matched.", + "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exclusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because there were not enough Egress credits AND the RBT bit was not set, but the RBT tag matched.", "UMask": "0x8", "Unit": "QPI LL" }, @@ -54,7 +54,7 @@ "EventCode": "0x13", "EventName": "UNC_Q_DIRECT2CORE.FAILURE_CREDITS_RBT_MISS", "PerPkg": "1", - "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because the RBT tag did not match, the valid bit was not set and there weren't enough Egress credits.", + "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exclusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because the RBT tag did not match, the valid bit was not set and there weren't enough Egress credits.", "UMask": "0x80", "Unit": "QPI LL" }, @@ -64,7 +64,7 @@ "EventCode": "0x13", "EventName": "UNC_Q_DIRECT2CORE.FAILURE_MISS", "PerPkg": "1", - "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because the RBT tag did not match although the valid bit was set and there were enough Egress credits.", + "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exclusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because the RBT tag did not match although the valid bit was set and there were enough Egress credits.", "UMask": "0x10", "Unit": "QPI LL" }, @@ -74,7 +74,7 @@ "EventCode": "0x13", "EventName": "UNC_Q_DIRECT2CORE.FAILURE_RBT_HIT", "PerPkg": "1", - "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because the route-back table (RBT) specified that the transaction should not trigger a direct2core tranaction. This is common for IO transactions. There were enough Egress credits and the RBT tag matched but the valid bit was not set.", + "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exclusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because the route-back table (RBT) specified that the transaction should not trigger a direct2core transaction. This is common for IO transactions. There were enough Egress credits and the RBT tag matched but the valid bit was not set.", "UMask": "0x4", "Unit": "QPI LL" }, @@ -84,7 +84,7 @@ "EventCode": "0x13", "EventName": "UNC_Q_DIRECT2CORE.FAILURE_RBT_MISS", "PerPkg": "1", - "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because the RBT tag did not match and the valid bit was not set although there were enough Egress credits.", + "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exclusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn failed because the RBT tag did not match and the valid bit was not set although there were enough Egress credits.", "UMask": "0x40", "Unit": "QPI LL" }, @@ -94,7 +94,7 @@ "EventCode": "0x13", "EventName": "UNC_Q_DIRECT2CORE.SUCCESS_RBT_HIT", "PerPkg": "1", - "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn was successful. There were sufficient credits, the RBT valid bit was set and there was an RBT tag match. The message was marked to spawn direct2core.", + "PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exclusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.; The spawn was successful. There were sufficient credits, the RBT valid bit was set and there was an RBT tag match. The message was marked to spawn direct2core.", "UMask": "0x1", "Unit": "QPI LL" }, @@ -131,7 +131,7 @@ "EventCode": "0x9", "EventName": "UNC_Q_RxL_BYPASSED", "PerPkg": "1", - "PublicDescription": "Counts the number of times that an incoming flit was able to bypass the flit buffer and pass directly across the BGF and into the Egress. This is a latency optimization, and should generally be the common case. If this value is less than the number of flits transfered, it implies that there was queueing getting onto the ring, and thus the transactions saw higher latency.", + "PublicDescription": "Counts the number of times that an incoming flit was able to bypass the flit buffer and pass directly across the BGF and into the Egress. This is a latency optimization, and should generally be the common case. If this value is less than the number of flits transferred, it implies that there was queueing getting onto the ring, and thus the transactions saw higher latency.", "Unit": "QPI LL" }, { @@ -443,7 +443,7 @@ "EventCode": "0x1", "EventName": "UNC_Q_RxL_FLITS_G0.DATA", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.; Number of data flitsreceived over QPI. Each flit contains 64b of data. This includes both DRS and NCB data flits (coherent and non-coherent). This can be used to calculate the data bandwidth of the QPI link. One can get a good picture of the QPI-link characteristics by evaluating the protocol flits, data flits, and idle/null flits. This does not include the header flits that go in data packets.", + "PublicDescription": "Counts the number of flits received from the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.; Number of data flits received over QPI. Each flit contains 64b of data. This includes both DRS and NCB data flits (coherent and non-coherent). This can be used to calculate the data bandwidth of the QPI link. One can get a good picture of the QPI-link characteristics by evaluating the protocol flits, data flits, and idle/null flits. This does not include the header flits that go in data packets.", "UMask": "0x2", "Unit": "QPI LL" }, @@ -453,7 +453,7 @@ "EventCode": "0x1", "EventName": "UNC_Q_RxL_FLITS_G0.IDLE", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.; Number of flits received over QPI that do not hold protocol payload. When QPI is not in a power saving state, it continuously transmits flits across the link. When there are no protocol flits to send, it will send IDLE and NULL flits across. These flits sometimes do carry a payload, such as credit returns, but are generall not considered part of the QPI bandwidth.", + "PublicDescription": "Counts the number of flits received from the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.; Number of flits received over QPI that do not hold protocol payload. When QPI is not in a power saving state, it continuously transmits flits across the link. When there are no protocol flits to send, it will send IDLE and NULL flits across. These flits sometimes do carry a payload, such as credit returns, but are generally not considered part of the QPI bandwidth.", "UMask": "0x1", "Unit": "QPI LL" }, @@ -463,7 +463,7 @@ "EventCode": "0x1", "EventName": "UNC_Q_RxL_FLITS_G0.NON_DATA", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.; Number of non-NULL non-data flits received across QPI. This basically tracks the protocol overhead on the QPI link. One can get a good picture of the QPI-link characteristics by evaluating the protocol flits, data flits, and idle/null flits. This includes the header flits for data packets.", + "PublicDescription": "Counts the number of flits received from the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.; Number of non-NULL non-data flits received across QPI. This basically tracks the protocol overhead on the QPI link. One can get a good picture of the QPI-link characteristics by evaluating the protocol flits, data flits, and idle/null flits. This includes the header flits for data packets.", "UMask": "0x4", "Unit": "QPI LL" }, @@ -474,7 +474,7 @@ "EventName": "UNC_Q_RxL_FLITS_G1.DRS", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits received over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency. This does not count data flits received over the NCB channel which transmits non-coherent data.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits received over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency. This does not count data flits received over the NCB channel which transmits non-coherent data.", "UMask": "0x18", "Unit": "QPI LL" }, @@ -485,7 +485,7 @@ "EventName": "UNC_Q_RxL_FLITS_G1.DRS_DATA", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of data flits received over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency. This does not count data flits received over the NCB channel which transmits non-coherent data. This includes only the data flits (not the header).", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of data flits received over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency. This does not count data flits received over the NCB channel which transmits non-coherent data. This includes only the data flits (not the header).", "UMask": "0x8", "Unit": "QPI LL" }, @@ -496,7 +496,7 @@ "EventName": "UNC_Q_RxL_FLITS_G1.DRS_NONDATA", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of protocol flits received over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency. This does not count data flits received over the NCB channel which transmits non-coherent data. This includes only the header flits (not the data). This includes extended headers.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of protocol flits received over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency. This does not count data flits received over the NCB channel which transmits non-coherent data. This includes only the header flits (not the data). This includes extended headers.", "UMask": "0x10", "Unit": "QPI LL" }, @@ -507,7 +507,7 @@ "EventName": "UNC_Q_RxL_FLITS_G1.HOM", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of flits received over QPI on the home channel.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of flits received over QPI on the home channel.", "UMask": "0x6", "Unit": "QPI LL" }, @@ -518,7 +518,7 @@ "EventName": "UNC_Q_RxL_FLITS_G1.HOM_NONREQ", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of non-request flits received over QPI on the home channel. These are most commonly snoop responses, and this event can be used as a proxy for that.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of non-request flits received over QPI on the home channel. These are most commonly snoop responses, and this event can be used as a proxy for that.", "UMask": "0x4", "Unit": "QPI LL" }, @@ -529,7 +529,7 @@ "EventName": "UNC_Q_RxL_FLITS_G1.HOM_REQ", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of data request received over QPI on the home channel. This basically counts the number of remote memory requests received over QPI. In conjunction with the local read count in the Home Agent, one can calculate the number of LLC Misses.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of data request received over QPI on the home channel. This basically counts the number of remote memory requests received over QPI. In conjunction with the local read count in the Home Agent, one can calculate the number of LLC Misses.", "UMask": "0x2", "Unit": "QPI LL" }, @@ -540,7 +540,7 @@ "EventName": "UNC_Q_RxL_FLITS_G1.SNP", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of snoop request flits received over QPI. These requests are contained in the snoop channel. This does not include snoop responses, which are received on the home channel.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of snoop request flits received over QPI. These requests are contained in the snoop channel. This does not include snoop responses, which are received on the home channel.", "UMask": "0x1", "Unit": "QPI LL" }, @@ -551,7 +551,7 @@ "EventName": "UNC_Q_RxL_FLITS_G2.NCB", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass flits. These packets are generally used to transmit non-coherent data across QPI.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass flits. These packets are generally used to transmit non-coherent data across QPI.", "UMask": "0xC", "Unit": "QPI LL" }, @@ -562,7 +562,7 @@ "EventName": "UNC_Q_RxL_FLITS_G2.NCB_DATA", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass data flits. These flits are generally used to transmit non-coherent data across QPI. This does not include a count of the DRS (coherent) data flits. This only counts the data flits, not the NCB headers.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass data flits. These flits are generally used to transmit non-coherent data across QPI. This does not include a count of the DRS (coherent) data flits. This only counts the data flits, not the NCB headers.", "UMask": "0x4", "Unit": "QPI LL" }, @@ -573,7 +573,7 @@ "EventName": "UNC_Q_RxL_FLITS_G2.NCB_NONDATA", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass non-data flits. These packets are generally used to transmit non-coherent data across QPI, and the flits counted here are for headers and other non-data flits. This includes extended headers.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass non-data flits. These packets are generally used to transmit non-coherent data across QPI, and the flits counted here are for headers and other non-data flits. This includes extended headers.", "UMask": "0x8", "Unit": "QPI LL" }, @@ -584,7 +584,7 @@ "EventName": "UNC_Q_RxL_FLITS_G2.NCS", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of NCS (non-coherent standard) flits received over QPI. This includes extended headers.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of NCS (non-coherent standard) flits received over QPI. This includes extended headers.", "UMask": "0x10", "Unit": "QPI LL" }, @@ -595,7 +595,7 @@ "EventName": "UNC_Q_RxL_FLITS_G2.NDR_AD", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits received over the NDR (Non-Data Response) channel. This channel is used to send a variety of protocol flits including grants and completions. This is only for NDR packets to the local socket which use the AK ring.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits received over the NDR (Non-Data Response) channel. This channel is used to send a variety of protocol flits including grants and completions. This is only for NDR packets to the local socket which use the AK ring.", "UMask": "0x1", "Unit": "QPI LL" }, @@ -606,7 +606,7 @@ "EventName": "UNC_Q_RxL_FLITS_G2.NDR_AK", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits received over the NDR (Non-Data Response) channel. This channel is used to send a variety of protocol flits including grants and completions. This is only for NDR packets destined for Route-thru to a remote socket.", + "PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits received over the NDR (Non-Data Response) channel. This channel is used to send a variety of protocol flits including grants and completions. This is only for NDR packets destined for Route-thru to a remote socket.", "UMask": "0x2", "Unit": "QPI LL" }, @@ -1227,7 +1227,7 @@ "Counter": "0,1,2,3", "EventName": "UNC_Q_TxL_FLITS_G0.DATA", "PerPkg": "1", - "PublicDescription": "Counts the number of flits transmitted across the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.; Number of data flits transmitted over QPI. Each flit contains 64b of data. This includes both DRS and NCB data flits (coherent and non-coherent). This can be used to calculate the data bandwidth of the QPI link. One can get a good picture of the QPI-link characteristics by evaluating the protocol flits, data flits, and idle/null flits. This does not include the header flits that go in data packets.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.; Number of data flits transmitted over QPI. Each flit contains 64b of data. This includes both DRS and NCB data flits (coherent and non-coherent). This can be used to calculate the data bandwidth of the QPI link. One can get a good picture of the QPI-link characteristics by evaluating the protocol flits, data flits, and idle/null flits. This does not include the header flits that go in data packets.", "UMask": "0x2", "Unit": "QPI LL" }, @@ -1236,7 +1236,7 @@ "Counter": "0,1,2,3", "EventName": "UNC_Q_TxL_FLITS_G0.NON_DATA", "PerPkg": "1", - "PublicDescription": "Counts the number of flits transmitted across the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.; Number of non-NULL non-data flits transmitted across QPI. This basically tracks the protocol overhead on the QPI link. One can get a good picture of the QPI-link characteristics by evaluating the protocol flits, data flits, and idle/null flits. This includes the header flits for data packets.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.; Number of non-NULL non-data flits transmitted across QPI. This basically tracks the protocol overhead on the QPI link. One can get a good picture of the QPI-link characteristics by evaluating the protocol flits, data flits, and idle/null flits. This includes the header flits for data packets.", "UMask": "0x4", "Unit": "QPI LL" }, @@ -1246,7 +1246,7 @@ "EventName": "UNC_Q_TxL_FLITS_G1.DRS", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits transmitted over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits transmitted over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency.", "UMask": "0x18", "Unit": "QPI LL" }, @@ -1256,7 +1256,7 @@ "EventName": "UNC_Q_TxL_FLITS_G1.DRS_DATA", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of data flits transmitted over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency. This does not count data flits transmitted over the NCB channel which transmits non-coherent data. This includes only the data flits (not the header).", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of data flits transmitted over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency. This does not count data flits transmitted over the NCB channel which transmits non-coherent data. This includes only the data flits (not the header).", "UMask": "0x8", "Unit": "QPI LL" }, @@ -1266,7 +1266,7 @@ "EventName": "UNC_Q_TxL_FLITS_G1.DRS_NONDATA", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of protocol flits transmitted over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency. This does not count data flits transmitted over the NCB channel which transmits non-coherent data. This includes only the header flits (not the data). This includes extended headers.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of protocol flits transmitted over QPI on the DRS (Data Response) channel. DRS flits are used to transmit data with coherency. This does not count data flits transmitted over the NCB channel which transmits non-coherent data. This includes only the header flits (not the data). This includes extended headers.", "UMask": "0x10", "Unit": "QPI LL" }, @@ -1276,7 +1276,7 @@ "EventName": "UNC_Q_TxL_FLITS_G1.HOM", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of flits transmitted over QPI on the home channel.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of flits transmitted over QPI on the home channel.", "UMask": "0x6", "Unit": "QPI LL" }, @@ -1286,7 +1286,7 @@ "EventName": "UNC_Q_TxL_FLITS_G1.HOM_NONREQ", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of non-request flits transmitted over QPI on the home channel. These are most commonly snoop responses, and this event can be used as a proxy for that.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of non-request flits transmitted over QPI on the home channel. These are most commonly snoop responses, and this event can be used as a proxy for that.", "UMask": "0x4", "Unit": "QPI LL" }, @@ -1296,7 +1296,7 @@ "EventName": "UNC_Q_TxL_FLITS_G1.HOM_REQ", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of data request transmitted over QPI on the home channel. This basically counts the number of remote memory requests transmitted over QPI. In conjunction with the local read count in the Home Agent, one can calculate the number of LLC Misses.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of data request transmitted over QPI on the home channel. This basically counts the number of remote memory requests transmitted over QPI. In conjunction with the local read count in the Home Agent, one can calculate the number of LLC Misses.", "UMask": "0x2", "Unit": "QPI LL" }, @@ -1306,7 +1306,7 @@ "EventName": "UNC_Q_TxL_FLITS_G1.SNP", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of snoop request flits transmitted over QPI. These requests are contained in the snoop channel. This does not include snoop responses, which are transmitted on the home channel.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the number of snoop request flits transmitted over QPI. These requests are contained in the snoop channel. This does not include snoop responses, which are transmitted on the home channel.", "UMask": "0x1", "Unit": "QPI LL" }, @@ -1317,7 +1317,7 @@ "EventName": "UNC_Q_TxL_FLITS_G2.NCB", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass flits. These packets are generally used to transmit non-coherent data across QPI.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass flits. These packets are generally used to transmit non-coherent data across QPI.", "UMask": "0xC", "Unit": "QPI LL" }, @@ -1328,7 +1328,7 @@ "EventName": "UNC_Q_TxL_FLITS_G2.NCB_DATA", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass data flits. These flits are generally used to transmit non-coherent data across QPI. This does not include a count of the DRS (coherent) data flits. This only counts the data flits, not te NCB headers.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass data flits. These flits are generally used to transmit non-coherent data across QPI. This does not include a count of the DRS (coherent) data flits. This only counts the data flits, not the NCB headers.", "UMask": "0x4", "Unit": "QPI LL" }, @@ -1339,7 +1339,7 @@ "EventName": "UNC_Q_TxL_FLITS_G2.NCB_NONDATA", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass non-data flits. These packets are generally used to transmit non-coherent data across QPI, and the flits counted here are for headers and other non-data flits. This includes extended headers.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of Non-Coherent Bypass non-data flits. These packets are generally used to transmit non-coherent data across QPI, and the flits counted here are for headers and other non-data flits. This includes extended headers.", "UMask": "0x8", "Unit": "QPI LL" }, @@ -1350,7 +1350,7 @@ "EventName": "UNC_Q_TxL_FLITS_G2.NCS", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of NCS (non-coherent standard) flits transmitted over QPI. This includes extended headers.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Number of NCS (non-coherent standard) flits transmitted over QPI. This includes extended headers.", "UMask": "0x10", "Unit": "QPI LL" }, @@ -1361,7 +1361,7 @@ "EventName": "UNC_Q_TxL_FLITS_G2.NDR_AD", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits transmitted over the NDR (Non-Data Response) channel. This channel is used to send a variety of protocol flits including grants and completions. This is only for NDR packets to the local socket which use the AK ring.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits transmitted over the NDR (Non-Data Response) channel. This channel is used to send a variety of protocol flits including grants and completions. This is only for NDR packets to the local socket which use the AK ring.", "UMask": "0x1", "Unit": "QPI LL" }, @@ -1372,7 +1372,7 @@ "EventName": "UNC_Q_TxL_FLITS_G2.NDR_AK", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits transmitted over the NDR (Non-Data Response) channel. This channel is used to send a variety of protocol flits including grants and completions. This is only for NDR packets destined for Route-thru to a remote socket.", + "PublicDescription": "Counts the number of flits transmitted across the QPI Link. This is one of three groups that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each flit is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four fits, each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI speed (for example, 8.0 GT/s), the transfers here refer to fits. Therefore, in L0, the system will transfer 1 flit at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as data bandwidth. For example, when we are transferring a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual data and an additional 16 bits of other information. To calculate data bandwidth, one should therefore do: data flits * 8B / time.; Counts the total number of flits transmitted over the NDR (Non-Data Response) channel. This channel is used to send a variety of protocol flits including grants and completions. This is only for NDR packets destined for Route-thru to a remote socket.", "UMask": "0x2", "Unit": "QPI LL" }, @@ -1511,7 +1511,7 @@ "EventName": "UNC_Q_TxR_AD_SNP_CREDIT_OCCUPANCY.VN0", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Occupancy event that tracks the number of link layer credits into the R3 (for transactions across the BGF) available in each cycle. Flow Control FIFO fro Snoop messages on AD.", + "PublicDescription": "Occupancy event that tracks the number of link layer credits into the R3 (for transactions across the BGF) available in each cycle. Flow Control FIFO for Snoop messages on AD.", "UMask": "0x1", "Unit": "QPI LL" }, @@ -1522,7 +1522,7 @@ "EventName": "UNC_Q_TxR_AD_SNP_CREDIT_OCCUPANCY.VN1", "ExtSel": "1", "PerPkg": "1", - "PublicDescription": "Occupancy event that tracks the number of link layer credits into the R3 (for transactions across the BGF) available in each cycle. Flow Control FIFO fro Snoop messages on AD.", + "PublicDescription": "Occupancy event that tracks the number of link layer credits into the R3 (for transactions across the BGF) available in each cycle. Flow Control FIFO for Snoop messages on AD.", "UMask": "0x2", "Unit": "QPI LL" }, diff --git a/tools/perf/pmu-events/arch/x86/ivytown/uncore-memory.json b/tools/perf/pmu-events/arch/x86/ivytown/uncore-memory.json index 63b49b712c62..ed60ebca35cb 100644 --- a/tools/perf/pmu-events/arch/x86/ivytown/uncore-memory.json +++ b/tools/perf/pmu-events/arch/x86/ivytown/uncore-memory.json @@ -188,7 +188,7 @@ "EventCode": "0x9", "EventName": "UNC_M_ECC_CORRECTABLE_ERRORS", "PerPkg": "1", - "PublicDescription": "Counts the number of ECC errors detected and corrected by the iMC on this channel. This counter is only useful with ECC DRAM devices. This count will increment one time for each correction regardless of the number of bits corrected. The iMC can correct up to 4 bit errors in independent channel mode and 8 bit erros in lockstep mode.", + "PublicDescription": "Counts the number of ECC errors detected and corrected by the iMC on this channel. This counter is only useful with ECC DRAM devices. This count will increment one time for each correction regardless of the number of bits corrected. The iMC can correct up to 4 bit errors in independent channel mode and 8 bit errors in lockstep mode.", "Unit": "iMC" }, { diff --git a/tools/perf/pmu-events/arch/x86/ivytown/uncore-other.json b/tools/perf/pmu-events/arch/x86/ivytown/uncore-other.json index af289aa6c98e..6c7ddf642fc3 100644 --- a/tools/perf/pmu-events/arch/x86/ivytown/uncore-other.json +++ b/tools/perf/pmu-events/arch/x86/ivytown/uncore-other.json @@ -2097,7 +2097,7 @@ "EventCode": "0x33", "EventName": "UNC_R3_VNA_CREDITS_ACQUIRED", "PerPkg": "1", - "PublicDescription": "Number of QPI VNA Credit acquisitions. This event can be used in conjunction with the VNA In-Use Accumulator to calculate the average lifetime of a credit holder. VNA credits are used by all message classes in order to communicate across QPI. If a packet is unable to acquire credits, it will then attempt to use credts from the VN0 pool. Note that a single packet may require multiple flit buffers (i.e. when data is being transfered). Therefore, this event will increment by the number of credits acquired in each cycle. Filtering based on message class is not provided. One can count the number of packets transfered in a given message class using an qfclk event.", + "PublicDescription": "Number of QPI VNA Credit acquisitions. This event can be used in conjunction with the VNA In-Use Accumulator to calculate the average lifetime of a credit holder. VNA credits are used by all message classes in order to communicate across QPI. If a packet is unable to acquire credits, it will then attempt to use credits from the VN0 pool. Note that a single packet may require multiple flit buffers (i.e. when data is being transferred). Therefore, this event will increment by the number of credits acquired in each cycle. Filtering based on message class is not provided. One can count the number of packets transferred in a given message class using an qfclk event.", "Unit": "R3QPI" }, { @@ -2106,7 +2106,7 @@ "EventCode": "0x33", "EventName": "UNC_R3_VNA_CREDITS_ACQUIRED.AD", "PerPkg": "1", - "PublicDescription": "Number of QPI VNA Credit acquisitions. This event can be used in conjunction with the VNA In-Use Accumulator to calculate the average lifetime of a credit holder. VNA credits are used by all message classes in order to communicate across QPI. If a packet is unable to acquire credits, it will then attempt to use credts from the VN0 pool. Note that a single packet may require multiple flit buffers (i.e. when data is being transfered). Therefore, this event will increment by the number of credits acquired in each cycle. Filtering based on message class is not provided. One can count the number of packets transfered in a given message class using an qfclk event.; Filter for the Home (HOM) message class. HOM is generally used to send requests, request responses, and snoop responses.", + "PublicDescription": "Number of QPI VNA Credit acquisitions. This event can be used in conjunction with the VNA In-Use Accumulator to calculate the average lifetime of a credit holder. VNA credits are used by all message classes in order to communicate across QPI. If a packet is unable to acquire credits, it will then attempt to use credits from the VN0 pool. Note that a single packet may require multiple flit buffers (i.e. when data is being transferred). Therefore, this event will increment by the number of credits acquired in each cycle. Filtering based on message class is not provided. One can count the number of packets transferred in a given message class using an qfclk event.; Filter for the Home (HOM) message class. HOM is generally used to send requests, request responses, and snoop responses.", "UMask": "0x1", "Unit": "R3QPI" }, @@ -2116,7 +2116,7 @@ "EventCode": "0x33", "EventName": "UNC_R3_VNA_CREDITS_ACQUIRED.BL", "PerPkg": "1", - "PublicDescription": "Number of QPI VNA Credit acquisitions. This event can be used in conjunction with the VNA In-Use Accumulator to calculate the average lifetime of a credit holder. VNA credits are used by all message classes in order to communicate across QPI. If a packet is unable to acquire credits, it will then attempt to use credts from the VN0 pool. Note that a single packet may require multiple flit buffers (i.e. when data is being transfered). Therefore, this event will increment by the number of credits acquired in each cycle. Filtering based on message class is not provided. One can count the number of packets transfered in a given message class using an qfclk event.; Filter for the Home (HOM) message class. HOM is generally used to send requests, request responses, and snoop responses.", + "PublicDescription": "Number of QPI VNA Credit acquisitions. This event can be used in conjunction with the VNA In-Use Accumulator to calculate the average lifetime of a credit holder. VNA credits are used by all message classes in order to communicate across QPI. If a packet is unable to acquire credits, it will then attempt to use credits from the VN0 pool. Note that a single packet may require multiple flit buffers (i.e. when data is being transferred). Therefore, this event will increment by the number of credits acquired in each cycle. Filtering based on message class is not provided. One can count the number of packets transferred in a given message class using an qfclk event.; Filter for the Home (HOM) message class. HOM is generally used to send requests, request responses, and snoop responses.", "UMask": "0x4", "Unit": "R3QPI" }, diff --git a/tools/perf/pmu-events/arch/x86/ivytown/uncore-power.json b/tools/perf/pmu-events/arch/x86/ivytown/uncore-power.json index 0ba63a97ddfa..74c87217d75c 100644 --- a/tools/perf/pmu-events/arch/x86/ivytown/uncore-power.json +++ b/tools/perf/pmu-events/arch/x86/ivytown/uncore-power.json @@ -601,7 +601,7 @@ "EventCode": "0x80", "EventName": "UNC_P_POWER_STATE_OCCUPANCY.CORES_C0", "PerPkg": "1", - "PublicDescription": "This is an occupancy event that tracks the number of cores that are in the chosen C-State. It can be used by itself to get the average number of cores in that C-state with threshholding to generate histograms, or with other PCU events and occupancy triggering to capture other details.", + "PublicDescription": "This is an occupancy event that tracks the number of cores that are in the chosen C-State. It can be used by itself to get the average number of cores in that C-state with thresholding to generate histograms, or with other PCU events and occupancy triggering to capture other details.", "Unit": "PCU" }, { @@ -610,7 +610,7 @@ "EventCode": "0x80", "EventName": "UNC_P_POWER_STATE_OCCUPANCY.CORES_C3", "PerPkg": "1", - "PublicDescription": "This is an occupancy event that tracks the number of cores that are in the chosen C-State. It can be used by itself to get the average number of cores in that C-state with threshholding to generate histograms, or with other PCU events and occupancy triggering to capture other details.", + "PublicDescription": "This is an occupancy event that tracks the number of cores that are in the chosen C-State. It can be used by itself to get the average number of cores in that C-state with thresholding to generate histograms, or with other PCU events and occupancy triggering to capture other details.", "Unit": "PCU" }, { @@ -619,7 +619,7 @@ "EventCode": "0x80", "EventName": "UNC_P_POWER_STATE_OCCUPANCY.CORES_C6", "PerPkg": "1", - "PublicDescription": "This is an occupancy event that tracks the number of cores that are in the chosen C-State. It can be used by itself to get the average number of cores in that C-state with threshholding to generate histograms, or with other PCU events and occupancy triggering to capture other details.", + "PublicDescription": "This is an occupancy event that tracks the number of cores that are in the chosen C-State. It can be used by itself to get the average number of cores in that C-state with thresholding to generate histograms, or with other PCU events and occupancy triggering to capture other details.", "Unit": "PCU" }, { @@ -637,7 +637,7 @@ "EventCode": "0x9", "EventName": "UNC_P_PROCHOT_INTERNAL_CYCLES", "PerPkg": "1", - "PublicDescription": "Counts the number of cycles that we are in Interal PROCHOT mode. This mode is triggered when a sensor on the die determines that we are too hot and must throttle to avoid damaging the chip.", + "PublicDescription": "Counts the number of cycles that we are in Internal PROCHOT mode. This mode is triggered when a sensor on the die determines that we are too hot and must throttle to avoid damaging the chip.", "Unit": "PCU" }, { diff --git a/tools/perf/pmu-events/arch/x86/jaketown/jkt-metrics.json b/tools/perf/pmu-events/arch/x86/jaketown/jkt-metrics.json index c0fbb4f31241..554f87c03c05 100644 --- a/tools/perf/pmu-events/arch/x86/jaketown/jkt-metrics.json +++ b/tools/perf/pmu-events/arch/x86/jaketown/jkt-metrics.json @@ -1,64 +1,247 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * min(CPU_CLK_UNHALTED.THREAD, IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE) / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: RS_EVENTS.EMPTY_END", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "(12 * ITLB_MISSES.STLB_HIT + ITLB_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: ITLB_MISSES.WALK_COMPLETED", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "12 * (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY) / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "3 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - (tma_frontend_bound + tma_bad_speculation + tma_retiring)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_L1D_PENDING) + RESOURCE_STALLS.SB) / (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_DISPATCH) + cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=1@ - cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=2@ - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "(7 * DTLB_LOAD_MISSES.STLB_HIT + DTLB_LOAD_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_UOPS_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEM_LOAD_UOPS_RETIRED.LLC_HIT / (MEM_LOAD_UOPS_RETIRED.LLC_HIT + 7 * MEM_LOAD_UOPS_RETIRED.LLC_MISS)) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(1 - (MEM_LOAD_UOPS_RETIRED.LLC_HIT / (MEM_LOAD_UOPS_RETIRED.LLC_HIT + 7 * MEM_LOAD_UOPS_RETIRED.LLC_MISS))) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=6@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "RESOURCE_STALLS.SB / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_UOPS_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.FPU_DIV_ACTIVE / CORE_CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_UOPS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_DISPATCH) + cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=1@ - cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=2@ - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) - RESOURCE_STALLS.SB - min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_L1D_PENDING)) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS * FP_COMP_OPS_EXE.X87 / UOPS_DISPATCHED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) / UOPS_DISPATCHED.THREAD", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) / UOPS_DISPATCHED.THREAD", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" }, { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -70,8 +253,8 @@ }, { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI" }, { @@ -82,17 +265,11 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "The ratio of Executed- by Issued-Uops", "MetricExpr": "UOPS_DISPATCHED.THREAD / UOPS_ISSUED.ANY", "MetricGroup": "Cor;Pipeline", @@ -101,44 +278,32 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * (FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) + 8 * SIMD_FP_256.PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { - "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;Flops_SMT", - "MetricName": "FLOPc_SMT" - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_DISPATCHED.THREAD / (( cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@)", + "MetricExpr": "UOPS_DISPATCHED.THREAD / ((cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@ / 2) if #SMT_on else cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -149,7 +314,7 @@ }, { "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)", - "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )", + "MetricExpr": "IDQ.DSB_UOPS / ((IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS))", "MetricGroup": "DSB;Fed;FetchBW", "MetricName": "DSB_Coverage" }, @@ -161,26 +326,26 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * (FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) + 8 * SIMD_FP_256.PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -198,7 +363,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "( 64 * ( uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@ ) / 1000000000 ) / duration_time", + "MetricExpr": "(64 * (uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@) / 1000000000) / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, @@ -209,12 +374,6 @@ "MetricName": "Socket_CLKS" }, { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "cbox_0@event\\=0x0@ / #num_dies / duration_time / 1000000000", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, - { "BriefDescription": "Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]", "MetricExpr": "INST_RETIRED.ANY / BR_INST_RETIRED.FAR_BRANCH:u", "MetricGroup": "Branches;OS", @@ -261,5 +420,11 @@ "MetricExpr": "(cstate_pkg@c7\\-residency@ / msr@tsc@) * 100", "MetricGroup": "Power", "MetricName": "C7_Pkg_Residency" + }, + { + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "Socket_CLKS / #num_dies / duration_time / 1000000000", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ" } ] diff --git a/tools/perf/pmu-events/arch/x86/mapfile.csv b/tools/perf/pmu-events/arch/x86/mapfile.csv index 7f2d777fd97f..5e609b876790 100644 --- a/tools/perf/pmu-events/arch/x86/mapfile.csv +++ b/tools/perf/pmu-events/arch/x86/mapfile.csv @@ -1,27 +1,27 @@ Family-model,Version,Filename,EventType -GenuineIntel-6-9[7A],v1.13,alderlake,core +GenuineIntel-6-(97|9A|B7|BA|BE|BF),v1.15,alderlake,core GenuineIntel-6-(1C|26|27|35|36),v4,bonnell,core GenuineIntel-6-(3D|47),v26,broadwell,core GenuineIntel-6-56,v23,broadwellde,core GenuineIntel-6-4F,v19,broadwellx,core GenuineIntel-6-55-[56789ABCDEF],v1.16,cascadelakex,core -GenuineIntel-6-96,v1.03,elkhartlake,core +GenuineIntel-6-9[6C],v1.03,elkhartlake,core GenuineIntel-6-5[CF],v13,goldmont,core GenuineIntel-6-7A,v1.01,goldmontplus,core -GenuineIntel-6-(3C|45|46),v31,haswell,core -GenuineIntel-6-3F,v25,haswellx,core -GenuineIntel-6-(7D|7E|A7),v1.14,icelake,core -GenuineIntel-6-6[AC],v1.15,icelakex,core +GenuineIntel-6-(3C|45|46),v32,haswell,core +GenuineIntel-6-3F,v26,haswellx,core +GenuineIntel-6-(7D|7E|A7),v1.15,icelake,core +GenuineIntel-6-6[AC],v1.16,icelakex,core GenuineIntel-6-3A,v22,ivybridge,core -GenuineIntel-6-3E,v21,ivytown,core +GenuineIntel-6-3E,v22,ivytown,core GenuineIntel-6-2D,v21,jaketown,core GenuineIntel-6-(57|85),v9,knightslanding,core GenuineIntel-6-AA,v1.00,meteorlake,core GenuineIntel-6-1[AEF],v3,nehalemep,core GenuineIntel-6-2E,v3,nehalemex,core GenuineIntel-6-2A,v17,sandybridge,core -GenuineIntel-6-8F,v1.04,sapphirerapids,core -GenuineIntel-6-(37|4C|4D),v14,silvermont,core +GenuineIntel-6-8F,v1.06,sapphirerapids,core +GenuineIntel-6-(37|4A|4C|4D|5A),v14,silvermont,core GenuineIntel-6-(4E|5E|8E|9E|A5|A6),v53,skylake,core GenuineIntel-6-55-[01234],v1.28,skylakex,core GenuineIntel-6-86,v1.20,snowridgex,core diff --git a/tools/perf/pmu-events/arch/x86/sandybridge/snb-metrics.json b/tools/perf/pmu-events/arch/x86/sandybridge/snb-metrics.json index ae7ed267b2a2..5d5a6d6f3bda 100644 --- a/tools/perf/pmu-events/arch/x86/sandybridge/snb-metrics.json +++ b/tools/perf/pmu-events/arch/x86/sandybridge/snb-metrics.json @@ -1,64 +1,247 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * min(CPU_CLK_UNHALTED.THREAD, IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE) / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: RS_EVENTS.EMPTY_END", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "(12 * ITLB_MISSES.STLB_HIT + ITLB_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: ITLB_MISSES.WALK_COMPLETED", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "12 * (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY) / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "3 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - (tma_frontend_bound + tma_bad_speculation + tma_retiring)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_L1D_PENDING) + RESOURCE_STALLS.SB) / (min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_DISPATCH) + cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=1@ - cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=2@ - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "(7 * DTLB_LOAD_MISSES.STLB_HIT + DTLB_LOAD_MISSES.WALK_DURATION) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_UOPS_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEM_LOAD_UOPS_RETIRED.LLC_HIT / (MEM_LOAD_UOPS_RETIRED.LLC_HIT + 7 * MEM_LOAD_UOPS_MISC_RETIRED.LLC_MISS)) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(1 - (MEM_LOAD_UOPS_RETIRED.LLC_HIT / (MEM_LOAD_UOPS_RETIRED.LLC_HIT + 7 * MEM_LOAD_UOPS_MISC_RETIRED.LLC_MISS))) * CYCLE_ACTIVITY.STALLS_L2_PENDING / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_UOPS_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=6@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "RESOURCE_STALLS.SB / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_UOPS_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.FPU_DIV_ACTIVE / CORE_CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "((min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.CYCLES_NO_DISPATCH) + cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=1@ - cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=3@ if (IPC > 1.8) else cpu@UOPS_DISPATCHED.THREAD\\,cmask\\=2@ - RS_EVENTS.EMPTY_CYCLES if (tma_fetch_latency > 0.1) else RESOURCE_STALLS.SB) - RESOURCE_STALLS.SB - min(CPU_CLK_UNHALTED.THREAD, CYCLE_ACTIVITY.STALLS_L1D_PENDING)) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS * FP_COMP_OPS_EXE.X87 / UOPS_DISPATCHED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) / UOPS_DISPATCHED.THREAD", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) / UOPS_DISPATCHED.THREAD", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" }, { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -70,8 +253,8 @@ }, { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI" }, { @@ -82,17 +265,11 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "The ratio of Executed- by Issued-Uops", "MetricExpr": "UOPS_DISPATCHED.THREAD / UOPS_ISSUED.ANY", "MetricGroup": "Cor;Pipeline", @@ -101,44 +278,32 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * (FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) + 8 * SIMD_FP_256.PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { - "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;Flops_SMT", - "MetricName": "FLOPc_SMT" - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_DISPATCHED.THREAD / (( cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@)", + "MetricExpr": "UOPS_DISPATCHED.THREAD / ((cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@ / 2) if #SMT_on else cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -149,7 +314,7 @@ }, { "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)", - "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )", + "MetricExpr": "IDQ.DSB_UOPS / ((IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS))", "MetricGroup": "DSB;Fed;FetchBW", "MetricName": "DSB_Coverage" }, @@ -161,26 +326,26 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * (FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE) + 8 * SIMD_FP_256.PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -198,7 +363,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * ( arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@ ) / 1000000 / duration_time / 1000", + "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1000000 / duration_time / 1000", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, diff --git a/tools/perf/pmu-events/arch/x86/sapphirerapids/cache.json b/tools/perf/pmu-events/arch/x86/sapphirerapids/cache.json index 348476ce8107..c05c741e22db 100644 --- a/tools/perf/pmu-events/arch/x86/sapphirerapids/cache.json +++ b/tools/perf/pmu-events/arch/x86/sapphirerapids/cache.json @@ -35,7 +35,7 @@ "UMask": "0x2" }, { - "BriefDescription": "Number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailablability.", + "BriefDescription": "Number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailability.", "CollectPEBSRecord": "2", "Counter": "0,1,2,3", "CounterMask": "1", @@ -43,7 +43,7 @@ "EventCode": "0x48", "EventName": "L1D_PEND_MISS.FB_FULL_PERIODS", "PEBScounters": "0,1,2,3", - "PublicDescription": "Counts number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailablability. Demand requests include cacheable/uncacheable demand load, store, lock or SW prefetch accesses.", + "PublicDescription": "Counts number of phases a demand request has waited due to L1D Fill Buffer (FB) unavailability. Demand requests include cacheable/uncacheable demand load, store, lock or SW prefetch accesses.", "SampleAfterValue": "1000003", "Speculative": "1", "UMask": "0x2" diff --git a/tools/perf/pmu-events/arch/x86/sapphirerapids/frontend.json b/tools/perf/pmu-events/arch/x86/sapphirerapids/frontend.json index 44ecf38ad970..ff0d47ce8e9a 100644 --- a/tools/perf/pmu-events/arch/x86/sapphirerapids/frontend.json +++ b/tools/perf/pmu-events/arch/x86/sapphirerapids/frontend.json @@ -12,6 +12,17 @@ "UMask": "0x1" }, { + "BriefDescription": "Cycles the Microcode Sequencer is busy.", + "CollectPEBSRecord": "2", + "Counter": "0,1,2,3", + "EventCode": "0x87", + "EventName": "DECODE.MS_BUSY", + "PEBScounters": "0,1,2,3", + "SampleAfterValue": "500009", + "Speculative": "1", + "UMask": "0x2" + }, + { "BriefDescription": "DSB-to-MITE switch true penalty cycles.", "CollectPEBSRecord": "2", "Counter": "0,1,2,3", diff --git a/tools/perf/pmu-events/arch/x86/sapphirerapids/pipeline.json b/tools/perf/pmu-events/arch/x86/sapphirerapids/pipeline.json index df4f3d714e6e..b2f0d9393d3c 100644 --- a/tools/perf/pmu-events/arch/x86/sapphirerapids/pipeline.json +++ b/tools/perf/pmu-events/arch/x86/sapphirerapids/pipeline.json @@ -80,10 +80,10 @@ "EventCode": "0xc1", "EventName": "ASSISTS.ANY", "PEBScounters": "0,1,2,3,4,5,6,7", - "PublicDescription": "Counts the number of occurrences where a microcode assist is invoked by hardware Examples include AD (page Access Dirty), FP and AVX related assists.", + "PublicDescription": "Counts the number of occurrences where a microcode assist is invoked by hardware. Examples include AD (page Access Dirty), FP and AVX related assists.", "SampleAfterValue": "100003", "Speculative": "1", - "UMask": "0x1f" + "UMask": "0x1b" }, { "BriefDescription": "All branch instructions retired.", diff --git a/tools/perf/pmu-events/arch/x86/sapphirerapids/spr-metrics.json b/tools/perf/pmu-events/arch/x86/sapphirerapids/spr-metrics.json index e194dfc5c25b..9ec42a68c160 100644 --- a/tools/perf/pmu-events/arch/x86/sapphirerapids/spr-metrics.json +++ b/tools/perf/pmu-events/arch/x86/sapphirerapids/spr-metrics.json @@ -1,17 +1,819 @@ [ { + "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", + "MetricExpr": "topdown\\-fe\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) - INT_MISC.UOP_DROPPING / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. Sample with: FRONTEND_RETIRED.LATENCY_GE_4_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "(topdown\\-fetch\\-lat / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) - INT_MISC.UOP_DROPPING / SLOTS)", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: FRONTEND_RETIRED.LATENCY_GE_16_PS;FRONTEND_RETIRED.LATENCY_GE_8_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses", + "MetricExpr": "ICACHE_DATA.STALLS / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses. Sample with: FRONTEND_RETIRED.L2_MISS_PS;FRONTEND_RETIRED.L1I_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "ICACHE_TAG.STALLS / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: FRONTEND_RETIRED.STLB_MISS_PS;FRONTEND_RETIRED.ITLB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "INT_MISC.CLEAR_RESTEER_CYCLES / CLKS + tma_unknown_branches", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage", + "MetricExpr": "(tma_branch_mispredicts / tma_bad_speculation) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears", + "MetricExpr": "(1 - (tma_branch_mispredicts / tma_bad_speculation)) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "INT_MISC.UNKNOWN_BRANCH_CYCLES / CLKS", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: FRONTEND_RETIRED.UNKNOWN_BRANCH", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty. Sample with: FRONTEND_RETIRED.DSB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "DECODE.