| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
* Add support for RVA fields to crossgen/R2R
RVA fields are became more common with pre-inititialized ReadOnlySpan<byte>. Fix crossgen to deal with them for R2R.
* Fix tests, map new JIT helper for R2R
|
|
attributes) (#21462)
* Fix for #21456 - restrict increased generic attribute allocations to only generic attributes
This is a trivial quick-fix for #21456 where regressions between 2.1 and 3.0 were discovered on most attibute pathways due to the allocation overhead in the generic-supporting pathways. The workaround is to simply not take that slow/expensive path for non-generics.
While I'd like to optimize `RuntimeModule.ResolveMethod` further, there's a public surface area in play there that makes the changes non-trivial. There, we'll have to choose overhead on the public path (which may still be a net win), or duplication in code for another path.
* Update comments
|
|
* Delete vm/context.*
Leftover from remoting
|
|
* Added Close On Exec Attribute for mutex files (#21455)
|
|
(#21457)
In additional to adding "DCI_NUMA" to dbg_channel_names, a static assertion was added to verify that dbg_channel_names length is valid. Also added some comments to help developpers not to forget to add corresponding entries in dbg_channel_names.
FIxes #21454
|
|
This seems to fix an AV in RyuJIT.
|
|
* Leap Seconds Support
* Address Feedback
* More feedback addressing
|
|
* Remove `TryGetGuid()` and `GetDefaultInterfaceForCoclass()` from the
stdinterfaces API surface.
|
|
* Use string.Create in ConvertFromUtf32
Removes the unsafe code from the method. Also happens to make it a bit faster.
* Improve Rune.ToString performance
|
|
Most of the use of SpinWait in CoreLib involves waiting for some short-lived operation to complete on another thread, in which case the spinning thread should backoff as it's unable to make forward progress until the other operation completes. In a few cases, however, SpinWait is being used just around CompareExchange operations, such that at least one thread running this code path is guaranteed to make forward progress, and the backoff in the spinning doesn't actually help (in theory it could help to reduce contention if lots of threads were all trying to CompareExchange concurrently, but in such cases you'd actually want more randomized backoff, as otherwise it's likely all the threads would re-attempt at around the same time and similarly re-encounter contention).
|
|
intrinsics (#21432)
* Updating Vector128.CreateScalar and Vector256.CreateScalar to be implemented using other intrinsics
* Updating Vector128.Equals and Vector256.Equals to be implemented using other intrinsics
* Updating Vector256.WithLower, Vector256.GetUpper, and Vector256.WithUpper to be implemented using other intrinsics
* Updating Vector128.Create(T) and Vector256.Create(T) to be implemented using other intrinsics
* Fixing the `NI_Base_Vector256_As` intrinsics to only fold the cast if AVX is supported and add a clarifying comment to the Vector128/256.Equals code
* Changing the various `*Software()` local functions in Vector128/256 to be `SoftwareFallback()`
|
|
* Add support for loading registered TLBs
|
|
|
|
assumptions (#21439)
|
|
The recent removal of SharedDomain has broken some SOS commands, like
Name2EE or bpmd. There was a code that was enumerating domains and
obtaining some information on them. And the shared domain pointer from
DacpAppDomainStoreData was being included in the list of domains. As it
is NULL now, we have failed to get the information and the domain
iteration loop was exited prematurely.
I have made SOS resilient to the possibility of missing shared domain.
On older runtimes, the shared domain is still being reported.
|
|
|
|
TryResolveStateMachineMethod (#21396)
|
|
good error messages on PIAs that could possibly directly use them. (#21429)
|
|
Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
|
|
Fix a bug that disabled Bmi1/2.X64
|
|
This is blocking TFS integrations.
Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
|
|
I've noticed that when dumping an object with thread locals using SOS
dumpobj command, I never get the thread locals. I've stepped through the
code in SOS that displays thread statics and found that for some strange
reason, it dismisses threads where DomainLocalModule doesn't have
"initialized" flag set for the given class. However, when runtime
accesses thread statics, it just checks "allocated" flag set for the
given class in the ThreadLocalModule. Removing the extra check for
DomainLocalModule fixed the problem and now thread statics are displayed
properly.
|
|
Follow-up from #21270 and #21414.