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "3 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: FRONTEND_RETIRED.MS_FLOWS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "max(0, tma_frontend_bound - tma_fetch_latency)", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend. Sample with: FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_2_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.MITE_CYCLES_ANY - IDQ.MITE_CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck. Sample with: FRONTEND_RETIRED.ANY_DSB_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where decoder-0 was the only active decoder", + "MetricExpr": "(cpu@INST_DECODED.DECODERS\\,cmask\\=1@ - cpu@INST_DECODED.DECODERS\\,cmask\\=2@) / CORE_CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_decoder0_alone", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.DSB_CYCLES_ANY - IDQ.DSB_CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", + "MetricExpr": "max(1 - (tma_frontend_bound + tma_backend_bound + tma_retiring), 0)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "topdown\\-br\\-mispredict / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: TOPDOWN.BR_MISPREDICT_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "max(0, tma_bad_speculation - tma_branch_mispredicts)", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", + "MetricExpr": "topdown\\-be\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. Sample with: TOPDOWN.BACKEND_BOUND_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "topdown\\-mem\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((EXE_ACTIVITY.BOUND_ON_LOADS - MEMORY_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_RETIRED.L1_HIT_PS;MEM_LOAD_RETIRED.FB_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "min(7 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE, max(CYCLE_ACTIVITY.CYCLES_MEM_ANY - MEMORY_ACTIVITY.CYCLES_L1D_MISS, 0)) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_INST_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_load - tma_load_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk", + "MetricExpr": "DTLB_LOAD_MISSES.WALK_ACTIVE / CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(16 * max(0, MEM_INST_RETIRED.LOCK_LOADS - L2_RQSTS.ALL_RFO) + (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES) * (10 * L2_RQSTS.RFO_HIT + min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO))) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_INST_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_INST_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "L1D_PEND_MISS.FB_FULL / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "(MEMORY_ACTIVITY.STALLS_L1D_MISS - MEMORY_ACTIVITY.STALLS_L2_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(MEMORY_ACTIVITY.STALLS_L2_MISS - MEMORY_ACTIVITY.STALLS_L3_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "((25 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD * (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM / (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM + OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD))) + (24 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "(24 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD + MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD * (1 - (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM / (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM + OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD)))) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "(9 * Average_Frequency) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "(XQ.FULL_CYCLES + L1D_PEND_MISS.L2_STALLS) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "((MEMORY_ACTIVITY.STALLS_L3_MISS / CLKS) - tma_pmm_bound)", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to memory bandwidth Allocation feature (RDT's memory bandwidth throttling).", + "MetricExpr": "INT_MISC.MBA_STALLS / CLKS", + "MetricGroup": "MemoryBW;Offcore;Server;TopdownL5;tma_mem_bandwidth_group", + "MetricName": "tma_mba_stalls", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory", + "MetricExpr": "(54.5 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Server;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_local_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory. Caching will improve the latency and increase performance. Sample with: MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory", + "MetricExpr": "(119 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues", + "MetricExpr": "((108 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM + (108 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_cache", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM_PS;MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates (based on idle latencies) how often the CPU was stalled on accesses to external 3D-Xpoint (Crystal Ridge, a.k.a", + "MetricExpr": "(((1 - ((19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + 10 * ((MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))))) / ((19 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + 10 * ((MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + (MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))))) + (25 * (MEM_LOAD_RETIRED.LOCAL_PMM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + 33 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))))))) * (MEMORY_ACTIVITY.STALLS_L3_MISS / CLKS)) if (1000000 * (MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM) > MEM_LOAD_RETIRED.L1_MISS) else 0)", + "MetricGroup": "MemoryBound;Server;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_pmm_bound", + "PublicDescription": "This metric roughly estimates (based on idle latencies) how often the CPU was stalled on accesses to external 3D-Xpoint (Crystal Ridge, a.k.a. IXP) memory by loads, PMM stands for Persistent Memory Module. ", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "EXE_ACTIVITY.BOUND_ON_STORES / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_INST_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((MEM_STORE_RETIRED.L2_HIT * 10 * (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES))) + (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "(28 * Average_Frequency) * OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "MEM_INST_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_INST_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores", + "MetricExpr": "9 * OCR.STREAMING_WR.ANY_RESPONSE / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_streaming_stores", + "PublicDescription": "This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should Streaming stores be a bottleneck. Sample with: OCR.STREAMING_WR.ANY_RESPONSE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(7 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE) / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_INST_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_store - tma_store_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk", + "MetricExpr": "DTLB_STORE_MISSES.WALK_ACTIVE / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "max(0, tma_backend_bound - tma_memory_bound)", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.DIVIDER_ACTIVE / CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_ACTIVE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "(cpu@EXE_ACTIVITY.3_PORTS_UTIL\\,umask\\=0x80@ + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * cpu@EXE_ACTIVITY.2_PORTS_UTIL\\,umask\\=0xc@)) / CLKS if (ARITH.DIVIDER_ACTIVE < (CYCLE_ACTIVITY.STALLS_TOTAL - EXE_ACTIVITY.BOUND_ON_LOADS)) else (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * cpu@EXE_ACTIVITY.2_PORTS_UTIL\\,umask\\=0xc@) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "cpu@EXE_ACTIVITY.3_PORTS_UTIL\\,umask\\=0x80@ / CLKS + tma_serializing_operation * (CYCLE_ACTIVITY.STALLS_TOTAL - EXE_ACTIVITY.BOUND_ON_LOADS) / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations", + "MetricExpr": "RESOURCE_STALLS.SCOREBOARD / CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_serializing_operation", + "PublicDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out-of-order execution which may limit performance. Sample with: RESOURCE_STALLS.SCOREBOARD", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions", + "MetricExpr": "CPU_CLK_UNHALTED.PAUSE / CLKS", + "MetricGroup": "TopdownL6;tma_serializing_operation_group", + "MetricName": "tma_slow_pause", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions. Sample with: CPU_CLK_UNHALTED.PAUSE_INST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to LFENCE Instructions.", + "MetricExpr": "13 * MISC2_RETIRED.LFENCE / CLKS", + "MetricGroup": "TopdownL6;tma_serializing_operation_group", + "MetricName": "tma_memory_fence", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued", + "MetricExpr": "160 * ASSISTS.SSE_AVX_MIX / CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_mixing_vectors", + "PublicDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the Advanced Matrix Extensions (AMX) execution engine was busy with tile (arithmetic) operations", + "MetricExpr": "EXE.AMX_BUSY / CORE_CLKS", + "MetricGroup": "Compute;HPC;Server;TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_amx_busy", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "EXE_ACTIVITY.1_PORTS_UTIL / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful. Sample with: EXE_ACTIVITY.1_PORTS_UTIL", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "EXE_ACTIVITY.2_PORTS_UTIL / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop. Sample with: EXE_ACTIVITY.2_PORTS_UTIL", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "UOPS_EXECUTED.CYCLES_GE_3 / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Sample with: UOPS_EXECUTED.CYCLES_GE_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED.PORT_0 + UOPS_DISPATCHED.PORT_1 + UOPS_DISPATCHED.PORT_5_11 + UOPS_DISPATCHED.PORT_6) / (5 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED.PORT_0", + "MetricExpr": "UOPS_DISPATCHED.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED.PORT_1", + "MetricExpr": "UOPS_DISPATCHED.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED.PORT_6", + "MetricExpr": "UOPS_DISPATCHED.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3_10", + "MetricExpr": "UOPS_DISPATCHED.PORT_2_3_10 / (3 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations Sample with: UOPS_DISPATCHED.PORT_7_8", + "MetricExpr": "(UOPS_DISPATCHED.PORT_4_9 + UOPS_DISPATCHED.PORT_7_8) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", + "MetricExpr": "topdown\\-retiring / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "max(0, tma_retiring - tma_heavy_operations)", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector + tma_fp_amx", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "tma_retiring * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.VECTOR) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.128B_PACKED_HALF) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.256B_PACKED_HALF) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.512B_PACKED_HALF) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_512b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) matrix uops fraction the CPU has retired (aggregated across all supported FP datatypes in AMX engine)", + "MetricExpr": "cpu@AMX_OPS_RETIRED.BF16\\,cmask\\=1@ / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;HPC;Pipeline;Server;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_amx", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) matrix uops fraction the CPU has retired (aggregated across all supported FP datatypes in AMX engine). Refer to AMX_Busy and GFLOPs metrics for actual AMX utilization and FP performance, resp.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall Integer (Int) select operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_int_vector_128b + tma_int_vector_256b + tma_shuffles", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_int_operations", + "PublicDescription": "This metric represents overall Integer (Int) select operations fraction the CPU has executed (retired). Vector/Matrix Int operations and shuffles are counted. Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents 128-bit vector Integer ADD/SUB/SAD or VNNI (Vector Neural Network Instructions) uops fraction the CPU has retired.", + "MetricExpr": "(INT_VEC_RETIRED.ADD_128 + INT_VEC_RETIRED.VNNI_128) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;IntVector;Pipeline;TopdownL4;tma_int_operations_group", + "MetricName": "tma_int_vector_128b", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents 256-bit vector Integer ADD/SUB/SAD or VNNI (Vector Neural Network Instructions) uops fraction the CPU has retired.", + "MetricExpr": "(INT_VEC_RETIRED.ADD_256 + INT_VEC_RETIRED.MUL_256 + INT_VEC_RETIRED.VNNI_256) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;IntVector;Pipeline;TopdownL4;tma_int_operations_group", + "MetricName": "tma_int_vector_256b", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic Integer (Int) matrix uops fraction the CPU has retired (aggregated across all supported Int datatypes in AMX engine)", + "MetricExpr": "cpu@AMX_OPS_RETIRED.INT8\\,cmask\\=1@ / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;HPC;IntVector;Pipeline;Server;TopdownL4;tma_int_operations_group", + "MetricName": "tma_int_amx", + "PublicDescription": "This metric approximates arithmetic Integer (Int) matrix uops fraction the CPU has retired (aggregated across all supported Int datatypes in AMX engine). Refer to AMX_Busy and TIOPs metrics for actual AMX utilization and Int performance, resp.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Shuffle (cross \"vector lane\" data transfers) uops fraction the CPU has retired.", + "MetricExpr": "INT_VEC_RETIRED.SHUFFLES / (tma_retiring * SLOTS)", + "MetricGroup": "HPC;Pipeline;TopdownL4;tma_int_operations_group", + "MetricName": "tma_shuffles", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", + "MetricExpr": "tma_light_operations * MEM_UOP_RETIRED.ANY / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_memory_operations", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions", + "MetricExpr": "tma_light_operations * INST_RETIRED.MACRO_FUSED / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fused_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions. The instruction pairs of CMP+JCC or DEC+JCC are commonly used examples.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused", + "MetricExpr": "tma_light_operations * (BR_INST_RETIRED.ALL_BRANCHES - INST_RETIRED.MACRO_FUSED) / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_non_fused_branches", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused. Non-conditional branches like direct JMP or CALL would count here. Can be used to examine fusible conditional jumps that were not fused.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions", + "MetricExpr": "tma_light_operations * INST_RETIRED.NOP / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_nop_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body. Sample with: INST_RETIRED.NOP", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", + "MetricExpr": "max(0, tma_light_operations - (tma_fp_arith + tma_int_operations + tma_memory_operations + tma_fused_instructions + tma_non_fused_branches + tma_nop_instructions))", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_other_light_ops", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "topdown\\-heavy\\-ops / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences. Sample with: UOPS_RETIRED.HEAVY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops", + "MetricExpr": "tma_heavy_operations - tma_microcode_sequencer", + "MetricGroup": "TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_few_uops_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "UOPS_RETIRED.MS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: UOPS_RETIRED.MS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * cpu@ASSISTS.ANY\\,umask\\=0x1B@ / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Page Faults", + "MetricExpr": "99 * ASSISTS.PAGE_FAULT / SLOTS", + "MetricGroup": "TopdownL5;tma_assists_group", + "MetricName": "tma_page_faults", + "PublicDescription": "This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Page Faults. A Page Fault may apply on first application access to a memory page. Note operating system handling of page faults accounts for the majority of its cost.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Floating Point (FP) Assists", + "MetricExpr": "30 * ASSISTS.FP / SLOTS", + "MetricGroup": "HPC;TopdownL5;tma_assists_group", + "MetricName": "tma_fp_assists", + "PublicDescription": "This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Floating Point (FP) Assists. FP Assist may apply when working with very small floating point values (so-called denormals).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops as a result of handing SSE to AVX* or AVX* to SSE transition Assists. ", + "MetricExpr": "63 * ASSISTS.SSE_AVX_MIX / SLOTS", + "MetricGroup": "HPC;TopdownL5;tma_assists_group", + "MetricName": "tma_avx_assists", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources. Sample with: FRONTEND_RETIRED.MS_FLOWS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", + "MetricExpr": "100 * (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches))", + "MetricGroup": "Bad;BadSpec;BrMispredicts", + "MetricName": "Mispredictions" + }, + { + "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_mem_bandwidth / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_sq_full / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full))) + (tma_l1_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_fb_full / (tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) ", + "MetricGroup": "Mem;MemoryBW;Offcore", + "MetricName": "Memory_Bandwidth" + }, + { + "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_mem_latency / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_l3_hit_latency / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full)) + (tma_l2_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)))", + "MetricGroup": "Mem;MemoryLat;Offcore", + "MetricName": "Memory_Latency" + }, + { + "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", + "MetricExpr": "100 * tma_memory_bound * ((tma_l1_bound / max(tma_memory_bound, tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_dtlb_load / max(tma_l1_bound, tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) + (tma_store_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_pmm_bound + tma_store_bound)) * (tma_dtlb_store / (tma_dtlb_store + tma_false_sharing + tma_split_stores + tma_store_latency + tma_streaming_stores))) ", + "MetricGroup": "Mem;MemoryTLB;Offcore", + "MetricName": "Memory_Data_TLBs" + }, + { "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.COND + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL) ) / TOPDOWN.SLOTS)", + "MetricExpr": "100 * ((BR_INST_RETIRED.COND + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL)) / SLOTS)", "MetricGroup": "Ret", "MetricName": "Branching_Overhead" }, { + "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", + "MetricExpr": "100 * tma_fetch_latency * (tma_itlb_misses + tma_icache_misses + tma_unknown_branches) / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)", + "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB", + "MetricName": "Big_Code" + }, + { + "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", + "MetricExpr": "100 * (tma_frontend_bound - tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) - Big_Code", + "MetricGroup": "Fed;FetchBW;Frontend", + "MetricName": "Instruction_Fetch_BW" + }, + { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, { + "BriefDescription": "Uops Per Instruction", + "MetricExpr": "(tma_retiring * SLOTS) / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;Ret;Retire", + "MetricName": "UPI" + }, + { + "BriefDescription": "Instruction per taken branch", + "MetricExpr": "(tma_retiring * SLOTS) / BR_INST_RETIRED.NEAR_TAKEN", + "MetricGroup": "Branches;Fed;FetchBW", + "MetricName": "UpTB" + }, + { + "BriefDescription": "Cycles Per Instruction (per Logical Processor)", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", + "MetricName": "CPI" + }, + { "BriefDescription": "Per-Logical Processor actual clocks when the Logical Processor is active.", "MetricExpr": "CPU_CLK_UNHALTED.THREAD", "MetricGroup": "Pipeline", @@ -20,13 +822,13 @@ { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", "MetricExpr": "TOPDOWN.SLOTS", - "MetricGroup": "TmaL1", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { "BriefDescription": "Fraction of Physical Core issue-slots utilized by this Logical Processor", - "MetricExpr": "TOPDOWN.SLOTS / ( TOPDOWN.SLOTS / 2 ) if #SMT_on else 1", - "MetricGroup": "SMT;TmaL1", + "MetricExpr": "SLOTS / (TOPDOWN.SLOTS / 2) if #SMT_on else 1", + "MetricGroup": "SMT;tma_L1_group", "MetricName": "Slots_Utilization" }, { @@ -38,30 +840,36 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.DISTRIBUTED", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR_HALF ) + 2 * ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED2.COMPLEX_SCALAR_HALF ) + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED2.128B_PACKED_HALF + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * ( FP_ARITH_INST_RETIRED2.256B_PACKED_HALF + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) + 32 * FP_ARITH_INST_RETIRED2.512B_PACKED_HALF + 4 * AMX_OPS_RETIRED.BF16 ) / CPU_CLK_UNHALTED.DISTRIBUTED", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR_HALF) + 2 * (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED2.COMPLEX_SCALAR_HALF) + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED2.128B_PACKED_HALF + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * (FP_ARITH_INST_RETIRED2.256B_PACKED_HALF + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) + 32 * FP_ARITH_INST_RETIRED2.512B_PACKED_HALF + 4 * AMX_OPS_RETIRED.BF16) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( FP_ARITH_DISPATCHED.PORT_0 + FP_ARITH_DISPATCHED.PORT_1 + FP_ARITH_DISPATCHED.PORT_5 ) / ( 2 * CPU_CLK_UNHALTED.DISTRIBUTED )", + "MetricExpr": "(FP_ARITH_DISPATCHED.PORT_0 + FP_ARITH_DISPATCHED.PORT_1 + FP_ARITH_DISPATCHED.PORT_5) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)." }, { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2 ) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { + "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", + "MetricExpr": "(1 - tma_core_bound / tma_ports_utilization if tma_core_bound < tma_ports_utilization else 1) if SMT_2T_Utilization > 0.5 else 0", + "MetricGroup": "Cor;SMT", + "MetricName": "Core_Bound_Likely" + }, + { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", "MetricExpr": "CPU_CLK_UNHALTED.DISTRIBUTED", "MetricGroup": "SMT", @@ -105,13 +913,13 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR_HALF ) + 2 * ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED2.COMPLEX_SCALAR_HALF ) + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED2.128B_PACKED_HALF + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * ( FP_ARITH_INST_RETIRED2.256B_PACKED_HALF + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) + 32 * FP_ARITH_INST_RETIRED2.512B_PACKED_HALF + 4 * AMX_OPS_RETIRED.BF16 )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR_HALF) + 2 * (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED2.COMPLEX_SCALAR_HALF) + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED2.128B_PACKED_HALF + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * (FP_ARITH_INST_RETIRED2.256B_PACKED_HALF + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) + 32 * FP_ARITH_INST_RETIRED2.512B_PACKED_HALF + 4 * AMX_OPS_RETIRED.BF16)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.VECTOR) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.VECTOR))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." @@ -132,21 +940,21 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.128B_PACKED_HALF )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.128B_PACKED_HALF)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.256B_PACKED_HALF )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.256B_PACKED_HALF)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.512B_PACKED_HALF )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.512B_PACKED_HALF)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX512", "PublicDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." @@ -161,7 +969,7 @@ { "BriefDescription": "Instructions per Integer Arithmetic AMX operation (lower number means higher occurrence rate)", "MetricExpr": "INST_RETIRED.ANY / AMX_OPS_RETIRED.INT8", - "MetricGroup": "IntVector;InsType;Server", + "MetricGroup": "InsType;IntVector;Server", "MetricName": "IpArith_AMX_Int8", "PublicDescription": "Instructions per Integer Arithmetic AMX operation (lower number means higher occurrence rate). Operations factored per matrices' sizes of the AMX instructions." }, @@ -172,12 +980,18 @@ "MetricName": "IpSWPF" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { + "BriefDescription": "Average number of Uops retired in cycles where at least one uop has retired.", + "MetricExpr": "(tma_retiring * SLOTS) / cpu@UOPS_RETIRED.SLOTS\\,cmask\\=1@", + "MetricGroup": "Pipeline;Ret", + "MetricName": "Retire" + }, + { "BriefDescription": "Estimated fraction of retirement-cycles dealing with repeat instructions", "MetricExpr": "INST_RETIRED.REP_ITERATION / cpu@UOPS_RETIRED.SLOTS\\,cmask\\=1@", "MetricGroup": "Pipeline;Ret", @@ -214,6 +1028,12 @@ "MetricName": "DSB_Switch_Cost" }, { + "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", + "MetricExpr": "100 * (tma_fetch_latency * tma_dsb_switches / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches) + tma_fetch_bandwidth * tma_mite / (tma_dsb + tma_mite))", + "MetricGroup": "DSBmiss;Fed", + "MetricName": "DSB_Misses" + }, + { "BriefDescription": "Number of Instructions per non-speculative DSB miss (lower number means higher occurrence rate)", "MetricExpr": "INST_RETIRED.ANY / FRONTEND_RETIRED.ANY_DSB_MISS", "MetricGroup": "DSBmiss;Fed", @@ -226,6 +1046,12 @@ "MetricName": "IpMispredict" }, { + "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricGroup": "Bad;BrMispredicts", + "MetricName": "Branch_Misprediction_Cost" + }, + { "BriefDescription": "Fraction of branches that are non-taken conditionals", "MetricExpr": "BR_INST_RETIRED.COND_NTAKEN / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", @@ -239,7 +1065,7 @@ }, { "BriefDescription": "Fraction of branches that are CALL or RET", - "MetricExpr": "( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "CallRet" }, @@ -251,7 +1077,7 @@ }, { "BriefDescription": "Fraction of branches of other types (not individually covered by other metrics in Info.Branches group)", - "MetricExpr": "1 - ( (BR_INST_RETIRED.COND_NTAKEN / BR_INST_RETIRED.ALL_BRANCHES) + (BR_INST_RETIRED.COND_TAKEN / BR_INST_RETIRED.ALL_BRANCHES) + (( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES) + ((BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES) )", + "MetricExpr": "1 - (Cond_NT + Cond_TK + CallRet + Jump)", "MetricGroup": "Bad;Branches", "MetricName": "Other_Branches" }, @@ -264,67 +1090,67 @@ { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L1 cache true misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.ALL_DEMAND_DATA_RD / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI_Load" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load" }, { "BriefDescription": "L2 cache hits per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( L2_RQSTS.REFERENCES - L2_RQSTS.MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricExpr": "1000 * (L2_RQSTS.REFERENCES - L2_RQSTS.MISS) / INST_RETIRED.ANY", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_All" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries)", "MetricExpr": "1000 * MEM_LOAD_RETIRED.FB_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "FB_HPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING ) / ( 4 * CPU_CLK_UNHALTED.DISTRIBUTED )", + "MetricExpr": "(ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING) / (4 * CORE_CLKS)", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, @@ -354,37 +1180,37 @@ }, { "BriefDescription": "Rate of silent evictions from the L2 cache per Kilo instruction where the evicted lines are dropped (no writeback to L3 or memory)", - "MetricExpr": "1000 * L2_LINES_OUT.SILENT / INST_RETIRED.ANY", + "MetricExpr": "1000 * L2_LINES_OUT.SILENT / Instructions", "MetricGroup": "L2Evicts;Mem;Server", "MetricName": "L2_Evictions_Silent_PKI" }, { "BriefDescription": "Rate of non silent evictions from the L2 cache per Kilo instruction", - "MetricExpr": "1000 * L2_LINES_OUT.NON_SILENT / INST_RETIRED.ANY", + "MetricExpr": "1000 * L2_LINES_OUT.NON_SILENT / Instructions", "MetricGroup": "L2Evicts;Mem;Server", "MetricName": "L2_Evictions_NonSilent_PKI" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data access bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * OFFCORE_REQUESTS.ALL_REQUESTS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Access_BW", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "L3_Cache_Access_BW_1T" }, @@ -396,26 +1222,26 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR_HALF ) + 2 * ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED2.COMPLEX_SCALAR_HALF ) + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED2.128B_PACKED_HALF + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * ( FP_ARITH_INST_RETIRED2.256B_PACKED_HALF + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) + 32 * FP_ARITH_INST_RETIRED2.512B_PACKED_HALF + 4 * AMX_OPS_RETIRED.BF16 ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR_HALF) + 2 * (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED2.COMPLEX_SCALAR_HALF) + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED2.128B_PACKED_HALF + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * (FP_ARITH_INST_RETIRED2.256B_PACKED_HALF + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) + 32 * FP_ARITH_INST_RETIRED2.512B_PACKED_HALF + 4 * AMX_OPS_RETIRED.BF16) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Tera Integer (matrix) Operations Per Second", - "MetricExpr": "( 8 * AMX_OPS_RETIRED.INT8 / 1000000000000 ) / duration_time", + "MetricExpr": "(8 * AMX_OPS_RETIRED.INT8 / 1e12) / duration_time", "MetricGroup": "Cor;HPC;IntVector;Server", "MetricName": "TIOPS" }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, @@ -439,13 +1265,13 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "( 64 * ( uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@ ) / 1000000000 ) / duration_time", + "MetricExpr": "(64 * (uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@) / 1000000000) / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, { "BriefDescription": "Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches", - "MetricExpr": "1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD / UNC_CHA_TOR_INSERTS.IA_MISS_DRD ) / ( uncore_cha_0@event\\=0x1@ / duration_time )", + "MetricExpr": "1000000000 * (UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD / UNC_CHA_TOR_INSERTS.IA_MISS_DRD) / (Socket_CLKS / duration_time)", "MetricGroup": "Mem;MemoryLat;SoC", "MetricName": "MEM_Read_Latency" }, @@ -457,32 +1283,32 @@ }, { "BriefDescription": "Average latency of data read request to external 3D X-Point memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches", - "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PMM ) / uncore_cha_0@event\\=0x1@ )", - "MetricGroup": "Mem;MemoryLat;SoC;Server", + "MetricExpr": "(1000000000 * (UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PMM) / uncore_cha_0@event\\=0x1@)", + "MetricGroup": "Mem;MemoryLat;Server;SoC", "MetricName": "MEM_PMM_Read_Latency" }, { "BriefDescription": "Average latency of data read request to external DRAM memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches", - "MetricExpr": " 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_DDR ) / uncore_cha_0@event\\=0x1@", - "MetricGroup": "Mem;MemoryLat;SoC;Server", + "MetricExpr": " 1000000000 * (UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_DDR) / uncore_cha_0@event\\=0x1@", + "MetricGroup": "Mem;MemoryLat;Server;SoC", "MetricName": "MEM_DRAM_Read_Latency" }, { "BriefDescription": "Average 3DXP Memory Bandwidth Use for reads [GB / sec]", - "MetricExpr": "( ( 64 * UNC_M_PMM_RPQ_INSERTS / 1000000000 ) / duration_time )", - "MetricGroup": "Mem;MemoryBW;SoC;Server", + "MetricExpr": "((64 * UNC_M_PMM_RPQ_INSERTS / 1000000000) / duration_time)", + "MetricGroup": "Mem;MemoryBW;Server;SoC", "MetricName": "PMM_Read_BW" }, { "BriefDescription": "Average 3DXP Memory Bandwidth Use for Writes [GB / sec]", - "MetricExpr": "( ( 64 * UNC_M_PMM_WPQ_INSERTS / 1000000000 ) / duration_time )", - "MetricGroup": "Mem;MemoryBW;SoC;Server", + "MetricExpr": "((64 * UNC_M_PMM_WPQ_INSERTS / 1000000000) / duration_time)", + "MetricGroup": "Mem;MemoryBW;Server;SoC", "MetricName": "PMM_Write_BW" }, { "BriefDescription": "Average IO (network or disk) Bandwidth Use for Writes [GB / sec]", "MetricExpr": "UNC_CHA_TOR_INSERTS.IO_PCIRDCUR * 64 / 1000000000 / duration_time", - "MetricGroup": "IoBW;Mem;SoC;Server", + "MetricGroup": "IoBW;Mem;Server;SoC", "MetricName": "IO_Write_BW" }, { @@ -492,12 +1318,6 @@ "MetricName": "Socket_CLKS" }, { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "uncore_cha_0@event\\=0x1@ / #num_dies / duration_time / 1000000000", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, - { "BriefDescription": "Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]", "MetricExpr": "INST_RETIRED.ANY / BR_INST_RETIRED.FAR_BRANCH:u", "MetricGroup": "Branches;OS", @@ -528,11 +1348,10 @@ "MetricName": "C6_Pkg_Residency" }, { - "BriefDescription": "Percentage of time spent in the active CPU power state C0", - "MetricExpr": "100 * CPU_CLK_UNHALTED.REF_TSC / TSC", - "MetricGroup": "", - "MetricName": "cpu_utilization_percent", - "ScaleUnit": "1%" + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "Socket_CLKS / #num_dies / duration_time / 1000000000", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ" }, { "BriefDescription": "CPU operating frequency (in GHz)", @@ -542,13 +1361,6 @@ "ScaleUnit": "1GHz" }, { - "BriefDescription": "Cycles per instruction retired; indicating how much time each executed instruction took; in units of cycles.", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY", - "MetricGroup": "", - "MetricName": "cpi", - "ScaleUnit": "1per_instr" - }, - { "BriefDescription": "The ratio of number of completed memory load instructions to the total number completed instructions", "MetricExpr": "MEM_INST_RETIRED.ALL_LOADS / INST_RETIRED.ANY", "MetricGroup": "", @@ -566,7 +1378,7 @@ "BriefDescription": "Ratio of number of requests missing L1 data cache (includes data+rfo w/ prefetches) to the total number of completed instructions", "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l1d_mpi_includes_data_plus_rfo_with_prefetches", + "MetricName": "l1d_mpi", "ScaleUnit": "1per_instr" }, { @@ -594,7 +1406,7 @@ "BriefDescription": "Ratio of number of requests missing L2 cache (includes code+data+rfo w/ prefetches) to the total number of completed instructions", "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches", + "MetricName": "l2_mpi", "ScaleUnit": "1per_instr" }, { @@ -620,42 +1432,42 @@ }, { "BriefDescription": "Ratio of number of code read requests missing last level core cache (includes demand w/ prefetches) to the total number of completed instructions", - "MetricExpr": "( UNC_CHA_TOR_INSERTS.IA_MISS_CRD ) / INST_RETIRED.ANY", + "MetricExpr": "UNC_CHA_TOR_INSERTS.IA_MISS_CRD / INST_RETIRED.ANY", "MetricGroup": "", "MetricName": "llc_code_read_mpi_demand_plus_prefetch", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand data read miss (read memory access) in nano seconds", - "MetricExpr": "( ( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD / UNC_CHA_TOR_INSERTS.IA_MISS_DRD ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD / UNC_CHA_TOR_INSERTS.IA_MISS_DRD ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_demand_data_read_miss_latency", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand data read miss (read memory access) addressed to local memory in nano seconds", - "MetricExpr": "( ( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_LOCAL / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_LOCAL ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_LOCAL) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_LOCAL / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_LOCAL ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_LOCAL) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_demand_data_read_miss_latency_for_local_requests", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand data read miss (read memory access) addressed to remote memory in nano seconds", - "MetricExpr": "( ( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_REMOTE / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_REMOTE ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_REMOTE) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_REMOTE / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_REMOTE ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_REMOTE) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_demand_data_read_miss_latency_for_remote_requests", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand data read miss (read memory access) addressed to Intel(R) Optane(TM) Persistent Memory(PMEM) in nano seconds", - "MetricExpr": "( ( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PMM ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PMM ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_demand_data_read_miss_to_pmem_latency", "ScaleUnit": "1ns" }, { "BriefDescription": "Average latency of a last level cache (LLC) demand data read miss (read memory access) addressed to DRAM in nano seconds", - "MetricExpr": "( ( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_DDR ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR) * #num_packages ) ) ) * duration_time )", + "MetricExpr": "( 1000000000 * ( UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR / UNC_CHA_TOR_INSERTS.IA_MISS_DRD_DDR ) / ( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR) * #num_packages ) ) ) * duration_time", "MetricGroup": "", "MetricName": "llc_demand_data_read_miss_to_dram_latency", "ScaleUnit": "1ns" @@ -699,14 +1511,14 @@ "BriefDescription": "Memory read that miss the last level cache (LLC) addressed to local DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", "MetricExpr": "100 * ( UNC_CHA_TOR_INSERTS.IA_MISS_DRD_LOCAL + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_LOCAL ) / ( UNC_CHA_TOR_INSERTS.IA_MISS_DRD_LOCAL + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_LOCAL + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_REMOTE + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_REMOTE )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_local_dram", + "MetricName": "numa_reads_addressed_to_local_dram", "ScaleUnit": "1%" }, { "BriefDescription": "Memory reads that miss the last level cache (LLC) addressed to remote DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", "MetricExpr": "100 * ( UNC_CHA_TOR_INSERTS.IA_MISS_DRD_REMOTE + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_REMOTE ) / ( UNC_CHA_TOR_INSERTS.IA_MISS_DRD_LOCAL + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_LOCAL + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_REMOTE + UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PREF_REMOTE )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_remote_dram", + "MetricName": "numa_reads_addressed_to_remote_dram", "ScaleUnit": "1%" }, { @@ -720,7 +1532,7 @@ "BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data transmit bandwidth (MB/sec)", "MetricExpr": "( UNC_UPI_TxL_FLITS.ALL_DATA * (64 / 9.0) / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "upi_data_transmit_bw_only_data", + "MetricName": "upi_data_transmit_bw", "ScaleUnit": "1MB/s" }, { @@ -769,35 +1581,35 @@ "BriefDescription": "Bandwidth of IO reads that are initiated by end device controllers that are requesting memory from the CPU.", "MetricExpr": "( UNC_CHA_TOR_INSERTS.IO_PCIRDCUR * 64 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_read", + "MetricName": "io_bandwidth_disk_or_network_writes", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Bandwidth of IO writes that are initiated by end device controllers that are writing memory to the CPU.", "MetricExpr": "(( UNC_CHA_TOR_INSERTS.IO_ITOM + UNC_CHA_TOR_INSERTS.IO_ITOMCACHENEAR ) * 64 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_write", + "MetricName": "io_bandwidth_disk_or_network_reads", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Uops delivered from decoded instruction cache (decoded stream buffer or DSB) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.DSB_UOPS / ( IDQ.DSB_UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS + LSD.UOPS ) )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_decoded_icache_dsb", + "MetricName": "percent_uops_delivered_from_decoded_icache", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from legacy decode pipeline (Micro-instruction Translation Engine or MITE) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.MITE_UOPS / ( IDQ.DSB_UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS + LSD.UOPS ) )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline_mite", + "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from microcode sequencer (MS) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.MS_UOPS / ( IDQ.DSB_UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS + LSD.UOPS ) )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_microcode_sequencer_ms", + "MetricName": "percent_uops_delivered_from_microcode_sequencer", "ScaleUnit": "1%" }, { @@ -827,264 +1639,5 @@ "MetricGroup": "", "MetricName": "llc_miss_remote_memory_bandwidth_write", "ScaleUnit": "1MB/s" - }, - { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", - "MetricExpr": "100 * ( topdown\\-fe\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) )", - "MetricGroup": "TmaL1;PGO", - "MetricName": "tma_frontend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.", - "MetricExpr": "100 * ( ( topdown\\-fetch\\-lat / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) ) )", - "MetricGroup": "Frontend;TmaL2;m_tma_frontend_bound_percent", - "MetricName": "tma_fetch_latency_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", - "MetricExpr": "100 * ( ICACHE_DATA.STALLS / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_icache_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.", - "MetricExpr": "100 * ( ICACHE_TAG.STALLS / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_itlb_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.", - "MetricExpr": "100 * ( INT_MISC.CLEAR_RESTEER_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) + ( INT_MISC.UNKNOWN_BRANCH_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) ) )", - "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_branch_resteers_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", - "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_dsb_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", - "MetricExpr": "100 * ( DECODE.LCP / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_lcp_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.", - "MetricExpr": "100 * ( ( 3 ) * IDQ.MS_SWITCHES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_ms_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", - "MetricExpr": "100 * ( max( 0 , ( topdown\\-fe\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) ) - ( ( topdown\\-fetch\\-lat / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) ) ) ) )", - "MetricGroup": "FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent", - "MetricName": "tma_fetch_bandwidth_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", - "MetricExpr": "100 * ( ( IDQ.MITE_CYCLES_ANY - IDQ.MITE_CYCLES_OK ) / ( CPU_CLK_UNHALTED.DISTRIBUTED ) / 2 )", - "MetricGroup": "DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent", - "MetricName": "tma_mite_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", - "MetricExpr": "100 * ( ( IDQ.DSB_CYCLES_ANY - IDQ.DSB_CYCLES_OK ) / ( CPU_CLK_UNHALTED.DISTRIBUTED ) / 2 )", - "MetricGroup": "DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent", - "MetricName": "tma_dsb_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", - "MetricExpr": "100 * ( max( 1 - ( ( topdown\\-fe\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) ) + ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) , 0 ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_bad_speculation_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.", - "MetricExpr": "( 100 * ( topdown\\-br\\-mispredict / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) + ( 0 * slots )", - "MetricGroup": "BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent", - "MetricName": "tma_branch_mispredicts_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.", - "MetricExpr": "100 * ( max( 0 , ( max( 1 - ( ( topdown\\-fe\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) - INT_MISC.UOP_DROPPING / ( slots ) ) + ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) , 0 ) ) - ( topdown\\-br\\-mispredict / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) )", - "MetricGroup": "BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent", - "MetricName": "tma_machine_clears_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", - "MetricExpr": "( 100 * ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) + ( 0 * slots )", - "MetricGroup": "TmaL1", - "MetricName": "tma_backend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", - "MetricExpr": "( 100 * ( topdown\\-mem\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) + ( 0 * slots )", - "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent", - "MetricName": "tma_memory_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.", - "MetricExpr": "100 * ( max( ( EXE_ACTIVITY.BOUND_ON_LOADS - MEMORY_ACTIVITY.STALLS_L1D_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) , 0 ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l1_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( MEMORY_ACTIVITY.STALLS_L1D_MISS - MEMORY_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l2_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( MEMORY_ACTIVITY.STALLS_L2_MISS - MEMORY_ACTIVITY.STALLS_L3_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l3_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.", - "MetricExpr": "100 * ( min( ( ( ( MEMORY_ACTIVITY.STALLS_L3_MISS / ( CPU_CLK_UNHALTED.THREAD ) ) - ( min( ( ( ( ( 1 - ( ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) / ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) + ( 25 * ( ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) + 33 * ( ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) ) ) ) ) * ( MEMORY_ACTIVITY.STALLS_L3_MISS / ( CPU_CLK_UNHALTED.THREAD ) ) ) if ( ( 1000000 ) * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) , ( 1 ) ) ) ) ) , ( 1 ) ) )", - "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_dram_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric roughly estimates (based on idle latencies) how often the CPU was stalled on accesses to external 3D-Xpoint (Crystal Ridge, a.k.a. IXP) memory by loads, PMM stands for Persistent Memory Module. ", - "MetricExpr": "100 * ( min( ( ( ( ( 1 - ( ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) / ( ( 19 * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + 10 * ( ( MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) + ( 25 * ( ( MEM_LOAD_RETIRED.LOCAL_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) + 33 * ( ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) ) ) ) ) ) ) * ( MEMORY_ACTIVITY.STALLS_L3_MISS / ( CPU_CLK_UNHALTED.THREAD ) ) ) if ( ( 1000000 ) * ( MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM + MEM_LOAD_RETIRED.LOCAL_PMM ) > MEM_LOAD_RETIRED.L1_MISS ) else 0 ) ) , ( 1 ) ) )", - "MetricGroup": "MemoryBound;Server;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_pmm_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.", - "MetricExpr": "100 * ( EXE_ACTIVITY.BOUND_ON_STORES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_store_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", - "MetricExpr": "( 100 * ( max( 0 , ( topdown\\-be\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-mem\\-bound / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) ) + ( 0 * slots )", - "MetricGroup": "Backend;TmaL2;Compute;m_tma_backend_bound_percent", - "MetricName": "tma_core_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.", - "MetricExpr": "100 * ( ARITH.DIVIDER_ACTIVE / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "TmaL3;m_tma_core_bound_percent", - "MetricName": "tma_divider_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", - "MetricExpr": "( 100 * ( ( EXE_ACTIVITY.EXE_BOUND_0_PORTS + ( EXE_ACTIVITY.1_PORTS_UTIL + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * cpu@EXE_ACTIVITY.2_PORTS_UTIL\\,umask\\=0xc@ ) ) / ( CPU_CLK_UNHALTED.THREAD ) if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - EXE_ACTIVITY.BOUND_ON_LOADS ) ) else ( EXE_ACTIVITY.1_PORTS_UTIL + ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * cpu@EXE_ACTIVITY.2_PORTS_UTIL\\,umask\\=0xc@ ) / ( CPU_CLK_UNHALTED.THREAD ) ) ) + ( 0 * slots )", - "MetricGroup": "PortsUtil;TmaL3;m_tma_core_bound_percent", - "MetricName": "tma_ports_utilization_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. ", - "MetricExpr": "( 100 * ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) + ( 0 * slots )", - "MetricGroup": "TmaL1", - "MetricName": "tma_retiring_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved.", - "MetricExpr": "( 100 * ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) ) + ( 0 * slots )", - "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", - "MetricName": "tma_light_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", - "MetricExpr": "100 * ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.VECTOR ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) , ( 1 ) ) ) + ( cpu@AMX_OPS_RETIRED.BF16\\,cmask\\=0x1@ / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) )", - "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_fp_arith_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents overall Integer (Int) select operations fraction the CPU has executed (retired). Vector/Matrix Int operations and shuffles are counted. Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain.", - "MetricExpr": "100 * ( ( ( INT_VEC_RETIRED.ADD_128 + INT_VEC_RETIRED.VNNI_128 ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( ( INT_VEC_RETIRED.ADD_256 + INT_VEC_RETIRED.MUL_256 + INT_VEC_RETIRED.VNNI_256 ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( INT_VEC_RETIRED.SHUFFLES / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_int_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", - "MetricExpr": "100 * ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) * MEM_UOP_RETIRED.ANY / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_memory_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions. The instruction pairs of CMP+JCC or DEC+JCC are commonly used examples.", - "MetricExpr": "100 * ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) * INST_RETIRED.MACRO_FUSED / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_fused_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused. Non-conditional branches like direct JMP or CALL would count here. Can be used to examine fusible conditional jumps that were not fused.", - "MetricExpr": "100 * ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) * ( BR_INST_RETIRED.ALL_BRANCHES - INST_RETIRED.MACRO_FUSED ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_non_fused_branches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body.", - "MetricExpr": "100 * ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) * INST_RETIRED.NOP / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_nop_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", - "MetricExpr": "100 * ( max( 0 , ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) - ( ( ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE + FP_ARITH_INST_RETIRED2.SCALAR ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE + FP_ARITH_INST_RETIRED2.VECTOR ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) , ( 1 ) ) ) + ( cpu@AMX_OPS_RETIRED.BF16\\,cmask\\=0x1@ / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) ) + ( ( ( INT_VEC_RETIRED.ADD_128 + INT_VEC_RETIRED.VNNI_128 ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( ( INT_VEC_RETIRED.ADD_256 + INT_VEC_RETIRED.MUL_256 + INT_VEC_RETIRED.VNNI_256 ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( INT_VEC_RETIRED.SHUFFLES / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) ) + ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) * MEM_UOP_RETIRED.ANY / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) * INST_RETIRED.MACRO_FUSED / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) * ( BR_INST_RETIRED.ALL_BRANCHES - INST_RETIRED.MACRO_FUSED ) / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) + ( ( max( 0 , ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) ) * INST_RETIRED.NOP / ( ( topdown\\-retiring / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) * ( slots ) ) ) ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_other_light_ops_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", - "MetricExpr": "( 100 * ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) ) + ( 0 * slots )", - "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", - "MetricName": "tma_heavy_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", - "MetricExpr": "100 * ( ( topdown\\-heavy\\-ops / ( topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound ) ) - ( UOPS_RETIRED.MS / ( slots ) ) )", - "MetricGroup": "TmaL3;m_tma_heavy_operations_percent", - "MetricName": "tma_few_uops_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided.", - "MetricExpr": "100 * ( UOPS_RETIRED.MS / ( slots ) )", - "MetricGroup": "MicroSeq;TmaL3;m_tma_heavy_operations_percent", - "MetricName": "tma_microcode_sequencer_percent", - "ScaleUnit": "1%" } ] diff --git a/tools/perf/pmu-events/arch/x86/skylake/skl-metrics.json b/tools/perf/pmu-events/arch/x86/skylake/skl-metrics.json index 73fa72d3dcb1..f138b9836b51 100644 --- a/tools/perf/pmu-events/arch/x86/skylake/skl-metrics.json +++ b/tools/perf/pmu-events/arch/x86/skylake/skl-metrics.json @@ -1,148 +1,694 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. Sample with: FRONTEND_RETIRED.LATENCY_GE_4_PS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: FRONTEND_RETIRED.LATENCY_GE_16_PS;FRONTEND_RETIRED.LATENCY_GE_8_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses", + "MetricExpr": "(ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@) / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses. Sample with: FRONTEND_RETIRED.L2_MISS_PS;FRONTEND_RETIRED.L1I_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "ICACHE_64B.IFTAG_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: FRONTEND_RETIRED.STLB_MISS_PS;FRONTEND_RETIRED.ITLB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "INT_MISC.CLEAR_RESTEER_CYCLES / CLKS + tma_unknown_branches", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears", + "MetricExpr": "(1 - (BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT))) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "9 * BACLEARS.ANY / CLKS", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: BACLEARS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty. Sample with: FRONTEND_RETIRED.DSB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "2 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend. Sample with: FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_2_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck. Sample with: FRONTEND_RETIRED.ANY_DSB_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where decoder-0 was the only active decoder", + "MetricExpr": "(cpu@INST_DECODED.DECODERS\\,cmask\\=1@ - cpu@INST_DECODED.DECODERS\\,cmask\\=2@) / CORE_CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_decoder0_alone", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - tma_frontend_bound - (UOPS_ISSUED.ANY + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_RETIRED.L1_HIT_PS;MEM_LOAD_RETIRED.FB_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "min(9 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE, max(CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS, 0)) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_INST_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_load - tma_load_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk", + "MetricExpr": "DTLB_LOAD_MISSES.WALK_ACTIVE / CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(12 * max(0, MEM_INST_RETIRED.LOCK_LOADS - L2_RQSTS.ALL_RFO) + (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES) * (9 * L2_RQSTS.RFO_HIT + min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO))) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_INST_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_INST_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "Load_Miss_Real_Latency * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) / ((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@)) * ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "((18.5 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM + (16.5 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "(16.5 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "(6.5 * Average_Frequency) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "((OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2) if #SMT_on else OFFCORE_REQUESTS_BUFFER.SQ_FULL) / CORE_CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L3_MISS / CLKS + ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS) - tma_l2_bound)", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "EXE_ACTIVITY.BOUND_ON_STORES / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_INST_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 9 * (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES))) + (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "(22 * Average_Frequency) * OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "MEM_INST_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_INST_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(9 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE) / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_INST_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_store - tma_store_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk", + "MetricExpr": "DTLB_STORE_MISSES.WALK_ACTIVE / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.DIVIDER_ACTIVE / CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_ACTIVE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "(EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL)) / CLKS if (ARITH.DIVIDER_ACTIVE < (CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY)) else (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(UOPS_EXECUTED.CORE_CYCLES_NONE / 2 if #SMT_on else CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations", + "MetricExpr": "PARTIAL_RAT_STALLS.SCOREBOARD / CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_serializing_operation", + "PublicDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out-of-order execution which may limit performance. Sample with: PARTIAL_RAT_STALLS.SCOREBOARD", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued", + "MetricExpr": "CLKS * UOPS_ISSUED.VECTOR_WIDTH_MISMATCH / UOPS_ISSUED.ANY", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_mixing_vectors", + "PublicDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "((UOPS_EXECUTED.CORE_CYCLES_GE_1 - UOPS_EXECUTED.CORE_CYCLES_GE_2) / 2 if #SMT_on else EXE_ACTIVITY.1_PORTS_UTIL) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "((UOPS_EXECUTED.CORE_CYCLES_GE_2 - UOPS_EXECUTED.CORE_CYCLES_GE_3) / 2 if #SMT_on else EXE_ACTIVITY.2_PORTS_UTIL) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).", + "MetricExpr": "(UOPS_EXECUTED.CORE_CYCLES_GE_3 / 2 if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_3) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_0 + UOPS_DISPATCHED_PORT.PORT_1 + UOPS_DISPATCHED_PORT.PORT_5 + UOPS_DISPATCHED_PORT.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED_PORT.PORT_0", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_1", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_6", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_2 + UOPS_DISPATCHED_PORT.PORT_3 + UOPS_DISPATCHED_PORT.PORT_7 - UOPS_DISPATCHED_PORT.PORT_4) / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_2", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_2 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_2", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_3", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_3 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data) Sample with: UOPS_DISPATCHED_PORT.PORT_4", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_4", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 7 ([HSW+]simple Store-address) Sample with: UOPS_DISPATCHED_PORT.PORT_7", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_7 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_7", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" }, { - "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", - "MetricExpr": "100 * ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "Bad;BadSpec;BrMispredicts", - "MetricName": "Mispredictions" + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "tma_retiring * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", + "MetricExpr": "tma_light_operations * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_memory_operations", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions", + "MetricExpr": "tma_light_operations * UOPS_RETIRED.MACRO_FUSED / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fused_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions. The instruction pairs of CMP+JCC or DEC+JCC are commonly used examples.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused", + "MetricExpr": "tma_light_operations * (BR_INST_RETIRED.ALL_BRANCHES - UOPS_RETIRED.MACRO_FUSED) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_non_fused_branches", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused. Non-conditional branches like direct JMP or CALL would count here. Can be used to examine fusible conditional jumps that were not fused.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions", + "MetricExpr": "tma_light_operations * INST_RETIRED.NOP / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_nop_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body. Sample with: INST_RETIRED.NOP", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", + "MetricExpr": "max(0, tma_light_operations - (tma_fp_arith + tma_memory_operations + tma_fused_instructions + tma_non_fused_branches + tma_nop_instructions))", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_other_light_ops", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY) / SLOTS", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops", + "MetricExpr": "tma_heavy_operations - tma_microcode_sequencer", + "MetricGroup": "TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_few_uops_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * (FP_ASSIST.ANY + OTHER_ASSISTS.ANY) / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: OTHER_ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" }, { "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", - "MetricExpr": "100 * ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "Bad;BadSpec;BrMispredicts_SMT", - "MetricName": "Mispredictions_SMT" + "MetricExpr": "100 * (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches))", + "MetricGroup": "Bad;BadSpec;BrMispredicts", + "MetricName": "Mispredictions" }, { "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) * ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD) / #(CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (OFFCORE_REQUESTS_BUFFER.SQ_FULL / CPU_CLK_UNHALTED.THREAD) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) ) + ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( ((L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )) * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CPU_CLK_UNHALTED.THREAD) / #(max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) ", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_mem_bandwidth / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_sq_full / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full))) + (tma_l1_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_fb_full / (tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) ", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "Memory_Bandwidth" }, { - "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD) / #(CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (( OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2 ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) ) + ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( ((L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )) * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CPU_CLK_UNHALTED.THREAD) / #(max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) ", - "MetricGroup": "Mem;MemoryBW;Offcore_SMT", - "MetricName": "Memory_Bandwidth_SMT" - }, - { "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) * ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD ) / CPU_CLK_UNHALTED.THREAD - (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD)) / #(CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (( (10 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) - (3.5 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) ) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CPU_CLK_UNHALTED.THREAD) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) + ( (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD)) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) )", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_mem_latency / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_l3_hit_latency / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full)) + (tma_l2_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)))", "MetricGroup": "Mem;MemoryLat;Offcore", "MetricName": "Memory_Latency" }, { - "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD ) / CPU_CLK_UNHALTED.THREAD - (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD)) / #(CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (( (10 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) - (3.5 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) ) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CPU_CLK_UNHALTED.THREAD) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) + ( (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD)) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) )", - "MetricGroup": "Mem;MemoryLat;Offcore_SMT", - "MetricName": "Memory_Latency_SMT" - }, - { "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) * ( ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (min( 9 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE , max( CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS , 0 ) ) / CPU_CLK_UNHALTED.THREAD) / (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) + ( (EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (( 9 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE ) / CPU_CLK_UNHALTED.THREAD) / #(EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) ) ) ", + "MetricExpr": "100 * tma_memory_bound * ((tma_l1_bound / max(tma_memory_bound, tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_dtlb_load / max(tma_l1_bound, tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) + (tma_store_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_dtlb_store / (tma_dtlb_store + tma_false_sharing + tma_split_stores + tma_store_latency))) ", "MetricGroup": "Mem;MemoryTLB;Offcore", "MetricName": "Memory_Data_TLBs" }, { - "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * ( ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (min( 9 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE , max( CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS , 0 ) ) / CPU_CLK_UNHALTED.THREAD) / (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) + ( (EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (( 9 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / #(EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) ) ) ", - "MetricGroup": "Mem;MemoryTLB;Offcore_SMT", - "MetricName": "Memory_Data_TLBs_SMT" - }, - { "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.CONDITIONAL + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - ( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) - 2 * BR_INST_RETIRED.NEAR_CALL) ) / (4 * CPU_CLK_UNHALTED.THREAD))", + "MetricExpr": "100 * ((BR_INST_RETIRED.CONDITIONAL + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - (BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN) - 2 * BR_INST_RETIRED.NEAR_CALL)) / SLOTS)", "MetricGroup": "Ret", "MetricName": "Branching_Overhead" }, { - "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.CONDITIONAL + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - ( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) - 2 * BR_INST_RETIRED.NEAR_CALL) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))", - "MetricGroup": "Ret_SMT", - "MetricName": "Branching_Overhead_SMT" - }, - { "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", - "MetricExpr": "100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD))", + "MetricExpr": "100 * tma_fetch_latency * (tma_itlb_misses + tma_icache_misses + tma_unknown_branches) / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)", "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB", "MetricName": "Big_Code" }, { - "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", - "MetricExpr": "100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))", - "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB_SMT", - "MetricName": "Big_Code_SMT" - }, - { "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", - "MetricExpr": "100 * ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) - (100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)))", + "MetricExpr": "100 * (tma_frontend_bound - tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) - Big_Code", "MetricGroup": "Fed;FetchBW;Frontend", "MetricName": "Instruction_Fetch_BW" }, { - "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", - "MetricExpr": "100 * ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) - (100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))))", - "MetricGroup": "Fed;FetchBW;Frontend_SMT", - "MetricName": "Instruction_Fetch_BW_SMT" - }, - { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -160,8 +706,8 @@ }, { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI" }, { @@ -172,17 +718,11 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "The ratio of Executed- by Issued-Uops", "MetricExpr": "UOPS_EXECUTED.THREAD / UOPS_ISSUED.ANY", "MetricGroup": "Cor;Pipeline", @@ -191,63 +731,38 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { - "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;Flops_SMT", - "MetricName": "FLOPc_SMT" - }, - { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.THREAD )", + "MetricExpr": "((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)." }, { - "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )", - "MetricGroup": "Cor;Flops;HPC_SMT", - "MetricName": "FP_Arith_Utilization_SMT", - "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common). SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2 ) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", - "MetricExpr": "( 1 - ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)))) / ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) if ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)))) < ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) else 1 ) if 0 > 0.5 else 0", + "MetricExpr": "(1 - tma_core_bound / tma_ports_utilization if tma_core_bound < tma_ports_utilization else 1) if SMT_2T_Utilization > 0.5 else 0", "MetricGroup": "Cor;SMT", "MetricName": "Core_Bound_Likely" }, { - "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", - "MetricExpr": "( 1 - ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))))) / ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) if ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))))) < ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) else 1 ) if (1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 )) > 0.5 else 0", - "MetricGroup": "Cor;SMT_SMT", - "MetricName": "Core_Bound_Likely_SMT" - }, - { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, @@ -289,13 +804,13 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." @@ -316,14 +831,14 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." @@ -335,9 +850,9 @@ "MetricName": "IpSWPF" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -372,17 +887,11 @@ }, { "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", - "MetricExpr": "100 * ( (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (DSB2MITE_SWITCHES.PENALTY_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + ((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD))) * (( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / CPU_CLK_UNHALTED.THREAD / 2) / #((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD))) )", + "MetricExpr": "100 * (tma_fetch_latency * tma_dsb_switches / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches) + tma_fetch_bandwidth * tma_mite / (tma_dsb + tma_mite))", "MetricGroup": "DSBmiss;Fed", "MetricName": "DSB_Misses" }, { - "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", - "MetricExpr": "100 * ( (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * (DSB2MITE_SWITCHES.PENALTY_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + ((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * (( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) / 2) / #((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) )", - "MetricGroup": "DSBmiss;Fed_SMT", - "MetricName": "DSB_Misses_SMT" - }, - { "BriefDescription": "Number of Instructions per non-speculative DSB miss (lower number means higher occurrence rate)", "MetricExpr": "INST_RETIRED.ANY / FRONTEND_RETIRED.ANY_DSB_MISS", "MetricGroup": "DSBmiss;Fed", @@ -396,17 +905,11 @@ }, { "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;BrMispredicts", "MetricName": "Branch_Misprediction_Cost" }, { - "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES", - "MetricGroup": "Bad;BrMispredicts_SMT", - "MetricName": "Branch_Misprediction_Cost_SMT" - }, - { "BriefDescription": "Fraction of branches that are non-taken conditionals", "MetricExpr": "BR_INST_RETIRED.NOT_TAKEN / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", @@ -414,102 +917,96 @@ }, { "BriefDescription": "Fraction of branches that are taken conditionals", - "MetricExpr": "( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", "MetricName": "Cond_TK" }, { "BriefDescription": "Fraction of branches that are CALL or RET", - "MetricExpr": "( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "CallRet" }, { "BriefDescription": "Fraction of branches that are unconditional (direct or indirect) jumps", - "MetricExpr": "(BR_INST_RETIRED.NEAR_TAKEN - ( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_TAKEN - (BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN) - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "Jump" }, { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L1 cache true misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.ALL_DEMAND_DATA_RD / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI_Load" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load" }, { "BriefDescription": "L2 cache hits per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( L2_RQSTS.REFERENCES - L2_RQSTS.MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricExpr": "1000 * (L2_RQSTS.REFERENCES - L2_RQSTS.MISS) / INST_RETIRED.ANY", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_All" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries)", "MetricExpr": "1000 * MEM_LOAD_RETIRED.FB_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "FB_HPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING + EPT.WALK_PENDING ) / ( 2 * CPU_CLK_UNHALTED.THREAD )", + "MetricExpr": "(ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING + EPT.WALK_PENDING) / (2 * CORE_CLKS)", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, { - "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING + EPT.WALK_PENDING ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )", - "MetricGroup": "Mem;MemoryTLB_SMT", - "MetricName": "Page_Walks_Utilization_SMT" - }, - { "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]", "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time", "MetricGroup": "Mem;MemoryBW", @@ -535,25 +1032,25 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data access bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * OFFCORE_REQUESTS.ALL_REQUESTS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Access_BW", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "L3_Cache_Access_BW_1T" }, @@ -565,26 +1062,26 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -602,7 +1099,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * ( arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@ ) / 1000000 / duration_time / 1000", + "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1000000 / duration_time / 1000", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, diff --git a/tools/perf/pmu-events/arch/x86/skylakex/skx-metrics.json b/tools/perf/pmu-events/arch/x86/skylakex/skx-metrics.json index 6a6764e1504b..bc8e42554096 100644 --- a/tools/perf/pmu-events/arch/x86/skylakex/skx-metrics.json +++ b/tools/perf/pmu-events/arch/x86/skylakex/skx-metrics.json @@ -1,148 +1,726 @@ [ { "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Frontend_Bound", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound." + "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. Sample with: FRONTEND_RETIRED.LATENCY_GE_4_PS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Frontend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: FRONTEND_RETIRED.LATENCY_GE_16_PS;FRONTEND_RETIRED.LATENCY_GE_8_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses", + "MetricExpr": "(ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@) / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses. Sample with: FRONTEND_RETIRED.L2_MISS_PS;FRONTEND_RETIRED.L1I_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "ICACHE_64B.IFTAG_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: FRONTEND_RETIRED.STLB_MISS_PS;FRONTEND_RETIRED.ITLB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "INT_MISC.CLEAR_RESTEER_CYCLES / CLKS + tma_unknown_branches", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears", + "MetricExpr": "(1 - (BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT))) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "9 * BACLEARS.ANY / CLKS", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: BACLEARS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty. Sample with: FRONTEND_RETIRED.DSB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "2 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "tma_frontend_bound - tma_fetch_latency", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend. Sample with: FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_2_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck. Sample with: FRONTEND_RETIRED.ANY_DSB_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where decoder-0 was the only active decoder", + "MetricExpr": "(cpu@INST_DECODED.DECODERS\\,cmask\\=1@ - cpu@INST_DECODED.DECODERS\\,cmask\\=2@) / CORE_CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_decoder0_alone", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Bad_Speculation", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example." + "MetricExpr": "(UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Bad_Speculation_SMT", - "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "tma_bad_speculation - tma_branch_mispredicts", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", - "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Backend_Bound", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound." + "MetricExpr": "1 - tma_frontend_bound - (UOPS_ISSUED.ANY + 4 * ((INT_MISC.RECOVERY_CYCLES_ANY / 2) if #SMT_on else INT_MISC.RECOVERY_CYCLES)) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_RETIRED.L1_HIT_PS;MEM_LOAD_RETIRED.FB_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "min(9 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE, max(CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS, 0)) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_INST_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_load - tma_load_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk", + "MetricExpr": "DTLB_LOAD_MISSES.WALK_ACTIVE / CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(12 * max(0, MEM_INST_RETIRED.LOCK_LOADS - L2_RQSTS.ALL_RFO) + (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES) * (11 * L2_RQSTS.RFO_HIT + min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO))) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_INST_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_INST_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "Load_Miss_Real_Latency * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) / ((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@)) * ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "((44 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM * (OFFCORE_RESPONSE.DEMAND_DATA_RD.L3_HIT.HITM_OTHER_CORE / (OFFCORE_RESPONSE.DEMAND_DATA_RD.L3_HIT.HITM_OTHER_CORE + OFFCORE_RESPONSE.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD))) + (44 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "(44 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT + MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM * (1 - (OFFCORE_RESPONSE.DEMAND_DATA_RD.L3_HIT.HITM_OTHER_CORE / (OFFCORE_RESPONSE.DEMAND_DATA_RD.L3_HIT.HITM_OTHER_CORE + OFFCORE_RESPONSE.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD)))) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "(17 * Average_Frequency) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "((OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2) if #SMT_on else OFFCORE_REQUESTS_BUFFER.SQ_FULL) / CORE_CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L3_MISS / CLKS + ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS) - tma_l2_bound)", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory", + "MetricExpr": "(59.5 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Server;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_local_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory. Caching will improve the latency and increase performance. Sample with: MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory", + "MetricExpr": "(127 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_dram", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues", + "MetricExpr": "((89.5 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM + (89.5 * Average_Frequency) * MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Server;Snoop;TopdownL5;tma_mem_latency_group", + "MetricName": "tma_remote_cache", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues. This is caused often due to non-optimal NUMA allocations. #link to NUMA article Sample with: MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM_PS;MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "EXE_ACTIVITY.BOUND_ON_STORES / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_INST_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 11 * (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES))) + (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "((110 * Average_Frequency) * (OFFCORE_RESPONSE.DEMAND_RFO.L3_MISS.REMOTE_HITM + OFFCORE_RESPONSE.PF_L2_RFO.L3_MISS.REMOTE_HITM) + (47.5 * Average_Frequency) * (OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.HITM_OTHER_CORE + OFFCORE_RESPONSE.PF_L2_RFO.L3_HIT.HITM_OTHER_CORE)) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM_PS;OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "MEM_INST_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_INST_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(9 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE) / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_INST_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_store - tma_store_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk", + "MetricExpr": "DTLB_STORE_MISSES.WALK_ACTIVE / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "tma_backend_bound - tma_memory_bound", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.DIVIDER_ACTIVE / CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_ACTIVE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "(EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL)) / CLKS if (ARITH.DIVIDER_ACTIVE < (CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY)) else (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "(UOPS_EXECUTED.CORE_CYCLES_NONE / 2 if #SMT_on else CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations", + "MetricExpr": "PARTIAL_RAT_STALLS.SCOREBOARD / CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_serializing_operation", + "PublicDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out-of-order execution which may limit performance. Sample with: PARTIAL_RAT_STALLS.SCOREBOARD", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Backend_Bound_SMT", - "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued", + "MetricExpr": "CLKS * UOPS_ISSUED.VECTOR_WIDTH_MISMATCH / UOPS_ISSUED.ANY", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_mixing_vectors", + "PublicDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "((UOPS_EXECUTED.CORE_CYCLES_GE_1 - UOPS_EXECUTED.CORE_CYCLES_GE_2) / 2 if #SMT_on else EXE_ACTIVITY.1_PORTS_UTIL) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "((UOPS_EXECUTED.CORE_CYCLES_GE_2 - UOPS_EXECUTED.CORE_CYCLES_GE_3) / 2 if #SMT_on else EXE_ACTIVITY.2_PORTS_UTIL) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).", + "MetricExpr": "(UOPS_EXECUTED.CORE_CYCLES_GE_3 / 2 if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_3) / CORE_CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_0 + UOPS_DISPATCHED_PORT.PORT_1 + UOPS_DISPATCHED_PORT.PORT_5 + UOPS_DISPATCHED_PORT.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED_PORT.PORT_0", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_1", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED_PORT.PORT_6", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "(UOPS_DISPATCHED_PORT.PORT_2 + UOPS_DISPATCHED_PORT.PORT_3 + UOPS_DISPATCHED_PORT.PORT_7 - UOPS_DISPATCHED_PORT.PORT_4) / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_2", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_2 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_2", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 ([SNB+]Loads and Store-address; [ICL+] Loads) Sample with: UOPS_DISPATCHED_PORT.PORT_3", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_3 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_load_op_utilization_group", + "MetricName": "tma_port_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data) Sample with: UOPS_DISPATCHED_PORT.PORT_4", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_4 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_4", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 7 ([HSW+]simple Store-address) Sample with: UOPS_DISPATCHED_PORT.PORT_7", + "MetricExpr": "UOPS_DISPATCHED_PORT.PORT_7 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_store_op_utilization_group", + "MetricName": "tma_port_7", + "ScaleUnit": "100%" }, { "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "TopdownL1", - "MetricName": "Retiring", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. " + "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.RETIRE_SLOTS", + "ScaleUnit": "100%" }, { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))", - "MetricGroup": "TopdownL1_SMT", - "MetricName": "Retiring_SMT", - "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. SMT version; use when SMT is enabled and measuring per logical CPU." + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "tma_retiring - tma_heavy_operations", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" }, { - "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", - "MetricExpr": "100 * ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) )", - "MetricGroup": "Bad;BadSpec;BrMispredicts", - "MetricName": "Mispredictions" + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "tma_retiring * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_512b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", + "MetricExpr": "tma_light_operations * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_memory_operations", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions", + "MetricExpr": "tma_light_operations * UOPS_RETIRED.MACRO_FUSED / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fused_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions. The instruction pairs of CMP+JCC or DEC+JCC are commonly used examples.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused", + "MetricExpr": "tma_light_operations * (BR_INST_RETIRED.ALL_BRANCHES - UOPS_RETIRED.MACRO_FUSED) / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_non_fused_branches", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused. Non-conditional branches like direct JMP or CALL would count here. Can be used to examine fusible conditional jumps that were not fused.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions", + "MetricExpr": "tma_light_operations * INST_RETIRED.NOP / UOPS_RETIRED.RETIRE_SLOTS", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_nop_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body. Sample with: INST_RETIRED.NOP", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", + "MetricExpr": "max(0, tma_light_operations - (tma_fp_arith + tma_memory_operations + tma_fused_instructions + tma_non_fused_branches + tma_nop_instructions))", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_other_light_ops", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY) / SLOTS", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops", + "MetricExpr": "tma_heavy_operations - tma_microcode_sequencer", + "MetricGroup": "TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_few_uops_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "(UOPS_RETIRED.RETIRE_SLOTS / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * (FP_ASSIST.ANY + OTHER_ASSISTS.ANY) / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: OTHER_ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" }, { "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", - "MetricExpr": "100 * ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) )", - "MetricGroup": "Bad;BadSpec;BrMispredicts_SMT", - "MetricName": "Mispredictions_SMT" + "MetricExpr": "100 * (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches))", + "MetricGroup": "Bad;BadSpec;BrMispredicts", + "MetricName": "Mispredictions" }, { "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) * ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD) / #(CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (OFFCORE_REQUESTS_BUFFER.SQ_FULL / CPU_CLK_UNHALTED.THREAD) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) ) + ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( ((L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )) * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CPU_CLK_UNHALTED.THREAD) / #(max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) ", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_mem_bandwidth / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_sq_full / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full))) + (tma_l1_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_fb_full / (tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) ", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "Memory_Bandwidth" }, { - "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD) / #(CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (( OFFCORE_REQUESTS_BUFFER.SQ_FULL / 2 ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) ) + ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( ((L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )) * cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ / CPU_CLK_UNHALTED.THREAD) / #(max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) ", - "MetricGroup": "Mem;MemoryBW;Offcore_SMT", - "MetricName": "Memory_Bandwidth_SMT" - }, - { "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) * ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD ) / CPU_CLK_UNHALTED.THREAD - (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD)) / #(CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (( (20.5 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) - (3.5 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) ) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CPU_CLK_UNHALTED.THREAD) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) + ( (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD)) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) )", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_mem_latency / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_l3_hit_latency / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full)) + (tma_l2_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)))", "MetricGroup": "Mem;MemoryLat;Offcore", "MetricName": "Memory_Latency" }, { - "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * ( ( (CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (min( CPU_CLK_UNHALTED.THREAD , OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD ) / CPU_CLK_UNHALTED.THREAD - (min( CPU_CLK_UNHALTED.THREAD , cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@ ) / CPU_CLK_UNHALTED.THREAD)) / #(CYCLE_ACTIVITY.STALLS_L3_MISS / CPU_CLK_UNHALTED.THREAD + (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD) - (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD))) ) + ( (( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (( (20.5 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) - (3.5 * ((CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time)) ) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CPU_CLK_UNHALTED.THREAD) / #(( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / CPU_CLK_UNHALTED.THREAD) ) + ( (( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) / ( (MEM_LOAD_RETIRED.L2_HIT * ( 1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) )) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=1@ ) ) * (( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / CPU_CLK_UNHALTED.THREAD)) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) )", - "MetricGroup": "Mem;MemoryLat;Offcore_SMT", - "MetricName": "Memory_Latency_SMT" - }, - { "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) * ( ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (min( 9 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE , max( CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS , 0 ) ) / CPU_CLK_UNHALTED.THREAD) / (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) + ( (EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) ) * ( (( 9 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE ) / CPU_CLK_UNHALTED.THREAD) / #(EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) ) ) ", + "MetricExpr": "100 * tma_memory_bound * ((tma_l1_bound / max(tma_memory_bound, tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_dtlb_load / max(tma_l1_bound, tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) + (tma_store_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_dtlb_store / (tma_dtlb_store + tma_false_sharing + tma_split_stores + tma_store_latency))) ", "MetricGroup": "Mem;MemoryTLB;Offcore", "MetricName": "Memory_Data_TLBs" }, { - "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", - "MetricExpr": "100 * ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * ( ( (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) / ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (min( 9 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE , max( CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS , 0 ) ) / CPU_CLK_UNHALTED.THREAD) / (max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / CPU_CLK_UNHALTED.THREAD , 0 )) ) + ( (EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) / #((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) ) * ( (( 9 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / #(EXE_ACTIVITY.BOUND_ON_STORES / CPU_CLK_UNHALTED.THREAD) ) ) ", - "MetricGroup": "Mem;MemoryTLB;Offcore_SMT", - "MetricName": "Memory_Data_TLBs_SMT" - }, - { "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.CONDITIONAL + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - ( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) - 2 * BR_INST_RETIRED.NEAR_CALL) ) / (4 * CPU_CLK_UNHALTED.THREAD))", + "MetricExpr": "100 * ((BR_INST_RETIRED.CONDITIONAL + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - (BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN) - 2 * BR_INST_RETIRED.NEAR_CALL)) / SLOTS)", "MetricGroup": "Ret", "MetricName": "Branching_Overhead" }, { - "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.CONDITIONAL + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - ( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) - 2 * BR_INST_RETIRED.NEAR_CALL) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))", - "MetricGroup": "Ret_SMT", - "MetricName": "Branching_Overhead_SMT" - }, - { "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", - "MetricExpr": "100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD))", + "MetricExpr": "100 * tma_fetch_latency * (tma_itlb_misses + tma_icache_misses + tma_unknown_branches) / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)", "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB", "MetricName": "Big_Code" }, { - "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", - "MetricExpr": "100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))", - "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB_SMT", - "MetricName": "Big_Code_SMT" - }, - { "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", - "MetricExpr": "100 * ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) - (100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)))", + "MetricExpr": "100 * (tma_frontend_bound - tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) - Big_Code", "MetricGroup": "Fed;FetchBW;Frontend", "MetricName": "Instruction_Fetch_BW" }, { - "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", - "MetricExpr": "100 * ( (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) - (100 * (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=1\\,edge@ ) / CPU_CLK_UNHALTED.THREAD) + (9 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))))", - "MetricGroup": "Fed;FetchBW;Frontend_SMT", - "MetricName": "Instruction_Fetch_BW_SMT" - }, - { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, @@ -159,6 +737,12 @@ "MetricName": "UpTB" }, { + "BriefDescription": "Cycles Per Instruction (per Logical Processor)", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", + "MetricName": "CPI" + }, + { "BriefDescription": "Per-Logical Processor actual clocks when the Logical Processor is active.", "MetricExpr": "CPU_CLK_UNHALTED.THREAD", "MetricGroup": "Pipeline", @@ -166,17 +750,11 @@ }, { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "TmaL1", + "MetricExpr": "4 * CORE_CLKS", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { - "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", - "MetricExpr": "4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "TmaL1_SMT", - "MetricName": "SLOTS_SMT" - }, - { "BriefDescription": "The ratio of Executed- by Issued-Uops", "MetricExpr": "UOPS_EXECUTED.THREAD / UOPS_ISSUED.ANY", "MetricGroup": "Cor;Pipeline", @@ -185,63 +763,38 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { - "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;SMT;TmaL1_SMT", - "MetricName": "CoreIPC_SMT" - }, - { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / CPU_CLK_UNHALTED.THREAD", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { - "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Ret;Flops_SMT", - "MetricName": "FLOPc_SMT" - }, - { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.THREAD )", + "MetricExpr": "((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)." }, { - "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )", - "MetricGroup": "Cor;Flops;HPC_SMT", - "MetricName": "FP_Arith_Utilization_SMT", - "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common). SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2 ) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", - "MetricExpr": "( 1 - ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)))) / ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) if ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - ( UOPS_ISSUED.ANY + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD)))) < ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * CPU_CLK_UNHALTED.THREAD)) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) else 1 ) if 0 > 0.5 else 0", + "MetricExpr": "(1 - tma_core_bound / tma_ports_utilization if tma_core_bound < tma_ports_utilization else 1) if SMT_2T_Utilization > 0.5 else 0", "MetricGroup": "Cor;SMT", "MetricName": "Core_Bound_Likely" }, { - "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", - "MetricExpr": "( 1 - ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))))) / ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) if ((1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ((( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * (1 - (IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - ( UOPS_ISSUED.ANY + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))))) < ((EXE_ACTIVITY.EXE_BOUND_0_PORTS + (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL)) / CPU_CLK_UNHALTED.THREAD if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else (EXE_ACTIVITY.1_PORTS_UTIL + (UOPS_RETIRED.RETIRE_SLOTS / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * EXE_ACTIVITY.2_PORTS_UTIL) / CPU_CLK_UNHALTED.THREAD) else 1 ) if (1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 )) > 0.5 else 0", - "MetricGroup": "Cor;SMT_SMT", - "MetricName": "Core_Bound_Likely_SMT" - }, - { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", - "MetricExpr": "( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "((CPU_CLK_UNHALTED.THREAD / 2) * (1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK)) if #core_wide < 1 else (CPU_CLK_UNHALTED.THREAD_ANY / 2) if #SMT_on else CLKS", "MetricGroup": "SMT", "MetricName": "CORE_CLKS" }, @@ -283,13 +836,13 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." @@ -310,21 +863,21 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX512", "PublicDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." @@ -336,9 +889,9 @@ "MetricName": "IpSWPF" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { @@ -373,17 +926,11 @@ }, { "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", - "MetricExpr": "100 * ( (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (DSB2MITE_SWITCHES.PENALTY_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) + ((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD))) * (( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / CPU_CLK_UNHALTED.THREAD / 2) / #((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD))) )", + "MetricExpr": "100 * (tma_fetch_latency * tma_dsb_switches / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches) + tma_fetch_bandwidth * tma_mite / (tma_dsb + tma_mite))", "MetricGroup": "DSBmiss;Fed", "MetricName": "DSB_Misses" }, { - "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", - "MetricExpr": "100 * ( (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * (DSB2MITE_SWITCHES.PENALTY_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) + ((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) * (( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) / 2) / #((IDQ_UOPS_NOT_DELIVERED.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) - (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) )", - "MetricGroup": "DSBmiss;Fed_SMT", - "MetricName": "DSB_Misses_SMT" - }, - { "BriefDescription": "Number of Instructions per non-speculative DSB miss (lower number means higher occurrence rate)", "MetricExpr": "INST_RETIRED.ANY / FRONTEND_RETIRED.ANY_DSB_MISS", "MetricGroup": "DSBmiss;Fed", @@ -397,17 +944,11 @@ }, { "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;BrMispredicts", "MetricName": "Branch_Misprediction_Cost" }, { - "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", - "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * INT_MISC.CLEAR_RESTEER_CYCLES / CPU_CLK_UNHALTED.THREAD) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES", - "MetricGroup": "Bad;BrMispredicts_SMT", - "MetricName": "Branch_Misprediction_Cost_SMT" - }, - { "BriefDescription": "Fraction of branches that are non-taken conditionals", "MetricExpr": "BR_INST_RETIRED.NOT_TAKEN / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", @@ -415,102 +956,96 @@ }, { "BriefDescription": "Fraction of branches that are taken conditionals", - "MetricExpr": "( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", "MetricName": "Cond_TK" }, { "BriefDescription": "Fraction of branches that are CALL or RET", - "MetricExpr": "( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "CallRet" }, { "BriefDescription": "Fraction of branches that are unconditional (direct or indirect) jumps", - "MetricExpr": "(BR_INST_RETIRED.NEAR_TAKEN - ( BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN ) - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_TAKEN - (BR_INST_RETIRED.CONDITIONAL - BR_INST_RETIRED.NOT_TAKEN) - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "Jump" }, { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L1 cache true misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.ALL_DEMAND_DATA_RD / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI_Load" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load" }, { "BriefDescription": "L2 cache hits per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( L2_RQSTS.REFERENCES - L2_RQSTS.MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricExpr": "1000 * (L2_RQSTS.REFERENCES - L2_RQSTS.MISS) / INST_RETIRED.ANY", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_All" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries)", "MetricExpr": "1000 * MEM_LOAD_RETIRED.FB_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "FB_HPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING + EPT.WALK_PENDING ) / ( 2 * CPU_CLK_UNHALTED.THREAD )", + "MetricExpr": "(ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING + EPT.WALK_PENDING) / (2 * CORE_CLKS)", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, { - "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING + EPT.WALK_PENDING ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )", - "MetricGroup": "Mem;MemoryTLB_SMT", - "MetricName": "Page_Walks_Utilization_SMT" - }, - { "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]", "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time", "MetricGroup": "Mem;MemoryBW", @@ -536,37 +1071,37 @@ }, { "BriefDescription": "Rate of silent evictions from the L2 cache per Kilo instruction where the evicted lines are dropped (no writeback to L3 or memory)", - "MetricExpr": "1000 * L2_LINES_OUT.SILENT / INST_RETIRED.ANY", + "MetricExpr": "1000 * L2_LINES_OUT.SILENT / Instructions", "MetricGroup": "L2Evicts;Mem;Server", "MetricName": "L2_Evictions_Silent_PKI" }, { "BriefDescription": "Rate of non silent evictions from the L2 cache per Kilo instruction", - "MetricExpr": "1000 * L2_LINES_OUT.NON_SILENT / INST_RETIRED.ANY", + "MetricExpr": "1000 * L2_LINES_OUT.NON_SILENT / Instructions", "MetricGroup": "L2Evicts;Mem;Server", "MetricName": "L2_Evictions_NonSilent_PKI" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data access bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * OFFCORE_REQUESTS.ALL_REQUESTS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Access_BW", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "L3_Cache_Access_BW_1T" }, @@ -578,68 +1113,47 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0", - "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / 2 / CORE_CLKS if #SMT_on else CORE_POWER.LVL0_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License0_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. This includes non-AVX codes, SSE, AVX 128-bit, and low-current AVX 256-bit codes." }, { - "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / 2 / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Power_SMT", - "MetricName": "Power_License0_Utilization_SMT", - "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. This includes non-AVX codes, SSE, AVX 128-bit, and low-current AVX 256-bit codes. SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1", - "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / 2 / CORE_CLKS if #SMT_on else CORE_POWER.LVL1_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License1_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1. This includes high current AVX 256-bit instructions as well as low current AVX 512-bit instructions." }, { - "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1. SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / 2 / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Power_SMT", - "MetricName": "Power_License1_Utilization_SMT", - "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1. This includes high current AVX 256-bit instructions as well as low current AVX 512-bit instructions. SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX)", - "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / 2 / CORE_CLKS if #SMT_on else CORE_POWER.LVL2_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License2_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). This includes high current AVX 512-bit instructions." }, { - "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). SMT version; use when SMT is enabled and measuring per logical CPU.", - "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / 2 / ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", - "MetricGroup": "Power_SMT", - "MetricName": "Power_License2_Utilization_SMT", - "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). This includes high current AVX 512-bit instructions. SMT version; use when SMT is enabled and measuring per logical CPU." - }, - { "BriefDescription": "Fraction of cycles where both hardware Logical Processors were active", - "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / ( CPU_CLK_UNHALTED.REF_XCLK_ANY / 2 ) if #SMT_on else 0", + "MetricExpr": "1 - CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / (CPU_CLK_UNHALTED.REF_XCLK_ANY / 2) if #SMT_on else 0", "MetricGroup": "SMT", "MetricName": "SMT_2T_Utilization" }, @@ -657,13 +1171,13 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "( 64 * ( uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@ ) / 1000000000 ) / duration_time", + "MetricExpr": "(64 * (uncore_imc@cas_count_read@ + uncore_imc@cas_count_write@) / 1000000000) / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, { "BriefDescription": "Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches", - "MetricExpr": "1000000000 * ( cha@event\\=0x36\\,umask\\=0x21\\,config\\=0x40433@ / cha@event\\=0x35\\,umask\\=0x21\\,config\\=0x40433@ ) / ( cha_0@event\\=0x0@ / duration_time )", + "MetricExpr": "1000000000 * (cha@event\\=0x36\\,umask\\=0x21\\,config\\=0x40433@ / cha@event\\=0x35\\,umask\\=0x21\\,config\\=0x40433@) / (Socket_CLKS / duration_time)", "MetricGroup": "Mem;MemoryLat;SoC", "MetricName": "MEM_Read_Latency" }, @@ -675,20 +1189,20 @@ }, { "BriefDescription": "Average latency of data read request to external DRAM memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches", - "MetricExpr": "1000000000 * ( UNC_M_RPQ_OCCUPANCY / UNC_M_RPQ_INSERTS ) / imc_0@event\\=0x0@", - "MetricGroup": "Mem;MemoryLat;SoC;Server", + "MetricExpr": "1000000000 * (UNC_M_RPQ_OCCUPANCY / UNC_M_RPQ_INSERTS) / imc_0@event\\=0x0@", + "MetricGroup": "Mem;MemoryLat;Server;SoC", "MetricName": "MEM_DRAM_Read_Latency" }, { "BriefDescription": "Average IO (network or disk) Bandwidth Use for Writes [GB / sec]", - "MetricExpr": "( UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART0 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART1 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART2 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART3 ) * 4 / 1000000000 / duration_time", - "MetricGroup": "IoBW;Mem;SoC;Server", + "MetricExpr": "(UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART0 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART1 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART2 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART3) * 4 / 1000000000 / duration_time", + "MetricGroup": "IoBW;Mem;Server;SoC", "MetricName": "IO_Write_BW" }, { "BriefDescription": "Average IO (network or disk) Bandwidth Use for Reads [GB / sec]", - "MetricExpr": "( UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART0 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART1 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART2 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART3 ) * 4 / 1000000000 / duration_time", - "MetricGroup": "IoBW;Mem;SoC;Server", + "MetricExpr": "(UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART0 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART1 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART2 + UNC_IIO_DATA_REQ_OF_CPU.MEM_WRITE.PART3) * 4 / 1000000000 / duration_time", + "MetricGroup": "IoBW;Mem;Server;SoC", "MetricName": "IO_Read_BW" }, { @@ -698,12 +1212,6 @@ "MetricName": "Socket_CLKS" }, { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "cha_0@event\\=0x0@ / #num_dies / duration_time / 1000000000", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, - { "BriefDescription": "Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]", "MetricExpr": "INST_RETIRED.ANY / BR_INST_RETIRED.FAR_BRANCH:u", "MetricGroup": "Branches;OS", @@ -752,11 +1260,10 @@ "MetricName": "C7_Pkg_Residency" }, { - "BriefDescription": "Percentage of time spent in the active CPU power state C0", - "MetricExpr": "100 * CPU_CLK_UNHALTED.REF_TSC / TSC", - "MetricGroup": "", - "MetricName": "cpu_utilization_percent", - "ScaleUnit": "1%" + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "Socket_CLKS / #num_dies / duration_time / 1000000000", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ" }, { "BriefDescription": "CPU operating frequency (in GHz)", @@ -766,13 +1273,6 @@ "ScaleUnit": "1GHz" }, { - "BriefDescription": "Cycles per instruction retired; indicating how much time each executed instruction took; in units of cycles.", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY", - "MetricGroup": "", - "MetricName": "cpi", - "ScaleUnit": "1per_instr" - }, - { "BriefDescription": "The ratio of number of completed memory load instructions