Block general cloning from inadvertently cloning a candidate call.
Add a separate path for cloning calls that are inline and guarded
devirtualization candidates for use by guarded devirtualization.
Callers need to take extra steps after cloning one of these calls
to properly fix everything up.
Also fix up some issues in the large comment for the fat calli
transformation.
|
|
|
|
Vector256.GetLower to be intrinsics (#21351)
* Moving CreateScalarUnsafe, ToScalar, Vector128.ToVector256Unsafe, and Vector256.GetLower to be intrinsics
* Adding containment support to the helper intrinsics
|
|
Don't struct-promote opaque vectors
|
|
* Use Span instead of manual copy in MetadataImport.GetUserString
* Address PR feedback
|
|
|
|
* Use string.Create in Path.GetRandomFilePath
Removes a good chunk of the unsafe code.
* Address PR feedback
|
|
Follow-up from a review comment in #21270.
|
|
Slightly cleaner, slightly faster.
|
|
CTR.Token should never throw, but it's currently throwing an ObjectDisposedException if the associated CancellationTokenSource has been disposed.
|
|
* Import CustomMarshalers from corefx PR.
* Add tests for marshalling IEnumerator and IEnumerable using the System.Runtime.InteropServices.CustomMarshalers.
* Verify that the marshaller is used in the managed -> native direction in the tests.
* Remove #if's
* PR Feedback.
* Move ComHelpers.h into interop common headers.
* Add some template machinery to ensure that we don't accidentally pass in an incorrect type into DoQueryInterface.
* Add VariantClear
* Make marshalers internal and root them in ILLinkTrim.xml
* Revert enable-if machinery.
* Add tests for the pass-through unwrapping done by the EnumeratorToEnumVariantMarshaler.
* Remove some of my helper code.
* Added todo about exposing EnumerableToDispatchMarshaler.
* PR Feedback.
* Removed the last placement new.
* Use the runtime hooks to enable us to not have to publically expose the custom marshalers.
* Remove out of date comments.
|
|
Lay the groundwork for guarded devirtualization of virtual and interface
calls in the jit.
Introduce the notion of a guarded devirtualization candidate and identify
these if regular devirtualization fails. Use simple heuristics to produce
a class to guess for. Require that the method that would be invoked if the class
guess is correct be a plausible inline candidate.
Generalize the calli transformer to become an indirect call transformer.
This runs after importation because it needs to introduce control flow and
runs before inlining so that the new direct calls it introduces can be inlined.
Implement the transformation to duplicate the call site, devirtualize on the side
where the class is now known exactly, and turn the resulting direct call into an
inline candidate.
Add a motivation and design document.
|
|
Always defined in .NET Core
|
|
* Disable native image usage for collectible assemblies
There are couple of issues with using native images as collectible
assemblies. The most important one is that statics handling is different
for collectible and non-collectible classes. Collectible classes have
statics stored in per class allocated blocks while the non-collectible ones
are all stored per module. So the code that accesses them needs to use
different offsets relative to a statics base address in each of these cases.
The crossgen generates just a single form of code that corresponds to the
non-collectible case and so such code cannot be used when a module is
loaded as collectible.
To mitigate that, we disable using the native code from crossgen-ed
assemblies and JIT all the code instead.
We may want to revisit this in the future and e.g. unify the statics
handling for collectible and non-collectible classes.
|
|
|
|
Fix lclvar "cloning" in const division lowering
|
|
* Remove FromGAC & OnTPAList
* PR feedback
|
|
Fixes #21046
|
|
|
|
Add AsyncIteratorStateMachineAttribute
|
|
|
|
Fix lots of malformed XML comments
|
|
|
|
|
|
These are causing errors in the source.dot.net build. Turning on DocumentationFile and suppressing 1591, 1573, and 0419, Corelib builds clean after these changes.
|
|
the formatProvider (#21350)
|
|
|
|
|