to the total number completed instructions", "MetricExpr": "MEM_INST_RETIRED.ALL_LOADS / INST_RETIRED.ANY", "MetricGroup": "", @@ -790,7 +1290,7 @@ "BriefDescription": "Ratio of number of requests missing L1 data cache (includes data+rfo w/ prefetches) to the total number of completed instructions", "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l1d_mpi_includes_data_plus_rfo_with_prefetches", + "MetricName": "l1d_mpi", "ScaleUnit": "1per_instr" }, { @@ -818,7 +1318,7 @@ "BriefDescription": "Ratio of number of requests missing L2 cache (includes code+data+rfo w/ prefetches) to the total number of completed instructions", "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches", + "MetricName": "l2_mpi", "ScaleUnit": "1per_instr" }, { @@ -850,57 +1350,78 @@ "ScaleUnit": "1per_instr" }, { + "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) in nano seconds", + "MetricExpr": "( 1000000000 * ( cha@unc_cha_tor_occupancy.ia_miss\\,config1\\=0x4043300000000@ / cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043300000000@ ) / ( UNC_CHA_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) ) ) * duration_time", + "MetricGroup": "", + "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency", + "ScaleUnit": "1ns" + }, + { + "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to local memory in nano seconds", + "MetricExpr": "( 1000000000 * ( cha@unc_cha_tor_occupancy.ia_miss\\,config1\\=0x4043200000000@ / cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043200000000@ ) / ( UNC_CHA_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) ) ) * duration_time", + "MetricGroup": "", + "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_local_requests", + "ScaleUnit": "1ns" + }, + { + "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to remote memory in nano seconds", + "MetricExpr": "( 1000000000 * ( cha@unc_cha_tor_occupancy.ia_miss\\,config1\\=0x4043100000000@ / cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043100000000@ ) / ( UNC_CHA_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) ) ) * duration_time", + "MetricGroup": "", + "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_remote_requests", + "ScaleUnit": "1ns" + }, + { "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the ITLB (Instruction TLB) and further levels of TLB.", "MetricExpr": "ITLB_MISSES.WALK_COMPLETED / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "itlb_2nd_level_mpi", + "MetricName": "itlb_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for 2 megabyte and 4 megabyte page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the Instruction Translation Lookaside Buffer (ITLB) and further levels of TLB.", "MetricExpr": "ITLB_MISSES.WALK_COMPLETED_2M_4M / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "itlb_2nd_level_large_page_mpi", + "MetricName": "itlb_large_page_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by demand data loads to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.", "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "dtlb_2nd_level_load_mpi", + "MetricName": "dtlb_load_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for 2 megabyte page sizes) caused by demand data loads to the total number of completed instructions. This implies it missed in the Data Translation Lookaside Buffer (DTLB) and further levels of TLB.", "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED_2M_4M / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "dtlb_2nd_level_2mb_large_page_load_mpi", + "MetricName": "dtlb_2mb_large_page_load_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by demand data stores to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.", "MetricExpr": "DTLB_STORE_MISSES.WALK_COMPLETED / INST_RETIRED.ANY", "MetricGroup": "", - "MetricName": "dtlb_2nd_level_store_mpi", + "MetricName": "dtlb_store_mpi", "ScaleUnit": "1per_instr" }, { "BriefDescription": "Memory read that miss the last level cache (LLC) addressed to local DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", "MetricExpr": "100 * cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043200000000@ / ( cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043200000000@ + cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043100000000@ )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_local_dram", + "MetricName": "numa_reads_addressed_to_local_dram", "ScaleUnit": "1%" }, { "BriefDescription": "Memory reads that miss the last level cache (LLC) addressed to remote DRAM as a percentage of total memory read accesses, does not include LLC prefetches.", "MetricExpr": "100 * cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043100000000@ / ( cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043200000000@ + cha@unc_cha_tor_inserts.ia_miss\\,config1\\=0x4043100000000@ )", "MetricGroup": "", - "MetricName": "numa_percent_reads_addressed_to_remote_dram", + "MetricName": "numa_reads_addressed_to_remote_dram", "ScaleUnit": "1%" }, { "BriefDescription": "Uncore operating frequency in GHz", - "MetricExpr": "( UNC_CHA_CLOCKTICKS / ( source_count(UNC_CHA_CLOCKTICKS) * #num_packages ) / 1000000000) / duration_time", + "MetricExpr": "( UNC_CHA_CLOCKTICKS / ( #num_cores / #num_packages * #num_packages ) / 1000000000) / duration_time", "MetricGroup": "", "MetricName": "uncore_frequency", "ScaleUnit": "1GHz" @@ -909,7 +1430,7 @@ "BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data transmit bandwidth (MB/sec)", "MetricExpr": "( UNC_UPI_TxL_FLITS.ALL_DATA * (64 / 9.0) / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "upi_data_transmit_bw_only_data", + "MetricName": "upi_data_transmit_bw", "ScaleUnit": "1MB/s" }, { @@ -937,35 +1458,35 @@ "BriefDescription": "Bandwidth of IO reads that are initiated by end device controllers that are requesting memory from the CPU.", "MetricExpr": "(( UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART0 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART1 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART2 + UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART3 ) * 4 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_read", + "MetricName": "io_bandwidth_disk_or_network_writes", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Bandwidth of IO writes that are initiated by end device controllers that are writing memory to the CPU.", "MetricExpr": "(( UNC_IIO_PAYLOAD_BYTES_IN.MEM_WRITE.PART0 + UNC_IIO_PAYLOAD_BYTES_IN.MEM_WRITE.PART1 + UNC_IIO_PAYLOAD_BYTES_IN.MEM_WRITE.PART2 + UNC_IIO_PAYLOAD_BYTES_IN.MEM_WRITE.PART3 ) * 4 / 1000000) / duration_time", "MetricGroup": "", - "MetricName": "io_bandwidth_write", + "MetricName": "io_bandwidth_disk_or_network_reads", "ScaleUnit": "1MB/s" }, { "BriefDescription": "Uops delivered from decoded instruction cache (decoded stream buffer or DSB) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.DSB_UOPS / UOPS_ISSUED.ANY )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_decoded_icache_dsb", + "MetricName": "percent_uops_delivered_from_decoded_icache", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from legacy decode pipeline (Micro-instruction Translation Engine or MITE) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.MITE_UOPS / UOPS_ISSUED.ANY )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline_mite", + "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline", "ScaleUnit": "1%" }, { "BriefDescription": "Uops delivered from microcode sequencer (MS) as a percent of total uops delivered to Instruction Decode Queue", "MetricExpr": "100 * ( IDQ.MS_UOPS / UOPS_ISSUED.ANY )", "MetricGroup": "", - "MetricName": "percent_uops_delivered_from_microcode_sequencer_ms", + "MetricName": "percent_uops_delivered_from_microcode_sequencer", "ScaleUnit": "1%" }, { @@ -988,250 +1509,5 @@ "MetricGroup": "", "MetricName": "llc_miss_remote_memory_bandwidth_read", "ScaleUnit": "1MB/s" - }, - { - "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.", - "MetricExpr": "100 * ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1;PGO", - "MetricName": "tma_frontend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.", - "MetricExpr": "100 * ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "Frontend;TmaL2;m_tma_frontend_bound_percent", - "MetricName": "tma_fetch_latency_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.", - "MetricExpr": "100 * ( ( ICACHE_16B.IFDATA_STALL + 2 * cpu@ICACHE_16B.IFDATA_STALL\\,cmask\\=0x1\\,edge\\=0x1@ ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_icache_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.", - "MetricExpr": "100 * ( ICACHE_64B.IFTAG_STALL / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_itlb_misses_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.", - "MetricExpr": "100 * ( INT_MISC.CLEAR_RESTEER_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) + ( ( 9 ) * BACLEARS.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) )", - "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_branch_resteers_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.", - "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_dsb_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", - "MetricExpr": "100 * ( ILD_STALL.LCP / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_lcp_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.", - "MetricExpr": "100 * ( ( 2 ) * IDQ.MS_SWITCHES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent", - "MetricName": "tma_ms_switches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.", - "MetricExpr": "100 * ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent", - "MetricName": "tma_fetch_bandwidth_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.", - "MetricExpr": "100 * ( ( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )", - "MetricGroup": "DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent", - "MetricName": "tma_mite_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", - "MetricExpr": "100 * ( ( IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )", - "MetricGroup": "DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent", - "MetricName": "tma_dsb_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", - "MetricExpr": "100 * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_bad_speculation_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.", - "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent", - "MetricName": "tma_branch_mispredicts_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.", - "MetricExpr": "100 * ( ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )", - "MetricGroup": "BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent", - "MetricName": "tma_machine_clears_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.", - "MetricExpr": "100 * ( 1 - ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( UOPS_ISSUED.ANY + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_backend_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", - "MetricExpr": "100 * ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / ( CYCLE_ACTIVITY.STALLS_TOTAL + ( EXE_ACTIVITY.1_PORTS_UTIL + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * EXE_ACTIVITY.2_PORTS_UTIL ) + EXE_ACTIVITY.BOUND_ON_STORES ) ) * ( 1 - ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( UOPS_ISSUED.ANY + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent", - "MetricName": "tma_memory_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.", - "MetricExpr": "100 * ( max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) , 0 ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l1_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) / ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=0x1@ ) ) * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l2_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.", - "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_l3_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.", - "MetricExpr": "100 * ( min( ( ( CYCLE_ACTIVITY.STALLS_L3_MISS / ( CPU_CLK_UNHALTED.THREAD ) + ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) - ( ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) / ( ( MEM_LOAD_RETIRED.L2_HIT * ( 1 + ( MEM_LOAD_RETIRED.FB_HIT / ( MEM_LOAD_RETIRED.L1_MISS ) ) ) ) + cpu@L1D_PEND_MISS.FB_FULL\\,cmask\\=0x1@ ) ) * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) , ( 1 ) ) )", - "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_dram_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.", - "MetricExpr": "100 * ( EXE_ACTIVITY.BOUND_ON_STORES / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent", - "MetricName": "tma_store_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", - "MetricExpr": "100 * ( ( 1 - ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( UOPS_ISSUED.ANY + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES ) / ( CYCLE_ACTIVITY.STALLS_TOTAL + ( EXE_ACTIVITY.1_PORTS_UTIL + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * EXE_ACTIVITY.2_PORTS_UTIL ) + EXE_ACTIVITY.BOUND_ON_STORES ) ) * ( 1 - ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( UOPS_ISSUED.ANY + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )", - "MetricGroup": "Backend;TmaL2;Compute;m_tma_backend_bound_percent", - "MetricName": "tma_core_bound_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.", - "MetricExpr": "100 * ( ARITH.DIVIDER_ACTIVE / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "TmaL3;m_tma_core_bound_percent", - "MetricName": "tma_divider_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", - "MetricExpr": "100 * ( ( EXE_ACTIVITY.EXE_BOUND_0_PORTS + ( EXE_ACTIVITY.1_PORTS_UTIL + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * EXE_ACTIVITY.2_PORTS_UTIL ) ) / ( CPU_CLK_UNHALTED.THREAD ) if ( ARITH.DIVIDER_ACTIVE < ( CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY ) ) else ( EXE_ACTIVITY.1_PORTS_UTIL + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * EXE_ACTIVITY.2_PORTS_UTIL ) / ( CPU_CLK_UNHALTED.THREAD ) )", - "MetricGroup": "PortsUtil;TmaL3;m_tma_core_bound_percent", - "MetricName": "tma_ports_utilization_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. ", - "MetricExpr": "100 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "TmaL1", - "MetricName": "tma_retiring_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved.", - "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", - "MetricName": "tma_light_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) )", - "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_fp_arith_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_memory_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions. The instruction pairs of CMP+JCC or DEC+JCC are commonly used examples.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * UOPS_RETIRED.MACRO_FUSED / ( UOPS_RETIRED.RETIRE_SLOTS ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_fused_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused. Non-conditional branches like direct JMP or CALL would count here. Can be used to examine fusible conditional jumps that were not fused.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * ( BR_INST_RETIRED.ALL_BRANCHES - UOPS_RETIRED.MACRO_FUSED ) / ( UOPS_RETIRED.RETIRE_SLOTS ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_non_fused_branches_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * INST_RETIRED.NOP / ( UOPS_RETIRED.RETIRE_SLOTS ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_nop_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", - "MetricExpr": "100 * ( max( 0 , ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) - ( ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) ) + ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY ) + ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * UOPS_RETIRED.MACRO_FUSED / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * ( BR_INST_RETIRED.ALL_BRANCHES - UOPS_RETIRED.MACRO_FUSED ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) * INST_RETIRED.NOP / ( UOPS_RETIRED.RETIRE_SLOTS ) ) ) ) )", - "MetricGroup": "Pipeline;TmaL3;m_tma_light_operations_percent", - "MetricName": "tma_other_light_ops_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", - "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", - "MetricName": "tma_heavy_operations_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", - "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) + UOPS_RETIRED.MACRO_FUSED - INST_RETIRED.ANY ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )", - "MetricGroup": "TmaL3;m_tma_heavy_operations_percent", - "MetricName": "tma_few_uops_instructions_percent", - "ScaleUnit": "1%" - }, - { - "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided.", - "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )", - "MetricGroup": "MicroSeq;TmaL3;m_tma_heavy_operations_percent", - "MetricName": "tma_microcode_sequencer_percent", - "ScaleUnit": "1%" } ] diff --git a/tools/perf/pmu-events/arch/x86/skylakex/uncore-memory.json b/tools/perf/pmu-events/arch/x86/skylakex/uncore-memory.json index 0746fcf2ebd9..62941146e396 100644 --- a/tools/perf/pmu-events/arch/x86/skylakex/uncore-memory.json +++ b/tools/perf/pmu-events/arch/x86/skylakex/uncore-memory.json @@ -27,20 +27,19 @@ "Unit": "iMC" }, { - "BriefDescription": "read requests to memory controller. Derived from unc_m_cas_count.rd", + "BriefDescription": "All DRAM Read CAS Commands issued (including underfills)", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "LLC_MISSES.MEM_READ", + "EventName": "UNC_M_CAS_COUNT.RD", "PerPkg": "1", - "ScaleUnit": "64Bytes", "UMask": "0x3", "Unit": "iMC" }, { - "BriefDescription": "read requests to memory controller", + "BriefDescription": "read requests to memory controller. Derived from unc_m_cas_count.rd", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "UNC_M_CAS_COUNT.RD", + "EventName": "LLC_MISSES.MEM_READ", "PerPkg": "1", "ScaleUnit": "64Bytes", "UMask": "0x3", @@ -56,20 +55,19 @@ "Unit": "iMC" }, { - "BriefDescription": "write requests to memory controller. Derived from unc_m_cas_count.wr", + "BriefDescription": "All DRAM Write CAS commands issued", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "LLC_MISSES.MEM_WRITE", + "EventName": "UNC_M_CAS_COUNT.WR", "PerPkg": "1", - "ScaleUnit": "64Bytes", "UMask": "0xC", "Unit": "iMC" }, { - "BriefDescription": "write requests to memory controller", + "BriefDescription": "write requests to memory controller. Derived from unc_m_cas_count.wr", "Counter": "0,1,2,3", "EventCode": "0x4", - "EventName": "UNC_M_CAS_COUNT.WR", + "EventName": "LLC_MISSES.MEM_WRITE", "PerPkg": "1", "ScaleUnit": "64Bytes", "UMask": "0xC", diff --git a/tools/perf/pmu-events/arch/x86/skylakex/uncore-other.json b/tools/perf/pmu-events/arch/x86/skylakex/uncore-other.json index f55aeadc630f..0d106fe7aae3 100644 --- a/tools/perf/pmu-events/arch/x86/skylakex/uncore-other.json +++ b/tools/perf/pmu-events/arch/x86/skylakex/uncore-other.json @@ -1089,7 +1089,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x01", - "ScaleUnit": "4Bytes", "UMask": "0x01", "Unit": "IIO" }, @@ -1101,7 +1100,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x02", - "ScaleUnit": "4Bytes", "UMask": "0x01", "Unit": "IIO" }, @@ -1113,7 +1111,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x04", - "ScaleUnit": "4Bytes", "UMask": "0x01", "Unit": "IIO" }, @@ -1125,7 +1122,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x08", - "ScaleUnit": "4Bytes", "UMask": "0x01", "Unit": "IIO" }, @@ -1196,7 +1192,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x01", - "ScaleUnit": "4Bytes", "UMask": "0x04", "Unit": "IIO" }, @@ -1208,7 +1203,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x02", - "ScaleUnit": "4Bytes", "UMask": "0x04", "Unit": "IIO" }, @@ -1220,7 +1214,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x04", - "ScaleUnit": "4Bytes", "UMask": "0x04", "Unit": "IIO" }, @@ -1232,7 +1225,6 @@ "FCMask": "0x07", "PerPkg": "1", "PortMask": "0x08", - "ScaleUnit": "4Bytes", "UMask": "0x04", "Unit": "IIO" }, @@ -1974,20 +1966,19 @@ "Unit": "UPI LL" }, { - "BriefDescription": "UPI interconnect send bandwidth for payload. Derived from unc_upi_txl_flits.all_data", + "BriefDescription": "Valid data FLITs transmitted via any slot", "Counter": "0,1,2,3", "EventCode": "0x2", - "EventName": "UPI_DATA_BANDWIDTH_TX", + "EventName": "UNC_UPI_TxL_FLITS.ALL_DATA", "PerPkg": "1", - "ScaleUnit": "7.11E-06Bytes", - "UMask": "0xf", + "UMask": "0x0F", "Unit": "UPI LL" }, { - "BriefDescription": "UPI interconnect send bandwidth for payload", + "BriefDescription": "UPI interconnect send bandwidth for payload. Derived from unc_upi_txl_flits.all_data", "Counter": "0,1,2,3", "EventCode": "0x2", - "EventName": "UNC_UPI_TxL_FLITS.ALL_DATA", + "EventName": "UPI_DATA_BANDWIDTH_TX", "PerPkg": "1", "ScaleUnit": "7.11E-06Bytes", "UMask": "0xf", diff --git a/tools/perf/pmu-events/arch/x86/tigerlake/tgl-metrics.json b/tools/perf/pmu-events/arch/x86/tigerlake/tgl-metrics.json index 03c97bd74ad9..79b8b101b68f 100644 --- a/tools/perf/pmu-events/arch/x86/tigerlake/tgl-metrics.json +++ b/tools/perf/pmu-events/arch/x86/tigerlake/tgl-metrics.json @@ -1,26 +1,716 @@ [ { + "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend", + "MetricExpr": "topdown\\-fe\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) - INT_MISC.UOP_DROPPING / SLOTS", + "MetricGroup": "PGO;TopdownL1;tma_L1_group", + "MetricName": "tma_frontend_bound", + "PublicDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound. Sample with: FRONTEND_RETIRED.LATENCY_GE_4_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues", + "MetricExpr": "(5 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING) / SLOTS", + "MetricGroup": "Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_latency", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period. Sample with: FRONTEND_RETIRED.LATENCY_GE_16_PS;FRONTEND_RETIRED.LATENCY_GE_8_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses", + "MetricExpr": "ICACHE_16B.IFDATA_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;IcMiss;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_icache_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses. Sample with: FRONTEND_RETIRED.L2_MISS_PS;FRONTEND_RETIRED.L1I_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses", + "MetricExpr": "ICACHE_64B.IFTAG_STALL / CLKS", + "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_itlb_misses", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses. Sample with: FRONTEND_RETIRED.STLB_MISS_PS;FRONTEND_RETIRED.ITLB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers", + "MetricExpr": "INT_MISC.CLEAR_RESTEER_CYCLES / CLKS + tma_unknown_branches", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_branch_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_mispredicts_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears", + "MetricExpr": "(1 - (BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT))) * INT_MISC.CLEAR_RESTEER_CYCLES / CLKS", + "MetricGroup": "BadSpec;MachineClears;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_clears_resteers", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears. Sample with: INT_MISC.CLEAR_RESTEER_CYCLES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears", + "MetricExpr": "10 * BACLEARS.ANY / CLKS", + "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group", + "MetricName": "tma_unknown_branches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (First fetch or hitting BPU capacity limit). Sample with: BACLEARS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines", + "MetricExpr": "DSB2MITE_SWITCHES.PENALTY_CYCLES / CLKS", + "MetricGroup": "DSBmiss;FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_dsb_switches", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty. Sample with: FRONTEND_RETIRED.DSB_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs)", + "MetricExpr": "ILD_STALL.LCP / CLKS", + "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_lcp", + "PublicDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS)", + "MetricExpr": "3 * IDQ.MS_SWITCHES / CLKS", + "MetricGroup": "FetchLat;MicroSeq;TopdownL3;tma_fetch_latency_group", + "MetricName": "tma_ms_switches", + "PublicDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals. Sample with: IDQ.MS_SWITCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues", + "MetricExpr": "max(0, tma_frontend_bound - tma_fetch_latency)", + "MetricGroup": "FetchBW;Frontend;TopdownL2;tma_L2_group;tma_frontend_bound_group", + "MetricName": "tma_fetch_bandwidth", + "PublicDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend. Sample with: FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_1_PS;FRONTEND_RETIRED.LATENCY_GE_2_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline)", + "MetricExpr": "(IDQ.MITE_CYCLES_ANY - IDQ.MITE_CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "DSBmiss;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_mite", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck. Sample with: FRONTEND_RETIRED.ANY_DSB_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where decoder-0 was the only active decoder", + "MetricExpr": "(cpu@INST_DECODED.DECODERS\\,cmask\\=1@ - cpu@INST_DECODED.DECODERS\\,cmask\\=2@) / CORE_CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_decoder0_alone", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where (only) 4 uops were delivered by the MITE pipeline", + "MetricExpr": "(cpu@IDQ.MITE_UOPS\\,cmask\\=4@ - cpu@IDQ.MITE_UOPS\\,cmask\\=5@) / CLKS", + "MetricGroup": "DSBmiss;FetchBW;TopdownL4;tma_mite_group", + "MetricName": "tma_mite_4wide", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline", + "MetricExpr": "(IDQ.DSB_CYCLES_ANY - IDQ.DSB_CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "DSB;FetchBW;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_dsb", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to LSD (Loop Stream Detector) unit", + "MetricExpr": "(LSD.CYCLES_ACTIVE - LSD.CYCLES_OK) / CORE_CLKS / 2", + "MetricGroup": "FetchBW;LSD;TopdownL3;tma_fetch_bandwidth_group", + "MetricName": "tma_lsd", + "PublicDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to LSD (Loop Stream Detector) unit. LSD typically does well sustaining Uop supply. However; in some rare cases; optimal uop-delivery could not be reached for small loops whose size (in terms of number of uops) does not suit well the LSD structure.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations", + "MetricExpr": "max(1 - (tma_frontend_bound + tma_backend_bound + tma_retiring), 0)", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_bad_speculation", + "PublicDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction", + "MetricExpr": "(BR_MISP_RETIRED.ALL_BRANCHES / (BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT)) * tma_bad_speculation", + "MetricGroup": "BadSpec;BrMispredicts;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_branch_mispredicts", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path. Sample with: BR_MISP_RETIRED.ALL_BRANCHES", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears", + "MetricExpr": "max(0, tma_bad_speculation - tma_branch_mispredicts)", + "MetricGroup": "BadSpec;MachineClears;TopdownL2;tma_L2_group;tma_bad_speculation_group", + "MetricName": "tma_machine_clears", + "PublicDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes. Sample with: MACHINE_CLEARS.COUNT", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend", + "MetricExpr": "topdown\\-be\\-bound / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + (5 * cpu@INT_MISC.RECOVERY_CYCLES\\,cmask\\=1\\,edge@) / SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_backend_bound", + "PublicDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound. Sample with: TOPDOWN.BACKEND_BOUND_SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck", + "MetricExpr": "((CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES)) * tma_backend_bound", + "MetricGroup": "Backend;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_memory_bound", + "PublicDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache", + "MetricExpr": "max((CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS) / CLKS, 0)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l1_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache. Sample with: MEM_LOAD_RETIRED.L1_HIT_PS;MEM_LOAD_RETIRED.FB_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses", + "MetricExpr": "min(7 * cpu@DTLB_LOAD_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_LOAD_MISSES.WALK_ACTIVE, max(CYCLE_ACTIVITY.CYCLES_MEM_ANY - CYCLE_ACTIVITY.CYCLES_L1D_MISS, 0)) / CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_dtlb_load", + "PublicDescription": "This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss. Sample with: MEM_INST_RETIRED.STLB_MISS_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_load - tma_load_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk", + "MetricExpr": "DTLB_LOAD_MISSES.WALK_ACTIVE / CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_load_group", + "MetricName": "tma_load_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores", + "MetricExpr": "13 * LD_BLOCKS.STORE_FORWARD / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_store_fwd_blk", + "PublicDescription": "This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations", + "MetricExpr": "(16 * max(0, MEM_INST_RETIRED.LOCK_LOADS - L2_RQSTS.ALL_RFO) + (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES) * (10 * L2_RQSTS.RFO_HIT + min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO))) / CLKS", + "MetricGroup": "Offcore;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_lock_latency", + "PublicDescription": "This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them. Sample with: MEM_INST_RETIRED.LOCK_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary", + "MetricExpr": "Load_Miss_Real_Latency * LD_BLOCKS.NO_SR / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_split_loads", + "PublicDescription": "This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. Sample with: MEM_INST_RETIRED.SPLIT_LOADS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset", + "MetricExpr": "LD_BLOCKS_PARTIAL.ADDRESS_ALIAS / CLKS", + "MetricGroup": "TopdownL4;tma_l1_bound_group", + "MetricName": "tma_4k_aliasing", + "PublicDescription": "This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed", + "MetricExpr": "L1D_PEND_MISS.FB_FULL / CLKS", + "MetricGroup": "MemoryBW;TopdownL4;tma_l1_bound_group", + "MetricName": "tma_fb_full", + "PublicDescription": "This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads", + "MetricExpr": "((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) / ((MEM_LOAD_RETIRED.L2_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS))) + L1D_PEND_MISS.FB_FULL_PERIODS)) * ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS)", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l2_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L2_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L2_MISS - CYCLE_ACTIVITY.STALLS_L3_MISS) / CLKS", + "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_l3_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses", + "MetricExpr": "((49 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD * (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM / (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM + OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD))) + (48 * Average_Frequency) * MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_contested_accesses", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD;MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses", + "MetricExpr": "(48 * Average_Frequency) * (MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD + MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD * (1 - (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM / (OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM + OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD)))) * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "Offcore;Snoop;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_data_sharing", + "PublicDescription": "This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance. Sample with: MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited)", + "MetricExpr": "(17.5 * Average_Frequency) * MEM_LOAD_RETIRED.L3_HIT * (1 + (MEM_LOAD_RETIRED.FB_HIT / MEM_LOAD_RETIRED.L1_MISS) / 2) / CLKS", + "MetricGroup": "MemoryLat;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_l3_hit_latency", + "PublicDescription": "This metric represents fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings. Sample with: MEM_LOAD_RETIRED.L3_HIT_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors)", + "MetricExpr": "L1D_PEND_MISS.L2_STALL / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_l3_bound_group", + "MetricName": "tma_sq_full", + "PublicDescription": "This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors). The Super Queue is used for requests to access the L2 cache or to go out to the Uncore.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads", + "MetricExpr": "(CYCLE_ACTIVITY.STALLS_L3_MISS / CLKS + ((CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS) / CLKS) - tma_l2_bound)", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_dram_bound", + "PublicDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance. Sample with: MEM_LOAD_RETIRED.L3_MISS_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, cpu@OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD\\,cmask\\=4@) / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_bandwidth", + "PublicDescription": "This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory (DRAM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM)", + "MetricExpr": "min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD) / CLKS - tma_mem_bandwidth", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_dram_bound_group", + "MetricName": "tma_mem_latency", + "PublicDescription": "This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory (DRAM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write", + "MetricExpr": "EXE_ACTIVITY.BOUND_ON_STORES / CLKS", + "MetricGroup": "MemoryBound;TmaL3mem;TopdownL3;tma_memory_bound_group", + "MetricName": "tma_store_bound", + "PublicDescription": "This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck. Sample with: MEM_INST_RETIRED.ALL_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses", + "MetricExpr": "((L2_RQSTS.RFO_HIT * 10 * (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES))) + (1 - (MEM_INST_RETIRED.LOCK_LOADS / MEM_INST_RETIRED.ALL_STORES)) * min(CPU_CLK_UNHALTED.THREAD, OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO)) / CLKS", + "MetricGroup": "MemoryLat;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_store_latency", + "PublicDescription": "This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full)", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing", + "MetricExpr": "(54 * Average_Frequency) * OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM / CLKS", + "MetricGroup": "DataSharing;Offcore;Snoop;TopdownL4;tma_store_bound_group", + "MetricName": "tma_false_sharing", + "PublicDescription": "This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. Sample with: OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents rate of split store accesses", + "MetricExpr": "MEM_INST_RETIRED.SPLIT_STORES / CORE_CLKS", + "MetricGroup": "TopdownL4;tma_store_bound_group", + "MetricName": "tma_split_stores", + "PublicDescription": "This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity. Sample with: MEM_INST_RETIRED.SPLIT_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores", + "MetricExpr": "9 * OCR.STREAMING_WR.ANY_RESPONSE / CLKS", + "MetricGroup": "MemoryBW;Offcore;TopdownL4;tma_store_bound_group", + "MetricName": "tma_streaming_stores", + "PublicDescription": "This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should Streaming stores be a bottleneck. Sample with: OCR.STREAMING_WR.ANY_RESPONSE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses", + "MetricExpr": "(7 * cpu@DTLB_STORE_MISSES.STLB_HIT\\,cmask\\=1@ + DTLB_STORE_MISSES.WALK_ACTIVE) / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL4;tma_store_bound_group", + "MetricName": "tma_dtlb_store", + "PublicDescription": "This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently-used data. Sample with: MEM_INST_RETIRED.STLB_MISS_STORES_PS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second-level TLB (STLB)", + "MetricExpr": "tma_dtlb_store - tma_store_stlb_miss", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_hit", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk", + "MetricExpr": "DTLB_STORE_MISSES.WALK_ACTIVE / CORE_CLKS", + "MetricGroup": "MemoryTLB;TopdownL5;tma_dtlb_store_group", + "MetricName": "tma_store_stlb_miss", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck", + "MetricExpr": "max(0, tma_backend_bound - tma_memory_bound)", + "MetricGroup": "Backend;Compute;TopdownL2;tma_L2_group;tma_backend_bound_group", + "MetricName": "tma_core_bound", + "PublicDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active", + "MetricExpr": "ARITH.DIVIDER_ACTIVE / CLKS", + "MetricGroup": "TopdownL3;tma_core_bound_group", + "MetricName": "tma_divider", + "PublicDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication. Sample with: ARITH.DIVIDER_ACTIVE", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related)", + "MetricExpr": "(cpu@EXE_ACTIVITY.3_PORTS_UTIL\\,umask\\=0x80@ + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL)) / CLKS if (ARITH.DIVIDER_ACTIVE < (CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY)) else (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) / CLKS", + "MetricGroup": "PortsUtil;TopdownL3;tma_core_bound_group", + "MetricName": "tma_ports_utilization", + "PublicDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "cpu@EXE_ACTIVITY.3_PORTS_UTIL\\,umask\\=0x80@ / CLKS + tma_serializing_operation * (CYCLE_ACTIVITY.STALLS_TOTAL - CYCLE_ACTIVITY.STALLS_MEM_ANY) / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_0", + "PublicDescription": "This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations", + "MetricExpr": "RESOURCE_STALLS.SCOREBOARD / CLKS", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_serializing_operation", + "PublicDescription": "This metric represents fraction of cycles the CPU issue-pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out-of-order execution which may limit performance. Sample with: RESOURCE_STALLS.SCOREBOARD", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions", + "MetricExpr": "140 * MISC_RETIRED.PAUSE_INST / CLKS", + "MetricGroup": "TopdownL6;tma_serializing_operation_group", + "MetricName": "tma_slow_pause", + "PublicDescription": "This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions. Sample with: MISC_RETIRED.PAUSE_INST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued", + "MetricExpr": "CLKS * UOPS_ISSUED.VECTOR_WIDTH_MISMATCH / UOPS_ISSUED.ANY", + "MetricGroup": "TopdownL5;tma_ports_utilized_0_group", + "MetricName": "tma_mixing_vectors", + "PublicDescription": "The Mixing_Vectors metric gives the percentage of injected blend uops out of all uops issued. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "EXE_ACTIVITY.1_PORTS_UTIL / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_1", + "PublicDescription": "This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful. Sample with: EXE_ACTIVITY.1_PORTS_UTIL", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "EXE_ACTIVITY.2_PORTS_UTIL / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_2", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop. Sample with: EXE_ACTIVITY.2_PORTS_UTIL", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise)", + "MetricExpr": "UOPS_EXECUTED.CYCLES_GE_3 / CLKS", + "MetricGroup": "PortsUtil;TopdownL4;tma_ports_utilization_group", + "MetricName": "tma_ports_utilized_3m", + "PublicDescription": "This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Sample with: UOPS_EXECUTED.CYCLES_GE_3", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", + "MetricExpr": "(UOPS_DISPATCHED.PORT_0 + UOPS_DISPATCHED.PORT_1 + UOPS_DISPATCHED.PORT_5 + UOPS_DISPATCHED.PORT_6) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_alu_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ([SNB+] ALU; [HSW+] ALU and 2nd branch) Sample with: UOPS_DISPATCHED.PORT_0", + "MetricExpr": "UOPS_DISPATCHED.PORT_0 / CORE_CLKS", + "MetricGroup": "Compute;TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_0", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU) Sample with: UOPS_DISPATCHED.PORT_1", + "MetricExpr": "UOPS_DISPATCHED.PORT_1 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_1", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ([SNB+] Branches and ALU; [HSW+] ALU) Sample with: UOPS_DISPATCHED.PORT_5", + "MetricExpr": "UOPS_DISPATCHED.PORT_5 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_5", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 ([HSW+]Primary Branch and simple ALU) Sample with: UOPS_DISPATCHED.PORT_6", + "MetricExpr": "UOPS_DISPATCHED.PORT_6 / CORE_CLKS", + "MetricGroup": "TopdownL6;tma_alu_op_utilization_group", + "MetricName": "tma_port_6", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations Sample with: UOPS_DISPATCHED.PORT_2_3", + "MetricExpr": "UOPS_DISPATCHED.PORT_2_3 / (2 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_load_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations Sample with: UOPS_DISPATCHED.PORT_7_8", + "MetricExpr": "(UOPS_DISPATCHED.PORT_4_9 + UOPS_DISPATCHED.PORT_7_8) / (4 * CORE_CLKS)", + "MetricGroup": "TopdownL5;tma_ports_utilized_3m_group", + "MetricName": "tma_store_op_utilization", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired", + "MetricExpr": "topdown\\-retiring / (topdown\\-fe\\-bound + topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0*SLOTS", + "MetricGroup": "TopdownL1;tma_L1_group", + "MetricName": "tma_retiring", + "PublicDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. Sample with: UOPS_RETIRED.SLOTS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation)", + "MetricExpr": "max(0, tma_retiring - tma_heavy_operations)", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_light_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Sample with: INST_RETIRED.PREC_DIST", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired)", + "MetricExpr": "tma_x87_use + tma_fp_scalar + tma_fp_vector", + "MetricGroup": "HPC;TopdownL3;tma_light_operations_group", + "MetricName": "tma_fp_arith", + "PublicDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric serves as an approximation of legacy x87 usage", + "MetricExpr": "tma_retiring * UOPS_EXECUTED.X87 / UOPS_EXECUTED.THREAD", + "MetricGroup": "Compute;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_x87_use", + "PublicDescription": "This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired", + "MetricExpr": "(FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_scalar", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL4;tma_fp_arith_group", + "MetricName": "tma_fp_vector", + "PublicDescription": "This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_128b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_256b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors", + "MetricExpr": "(FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / (tma_retiring * SLOTS)", + "MetricGroup": "Compute;Flops;TopdownL5;tma_fp_vector_group", + "MetricName": "tma_fp_vector_512b", + "PublicDescription": "This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors. May overcount due to FMA double counting.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.", + "MetricExpr": "tma_light_operations * MEM_INST_RETIRED.ANY / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_memory_operations", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring branch instructions.", + "MetricExpr": "tma_light_operations * BR_INST_RETIRED.ALL_BRANCHES / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_branch_instructions", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions", + "MetricExpr": "tma_light_operations * INST_RETIRED.NOP / (tma_retiring * SLOTS)", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_nop_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body. Sample with: INST_RETIRED.NOP", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting", + "MetricExpr": "max(0, tma_light_operations - (tma_fp_arith + tma_memory_operations + tma_branch_instructions + tma_nop_instructions))", + "MetricGroup": "Pipeline;TopdownL3;tma_light_operations_group", + "MetricName": "tma_other_light_ops", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences", + "MetricExpr": "tma_microcode_sequencer + tma_retiring * (UOPS_DECODED.DEC0 - cpu@UOPS_DECODED.DEC0\\,cmask\\=1@) / IDQ.MITE_UOPS", + "MetricGroup": "Retire;TopdownL2;tma_L2_group;tma_retiring_group", + "MetricName": "tma_heavy_operations", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops", + "MetricExpr": "tma_heavy_operations - tma_microcode_sequencer", + "MetricGroup": "TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_few_uops_instructions", + "PublicDescription": "This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to ([SNB+] four; [ADL+] five) uops. This highly-correlates with the number of uops in such instructions.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit", + "MetricExpr": "((tma_retiring * SLOTS) / UOPS_ISSUED.ANY) * IDQ.MS_UOPS / SLOTS", + "MetricGroup": "MicroSeq;TopdownL3;tma_heavy_operations_group", + "MetricName": "tma_microcode_sequencer", + "PublicDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided. Sample with: IDQ.MS_UOPS", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists", + "MetricExpr": "100 * ASSISTS.ANY / SLOTS", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_assists", + "PublicDescription": "This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases. Sample with: ASSISTS.ANY", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction", + "MetricExpr": "max(0, tma_microcode_sequencer - tma_assists)", + "MetricGroup": "TopdownL4;tma_microcode_sequencer_group", + "MetricName": "tma_cisc", + "PublicDescription": "This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.", + "ScaleUnit": "100%" + }, + { + "BriefDescription": "Total pipeline cost of Branch Misprediction related bottlenecks", + "MetricExpr": "100 * (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches))", + "MetricGroup": "Bad;BadSpec;BrMispredicts", + "MetricName": "Mispredictions" + }, + { + "BriefDescription": "Total pipeline cost of (external) Memory Bandwidth related bottlenecks", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_mem_bandwidth / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_sq_full / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full))) + (tma_l1_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_fb_full / (tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) ", + "MetricGroup": "Mem;MemoryBW;Offcore", + "MetricName": "Memory_Bandwidth" + }, + { + "BriefDescription": "Total pipeline cost of Memory Latency related bottlenecks (external memory and off-core caches)", + "MetricExpr": "100 * tma_memory_bound * ((tma_dram_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_mem_latency / (tma_mem_bandwidth + tma_mem_latency)) + (tma_l3_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_l3_hit_latency / (tma_contested_accesses + tma_data_sharing + tma_l3_hit_latency + tma_sq_full)) + (tma_l2_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)))", + "MetricGroup": "Mem;MemoryLat;Offcore", + "MetricName": "Memory_Latency" + }, + { + "BriefDescription": "Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)", + "MetricExpr": "100 * tma_memory_bound * ((tma_l1_bound / max(tma_memory_bound, tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_dtlb_load / max(tma_l1_bound, tma_4k_aliasing + tma_dtlb_load + tma_fb_full + tma_lock_latency + tma_split_loads + tma_store_fwd_blk)) + (tma_store_bound / (tma_dram_bound + tma_l1_bound + tma_l2_bound + tma_l3_bound + tma_store_bound)) * (tma_dtlb_store / (tma_dtlb_store + tma_false_sharing + tma_split_stores + tma_store_latency + tma_streaming_stores))) ", + "MetricGroup": "Mem;MemoryTLB;Offcore", + "MetricName": "Memory_Data_TLBs" + }, + { "BriefDescription": "Total pipeline cost of branch related instructions (used for program control-flow including function calls)", - "MetricExpr": "100 * (( BR_INST_RETIRED.COND + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL) ) / TOPDOWN.SLOTS)", + "MetricExpr": "100 * ((BR_INST_RETIRED.COND + 3 * BR_INST_RETIRED.NEAR_CALL + (BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL)) / SLOTS)", "MetricGroup": "Ret", "MetricName": "Branching_Overhead" }, { "BriefDescription": "Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)", - "MetricExpr": "100 * (( 5 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING ) / TOPDOWN.SLOTS) * ( (ICACHE_64B.IFTAG_STALL / CPU_CLK_UNHALTED.THREAD) + (ICACHE_16B.IFDATA_STALL / CPU_CLK_UNHALTED.THREAD) + (10 * BACLEARS.ANY / CPU_CLK_UNHALTED.THREAD) ) / #(( 5 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE - INT_MISC.UOP_DROPPING ) / TOPDOWN.SLOTS)", + "MetricExpr": "100 * tma_fetch_latency * (tma_itlb_misses + tma_icache_misses + tma_unknown_branches) / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)", "MetricGroup": "BigFoot;Fed;Frontend;IcMiss;MemoryTLB", "MetricName": "Big_Code" }, { + "BriefDescription": "Total pipeline cost of instruction fetch bandwidth related bottlenecks", + "MetricExpr": "100 * (tma_frontend_bound - tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) - Big_Code", + "MetricGroup": "Fed;FetchBW;Frontend", + "MetricName": "Instruction_Fetch_BW" + }, + { "BriefDescription": "Instructions Per Cycle (per Logical Processor)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD", + "MetricExpr": "INST_RETIRED.ANY / CLKS", "MetricGroup": "Ret;Summary", "MetricName": "IPC" }, { + "BriefDescription": "Uops Per Instruction", + "MetricExpr": "(tma_retiring * SLOTS) / INST_RETIRED.ANY", + "MetricGroup": "Pipeline;Ret;Retire", + "MetricName": "UPI" + }, + { + "BriefDescription": "Instruction per taken branch", + "MetricExpr": "(tma_retiring * SLOTS) / BR_INST_RETIRED.NEAR_TAKEN", + "MetricGroup": "Branches;Fed;FetchBW", + "MetricName": "UpTB" + }, + { "BriefDescription": "Cycles Per Instruction (per Logical Processor)", - "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)", - "MetricGroup": "Pipeline;Mem", + "MetricExpr": "1 / IPC", + "MetricGroup": "Mem;Pipeline", "MetricName": "CPI" }, { @@ -32,13 +722,13 @@ { "BriefDescription": "Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward)", "MetricExpr": "TOPDOWN.SLOTS", - "MetricGroup": "TmaL1", + "MetricGroup": "tma_L1_group", "MetricName": "SLOTS" }, { "BriefDescription": "Fraction of Physical Core issue-slots utilized by this Logical Processor", - "MetricExpr": "TOPDOWN.SLOTS / ( TOPDOWN.SLOTS / 2 ) if #SMT_on else 1", - "MetricGroup": "SMT;TmaL1", + "MetricExpr": "SLOTS / (TOPDOWN.SLOTS / 2) if #SMT_on else 1", + "MetricGroup": "SMT;tma_L1_group", "MetricName": "Slots_Utilization" }, { @@ -50,30 +740,36 @@ }, { "BriefDescription": "Instructions Per Cycle across hyper-threads (per physical core)", - "MetricExpr": "INST_RETIRED.ANY / CPU_CLK_UNHALTED.DISTRIBUTED", - "MetricGroup": "Ret;SMT;TmaL1", + "MetricExpr": "INST_RETIRED.ANY / CORE_CLKS", + "MetricGroup": "Ret;SMT;tma_L1_group", "MetricName": "CoreIPC" }, { "BriefDescription": "Floating Point Operations Per Cycle", - "MetricExpr": "( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / CPU_CLK_UNHALTED.DISTRIBUTED", - "MetricGroup": "Ret;Flops", + "MetricExpr": "(1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / CORE_CLKS", + "MetricGroup": "Flops;Ret", "MetricName": "FLOPc" }, { "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)", - "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.DISTRIBUTED )", + "MetricExpr": "((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)) / (2 * CORE_CLKS)", "MetricGroup": "Cor;Flops;HPC", "MetricName": "FP_Arith_Utilization", "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)." }, { "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core", - "MetricExpr": "UOPS_EXECUTED.THREAD / (( UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2 ) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", + "MetricExpr": "UOPS_EXECUTED.THREAD / ((UOPS_EXECUTED.CORE_CYCLES_GE_1 / 2) if #SMT_on else UOPS_EXECUTED.CORE_CYCLES_GE_1)", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricName": "ILP" }, { + "BriefDescription": "Probability of Core Bound bottleneck hidden by SMT-profiling artifacts", + "MetricExpr": "(1 - tma_core_bound / tma_ports_utilization if tma_core_bound < tma_ports_utilization else 1) if SMT_2T_Utilization > 0.5 else 0", + "MetricGroup": "Cor;SMT", + "MetricName": "Core_Bound_Likely" + }, + { "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core", "MetricExpr": "CPU_CLK_UNHALTED.DISTRIBUTED", "MetricGroup": "SMT", @@ -117,13 +813,13 @@ }, { "BriefDescription": "Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)", "MetricGroup": "Flops;InsType", "MetricName": "IpFLOP" }, { "BriefDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) )", + "MetricExpr": "INST_RETIRED.ANY / ((FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE))", "MetricGroup": "Flops;InsType", "MetricName": "IpArith", "PublicDescription": "Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). May undercount due to FMA double counting. Approximated prior to BDW." @@ -144,21 +840,21 @@ }, { "BriefDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX128", "PublicDescription": "Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX256", "PublicDescription": "Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." }, { "BriefDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate)", - "MetricExpr": "INST_RETIRED.ANY / ( FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE )", + "MetricExpr": "INST_RETIRED.ANY / (FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE)", "MetricGroup": "Flops;FpVector;InsType", "MetricName": "IpArith_AVX512", "PublicDescription": "Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate). May undercount due to FMA double counting." @@ -170,12 +866,18 @@ "MetricName": "IpSWPF" }, { - "BriefDescription": "Total number of retired Instructions, Sample with: INST_RETIRED.PREC_DIST", + "BriefDescription": "Total number of retired Instructions Sample with: INST_RETIRED.PREC_DIST", "MetricExpr": "INST_RETIRED.ANY", - "MetricGroup": "Summary;TmaL1", + "MetricGroup": "Summary;tma_L1_group", "MetricName": "Instructions" }, { + "BriefDescription": "Average number of Uops retired in cycles where at least one uop has retired.", + "MetricExpr": "(tma_retiring * SLOTS) / cpu@UOPS_RETIRED.SLOTS\\,cmask\\=1@", + "MetricGroup": "Pipeline;Ret", + "MetricName": "Retire" + }, + { "BriefDescription": "", "MetricExpr": "UOPS_EXECUTED.THREAD / cpu@UOPS_EXECUTED.THREAD\\,cmask\\=1@", "MetricGroup": "Cor;Pipeline;PortsUtil;SMT", @@ -206,6 +908,12 @@ "MetricName": "DSB_Switch_Cost" }, { + "BriefDescription": "Total penalty related to DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.", + "MetricExpr": "100 * (tma_fetch_latency * tma_dsb_switches / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches) + tma_fetch_bandwidth * tma_mite / (tma_dsb + tma_lsd + tma_mite))", + "MetricGroup": "DSBmiss;Fed", + "MetricName": "DSB_Misses" + }, + { "BriefDescription": "Number of Instructions per non-speculative DSB miss (lower number means higher occurrence rate)", "MetricExpr": "INST_RETIRED.ANY / FRONTEND_RETIRED.ANY_DSB_MISS", "MetricGroup": "DSBmiss;Fed", @@ -218,6 +926,12 @@ "MetricName": "IpMispredict" }, { + "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)", + "MetricExpr": " (tma_branch_mispredicts + tma_fetch_latency * tma_mispredicts_resteers / (tma_branch_resteers + tma_dsb_switches + tma_icache_misses + tma_itlb_misses + tma_lcp + tma_ms_switches)) * SLOTS / BR_MISP_RETIRED.ALL_BRANCHES", + "MetricGroup": "Bad;BrMispredicts", + "MetricName": "Branch_Misprediction_Cost" + }, + { "BriefDescription": "Fraction of branches that are non-taken conditionals", "MetricExpr": "BR_INST_RETIRED.COND_NTAKEN / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches;CodeGen;PGO", @@ -231,7 +945,7 @@ }, { "BriefDescription": "Fraction of branches that are CALL or RET", - "MetricExpr": "( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES", + "MetricExpr": "(BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN) / BR_INST_RETIRED.ALL_BRANCHES", "MetricGroup": "Bad;Branches", "MetricName": "CallRet" }, @@ -243,80 +957,80 @@ }, { "BriefDescription": "Fraction of branches of other types (not individually covered by other metrics in Info.Branches group)", - "MetricExpr": "1 - ( (BR_INST_RETIRED.COND_NTAKEN / BR_INST_RETIRED.ALL_BRANCHES) + (BR_INST_RETIRED.COND_TAKEN / BR_INST_RETIRED.ALL_BRANCHES) + (( BR_INST_RETIRED.NEAR_CALL + BR_INST_RETIRED.NEAR_RETURN ) / BR_INST_RETIRED.ALL_BRANCHES) + ((BR_INST_RETIRED.NEAR_TAKEN - BR_INST_RETIRED.COND_TAKEN - 2 * BR_INST_RETIRED.NEAR_CALL) / BR_INST_RETIRED.ALL_BRANCHES) )", + "MetricExpr": "1 - (Cond_NT + Cond_TK + CallRet + Jump)", "MetricGroup": "Bad;Branches", "MetricName": "Other_Branches" }, { "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)", - "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT )", + "MetricExpr": "L1D_PEND_MISS.PENDING / (MEM_LOAD_RETIRED.L1_MISS + MEM_LOAD_RETIRED.FB_HIT)", "MetricGroup": "Mem;MemoryBound;MemoryLat", "MetricName": "Load_Miss_Real_Latency" }, { "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)", "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES", - "MetricGroup": "Mem;MemoryBound;MemoryBW", + "MetricGroup": "Mem;MemoryBW;MemoryBound", "MetricName": "MLP" }, { "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L1_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI" }, { "BriefDescription": "L1 cache true misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.ALL_DEMAND_DATA_RD / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L1MPKI_Load" }, { "BriefDescription": "L2 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L2_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;Backend;CacheMisses", + "MetricGroup": "Backend;CacheMisses;Mem", "MetricName": "L2MPKI" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)", "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses;Offcore", + "MetricGroup": "CacheMisses;Mem;Offcore", "MetricName": "L2MPKI_All" }, { "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2MPKI_Load" }, { "BriefDescription": "L2 cache hits per kilo instruction for all request types (including speculative)", - "MetricExpr": "1000 * ( L2_RQSTS.REFERENCES - L2_RQSTS.MISS ) / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricExpr": "1000 * (L2_RQSTS.REFERENCES - L2_RQSTS.MISS) / INST_RETIRED.ANY", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_All" }, { "BriefDescription": "L2 cache hits per kilo instruction for all demand loads (including speculative)", "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L2HPKI_Load" }, { "BriefDescription": "L3 cache true misses per kilo instruction for retired demand loads", "MetricExpr": "1000 * MEM_LOAD_RETIRED.L3_MISS / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "L3MPKI" }, { "BriefDescription": "Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries)", "MetricExpr": "1000 * MEM_LOAD_RETIRED.FB_HIT / INST_RETIRED.ANY", - "MetricGroup": "Mem;CacheMisses", + "MetricGroup": "CacheMisses;Mem", "MetricName": "FB_HPKI" }, { "BriefDescription": "Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses", "MetricConstraint": "NO_NMI_WATCHDOG", - "MetricExpr": "( ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING ) / ( 2 * CPU_CLK_UNHALTED.DISTRIBUTED )", + "MetricExpr": "(ITLB_MISSES.WALK_PENDING + DTLB_LOAD_MISSES.WALK_PENDING + DTLB_STORE_MISSES.WALK_PENDING) / (2 * CORE_CLKS)", "MetricGroup": "Mem;MemoryTLB", "MetricName": "Page_Walks_Utilization" }, @@ -346,25 +1060,25 @@ }, { "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]", - "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)", + "MetricExpr": "L1D_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L1D_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]", - "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)", + "MetricExpr": "L2_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L2_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Fill_BW", "MetricGroup": "Mem;MemoryBW", "MetricName": "L3_Cache_Fill_BW_1T" }, { "BriefDescription": "Average per-thread data access bandwidth to the L3 cache [GB / sec]", - "MetricExpr": "(64 * OFFCORE_REQUESTS.ALL_REQUESTS / 1000000000 / duration_time)", + "MetricExpr": "L3_Cache_Access_BW", "MetricGroup": "Mem;MemoryBW;Offcore", "MetricName": "L3_Cache_Access_BW_1T" }, @@ -376,40 +1090,40 @@ }, { "BriefDescription": "Measured Average Frequency for unhalted processors [GHz]", - "MetricExpr": "(CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC) * msr@tsc@ / 1000000000 / duration_time", - "MetricGroup": "Summary;Power", + "MetricExpr": "Turbo_Utilization * msr@tsc@ / 1000000000 / duration_time", + "MetricGroup": "Power;Summary", "MetricName": "Average_Frequency" }, { "BriefDescription": "Giga Floating Point Operations Per Second", - "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * ( FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE ) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE ) / 1000000000 ) / duration_time", + "MetricExpr": "((1 * (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE) + 8 * (FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE) + 16 * FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE) / 1000000000) / duration_time", "MetricGroup": "Cor;Flops;HPC", "MetricName": "GFLOPs", "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine." }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", - "MetricExpr": "CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC", + "MetricExpr": "CLKS / CPU_CLK_UNHALTED.REF_TSC", "MetricGroup": "Power", "MetricName": "Turbo_Utilization" }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0", - "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / CPU_CLK_UNHALTED.DISTRIBUTED", + "MetricExpr": "CORE_POWER.LVL0_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License0_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. This includes non-AVX codes, SSE, AVX 128-bit, and low-current AVX 256-bit codes." }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1", - "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / CPU_CLK_UNHALTED.DISTRIBUTED", + "MetricExpr": "CORE_POWER.LVL1_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License1_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 1. This includes high current AVX 256-bit instructions as well as low current AVX 512-bit instructions." }, { "BriefDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX)", - "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / CPU_CLK_UNHALTED.DISTRIBUTED", + "MetricExpr": "CORE_POWER.LVL2_TURBO_LICENSE / CORE_CLKS", "MetricGroup": "Power", "MetricName": "Power_License2_Utilization", "PublicDescription": "Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). This includes high current AVX 512-bit instructions." @@ -434,7 +1148,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * ( arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@ ) / 1000000 / duration_time / 1000", + "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1000000 / duration_time / 1000", "MetricGroup": "HPC;Mem;MemoryBW;SoC", "MetricName": "DRAM_BW_Use" }, diff --git a/tools/perf/pmu-events/empty-pmu-events.c b/tools/perf/pmu-events/empty-pmu-events.c index 5ed8c0aa4817..480e8f0d30c8 100644 --- a/tools/perf/pmu-events/empty-pmu-events.c +++ b/tools/perf/pmu-events/empty-pmu-events.c @@ -142,15 +142,15 @@ static const struct pmu_event pme_test_soc_cpu[] = { .metric_name = "DCache_L2_All_Miss", }, { - .metric_expr = "dcache_l2_all_hits + dcache_l2_all_miss", + .metric_expr = "DCache_L2_All_Hits + DCache_L2_All_Miss", .metric_name = "DCache_L2_All", }, { - .metric_expr = "d_ratio(dcache_l2_all_hits, dcache_l2_all)", + .metric_expr = "d_ratio(DCache_L2_All_Hits, DCache_L2_All)", .metric_name = "DCache_L2_Hits", }, { - .metric_expr = "d_ratio(dcache_l2_all_miss, dcache_l2_all)", + .metric_expr = "d_ratio(DCache_L2_All_Miss, DCache_L2_All)", .metric_name = "DCache_L2_Misses", }, { diff --git a/tools/perf/tests/cpumap.c b/tools/perf/tests/cpumap.c index 7ea150cdc137..7c873c6ae3eb 100644 --- a/tools/perf/tests/cpumap.c +++ b/tools/perf/tests/cpumap.c @@ -19,7 +19,6 @@ static int process_event_mask(struct perf_tool *tool __maybe_unused, struct perf_record_cpu_map *map_event = &event->cpu_map; struct perf_record_cpu_map_data *data; struct perf_cpu_map *map; - int i; unsigned int long_size; data = &map_event->data; @@ -32,16 +31,17 @@ static int process_event_mask(struct perf_tool *tool __maybe_unused, TEST_ASSERT_VAL("wrong nr", data->mask32_data.nr == 1); - for (i = 0; i < 20; i++) { + TEST_ASSERT_VAL("wrong cpu", perf_record_cpu_map_data__test_bit(0, data)); + TEST_ASSERT_VAL("wrong cpu", !perf_record_cpu_map_data__test_bit(1, data)); + for (int i = 2; i <= 20; i++) TEST_ASSERT_VAL("wrong cpu", perf_record_cpu_map_data__test_bit(i, data)); - } map = cpu_map__new_data(data); TEST_ASSERT_VAL("wrong nr", perf_cpu_map__nr(map) == 20); - for (i = 0; i < 20; i++) { - TEST_ASSERT_VAL("wrong cpu", perf_cpu_map__cpu(map, i).cpu == i); - } + TEST_ASSERT_VAL("wrong cpu", perf_cpu_map__cpu(map, 0).cpu == 0); + for (int i = 2; i <= 20; i++) + TEST_ASSERT_VAL("wrong cpu", perf_cpu_map__cpu(map, i - 1).cpu == i); perf_cpu_map__put(map); return 0; @@ -73,26 +73,60 @@ static int process_event_cpus(struct perf_tool *tool __maybe_unused, return 0; } +static int process_event_range_cpus(struct perf_tool *tool __maybe_unused, + union perf_event *event, + struct perf_sample *sample __maybe_unused, + struct machine *machine __maybe_unused) +{ + struct perf_record_cpu_map *map_event = &event->cpu_map; + struct perf_record_cpu_map_data *data; + struct perf_cpu_map *map; + + data = &map_event->data; + + TEST_ASSERT_VAL("wrong type", data->type == PERF_CPU_MAP__RANGE_CPUS); + + TEST_ASSERT_VAL("wrong any_cpu", data->range_cpu_data.any_cpu == 0); + TEST_ASSERT_VAL("wrong start_cpu", data->range_cpu_data.start_cpu == 1); + TEST_ASSERT_VAL("wrong end_cpu", data->range_cpu_data.end_cpu == 256); + + map = cpu_map__new_data(data); + TEST_ASSERT_VAL("wrong nr", perf_cpu_map__nr(map) == 256); + TEST_ASSERT_VAL("wrong cpu", perf_cpu_map__cpu(map, 0).cpu == 1); + TEST_ASSERT_VAL("wrong cpu", perf_cpu_map__max(map).cpu == 256); + TEST_ASSERT_VAL("wrong refcnt", refcount_read(&map->refcnt) == 1); + perf_cpu_map__put(map); + return 0; +} + static int test__cpu_map_synthesize(struct test_suite *test __maybe_unused, int subtest __maybe_unused) { struct perf_cpu_map *cpus; - /* This one is better stores in mask. */ - cpus = perf_cpu_map__new("0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19"); + /* This one is better stored in a mask. */ + cpus = perf_cpu_map__new("0,2-20"); TEST_ASSERT_VAL("failed to synthesize map", !perf_event__synthesize_cpu_map(NULL, cpus, process_event_mask, NULL)); perf_cpu_map__put(cpus); - /* This one is better stores in cpu values. */ + /* This one is better stored in cpu values. */ cpus = perf_cpu_map__new("1,256"); TEST_ASSERT_VAL("failed to synthesize map", !perf_event__synthesize_cpu_map(NULL, cpus, process_event_cpus, NULL)); perf_cpu_map__put(cpus); + + /* This one is better stored as a range. */ + cpus = perf_cpu_map__new("1-256"); + + TEST_ASSERT_VAL("failed to synthesize map", + !perf_event__synthesize_cpu_map(NULL, cpus, process_event_range_cpus, NULL)); + + perf_cpu_map__put(cpus); return 0; } diff --git a/tools/perf/tests/event_update.c b/tools/perf/tests/event_update.c index 78db4d704e76..d093a9b878d1 100644 --- a/tools/perf/tests/event_update.c +++ b/tools/perf/tests/event_update.c @@ -21,7 +21,7 @@ static int process_event_unit(struct perf_tool *tool __maybe_unused, TEST_ASSERT_VAL("wrong id", ev->id == 123); TEST_ASSERT_VAL("wrong id", ev->type == PERF_EVENT_UPDATE__UNIT); - TEST_ASSERT_VAL("wrong unit", !strcmp(ev->data, "KRAVA")); + TEST_ASSERT_VAL("wrong unit", !strcmp(ev->unit, "KRAVA")); return 0; } @@ -31,13 +31,10 @@ static int process_event_scale(struct perf_tool *tool __maybe_unused, struct machine *machine __maybe_unused) { struct perf_record_event_update *ev = (struct perf_record_event_update *)event; - struct perf_record_event_update_scale *ev_data; - - ev_data = (struct perf_record_event_update_scale *)ev->data; TEST_ASSERT_VAL("wrong id", ev->id == 123); TEST_ASSERT_VAL("wrong id", ev->type == PERF_EVENT_UPDATE__SCALE); - TEST_ASSERT_VAL("wrong scale", ev_data->scale == 0.123); + TEST_ASSERT_VAL("wrong scale", ev->scale.scale == 0.123); return 0; } @@ -56,7 +53,7 @@ static int process_event_name(struct perf_tool *tool, TEST_ASSERT_VAL("wrong id", ev->id == 123); TEST_ASSERT_VAL("wrong id", ev->type == PERF_EVENT_UPDATE__NAME); - TEST_ASSERT_VAL("wrong name", !strcmp(ev->data, tmp->name)); + TEST_ASSERT_VAL("wrong name", !strcmp(ev->name, tmp->name)); return 0; } @@ -66,12 +63,9 @@ static int process_event_cpus(struct perf_tool *tool __maybe_unused, struct machine *machine __maybe_unused) { struct perf_record_event_update *ev = (struct perf_record_event_update *)event; - struct perf_record_event_update_cpus *ev_data; struct perf_cpu_map *map; - ev_data = (struct perf_record_event_update_cpus *) ev->data; - - map = cpu_map__new_data(&ev_data->cpus); + map = cpu_map__new_data(&ev->cpus.cpus); TEST_ASSERT_VAL("wrong id", ev->id == 123); TEST_ASSERT_VAL("wrong type", ev->type == PERF_EVENT_UPDATE__CPUS); diff --git a/tools/perf/tests/expr.c b/tools/perf/tests/expr.c index 2efe9e3a63b8..6512f5e22045 100644 --- a/tools/perf/tests/expr.c +++ b/tools/perf/tests/expr.c @@ -1,4 +1,5 @@ // SPDX-License-Identifier: GPL-2.0 +#include "util/cputopo.h" #include "util/debug.h" #include "util/expr.h" #include "util/header.h" @@ -94,6 +95,10 @@ static int test__expr(struct test_suite *t __maybe_unused, int subtest __maybe_u ret |= test(ctx, "min(1,2) + 1", 2); ret |= test(ctx, "max(1,2) + 1", 3); ret |= test(ctx, "1+1 if 3*4 else 0", 2); + ret |= test(ctx, "100 if 1 else 200 if 1 else 300", 100); + ret |= test(ctx, "100 if 0 else 200 if 1 else 300", 200); + ret |= test(ctx, "100 if 1 else 200 if 0 else 300", 100); + ret |= test(ctx, "100 if 0 else 200 if 0 else 300", 300); ret |= test(ctx, "1.1 + 2.1", 3.2); ret |= test(ctx, ".1 + 2.", 2.1); ret |= test(ctx, "d_ratio(1, 2)", 0.5); @@ -133,7 +138,7 @@ static int test__expr(struct test_suite *t __maybe_unused, int subtest __maybe_u (void **)&val_ptr)); expr__ctx_clear(ctx); - ctx->runtime = 3; + ctx->sctx.runtime = 3; TEST_ASSERT_VAL("find ids", expr__find_ids("EVENT1\\,param\\=?@ + EVENT2\\,param\\=?@", NULL, ctx) == 0); @@ -154,15 +159,33 @@ static int test__expr(struct test_suite *t __maybe_unused, int subtest __maybe_u (void **)&val_ptr)); /* Only EVENT1 or EVENT2 need be measured depending on the value of smt_on. */ - expr__ctx_clear(ctx); - TEST_ASSERT_VAL("find ids", - expr__find_ids("EVENT1 if #smt_on else EVENT2", - NULL, ctx) == 0); - TEST_ASSERT_VAL("find ids", hashmap__size(ctx->ids) == 1); - TEST_ASSERT_VAL("find ids", hashmap__find(ctx->ids, - smt_on() ? "EVENT1" : "EVENT2", - (void **)&val_ptr)); + { + struct cpu_topology *topology = cpu_topology__new(); + bool smton = smt_on(topology); + bool corewide = core_wide(/*system_wide=*/false, + /*user_requested_cpus=*/false, + topology); + + cpu_topology__delete(topology); + expr__ctx_clear(ctx); + TEST_ASSERT_VAL("find ids", + expr__find_ids("EVENT1 if #smt_on else EVENT2", + NULL, ctx) == 0); + TEST_ASSERT_VAL("find ids", hashmap__size(ctx->ids) == 1); + TEST_ASSERT_VAL("find ids", hashmap__find(ctx->ids, + smton ? "EVENT1" : "EVENT2", + (void **)&val_ptr)); + + expr__ctx_clear(ctx); + TEST_ASSERT_VAL("find ids", + expr__find_ids("EVENT1 if #core_wide else EVENT2", + NULL, ctx) == 0); + TEST_ASSERT_VAL("find ids", hashmap__size(ctx->ids) == 1); + TEST_ASSERT_VAL("find ids", hashmap__find(ctx->ids, + corewide ? "EVENT1" : "EVENT2", + (void **)&val_ptr)); + } /* The expression is a constant 1.0 without needing to evaluate EVENT1. */ expr__ctx_clear(ctx); TEST_ASSERT_VAL("find ids", diff --git a/tools/perf/tests/mmap-basic.c b/tools/perf/tests/mmap-basic.c index 9e9a2b67de19..8322fc2295fa 100644 --- a/tools/perf/tests/mmap-basic.c +++ b/tools/perf/tests/mmap-basic.c @@ -1,8 +1,6 @@ // SPDX-License-Identifier: GPL-2.0 #include <errno.h> #include <inttypes.h> -/* For the CLR_() macros */ -#include <pthread.h> #include <stdlib.h> #include <perf/cpumap.h> diff --git a/tools/perf/tests/openat-syscall-all-cpus.c b/tools/perf/tests/openat-syscall-all-cpus.c index 90828ae03ef5..f3275be83a33 100644 --- a/tools/perf/tests/openat-syscall-all-cpus.c +++ b/tools/perf/tests/openat-syscall-all-cpus.c @@ -2,7 +2,7 @@ #include <errno.h> #include <inttypes.h> /* For the CPU_* macros */ -#include <pthread.h> +#include <sched.h> #include <sys/types.h> #include <sys/stat.h> diff --git a/tools/perf/tests/perf-record.c b/tools/perf/tests/perf-record.c index 4952abe716f3..7aa946aa886d 100644 --- a/tools/perf/tests/perf-record.c +++ b/tools/perf/tests/perf-record.c @@ -2,8 +2,6 @@ #include <errno.h> #include <inttypes.h> #include <linux/string.h> -/* For the CLR_() macros */ -#include <pthread.h> #include <sched.h> #include <perf/mmap.h> diff --git a/tools/perf/tests/shell/coresight/Makefile b/tools/perf/tests/shell/coresight/Makefile new file mode 100644 index 000000000000..b070e779703e --- /dev/null +++ b/tools/perf/tests/shell/coresight/Makefile @@ -0,0 +1,29 @@ +# SPDX-License-Identifier: GPL-2.0-only +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 +include ../../../../../tools/scripts/Makefile.include +include ../../../../../tools/scripts/Makefile.arch +include ../../../../../tools/scripts/utilities.mak + +SUBDIRS = \ + asm_pure_loop \ + memcpy_thread \ + thread_loop \ + unroll_loop_thread + +all: $(SUBDIRS) +$(SUBDIRS): + @$(MAKE) -C $@ >/dev/null + +INSTALLDIRS = $(SUBDIRS:%=install-%) + +install-tests: $(INSTALLDIRS) +$(INSTALLDIRS): + @$(MAKE) -C $(@:install-%=%) install-tests >/dev/null + +CLEANDIRS = $(SUBDIRS:%=clean-%) + +clean: $(CLEANDIRS) +$(CLEANDIRS): + $(call QUIET_CLEAN, test-$(@:clean-%=%)) $(Q)$(MAKE) -C $(@:clean-%=%) clean >/dev/null + +.PHONY: all clean $(SUBDIRS) $(CLEANDIRS) $(INSTALLDIRS) diff --git a/tools/perf/tests/shell/coresight/Makefile.miniconfig b/tools/perf/tests/shell/coresight/Makefile.miniconfig new file mode 100644 index 000000000000..5f72a9cb43f3 --- /dev/null +++ b/tools/perf/tests/shell/coresight/Makefile.miniconfig @@ -0,0 +1,14 @@ +# SPDX-License-Identifier: GPL-2.0-only +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 + +ifndef DESTDIR +prefix ?= $(HOME) +endif + +DESTDIR_SQ = $(subst ','\'',$(DESTDIR)) +INSTALL = install +INSTDIR_SUB = tests/shell/coresight + +include ../../../../../scripts/Makefile.include +include ../../../../../scripts/Makefile.arch +include ../../../../../scripts/utilities.mak diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop.sh b/tools/perf/tests/shell/coresight/asm_pure_loop.sh new file mode 100755 index 000000000000..569e9d46162b --- /dev/null +++ b/tools/perf/tests/shell/coresight/asm_pure_loop.sh @@ -0,0 +1,18 @@ +#!/bin/sh -e +# CoreSight / ASM Pure Loop + +# SPDX-License-Identifier: GPL-2.0 +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 + +TEST="asm_pure_loop" +. $(dirname $0)/../lib/coresight.sh +ARGS="" +DATV="out" +DATA="$DATD/perf-$TEST-$DATV.data" + +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS + +perf_dump_aux_verify "$DATA" 10 10 10 + +err=$? +exit $err diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore new file mode 100644 index 000000000000..468673ac32e8 --- /dev/null +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/.gitignore @@ -0,0 +1 @@ +asm_pure_loop diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile new file mode 100644 index 000000000000..206849e92bc9 --- /dev/null +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/Makefile @@ -0,0 +1,34 @@ +# SPDX-License-Identifier: GPL-2.0 +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 + +include ../Makefile.miniconfig + +# Binary to produce +BIN=asm_pure_loop +# Any linking/libraries needed for the binary - empty if none needed +LIB= + +all: $(BIN) + +$(BIN): $(BIN).S +ifdef CORESIGHT +ifeq ($(ARCH),arm64) +# Build line - this is raw asm with no libc to have an always exact binary + $(Q)$(CC) $(BIN).S -nostdlib -static -o $(BIN) $(LIB) +endif +endif + +install-tests: all +ifdef CORESIGHT +ifeq ($(ARCH),arm64) +# Install the test tool in the right place + $(call QUIET_INSTALL, tests) \ + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/$(INSTDIR_SUB)/$(BIN)'; \ + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/$(INSTDIR_SUB)/$(BIN)/$(BIN)' +endif +endif + +clean: + $(Q)$(RM) -f $(BIN) + +.PHONY: all clean install-tests diff --git a/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S new file mode 100644 index 000000000000..75cf084a927d --- /dev/null +++ b/tools/perf/tests/shell/coresight/asm_pure_loop/asm_pure_loop.S @@ -0,0 +1,28 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* Tamas Zsoldos <tamas.zsoldos@arm.com>, 2021 */ + +.globl _start +_start: + mov x0, 0x0000ffff + mov x1, xzr +loop: + nop + nop + cbnz x1, noskip + nop + nop + adrp x2, skip + add x2, x2, :lo12:skip + br x2 + nop + nop +noskip: + nop + nop +skip: + sub x0, x0, 1 + cbnz x0, loop + + mov x0, #0 + mov x8, #93 // __NR_exit syscall + svc #0 diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore new file mode 100644 index 000000000000..f8217e56091e --- /dev/null +++ b/tools/perf/tests/shell/coresight/memcpy_thread/.gitignore @@ -0,0 +1 @@ +memcpy_thread diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/Makefile b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile new file mode 100644 index 000000000000..2db637eb2c26 --- /dev/null +++ b/tools/perf/tests/shell/coresight/memcpy_thread/Makefile @@ -0,0 +1,33 @@ +# SPDX-License-Identifier: GPL-2.0 +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 +include ../Makefile.miniconfig + +# Binary to produce +BIN=memcpy_thread +# Any linking/libraries needed for the binary - empty if none needed +LIB=-pthread + +all: $(BIN) + +$(BIN): $(BIN).c +ifdef CORESIGHT +ifeq ($(ARCH),arm64) +# Build line + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB) +endif +endif + +install-tests: all +ifdef CORESIGHT +ifeq ($(ARCH),arm64) +# Install the test tool in the right place + $(call QUIET_INSTALL, tests) \ + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/$(INSTDIR_SUB)/$(BIN)'; \ + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/$(INSTDIR_SUB)/$(BIN)/$(BIN)' +endif +endif + +clean: + $(Q)$(RM) -f $(BIN) + +.PHONY: all clean install-tests diff --git a/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c new file mode 100644 index 000000000000..a7e169d1bf64 --- /dev/null +++ b/tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c @@ -0,0 +1,79 @@ +// SPDX-License-Identifier: GPL-2.0 +// Carsten Haitzler <carsten.haitzler@arm.com>, 2021 +#include <stdio.h> +#include <stdlib.h> +#include <unistd.h> +#include <string.h> +#include <pthread.h> + +struct args { + unsigned long loops; + unsigned long size; + pthread_t th; + void *ret; +}; + +static void *thrfn(void *arg) +{ + struct args *a = arg; + unsigned long i, len = a->loops; + unsigned char *src, *dst; + + src = malloc(a->size * 1024); + dst = malloc(a->size * 1024); + if ((!src) || (!dst)) { + printf("ERR: Can't allocate memory\n"); + exit(1); + } + for (i = 0; i < len; i++) + memcpy(dst, src, a->size * 1024); +} + +static pthread_t new_thr(void *(*fn) (void *arg), void *arg) +{ + pthread_t t; + pthread_attr_t attr; + + pthread_attr_init(&attr); + pthread_create(&t, &attr, fn, arg); + return t; +} + +int main(int argc, char **argv) +{ + unsigned long i, len, size, thr; + pthread_t threads[256]; + struct args args[256]; + long long v; + + if (argc < 4) { + printf("ERR: %s [copysize Kb] [numthreads] [numloops (hundreds)]\n", argv[0]); + exit(1); + } + + v = atoll(argv[1]); + if ((v < 1) || (v > (1024 * 1024))) { + printf("ERR: max memory 1GB (1048576 KB)\n"); + exit(1); + } + size = v; + thr = atol(argv[2]); + if ((thr < 1) || (thr > 256)) { + printf("ERR: threads 1-256\n"); + exit(1); + } + v = atoll(argv[3]); + if ((v < 1) || (v > 40000000000ll)) { + printf("ERR: loops 1-40000000000 (hundreds)\n"); + exit(1); + } + len = v * 100; + for (i = 0; i < thr; i++) { + args[i].loops = len; + args[i].size = size; + args[i].th = new_thr(thrfn, &(args[i])); + } + for (i = 0; i < thr; i++) + pthread_join(args[i].th, &(args[i].ret)); + return 0; +} diff --git a/tools/perf/tests/shell/coresight/memcpy_thread_16k_10.sh b/tools/perf/tests/shell/coresight/memcpy_thread_16k_10.sh new file mode 100755 index 000000000000..d21ba8545938 --- /dev/null +++ b/tools/perf/tests/shell/coresight/memcpy_thread_16k_10.sh @@ -0,0 +1,18 @@ +#!/bin/sh -e +# CoreSight / Memcpy 16k 10 Threads + +# SPDX-License-Identifier: GPL-2.0 +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 + +TEST="memcpy_thread" +. $(dirname $0)/../lib/coresight.sh +ARGS="16 10 1" +DATV="16k_10" +DATA="$DATD/perf-$TEST-$DATV.data" + +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS + +perf_dump_aux_verify "$DATA" 10 10 10 + +err=$? +exit $err diff --git a/tools/perf/tests/shell/coresight/thread_loop/.gitignore b/tools/perf/tests/shell/coresight/thread_loop/.gitignore new file mode 100644 index 000000000000..6d4c33eaa9e8 --- /dev/null +++ b/tools/perf/tests/shell/coresight/thread_loop/.gitignore @@ -0,0 +1 @@ +thread_loop diff --git a/tools/perf/tests/shell/coresight/thread_loop/Makefile b/tools/perf/tests/shell/coresight/thread_loop/Makefile new file mode 100644 index 000000000000..ea846c038e7a --- /dev/null +++ b/tools/perf/tests/shell/coresight/thread_loop/Makefile @@ -0,0 +1,33 @@ +# SPDX-License-Identifier: GPL-2.0 +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 +include ../Makefile.miniconfig + +# Binary to produce +BIN=thread_loop +# Any linking/libraries needed for the binary - empty if none needed +LIB=-pthread + +all: $(BIN) + +$(BIN): $(BIN).c +ifdef CORESIGHT +ifeq ($(ARCH),arm64) +# Build line + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB) +endif +endif + +install-tests: all +ifdef CORESIGHT +ifeq ($(ARCH),arm64) +# Install the test tool in the right place + $(call QUIET_INSTALL, tests) \ + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/$(INSTDIR_SUB)/$(BIN)'; \ + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/$(INSTDIR_SUB)/$(BIN)/$(BIN)' +endif +endif + +clean: + $(Q)$(RM) -f $(BIN) + +.PHONY: all clean install-tests diff --git a/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c new file mode 100644 index 000000000000..c0158fac7d0b --- /dev/null +++ b/tools/perf/tests/shell/coresight/thread_loop/thread_loop.c @@ -0,0 +1,86 @@ +// SPDX-License-Identifier: GPL-2.0 +// Carsten Haitzler <carsten.haitzler@arm.com>, 2021 + +// define this for gettid() +#define _GNU_SOURCE + +#include <stdio.h> +#include <stdlib.h> +#include <unistd.h> +#include <string.h> +#include <pthread.h> +#include <sys/syscall.h> +#ifndef SYS_gettid +// gettid is 178 on arm64 +# define SYS_gettid 178 +#endif +#define gettid() syscall(SYS_gettid) + +struct args { + unsigned int loops; + pthread_t th; + void *ret; +}; + +static void *thrfn(void *arg) +{ + struct args *a = arg; + int i = 0, len = a->loops; + + if (getenv("SHOW_TID")) { + unsigned long long tid = gettid(); + + printf("%llu\n", tid); + } + asm volatile( + "loop:\n" + "add %[i], %[i], #1\n" + "cmp %[i], %[len]\n" + "blt loop\n" + : /* out */ + : /* in */ [i] "r" (i), [len] "r" (len) + : /* clobber */ + ); + return (void *)(long)i; +} + +static pthread_t new_thr(void *(*fn) (void *arg), void *arg) +{ + pthread_t t; + pthread_attr_t attr; + + pthread_attr_init(&attr); + pthread_create(&t, &attr, fn, arg); + return t; +} + +int main(int argc, char **argv) +{ + unsigned int i, len, thr; + pthread_t threads[256]; + struct args args[256]; + + if (argc < 3) { + printf("ERR: %s [numthreads] [numloops (millions)]\n", argv[0]); + exit(1); + } + + thr = atoi(argv[1]); + if ((thr < 1) || (thr > 256)) { + printf("ERR: threads 1-256\n"); + exit(1); + } + len = atoi(argv[2]); + if ((len < 1) || (len > 4000)) { + printf("ERR: max loops 4000 (millions)\n"); + exit(1); + } + len *= 1000000; + for (i = 0; i < thr; i++) { + args[i].loops = len; + args[i].th = new_thr(thrfn, &(args[i])); + } + for (i = 0; i < thr; i++) + pthread_join(args[i].th, &(args[i].ret)); + return 0; +} diff --git a/tools/perf/tests/shell/coresight/thread_loop_check_tid_10.sh b/tools/perf/tests/shell/coresight/thread_loop_check_tid_10.sh new file mode 100755 index 000000000000..7c13636fc778 --- /dev/null +++ b/tools/perf/tests/shell/coresight/thread_loop_check_tid_10.sh @@ -0,0 +1,19 @@ +#!/bin/sh -e +# CoreSight / Thread Loop 10 Threads - Check TID + +# SPDX-License-Identifier: GPL-2.0 +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 + +TEST="thread_loop" +. $(dirname $0)/../lib/coresight.sh +ARGS="10 1" +DATV="check-tid-10th" +DATA="$DATD/perf-$TEST-$DATV.data" +STDO="$DATD/perf-$TEST-$DATV.stdout" + +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO + +perf_dump_aux_tid_verify "$DATA" "$STDO" + +err=$? +exit $err diff --git a/tools/perf/tests/shell/coresight/thread_loop_check_tid_2.sh b/tools/perf/tests/shell/coresight/thread_loop_check_tid_2.sh new file mode 100755 index 000000000000..a067145af43c --- /dev/null +++ b/tools/perf/tests/shell/coresight/thread_loop_check_tid_2.sh @@ -0,0 +1,19 @@ +#!/bin/sh -e +# CoreSight / Thread Loop 2 Threads - Check TID + +# SPDX-License-Identifier: GPL-2.0 +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 + +TEST="thread_loop" +. $(dirname $0)/../lib/coresight.sh +ARGS="2 20" +DATV="check-tid-2th" +DATA="$DATD/perf-$TEST-$DATV.data" +STDO="$DATD/perf-$TEST-$DATV.stdout" + +SHOW_TID=1 perf record -s $PERFRECOPT -o "$DATA" "$BIN" $ARGS > $STDO + +perf_dump_aux_tid_verify "$DATA" "$STDO" + +err=$? +exit $err diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore new file mode 100644 index 000000000000..2cb4e996dbf3 --- /dev/null +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/.gitignore @@ -0,0 +1 @@ +unroll_loop_thread diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile new file mode 100644 index 000000000000..6264c4e3abd1 --- /dev/null +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/Makefile @@ -0,0 +1,33 @@ +# SPDX-License-Identifier: GPL-2.0 +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 +include ../Makefile.miniconfig + +# Binary to produce +BIN=unroll_loop_thread +# Any linking/libraries needed for the binary - empty if none needed +LIB=-pthread + +all: $(BIN) + +$(BIN): $(BIN).c +ifdef CORESIGHT +ifeq ($(ARCH),arm64) +# Build line + $(Q)$(CC) $(BIN).c -o $(BIN) $(LIB) +endif +endif + +install-tests: all +ifdef CORESIGHT +ifeq ($(ARCH),arm64) +# Install the test tool in the right place + $(call QUIET_INSTALL, tests) \ + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/$(INSTDIR_SUB)/$(BIN)'; \ + $(INSTALL) $(BIN) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/$(INSTDIR_SUB)/$(BIN)/$(BIN)' +endif +endif + +clean: + $(Q)$(RM) -f $(BIN) + +.PHONY: all clean install-tests diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c new file mode 100644 index 000000000000..8f6d384208ed --- /dev/null +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c @@ -0,0 +1,74 @@ +// SPDX-License-Identifier: GPL-2.0 +// Carsten Haitzler <carsten.haitzler@arm.com>, 2021 +#include <stdio.h> +#include <stdlib.h> +#include <unistd.h> +#include <string.h> +#include <pthread.h> + +struct args { + pthread_t th; + unsigned int in; + void *ret; +}; + +static void *thrfn(void *arg) +{ + struct args *a = arg; + unsigned int i, in = a->in; + + for (i = 0; i < 10000; i++) { + asm volatile ( +// force an unroll of thia add instruction so we can test long runs of code +#define SNIP1 "add %[in], %[in], #1\n" +// 10 +#define SNIP2 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 SNIP1 +// 100 +#define SNIP3 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 SNIP2 +// 1000 +#define SNIP4 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 SNIP3 +// 10000 +#define SNIP5 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 SNIP4 +// 100000 + SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 SNIP5 + : /* out */ + : /* in */ [in] "r" (in) + : /* clobber */ + ); + } +} + +static pthread_t new_thr(void *(*fn) (void *arg), void *arg) +{ + pthread_t t; + pthread_attr_t attr; + + pthread_attr_init(&attr); + pthread_create(&t, &attr, fn, arg); + return t; +} + +int main(int argc, char **argv) +{ + unsigned int i, thr; + pthread_t threads[256]; + struct args args[256]; + + if (argc < 2) { + printf("ERR: %s [numthreads]\n", argv[0]); + exit(1); + } + + thr = atoi(argv[1]); + if ((thr > 256) || (thr < 1)) { + printf("ERR: threads 1-256\n"); + exit(1); + } + for (i = 0; i < thr; i++) { + args[i].in = rand(); + args[i].th = new_thr(thrfn, &(args[i])); + } + for (i = 0; i < thr; i++) + pthread_join(args[i].th, &(args[i].ret)); + return 0; +} diff --git a/tools/perf/tests/shell/coresight/unroll_loop_thread_10.sh b/tools/perf/tests/shell/coresight/unroll_loop_thread_10.sh new file mode 100755 index 000000000000..f48c85230b15 --- /dev/null +++ b/tools/perf/tests/shell/coresight/unroll_loop_thread_10.sh @@ -0,0 +1,18 @@ +#!/bin/sh -e +# CoreSight / Unroll Loop Thread 10 + +# SPDX-License-Identifier: GPL-2.0 +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 + +TEST="unroll_loop_thread" +. $(dirname $0)/../lib/coresight.sh +ARGS="10" +DATV="10" +DATA="$DATD/perf-$TEST-$DATV.data" + +perf record $PERFRECOPT -o "$DATA" "$BIN" $ARGS + +perf_dump_aux_verify "$DATA" 10 10 10 + +err=$? +exit $err diff --git a/tools/perf/tests/shell/lib/coresight.sh b/tools/perf/tests/shell/lib/coresight.sh new file mode 100644 index 000000000000..45a1477256b6 --- /dev/null +++ b/tools/perf/tests/shell/lib/coresight.sh @@ -0,0 +1,132 @@ +# SPDX-License-Identifier: GPL-2.0 +# Carsten Haitzler <carsten.haitzler@arm.com>, 2021 + +# This is sourced from a driver script so no need for #!/bin... etc. at the +# top - the assumption below is that it runs as part of sourcing after the +# test sets up some basic env vars to say what it is. + +# This currently works with ETMv4 / ETF not any other packet types at thi +# point. This will need changes if that changes. + +# perf record options for the perf tests to use +PERFRECMEM="-m ,16M" +PERFRECOPT="$PERFRECMEM -e cs_etm//u" + +TOOLS=$(dirname $0) +DIR="$TOOLS/$TEST" +BIN="$DIR/$TEST" +# If the test tool/binary does not exist and is executable then skip the test +if ! test -x "$BIN"; then exit 2; fi +DATD="." +# If the data dir env is set then make the data dir use that instead of ./ +if test -n "$PERF_TEST_CORESIGHT_DATADIR"; then + DATD="$PERF_TEST_CORESIGHT_DATADIR"; +fi +# If the stat dir env is set then make the data dir use that instead of ./ +STATD="." +if test -n "$PERF_TEST_CORESIGHT_STATDIR"; then + STATD="$PERF_TEST_CORESIGHT_STATDIR"; +fi + +# Called if the test fails - error code 1 +err() { + echo "$1" + exit 1 +} + +# Check that some statistics from our perf +check_val_min() { + STATF="$4" + if test "$2" -lt "$3"; then + echo ", FAILED" >> "$STATF" + err "Sanity check number of $1 is too low ($2 < $3)" + fi +} + +perf_dump_aux_verify() { + # Some basic checking that the AUX chunk contains some sensible data + # to see that we are recording something and at least a minimum + # amount of it. We should almost always see Fn packets in just about + # anything but certainly we will see some trace info and async + # packets + DUMP="$DATD/perf-tmp-aux-dump.txt" + perf report --stdio --dump -i "$1" | \ + grep -o -e I_ATOM_F -e I_ASYNC -e I_TRACE_INFO > "$DUMP" + # Simply count how many of these packets we find to see that we are + # producing a reasonable amount of data - exact checks are not sane + # as this is a lossy process where we may lose some blocks and the + # compiler may produce different code depending on the compiler and + # optimization options, so this is rough just to see if we're + # either missing almost all the data or all of it + ATOM_FX_NUM=`grep I_ATOM_F "$DUMP" | wc -l` + ASYNC_NUM=`grep I_ASYNC "$DUMP" | wc -l` + TRACE_INFO_NUM=`grep I_TRACE_INFO "$DUMP" | wc -l` + rm -f "$DUMP" + + # Arguments provide minimums for a pass + CHECK_FX_MIN="$2" + CHECK_ASYNC_MIN="$3" + CHECK_TRACE_INFO_MIN="$4" + + # Write out statistics, so over time you can track results to see if + # there is a pattern - for example we have less "noisy" results that + # produce more consistent amounts of data each run, to see if over + # time any techinques to minimize data loss are having an effect or + # not + STATF="$STATD/stats-$TEST-$DATV.csv" + if ! test -f "$STATF"; then + echo "ATOM Fx Count, Minimum, ASYNC Count, Minimum, TRACE INFO Count, Minimum" > "$STATF" + fi + echo -n "$ATOM_FX_NUM, $CHECK_FX_MIN, $ASYNC_NUM, $CHECK_ASYNC_MIN, $TRACE_INFO_NUM, $CHECK_TRACE_INFO_MIN" >> "$STATF" + + # Actually check to see if we passed or failed. + check_val_min "ATOM_FX" "$ATOM_FX_NUM" "$CHECK_FX_MIN" "$STATF" + check_val_min "ASYNC" "$ASYNC_NUM" "$CHECK_ASYNC_MIN" "$STATF" + check_val_min "TRACE_INFO" "$TRACE_INFO_NUM" "$CHECK_TRACE_INFO_MIN" "$STATF" + echo ", Ok" >> "$STATF" +} + +perf_dump_aux_tid_verify() { + # Specifically crafted test will produce a list of Tread ID's to + # stdout that need to be checked to see that they have had trace + # info collected in AUX blocks in the perf data. This will go + # through all the TID's that are listed as CID=0xabcdef and see + # that all the Thread IDs the test tool reports are in the perf + # data AUX chunks + + # The TID test tools will print a TID per stdout line that are being + # tested + TIDS=`cat "$2"` + # Scan the perf report to find the TIDs that are actually CID in hex + # and build a list of the ones found + FOUND_TIDS=`perf report --stdio --dump -i "$1" | \ + grep -o "CID=0x[0-9a-z]\+" | sed 's/CID=//g' | \ + uniq | sort | uniq` + # No CID=xxx found - maybe your kernel is reporting these as + # VMID=xxx so look there + if test -z "$FOUND_TIDS"; then + FOUND_TIDS=`perf report --stdio --dump -i "$1" | \ + grep -o "VMID=0x[0-9a-z]\+" | sed 's/VMID=//g' | \ + uniq | sort | uniq` + fi + + # Iterate over the list of TIDs that the test says it has and find + # them in the TIDs found in the perf report + MISSING="" + for TID2 in $TIDS; do + FOUND="" + for TIDHEX in $FOUND_TIDS; do + TID=`printf "%i" $TIDHEX` + if test "$TID" -eq "$TID2"; then + FOUND="y" + break + fi + done + if test -z "$FOUND"; then + MISSING="$MISSING $TID" + fi + done + if test -n "$MISSING"; then + err "Thread IDs $MISSING not found in perf AUX data" + fi +} diff --git a/tools/perf/tests/shell/lib/probe_vfs_getname.sh b/tools/perf/tests/shell/lib/probe_vfs_getname.sh index 5b17d916c555..b616d42bd19d 100644 --- a/tools/perf/tests/shell/lib/probe_vfs_getname.sh +++ b/tools/perf/tests/shell/lib/probe_vfs_getname.sh @@ -19,6 +19,6 @@ add_probe_vfs_getname() { } skip_if_no_debuginfo() { - add_probe_vfs_getname -v 2>&1 | egrep -q "^(Failed to find the path for the kernel|Debuginfo-analysis is not supported)" && return 2 + add_probe_vfs_getname -v 2>&1 | egrep -q "^(Failed to find the path for the kernel|Debuginfo-analysis is not supported)|(file has no debug information)" && return 2 return 1 } diff --git a/tools/perf/tests/shell/lib/waiting.sh b/tools/perf/tests/shell/lib/waiting.sh new file mode 100644 index 000000000000..e7a39134a68e --- /dev/null +++ b/tools/perf/tests/shell/lib/waiting.sh @@ -0,0 +1,77 @@ +# SPDX-License-Identifier: GPL-2.0 + +tenths=date\ +%s%1N + +# Wait for PID $1 to have $2 number of threads started +# Time out after $3 tenths of a second or 5 seconds if $3 is "" +wait_for_threads() +{ + tm_out=$3 ; [ -n "${tm_out}" ] || tm_out=50 + start_time=$($tenths) + while [ -e "/proc/$1/task" ] ; do + th_cnt=$(find "/proc/$1/task" -mindepth 1 -maxdepth 1 -printf x | wc -c) + if [ "${th_cnt}" -ge "$2" ] ; then + return 0 + fi + # Wait at most tm_out tenths of a second + if [ $(($($tenths) - start_time)) -ge $tm_out ] ; then + echo "PID $1 does not have $2 threads" + return 1 + fi + done + return 1 +} + +# Wait for perf record -vvv 2>$2 with PID $1 to start by looking at file $2 +# It depends on capturing perf record debug message "perf record has started" +# Time out after $3 tenths of a second or 5 seconds if $3 is "" +wait_for_perf_to_start() +{ + tm_out=$3 ; [ -n "${tm_out}" ] || tm_out=50 + echo "Waiting for \"perf record has started\" message" + start_time=$($tenths) + while [ -e "/proc/$1" ] ; do + if grep -q "perf record has started" "$2" ; then + echo OK + break + fi + # Wait at most tm_out tenths of a second + if [ $(($($tenths) - start_time)) -ge $tm_out ] ; then + echo "perf recording did not start" + return 1 + fi + done + return 0 +} + +# Wait for process PID %1 to exit +# Time out after $2 tenths of a second or 5 seconds if $2 is "" +wait_for_process_to_exit() +{ + tm_out=$2 ; [ -n "${tm_out}" ] || tm_out=50 + start_time=$($tenths) + while [ -e "/proc/$1" ] ; do + # Wait at most tm_out tenths of a second + if [ $(($($tenths) - start_time)) -ge $tm_out ] ; then + echo "PID $1 did not exit as expected" + return 1 + fi + done + return 0 +} + +# Check if PID $1 is still running after $2 tenths of a second +# or 0.3 seconds if $2 is "" +is_running() +{ + tm_out=$2 ; [ -n "${tm_out}" ] || tm_out=3 + start_time=$($tenths) + while [ -e "/proc/$1" ] ; do + # Check for at least tm_out tenths of a second + if [ $(($($tenths) - start_time)) -gt $tm_out ] ; then + return 0 + fi + done + echo "PID $1 exited prematurely" + return 1 +} diff --git a/tools/perf/tests/shell/lock_contention.sh b/tools/perf/tests/shell/lock_contention.sh new file mode 100755 index 000000000000..04bf604e3c6f --- /dev/null +++ b/tools/perf/tests/shell/lock_contention.sh @@ -0,0 +1,73 @@ +#!/bin/sh +# kernel lock contention analysis test +# SPDX-License-Identifier: GPL-2.0 + +set -e + +err=0 +perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX) +result=$(mktemp /tmp/__perf_test.result.XXXXX) + +cleanup() { + rm -f ${perfdata} + rm -f ${result} + trap - exit term int +} + +trap_cleanup() { + cleanup + exit ${err} +} +trap trap_cleanup exit term int + +check() { + if [ `id -u` != 0 ]; then + echo "[Skip] No root permission" + err=2 + exit + fi + + if ! perf list | grep -q lock:contention_begin; then + echo "[Skip] No lock contention tracepoints" + err=2 + exit + fi +} + +test_record() +{ + echo "Testing perf lock record and perf lock contention" + perf lock record -o ${perfdata} -- perf bench sched messaging > /dev/null 2>&1 + # the output goes to the stderr and we expect only 1 output (-E 1) + perf lock contention -i ${perfdata} -E 1 -q 2> ${result} + if [ $(cat "${result}" | wc -l) != "1" ]; then + echo "[Fail] Recorded result count is not 1:" $(cat "${result}" | wc -l) + err=1 + exit + fi +} + +test_bpf() +{ + echo "Testing perf lock contention --use-bpf" + + if ! perf lock con -b true > /dev/null 2>&1 ; then + echo "[Skip] No BPF support" + exit + fi + + # the perf lock contention output goes to the stderr + perf lock con -a -b -E 1 -q -- perf bench sched messaging > /dev/null 2> ${result} + if [ $(cat "${result}" | wc -l) != "1" ]; then + echo "[Fail] BPF result count is not 1:" $(cat "${result}" | wc -l) + err=1 + exit + fi +} + +check + +test_record +test_bpf + +exit ${err} diff --git a/tools/perf/tests/shell/test_data_symbol.sh b/tools/perf/tests/shell/test_data_symbol.sh new file mode 100755 index 000000000000..cd6eb54d235d --- /dev/null +++ b/tools/perf/tests/shell/test_data_symbol.sh @@ -0,0 +1,93 @@ +#!/bin/bash +# Test data symbol + +# SPDX-License-Identifier: GPL-2.0 +# Leo Yan <leo.yan@linaro.org>, 2022 + +skip_if_no_mem_event() { + perf mem record -e list 2>&1 | egrep -q 'available' && return 0 + return 2 +} + +skip_if_no_mem_event || exit 2 + +# skip if there's no compiler +if ! [ -x "$(command -v cc)" ]; then + echo "skip: no compiler, install gcc" + exit 2 +fi + +TEST_PROGRAM=$(mktemp /tmp/__perf_test.program.XXXXX) +PERF_DATA=$(mktemp /tmp/__perf_test.perf.data.XXXXX) + +check_result() { + # The memory report format is as below: + # 99.92% ... [.] buf1+0x38 + result=$(perf mem report -i ${PERF_DATA} -s symbol_daddr -q 2>&1 | + awk '/buf1/ { print $4 }') + + # Testing is failed if has no any sample for "buf1" + [ -z "$result" ] && return 1 + + while IFS= read -r line; do + # The "data1" and "data2" fields in structure "buf1" have + # offset "0x0" and "0x38", returns failure if detect any + # other offset value. + if [ "$line" != "buf1+0x0" ] && [ "$line" != "buf1+0x38" ]; then + return 1 + fi + done <<< "$result" + + return 0 +} + +cleanup_files() +{ + echo "Cleaning up files..." + rm -f ${PERF_DATA} + rm -f ${TEST_PROGRAM} +} + +trap cleanup_files exit term int + +# compile test program +echo "Compiling test program..." +cat << EOF | cc -o ${TEST_PROGRAM} -x c - +typedef struct _buf { + char data1; + char reserved[55]; + char data2; +} buf __attribute__((aligned(64))); + +static buf buf1; + +int main(void) { + for (;;) { + buf1.data1++; + buf1.data2 += buf1.data1; + } + return 0; +} +EOF + +echo "Recording workload..." + +# perf mem/c2c internally uses IBS PMU on AMD CPU which doesn't support +# user/kernel filtering and per-process monitoring, spin program on +# specific CPU and test in per-CPU mode. +is_amd=$(egrep -c 'vendor_id.*AuthenticAMD' /proc/cpuinfo) +if (($is_amd >= 1)); then + perf mem record -o ${PERF_DATA} -C 0 -- taskset -c 0 $TEST_PROGRAM & +else + perf mem record --all-user -o ${PERF_DATA} -- $TEST_PROGRAM & +fi + +PERFPID=$! + +sleep 1 + +kill $PERFPID +wait $PERFPID + +check_result +exit $? diff --git a/tools/perf/tests/shell/test_intel_pt.sh b/tools/perf/tests/shell/test_intel_pt.sh index a3298643884d..efaad9566c34 100755 --- a/tools/perf/tests/shell/test_intel_pt.sh +++ b/tools/perf/tests/shell/test_intel_pt.sh @@ -7,32 +7,99 @@ set -e # Skip if no Intel PT perf list | grep -q 'intel_pt//' || exit 2 +shelldir=$(dirname "$0") +. "${shelldir}"/lib/waiting.sh + skip_cnt=0 ok_cnt=0 err_cnt=0 -tmpfile=`mktemp` -perfdatafile=`mktemp` +temp_dir=$(mktemp -d /tmp/perf-test-intel-pt-sh.XXXXXXXXXX) + +tmpfile="${temp_dir}/tmp-perf.data" +perfdatafile="${temp_dir}/test-perf.data" +outfile="${temp_dir}/test-out.txt" +errfile="${temp_dir}/test-err.txt" +workload="${temp_dir}/workload" +awkscript="${temp_dir}/awkscript" + +cleanup() +{ + trap - EXIT TERM INT + sane=$(echo "${temp_dir}" | cut -b 1-26) + if [ "${sane}" = "/tmp/perf-test-intel-pt-sh" ] ; then + echo "--- Cleaning up ---" + rm -f "${temp_dir}/"* + rmdir "${temp_dir}" + fi +} + +trap_cleanup() +{ + cleanup + exit 1 +} + +trap trap_cleanup EXIT TERM INT + +have_workload=false +cat << _end_of_file_ | /usr/bin/cc -o "${workload}" -xc - -pthread && have_workload=true +#include <time.h> +#include <pthread.h> + +void work(void) { + struct timespec tm = { + .tv_nsec = 1000000, + }; + int i; + + /* Run for about 30 seconds */ + for (i = 0; i < 30000; i++) + nanosleep(&tm, NULL); +} + +void *threadfunc(void *arg) { + work(); + return NULL; +} + +int main(void) { + pthread_t th; + + pthread_create(&th, NULL, threadfunc, NULL); + work(); + pthread_join(th, NULL); + return 0; +} +_end_of_file_ can_cpu_wide() { - perf record -o ${tmpfile} -B -N --no-bpf-event -e dummy:u -C $1 true 2>&1 >/dev/null || return 2 + echo "Checking for CPU-wide recording on CPU $1" + if ! perf record -o "${tmpfile}" -B -N --no-bpf-event -e dummy:u -C "$1" true >/dev/null 2>&1 ; then + echo "No so skipping" + return 2 + fi + echo OK return 0 } test_system_wide_side_band() { + echo "--- Test system-wide sideband ---" + # Need CPU 0 and CPU 1 can_cpu_wide 0 || return $? can_cpu_wide 1 || return $? # Record on CPU 0 a task running on CPU 1 - perf record -B -N --no-bpf-event -o ${perfdatafile} -e intel_pt//u -C 0 -- taskset --cpu-list 1 uname + perf record -B -N --no-bpf-event -o "${perfdatafile}" -e intel_pt//u -C 0 -- taskset --cpu-list 1 uname # Should get MMAP events from CPU 1 because they can be needed to decode - mmap_cnt=`perf script -i ${perfdatafile} --no-itrace --show-mmap-events -C 1 2>/dev/null | grep MMAP | wc -l` + mmap_cnt=$(perf script -i "${perfdatafile}" --no-itrace --show-mmap-events -C 1 2>/dev/null | grep -c MMAP) - if [ ${mmap_cnt} -gt 0 ] ; then + if [ "${mmap_cnt}" -gt 0 ] ; then + echo OK return 0 fi @@ -40,25 +107,175 @@ test_system_wide_side_band() return 1 } +can_kernel() +{ + perf record -o "${tmpfile}" -B -N --no-bpf-event -e dummy:k true >/dev/null 2>&1 || return 2 + return 0 +} + +test_per_thread() +{ + k="$1" + desc="$2" + + echo "--- Test per-thread ${desc}recording ---" + + if ! $have_workload ; then + echo "No workload, so skipping" + return 2 + fi + + if [ "${k}" = "k" ] ; then + can_kernel || return 2 + fi + + cat <<- "_end_of_file_" > "${awkscript}" + BEGIN { + s = "[ ]*" + u = s"[0-9]+"s + d = s"[0-9-]+"s + x = s"[0-9a-fA-FxX]+"s + mmapping = "idx"u": mmapping fd"u + set_output = "idx"u": set output fd"u"->"u + perf_event_open = "sys_perf_event_open: pid"d"cpu"d"group_fd"d"flags"x"="u + } + + /perf record opening and mmapping events/ { + if (!done) + active = 1 + } + + /perf record done opening and mmapping events/ { + active = 0 + done = 1 + } + + $0 ~ perf_event_open && active { + match($0, perf_event_open) + $0 = substr($0, RSTART, RLENGTH) + pid = $3 + cpu = $5 + fd = $11 + print "pid " pid " cpu " cpu " fd " fd " : " $0 + fd_array[fd] = fd + pid_array[fd] = pid + cpu_array[fd] = cpu + } + + $0 ~ mmapping && active { + match($0, mmapping) + $0 = substr($0, RSTART, RLENGTH) + fd = $5 + print "fd " fd " : " $0 + if (fd in fd_array) { + mmap_array[fd] = 1 + } else { + print "Unknown fd " fd + exit 1 + } + } + + $0 ~ set_output && active { + match($0, set_output) + $0 = substr($0, RSTART, RLENGTH) + fd = $6 + fd_to = $8 + print "fd " fd " fd_to " fd_to " : " $0 + if (fd in fd_array) { + if (fd_to in fd_array) { + set_output_array[fd] = fd_to + } else { + print "Unknown fd " fd_to + exit 1 + } + } else { + print "Unknown fd " fd + exit 1 + } + } + + END { + print "Checking " length(fd_array) " fds" + for (fd in fd_array) { + if (fd in mmap_array) { + pid = pid_array[fd] + if (pid != -1) { + if (pid in pids) { + print "More than 1 mmap for PID " pid + exit 1 + } + pids[pid] = 1 + } + cpu = cpu_array[fd] + if (cpu != -1) { + if (cpu in cpus) { + print "More than 1 mmap for CPU " cpu + exit 1 + } + cpus[cpu] = 1 + } + } else if (!(fd in set_output_array)) { + print "No mmap for fd " fd + exit 1 + } + } + n = length(pids) + if (n != thread_cnt) { + print "Expected " thread_cnt " per-thread mmaps - found " n + exit 1 + } + } + _end_of_file_ + + $workload & + w1=$! + $workload & + w2=$! + echo "Workload PIDs are $w1 and $w2" + wait_for_threads ${w1} 2 + wait_for_threads ${w2} 2 + + perf record -B -N --no-bpf-event -o "${perfdatafile}" -e intel_pt//u"${k}" -vvv --per-thread -p "${w1},${w2}" 2>"${errfile}" >"${outfile}" & + ppid=$! + echo "perf PID is $ppid" + wait_for_perf_to_start ${ppid} "${errfile}" || return 1 + + kill ${w1} + wait_for_process_to_exit ${w1} || return 1 + is_running ${ppid} || return 1 + + kill ${w2} + wait_for_process_to_exit ${w2} || return 1 + wait_for_process_to_exit ${ppid} || return 1 + + awk -v thread_cnt=4 -f "${awkscript}" "${errfile}" || return 1 + + echo OK + return 0 +} + count_result() { - if [ $1 -eq 2 ] ; then - skip_cnt=`expr ${skip_cnt} \+ 1` + if [ "$1" -eq 2 ] ; then + skip_cnt=$((skip_cnt + 1)) return fi - if [ $1 -eq 0 ] ; then - ok_cnt=`expr ${ok_cnt} \+ 1` + if [ "$1" -eq 0 ] ; then + ok_cnt=$((ok_cnt + 1)) return fi - err_cnt=`expr ${err_cnt} \+ 1` + err_cnt=$((err_cnt + 1)) + ret=0 } -test_system_wide_side_band +ret=0 +test_system_wide_side_band || ret=$? ; count_result $ret +test_per_thread "" "" || ret=$? ; count_result $ret +test_per_thread "k" "(incl. kernel) " || ret=$? ; count_result $ret -count_result $? +cleanup -rm -f ${tmpfile} -rm -f ${perfdatafile} +echo "--- Done ---" if [ ${err_cnt} -gt 0 ] ; then exit 1 diff --git a/tools/perf/tests/shell/test_java_symbol.sh b/tools/perf/tests/shell/test_java_symbol.sh new file mode 100755 index 000000000000..f221225808a3 --- /dev/null +++ b/tools/perf/tests/shell/test_java_symbol.sh @@ -0,0 +1,75 @@ +#!/bin/bash +# Test java symbol + +# SPDX-License-Identifier: GPL-2.0 +# Leo Yan <leo.yan@linaro.org>, 2022 + +# skip if there's no jshell +if ! [ -x "$(command -v jshell)" ]; then + echo "skip: no jshell, install JDK" + exit 2 +fi + +PERF_DATA=$(mktemp /tmp/__perf_test.perf.data.XXXXX) +PERF_INJ_DATA=$(mktemp /tmp/__perf_test.perf.data.inj.XXXXX) + +cleanup_files() +{ + echo "Cleaning up files..." + rm -f ${PERF_DATA} + rm -f ${PERF_INJ_DATA} +} + +trap cleanup_files exit term int + +if [ -e "$PWD/tools/perf/libperf-jvmti.so" ]; then + LIBJVMTI=$PWD/tools/perf/libperf-jvmti.so +elif [ -e "$PWD/libperf-jvmti.so" ]; then + LIBJVMTI=$PWD/libperf-jvmti.so +elif [ -e "$PREFIX/lib64/libperf-jvmti.so" ]; then + LIBJVMTI=$PREFIX/lib64/libperf-jvmti.so +elif [ -e "$PREFIX/lib/libperf-jvmti.so" ]; then + LIBJVMTI=$PREFIX/lib/libperf-jvmti.so +elif [ -e "/usr/lib/linux-tools-$(uname -a | awk '{ print $3 }' | sed -r 's/-generic//')/libperf-jvmti.so" ]; then + LIBJVMTI=/usr/lib/linux-tools-$(uname -a | awk '{ print $3 }' | sed -r 's/-generic//')/libperf-jvmti.so +else + echo "Fail to find libperf-jvmti.so" + # JVMTI is a build option, skip the test if fail to find lib + exit 2 +fi + +cat <<EOF | perf record -k 1 -o $PERF_DATA jshell -s -J-agentpath:$LIBJVMTI +int fib(int x) { + return x > 1 ? fib(x - 2) + fib(x - 1) : 1; +} + +int q = 0; + +for (int i = 0; i < 10; i++) + q += fib(i); + +System.out.println(q); +EOF + +if [ $? -ne 0 ]; then + echo "Fail to record for java program" + exit 1 +fi + +if ! perf inject -i $PERF_DATA -o $PERF_INJ_DATA -j; then + echo "Fail to inject samples" + exit 1 +fi + +# Below is an example of the instruction samples reporting: +# 8.18% jshell jitted-50116-29.so [.] Interpreter +# 0.75% Thread-1 jitted-83602-1670.so [.] jdk.internal.jimage.BasicImageReader.getString(int) +perf report --stdio -i ${PERF_INJ_DATA} 2>&1 | \ + egrep " +[0-9]+\.[0-9]+% .* (Interpreter|jdk\.internal).*" > /dev/null 2>&1 + +if [ $? -ne 0 ]; then + echo "Fail to find java symbols" + exit 1 +fi + +exit 0 diff --git a/tools/perf/tests/sigtrap.c b/tools/perf/tests/sigtrap.c index e32ece90e164..1de7478ec189 100644 --- a/tools/perf/tests/sigtrap.c +++ b/tools/perf/tests/sigtrap.c @@ -54,6 +54,63 @@ static struct perf_event_attr make_event_attr(void) return attr; } +#ifdef HAVE_BPF_SKEL +#include <bpf/btf.h> + +static bool attr_has_sigtrap(void) +{ + bool ret = false; + struct btf *btf; + const struct btf_type *t; + const struct btf_member *m; + const char *name; + int i, id; + + btf = btf__load_vmlinux_btf(); + if (btf == NULL) { + /* should be an old kernel */ + return false; + } + + id = btf__find_by_name_kind(btf, "perf_event_attr", BTF_KIND_STRUCT); + if (id < 0) + goto out; + + t = btf__type_by_id(btf, id); + for (i = 0, m = btf_members(t); i < btf_vlen(t); i++, m++) { + name = btf__name_by_offset(btf, m->name_off); + if (!strcmp(name, "sigtrap")) { + ret = true; + break; + } + } +out: + btf__free(btf); + return ret; +} +#else /* !HAVE_BPF_SKEL */ +static bool attr_has_sigtrap(void) +{ + struct perf_event_attr attr = { + .type = PERF_TYPE_SOFTWARE, + .config = PERF_COUNT_SW_DUMMY, + .size = sizeof(attr), + .remove_on_exec = 1, /* Required by sigtrap. */ + .sigtrap = 1, /* Request synchronous SIGTRAP on event. */ + }; + int fd; + bool ret = false; + + fd = sys_perf_event_open(&attr, 0, -1, -1, perf_event_open_cloexec_flag()); + if (fd >= 0) { + ret = true; + close(fd); + } + + return ret; +} +#endif /* HAVE_BPF_SKEL */ + static void sigtrap_handler(int signum __maybe_unused, siginfo_t *info, void *ucontext __maybe_unused) { @@ -139,7 +196,13 @@ static int test__sigtrap(struct test_suite *test __maybe_unused, int subtest __m fd = sys_perf_event_open(&attr, 0, -1, -1, perf_event_open_cloexec_flag()); if (fd < 0) { - pr_debug("FAILED sys_perf_event_open(): %s\n", str_error_r(errno, sbuf, sizeof(sbuf))); + if (attr_has_sigtrap()) { + pr_debug("FAILED sys_perf_event_open(): %s\n", + str_error_r(errno, sbuf, sizeof(sbuf))); + } else { + pr_debug("perf_event_attr doesn't have sigtrap\n"); + ret = TEST_SKIP; + } goto out_restore_sigaction; } diff --git a/tools/perf/tests/switch-tracking.c b/tools/perf/tests/switch-tracking.c index 2d46af9ef935..87f565c7f650 100644 --- a/tools/perf/tests/switch-tracking.c +++ b/tools/perf/tests/switch-tracking.c @@ -6,6 +6,7 @@ #include <time.h> #include <stdlib.h> #include <linux/zalloc.h> +#include <linux/err.h> #include <perf/cpumap.h> #include <perf/evlist.h> #include <perf/mmap.h> @@ -398,19 +399,13 @@ static int test__switch_tracking(struct test_suite *test __maybe_unused, int sub goto out; } - err = parse_event(evlist, sched_switch); - if (err) { - pr_debug("Failed to parse event %s\n", sched_switch); + switch_evsel = evlist__add_sched_switch(evlist, true); + if (IS_ERR(switch_evsel)) { + err = PTR_ERR(switch_evsel); + pr_debug("Failed to create event %s\n", sched_switch); goto out_err; } - switch_evsel = evlist__last(evlist); - - evsel__set_sample_bit(switch_evsel, CPU); - evsel__set_sample_bit(switch_evsel, TIME); - - switch_evsel->core.system_wide = true; - switch_evsel->no_aux_samples = true; switch_evsel->immediate = true; /* Test moving an event to the front */ diff --git a/tools/perf/tests/topology.c b/tools/perf/tests/topology.c index 0b4f61b6cc6b..c4630cfc80ea 100644 --- a/tools/perf/tests/topology.c +++ b/tools/perf/tests/topology.c @@ -147,7 +147,7 @@ static int check_cpu_topology(char *path, struct perf_cpu_map *map) TEST_ASSERT_VAL("Cpu map - Die ID doesn't match", session->header.env.cpu[perf_cpu_map__cpu(map, i).cpu].die_id == id.die); TEST_ASSERT_VAL("Cpu map - Node ID is set", id.node == -1); - TEST_ASSERT_VAL("Cpu map - Thread is set", id.thread == -1); + TEST_ASSERT_VAL("Cpu map - Thread IDX is set", id.thread_idx == -1); } // Test that core ID contains socket, die and core @@ -163,7 +163,7 @@ static int check_cpu_topology(char *path, struct perf_cpu_map *map) TEST_ASSERT_VAL("Core map - Die ID doesn't match", session->header.env.cpu[perf_cpu_map__cpu(map, i).cpu].die_id == id.die); TEST_ASSERT_VAL("Core map - Node ID is set", id.node == -1); - TEST_ASSERT_VAL("Core map - Thread is set", id.thread == -1); + TEST_ASSERT_VAL("Core map - Thread IDX is set", id.thread_idx == -1); } // Test that die ID contains socket and die @@ -179,7 +179,7 @@ static int check_cpu_topology(char *path, struct perf_cpu_map *map) TEST_ASSERT_VAL("Die map - Node ID is set", id.node == -1); TEST_ASSERT_VAL("Die map - Core is set", id.core == -1); TEST_ASSERT_VAL("Die map - CPU is set", id.cpu.cpu == -1); - TEST_ASSERT_VAL("Die map - Thread is set", id.thread == -1); + TEST_ASSERT_VAL("Die map - Thread IDX is set", id.thread_idx == -1); } // Test that socket ID contains only socket @@ -193,7 +193,7 @@ static int check_cpu_topology(char *path, struct perf_cpu_map *map) TEST_ASSERT_VAL("Socket map - Die ID is set", id.die == -1); TEST_ASSERT_VAL("Socket map - Core is set", id.core == -1); TEST_ASSERT_VAL("Socket map - CPU is set", id.cpu.cpu == -1); - TEST_ASSERT_VAL("Socket map - Thread is set", id.thread == -1); + TEST_ASSERT_VAL("Socket map - Thread IDX is set", id.thread_idx == -1); } // Test that node ID contains only node @@ -205,7 +205,7 @@ static int check_cpu_topology(char *path, struct perf_cpu_map *map) TEST_ASSERT_VAL("Node map - Die ID is set", id.die == -1); TEST_ASSERT_VAL("Node map - Core is set", id.core == -1); TEST_ASSERT_VAL("Node map - CPU is set", id.cpu.cpu == -1); - TEST_ASSERT_VAL("Node map - Thread is set", id.thread == -1); + TEST_ASSERT_VAL("Node map - Thread IDX is set", id.thread_idx == -1); } perf_session__delete(session); diff --git a/tools/perf/ui/browser.c b/tools/perf/ui/browser.c index fa5bd5c20e96..78fb01d6ad63 100644 --- a/tools/perf/ui/browser.c +++ b/tools/perf/ui/browser.c @@ -268,9 +268,9 @@ void __ui_browser__show_title(struct ui_browser *browser, const char *title) void ui_browser__show_title(struct ui_browser *browser, const char *title) { - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); __ui_browser__show_title(browser, title); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); } int ui_browser__show(struct ui_browser *browser, const char *title, @@ -284,7 +284,7 @@ int ui_browser__show(struct ui_browser *browser, const char *title, browser->refresh_dimensions(browser); - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); __ui_browser__show_title(browser, title); browser->title = title; @@ -295,16 +295,16 @@ int ui_browser__show(struct ui_browser *browser, const char *title, va_end(ap); if (err > 0) ui_helpline__push(browser->helpline); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); return err ? 0 : -1; } void ui_browser__hide(struct ui_browser *browser) { - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); ui_helpline__pop(); zfree(&browser->helpline); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); } static void ui_browser__scrollbar_set(struct ui_browser *browser) @@ -352,9 +352,9 @@ static int __ui_browser__refresh(struct ui_browser *browser) int ui_browser__refresh(struct ui_browser *browser) { - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); __ui_browser__refresh(browser); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); return 0; } @@ -390,10 +390,10 @@ int ui_browser__run(struct ui_browser *browser, int delay_secs) while (1) { off_t offset; - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); err = __ui_browser__refresh(browser); SLsmg_refresh(); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); if (err < 0) break; diff --git a/tools/perf/ui/browsers/annotate.c b/tools/perf/ui/browsers/annotate.c index 44ba900828f6..c03fa76c02ff 100644 --- a/tools/perf/ui/browsers/annotate.c +++ b/tools/perf/ui/browsers/annotate.c @@ -8,22 +8,17 @@ #include "../../util/hist.h" #include "../../util/sort.h" #include "../../util/map.h" +#include "../../util/mutex.h" #include "../../util/symbol.h" #include "../../util/evsel.h" #include "../../util/evlist.h" #include <inttypes.h> -#include <pthread.h> #include <linux/kernel.h> #include <linux/string.h> #include <linux/zalloc.h> #include <sys/ttydefaults.h> #include <asm/bug.h> -struct disasm_line_samples { - double percent; - struct sym_hist_entry he; -}; - struct arch; struct annotate_browser { @@ -319,7 +314,7 @@ static void annotate_browser__calc_percent(struct annotate_browser *browser, browser->entries = RB_ROOT; - pthread_mutex_lock(¬es->lock); + mutex_lock(¬es->lock); symbol__calc_percent(sym, evsel); @@ -348,7 +343,7 @@ static void annotate_browser__calc_percent(struct annotate_browser *browser, } disasm_rb_tree__insert(browser, &pos->al); } - pthread_mutex_unlock(¬es->lock); + mutex_unlock(¬es->lock); browser->curr_hot = rb_last(&browser->entries); } @@ -474,10 +469,10 @@ static bool annotate_browser__callq(struct annotate_browser *browser, } notes = symbol__annotation(dl->ops.target.sym); - pthread_mutex_lock(¬es->lock); + mutex_lock(¬es->lock); if (!symbol__hists(dl->ops.target.sym, evsel->evlist->core.nr_entries)) { - pthread_mutex_unlock(¬es->lock); + mutex_unlock(¬es->lock); ui__warning("Not enough memory for annotating '%s' symbol!\n", dl->ops.target.sym->name); return true; @@ -486,7 +481,7 @@ static bool annotate_browser__callq(struct annotate_browser *browser, target_ms.maps = ms->maps; target_ms.map = ms->map; target_ms.sym = dl->ops.target.sym; - pthread_mutex_unlock(¬es->lock); + mutex_unlock(¬es->lock); symbol__tui_annotate(&target_ms, evsel, hbt, browser->opts); sym_title(ms->sym, ms->map, title, sizeof(title), browser->opts->percent_type); ui_browser__show_title(&browser->b, title); @@ -805,7 +800,8 @@ static int annotate_browser__run(struct annotate_browser *browser, "r Run available scripts\n" "p Toggle percent type [local/global]\n" "b Toggle percent base [period/hits]\n" - "? Search string backwards\n"); + "? Search string backwards\n" + "f Toggle showing offsets to full address\n"); continue; case 'r': script_browse(NULL, NULL); @@ -912,6 +908,9 @@ show_sup_ins: hists__scnprintf_title(hists, title, sizeof(title)); annotate_browser__show(&browser->b, title, help); continue; + case 'f': + annotation__toggle_full_addr(notes, ms); + continue; case K_LEFT: case K_ESC: case 'q': diff --git a/tools/perf/ui/setup.c b/tools/perf/ui/setup.c index 700335cde618..25ded88801a3 100644 --- a/tools/perf/ui/setup.c +++ b/tools/perf/ui/setup.c @@ -1,5 +1,4 @@ // SPDX-License-Identifier: GPL-2.0 -#include <pthread.h> #include <dlfcn.h> #include <unistd.h> @@ -8,7 +7,7 @@ #include "../util/hist.h" #include "ui.h" -pthread_mutex_t ui__lock = PTHREAD_MUTEX_INITIALIZER; +struct mutex ui__lock; void *perf_gtk_handle; int use_browser = -1; @@ -76,6 +75,7 @@ int stdio__config_color(const struct option *opt __maybe_unused, void setup_browser(bool fallback_to_pager) { + mutex_init(&ui__lock); if (use_browser < 2 && (!isatty(1) || dump_trace)) use_browser = 0; @@ -118,4 +118,5 @@ void exit_browser(bool wait_for_ok) default: break; } + mutex_destroy(&ui__lock); } diff --git a/tools/perf/ui/tui/helpline.c b/tools/perf/ui/tui/helpline.c index 298d6af82fdd..db4952f5990b 100644 --- a/tools/perf/ui/tui/helpline.c +++ b/tools/perf/ui/tui/helpline.c @@ -2,7 +2,6 @@ #include <stdio.h> #include <stdlib.h> #include <string.h> -#include <pthread.h> #include <linux/kernel.h> #include <linux/string.h> @@ -33,7 +32,7 @@ static int tui_helpline__show(const char *format, va_list ap) int ret; static int backlog; - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); ret = vscnprintf(ui_helpline__last_msg + backlog, sizeof(ui_helpline__last_msg) - backlog, format, ap); backlog += ret; @@ -45,7 +44,7 @@ static int tui_helpline__show(const char *format, va_list ap) SLsmg_refresh(); backlog = 0; } - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); return ret; } diff --git a/tools/perf/ui/tui/progress.c b/tools/perf/ui/tui/progress.c index 3d74af5a7ece..71b6c8d9474f 100644 --- a/tools/perf/ui/tui/progress.c +++ b/tools/perf/ui/tui/progress.c @@ -45,7 +45,7 @@ static void tui_progress__update(struct ui_progress *p) } ui__refresh_dimensions(false); - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); y = SLtt_Screen_Rows / 2 - 2; SLsmg_set_color(0); SLsmg_draw_box(y, 0, 3, SLtt_Screen_Cols); @@ -56,7 +56,7 @@ static void tui_progress__update(struct ui_progress *p) bar = ((SLtt_Screen_Cols - 2) * p->curr) / p->total; SLsmg_fill_region(y, 1, 1, bar, ' '); SLsmg_refresh(); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); } static void tui_progress__finish(void) @@ -67,12 +67,12 @@ static void tui_progress__finish(void) return; ui__refresh_dimensions(false); - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); y = SLtt_Screen_Rows / 2 - 2; SLsmg_set_color(0); SLsmg_fill_region(y, 0, 3, SLtt_Screen_Cols, ' '); SLsmg_refresh(); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); } static struct ui_progress_ops tui_progress__ops = { diff --git a/tools/perf/ui/tui/setup.c b/tools/perf/ui/tui/setup.c index b1be59b4e2a4..a3b8c397c24d 100644 --- a/tools/perf/ui/tui/setup.c +++ b/tools/perf/ui/tui/setup.c @@ -29,10 +29,10 @@ void ui__refresh_dimensions(bool force) { if (force || ui__need_resize) { ui__need_resize = 0; - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); SLtt_get_screen_size(); SLsmg_reinit_smg(); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); } } @@ -170,10 +170,10 @@ void ui__exit(bool wait_for_ok) "Press any key...", 0); SLtt_set_cursor_visibility(1); - if (!pthread_mutex_trylock(&ui__lock)) { + if (mutex_trylock(&ui__lock)) { SLsmg_refresh(); SLsmg_reset_smg(); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); } SLang_reset_tty(); perf_error__unregister(&perf_tui_eops); diff --git a/tools/perf/ui/tui/util.c b/tools/perf/ui/tui/util.c index 0f562e2cb1e8..3c5174854ac8 100644 --- a/tools/perf/ui/tui/util.c +++ b/tools/perf/ui/tui/util.c @@ -95,7 +95,7 @@ int ui_browser__input_window(const char *title, const char *text, char *input, t = sep + 1; } - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); max_len += 2; nr_lines += 8; @@ -125,17 +125,17 @@ int ui_browser__input_window(const char *title, const char *text, char *input, SLsmg_write_nstring((char *)exit_msg, max_len); SLsmg_refresh(); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); x += 2; len = 0; key = ui__getch(delay_secs); while (key != K_TIMER && key != K_ENTER && key != K_ESC) { - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); if (key == K_BKSPC) { if (len == 0) { - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); goto next_key; } SLsmg_gotorc(y, x + --len); @@ -147,7 +147,7 @@ int ui_browser__input_window(const char *title, const char *text, char *input, } SLsmg_refresh(); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); /* XXX more graceful overflow handling needed */ if (len == sizeof(buf) - 1) { @@ -215,19 +215,19 @@ void __ui__info_window(const char *title, const char *text, const char *exit_msg void ui__info_window(const char *title, const char *text) { - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); __ui__info_window(title, text, NULL); SLsmg_refresh(); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); } int ui__question_window(const char *title, const char *text, const char *exit_msg, int delay_secs) { - pthread_mutex_lock(&ui__lock); + mutex_lock(&ui__lock); __ui__info_window(title, text, exit_msg); SLsmg_refresh(); - pthread_mutex_unlock(&ui__lock); + mutex_unlock(&ui__lock); return ui__getch(delay_secs); } diff --git a/tools/perf/ui/ui.h b/tools/perf/ui/ui.h index 9b6fdf06e1d2..99f8d2fe9bc5 100644 --- a/tools/perf/ui/ui.h +++ b/tools/perf/ui/ui.h @@ -2,11 +2,11 @@ #ifndef _PERF_UI_H_ #define _PERF_UI_H_ 1 -#include <pthread.h> +#include "../util/mutex.h" #include <stdbool.h> #include <linux/compiler.h> -extern pthread_mutex_t ui__lock; +extern struct mutex ui__lock; extern void *perf_gtk_handle; extern int use_browser; diff --git a/tools/perf/util/Build b/tools/perf/util/Build index 485e1a343165..815d235466d0 100644 --- a/tools/perf/util/Build +++ b/tools/perf/util/Build @@ -143,6 +143,7 @@ perf-y += branch.o perf-y += mem2node.o perf-y += clockid.o perf-y += list_sort.o +perf-y += mutex.o perf-$(CONFIG_LIBBPF) += bpf-loader.o perf-$(CONFIG_LIBBPF) += bpf_map.o diff --git a/tools/perf/util/PERF-VERSION-GEN b/tools/perf/util/PERF-VERSION-GEN index 0ee5af529238..3cc42821d9b3 100755 --- a/tools/perf/util/PERF-VERSION-GEN +++ b/tools/perf/util/PERF-VERSION-GEN @@ -11,7 +11,8 @@ LF=' ' # -# Always try first to get the version from the kernel Makefile +# Use version from kernel Makefile unless not in a git repository and +# PERF-VERSION-FILE exists # CID= TAG= @@ -19,9 +20,14 @@ if test -d ../../.git -o -f ../../.git then TAG=$(MAKEFLAGS= make -sC ../.. kernelversion) CID=$(git log -1 --abbrev=12 --pretty=format:"%h" 2>/dev/null) && CID="-g$CID" -else +elif test -f ../../PERF-VERSION-FILE +then TAG=$(cut -d' ' -f3 ../../PERF-VERSION-FILE | sed -e 's/\"//g') fi +if test -z "$TAG" +then + TAG=$(MAKEFLAGS= make -sC ../.. kernelversion) +fi VN="$TAG$CID" if test -n "$CID" diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index 2c6a485c3de5..db475e44f42f 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -35,7 +35,6 @@ #include "arch/common.h" #include "namespaces.h" #include <regex.h> -#include <pthread.h> #include <linux/bitops.h> #include <linux/kernel.h> #include <linux/string.h> @@ -821,7 +820,7 @@ void symbol__annotate_zero_histograms(struct symbol *sym) { struct annotation *notes = symbol__annotation(sym); - pthread_mutex_lock(¬es->lock); + mutex_lock(¬es->lock); if (notes->src != NULL) { memset(notes->src->histograms, 0, notes->src->nr_histograms * notes->src->sizeof_sym_hist); @@ -829,7 +828,7 @@ void symbol__annotate_zero_histograms(struct symbol *sym) memset(notes->src->cycles_hist, 0, symbol__size(sym) * sizeof(struct cyc_hist)); } - pthread_mutex_unlock(¬es->lock); + mutex_unlock(¬es->lock); } static int __symbol__account_cycles(struct cyc_hist *ch, @@ -1086,7 +1085,7 @@ void annotation__compute_ipc(struct annotation *notes, size_t size) notes->hit_insn = 0; notes->cover_insn = 0; - pthread_mutex_lock(¬es->lock); + mutex_lock(¬es->lock); for (offset = size - 1; offset >= 0; --offset) { struct cyc_hist *ch; @@ -1105,7 +1104,7 @@ void annotation__compute_ipc(struct annotation *notes, size_t size) notes->have_cycles = true; } } - pthread_mutex_unlock(¬es->lock); + mutex_unlock(¬es->lock); } int addr_map_symbol__inc_samples(struct addr_map_symbol *ams, struct perf_sample *sample, @@ -1258,13 +1257,13 @@ int disasm_line__scnprintf(struct disasm_line *dl, char *bf, size_t size, bool r void annotation__init(struct annotation *notes) { - pthread_mutex_init(¬es->lock, NULL); + mutex_init(¬es->lock); } void annotation__exit(struct annotation *notes) { annotated_source__delete(notes->src); - pthread_mutex_destroy(¬es->lock); + mutex_destroy(¬es->lock); } static void annotation_line__add(struct annotation_line *al, struct list_head *head) @@ -1698,6 +1697,7 @@ fallback: */ __symbol__join_symfs(filename, filename_size, dso->long_name); + mutex_lock(&dso->lock); if (access(filename, R_OK) && errno == ENOENT && dso->nsinfo) { char *new_name = filename_with_chroot(dso->nsinfo->pid, filename); @@ -1706,6 +1706,7 @@ fallback: free(new_name); } } + mutex_unlock(&dso->lock); } free(build_id_path); @@ -2238,7 +2239,10 @@ int symbol__annotate(struct map_symbol *ms, struct evsel *evsel, } args.ms = *ms; - notes->start = map__rip_2objdump(ms->map, sym->start); + if (notes->options && notes->options->full_addr) + notes->start = map__objdump_2mem(ms->map, ms->sym->start); + else + notes->start = map__rip_2objdump(ms->map, ms->sym->start); return symbol__disassemble(sym, &args); } @@ -2761,6 +2765,8 @@ void annotation__update_column_widths(struct annotation *notes) { if (notes->options->use_offset) notes->widths.target = notes->widths.min_addr; + else if (notes->options->full_addr) + notes->widths.target = BITS_PER_LONG / 4; else notes->widths.target = notes->widths.max_addr; @@ -2770,6 +2776,18 @@ void annotation__update_column_widths(struct annotation *notes) notes->widths.addr += notes->widths.jumps + 1; } +void annotation__toggle_full_addr(struct annotation *notes, struct map_symbol *ms) +{ + notes->options->full_addr = !notes->options->full_addr; + + if (notes->options->full_addr) + notes->start = map__objdump_2mem(ms->map, ms->sym->start); + else + notes->start = map__rip_2objdump(ms->map, ms->sym->start); + + annotation__update_column_widths(notes); +} + static void annotation__calc_lines(struct annotation *notes, struct map *map, struct rb_root *root, struct annotation_options *opts) diff --git a/tools/perf/util/annotate.h b/tools/perf/util/annotate.h index 986f2bbe4870..8934072c39e6 100644 --- a/tools/perf/util/annotate.h +++ b/tools/perf/util/annotate.h @@ -8,9 +8,9 @@ #include <linux/types.h> #include <linux/list.h> #include <linux/rbtree.h> -#include <pthread.h> #include <asm/bug.h> #include "symbol_conf.h" +#include "mutex.h" #include "spark.h" struct hist_browser_timer; @@ -88,7 +88,8 @@ struct annotation_options { show_nr_jumps, show_minmax_cycle, show_asm_raw, - annotate_src; + annotate_src, + full_addr; u8 offset_level; int min_pcnt; int max_lines; @@ -273,7 +274,7 @@ struct annotated_source { }; struct annotation { - pthread_mutex_t lock; + struct mutex lock; u64 max_coverage; u64 start; u64 hit_cycles; @@ -325,6 +326,7 @@ void annotation__compute_ipc(struct annotation *notes, size_t size); void annotation__mark_jump_targets(struct annotation *notes, struct symbol *sym); void annotation__update_column_widths(struct annotation *notes); void annotation__init_column_widths(struct annotation *notes, struct symbol *sym); +void annotation__toggle_full_addr(struct annotation *notes, struct map_symbol *ms); static inline struct sym_hist *annotated_source__histogram(struct annotated_source *src, int idx) { diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c index 6edab8a16de6..b59c278fe9ed 100644 --- a/tools/perf/util/auxtrace.c +++ b/tools/perf/util/auxtrace.c @@ -26,6 +26,7 @@ #include <linux/list.h> #include <linux/zalloc.h> +#include "config.h" #include "evlist.h" #include "dso.h" #include "map.h" @@ -1434,6 +1435,16 @@ static int get_flags(const char **ptr, unsigned int *plus_flags, unsigned int *m } } +#define ITRACE_DFLT_LOG_ON_ERROR_SZ 16384 + +static unsigned int itrace_log_on_error_size(void) +{ + unsigned int sz = 0; + + perf_config_scan("itrace.debug-log-buffer-size", "%u", &sz); + return sz ?: ITRACE_DFLT_LOG_ON_ERROR_SZ; +} + /* * Please check tools/perf/Documentation/perf-script.txt for information * about the options parsed here, which is introduced after this cset, @@ -1532,6 +1543,8 @@ int itrace_do_parse_synth_opts(struct itrace_synth_opts *synth_opts, if (get_flags(&p, &synth_opts->log_plus_flags, &synth_opts->log_minus_flags)) goto out_err; + if (synth_opts->log_plus_flags & AUXTRACE_LOG_FLG_ON_ERROR) + synth_opts->log_on_error_size = itrace_log_on_error_size(); break; case 'c': synth_opts->branches = true; diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h index 6a4fbfd34c6b..cb8e0a01abb6 100644 --- a/tools/perf/util/auxtrace.h +++ b/tools/perf/util/auxtrace.h @@ -60,6 +60,7 @@ enum itrace_period_type { #define AUXTRACE_ERR_FLG_DATA_LOST (1 << ('l' - 'a')) #define AUXTRACE_LOG_FLG_ALL_PERF_EVTS (1 << ('a' - 'a')) +#define AUXTRACE_LOG_FLG_ON_ERROR (1 << ('e' - 'a')) #define AUXTRACE_LOG_FLG_USE_STDOUT (1 << ('o' - 'a')) /** @@ -110,6 +111,7 @@ enum itrace_period_type { * @log_plus_flags: flags to affect what is logged * @log_minus_flags: flags to affect what is logged * @quick: quicker (less detailed) decoding + * @log_on_error_size: size of log to keep for outputting log only on errors */ struct itrace_synth_opts { bool set; @@ -155,6 +157,7 @@ struct itrace_synth_opts { unsigned int log_plus_flags; unsigned int log_minus_flags; unsigned int quick; + unsigned int log_on_error_size; }; /** diff --git a/tools/perf/util/bpf-event.h b/tools/perf/util/bpf-event.h index 144a8a24cc69..1bcbd4fb6c66 100644 --- a/tools/perf/util/bpf-event.h +++ b/tools/perf/util/bpf-event.h @@ -4,7 +4,6 @@ #include <linux/compiler.h> #include <linux/rbtree.h> -#include <pthread.h> #include <api/fd/array.h> #include <stdio.h> diff --git a/tools/perf/util/bpf-loader.c b/tools/perf/util/bpf-loader.c index e2052f4fed33..d657594894cf 100644 --- a/tools/perf/util/bpf-loader.c +++ b/tools/perf/util/bpf-loader.c @@ -27,7 +27,11 @@ #include "util.h" #include "llvm-utils.h" #include "c++/clang-c.h" -#include "hashmap.h" +#ifdef HAVE_LIBBPF_SUPPORT +#include <bpf/hashmap.h> +#else +#include "util/hashmap.h" +#endif #include "asm/bug.h" #include <internal/xyarray.h> diff --git a/tools/perf/util/bpf_lock_contention.c b/tools/perf/util/bpf_lock_contention.c index c591a66733ef..fc4d613cb979 100644 --- a/tools/perf/util/bpf_lock_contention.c +++ b/tools/perf/util/bpf_lock_contention.c @@ -8,17 +8,13 @@ #include "util/thread_map.h" #include "util/lock-contention.h" #include <linux/zalloc.h> +#include <linux/string.h> #include <bpf/bpf.h> #include "bpf_skel/lock_contention.skel.h" static struct lock_contention_bpf *skel; -/* should be same as bpf_skel/lock_contention.bpf.c */ -struct lock_contention_key { - s32 stack_id; -}; - struct lock_contention_data { u64 total_time; u64 min_time; @@ -40,6 +36,7 @@ int lock_contention_prepare(struct lock_contention *con) return -1; } + bpf_map__set_value_size(skel->maps.stacks, con->max_stack * sizeof(u64)); bpf_map__set_max_entries(skel->maps.stacks, con->map_nr_entries); bpf_map__set_max_entries(skel->maps.lock_stat, con->map_nr_entries); @@ -91,6 +88,8 @@ int lock_contention_prepare(struct lock_contention *con) bpf_map_update_elem(fd, &pid, &val, BPF_ANY); } + skel->bss->stack_skip = con->stack_skip; + lock_contention_bpf__attach(skel); return 0; } @@ -114,7 +113,7 @@ int lock_contention_read(struct lock_contention *con) struct lock_contention_data data; struct lock_stat *st; struct machine *machine = con->machine; - u64 stack_trace[CONTENTION_STACK_DEPTH]; + u64 stack_trace[con->max_stack]; fd = bpf_map__fd(skel->maps.lock_stat); stack = bpf_map__fd(skel->maps.stacks); @@ -125,7 +124,7 @@ int lock_contention_read(struct lock_contention *con) while (!bpf_map_get_next_key(fd, &prev_key, &key)) { struct map *kmap; struct symbol *sym; - int idx; + int idx = 0; bpf_map_lookup_elem(fd, &key, &data); st = zalloc(sizeof(*st)); @@ -144,10 +143,9 @@ int lock_contention_read(struct lock_contention *con) bpf_map_lookup_elem(stack, &key, stack_trace); - /* skip BPF + lock internal functions */ - idx = CONTENTION_STACK_SKIP; + /* skip lock internal functions */ while (is_lock_function(machine, stack_trace[idx]) && - idx < CONTENTION_STACK_DEPTH - 1) + idx < con->max_stack - 1) idx++; st->addr = stack_trace[idx]; @@ -171,6 +169,14 @@ int lock_contention_read(struct lock_contention *con) return -1; } + if (verbose) { + st->callstack = memdup(stack_trace, sizeof(stack_trace)); + if (st->callstack == NULL) { + free(st); + return -1; + } + } + hlist_add_head(&st->hash_entry, con->result); prev_key = key; } diff --git a/tools/perf/util/bpf_skel/lock_contention.bpf.c b/tools/perf/util/bpf_skel/lock_contention.bpf.c index 9e8b94eb6320..1bb8628e7c9f 100644 --- a/tools/perf/util/bpf_skel/lock_contention.bpf.c +++ b/tools/perf/util/bpf_skel/lock_contention.bpf.c @@ -72,9 +72,10 @@ struct { int enabled; int has_cpu; int has_task; +int stack_skip; /* error stat */ -unsigned long lost; +int lost; static inline int can_record(void) { @@ -117,7 +118,7 @@ int contention_begin(u64 *ctx) pelem->timestamp = bpf_ktime_get_ns(); pelem->lock = (__u64)ctx[0]; pelem->flags = (__u32)ctx[1]; - pelem->stack_id = bpf_get_stackid(ctx, &stacks, BPF_F_FAST_STACK_CMP); + pelem->stack_id = bpf_get_stackid(ctx, &stacks, BPF_F_FAST_STACK_CMP | stack_skip); if (pelem->stack_id < 0) lost++; diff --git a/tools/perf/util/branch.c b/tools/perf/util/branch.c index a9a909db8cc7..6d38238481d3 100644 --- a/tools/perf/util/branch.c +++ b/tools/perf/util/branch.c @@ -21,7 +21,10 @@ void branch_type_count(struct branch_type_stat *st, struct branch_flags *flags, if (flags->type == PERF_BR_UNKNOWN || from == 0) return; - st->counts[flags->type]++; + if (flags->type == PERF_BR_EXTEND_ABI) + st->new_counts[flags->new_type]++; + else + st->counts[flags->type]++; if (flags->type == PERF_BR_COND) { if (to > from) @@ -36,6 +39,38 @@ void branch_type_count(struct branch_type_stat *st, struct branch_flags *flags, st->cross_4k++; } +const char *branch_new_type_name(int new_type) +{ + const char *branch_new_names[PERF_BR_NEW_MAX] = { + "FAULT_ALGN", + "FAULT_DATA", + "FAULT_INST", +/* + * TODO: This switch should happen on 'session->header.env.arch' + * instead, because an arm64 platform perf recording could be + * opened for analysis on other platforms as well. + */ +#ifdef __aarch64__ + "ARM64_FIQ", + "ARM64_DEBUG_HALT", + "ARM64_DEBUG_EXIT", + "ARM64_DEBUG_INST", + "ARM64_DEBUG_DATA" +#else + "ARCH_1", + "ARCH_2", + "ARCH_3", + "ARCH_4", + "ARCH_5" +#endif + }; + + if (new_type >= 0 && new_type < PERF_BR_NEW_MAX) + return branch_new_names[new_type]; + + return NULL; +} + const char *branch_type_name(int type) { const char *branch_names[PERF_BR_MAX] = { @@ -51,7 +86,10 @@ const char *branch_type_name(int type) "COND_CALL", "COND_RET", "ERET", - "IRQ" + "IRQ", + "SERROR", + "NO_TX", + "", // Needed for PERF_BR_EXTEND_ABI that ends up triggering some compiler warnings about NULL deref }; if (type >= 0 && type < PERF_BR_MAX) @@ -60,6 +98,17 @@ const char *branch_type_name(int type) return NULL; } +const char *get_branch_type(struct branch_entry *e) +{ + if (e->flags.type == PERF_BR_UNKNOWN) + return ""; + + if (e->flags.type == PERF_BR_EXTEND_ABI) + return branch_new_type_name(e->flags.new_type); + + return branch_type_name(e->flags.type); +} + void branch_type_stat_display(FILE *fp, struct branch_type_stat *st) { u64 total = 0; @@ -106,6 +155,15 @@ void branch_type_stat_display(FILE *fp, struct branch_type_stat *st) 100.0 * (double)st->counts[i] / (double)total); } + + for (i = 0; i < PERF_BR_NEW_MAX; i++) { + if (st->new_counts[i] > 0) + fprintf(fp, "\n%8s: %5.1f%%", + branch_new_type_name(i), + 100.0 * + (double)st->new_counts[i] / (double)total); + } + } static int count_str_scnprintf(int idx, const char *str, char *bf, int size) @@ -121,6 +179,9 @@ int branch_type_str(struct branch_type_stat *st, char *bf, int size) for (i = 0; i < PERF_BR_MAX; i++) total += st->counts[i]; + for (i = 0; i < PERF_BR_NEW_MAX; i++) + total += st->new_counts[i]; + if (total == 0) return 0; @@ -138,6 +199,11 @@ int branch_type_str(struct branch_type_stat *st, char *bf, int size) printed += count_str_scnprintf(j++, branch_type_name(i), bf + printed, size - printed); } + for (i = 0; i < PERF_BR_NEW_MAX; i++) { + if (st->new_counts[i] > 0) + printed += count_str_scnprintf(j++, branch_new_type_name(i), bf + printed, size - printed); + } + if (st->cross_4k > 0) printed += count_str_scnprintf(j++, "CROSS_4K", bf + printed, size - printed); diff --git a/tools/perf/util/branch.h b/tools/perf/util/branch.h index 17b2ccc61094..f838b23db180 100644 --- a/tools/perf/util/branch.h +++ b/tools/perf/util/branch.h @@ -24,7 +24,9 @@ struct branch_flags { u64 abort:1; u64 cycles:16; u64 type:4; - u64 reserved:40; + u64 new_type:4; + u64 priv:3; + u64 reserved:33; }; }; }; @@ -72,6 +74,7 @@ static inline struct branch_entry *perf_sample__branch_entries(struct perf_sampl struct branch_type_stat { bool branch_to; u64 counts[PERF_BR_MAX]; + u64 new_counts[PERF_BR_NEW_MAX]; u64 cond_fwd; u64 cond_bwd; u64 cross_4k; @@ -82,6 +85,8 @@ void branch_type_count(struct branch_type_stat *st, struct branch_flags *flags, u64 from, u64 to); const char *branch_type_name(int type); +const char *branch_new_type_name(int new_type); +const char *get_branch_type(struct branch_entry *e); void branch_type_stat_display(FILE *fp, struct branch_type_stat *st); int branch_type_str(struct branch_type_stat *st, char *bf, int bfsize); diff --git a/tools/perf/util/build-id.c b/tools/perf/util/build-id.c index ec18ed5caf3e..a839b30c981b 100644 --- a/tools/perf/util/build-id.c +++ b/tools/perf/util/build-id.c @@ -898,11 +898,15 @@ static int filename__read_build_id_ns(const char *filename, static bool dso__build_id_mismatch(struct dso *dso, const char *name) { struct build_id bid; + bool ret = false; - if (filename__read_build_id_ns(name, &bid, dso->nsinfo) < 0) - return false; + mutex_lock(&dso->lock); + if (filename__read_build_id_ns(name, &bid, dso->nsinfo) >= 0) + ret = !dso__build_id_equal(dso, &bid); - return !dso__build_id_equal(dso, &bid); + mutex_unlock(&dso->lock); + + return ret; } static int dso__cache_build_id(struct dso *dso, struct machine *machine, @@ -941,8 +945,10 @@ static int dso__cache_build_id(struct dso *dso, struct machine *machine, if (!is_kallsyms && dso__build_id_mismatch(dso, name)) goto out_free; + mutex_lock(&dso->lock); ret = build_id_cache__add_b(&dso->bid, name, dso->nsinfo, is_kallsyms, is_vdso, proper_name, root_dir); + mutex_unlock(&dso->lock); out_free: free(allocated_name); return ret; diff --git a/tools/perf/util/callchain.c b/tools/perf/util/callchain.c index 7e663673f79f..a093a15f048f 100644 --- a/tools/perf/util/callchain.c +++ b/tools/perf/util/callchain.c @@ -1307,24 +1307,16 @@ int callchain_branch_counts(struct callchain_root *root, static int count_pri64_printf(int idx, const char *str, u64 value, char *bf, int bfsize) { - int printed; - - printed = scnprintf(bf, bfsize, "%s%s:%" PRId64 "", (idx) ? " " : " (", str, value); - - return printed; + return scnprintf(bf, bfsize, "%s%s:%" PRId64 "", (idx) ? " " : " (", str, value); } static int count_float_printf(int idx, const char *str, float value, char *bf, int bfsize, float threshold) { - int printed; - if (threshold != 0.0 && value < threshold) return 0; - printed = scnprintf(bf, bfsize, "%s%s:%.1f%%", (idx) ? " " : " (", str, value); - - return printed; + return scnprintf(bf, bfsize, "%s%s:%.1f%%", (idx) ? " " : " (", str, value); } static int branch_to_str(char *bf, int bfsize, diff --git a/tools/perf/util/config.c b/tools/perf/util/config.c index 60ce5908c664..3f2ae19a1dd4 100644 --- a/tools/perf/util/config.c +++ b/tools/perf/util/config.c @@ -908,3 +908,34 @@ void set_buildid_dir(const char *dir) /* for communicating with external commands */ setenv("PERF_BUILDID_DIR", buildid_dir, 1); } + +struct perf_config_scan_data { + const char *name; + const char *fmt; + va_list args; + int ret; +}; + +static int perf_config_scan_cb(const char *var, const char *value, void *data) +{ + struct perf_config_scan_data *d = data; + + if (!strcmp(var, d->name)) + d->ret = vsscanf(value, d->fmt, d->args); + + return 0; +} + +int perf_config_scan(const char *name, const char *fmt, ...) +{ + struct perf_config_scan_data d = { + .name = name, + .fmt = fmt, + }; + + va_start(d.args, fmt); + perf_config(perf_config_scan_cb, &d); + va_end(d.args); + + return d.ret; +} diff --git a/tools/perf/util/config.h b/tools/perf/util/config.h index 2fd77aaff4d2..2e5e808928a5 100644 --- a/tools/perf/util/config.h +++ b/tools/perf/util/config.h @@ -29,6 +29,7 @@ typedef int (*config_fn_t)(const char *, const char *, void *); int perf_default_config(const char *, const char *, void *); int perf_config(config_fn_t fn, void *); +int perf_config_scan(const char *name, const char *fmt, ...) __scanf(2, 3); int perf_config_set(struct perf_config_set *set, config_fn_t fn, void *data); int perf_config_int(int *dest, const char *, const char *); diff --git a/tools/perf/util/cpumap.c b/tools/perf/util/cpumap.c index ae43fb88f444..8486ca3bec75 100644 --- a/tools/perf/util/cpumap.c +++ b/tools/perf/util/cpumap.c @@ -112,12 +112,39 @@ static struct perf_cpu_map *cpu_map__from_mask(const struct perf_record_cpu_map_ } +static struct perf_cpu_map *cpu_map__from_range(const struct perf_record_cpu_map_data *data) +{ + struct perf_cpu_map *map; + unsigned int i = 0; + + map = perf_cpu_map__empty_new(data->range_cpu_data.end_cpu - + data->range_cpu_data.start_cpu + 1 + data->range_cpu_data.any_cpu); + if (!map) + return NULL; + + if (data->range_cpu_data.any_cpu) + map->map[i++].cpu = -1; + + for (int cpu = data->range_cpu_data.start_cpu; cpu <= data->range_cpu_data.end_cpu; + i++, cpu++) + map->map[i].cpu = cpu; + + return map; +} + struct perf_cpu_map *cpu_map__new_data(const struct perf_record_cpu_map_data *data) { - if (data->type == PERF_CPU_MAP__CPUS) + switch (data->type) { + case PERF_CPU_MAP__CPUS: return cpu_map__from_entries(data); - else + case PERF_CPU_MAP__MASK: return cpu_map__from_mask(data); + case PERF_CPU_MAP__RANGE_CPUS: + return cpu_map__from_range(data); + default: + pr_err("cpu_map__new_data unknown type %d\n", data->type); + return NULL; + } } size_t cpu_map__fprintf(struct perf_cpu_map *map, FILE *fp) @@ -202,7 +229,7 @@ static int aggr_cpu_id__cmp(const void *a_pointer, const void *b_pointer) else if (a->core != b->core) return a->core - b->core; else - return a->thread - b->thread; + return a->thread_idx - b->thread_idx; } struct cpu_aggr_map *cpu_aggr_map__new(const struct perf_cpu_map *cpus, @@ -640,7 +667,7 @@ const struct perf_cpu_map *cpu_map__online(void) /* thread unsafe */ bool aggr_cpu_id__equal(const struct aggr_cpu_id *a, const struct aggr_cpu_id *b) { - return a->thread == b->thread && + return a->thread_idx == b->thread_idx && a->node == b->node && a->socket == b->socket && a->die == b->die && @@ -650,7 +677,7 @@ bool aggr_cpu_id__equal(const struct aggr_cpu_id *a, const struct aggr_cpu_id *b bool aggr_cpu_id__is_empty(const struct aggr_cpu_id *a) { - return a->thread == -1 && + return a->thread_idx == -1 && a->node == -1 && a->socket == -1 && a->die == -1 && @@ -661,7 +688,7 @@ bool aggr_cpu_id__is_empty(const struct aggr_cpu_id *a) struct aggr_cpu_id aggr_cpu_id__empty(void) { struct aggr_cpu_id ret = { - .thread = -1, + .thread_idx = -1, .node = -1, .socket = -1, .die = -1, diff --git a/tools/perf/util/cpumap.h b/tools/perf/util/cpumap.h index fa8a5acdcae1..4a6d029576ee 100644 --- a/tools/perf/util/cpumap.h +++ b/tools/perf/util/cpumap.h @@ -10,7 +10,7 @@ /** Identify where counts are aggregated, -1 implies not to aggregate. */ struct aggr_cpu_id { /** A value in the range 0 to number of threads. */ - int thread; + int thread_idx; /** The numa node X as read from /sys/devices/system/node/nodeX. */ int node; /** diff --git a/tools/perf/util/cputopo.c b/tools/perf/util/cputopo.c index d275d843c155..1a3ff6449158 100644 --- a/tools/perf/util/cputopo.c +++ b/tools/perf/util/cputopo.c @@ -157,6 +157,67 @@ void cpu_topology__delete(struct cpu_topology *tp) free(tp); } +bool cpu_topology__smt_on(const struct cpu_topology *topology) +{ + for (u32 i = 0; i < topology->core_cpus_lists; i++) { + const char *cpu_list = topology->core_cpus_list[i]; + + /* + * If there is a need to separate siblings in a core then SMT is + * enabled. + */ + if (strchr(cpu_list, ',') || strchr(cpu_list, '-')) + return true; + } + return false; +} + +bool cpu_topology__core_wide(const struct cpu_topology *topology, + const char *user_requested_cpu_list) +{ + struct perf_cpu_map *user_requested_cpus; + + /* + * If user_requested_cpu_list is empty then all CPUs are recorded and so + * core_wide is true. + */ + if (!user_requested_cpu_list) + return true; + + user_requested_cpus = perf_cpu_map__new(user_requested_cpu_list); + /* Check that every user requested CPU is the complete set of SMT threads on a core. */ + for (u32 i = 0; i < topology->core_cpus_lists; i++) { + const char *core_cpu_list = topology->core_cpus_list[i]; + struct perf_cpu_map *core_cpus = perf_cpu_map__new(core_cpu_list); + struct perf_cpu cpu; + int idx; + bool has_first, first = true; + + perf_cpu_map__for_each_cpu(cpu, idx, core_cpus) { + if (first) { + has_first = perf_cpu_map__has(user_requested_cpus, cpu); + first = false; + } else { + /* + * If the first core CPU is user requested then + * all subsequent CPUs in the core must be user + * requested too. If the first CPU isn't user + * requested then none of the others must be + * too. + */ + if (perf_cpu_map__has(user_requested_cpus, cpu) != has_first) { + perf_cpu_map__put(core_cpus); + perf_cpu_map__put(user_requested_cpus); + return false; + } + } + } + perf_cpu_map__put(core_cpus); + } + perf_cpu_map__put(user_requested_cpus); + return true; +} + static bool has_die_topology(void) { char filename[MAXPATHLEN]; diff --git a/tools/perf/util/cputopo.h b/tools/perf/util/cputopo.h index 854e18f9041e..969e5920a00e 100644 --- a/tools/perf/util/cputopo.h +++ b/tools/perf/util/cputopo.h @@ -58,6 +58,11 @@ struct hybrid_topology { struct cpu_topology *cpu_topology__new(void); void cpu_topology__delete(struct cpu_topology *tp); +/* Determine from the core list whether SMT was enabled. */ +bool cpu_topology__smt_on(const struct cpu_topology *topology); +/* Are the sets of SMT siblings all enabled or all disabled in user_requested_cpus. */ +bool cpu_topology__core_wide(const struct cpu_topology *topology, + const char *user_requested_cpu_list); struct numa_topology *numa_topology__new(void); void numa_topology__delete(struct numa_topology *tp); diff --git a/tools/perf/util/dso.c b/tools/perf/util/dso.c index 5ac13958d1bd..f1a14c0ad26d 100644 --- a/tools/perf/util/dso.c +++ b/tools/perf/util/dso.c @@ -501,6 +501,7 @@ static int __open_dso(struct dso *dso, struct machine *machine) if (!name) return -ENOMEM; + mutex_lock(&dso->lock); if (machine) root_dir = machine->root_dir; @@ -541,6 +542,7 @@ static int __open_dso(struct dso *dso, struct machine *machine) unlink(name); out: + mutex_unlock(&dso->lock); free(name); return fd; } @@ -559,8 +561,11 @@ static int open_dso(struct dso *dso, struct machine *machine) int fd; struct nscookie nsc; - if (dso->binary_type != DSO_BINARY_TYPE__BUILD_ID_CACHE) + if (dso->binary_type != DSO_BINARY_TYPE__BUILD_ID_CACHE) { + mutex_lock(&dso->lock); nsinfo__mountns_enter(dso->nsinfo, &nsc); + mutex_unlock(&dso->lock); + } fd = __open_dso(dso, machine); if (dso->binary_type != DSO_BINARY_TYPE__BUILD_ID_CACHE) nsinfo__mountns_exit(&nsc); @@ -795,7 +800,7 @@ dso_cache__free(struct dso *dso) struct rb_root *root = &dso->data.cache; struct rb_node *next = rb_first(root); - pthread_mutex_lock(&dso->lock); + mutex_lock(&dso->lock); while (next) { struct dso_cache *cache; @@ -804,7 +809,7 @@ dso_cache__free(struct dso *dso) rb_erase(&cache->rb_node, root); free(cache); } - pthread_mutex_unlock(&dso->lock); + mutex_unlock(&dso->lock); } static struct dso_cache *__dso_cache__find(struct dso *dso, u64 offset) @@ -841,7 +846,7 @@ dso_cache__insert(struct dso *dso, struct dso_cache *new) struct dso_cache *cache; u64 offset = new->offset; - pthread_mutex_lock(&dso->lock); + mutex_lock(&dso->lock); while (*p != NULL) { u64 end; @@ -862,7 +867,7 @@ dso_cache__insert(struct dso *dso, struct dso_cache *new) cache = NULL; out: - pthread_mutex_unlock(&dso->lock); + mutex_unlock(&dso->lock); return cache; } @@ -1297,7 +1302,7 @@ struct dso *dso__new_id(const char *name, struct dso_id *id) dso->root = NULL; INIT_LIST_HEAD(&dso->node); INIT_LIST_HEAD(&dso->data.open_entry); - pthread_mutex_init(&dso->lock, NULL); + mutex_init(&dso->lock); refcount_set(&dso->refcnt, 1); } @@ -1336,7 +1341,7 @@ void dso__delete(struct dso *dso) dso__free_a2l(dso); zfree(&dso->symsrc_filename); nsinfo__zput(dso->nsinfo); - pthread_mutex_destroy(&dso->lock); + mutex_destroy(&dso->lock); free(dso); } diff --git a/tools/perf/util/dso.h b/tools/perf/util/dso.h index 66981c7a9a18..58d94175e714 100644 --- a/tools/perf/util/dso.h +++ b/tools/perf/util/dso.h @@ -2,7 +2,6 @@ #ifndef __PERF_DSO #define __PERF_DSO -#include <pthread.h> #include <linux/refcount.h> #include <linux/types.h> #include <linux/rbtree.h> @@ -11,6 +10,7 @@ #include <stdio.h> #include <linux/bitops.h> #include "build-id.h" +#include "mutex.h" struct machine; struct map; @@ -145,7 +145,7 @@ struct dso_cache { struct auxtrace_cache; struct dso { - pthread_mutex_t lock; + struct mutex lock; struct list_head node; struct rb_node rb_node; /* rbtree node sorted by long name */ struct rb_root *root; /* root of rbtree that rb_node is in */ diff --git a/tools/perf/util/events_stats.h b/tools/perf/util/events_stats.h index 040ab9d0a803..8fecc9fbaecc 100644 --- a/tools/perf/util/events_stats.h +++ b/tools/perf/util/events_stats.h @@ -47,6 +47,7 @@ struct hists_stats { u64 total_non_filtered_period; u32 nr_samples; u32 nr_non_filtered_samples; + u32 nr_lost_samples; }; void events_stats__inc(struct events_stats *stats, u32 type); diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index 48167f3941a6..6612b00949e7 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -15,6 +15,7 @@ #include "target.h" #include "evlist.h" #include "evsel.h" +#include "record.h" #include "debug.h" #include "units.h" #include "bpf_counter.h" @@ -40,12 +41,14 @@ #include <sys/ioctl.h> #include <sys/mman.h> #include <sys/prctl.h> +#include <sys/timerfd.h> #include <linux/bitops.h> #include <linux/hash.h> #include <linux/log2.h> #include <linux/err.h> #include <linux/string.h> +#include <linux/time64.h> #include <linux/zalloc.h> #include <perf/evlist.h> #include <perf/evsel.h> @@ -147,6 +150,7 @@ static void evlist__purge(struct evlist *evlist) void evlist__exit(struct evlist *evlist) { + event_enable_timer__exit(&evlist->eet); zfree(&evlist->mmap); zfree(&evlist->overwrite_mmap); perf_evlist__exit(&evlist->core); @@ -264,28 +268,6 @@ int evlist__add_dummy(struct evlist *evlist) return 0; } -static void evlist__add_on_all_cpus(struct evlist *evlist, struct evsel *evsel) -{ - evsel->core.system_wide = true; - - /* - * All CPUs. - * - * Note perf_event_open() does not accept CPUs that are not online, so - * in fact this CPU list will include only all online CPUs. - */ - perf_cpu_map__put(evsel->core.own_cpus); - evsel->core.own_cpus = perf_cpu_map__new(NULL); - perf_cpu_map__put(evsel->core.cpus); - evsel->core.cpus = perf_cpu_map__get(evsel->core.own_cpus); - - /* No threads */ - perf_thread_map__put(evsel->core.threads); - evsel->core.threads = perf_thread_map__new_dummy(); - - evlist__add(evlist, evsel); -} - struct evsel *evlist__add_aux_dummy(struct evlist *evlist, bool system_wide) { struct evsel *evsel = evlist__dummy_event(evlist); @@ -298,17 +280,31 @@ struct evsel *evlist__add_aux_dummy(struct evlist *evlist, bool system_wide) evsel->core.attr.exclude_hv = 1; evsel->core.attr.freq = 0; evsel->core.attr.sample_period = 1; + evsel->core.system_wide = system_wide; evsel->no_aux_samples = true; evsel->name = strdup("dummy:u"); - if (system_wide) - evlist__add_on_all_cpus(evlist, evsel); - else - evlist__add(evlist, evsel); - + evlist__add(evlist, evsel); return evsel; } +struct evsel *evlist__add_sched_switch(struct evlist *evlist, bool system_wide) +{ + struct evsel *evsel = evsel__newtp_idx("sched", "sched_switch", 0); + + if (IS_ERR(evsel)) + return evsel; + + evsel__set_sample_bit(evsel, CPU); + evsel__set_sample_bit(evsel, TIME); + + evsel->core.system_wide = system_wide; + evsel->no_aux_samples = true; + + evlist__add(evlist, evsel); + return evsel; +}; + int evlist__add_attrs(struct evlist *evlist, struct perf_event_attr *attrs, size_t nr_attrs) { struct evsel *evsel, *n; @@ -480,7 +476,7 @@ static int evlist__is_enabled(struct evlist *evlist) return false; } -static void __evlist__disable(struct evlist *evlist, char *evsel_name) +static void __evlist__disable(struct evlist *evlist, char *evsel_name, bool excl_dummy) { struct evsel *pos; struct evlist_cpu_iterator evlist_cpu_itr; @@ -502,6 +498,8 @@ static void __evlist__disable(struct evlist *evlist, char *evsel_name) continue; if (pos->disabled || !evsel__is_group_leader(pos) || !pos->core.fd) continue; + if (excl_dummy && evsel__is_dummy_event(pos)) + continue; if (pos->immediate) has_imm = true; if (pos->immediate != imm) @@ -518,6 +516,8 @@ static void __evlist__disable(struct evlist *evlist, char *evsel_name) continue; if (!evsel__is_group_leader(pos) || !pos->core.fd) continue; + if (excl_dummy && evsel__is_dummy_event(pos)) + continue; pos->disabled = true; } @@ -533,15 +533,20 @@ static void __evlist__disable(struct evlist *evlist, char *evsel_name) void evlist__disable(struct evlist *evlist) { - __evlist__disable(evlist, NULL); + __evlist__disable(evlist, NULL, false); +} + +void evlist__disable_non_dummy(struct evlist *evlist) +{ + __evlist__disable(evlist, NULL, true); } void evlist__disable_evsel(struct evlist *evlist, char *evsel_name) { - __evlist__disable(evlist, evsel_name); + __evlist__disable(evlist, evsel_name, false); } -static void __evlist__enable(struct evlist *evlist, char *evsel_name) +static void __evlist__enable(struct evlist *evlist, char *evsel_name, bool excl_dummy) { struct evsel *pos; struct evlist_cpu_iterator evlist_cpu_itr; @@ -560,6 +565,8 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name) continue; if (!evsel__is_group_leader(pos) || !pos->core.fd) continue; + if (excl_dummy && evsel__is_dummy_event(pos)) + continue; evsel__enable_cpu(pos, evlist_cpu_itr.cpu_map_idx); } affinity__cleanup(affinity); @@ -568,6 +575,8 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name) continue; if (!evsel__is_group_leader(pos) || !pos->core.fd) continue; + if (excl_dummy && evsel__is_dummy_event(pos)) + continue; pos->disabled = false; } @@ -581,12 +590,17 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name) void evlist__enable(struct evlist *evlist) { - __evlist__enable(evlist, NULL); + __evlist__enable(evlist, NULL, false); +} + +void evlist__enable_non_dummy(struct evlist *evlist) +{ + __evlist__enable(evlist, NULL, true); } void evlist__enable_evsel(struct evlist *evlist, char *evsel_name) { - __evlist__enable(evlist, evsel_name); + __evlist__enable(evlist, evsel_name, false); } void evlist__toggle_enable(struct evlist *evlist) @@ -608,7 +622,8 @@ int evlist__filter_pollfd(struct evlist *evlist, short revents_and_mask) int evlist__add_wakeup_eventfd(struct evlist *evlist, int fd) { return perf_evlist__add_pollfd(&evlist->core, fd, NULL, POLLIN, - fdarray_flag__nonfilterable); + fdarray_flag__nonfilterable | + fdarray_flag__non_perf_event); } #endif @@ -1897,7 +1912,8 @@ int evlist__initialize_ctlfd(struct evlist *evlist, int fd, int ack) } evlist->ctl_fd.pos = perf_evlist__add_pollfd(&evlist->core, fd, NULL, POLLIN, - fdarray_flag__nonfilterable); + fdarray_flag__nonfilterable | + fdarray_flag__non_perf_event); if (evlist->ctl_fd.pos < 0) { evlist->ctl_fd.pos = -1; pr_err("Failed to add ctl fd entry: %m\n"); @@ -2147,20 +2163,234 @@ int evlist__ctlfd_process(struct evlist *evlist, enum evlist_ctl_cmd *cmd) return err; } -int evlist__ctlfd_update(struct evlist *evlist, struct pollfd *update) +/** + * struct event_enable_time - perf record -D/--delay single time range. + * @start: start of time range to enable events in milliseconds + * @end: end of time range to enable events in milliseconds + * + * N.B. this structure is also accessed as an array of int. + */ +struct event_enable_time { + int start; + int end; +}; + +static int parse_event_enable_time(const char *str, struct event_enable_time *range, bool first) { - int ctlfd_pos = evlist->ctl_fd.pos; - struct pollfd *entries = evlist->core.pollfd.entries; + const char *fmt = first ? "%u - %u %n" : " , %u - %u %n"; + int ret, start, end, n; - if (!evlist__ctlfd_initialized(evlist)) + ret = sscanf(str, fmt, &start, &end, &n); + if (ret != 2 || end <= start) + return -EINVAL; + if (range) { + range->start = start; + range->end = end; + } + return n; +} + +static ssize_t parse_event_enable_times(const char *str, struct event_enable_time *range) +{ + int incr = !!range; + bool first = true; + ssize_t ret, cnt; + + for (cnt = 0; *str; cnt++) { + ret = parse_event_enable_time(str, range, first); + if (ret < 0) + return ret; + /* Check no overlap */ + if (!first && range && range->start <= range[-1].end) + return -EINVAL; + str += ret; + range += incr; + first = false; + } + return cnt; +} + +/** + * struct event_enable_timer - control structure for perf record -D/--delay. + * @evlist: event list + * @times: time ranges that events are enabled (N.B. this is also accessed as an + * array of int) + * @times_cnt: number of time ranges + * @timerfd: timer file descriptor + * @pollfd_pos: position in @evlist array of file descriptors to poll (fdarray) + * @times_step: current position in (int *)@times)[], + * refer event_enable_timer__process() + * + * Note, this structure is only used when there are time ranges, not when there + * is only an initial delay. + */ +struct event_enable_timer { + struct evlist *evlist; + struct event_enable_time *times; + size_t times_cnt; + int timerfd; + int pollfd_pos; + size_t times_step; +}; + +static int str_to_delay(const char *str) +{ + char *endptr; + long d; + + d = strtol(str, &endptr, 10); + if (*endptr || d > INT_MAX || d < -1) return 0; + return d; +} - if (entries[ctlfd_pos].fd != update->fd || - entries[ctlfd_pos].events != update->events) - return -1; +int evlist__parse_event_enable_time(struct evlist *evlist, struct record_opts *opts, + const char *str, int unset) +{ + enum fdarray_flags flags = fdarray_flag__nonfilterable | fdarray_flag__non_perf_event; + struct event_enable_timer *eet; + ssize_t times_cnt; + ssize_t ret; + int err; + + if (unset) + return 0; + + opts->initial_delay = str_to_delay(str); + if (opts->initial_delay) + return 0; + + ret = parse_event_enable_times(str, NULL); + if (ret < 0) + return ret; + + times_cnt = ret; + if (times_cnt == 0) + return -EINVAL; + + eet = zalloc(sizeof(*eet)); + if (!eet) + return -ENOMEM; + + eet->times = calloc(times_cnt, sizeof(*eet->times)); + if (!eet->times) { + err = -ENOMEM; + goto free_eet; + } + + if (parse_event_enable_times(str, eet->times) != times_cnt) { + err = -EINVAL; + goto free_eet_times; + } + + eet->times_cnt = times_cnt; + + eet->timerfd = timerfd_create(CLOCK_MONOTONIC, TFD_CLOEXEC); + if (eet->timerfd == -1) { + err = -errno; + pr_err("timerfd_create failed: %s\n", strerror(errno)); + goto free_eet_times; + } + + eet->pollfd_pos = perf_evlist__add_pollfd(&evlist->core, eet->timerfd, NULL, POLLIN, flags); + if (eet->pollfd_pos < 0) { + err = eet->pollfd_pos; + goto close_timerfd; + } + + eet->evlist = evlist; + evlist->eet = eet; + opts->initial_delay = eet->times[0].start; - entries[ctlfd_pos].revents = update->revents; return 0; + +close_timerfd: + close(eet->timerfd); +free_eet_times: + free(eet->times); +free_eet: + free(eet); + return err; +} + +static int event_enable_timer__set_timer(struct event_enable_timer *eet, int ms) +{ + struct itimerspec its = { + .it_value.tv_sec = ms / MSEC_PER_SEC, + .it_value.tv_nsec = (ms % MSEC_PER_SEC) * NSEC_PER_MSEC, + }; + int err = 0; + + if (timerfd_settime(eet->timerfd, 0, &its, NULL) < 0) { + err = -errno; + pr_err("timerfd_settime failed: %s\n", strerror(errno)); + } + return err; +} + +int event_enable_timer__start(struct event_enable_timer *eet) +{ + int ms; + + if (!eet) + return 0; + + ms = eet->times[0].end - eet->times[0].start; + eet->times_step = 1; + + return event_enable_timer__set_timer(eet, ms); +} + +int event_enable_timer__process(struct event_enable_timer *eet) +{ + struct pollfd *entries; + short revents; + + if (!eet) + return 0; + + entries = eet->evlist->core.pollfd.entries; + revents = entries[eet->pollfd_pos].revents; + entries[eet->pollfd_pos].revents = 0; + + if (revents & POLLIN) { + size_t step = eet->times_step; + size_t pos = step / 2; + + if (step & 1) { + evlist__disable_non_dummy(eet->evlist); + pr_info(EVLIST_DISABLED_MSG); + if (pos >= eet->times_cnt - 1) { + /* Disarm timer */ + event_enable_timer__set_timer(eet, 0); + return 1; /* Stop */ + } + } else { + evlist__enable_non_dummy(eet->evlist); + pr_info(EVLIST_ENABLED_MSG); + } + + step += 1; + pos = step / 2; + + if (pos < eet->times_cnt) { + int *times = (int *)eet->times; /* Accessing 'times' as array of int */ + int ms = times[step] - times[step - 1]; + + eet->times_step = step; + return event_enable_timer__set_timer(eet, ms); + } + } + + return 0; +} + +void event_enable_timer__exit(struct event_enable_timer **ep) +{ + if (!ep || !*ep) + return; + free((*ep)->times); + zfree(ep); } struct evsel *evlist__find_evsel(struct evlist *evlist, int idx) diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index 351ba2887a79..16734c6756b3 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -48,6 +48,8 @@ enum bkw_mmap_state { BKW_MMAP_EMPTY, }; +struct event_enable_timer; + struct evlist { struct perf_evlist core; bool enabled; @@ -79,6 +81,7 @@ struct evlist { int ack; /* ack file descriptor for control commands */ int pos; /* index at evlist core object to check signals */ } ctl_fd; + struct event_enable_timer *eet; }; struct evsel_str_handler { @@ -124,6 +127,7 @@ static inline struct evsel *evlist__add_dummy_on_all_cpus(struct evlist *evlist) { return evlist__add_aux_dummy(evlist, true); } +struct evsel *evlist__add_sched_switch(struct evlist *evlist, bool system_wide); int evlist__add_sb_event(struct evlist *evlist, struct perf_event_attr *attr, evsel__sb_cb_t cb, void *data); @@ -205,6 +209,8 @@ void evlist__enable(struct evlist *evlist); void evlist__toggle_enable(struct evlist *evlist); void evlist__disable_evsel(struct evlist *evlist, char *evsel_name); void evlist__enable_evsel(struct evlist *evlist, char *evsel_name); +void evlist__disable_non_dummy(struct evlist *evlist); +void evlist__enable_non_dummy(struct evlist *evlist); void evlist__set_selected(struct evlist *evlist, struct evsel *evsel); @@ -418,13 +424,18 @@ void evlist__close_control(int ctl_fd, int ctl_fd_ack, bool *ctl_fd_close); int evlist__initialize_ctlfd(struct evlist *evlist, int ctl_fd, int ctl_fd_ack); int evlist__finalize_ctlfd(struct evlist *evlist); bool evlist__ctlfd_initialized(struct evlist *evlist); -int evlist__ctlfd_update(struct evlist *evlist, struct pollfd *update); int evlist__ctlfd_process(struct evlist *evlist, enum evlist_ctl_cmd *cmd); int evlist__ctlfd_ack(struct evlist *evlist); #define EVLIST_ENABLED_MSG "Events enabled\n" #define EVLIST_DISABLED_MSG "Events disabled\n" +int evlist__parse_event_enable_time(struct evlist *evlist, struct record_opts *opts, + const char *str, int unset); +int event_enable_timer__start(struct event_enable_timer *eet); +void event_enable_timer__exit(struct event_enable_timer **ep); +int event_enable_timer__process(struct event_enable_timer *eet); + struct evsel *evlist__find_evsel(struct evlist *evlist, int idx); int evlist__scnprintf_evsels(struct evlist *evlist, size_t size, char *bf); diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index 18c3eb864d55..76605fde3507 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -46,7 +46,11 @@ #include "string2.h" #include "memswap.h" #include "util.h" -#include "hashmap.h" +#ifdef HAVE_LIBBPF_SUPPORT +#include <bpf/hashmap.h> +#else +#include "util/hashmap.h" +#endif #include "pmu-hybrid.h" #include "off_cpu.h" #include "../perf-sys.h" @@ -1157,6 +1161,7 @@ void evsel__config(struct evsel *evsel, struct record_opts *opts, attr->sample_id_all = perf_missing_features.sample_id_all ? 0 : 1; attr->inherit = !opts->no_inherit; attr->write_backward = opts->overwrite ? 1 : 0; + attr->read_format = PERF_FORMAT_LOST; evsel__set_sample_bit(evsel, IP); evsel__set_sample_bit(evsel, TID); @@ -1808,7 +1813,7 @@ static struct perf_thread_map *empty_thread_map; static int __evsel__prepare_open(struct evsel *evsel, struct perf_cpu_map *cpus, struct perf_thread_map *threads) { - int nthreads; + int nthreads = perf_thread_map__nr(threads); if ((perf_missing_features.write_backward && evsel->core.attr.write_backward) || (perf_missing_features.aux_output && evsel->core.attr.aux_output)) @@ -1834,11 +1839,6 @@ static int __evsel__prepare_open(struct evsel *evsel, struct perf_cpu_map *cpus, threads = empty_thread_map; } - if (evsel->core.system_wide) - nthreads = 1; - else - nthreads = threads->nr; - if (evsel->core.fd == NULL && perf_evsel__alloc_fd(&evsel->core, perf_cpu_map__nr(cpus), nthreads) < 0) return -ENOMEM; @@ -1852,6 +1852,8 @@ static int __evsel__prepare_open(struct evsel *evsel, struct perf_cpu_map *cpus, static void evsel__disable_missing_features(struct evsel *evsel) { + if (perf_missing_features.read_lost) + evsel->core.attr.read_format &= ~PERF_FORMAT_LOST; if (perf_missing_features.weight_struct) { evsel__set_sample_bit(evsel, WEIGHT); evsel__reset_sample_bit(evsel, WEIGHT_STRUCT); @@ -1903,7 +1905,12 @@ bool evsel__detect_missing_features(struct evsel *evsel) * Must probe features in the order they were added to the * perf_event_attr interface. */ - if (!perf_missing_features.weight_struct && + if (!perf_missing_features.read_lost && + (evsel->core.attr.read_format & PERF_FORMAT_LOST)) { + perf_missing_features.read_lost = true; + pr_debug2("switching off PERF_FORMAT_LOST support\n"); + return true; + } else if (!perf_missing_features.weight_struct && (evsel->core.attr.sample_type & PERF_SAMPLE_WEIGHT_STRUCT)) { perf_missing_features.weight_struct = true; pr_debug2("switching off weight struct support\n"); @@ -2049,10 +2056,7 @@ static int evsel__open_cpu(struct evsel *evsel, struct perf_cpu_map *cpus, if (threads == NULL) threads = empty_thread_map; - if (evsel->core.system_wide) - nthreads = 1; - else - nthreads = threads->nr; + nthreads = perf_thread_map__nr(threads); if (evsel->cgrp) pid = evsel->cgrp->fd; @@ -2077,6 +2081,7 @@ retry_open: test_attr__ready(); + /* Debug message used by test scripts */ pr_debug2_peo("sys_perf_event_open: pid %d cpu %d group_fd %d flags %#lx", pid, perf_cpu_map__cpu(cpus, idx).cpu, group_fd, evsel->open_flags); @@ -2102,6 +2107,7 @@ retry_open: fd, group_fd, evsel->open_flags); } + /* Debug message used by test scripts */ pr_debug2_peo(" = %d\n", fd); if (evsel->bpf_fd >= 0) { diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h index d927713b513e..989865e16aad 100644 --- a/tools/perf/util/evsel.h +++ b/tools/perf/util/evsel.h @@ -188,6 +188,7 @@ struct perf_missing_features { bool data_page_size; bool code_page_size; bool weight_struct; + bool read_lost; }; extern struct perf_missing_features perf_missing_features; diff --git a/tools/perf/util/expr.c b/tools/perf/util/expr.c index c15a9852fa41..aaacf514dc09 100644 --- a/tools/perf/util/expr.c +++ b/tools/perf/util/expr.c @@ -182,7 +182,7 @@ int expr__add_ref(struct expr_parse_ctx *ctx, struct metric_ref *ref) { struct expr_id_data *data_ptr = NULL, *old_data = NULL; char *old_key = NULL; - char *name, *p; + char *name; int ret; data_ptr = zalloc(sizeof(*data_ptr)); @@ -196,15 +196,6 @@ int expr__add_ref(struct expr_parse_ctx *ctx, struct metric_ref *ref) } /* - * The jevents tool converts all metric expressions - * to lowercase, including metric references, hence - * we need to add lowercase name for metric, so it's - * properly found. - */ - for (p = name; *p; p++) - *p = tolower(*p); - - /* * Intentionally passing just const char pointers, * originally from 'struct pmu_event' object. * We don't need to change them, so there's no @@ -310,7 +301,9 @@ struct expr_parse_ctx *expr__ctx_new(void) free(ctx); return NULL; } - ctx->runtime = 0; + ctx->sctx.user_requested_cpu_list = NULL; + ctx->sctx.runtime = 0; + ctx->sctx.system_wide = false; return ctx; } @@ -332,6 +325,10 @@ void expr__ctx_free(struct expr_parse_ctx *ctx) struct hashmap_entry *cur; size_t bkt; + if (!ctx) + return; + + free(ctx->sctx.user_requested_cpu_list); hashmap__for_each_entry(ctx->ids, cur, bkt) { free((char *)cur->key); free(cur->value); @@ -344,16 +341,13 @@ static int __expr__parse(double *val, struct expr_parse_ctx *ctx, const char *expr, bool compute_ids) { - struct expr_scanner_ctx scanner_ctx = { - .runtime = ctx->runtime, - }; YY_BUFFER_STATE buffer; void *scanner; int ret; pr_debug2("parsing metric: %s\n", expr); - ret = expr_lex_init_extra(&scanner_ctx, &scanner); + ret = expr_lex_init_extra(&ctx->sctx, &scanner); if (ret) return ret; @@ -410,16 +404,11 @@ double arch_get_tsc_freq(void) } #endif -double expr__get_literal(const char *literal) +double expr__get_literal(const char *literal, const struct expr_scanner_ctx *ctx) { static struct cpu_topology *topology; double result = NAN; - if (!strcasecmp("#smt_on", literal)) { - result = smt_on() > 0 ? 1.0 : 0.0; - goto out; - } - if (!strcmp("#num_cpus", literal)) { result = cpu__max_present_cpu().cpu; goto out; @@ -443,6 +432,15 @@ double expr__get_literal(const char *literal) goto out; } } + if (!strcasecmp("#smt_on", literal)) { + result = smt_on(topology) ? 1.0 : 0.0; + goto out; + } + if (!strcmp("#core_wide", literal)) { + result = core_wide(ctx->system_wide, ctx->user_requested_cpu_list, topology) + ? 1.0 : 0.0; + goto out; + } if (!strcmp("#num_packages", literal)) { result = topology->package_cpus_lists; goto out; diff --git a/tools/perf/util/expr.h b/tools/perf/util/expr.h index bd2116983bbb..d6c1668dc1a0 100644 --- a/tools/perf/util/expr.h +++ b/tools/perf/util/expr.h @@ -2,28 +2,27 @@ #ifndef PARSE_CTX_H #define PARSE_CTX_H 1 -// There are fixes that need to land upstream before we can use libbpf's headers, -// for now use our copy unconditionally, since the data structures at this point -// are exactly the same, no problem. -//#ifdef HAVE_LIBBPF_SUPPORT -//#include <bpf/hashmap.h> -//#else +#ifdef HAVE_LIBBPF_SUPPORT +#include <bpf/hashmap.h> +#else #include "util/hashmap.h" -//#endif +#endif struct metric_ref; +struct expr_scanner_ctx { + char *user_requested_cpu_list; + int runtime; + bool system_wide; +}; + struct expr_parse_ctx { struct hashmap *ids; - int runtime; + struct expr_scanner_ctx sctx; }; struct expr_id_data; -struct expr_scanner_ctx { - int runtime; -}; - struct hashmap *ids__new(void); void ids__free(struct hashmap *ids); int ids__insert(struct hashmap *ids, const char *id); @@ -58,6 +57,6 @@ int expr__find_ids(const char *expr, const char *one, double expr_id_data__value(const struct expr_id_data *data); double expr_id_data__source_count(const struct expr_id_data *data); -double expr__get_literal(const char *literal); +double expr__get_literal(const char *literal, const struct expr_scanner_ctx *ctx); #endif diff --git a/tools/perf/util/expr.l b/tools/perf/util/expr.l index 4dc8edbfd9ce..0168a9637330 100644 --- a/tools/perf/util/expr.l +++ b/tools/perf/util/expr.l @@ -79,11 +79,11 @@ static int str(yyscan_t scanner, int token, int runtime) return token; } -static int literal(yyscan_t scanner) +static int literal(yyscan_t scanner, const struct expr_scanner_ctx *sctx) { YYSTYPE *yylval = expr_get_lval(scanner); - yylval->num = expr__get_literal(expr_get_text(scanner)); + yylval->num = expr__get_literal(expr_get_text(scanner), sctx); if (isnan(yylval->num)) return EXPR_ERROR; @@ -108,7 +108,7 @@ min { return MIN; } if { return IF; } else { return ELSE; } source_count { return SOURCE_COUNT; } -{literal} { return literal(yyscanner); } +{literal} { return literal(yyscanner, sctx); } {number} { return value(yyscanner); } {symbol} { return str(yyscanner, ID, sctx->runtime); } "|" { return '|'; } diff --git a/tools/perf/util/expr.y b/tools/perf/util/expr.y index a30b825adb7b..635e562350c5 100644 --- a/tools/perf/util/expr.y +++ b/tools/perf/util/expr.y @@ -156,7 +156,7 @@ start: if_expr } ; -if_expr: expr IF expr ELSE expr +if_expr: expr IF expr ELSE if_expr { if (fpclassify($3.val) == FP_ZERO) { /* diff --git a/tools/perf/util/genelf.c b/tools/perf/util/genelf.c index d81b54563e96..fefc72066c4e 100644 --- a/tools/perf/util/genelf.c +++ b/tools/perf/util/genelf.c @@ -345,6 +345,7 @@ jit_write_elf(int fd, uint64_t load_addr, const char *sym, eh_frame_base_offset); if (retval) goto error; + retval = -1; } /* diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index c30c29c51410..98dfaf84bd13 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -4295,8 +4295,6 @@ out: size_t perf_event__fprintf_event_update(union perf_event *event, FILE *fp) { struct perf_record_event_update *ev = &event->event_update; - struct perf_record_event_update_scale *ev_scale; - struct perf_record_event_update_cpus *ev_cpus; struct perf_cpu_map *map; size_t ret; @@ -4304,20 +4302,18 @@ size_t perf_event__fprintf_event_update(union perf_event *event, FILE *fp) switch (ev->type) { case PERF_EVENT_UPDATE__SCALE: - ev_scale = (struct perf_record_event_update_scale *)ev->data; - ret += fprintf(fp, "... scale: %f\n", ev_scale->scale); + ret += fprintf(fp, "... scale: %f\n", ev->scale.scale); break; case PERF_EVENT_UPDATE__UNIT: - ret += fprintf(fp, "... unit: %s\n", ev->data); + ret += fprintf(fp, "... unit: %s\n", ev->unit); break; case PERF_EVENT_UPDATE__NAME: - ret += fprintf(fp, "... name: %s\n", ev->data); + ret += fprintf(fp, "... name: %s\n", ev->name); break; case PERF_EVENT_UPDATE__CPUS: - ev_cpus = (struct perf_record_event_update_cpus *)ev->data; ret += fprintf(fp, "... "); - map = cpu_map__new_data(&ev_cpus->cpus); + map = cpu_map__new_data(&ev->cpus.cpus); if (map) ret += cpu_map__fprintf(map, fp); else @@ -4374,8 +4370,6 @@ int perf_event__process_event_update(struct perf_tool *tool __maybe_unused, struct evlist **pevlist) { struct perf_record_event_update *ev = &event->event_update; - struct perf_record_event_update_scale *ev_scale; - struct perf_record_event_update_cpus *ev_cpus; struct evlist *evlist; struct evsel *evsel; struct perf_cpu_map *map; @@ -4395,19 +4389,17 @@ int perf_event__process_event_update(struct perf_tool *tool __maybe_unused, switch (ev->type) { case PERF_EVENT_UPDATE__UNIT: free((char *)evsel->unit); - evsel->unit = strdup(ev->data); + evsel->unit = strdup(ev->unit); break; case PERF_EVENT_UPDATE__NAME: free(evsel->name); - evsel->name = strdup(ev->data); + evsel->name = strdup(ev->name); break; case PERF_EVENT_UPDATE__SCALE: - ev_scale = (struct perf_record_event_update_scale *)ev->data; - evsel->scale = ev_scale->scale; + evsel->scale = ev->scale.scale; break; case PERF_EVENT_UPDATE__CPUS: - ev_cpus = (struct perf_record_event_update_cpus *)ev->data; - map = cpu_map__new_data(&ev_cpus->cpus); + map = cpu_map__new_data(&ev->cpus.cpus); if (map) { perf_cpu_map__put(evsel->core.own_cpus); evsel->core.own_cpus = map; diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c index 1c085ab56534..17a05e943b44 100644 --- a/tools/perf/util/hist.c +++ b/tools/perf/util/hist.c @@ -215,6 +215,7 @@ void hists__calc_col_len(struct hists *hists, struct hist_entry *h) hists__new_col_len(hists, HISTC_GLOBAL_INS_LAT, 13); hists__new_col_len(hists, HISTC_LOCAL_P_STAGE_CYC, 13); hists__new_col_len(hists, HISTC_GLOBAL_P_STAGE_CYC, 13); + hists__new_col_len(hists, HISTC_ADDR, BITS_PER_LONG / 4 + 2); if (symbol_conf.nanosecs) hists__new_col_len(hists, HISTC_TIME, 16); @@ -1622,13 +1623,13 @@ struct rb_root_cached *hists__get_rotate_entries_in(struct hists *hists) { struct rb_root_cached *root; - pthread_mutex_lock(&hists->lock); + mutex_lock(&hists->lock); root = hists->entries_in; if (++hists->entries_in > &hists->entries_in_array[1]) hists->entries_in = &hists->entries_in_array[0]; - pthread_mutex_unlock(&hists->lock); + mutex_unlock(&hists->lock); return root; } @@ -2335,6 +2336,11 @@ void hists__inc_nr_samples(struct hists *hists, bool filtered) hists->stats.nr_non_filtered_samples++; } +void hists__inc_nr_lost_samples(struct hists *hists, u32 lost) +{ + hists->stats.nr_lost_samples += lost; +} + static struct hist_entry *hists__add_dummy_entry(struct hists *hists, struct hist_entry *pair) { @@ -2678,12 +2684,16 @@ size_t evlist__fprintf_nr_events(struct evlist *evlist, FILE *fp, evlist__for_each_entry(evlist, pos) { struct hists *hists = evsel__hists(pos); - if (skip_empty && !hists->stats.nr_samples) + if (skip_empty && !hists->stats.nr_samples && !hists->stats.nr_lost_samples) continue; ret += fprintf(fp, "%s stats:\n", evsel__name(pos)); - ret += fprintf(fp, "%16s events: %10d\n", - "SAMPLE", hists->stats.nr_samples); + if (hists->stats.nr_samples) + ret += fprintf(fp, "%16s events: %10d\n", + "SAMPLE", hists->stats.nr_samples); + if (hists->stats.nr_lost_samples) + ret += fprintf(fp, "%16s events: %10d\n", + "LOST_SAMPLES", hists->stats.nr_lost_samples); } return ret; @@ -2805,7 +2815,7 @@ int __hists__init(struct hists *hists, struct perf_hpp_list *hpp_list) hists->entries_in = &hists->entries_in_array[0]; hists->entries_collapsed = RB_ROOT_CACHED; hists->entries = RB_ROOT_CACHED; - pthread_mutex_init(&hists->lock, NULL); + mutex_init(&hists->lock); hists->socket_filter = -1; hists->hpp_list = hpp_list; INIT_LIST_HEAD(&hists->hpp_formats); diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h index 7ed4648d2fc2..ebd8a8f783ee 100644 --- a/tools/perf/util/hist.h +++ b/tools/perf/util/hist.h @@ -4,10 +4,10 @@ #include <linux/rbtree.h> #include <linux/types.h> -#include <pthread.h> #include "evsel.h" #include "color.h" #include "events_stats.h" +#include "mutex.h" struct hist_entry; struct hist_entry_ops; @@ -79,6 +79,7 @@ enum hist_column { HISTC_GLOBAL_P_STAGE_CYC, HISTC_ADDR_FROM, HISTC_ADDR_TO, + HISTC_ADDR, HISTC_NR_COLS, /* Last entry */ }; @@ -98,7 +99,7 @@ struct hists { const struct dso *dso_filter; const char *uid_filter_str; const char *symbol_filter_str; - pthread_mutex_t lock; + struct mutex lock; struct hists_stats stats; u64 event_stream; u16 col_len[HISTC_NR_COLS]; @@ -201,6 +202,7 @@ void hists__reset_stats(struct hists *hists); void hists__inc_stats(struct hists *hists, struct hist_entry *h); void hists__inc_nr_events(struct hists *hists); void hists__inc_nr_samples(struct hists *hists, bool filtered); +void hists__inc_nr_lost_samples(struct hists *hists, u32 lost); size_t hists__fprintf(struct hists *hists, bool show_header, int max_rows, int max_cols, float min_pcnt, FILE *fp, diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-log.c b/tools/perf/util/intel-pt-decoder/intel-pt-log.c index 5f5dfc8753f3..ef55d6232cf0 100644 --- a/tools/perf/util/intel-pt-decoder/intel-pt-log.c +++ b/tools/perf/util/intel-pt-decoder/intel-pt-log.c @@ -5,12 +5,16 @@ */ #include <stdio.h> +#include <stdlib.h> #include <stdint.h> #include <inttypes.h> #include <stdarg.h> #include <stdbool.h> #include <string.h> +#include <linux/zalloc.h> +#include <linux/kernel.h> + #include "intel-pt-log.h" #include "intel-pt-insn-decoder.h" @@ -18,18 +22,33 @@ #define MAX_LOG_NAME 256 +#define DFLT_BUF_SZ (16 * 1024) + +struct log_buf { + char *buf; + size_t buf_sz; + size_t head; + bool wrapped; + FILE *backend; +}; + static FILE *f; static char log_name[MAX_LOG_NAME]; bool intel_pt_enable_logging; +static bool intel_pt_dump_log_on_error; +static unsigned int intel_pt_log_on_error_size; +static struct log_buf log_buf; void *intel_pt_log_fp(void) { return f; } -void intel_pt_log_enable(void) +void intel_pt_log_enable(bool dump_log_on_error, unsigned int log_on_error_size) { intel_pt_enable_logging = true; + intel_pt_dump_log_on_error = dump_log_on_error; + intel_pt_log_on_error_size = log_on_error_size; } void intel_pt_log_disable(void) @@ -74,6 +93,100 @@ static void intel_pt_print_no_data(uint64_t pos, int indent) fprintf(f, " "); } +static ssize_t log_buf__write(void *cookie, const char *buf, size_t size) +{ + struct log_buf *b = cookie; + size_t sz = size; + + if (!b->buf) + return size; + + while (sz) { + size_t space = b->buf_sz - b->head; + size_t n = min(space, sz); + + memcpy(b->buf + b->head, buf, n); + sz -= n; + buf += n; + b->head += n; + if (sz && b->head >= b->buf_sz) { + b->head = 0; + b->wrapped = true; + } + } + return size; +} + +static int log_buf__close(void *cookie) +{ + struct log_buf *b = cookie; + + zfree(&b->buf); + return 0; +} + +static FILE *log_buf__open(struct log_buf *b, FILE *backend, unsigned int sz) +{ + cookie_io_functions_t fns = { + .write = log_buf__write, + .close = log_buf__close, + }; + FILE *file; + + memset(b, 0, sizeof(*b)); + b->buf_sz = sz; + b->buf = malloc(b->buf_sz); + b->backend = backend; + file = fopencookie(b, "a", fns); + if (!file) + zfree(&b->buf); + return file; +} + +static bool remove_first_line(const char **p, size_t *n) +{ + for (; *n && **p != '\n'; ++*p, --*n) + ; + if (*n) { + *p += 1; + *n -= 1; + return true; + } + return false; +} + +static void write_lines(const char *p, size_t n, FILE *fp, bool *remove_first) +{ + if (*remove_first) + *remove_first = !remove_first_line(&p, &n); + fwrite(p, n, 1, fp); +} + +static void log_buf__dump(struct log_buf *b) +{ + bool remove_first = false; + + if (!b->buf) + return; + + fflush(f); /* Could update b->head and b->wrapped */ + fprintf(b->backend, "Dumping debug log buffer\n"); + if (b->wrapped) { + remove_first = true; + write_lines(b->buf + b->head, b->buf_sz - b->head, b->backend, &remove_first); + } + write_lines(b->buf, b->head, b->backend, &remove_first); + fprintf(b->backend, "End of debug log buffer dump\n"); + + b->head = 0; + b->wrapped = false; +} + +void intel_pt_log_dump_buf(void) +{ + log_buf__dump(&log_buf); +} + static int intel_pt_log_open(void) { if (!intel_pt_enable_logging) @@ -86,6 +199,8 @@ static int intel_pt_log_open(void) f = fopen(log_name, "w+"); else f = stdout; + if (f && intel_pt_dump_log_on_error) + f = log_buf__open(&log_buf, f, intel_pt_log_on_error_size); if (!f) { intel_pt_enable_logging = false; return -1; diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-log.h b/tools/perf/util/intel-pt-decoder/intel-pt-log.h index d900aab24b21..354d7d23fc81 100644 --- a/tools/perf/util/intel-pt-decoder/intel-pt-log.h +++ b/tools/perf/util/intel-pt-decoder/intel-pt-log.h @@ -14,9 +14,10 @@ struct intel_pt_pkt; void *intel_pt_log_fp(void); -void intel_pt_log_enable(void); +void intel_pt_log_enable(bool dump_log_on_error, unsigned int log_on_error_size); void intel_pt_log_disable(void); void intel_pt_log_set_name(const char *name); +void intel_pt_log_dump_buf(void); void __intel_pt_log_packet(const struct intel_pt_pkt *packet, int pkt_len, uint64_t pos, const unsigned char *buf); diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c index d5e9fc8106dd..b34cb3dec1aa 100644 --- a/tools/perf/util/intel-pt.c +++ b/tools/perf/util/intel-pt.c @@ -842,7 +842,8 @@ static int intel_pt_walk_next_insn(struct intel_pt_insn *intel_pt_insn, offset, buf, INTEL_PT_INSN_BUF_SZ); if (len <= 0) { - intel_pt_log("ERROR: failed to read at %" PRIu64 " ", offset); + intel_pt_log("ERROR: failed to read at offset %#" PRIx64 " ", + offset); if (intel_pt_enable_logging) dso__fprintf(al.map->dso, intel_pt_log_fp()); return -EINVAL; @@ -2418,6 +2419,8 @@ static int intel_pt_synth_error(struct intel_pt *pt, int code, int cpu, pid_t pid, pid_t tid, u64 ip, u64 timestamp, pid_t machine_pid, int vcpu) { + bool dump_log_on_error = pt->synth_opts.log_plus_flags & AUXTRACE_LOG_FLG_ON_ERROR; + bool log_on_stdout = pt->synth_opts.log_plus_flags & AUXTRACE_LOG_FLG_USE_STDOUT; union perf_event event; char msg[MAX_AUXTRACE_ERROR_MSG]; int err; @@ -2437,6 +2440,16 @@ static int intel_pt_synth_error(struct intel_pt *pt, int code, int cpu, code, cpu, pid, tid, ip, msg, timestamp, machine_pid, vcpu); + if (intel_pt_enable_logging && !log_on_stdout) { + FILE *fp = intel_pt_log_fp(); + + if (fp) + perf_event__fprintf_auxtrace_error(&event, fp); + } + + if (code != INTEL_PT_ERR_LOST && dump_log_on_error) + intel_pt_log_dump_buf(); + err = perf_session__deliver_synth_event(pt->session, &event, NULL); if (err) pr_err("Intel Processor Trace: failed to deliver error event, error %d\n", @@ -4271,8 +4284,12 @@ int intel_pt_process_auxtrace_info(union perf_event *event, goto err_delete_thread; } - if (pt->synth_opts.log) - intel_pt_log_enable(); + if (pt->synth_opts.log) { + bool log_on_error = pt->synth_opts.log_plus_flags & AUXTRACE_LOG_FLG_ON_ERROR; + unsigned int log_on_error_size = pt->synth_opts.log_on_error_size; + + intel_pt_log_enable(log_on_error, log_on_error_size); + } /* Maximum non-turbo ratio is TSC freq / 100 MHz */ if (pt->tc.time_mult) { diff --git a/tools/perf/util/jitdump.c b/tools/perf/util/jitdump.c index 4e6632203704..0e033278fa12 100644 --- a/tools/perf/util/jitdump.c +++ b/tools/perf/util/jitdump.c @@ -56,13 +56,6 @@ struct jit_buf_desc { char dir[PATH_MAX]; }; -struct debug_line_info { - unsigned long vma; - unsigned int lineno; - /* The filename format is unspecified, absolute path, relative etc. */ - char const filename[]; -}; - struct jit_tool { struct perf_tool tool; struct perf_data output; diff --git a/tools/perf/util/lock-contention.h b/tools/perf/util/lock-contention.h index 2146efc33396..b8cb8830b7bc 100644 --- a/tools/perf/util/lock-contention.h +++ b/tools/perf/util/lock-contention.h @@ -11,6 +11,7 @@ struct lock_stat { u64 addr; /* address of lockdep_map, used as ID */ char *name; /* for strcpy(), we cannot use const */ + u64 *callstack; unsigned int nr_acquire; unsigned int nr_acquired; @@ -113,7 +114,9 @@ struct lock_contention { struct machine *machine; struct hlist_head *result; unsigned long map_nr_entries; - unsigned long lost; + int lost; + int max_stack; + int stack_skip; }; #ifdef HAVE_BPF_SKEL diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c index 2a16cae28407..76316e459c3d 100644 --- a/tools/perf/util/machine.c +++ b/tools/perf/util/machine.c @@ -1128,10 +1128,6 @@ static struct dso *machine__get_kernel(struct machine *machine) return kernel; } -struct process_args { - u64 start; -}; - void machine__get_kallsyms_filename(struct machine *machine, char *buf, size_t bufsz) { diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index e0aa4a254583..f3a3d9b3a40d 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -181,7 +181,10 @@ struct map *map__new(struct machine *machine, u64 start, u64 len, if (!(prot & PROT_EXEC)) dso__set_loaded(dso); } + mutex_lock(&dso->lock); + nsinfo__put(dso->nsinfo); dso->nsinfo = nsi; + mutex_unlock(&dso->lock); if (build_id__is_defined(bid)) { dso__set_build_id(dso, bid); diff --git a/tools/perf/util/mem-events.c b/tools/perf/util/mem-events.c index 764883183519..b3a91093069a 100644 --- a/tools/perf/util/mem-events.c +++ b/tools/perf/util/mem-events.c @@ -156,11 +156,12 @@ void perf_mem_events__list(void) for (j = 0; j < PERF_MEM_EVENTS__MAX; j++) { struct perf_mem_event *e = perf_mem_events__ptr(j); - fprintf(stderr, "%-13s%-*s%s\n", - e->tag ?: "", - verbose > 0 ? 25 : 0, - verbose > 0 ? perf_mem_events__name(j, NULL) : "", - e->supported ? ": available" : ""); + fprintf(stderr, "%-*s%-*s%s", + e->tag ? 13 : 0, + e->tag ? : "", + e->tag && verbose > 0 ? 25 : 0, + e->tag && verbose > 0 ? perf_mem_events__name(j, NULL) : "", + e->supported ? ": available\n" : ""); } } @@ -281,7 +282,7 @@ static const char * const mem_lvl[] = { "HIT", "MISS", "L1", - "LFB", + "LFB/MAB", "L2", "L3", "Local RAM", @@ -294,8 +295,10 @@ static const char * const mem_lvl[] = { }; static const char * const mem_lvlnum[] = { + [PERF_MEM_LVLNUM_CXL] = "CXL", + [PERF_MEM_LVLNUM_IO] = "I/O", [PERF_MEM_LVLNUM_ANY_CACHE] = "Any cache", - [PERF_MEM_LVLNUM_LFB] = "LFB", + [PERF_MEM_LVLNUM_LFB] = "LFB/MAB", [PERF_MEM_LVLNUM_RAM] = "RAM", [PERF_MEM_LVLNUM_PMEM] = "PMEM", [PERF_MEM_LVLNUM_NA] = "N/A", diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c index c93bcaf6d55d..4c98ac29ee13 100644 --- a/tools/perf/util/metricgroup.c +++ b/tools/perf/util/metricgroup.c @@ -22,6 +22,7 @@ #include <linux/list_sort.h> #include <linux/string.h> #include <linux/zalloc.h> +#include <perf/cpumap.h> #include <subcmd/parse-options.h> #include <api/fs/fs.h> #include "util.h" @@ -108,17 +109,6 @@ void metricgroup__rblist_exit(struct rblist *metric_events) rblist__exit(metric_events); } -/* - * A node in the list of referenced metrics. metric_expr - * is held as a convenience to avoid a search through the - * metric list. - */ -struct metric_ref_node { - const char *metric_name; - const char *metric_expr; - struct list_head list; -}; - /** * The metric under construction. The data held here will be placed in a * metric_expr. @@ -189,10 +179,24 @@ static bool metricgroup__has_constraint(const struct pmu_event *pe) return false; } +static void metric__free(struct metric *m) +{ + if (!m) + return; + + free(m->metric_refs); + expr__ctx_free(m->pctx); + free((char *)m->modifier); + evlist__delete(m->evlist); + free(m); +} + static struct metric *metric__new(const struct pmu_event *pe, const char *modifier, bool metric_no_group, - int runtime) + int runtime, + const char *user_requested_cpu_list, + bool system_wide) { struct metric *m; @@ -201,35 +205,34 @@ static struct metric *metric__new(const struct pmu_event *pe, return NULL; m->pctx = expr__ctx_new(); - if (!m->pctx) { - free(m); - return NULL; - } + if (!m->pctx) + goto out_err; m->metric_name = pe->metric_name; - m->modifier = modifier ? strdup(modifier) : NULL; - if (modifier && !m->modifier) { - expr__ctx_free(m->pctx); - free(m); - return NULL; + m->modifier = NULL; + if (modifier) { + m->modifier = strdup(modifier); + if (!m->modifier) + goto out_err; } m->metric_expr = pe->metric_expr; m->metric_unit = pe->unit; - m->pctx->runtime = runtime; + m->pctx->sctx.user_requested_cpu_list = NULL; + if (user_requested_cpu_list) { + m->pctx->sctx.user_requested_cpu_list = strdup(user_requested_cpu_list); + if (!m->pctx->sctx.user_requested_cpu_list) + goto out_err; + } + m->pctx->sctx.runtime = runtime; + m->pctx->sctx.system_wide = system_wide; m->has_constraint = metric_no_group || metricgroup__has_constraint(pe); m->metric_refs = NULL; m->evlist = NULL; return m; -} - -static void metric__free(struct metric *m) -{ - free(m->metric_refs); - expr__ctx_free(m->pctx); - free((char *)m->modifier); - evlist__delete(m->evlist); - free(m); +out_err: + metric__free(m); + return NULL; } static bool contains_metric_id(struct evsel **metric_events, int num_events, @@ -874,6 +877,8 @@ struct metricgroup_add_iter_data { int *ret; bool *has_match; bool metric_no_group; + const char *user_requested_cpu_list; + bool system_wide; struct metric *root_metric; const struct visited_metric *visited; const struct pmu_events_table *table; @@ -887,6 +892,8 @@ static int add_metric(struct list_head *metric_list, const struct pmu_event *pe, const char *modifier, bool metric_no_group, + const char *user_requested_cpu_list, + bool system_wide, struct metric *root_metric, const struct visited_metric *visited, const struct pmu_events_table *table); @@ -899,6 +906,8 @@ static int add_metric(struct list_head *metric_list, * @metric_no_group: Should events written to events be grouped "{}" or * global. Grouping is the default but due to multiplexing the * user may override. + * @user_requested_cpu_list: Command line specified CPUs to record on. + * @system_wide: Are events for all processes recorded. * @root_metric: Metrics may reference other metrics to form a tree. In this * case the root_metric holds all the IDs and a list of referenced * metrics. When adding a root this argument is NULL. @@ -910,6 +919,8 @@ static int add_metric(struct list_head *metric_list, static int resolve_metric(struct list_head *metric_list, const char *modifier, bool metric_no_group, + const char *user_requested_cpu_list, + bool system_wide, struct metric *root_metric, const struct visited_metric *visited, const struct pmu_events_table *table) @@ -956,7 +967,8 @@ static int resolve_metric(struct list_head *metric_list, */ for (i = 0; i < pending_cnt; i++) { ret = add_metric(metric_list, &pending[i].pe, modifier, metric_no_group, - root_metric, visited, table); + user_requested_cpu_list, system_wide, root_metric, visited, + table); if (ret) break; } @@ -974,6 +986,8 @@ static int resolve_metric(struct list_head *metric_list, * global. Grouping is the default but due to multiplexing the * user may override. * @runtime: A special argument for the parser only known at runtime. + * @user_requested_cpu_list: Command line specified CPUs to record on. + * @system_wide: Are events for all processes recorded. * @root_metric: Metrics may reference other metrics to form a tree. In this * case the root_metric holds all the IDs and a list of referenced * metrics. When adding a root this argument is NULL. @@ -987,6 +1001,8 @@ static int __add_metric(struct list_head *metric_list, const char *modifier, bool metric_no_group, int runtime, + const char *user_requested_cpu_list, + bool system_wide, struct metric *root_metric, const struct visited_metric *visited, const struct pmu_events_table *table) @@ -1011,7 +1027,8 @@ static int __add_metric(struct list_head *metric_list, * This metric is the root of a tree and may reference other * metrics that are added recursively. */ - root_metric = metric__new(pe, modifier, metric_no_group, runtime); + root_metric = metric__new(pe, modifier, metric_no_group, runtime, + user_requested_cpu_list, system_wide); if (!root_metric) return -ENOMEM; @@ -1060,8 +1077,9 @@ static int __add_metric(struct list_head *metric_list, ret = -EINVAL; } else { /* Resolve referenced metrics. */ - ret = resolve_metric(metric_list, modifier, metric_no_group, root_metric, - &visited_node, table); + ret = resolve_metric(metric_list, modifier, metric_no_group, + user_requested_cpu_list, system_wide, + root_metric, &visited_node, table); } if (ret) { @@ -1109,6 +1127,8 @@ static int add_metric(struct list_head *metric_list, const struct pmu_event *pe, const char *modifier, bool metric_no_group, + const char *user_requested_cpu_list, + bool system_wide, struct metric *root_metric, const struct visited_metric *visited, const struct pmu_events_table *table) @@ -1119,7 +1139,8 @@ static int add_metric(struct list_head *metric_list, if (!strstr(pe->metric_expr, "?")) { ret = __add_metric(metric_list, pe, modifier, metric_no_group, 0, - root_metric, visited, table); + user_requested_cpu_list, system_wide, root_metric, + visited, table); } else { int j, count; @@ -1132,7 +1153,8 @@ static int add_metric(struct list_head *metric_list, for (j = 0; j < count && !ret; j++) ret = __add_metric(metric_list, pe, modifier, metric_no_group, j, - root_metric, visited, table); + user_requested_cpu_list, system_wide, + root_metric, visited, table); } return ret; @@ -1149,6 +1171,7 @@ static int metricgroup__add_metric_sys_event_iter(const struct pmu_event *pe, return 0; ret = add_metric(d->metric_list, pe, d->modifier, d->metric_no_group, + d->user_requested_cpu_list, d->system_wide, d->root_metric, d->visited, d->table); if (ret) goto out; @@ -1191,7 +1214,9 @@ struct metricgroup__add_metric_data { struct list_head *list; const char *metric_name; const char *modifier; + const char *user_requested_cpu_list; bool metric_no_group; + bool system_wide; bool has_match; }; @@ -1208,8 +1233,8 @@ static int metricgroup__add_metric_callback(const struct pmu_event *pe, data->has_match = true; ret = add_metric(data->list, pe, data->modifier, data->metric_no_group, - /*root_metric=*/NULL, - /*visited_metrics=*/NULL, table); + data->user_requested_cpu_list, data->system_wide, + /*root_metric=*/NULL, /*visited_metrics=*/NULL, table); } return ret; } @@ -1223,12 +1248,16 @@ static int metricgroup__add_metric_callback(const struct pmu_event *pe, * @metric_no_group: Should events written to events be grouped "{}" or * global. Grouping is the default but due to multiplexing the * user may override. + * @user_requested_cpu_list: Command line specified CPUs to record on. + * @system_wide: Are events for all processes recorded. * @metric_list: The list that the metric or metric group are added to. * @table: The table that is searched for metrics, most commonly the table for the * architecture perf is running upon. */ static int metricgroup__add_metric(const char *metric_name, const char *modifier, bool metric_no_group, + const char *user_requested_cpu_list, + bool system_wide, struct list_head *metric_list, const struct pmu_events_table *table) { @@ -1242,6 +1271,8 @@ static int metricgroup__add_metric(const char *metric_name, const char *modifier .metric_name = metric_name, .modifier = modifier, .metric_no_group = metric_no_group, + .user_requested_cpu_list = user_requested_cpu_list, + .system_wide = system_wide, .has_match = false, }; /* @@ -1263,6 +1294,8 @@ static int metricgroup__add_metric(const char *metric_name, const char *modifier .metric_name = metric_name, .modifier = modifier, .metric_no_group = metric_no_group, + .user_requested_cpu_list = user_requested_cpu_list, + .system_wide = system_wide, .has_match = &has_match, .ret = &ret, .table = table, @@ -1293,12 +1326,15 @@ out: * @metric_no_group: Should events written to events be grouped "{}" or * global. Grouping is the default but due to multiplexing the * user may override. + * @user_requested_cpu_list: Command line specified CPUs to record on. + * @system_wide: Are events for all processes recorded. * @metric_list: The list that metrics are added to. * @table: The table that is searched for metrics, most commonly the table for the * architecture perf is running upon. */ static int metricgroup__add_metric_list(const char *list, bool metric_no_group, - struct list_head *metric_list, + const char *user_requested_cpu_list, + bool system_wide, struct list_head *metric_list, const struct pmu_events_table *table) { char *list_itr, *list_copy, *metric_name, *modifier; @@ -1315,8 +1351,8 @@ static int metricgroup__add_metric_list(const char *list, bool metric_no_group, *modifier++ = '\0'; ret = metricgroup__add_metric(metric_name, modifier, - metric_no_group, metric_list, - table); + metric_no_group, user_requested_cpu_list, + system_wide, metric_list, table); if (ret == -EINVAL) pr_err("Cannot find metric or group `%s'\n", metric_name); @@ -1505,6 +1541,8 @@ err_out: static int parse_groups(struct evlist *perf_evlist, const char *str, bool metric_no_group, bool metric_no_merge, + const char *user_requested_cpu_list, + bool system_wide, struct perf_pmu *fake_pmu, struct rblist *metric_events_list, const struct pmu_events_table *table) @@ -1518,7 +1556,8 @@ static int parse_groups(struct evlist *perf_evlist, const char *str, if (metric_events_list->nr_entries == 0) metricgroup__rblist_init(metric_events_list); ret = metricgroup__add_metric_list(str, metric_no_group, - &metric_list, table); + user_requested_cpu_list, + system_wide, &metric_list, table); if (ret) goto out; @@ -1626,7 +1665,7 @@ static int parse_groups(struct evlist *perf_evlist, const char *str, } expr->metric_unit = m->metric_unit; expr->metric_events = metric_events; - expr->runtime = m->pctx->runtime; + expr->runtime = m->pctx->sctx.runtime; list_add(&expr->nd, &me->head); } @@ -1646,20 +1685,22 @@ out: return ret; } -int metricgroup__parse_groups(const struct option *opt, +int metricgroup__parse_groups(struct evlist *perf_evlist, const char *str, bool metric_no_group, bool metric_no_merge, + const char *user_requested_cpu_list, + bool system_wide, struct rblist *metric_events) { - struct evlist *perf_evlist = *(struct evlist **)opt->value; const struct pmu_events_table *table = pmu_events_table__find(); if (!table) return -EINVAL; - return parse_groups(perf_evlist, str, metric_no_group, - metric_no_merge, NULL, metric_events, table); + return parse_groups(perf_evlist, str, metric_no_group, metric_no_merge, + user_requested_cpu_list, system_wide, + /*fake_pmu=*/NULL, metric_events, table); } int metricgroup__parse_groups_test(struct evlist *evlist, @@ -1669,8 +1710,10 @@ int metricgroup__parse_groups_test(struct evlist *evlist, bool metric_no_merge, struct rblist *metric_events) { - return parse_groups(evlist, str, metric_no_group, - metric_no_merge, &perf_pmu__fake, metric_events, table); + return parse_groups(evlist, str, metric_no_group, metric_no_merge, + /*user_requested_cpu_list=*/NULL, + /*system_wide=*/false, + &perf_pmu__fake, metric_events, table); } static int metricgroup__has_metric_callback(const struct pmu_event *pe, @@ -1703,7 +1746,7 @@ int metricgroup__copy_metric_events(struct evlist *evlist, struct cgroup *cgrp, struct rblist *new_metric_events, struct rblist *old_metric_events) { - unsigned i; + unsigned int i; for (i = 0; i < rblist__nr_entries(old_metric_events); i++) { struct rb_node *nd; diff --git a/tools/perf/util/metricgroup.h b/tools/perf/util/metricgroup.h index 016b3b1a289a..732d3a0d3334 100644 --- a/tools/perf/util/metricgroup.h +++ b/tools/perf/util/metricgroup.h @@ -64,10 +64,12 @@ struct metric_expr { struct metric_event *metricgroup__lookup(struct rblist *metric_events, struct evsel *evsel, bool create); -int metricgroup__parse_groups(const struct option *opt, +int metricgroup__parse_groups(struct evlist *perf_evlist, const char *str, bool metric_no_group, bool metric_no_merge, + const char *user_requested_cpu_list, + bool system_wide, struct rblist *metric_events); int metricgroup__parse_groups_test(struct evlist *evlist, const struct pmu_events_table *table, diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h index cd8b0777473b..cd4ccec7f361 100644 --- a/tools/perf/util/mmap.h +++ b/tools/perf/util/mmap.h @@ -9,7 +9,6 @@ #include <linux/bitops.h> #include <perf/cpumap.h> #include <stdbool.h> -#include <pthread.h> // for cpu_set_t #ifdef HAVE_AIO_SUPPORT #include <aio.h> #endif diff --git a/tools/perf/util/mutex.c b/tools/perf/util/mutex.c new file mode 100644 index 000000000000..bca7f0717f35 --- /dev/null +++ b/tools/perf/util/mutex.c @@ -0,0 +1,119 @@ +// SPDX-License-Identifier: GPL-2.0 +#include "mutex.h" + +#include "debug.h" +#include <linux/string.h> +#include <errno.h> + +static void check_err(const char *fn, int err) +{ + char sbuf[STRERR_BUFSIZE]; + + if (err == 0) + return; + + pr_err("%s error: '%s'\n", fn, str_error_r(err, sbuf, sizeof(sbuf))); +} + +#define CHECK_ERR(err) check_err(__func__, err) + +static void __mutex_init(struct mutex *mtx, bool pshared) +{ + pthread_mutexattr_t attr; + + CHECK_ERR(pthread_mutexattr_init(&attr)); + +#ifndef NDEBUG + /* In normal builds enable error checking, such as recursive usage. */ + CHECK_ERR(pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_ERRORCHECK)); +#endif + if (pshared) + CHECK_ERR(pthread_mutexattr_setpshared(&attr, PTHREAD_PROCESS_SHARED)); + + CHECK_ERR(pthread_mutex_init(&mtx->lock, &attr)); + CHECK_ERR(pthread_mutexattr_destroy(&attr)); +} + +void mutex_init(struct mutex *mtx) +{ + __mutex_init(mtx, /*pshared=*/false); +} + +void mutex_init_pshared(struct mutex *mtx) +{ + __mutex_init(mtx, /*pshared=*/true); +} + +void mutex_destroy(struct mutex *mtx) +{ + CHECK_ERR(pthread_mutex_destroy(&mtx->lock)); +} + +void mutex_lock(struct mutex *mtx) + NO_THREAD_SAFETY_ANALYSIS +{ + CHECK_ERR(pthread_mutex_lock(&mtx->lock)); +} + +void mutex_unlock(struct mutex *mtx) + NO_THREAD_SAFETY_ANALYSIS +{ + CHECK_ERR(pthread_mutex_unlock(&mtx->lock)); +} + +bool mutex_trylock(struct mutex *mtx) +{ + int ret = pthread_mutex_trylock(&mtx->lock); + + if (ret == 0) + return true; /* Lock acquired. */ + + if (ret == EBUSY) + return false; /* Lock busy. */ + + /* Print error. */ + CHECK_ERR(ret); + return false; +} + +static void __cond_init(struct cond *cnd, bool pshared) +{ + pthread_condattr_t attr; + + CHECK_ERR(pthread_condattr_init(&attr)); + if (pshared) + CHECK_ERR(pthread_condattr_setpshared(&attr, PTHREAD_PROCESS_SHARED)); + + CHECK_ERR(pthread_cond_init(&cnd->cond, &attr)); + CHECK_ERR(pthread_condattr_destroy(&attr)); +} + +void cond_init(struct cond *cnd) +{ + __cond_init(cnd, /*pshared=*/false); +} + +void cond_init_pshared(struct cond *cnd) +{ + __cond_init(cnd, /*pshared=*/true); +} + +void cond_destroy(struct cond *cnd) +{ + CHECK_ERR(pthread_cond_destroy(&cnd->cond)); +} + +void cond_wait(struct cond *cnd, struct mutex *mtx) +{ + CHECK_ERR(pthread_cond_wait(&cnd->cond, &mtx->lock)); +} + +void cond_signal(struct cond *cnd) +{ + CHECK_ERR(pthread_cond_signal(&cnd->cond)); +} + +void cond_broadcast(struct cond *cnd) +{ + CHECK_ERR(pthread_cond_broadcast(&cnd->cond)); +} diff --git a/tools/perf/util/mutex.h b/tools/perf/util/mutex.h new file mode 100644 index 000000000000..40661120cacc --- /dev/null +++ b/tools/perf/util/mutex.h @@ -0,0 +1,108 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef __PERF_MUTEX_H +#define __PERF_MUTEX_H + +#include <pthread.h> +#include <stdbool.h> + +/* + * A function-like feature checking macro that is a wrapper around + * `__has_attribute`, which is defined by GCC 5+ and Clang and evaluates to a + * nonzero constant integer if the attribute is supported or 0 if not. + */ +#ifdef __has_attribute +#define HAVE_ATTRIBUTE(x) __has_attribute(x) +#else +#define HAVE_ATTRIBUTE(x) 0 +#endif + +#if HAVE_ATTRIBUTE(guarded_by) && HAVE_ATTRIBUTE(pt_guarded_by) && \ + HAVE_ATTRIBUTE(lockable) && HAVE_ATTRIBUTE(exclusive_lock_function) && \ + HAVE_ATTRIBUTE(exclusive_trylock_function) && HAVE_ATTRIBUTE(exclusive_locks_required) && \ + HAVE_ATTRIBUTE(no_thread_safety_analysis) + +/* Documents if a shared field or global variable needs to be protected by a mutex. */ +#define GUARDED_BY(x) __attribute__((guarded_by(x))) + +/* + * Documents if the memory location pointed to by a pointer should be guarded by + * a mutex when dereferencing the pointer. + */ +#define PT_GUARDED_BY(x) __attribute__((pt_guarded_by(x))) + +/* Documents if a type is a lockable type. */ +#define LOCKABLE __attribute__((lockable)) + +/* Documents functions that acquire a lock in the body of a function, and do not release it. */ +#define EXCLUSIVE_LOCK_FUNCTION(...) __attribute__((exclusive_lock_function(__VA_ARGS__))) + +/* + * Documents functions that expect a lock to be held on entry to the function, + * and release it in the body of the function. + */ +#define UNLOCK_FUNCTION(...) __attribute__((unlock_function(__VA_ARGS__))) + +/* Documents functions that try to acquire a lock, and return success or failure. */ +#define EXCLUSIVE_TRYLOCK_FUNCTION(...) \ + __attribute__((exclusive_trylock_function(__VA_ARGS__))) + +/* Documents a function that expects a mutex to be held prior to entry. */ +#define EXCLUSIVE_LOCKS_REQUIRED(...) __attribute__((exclusive_locks_required(__VA_ARGS__))) + +/* Turns off thread safety checking within the body of a particular function. */ +#define NO_THREAD_SAFETY_ANALYSIS __attribute__((no_thread_safety_analysis)) + +#else + +#define GUARDED_BY(x) +#define PT_GUARDED_BY(x) +#define LOCKABLE +#define EXCLUSIVE_LOCK_FUNCTION(...) +#define UNLOCK_FUNCTION(...) +#define EXCLUSIVE_TRYLOCK_FUNCTION(...) +#define EXCLUSIVE_LOCKS_REQUIRED(...) +#define NO_THREAD_SAFETY_ANALYSIS + +#endif + +/* + * A wrapper around the mutex implementation that allows perf to error check + * usage, etc. + */ +struct LOCKABLE mutex { + pthread_mutex_t lock; +}; + +/* A wrapper around the condition variable implementation. */ +struct cond { + pthread_cond_t cond; +}; + +/* Default initialize the mtx struct. */ +void mutex_init(struct mutex *mtx); +/* + * Initialize the mtx struct and set the process-shared rather than default + * process-private attribute. + */ +void mutex_init_pshared(struct mutex *mtx); +void mutex_destroy(struct mutex *mtx); + +void mutex_lock(struct mutex *mtx) EXCLUSIVE_LOCK_FUNCTION(*mtx); +void mutex_unlock(struct mutex *mtx) UNLOCK_FUNCTION(*mtx); +/* Tries to acquire the lock and returns true on success. */ +bool mutex_trylock(struct mutex *mtx) EXCLUSIVE_TRYLOCK_FUNCTION(true, *mtx); + +/* Default initialize the cond struct. */ +void cond_init(struct cond *cnd); +/* + * Initialize the cond struct and specify the process-shared rather than default + * process-private attribute. + */ +void cond_init_pshared(struct cond *cnd); +void cond_destroy(struct cond *cnd); + +void cond_wait(struct cond *cnd, struct mutex *mtx) EXCLUSIVE_LOCKS_REQUIRED(mtx); +void cond_signal(struct cond *cnd); +void cond_broadcast(struct cond *cnd); + +#endif /* __PERF_MUTEX_H */ diff --git a/tools/perf/util/parse-branch-options.c b/tools/perf/util/parse-branch-options.c index bb4aa88c50a8..00588b9db474 100644 --- a/tools/perf/util/parse-branch-options.c +++ b/tools/perf/util/parse-branch-options.c @@ -32,6 +32,7 @@ static const struct branch_mode branch_modes[] = { BRANCH_OPT("call", PERF_SAMPLE_BRANCH_CALL), BRANCH_OPT("save_type", PERF_SAMPLE_BRANCH_TYPE_SAVE), BRANCH_OPT("stack", PERF_SAMPLE_BRANCH_CALL_STACK), + BRANCH_OPT("priv", PERF_SAMPLE_BRANCH_PRIV_SAVE), BRANCH_END }; diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index f3b2c2a87456..437389dacf48 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -150,14 +150,6 @@ struct event_symbol event_symbols_sw[PERF_COUNT_SW_MAX] = { }, }; -#define __PERF_EVENT_FIELD(config, name) \ - ((config & PERF_EVENT_##name##_MASK) >> PERF_EVENT_##name##_SHIFT) - -#define PERF_EVENT_RAW(config) __PERF_EVENT_FIELD(config, RAW) -#define PERF_EVENT_CONFIG(config) __PERF_EVENT_FIELD(config, CONFIG) -#define PERF_EVENT_TYPE(config) __PERF_EVENT_FIELD(config, TYPE) -#define PERF_EVENT_ID(config) __PERF_EVENT_FIELD(config, EVENT) - bool is_event_supported(u8 type, u64 config) { bool ret = true; diff --git a/tools/perf/util/perf_event_attr_fprintf.c b/tools/perf/util/perf_event_attr_fprintf.c index 98af3fa4ea35..7e5e7b30510d 100644 --- a/tools/perf/util/perf_event_attr_fprintf.c +++ b/tools/perf/util/perf_event_attr_fprintf.c @@ -52,7 +52,7 @@ static void __p_branch_sample_type(char *buf, size_t size, u64 value) bit_name(ABORT_TX), bit_name(IN_TX), bit_name(NO_TX), bit_name(COND), bit_name(CALL_STACK), bit_name(IND_JUMP), bit_name(CALL), bit_name(NO_FLAGS), bit_name(NO_CYCLES), - bit_name(TYPE_SAVE), bit_name(HW_INDEX), + bit_name(TYPE_SAVE), bit_name(HW_INDEX), bit_name(PRIV_SAVE), { .name = NULL, } }; #undef bit_name @@ -64,7 +64,7 @@ static void __p_read_format(char *buf, size_t size, u64 value) #define bit_name(n) { PERF_FORMAT_##n, #n } struct bit_names bits[] = { bit_name(TOTAL_TIME_ENABLED), bit_name(TOTAL_TIME_RUNNING), - bit_name(ID), bit_name(GROUP), + bit_name(ID), bit_name(GROUP), bit_name(LOST), { .name = NULL, } }; #undef bit_name diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c index 89655d53117a..74a2cafb4e8d 100644 --- a/tools/perf/util/pmu.c +++ b/tools/perf/util/pmu.c @@ -1182,7 +1182,7 @@ static char *pmu_formats_string(struct list_head *formats) struct perf_pmu_format *format; char *str = NULL; struct strbuf buf = STRBUF_INIT; - unsigned i = 0; + unsigned int i = 0; if (!formats) return NULL; diff --git a/tools/perf/util/pmu.y b/tools/perf/util/pmu.y index bfd7e8509869..0dab0ec2eff7 100644 --- a/tools/perf/util/pmu.y +++ b/tools/perf/util/pmu.y @@ -10,8 +10,6 @@ #include <string.h> #include "pmu.h" -extern int perf_pmu_lex (void); - #define ABORT_ON(val) \ do { \ if (val) \ diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c index 785246ff4179..0c24bc7afbca 100644 --- a/tools/perf/util/probe-event.c +++ b/tools/perf/util/probe-event.c @@ -29,6 +29,7 @@ #include "color.h" #include "map.h" #include "maps.h" +#include "mutex.h" #include "symbol.h" #include <api/fs/fs.h> #include "trace-event.h" /* For __maybe_unused */ @@ -180,8 +181,10 @@ struct map *get_target_map(const char *target, struct nsinfo *nsi, bool user) map = dso__new_map(target); if (map && map->dso) { + mutex_lock(&map->dso->lock); nsinfo__put(map->dso->nsinfo); map->dso->nsinfo = nsinfo__get(nsi); + mutex_unlock(&map->dso->lock); } return map; } else { diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index 192c9274f7ad..1a4f10de29ff 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -943,6 +943,11 @@ static void perf_event__cpu_map_swap(union perf_event *event, default: pr_err("cpu_map swap: unsupported long size\n"); } + break; + case PERF_CPU_MAP__RANGE_CPUS: + data->range_cpu_data.start_cpu = bswap_16(data->range_cpu_data.start_cpu); + data->range_cpu_data.end_cpu = bswap_16(data->range_cpu_data.end_cpu); + break; default: break; } @@ -1180,7 +1185,7 @@ static void branch_stack__printf(struct perf_sample *sample, bool callstack) e->flags.abort ? "A" : " ", e->flags.in_tx ? "T" : " ", (unsigned)e->flags.reserved, - e->flags.type ? branch_type_name(e->flags.type) : ""); + get_branch_type(e)); } else { if (i == 0) { printf("..... %2"PRIu64": %016" PRIx64 "\n" diff --git a/tools/perf/util/smt.c b/tools/perf/util/smt.c index 2b0a36ebf27a..994e9e418227 100644 --- a/tools/perf/util/smt.c +++ b/tools/perf/util/smt.c @@ -1,99 +1,37 @@ -#include <stdio.h> -#include <stdlib.h> -#include <unistd.h> -#include <linux/bitops.h> +// SPDX-License-Identifier: GPL-2.0-only +#include <string.h> #include "api/fs/fs.h" +#include "cputopo.h" #include "smt.h" -/** - * hweight_str - Returns the number of bits set in str. Stops at first non-hex - * or ',' character. - */ -static int hweight_str(char *str) -{ - int result = 0; - - while (*str) { - switch (*str++) { - case '0': - case ',': - break; - case '1': - case '2': - case '4': - case '8': - result++; - break; - case '3': - case '5': - case '6': - case '9': - case 'a': - case 'A': - case 'c': - case 'C': - result += 2; - break; - case '7': - case 'b': - case 'B': - case 'd': - case 'D': - case 'e': - case 'E': - result += 3; - break; - case 'f': - case 'F': - result += 4; - break; - default: - goto done; - } - } -done: - return result; -} - -int smt_on(void) +bool smt_on(const struct cpu_topology *topology) { static bool cached; - static int cached_result; - int cpu; - int ncpu; + static bool cached_result; + int fs_value; if (cached) return cached_result; - if (sysfs__read_int("devices/system/cpu/smt/active", &cached_result) >= 0) { - cached = true; - return cached_result; - } - - cached_result = 0; - ncpu = sysconf(_SC_NPROCESSORS_CONF); - for (cpu = 0; cpu < ncpu; cpu++) { - unsigned long long siblings; - char *str; - size_t strlen; - char fn[256]; + if (sysfs__read_int("devices/system/cpu/smt/active", &fs_value) >= 0) + cached_result = (fs_value == 1); + else + cached_result = cpu_topology__smt_on(topology); - snprintf(fn, sizeof fn, - "devices/system/cpu/cpu%d/topology/thread_siblings", cpu); - if (sysfs__read_str(fn, &str, &strlen) < 0) { - snprintf(fn, sizeof fn, - "devices/system/cpu/cpu%d/topology/core_cpus", cpu); - if (sysfs__read_str(fn, &str, &strlen) < 0) - continue; - } - /* Entry is hex, but does not have 0x, so need custom parser */ - siblings = hweight_str(str); - free(str); - if (siblings > 1) { - cached_result = 1; - break; - } - } cached = true; return cached_result; } + +bool core_wide(bool system_wide, const char *user_requested_cpu_list, + const struct cpu_topology *topology) +{ + /* If not everything running on a core is being recorded then we can't use core_wide. */ + if (!system_wide) + return false; + + /* Cheap case that SMT is disabled and therefore we're inherently core_wide. */ + if (!smt_on(topology)) + return true; + + return cpu_topology__core_wide(topology, user_requested_cpu_list); +} diff --git a/tools/perf/util/smt.h b/tools/perf/util/smt.h index b8414b7bcbc8..ae9095f2c38c 100644 --- a/tools/perf/util/smt.h +++ b/tools/perf/util/smt.h @@ -1,6 +1,17 @@ -#ifndef SMT_H -#define SMT_H 1 +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef __SMT_H +#define __SMT_H 1 -int smt_on(void); +struct cpu_topology; -#endif +/* Returns true if SMT (aka hyperthreading) is enabled. */ +bool smt_on(const struct cpu_topology *topology); + +/* + * Returns true when system wide and all SMT threads for a core are in the + * user_requested_cpus map. + */ +bool core_wide(bool system_wide, const char *user_requested_cpu_list, + const struct cpu_topology *topology); + +#endif /* __SMT_H */ diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c index 6d5588e80935..2e7330867e2e 100644 --- a/tools/perf/util/sort.c +++ b/tools/perf/util/sort.c @@ -1948,6 +1948,43 @@ struct sort_entry sort_dso_size = { .se_width_idx = HISTC_DSO_SIZE, }; +/* --sort dso_size */ + +static int64_t +sort__addr_cmp(struct hist_entry *left, struct hist_entry *right) +{ + u64 left_ip = left->ip; + u64 right_ip = right->ip; + struct map *left_map = left->ms.map; + struct map *right_map = right->ms.map; + + if (left_map) + left_ip = left_map->unmap_ip(left_map, left_ip); + if (right_map) + right_ip = right_map->unmap_ip(right_map, right_ip); + + return _sort__addr_cmp(left_ip, right_ip); +} + +static int hist_entry__addr_snprintf(struct hist_entry *he, char *bf, + size_t size, unsigned int width) +{ + u64 ip = he->ip; + struct map *map = he->ms.map; + + if (map) + ip = map->unmap_ip(map, ip); + + return repsep_snprintf(bf, size, "%-#*llx", width, ip); +} + +struct sort_entry sort_addr = { + .se_header = "Address", + .se_cmp = sort__addr_cmp, + .se_snprintf = hist_entry__addr_snprintf, + .se_width_idx = HISTC_ADDR, +}; + struct sort_dimension { const char *name; @@ -1997,6 +2034,7 @@ static struct sort_dimension common_sort_dimensions[] = { DIM(SORT_GLOBAL_INS_LAT, "ins_lat", sort_global_ins_lat), DIM(SORT_LOCAL_PIPELINE_STAGE_CYC, "local_p_stage_cyc", sort_local_p_stage_cyc), DIM(SORT_GLOBAL_PIPELINE_STAGE_CYC, "p_stage_cyc", sort_global_p_stage_cyc), + DIM(SORT_ADDR, "addr", sort_addr), }; #undef DIM diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h index 2ddc00d1c464..04ff8b61a2a7 100644 --- a/tools/perf/util/sort.h +++ b/tools/perf/util/sort.h @@ -34,7 +34,6 @@ extern struct sort_entry sort_dso_to; extern struct sort_entry sort_sym_from; extern struct sort_entry sort_sym_to; extern struct sort_entry sort_srcline; -extern enum sort_type sort__first_dimension; extern const char default_mem_sort_order[]; struct res_sample { @@ -237,6 +236,7 @@ enum sort_type { SORT_GLOBAL_INS_LAT, SORT_LOCAL_PIPELINE_STAGE_CYC, SORT_GLOBAL_PIPELINE_STAGE_CYC, + SORT_ADDR, /* branch stack specific sort keys */ __SORT_BRANCH_STACK, @@ -295,7 +295,6 @@ struct block_hist { }; extern struct sort_entry sort_thread; -extern struct list_head hist_entry__sort_list; struct evlist; struct tep_handle; diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c index b82844cb0ce7..5c47ee9963a7 100644 --- a/tools/perf/util/stat-display.c +++ b/tools/perf/util/stat-display.c @@ -67,7 +67,7 @@ static void print_noise(struct perf_stat_config *config, return; ps = evsel->stats; - print_noise_pct(config, stddev_stats(&ps->res_stats[0]), avg); + print_noise_pct(config, stddev_stats(&ps->res_stats), avg); } static void print_cgroup(struct perf_stat_config *config, struct evsel *evsel) @@ -168,7 +168,7 @@ static void aggr_printout(struct perf_stat_config *config, id.socket, id.die, id.core); - } else if (id.core > -1) { + } else if (id.cpu.cpu > -1) { fprintf(config->output, "\"cpu\" : \"%d\", ", id.cpu.cpu); } @@ -179,7 +179,7 @@ static void aggr_printout(struct perf_stat_config *config, id.die, config->csv_output ? 0 : -3, id.core, config->csv_sep); - } else if (id.core > -1) { + } else if (id.cpu.cpu > -1) { fprintf(config->output, "CPU%*d%s", config->csv_output ? 0 : -7, id.cpu.cpu, config->csv_sep); @@ -189,14 +189,14 @@ static void aggr_printout(struct perf_stat_config *config, case AGGR_THREAD: if (config->json_output) { fprintf(config->output, "\"thread\" : \"%s-%d\", ", - perf_thread_map__comm(evsel->core.threads, id.thread), - perf_thread_map__pid(evsel->core.threads, id.thread)); + perf_thread_map__comm(evsel->core.threads, id.thread_idx), + perf_thread_map__pid(evsel->core.threads, id.thread_idx)); } else { fprintf(config->output, "%*s-%*d%s", config->csv_output ? 0 : 16, - perf_thread_map__comm(evsel->core.threads, id.thread), + perf_thread_map__comm(evsel->core.threads, id.thread_idx), config->csv_output ? 0 : -8, - perf_thread_map__pid(evsel->core.threads, id.thread), + perf_thread_map__pid(evsel->core.threads, id.thread_idx), config->csv_sep); } break; @@ -442,7 +442,7 @@ static void print_metric_header(struct perf_stat_config *config, fprintf(os->fh, "%*s ", config->metric_only_len, unit); } -static int first_shadow_cpu_map_idx(struct perf_stat_config *config, +static int first_shadow_map_idx(struct perf_stat_config *config, struct evsel *evsel, const struct aggr_cpu_id *id) { struct perf_cpu_map *cpus = evsel__cpus(evsel); @@ -452,6 +452,9 @@ static int first_shadow_cpu_map_idx(struct perf_stat_config *config, if (config->aggr_mode == AGGR_NONE) return perf_cpu_map__idx(cpus, id->cpu); + if (config->aggr_mode == AGGR_THREAD) + return id->thread_idx; + if (!config->aggr_get_id) return 0; @@ -646,7 +649,7 @@ static void printout(struct perf_stat_config *config, struct aggr_cpu_id id, int } perf_stat__print_shadow_stats(config, counter, uval, - first_shadow_cpu_map_idx(config, counter, &id), + first_shadow_map_idx(config, counter, &id), &out, &config->metric_events, st); if (!config->csv_output && !config->metric_only && !config->json_output) { print_noise(config, counter, noise); @@ -676,7 +679,7 @@ static void aggr_update_shadow(struct perf_stat_config *config, val += perf_counts(counter->counts, idx, 0)->val; } perf_stat__update_shadow_stats(counter, val, - first_shadow_cpu_map_idx(config, counter, &id), + first_shadow_map_idx(config, counter, &id), &rt_stat); } } @@ -943,7 +946,7 @@ static struct perf_aggr_thread_value *sort_aggr_thread( buf[i].counter = counter; buf[i].id = aggr_cpu_id__empty(); - buf[i].id.thread = thread; + buf[i].id.thread_idx = thread; buf[i].uval = uval; buf[i].val = val; buf[i].run = run; @@ -979,14 +982,9 @@ static void print_aggr_thread(struct perf_stat_config *config, fprintf(output, "%s", prefix); id = buf[thread].id; - if (config->stats) - printout(config, id, 0, buf[thread].counter, buf[thread].uval, - prefix, buf[thread].run, buf[thread].ena, 1.0, - &config->stats[id.thread]); - else - printout(config, id, 0, buf[thread].counter, buf[thread].uval, - prefix, buf[thread].run, buf[thread].ena, 1.0, - &rt_stat); + printout(config, id, 0, buf[thread].counter, buf[thread].uval, + prefix, buf[thread].run, buf[thread].ena, 1.0, + &rt_stat); fputc('\n', output); } diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c index 788ce5e46470..07b29fe272c7 100644 --- a/tools/perf/util/stat-shadow.c +++ b/tools/perf/util/stat-shadow.c @@ -33,9 +33,8 @@ struct saved_value { struct evsel *evsel; enum stat_type type; int ctx; - int cpu_map_idx; + int map_idx; /* cpu or thread map index */ struct cgroup *cgrp; - struct runtime_stat *stat; struct stats stats; u64 metric_total; int metric_other; @@ -48,8 +47,8 @@ static int saved_value_cmp(struct rb_node *rb_node, const void *entry) rb_node); const struct saved_value *b = entry; - if (a->cpu_map_idx != b->cpu_map_idx) - return a->cpu_map_idx - b->cpu_map_idx; + if (a->map_idx != b->map_idx) + return a->map_idx - b->map_idx; /* * Previously the rbtree was used to link generic metrics. @@ -67,16 +66,6 @@ static int saved_value_cmp(struct rb_node *rb_node, const void *entry) if (a->cgrp != b->cgrp) return (char *)a->cgrp < (char *)b->cgrp ? -1 : +1; - if (a->evsel == NULL && b->evsel == NULL) { - if (a->stat == b->stat) - return 0; - - if ((char *)a->stat < (char *)b->stat) - return -1; - - return 1; - } - if (a->evsel == b->evsel) return 0; if ((char *)a->evsel < (char *)b->evsel) @@ -106,7 +95,7 @@ static void saved_value_delete(struct rblist *rblist __maybe_unused, } static struct saved_value *saved_value_lookup(struct evsel *evsel, - int cpu_map_idx, + int map_idx, bool create, enum stat_type type, int ctx, @@ -116,11 +105,10 @@ static struct saved_value *saved_value_lookup(struct evsel *evsel, struct rblist *rblist; struct rb_node *nd; struct saved_value dm = { - .cpu_map_idx = cpu_map_idx, + .map_idx = map_idx, .evsel = evsel, .type = type, .ctx = ctx, - .stat = st, .cgrp = cgrp, }; @@ -215,10 +203,10 @@ struct runtime_stat_data { static void update_runtime_stat(struct runtime_stat *st, enum stat_type type, - int cpu_map_idx, u64 count, + int map_idx, u64 count, struct runtime_stat_data *rsd) { - struct saved_value *v = saved_value_lookup(NULL, cpu_map_idx, true, type, + struct saved_value *v = saved_value_lookup(NULL, map_idx, true, type, rsd->ctx, st, rsd->cgrp); if (v) @@ -231,7 +219,7 @@ static void update_runtime_stat(struct runtime_stat *st, * instruction rates, etc: */ void perf_stat__update_shadow_stats(struct evsel *counter, u64 count, - int cpu_map_idx, struct runtime_stat *st) + int map_idx, struct runtime_stat *st) { u64 count_ns = count; struct saved_value *v; @@ -243,88 +231,88 @@ void perf_stat__update_shadow_stats(struct evsel *counter, u64 count, count *= counter->scale; if (evsel__is_clock(counter)) - update_runtime_stat(st, STAT_NSECS, cpu_map_idx, count_ns, &rsd); + update_runtime_stat(st, STAT_NSECS, map_idx, count_ns, &rsd); else if (evsel__match(counter, HARDWARE, HW_CPU_CYCLES)) - update_runtime_stat(st, STAT_CYCLES, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_CYCLES, map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, CYCLES_IN_TX)) - update_runtime_stat(st, STAT_CYCLES_IN_TX, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_CYCLES_IN_TX, map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TRANSACTION_START)) - update_runtime_stat(st, STAT_TRANSACTION, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_TRANSACTION, map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, ELISION_START)) - update_runtime_stat(st, STAT_ELISION, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_ELISION, map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_TOTAL_SLOTS)) update_runtime_stat(st, STAT_TOPDOWN_TOTAL_SLOTS, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_SLOTS_ISSUED)) update_runtime_stat(st, STAT_TOPDOWN_SLOTS_ISSUED, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_SLOTS_RETIRED)) update_runtime_stat(st, STAT_TOPDOWN_SLOTS_RETIRED, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_FETCH_BUBBLES)) update_runtime_stat(st, STAT_TOPDOWN_FETCH_BUBBLES, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_RECOVERY_BUBBLES)) update_runtime_stat(st, STAT_TOPDOWN_RECOVERY_BUBBLES, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_RETIRING)) update_runtime_stat(st, STAT_TOPDOWN_RETIRING, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_BAD_SPEC)) update_runtime_stat(st, STAT_TOPDOWN_BAD_SPEC, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_FE_BOUND)) update_runtime_stat(st, STAT_TOPDOWN_FE_BOUND, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_BE_BOUND)) update_runtime_stat(st, STAT_TOPDOWN_BE_BOUND, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_HEAVY_OPS)) update_runtime_stat(st, STAT_TOPDOWN_HEAVY_OPS, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_BR_MISPREDICT)) update_runtime_stat(st, STAT_TOPDOWN_BR_MISPREDICT, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_FETCH_LAT)) update_runtime_stat(st, STAT_TOPDOWN_FETCH_LAT, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, TOPDOWN_MEM_BOUND)) update_runtime_stat(st, STAT_TOPDOWN_MEM_BOUND, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) update_runtime_stat(st, STAT_STALLED_CYCLES_FRONT, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND)) update_runtime_stat(st, STAT_STALLED_CYCLES_BACK, - cpu_map_idx, count, &rsd); + map_idx, count, &rsd); else if (evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS)) - update_runtime_stat(st, STAT_BRANCHES, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_BRANCHES, map_idx, count, &rsd); else if (evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES)) - update_runtime_stat(st, STAT_CACHEREFS, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_CACHEREFS, map_idx, count, &rsd); else if (evsel__match(counter, HW_CACHE, HW_CACHE_L1D)) - update_runtime_stat(st, STAT_L1_DCACHE, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_L1_DCACHE, map_idx, count, &rsd); else if (evsel__match(counter, HW_CACHE, HW_CACHE_L1I)) - update_runtime_stat(st, STAT_L1_ICACHE, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_L1_ICACHE, map_idx, count, &rsd); else if (evsel__match(counter, HW_CACHE, HW_CACHE_LL)) - update_runtime_stat(st, STAT_LL_CACHE, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_LL_CACHE, map_idx, count, &rsd); else if (evsel__match(counter, HW_CACHE, HW_CACHE_DTLB)) - update_runtime_stat(st, STAT_DTLB_CACHE, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_DTLB_CACHE, map_idx, count, &rsd); else if (evsel__match(counter, HW_CACHE, HW_CACHE_ITLB)) - update_runtime_stat(st, STAT_ITLB_CACHE, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_ITLB_CACHE, map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, SMI_NUM)) - update_runtime_stat(st, STAT_SMI_NUM, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_SMI_NUM, map_idx, count, &rsd); else if (perf_stat_evsel__is(counter, APERF)) - update_runtime_stat(st, STAT_APERF, cpu_map_idx, count, &rsd); + update_runtime_stat(st, STAT_APERF, map_idx, count, &rsd); if (counter->collect_stat) { - v = saved_value_lookup(counter, cpu_map_idx, true, STAT_NONE, 0, st, + v = saved_value_lookup(counter, map_idx, true, STAT_NONE, 0, st, rsd.cgrp); update_stats(&v->stats, count); if (counter->metric_leader) v->metric_total += count; } else if (counter->metric_leader) { v = saved_value_lookup(counter->metric_leader, - cpu_map_idx, true, STAT_NONE, 0, st, rsd.cgrp); + map_idx, true, STAT_NONE, 0, st, rsd.cgrp); v->metric_total += count; v->metric_other++; } @@ -466,12 +454,12 @@ void perf_stat__collect_metric_expr(struct evlist *evsel_list) } static double runtime_stat_avg(struct runtime_stat *st, - enum stat_type type, int cpu_map_idx, + enum stat_type type, int map_idx, struct runtime_stat_data *rsd) { struct saved_value *v; - v = saved_value_lookup(NULL, cpu_map_idx, false, type, rsd->ctx, st, rsd->cgrp); + v = saved_value_lookup(NULL, map_idx, false, type, rsd->ctx, st, rsd->cgrp); if (!v) return 0.0; @@ -479,12 +467,12 @@ static double runtime_stat_avg(struct runtime_stat *st, } static double runtime_stat_n(struct runtime_stat *st, - enum stat_type type, int cpu_map_idx, + enum stat_type type, int map_idx, struct runtime_stat_data *rsd) { struct saved_value *v; - v = saved_value_lookup(NULL, cpu_map_idx, false, type, rsd->ctx, st, rsd->cgrp); + v = saved_value_lookup(NULL, map_idx, false, type, rsd->ctx, st, rsd->cgrp); if (!v) return 0.0; @@ -492,7 +480,7 @@ static double runtime_stat_n(struct runtime_stat *st, } static void print_stalled_cycles_frontend(struct perf_stat_config *config, - int cpu_map_idx, double avg, + int map_idx, double avg, struct perf_stat_output_ctx *out, struct runtime_stat *st, struct runtime_stat_data *rsd) @@ -500,7 +488,7 @@ static void print_stalled_cycles_frontend(struct perf_stat_config *config, double total, ratio = 0.0; const char *color; - total = runtime_stat_avg(st, STAT_CYCLES, cpu_map_idx, rsd); + total = runtime_stat_avg(st, STAT_CYCLES, map_idx, rsd); if (total) ratio = avg / total * 100.0; @@ -515,7 +503,7 @@ static void print_stalled_cycles_frontend(struct perf_stat_config *config, } static void print_stalled_cycles_backend(struct perf_stat_config *config, - int cpu_map_idx, double avg, + int map_idx, double avg, struct perf_stat_output_ctx *out, struct runtime_stat *st, struct runtime_stat_data *rsd) @@ -523,7 +511,7 @@ static void print_stalled_cycles_backend(struct perf_stat_config *config, double total, ratio = 0.0; const char *color; - total = runtime_stat_avg(st, STAT_CYCLES, cpu_map_idx, rsd); + total = runtime_stat_avg(st, STAT_CYCLES, map_idx, rsd); if (total) ratio = avg / total * 100.0; @@ -534,7 +522,7 @@ static void print_stalled_cycles_backend(struct perf_stat_config *config, } static void print_branch_misses(struct perf_stat_config *config, - int cpu_map_idx, double avg, + int map_idx, double avg, struct perf_stat_output_ctx *out, struct runtime_stat *st, struct runtime_stat_data *rsd) @@ -542,7 +530,7 @@ static void print_branch_misses(struct perf_stat_config *config, double total, ratio = 0.0; const char *color; - total = runtime_stat_avg(st, STAT_BRANCHES, cpu_map_idx, rsd); + total = runtime_stat_avg(st, STAT_BRANCHES, map_idx, rsd); if (total) ratio = avg / total * 100.0; @@ -553,7 +541,7 @@ static void print_branch_misses(struct perf_stat_config *config, } static void print_l1_dcache_misses(struct perf_stat_config *config, - int cpu_map_idx, double avg, + int map_idx, double avg, struct perf_stat_output_ctx *out, struct runtime_stat *st, struct runtime_stat_data *rsd) @@ -561,7 +549,7 @@ static void print_l1_dcache_misses(struct perf_stat_config *config, double total, ratio = 0.0; const char *color; - total = runtime_stat_avg(st, STAT_L1_DCACHE, cpu_map_idx, rsd); + total = runtime_stat_avg(st, STAT_L1_DCACHE, map_idx, rsd); if (total) ratio = avg / total * 100.0; @@ -572,7 +560,7 @@ static void print_l1_dcache_misses(struct perf_stat_config *config, } static void print_l1_icache_misses(struct perf_stat_config *config, - int cpu_map_idx, double avg, + int map_idx, double avg, struct perf_stat_output_ctx *out, struct runtime_stat *st, struct runtime_stat_data *rsd) @@ -580,7 +568,7 @@ static void print_l1_icache_misses(struct perf_stat_config *config, double total, ratio = 0.0; const char *color; - total = runtime_stat_avg(st, STAT_L1_ICACHE, cpu_map_idx, rsd); + total = runtime_stat_avg(st, STAT_L1_ICACHE, map_idx, rsd); if (total) ratio = avg / total * 100.0; @@ -590,7 +578,7 @@ static void print_l1_icache_misses(struct perf_stat_config *config, } static void print_dtlb_cache_misses(struct perf_stat_config *config, - int cpu_map_idx, double avg, + int map_idx, double avg, struct perf_stat_output_ctx *out, struct runtime_stat *st, struct runtime_stat_data *rsd) @@ -598,7 +586,7 @@ static void print_dtlb_cache_misses(struct perf_stat_config *config, double total, ratio = 0.0; const char *color; - total = runtime_stat_avg(st, STAT_DTLB_CACHE, cpu_map_idx, rsd); + total = runtime_stat_avg(st, STAT_DTLB_CACHE, map_idx, rsd); if (total) ratio = avg / total * 100.0; @@ -608,7 +596,7 @@ static void print_dtlb_cache_misses(struct perf_stat_config *config, } static void print_itlb_cache_misses(struct perf_stat_config *config, - int cpu_map_idx, double avg, + int map_idx, double avg, struct perf_stat_output_ctx *out, struct runtime_stat *st, struct runtime_stat_data *rsd) @@ -616,7 +604,7 @@ static void print_itlb_cache_misses(struct perf_stat_config *config, double total, ratio = 0.0; const char *color; - total = runtime_stat_avg(st, STAT_ITLB_CACHE, cpu_map_idx, rsd); + total = runtime_stat_avg(st, STAT_ITLB_CACHE, map_idx, rsd); if (total) ratio = avg / total * 100.0; @@ -626,7 +614,7 @@ static void print_itlb_cache_misses(struct perf_stat_config *config, } static void print_ll_cache_misses(struct perf_stat_config *config, - int cpu_map_idx, double avg, + int map_idx, double avg, struct perf_stat_output_ctx *out, struct runtime_stat *st, struct runtime_stat_data *rsd) @@ -634,7 +622,7 @@ static void print_ll_cache_misses(struct perf_stat_config *config, double total, ratio = 0.0; const char *color; - total = runtime_stat_avg(st, STAT_LL_CACHE, cpu_map_idx, rsd); + total = runtime_stat_avg(st, STAT_LL_CACHE, map_idx, rsd); if (total) ratio = avg / total * 100.0; @@ -692,61 +680,61 @@ static double sanitize_val(double x) return x; } -static double td_total_slots(int cpu_map_idx, struct runtime_stat *st, +static double td_total_slots(int map_idx, struct runtime_stat *st, struct runtime_stat_data *rsd) { - return runtime_stat_avg(st, STAT_TOPDOWN_TOTAL_SLOTS, cpu_map_idx, rsd); + return runtime_stat_avg(st, STAT_TOPDOWN_TOTAL_SLOTS, map_idx, rsd); } -static double td_bad_spec(int cpu_map_idx, struct runtime_stat *st, +static double td_bad_spec(int map_idx, struct runtime_stat *st, struct runtime_stat_data *rsd) { double bad_spec = 0; double total_slots; double total; - total = runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_ISSUED, cpu_map_idx, rsd) - - runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_RETIRED, cpu_map_idx, rsd) + - runtime_stat_avg(st, STAT_TOPDOWN_RECOVERY_BUBBLES, cpu_map_idx, rsd); + total = runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_ISSUED, map_idx, rsd) - + runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_RETIRED, map_idx, rsd) + + runtime_stat_avg(st, STAT_TOPDOWN_RECOVERY_BUBBLES, map_idx, rsd); - total_slots = td_total_slots(cpu_map_idx, st, rsd); + total_slots = td_total_slots(map_idx, st, rsd); if (total_slots) bad_spec = total / total_slots; return sanitize_val(bad_spec); } -static double td_retiring(int cpu_map_idx, struct runtime_stat *st, +static double td_retiring(int map_idx, struct runtime_stat *st, struct runtime_stat_data *rsd) { double retiring = 0; - double total_slots = td_total_slots(cpu_map_idx, st, rsd); + double total_slots = td_total_slots(map_idx, st, rsd); double ret_slots = runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_RETIRED, - cpu_map_idx, rsd); + map_idx, rsd); if (total_slots) retiring = ret_slots / total_slots; return retiring; } -static double td_fe_bound(int cpu_map_idx, struct runtime_stat *st, +static double td_fe_bound(int map_idx, struct runtime_stat *st, struct runtime_stat_data *rsd) { double fe_bound = 0; - double total_slots = td_total_slots(cpu_map_idx, st, rsd); + double total_slots = td_total_slots(map_idx, st, rsd); double fetch_bub = runtime_stat_avg(st, STAT_TOPDOWN_FETCH_BUBBLES, - cpu_map_idx, rsd); + map_idx, rsd); if (total_slots) fe_bound = fetch_bub / total_slots; return fe_bound; } -static double td_be_bound(int cpu_map_idx, struct runtime_stat *st, +static double td_be_bound(int map_idx, struct runtime_stat *st, struct runtime_stat_data *rsd) { - double sum = (td_fe_bound(cpu_map_idx, st, rsd) + - td_bad_spec(cpu_map_idx, st, rsd) + - td_retiring(cpu_map_idx, st, rsd)); + double sum = (td_fe_bound(map_idx, st, rsd) + + td_bad_spec(map_idx, st, rsd) + + td_retiring(map_idx, st, rsd)); if (sum == 0) return 0; return sanitize_val(1.0 - sum); @@ -757,15 +745,15 @@ static double td_be_bound(int cpu_map_idx, struct runtime_stat *st, * the ratios we need to recreate the sum. */ -static double td_metric_ratio(int cpu_map_idx, enum stat_type type, +static double td_metric_ratio(int map_idx, enum stat_type type, struct runtime_stat *stat, struct runtime_stat_data *rsd) { - double sum = runtime_stat_avg(stat, STAT_TOPDOWN_RETIRING, cpu_map_idx, rsd) + - runtime_stat_avg(stat, STAT_TOPDOWN_FE_BOUND, cpu_map_idx, rsd) + - runtime_stat_avg(stat, STAT_TOPDOWN_BE_BOUND, cpu_map_idx, rsd) + - runtime_stat_avg(stat, STAT_TOPDOWN_BAD_SPEC, cpu_map_idx, rsd); - double d = runtime_stat_avg(stat, type, cpu_map_idx, rsd); + double sum = runtime_stat_avg(stat, STAT_TOPDOWN_RETIRING, map_idx, rsd) + + runtime_stat_avg(stat, STAT_TOPDOWN_FE_BOUND, map_idx, rsd) + + runtime_stat_avg(stat, STAT_TOPDOWN_BE_BOUND, map_idx, rsd) + + runtime_stat_avg(stat, STAT_TOPDOWN_BAD_SPEC, map_idx, rsd); + double d = runtime_stat_avg(stat, type, map_idx, rsd); if (sum) return d / sum; @@ -777,23 +765,23 @@ static double td_metric_ratio(int cpu_map_idx, enum stat_type type, * We allow two missing. */ -static bool full_td(int cpu_map_idx, struct runtime_stat *stat, +static bool full_td(int map_idx, struct runtime_stat *stat, struct runtime_stat_data *rsd) { int c = 0; - if (runtime_stat_avg(stat, STAT_TOPDOWN_RETIRING, cpu_map_idx, rsd) > 0) + if (runtime_stat_avg(stat, STAT_TOPDOWN_RETIRING, map_idx, rsd) > 0) c++; - if (runtime_stat_avg(stat, STAT_TOPDOWN_BE_BOUND, cpu_map_idx, rsd) > 0) + if (runtime_stat_avg(stat, STAT_TOPDOWN_BE_BOUND, map_idx, rsd) > 0) c++; - if (runtime_stat_avg(stat, STAT_TOPDOWN_FE_BOUND, cpu_map_idx, rsd) > 0) + if (runtime_stat_avg(stat, STAT_TOPDOWN_FE_BOUND, map_idx, rsd) > 0) c++; - if (runtime_stat_avg(stat, STAT_TOPDOWN_BAD_SPEC, cpu_map_idx, rsd) > 0) + if (runtime_stat_avg(stat, STAT_TOPDOWN_BAD_SPEC, map_idx, rsd) > 0) c++; return c >= 2; } -static void print_smi_cost(struct perf_stat_config *config, int cpu_map_idx, +static void print_smi_cost(struct perf_stat_config *config, int map_idx, struct perf_stat_output_ctx *out, struct runtime_stat *st, struct runtime_stat_data *rsd) @@ -801,9 +789,9 @@ static void print_smi_cost(struct perf_stat_config *config, int cpu_map_idx, double smi_num, aperf, cycles, cost = 0.0; const char *color = NULL; - smi_num = runtime_stat_avg(st, STAT_SMI_NUM, cpu_map_idx, rsd); - aperf = runtime_stat_avg(st, STAT_APERF, cpu_map_idx, rsd); - cycles = runtime_stat_avg(st, STAT_CYCLES, cpu_map_idx, rsd); + smi_num = runtime_stat_avg(st, STAT_SMI_NUM, map_idx, rsd); + aperf = runtime_stat_avg(st, STAT_APERF, map_idx, rsd); + cycles = runtime_stat_avg(st, STAT_CYCLES, map_idx, rsd); if ((cycles == 0) || (aperf == 0)) return; @@ -820,7 +808,7 @@ static void print_smi_cost(struct perf_stat_config *config, int cpu_map_idx, static int prepare_metric(struct evsel **metric_events, struct metric_ref *metric_refs, struct expr_parse_ctx *pctx, - int cpu_map_idx, + int map_idx, struct runtime_stat *st) { double scale; @@ -859,17 +847,22 @@ static int prepare_metric(struct evsel **metric_events, abort(); } } else { - v = saved_value_lookup(metric_events[i], cpu_map_idx, false, + v = saved_value_lookup(metric_events[i], map_idx, false, STAT_NONE, 0, st, metric_events[i]->cgrp); if (!v) break; stats = &v->stats; - scale = 1.0; + /* + * If an event was scaled during stat gathering, reverse + * the scale before computing the metric. + */ + scale = 1.0 / metric_events[i]->scale; + source_count = evsel__source_count(metric_events[i]); if (v->metric_other) - metric_total = v->metric_total; + metric_total = v->metric_total * scale; } n = strdup(evsel__metric_id(metric_events[i])); if (!n) @@ -897,7 +890,7 @@ static void generic_metric(struct perf_stat_config *config, const char *metric_name, const char *metric_unit, int runtime, - int cpu_map_idx, + int map_idx, struct perf_stat_output_ctx *out, struct runtime_stat *st) { @@ -911,8 +904,11 @@ static void generic_metric(struct perf_stat_config *config, if (!pctx) return; - pctx->runtime = runtime; - i = prepare_metric(metric_events, metric_refs, pctx, cpu_map_idx, st); + if (config->user_requested_cpu_list) + pctx->sctx.user_requested_cpu_list = strdup(config->user_requested_cpu_list); + pctx->sctx.runtime = runtime; + pctx->sctx.system_wide = config->system_wide; + i = prepare_metric(metric_events, metric_refs, pctx, map_idx, st); if (i < 0) { expr__ctx_free(pctx); return; @@ -957,7 +953,7 @@ static void generic_metric(struct perf_stat_config *config, expr__ctx_free(pctx); } -double test_generic_metric(struct metric_expr *mexp, int cpu_map_idx, struct runtime_stat *st) +double test_generic_metric(struct metric_expr *mexp, int map_idx, struct runtime_stat *st) { struct expr_parse_ctx *pctx; double ratio = 0.0; @@ -966,7 +962,7 @@ double test_generic_metric(struct metric_expr *mexp, int cpu_map_idx, struct run if (!pctx) return NAN; - if (prepare_metric(mexp->metric_events, mexp->metric_refs, pctx, cpu_map_idx, st) < 0) + if (prepare_metric(mexp->metric_events, mexp->metric_refs, pctx, map_idx, st) < 0) goto out; if (expr__parse(&ratio, pctx, mexp->metric_expr)) @@ -979,7 +975,7 @@ out: void perf_stat__print_shadow_stats(struct perf_stat_config *config, struct evsel *evsel, - double avg, int cpu_map_idx, + double avg, int map_idx, struct perf_stat_output_ctx *out, struct rblist *metric_events, struct runtime_stat *st) @@ -998,7 +994,7 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, if (config->iostat_run) { iostat_print_metric(config, evsel, out); } else if (evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) { - total = runtime_stat_avg(st, STAT_CYCLES, cpu_map_idx, &rsd); + total = runtime_stat_avg(st, STAT_CYCLES, map_idx, &rsd); if (total) { ratio = avg / total; @@ -1008,11 +1004,11 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, print_metric(config, ctxp, NULL, NULL, "insn per cycle", 0); } - total = runtime_stat_avg(st, STAT_STALLED_CYCLES_FRONT, cpu_map_idx, &rsd); + total = runtime_stat_avg(st, STAT_STALLED_CYCLES_FRONT, map_idx, &rsd); total = max(total, runtime_stat_avg(st, STAT_STALLED_CYCLES_BACK, - cpu_map_idx, &rsd)); + map_idx, &rsd)); if (total && avg) { out->new_line(config, ctxp); @@ -1022,8 +1018,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, ratio); } } else if (evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES)) { - if (runtime_stat_n(st, STAT_BRANCHES, cpu_map_idx, &rsd) != 0) - print_branch_misses(config, cpu_map_idx, avg, out, st, &rsd); + if (runtime_stat_n(st, STAT_BRANCHES, map_idx, &rsd) != 0) + print_branch_misses(config, map_idx, avg, out, st, &rsd); else print_metric(config, ctxp, NULL, NULL, "of all branches", 0); } else if ( @@ -1032,8 +1028,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, ((PERF_COUNT_HW_CACHE_OP_READ) << 8) | ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) { - if (runtime_stat_n(st, STAT_L1_DCACHE, cpu_map_idx, &rsd) != 0) - print_l1_dcache_misses(config, cpu_map_idx, avg, out, st, &rsd); + if (runtime_stat_n(st, STAT_L1_DCACHE, map_idx, &rsd) != 0) + print_l1_dcache_misses(config, map_idx, avg, out, st, &rsd); else print_metric(config, ctxp, NULL, NULL, "of all L1-dcache accesses", 0); } else if ( @@ -1042,8 +1038,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, ((PERF_COUNT_HW_CACHE_OP_READ) << 8) | ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) { - if (runtime_stat_n(st, STAT_L1_ICACHE, cpu_map_idx, &rsd) != 0) - print_l1_icache_misses(config, cpu_map_idx, avg, out, st, &rsd); + if (runtime_stat_n(st, STAT_L1_ICACHE, map_idx, &rsd) != 0) + print_l1_icache_misses(config, map_idx, avg, out, st, &rsd); else print_metric(config, ctxp, NULL, NULL, "of all L1-icache accesses", 0); } else if ( @@ -1052,8 +1048,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, ((PERF_COUNT_HW_CACHE_OP_READ) << 8) | ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) { - if (runtime_stat_n(st, STAT_DTLB_CACHE, cpu_map_idx, &rsd) != 0) - print_dtlb_cache_misses(config, cpu_map_idx, avg, out, st, &rsd); + if (runtime_stat_n(st, STAT_DTLB_CACHE, map_idx, &rsd) != 0) + print_dtlb_cache_misses(config, map_idx, avg, out, st, &rsd); else print_metric(config, ctxp, NULL, NULL, "of all dTLB cache accesses", 0); } else if ( @@ -1062,8 +1058,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, ((PERF_COUNT_HW_CACHE_OP_READ) << 8) | ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) { - if (runtime_stat_n(st, STAT_ITLB_CACHE, cpu_map_idx, &rsd) != 0) - print_itlb_cache_misses(config, cpu_map_idx, avg, out, st, &rsd); + if (runtime_stat_n(st, STAT_ITLB_CACHE, map_idx, &rsd) != 0) + print_itlb_cache_misses(config, map_idx, avg, out, st, &rsd); else print_metric(config, ctxp, NULL, NULL, "of all iTLB cache accesses", 0); } else if ( @@ -1072,27 +1068,27 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, ((PERF_COUNT_HW_CACHE_OP_READ) << 8) | ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) { - if (runtime_stat_n(st, STAT_LL_CACHE, cpu_map_idx, &rsd) != 0) - print_ll_cache_misses(config, cpu_map_idx, avg, out, st, &rsd); + if (runtime_stat_n(st, STAT_LL_CACHE, map_idx, &rsd) != 0) + print_ll_cache_misses(config, map_idx, avg, out, st, &rsd); else print_metric(config, ctxp, NULL, NULL, "of all LL-cache accesses", 0); } else if (evsel__match(evsel, HARDWARE, HW_CACHE_MISSES)) { - total = runtime_stat_avg(st, STAT_CACHEREFS, cpu_map_idx, &rsd); + total = runtime_stat_avg(st, STAT_CACHEREFS, map_idx, &rsd); if (total) ratio = avg * 100 / total; - if (runtime_stat_n(st, STAT_CACHEREFS, cpu_map_idx, &rsd) != 0) + if (runtime_stat_n(st, STAT_CACHEREFS, map_idx, &rsd) != 0) print_metric(config, ctxp, NULL, "%8.3f %%", "of all cache refs", ratio); else print_metric(config, ctxp, NULL, NULL, "of all cache refs", 0); } else if (evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) { - print_stalled_cycles_frontend(config, cpu_map_idx, avg, out, st, &rsd); + print_stalled_cycles_frontend(config, map_idx, avg, out, st, &rsd); } else if (evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) { - print_stalled_cycles_backend(config, cpu_map_idx, avg, out, st, &rsd); + print_stalled_cycles_backend(config, map_idx, avg, out, st, &rsd); } else if (evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) { - total = runtime_stat_avg(st, STAT_NSECS, cpu_map_idx, &rsd); + total = runtime_stat_avg(st, STAT_NSECS, map_idx, &rsd); if (total) { ratio = avg / total; @@ -1101,7 +1097,7 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, print_metric(config, ctxp, NULL, NULL, "Ghz", 0); } } else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) { - total = runtime_stat_avg(st, STAT_CYCLES, cpu_map_idx, &rsd); + total = runtime_stat_avg(st, STAT_CYCLES, map_idx, &rsd); if (total) print_metric(config, ctxp, NULL, @@ -1111,8 +1107,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, print_metric(config, ctxp, NULL, NULL, "transactional cycles", 0); } else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) { - total = runtime_stat_avg(st, STAT_CYCLES, cpu_map_idx, &rsd); - total2 = runtime_stat_avg(st, STAT_CYCLES_IN_TX, cpu_map_idx, &rsd); + total = runtime_stat_avg(st, STAT_CYCLES, map_idx, &rsd); + total2 = runtime_stat_avg(st, STAT_CYCLES_IN_TX, map_idx, &rsd); if (total2 < avg) total2 = avg; @@ -1122,19 +1118,19 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, else print_metric(config, ctxp, NULL, NULL, "aborted cycles", 0); } else if (perf_stat_evsel__is(evsel, TRANSACTION_START)) { - total = runtime_stat_avg(st, STAT_CYCLES_IN_TX, cpu_map_idx, &rsd); + total = runtime_stat_avg(st, STAT_CYCLES_IN_TX, map_idx, &rsd); if (avg) ratio = total / avg; - if (runtime_stat_n(st, STAT_CYCLES_IN_TX, cpu_map_idx, &rsd) != 0) + if (runtime_stat_n(st, STAT_CYCLES_IN_TX, map_idx, &rsd) != 0) print_metric(config, ctxp, NULL, "%8.0f", "cycles / transaction", ratio); else print_metric(config, ctxp, NULL, NULL, "cycles / transaction", 0); } else if (perf_stat_evsel__is(evsel, ELISION_START)) { - total = runtime_stat_avg(st, STAT_CYCLES_IN_TX, cpu_map_idx, &rsd); + total = runtime_stat_avg(st, STAT_CYCLES_IN_TX, map_idx, &rsd); if (avg) ratio = total / avg; @@ -1147,28 +1143,28 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, else print_metric(config, ctxp, NULL, NULL, "CPUs utilized", 0); } else if (perf_stat_evsel__is(evsel, TOPDOWN_FETCH_BUBBLES)) { - double fe_bound = td_fe_bound(cpu_map_idx, st, &rsd); + double fe_bound = td_fe_bound(map_idx, st, &rsd); if (fe_bound > 0.2) color = PERF_COLOR_RED; print_metric(config, ctxp, color, "%8.1f%%", "frontend bound", fe_bound * 100.); } else if (perf_stat_evsel__is(evsel, TOPDOWN_SLOTS_RETIRED)) { - double retiring = td_retiring(cpu_map_idx, st, &rsd); + double retiring = td_retiring(map_idx, st, &rsd); if (retiring > 0.7) color = PERF_COLOR_GREEN; print_metric(config, ctxp, color, "%8.1f%%", "retiring", retiring * 100.); } else if (perf_stat_evsel__is(evsel, TOPDOWN_RECOVERY_BUBBLES)) { - double bad_spec = td_bad_spec(cpu_map_idx, st, &rsd); + double bad_spec = td_bad_spec(map_idx, st, &rsd); if (bad_spec > 0.1) color = PERF_COLOR_RED; print_metric(config, ctxp, color, "%8.1f%%", "bad speculation", bad_spec * 100.); } else if (perf_stat_evsel__is(evsel, TOPDOWN_SLOTS_ISSUED)) { - double be_bound = td_be_bound(cpu_map_idx, st, &rsd); + double be_bound = td_be_bound(map_idx, st, &rsd); const char *name = "backend bound"; static int have_recovery_bubbles = -1; @@ -1181,14 +1177,14 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, if (be_bound > 0.2) color = PERF_COLOR_RED; - if (td_total_slots(cpu_map_idx, st, &rsd) > 0) + if (td_total_slots(map_idx, st, &rsd) > 0) print_metric(config, ctxp, color, "%8.1f%%", name, be_bound * 100.); else print_metric(config, ctxp, NULL, NULL, name, 0); } else if (perf_stat_evsel__is(evsel, TOPDOWN_RETIRING) && - full_td(cpu_map_idx, st, &rsd)) { - double retiring = td_metric_ratio(cpu_map_idx, + full_td(map_idx, st, &rsd)) { + double retiring = td_metric_ratio(map_idx, STAT_TOPDOWN_RETIRING, st, &rsd); if (retiring > 0.7) @@ -1196,8 +1192,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, print_metric(config, ctxp, color, "%8.1f%%", "Retiring", retiring * 100.); } else if (perf_stat_evsel__is(evsel, TOPDOWN_FE_BOUND) && - full_td(cpu_map_idx, st, &rsd)) { - double fe_bound = td_metric_ratio(cpu_map_idx, + full_td(map_idx, st, &rsd)) { + double fe_bound = td_metric_ratio(map_idx, STAT_TOPDOWN_FE_BOUND, st, &rsd); if (fe_bound > 0.2) @@ -1205,8 +1201,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, print_metric(config, ctxp, color, "%8.1f%%", "Frontend Bound", fe_bound * 100.); } else if (perf_stat_evsel__is(evsel, TOPDOWN_BE_BOUND) && - full_td(cpu_map_idx, st, &rsd)) { - double be_bound = td_metric_ratio(cpu_map_idx, + full_td(map_idx, st, &rsd)) { + double be_bound = td_metric_ratio(map_idx, STAT_TOPDOWN_BE_BOUND, st, &rsd); if (be_bound > 0.2) @@ -1214,8 +1210,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, print_metric(config, ctxp, color, "%8.1f%%", "Backend Bound", be_bound * 100.); } else if (perf_stat_evsel__is(evsel, TOPDOWN_BAD_SPEC) && - full_td(cpu_map_idx, st, &rsd)) { - double bad_spec = td_metric_ratio(cpu_map_idx, + full_td(map_idx, st, &rsd)) { + double bad_spec = td_metric_ratio(map_idx, STAT_TOPDOWN_BAD_SPEC, st, &rsd); if (bad_spec > 0.1) @@ -1223,11 +1219,11 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, print_metric(config, ctxp, color, "%8.1f%%", "Bad Speculation", bad_spec * 100.); } else if (perf_stat_evsel__is(evsel, TOPDOWN_HEAVY_OPS) && - full_td(cpu_map_idx, st, &rsd) && (config->topdown_level > 1)) { - double retiring = td_metric_ratio(cpu_map_idx, + full_td(map_idx, st, &rsd) && (config->topdown_level > 1)) { + double retiring = td_metric_ratio(map_idx, STAT_TOPDOWN_RETIRING, st, &rsd); - double heavy_ops = td_metric_ratio(cpu_map_idx, + double heavy_ops = td_metric_ratio(map_idx, STAT_TOPDOWN_HEAVY_OPS, st, &rsd); double light_ops = retiring - heavy_ops; @@ -1243,11 +1239,11 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, print_metric(config, ctxp, color, "%8.1f%%", "Light Operations", light_ops * 100.); } else if (perf_stat_evsel__is(evsel, TOPDOWN_BR_MISPREDICT) && - full_td(cpu_map_idx, st, &rsd) && (config->topdown_level > 1)) { - double bad_spec = td_metric_ratio(cpu_map_idx, + full_td(map_idx, st, &rsd) && (config->topdown_level > 1)) { + double bad_spec = td_metric_ratio(map_idx, STAT_TOPDOWN_BAD_SPEC, st, &rsd); - double br_mis = td_metric_ratio(cpu_map_idx, + double br_mis = td_metric_ratio(map_idx, STAT_TOPDOWN_BR_MISPREDICT, st, &rsd); double m_clears = bad_spec - br_mis; @@ -1263,11 +1259,11 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, print_metric(config, ctxp, color, "%8.1f%%", "Machine Clears", m_clears * 100.); } else if (perf_stat_evsel__is(evsel, TOPDOWN_FETCH_LAT) && - full_td(cpu_map_idx, st, &rsd) && (config->topdown_level > 1)) { - double fe_bound = td_metric_ratio(cpu_map_idx, + full_td(map_idx, st, &rsd) && (config->topdown_level > 1)) { + double fe_bound = td_metric_ratio(map_idx, STAT_TOPDOWN_FE_BOUND, st, &rsd); - double fetch_lat = td_metric_ratio(cpu_map_idx, + double fetch_lat = td_metric_ratio(map_idx, STAT_TOPDOWN_FETCH_LAT, st, &rsd); double fetch_bw = fe_bound - fetch_lat; @@ -1283,11 +1279,11 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, print_metric(config, ctxp, color, "%8.1f%%", "Fetch Bandwidth", fetch_bw * 100.); } else if (perf_stat_evsel__is(evsel, TOPDOWN_MEM_BOUND) && - full_td(cpu_map_idx, st, &rsd) && (config->topdown_level > 1)) { - double be_bound = td_metric_ratio(cpu_map_idx, + full_td(map_idx, st, &rsd) && (config->topdown_level > 1)) { + double be_bound = td_metric_ratio(map_idx, STAT_TOPDOWN_BE_BOUND, st, &rsd); - double mem_bound = td_metric_ratio(cpu_map_idx, + double mem_bound = td_metric_ratio(map_idx, STAT_TOPDOWN_MEM_BOUND, st, &rsd); double core_bound = be_bound - mem_bound; @@ -1304,12 +1300,13 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, core_bound * 100.); } else if (evsel->metric_expr) { generic_metric(config, evsel->metric_expr, evsel->metric_events, NULL, - evsel->name, evsel->metric_name, NULL, 1, cpu_map_idx, out, st); - } else if (runtime_stat_n(st, STAT_NSECS, cpu_map_idx, &rsd) != 0) { + evsel->name, evsel->metric_name, NULL, 1, + map_idx, out, st); + } else if (runtime_stat_n(st, STAT_NSECS, map_idx, &rsd) != 0) { char unit = ' '; char unit_buf[10] = "/sec"; - total = runtime_stat_avg(st, STAT_NSECS, cpu_map_idx, &rsd); + total = runtime_stat_avg(st, STAT_NSECS, map_idx, &rsd); if (total) ratio = convert_unit_double(1000000000.0 * avg / total, &unit); @@ -1317,7 +1314,7 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, snprintf(unit_buf, sizeof(unit_buf), "%c/sec", unit); print_metric(config, ctxp, NULL, "%8.3f", unit_buf, ratio); } else if (perf_stat_evsel__is(evsel, SMI_NUM)) { - print_smi_cost(config, cpu_map_idx, out, st, &rsd); + print_smi_cost(config, map_idx, out, st, &rsd); } else { num = 0; } @@ -1329,8 +1326,9 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config, if (num++ > 0) out->new_line(config, ctxp); generic_metric(config, mexp->metric_expr, mexp->metric_events, - mexp->metric_refs, evsel->name, mexp->metric_name, - mexp->metric_unit, mexp->runtime, cpu_map_idx, out, st); + mexp->metric_refs, evsel->name, mexp->metric_name, + mexp->metric_unit, mexp->runtime, + map_idx, out, st); } } if (num == 0) diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c index 0882b4754fcf..8ec8bb4a9912 100644 --- a/tools/perf/util/stat.c +++ b/tools/perf/util/stat.c @@ -14,7 +14,11 @@ #include "evlist.h" #include "evsel.h" #include "thread_map.h" -#include "hashmap.h" +#ifdef HAVE_LIBBPF_SUPPORT +#include <bpf/hashmap.h> +#else +#include "util/hashmap.h" +#endif #include <linux/zalloc.h> void update_stats(struct stats *stats, u64 val) @@ -128,13 +132,9 @@ static void perf_stat_evsel_id_init(struct evsel *evsel) static void evsel__reset_stat_priv(struct evsel *evsel) { - int i; struct perf_stat_evsel *ps = evsel->stats; - for (i = 0; i < 3; i++) - init_stats(&ps->res_stats[i]); - - perf_stat_evsel_id_init(evsel); + init_stats(&ps->res_stats); } static int evsel__alloc_stat_priv(struct evsel *evsel) @@ -142,6 +142,7 @@ static int evsel__alloc_stat_priv(struct evsel *evsel) evsel->stats = zalloc(sizeof(struct perf_stat_evsel)); if (evsel->stats == NULL) return -ENOMEM; + perf_stat_evsel_id_init(evsel); evsel__reset_stat_priv(evsel); return 0; } @@ -388,12 +389,8 @@ process_counter_values(struct perf_stat_config *config, struct evsel *evsel, } if (config->aggr_mode == AGGR_THREAD) { - if (config->stats) - perf_stat__update_shadow_stats(evsel, - count->val, 0, &config->stats[thread]); - else - perf_stat__update_shadow_stats(evsel, - count->val, 0, &rt_stat); + perf_stat__update_shadow_stats(evsel, count->val, + thread, &rt_stat); } break; case AGGR_GLOBAL: @@ -416,9 +413,6 @@ static int process_counter_maps(struct perf_stat_config *config, int ncpus = evsel__nr_cpus(counter); int idx, thread; - if (counter->core.system_wide) - nthreads = 1; - for (thread = 0; thread < nthreads; thread++) { for (idx = 0; idx < ncpus; idx++) { if (process_counter_values(config, counter, idx, thread, @@ -436,7 +430,7 @@ int perf_stat_process_counter(struct perf_stat_config *config, struct perf_counts_values *aggr = &counter->counts->aggr; struct perf_stat_evsel *ps = counter->stats; u64 *count = counter->counts->aggr.values; - int i, ret; + int ret; aggr->val = aggr->ena = aggr->run = 0; @@ -454,8 +448,7 @@ int perf_stat_process_counter(struct perf_stat_config *config, evsel__compute_deltas(counter, -1, -1, aggr); perf_counts_values__scale(aggr, config->scale, &counter->counts->scaled); - for (i = 0; i < 3; i++) - update_stats(&ps->res_stats[i], count[i]); + update_stats(&ps->res_stats, *count); if (verbose > 0) { fprintf(config->output, "%s: %" PRIu64 " %" PRIu64 " %" PRIu64 "\n", diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h index 668250022f8c..b0899c6e002f 100644 --- a/tools/perf/util/stat.h +++ b/tools/perf/util/stat.h @@ -43,7 +43,7 @@ enum perf_stat_evsel_id { }; struct perf_stat_evsel { - struct stats res_stats[3]; + struct stats res_stats; enum perf_stat_evsel_id id; u64 *group_data; }; @@ -141,6 +141,8 @@ struct perf_stat_config { bool stop_read_counter; bool quiet; bool iostat_run; + char *user_requested_cpu_list; + bool system_wide; FILE *output; unsigned int interval; unsigned int timeout; @@ -151,8 +153,6 @@ struct perf_stat_config { int run_count; int print_free_counters_hint; int print_mixed_hw_group_error; - struct runtime_stat *stats; - int stats_num; const char *csv_sep; struct stats *walltime_nsecs_stats; struct rusage ru_data; @@ -232,7 +232,7 @@ void perf_stat__init_shadow_stats(void); void perf_stat__reset_shadow_stats(void); void perf_stat__reset_shadow_per_stat(struct runtime_stat *st); void perf_stat__update_shadow_stats(struct evsel *counter, u64 count, - int cpu_map_idx, struct runtime_stat *st); + int map_idx, struct runtime_stat *st); struct perf_stat_output_ctx { void *ctx; print_metric_t print_metric; @@ -242,7 +242,7 @@ struct perf_stat_output_ctx { void perf_stat__print_shadow_stats(struct perf_stat_config *config, struct evsel *evsel, - double avg, int cpu, + double avg, int map_idx, struct perf_stat_output_ctx *out, struct rblist *metric_events, struct runtime_stat *st); @@ -277,5 +277,5 @@ void evlist__print_counters(struct evlist *evlist, struct perf_stat_config *conf struct target *_target, struct timespec *ts, int argc, const char **argv); struct metric_expr; -double test_generic_metric(struct metric_expr *mexp, int cpu_map_idx, struct runtime_stat *st); +double test_generic_metric(struct metric_expr *mexp, int map_idx, struct runtime_stat *st); #endif diff --git a/tools/perf/util/string.c b/tools/perf/util/string.c index f6d90cdd9225..4f12a96f33cc 100644 --- a/tools/perf/util/string.c +++ b/tools/perf/util/string.c @@ -15,7 +15,6 @@ const char *dots = "....................................................................." "....................................................................."; -#define K 1024LL /* * perf_atoll() * Parse (\d+)(b|B|kb|KB|mb|MB|gb|GB|tb|TB) (e.g. "256MB") diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index a4b22caa7c24..a3a165ae933a 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1791,6 +1791,7 @@ int dso__load(struct dso *dso, struct map *map) char newmapname[PATH_MAX]; const char *map_path = dso->long_name; + mutex_lock(&dso->lock); perfmap = strncmp(dso->name, "/tmp/perf-", 10) == 0; if (perfmap) { if (dso->nsinfo && (dso__find_perf_map(newmapname, @@ -1800,7 +1801,6 @@ int dso__load(struct dso *dso, struct map *map) } nsinfo__mountns_enter(dso->nsinfo, &nsc); - pthread_mutex_lock(&dso->lock); /* check again under the dso->lock */ if (dso__loaded(dso)) { @@ -1964,7 +1964,7 @@ out_free: ret = 0; out: dso__set_loaded(dso); - pthread_mutex_unlock(&dso->lock); + mutex_unlock(&dso->lock); nsinfo__mountns_exit(&nsc); return ret; diff --git a/tools/perf/util/synthetic-events.c b/tools/perf/util/synthetic-events.c index 538790758e24..cccd293b5312 100644 --- a/tools/perf/util/synthetic-events.c +++ b/tools/perf/util/synthetic-events.c @@ -364,11 +364,14 @@ static bool read_proc_maps_line(struct io *io, __u64 *start, __u64 *end, } static void perf_record_mmap2__read_build_id(struct perf_record_mmap2 *event, + struct machine *machine, bool is_kernel) { struct build_id bid; struct nsinfo *nsi; struct nscookie nc; + struct dso *dso = NULL; + struct dso_id id; int rc; if (is_kernel) { @@ -376,6 +379,18 @@ static void perf_record_mmap2__read_build_id(struct perf_record_mmap2 *event, goto out; } + id.maj = event->maj; + id.min = event->min; + id.ino = event->ino; + id.ino_generation = event->ino_generation; + + dso = dsos__findnew_id(&machine->dsos, event->filename, &id); + if (dso && dso->has_build_id) { + bid = dso->bid; + rc = 0; + goto out; + } + nsi = nsinfo__new(event->pid); nsinfo__mountns_enter(nsi, &nc); @@ -391,12 +406,16 @@ out: event->header.misc |= PERF_RECORD_MISC_MMAP_BUILD_ID; event->__reserved_1 = 0; event->__reserved_2 = 0; + + if (dso && !dso->has_build_id) + dso__set_build_id(dso, &bid); } else { if (event->filename[0] == '/') { pr_debug2("Failed to read build ID for %s\n", event->filename); } } + dso__put(dso); } int perf_event__synthesize_mmap_events(struct perf_tool *tool, @@ -507,7 +526,7 @@ out: event->mmap2.tid = pid; if (symbol_conf.buildid_mmap2) - perf_record_mmap2__read_build_id(&event->mmap2, false); + perf_record_mmap2__read_build_id(&event->mmap2, machine, false); if (perf_tool__process_synth_event(tool, event, machine, process) != 0) { rc = -1; @@ -690,7 +709,7 @@ int perf_event__synthesize_modules(struct perf_tool *tool, perf_event__handler_t memcpy(event->mmap2.filename, pos->dso->long_name, pos->dso->long_name_len + 1); - perf_record_mmap2__read_build_id(&event->mmap2, false); + perf_record_mmap2__read_build_id(&event->mmap2, machine, false); } else { size = PERF_ALIGN(pos->dso->long_name_len + 1, sizeof(u64)); event->mmap.header.type = PERF_RECORD_MMAP; @@ -1126,7 +1145,7 @@ static int __perf_event__synthesize_kernel_mmap(struct perf_tool *tool, event->mmap2.len = map->end - event->mmap.start; event->mmap2.pid = machine->pid; - perf_record_mmap2__read_build_id(&event->mmap2, true); + perf_record_mmap2__read_build_id(&event->mmap2, machine, true); } else { size = snprintf(event->mmap.filename, sizeof(event->mmap.filename), "%s%s", machine->mmap_name, kmap->ref_reloc_sym->name) + 1; @@ -1195,93 +1214,97 @@ int perf_event__synthesize_thread_map2(struct perf_tool *tool, return err; } -static void synthesize_cpus(struct perf_record_cpu_map_data *data, - const struct perf_cpu_map *map) -{ - int i, map_nr = perf_cpu_map__nr(map); - - data->cpus_data.nr = map_nr; +struct synthesize_cpu_map_data { + const struct perf_cpu_map *map; + int nr; + int min_cpu; + int max_cpu; + int has_any_cpu; + int type; + size_t size; + struct perf_record_cpu_map_data *data; +}; - for (i = 0; i < map_nr; i++) - data->cpus_data.cpu[i] = perf_cpu_map__cpu(map, i).cpu; +static void synthesize_cpus(struct synthesize_cpu_map_data *data) +{ + data->data->type = PERF_CPU_MAP__CPUS; + data->data->cpus_data.nr = data->nr; + for (int i = 0; i < data->nr; i++) + data->data->cpus_data.cpu[i] = perf_cpu_map__cpu(data->map, i).cpu; } -static void synthesize_mask(struct perf_record_cpu_map_data *data, - const struct perf_cpu_map *map, int max) +static void synthesize_mask(struct synthesize_cpu_map_data *data) { int idx; struct perf_cpu cpu; /* Due to padding, the 4bytes per entry mask variant is always smaller. */ - data->mask32_data.nr = BITS_TO_U32(max); - data->mask32_data.long_size = 4; + data->data->type = PERF_CPU_MAP__MASK; + data->data->mask32_data.nr = BITS_TO_U32(data->max_cpu); + data->data->mask32_data.long_size = 4; - perf_cpu_map__for_each_cpu(cpu, idx, map) { + perf_cpu_map__for_each_cpu(cpu, idx, data->map) { int bit_word = cpu.cpu / 32; - __u32 bit_mask = 1U << (cpu.cpu & 31); + u32 bit_mask = 1U << (cpu.cpu & 31); - data->mask32_data.mask[bit_word] |= bit_mask; + data->data->mask32_data.mask[bit_word] |= bit_mask; } } -static size_t cpus_size(const struct perf_cpu_map *map) -{ - return sizeof(struct cpu_map_entries) + perf_cpu_map__nr(map) * sizeof(u16); -} - -static size_t mask_size(const struct perf_cpu_map *map, int *max) +static void synthesize_range_cpus(struct synthesize_cpu_map_data *data) { - *max = perf_cpu_map__max(map).cpu; - return sizeof(struct perf_record_mask_cpu_map32) + BITS_TO_U32(*max) * sizeof(__u32); + data->data->type = PERF_CPU_MAP__RANGE_CPUS; + data->data->range_cpu_data.any_cpu = data->has_any_cpu; + data->data->range_cpu_data.start_cpu = data->min_cpu; + data->data->range_cpu_data.end_cpu = data->max_cpu; } -static void *cpu_map_data__alloc(const struct perf_cpu_map *map, size_t *size, - u16 *type, int *max) +static void *cpu_map_data__alloc(struct synthesize_cpu_map_data *syn_data, + size_t header_size) { size_t size_cpus, size_mask; - bool is_dummy = perf_cpu_map__empty(map); - /* - * Both array and mask data have variable size based - * on the number of cpus and their actual values. - * The size of the 'struct perf_record_cpu_map_data' is: - * - * array = size of 'struct cpu_map_entries' + - * number of cpus * sizeof(u64) - * - * mask = size of 'struct perf_record_record_cpu_map' + - * maximum cpu bit converted to size of longs - * - * and finally + the size of 'struct perf_record_cpu_map_data'. - */ - size_cpus = cpus_size(map); - size_mask = mask_size(map, max); + syn_data->nr = perf_cpu_map__nr(syn_data->map); + syn_data->has_any_cpu = (perf_cpu_map__cpu(syn_data->map, 0).cpu == -1) ? 1 : 0; - if (is_dummy || (size_cpus < size_mask)) { - *size += size_cpus; - *type = PERF_CPU_MAP__CPUS; - } else { - *size += size_mask; - *type = PERF_CPU_MAP__MASK; + syn_data->min_cpu = perf_cpu_map__cpu(syn_data->map, syn_data->has_any_cpu).cpu; + syn_data->max_cpu = perf_cpu_map__max(syn_data->map).cpu; + if (syn_data->max_cpu - syn_data->min_cpu + 1 == syn_data->nr - syn_data->has_any_cpu) { + /* A consecutive range of CPUs can be encoded using a range. */ + assert(sizeof(u16) + sizeof(struct perf_record_range_cpu_map) == sizeof(u64)); + syn_data->type = PERF_CPU_MAP__RANGE_CPUS; + syn_data->size = header_size + sizeof(u64); + return zalloc(syn_data->size); } - *size += sizeof(__u16); /* For perf_record_cpu_map_data.type. */ - *size = PERF_ALIGN(*size, sizeof(u64)); - return zalloc(*size); + size_cpus = sizeof(u16) + sizeof(struct cpu_map_entries) + syn_data->nr * sizeof(u16); + /* Due to padding, the 4bytes per entry mask variant is always smaller. */ + size_mask = sizeof(u16) + sizeof(struct perf_record_mask_cpu_map32) + + BITS_TO_U32(syn_data->max_cpu) * sizeof(__u32); + if (syn_data->has_any_cpu || size_cpus < size_mask) { + /* Follow the CPU map encoding. */ + syn_data->type = PERF_CPU_MAP__CPUS; + syn_data->size = header_size + PERF_ALIGN(size_cpus, sizeof(u64)); + return zalloc(syn_data->size); + } + /* Encode using a bitmask. */ + syn_data->type = PERF_CPU_MAP__MASK; + syn_data->size = header_size + PERF_ALIGN(size_mask, sizeof(u64)); + return zalloc(syn_data->size); } -static void cpu_map_data__synthesize(struct perf_record_cpu_map_data *data, - const struct perf_cpu_map *map, - u16 type, int max) +static void cpu_map_data__synthesize(struct synthesize_cpu_map_data *data) { - data->type = type; - - switch (type) { + switch (data->type) { case PERF_CPU_MAP__CPUS: - synthesize_cpus(data, map); + synthesize_cpus(data); break; case PERF_CPU_MAP__MASK: - synthesize_mask(data, map, max); + synthesize_mask(data); + break; + case PERF_CPU_MAP__RANGE_CPUS: + synthesize_range_cpus(data); + break; default: break; } @@ -1289,23 +1312,22 @@ static void cpu_map_data__synthesize(struct perf_record_cpu_map_data *data, static struct perf_record_cpu_map *cpu_map_event__new(const struct perf_cpu_map *map) { - size_t size = sizeof(struct perf_event_header); + struct synthesize_cpu_map_data syn_data = { .map = map }; struct perf_record_cpu_map *event; - int max; - u16 type; - event = cpu_map_data__alloc(map, &size, &type, &max); + + event = cpu_map_data__alloc(&syn_data, sizeof(struct perf_event_header)); if (!event) return NULL; + syn_data.data = &event->data; event->header.type = PERF_RECORD_CPU_MAP; - event->header.size = size; - event->data.type = type; - - cpu_map_data__synthesize(&event->data, map, type, max); + event->header.size = syn_data.size; + cpu_map_data__synthesize(&syn_data); return event; } + int perf_event__synthesize_cpu_map(struct perf_tool *tool, const struct perf_cpu_map *map, perf_event__handler_t process, @@ -1955,7 +1977,7 @@ int perf_event__synthesize_event_update_unit(struct perf_tool *tool, struct evse if (ev == NULL) return -ENOMEM; - strlcpy(ev->data, evsel->unit, size + 1); + strlcpy(ev->unit, evsel->unit, size + 1); err = process(tool, (union perf_event *)ev, NULL, NULL); free(ev); return err; @@ -1972,8 +1994,7 @@ int perf_event__synthesize_event_update_scale(struct perf_tool *tool, struct evs if (ev == NULL) return -ENOMEM; - ev_data = (struct perf_record_event_update_scale *)ev->data; - ev_data->scale = evsel->scale; + ev->scale.scale = evsel->scale; err = process(tool, (union perf_event *)ev, NULL, NULL); free(ev); return err; @@ -1990,7 +2011,7 @@ int perf_event__synthesize_event_update_name(struct perf_tool *tool, struct evse if (ev == NULL) return -ENOMEM; - strlcpy(ev->data, evsel->name, len + 1); + strlcpy(ev->name, evsel->name, len + 1); err = process(tool, (union perf_event *)ev, NULL, NULL); free(ev); return err; @@ -1999,25 +2020,20 @@ int perf_event__synthesize_event_update_name(struct perf_tool *tool, struct evse int perf_event__synthesize_event_update_cpus(struct perf_tool *tool, struct evsel *evsel, perf_event__handler_t process) { - size_t size = sizeof(struct perf_record_event_update); + struct synthesize_cpu_map_data syn_data = { .map = evsel->core.own_cpus }; struct perf_record_event_update *ev; - int max, err; - u16 type; - - if (!evsel->core.own_cpus) - return 0; + int err; - ev = cpu_map_data__alloc(evsel->core.own_cpus, &size, &type, &max); + ev = cpu_map_data__alloc(&syn_data, sizeof(struct perf_event_header) + 2 * sizeof(u64)); if (!ev) return -ENOMEM; + syn_data.data = &ev->cpus.cpus; ev->header.type = PERF_RECORD_EVENT_UPDATE; - ev->header.size = (u16)size; + ev->header.size = (u16)syn_data.size; ev->type = PERF_EVENT_UPDATE__CPUS; ev->id = evsel->core.id[0]; - - cpu_map_data__synthesize((struct perf_record_cpu_map_data *)ev->data, - evsel->core.own_cpus, type, max); + cpu_map_data__synthesize(&syn_data); err = process(tool, (union perf_event *)ev, NULL, NULL); free(ev); diff --git a/tools/perf/util/top.h b/tools/perf/util/top.h index 1c2c0a838430..a8b0d79bd96c 100644 --- a/tools/perf/util/top.h +++ b/tools/perf/util/top.h @@ -5,6 +5,7 @@ #include "tool.h" #include "evswitch.h" #include "annotate.h" +#include "mutex.h" #include "ordered-events.h" #include "record.h" #include <linux/types.h> @@ -53,8 +54,8 @@ struct perf_top { struct ordered_events *in; struct ordered_events data[2]; bool rotate; - pthread_mutex_t mutex; - pthread_cond_t cond; + struct mutex mutex; + struct cond cond; } qe; }; |