// Licensed to the .NET Foundation under one or more agreements. // The .NET Foundation licenses this file to you under the MIT license. // See the LICENSE file in the project root for more information. /*XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX Compiler XX XX XX XX Represents the method data we are currently JIT-compiling. XX XX An instance of this class is created for every method we JIT. XX XX This contains all the info needed for the method. So allocating a XX XX a new instance per method makes it thread-safe. XX XX It should be used to do all the memory management for the compiler run. XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ /*****************************************************************************/ #ifndef _COMPILER_H_ #define _COMPILER_H_ /*****************************************************************************/ #include "jit.h" #include "opcode.h" #include "varset.h" #include "jitstd.h" #include "jithashtable.h" #include "gentree.h" #include "lir.h" #include "block.h" #include "inline.h" #include "jiteh.h" #include "instr.h" #include "regalloc.h" #include "sm.h" #include "cycletimer.h" #include "blockset.h" #include "arraystack.h" #include "hashbv.h" #include "jitexpandarray.h" #include "tinyarray.h" #include "valuenum.h" #include "reglist.h" #include "jittelemetry.h" #include "namedintrinsiclist.h" #ifdef LATE_DISASM #include "disasm.h" #endif #include "codegeninterface.h" #include "regset.h" #include "jitgcinfo.h" #if DUMP_GC_TABLES && defined(JIT32_GCENCODER) #include "gcdump.h" #endif #include "emit.h" #include "hwintrinsic.h" #include "simd.h" // This is only used locally in the JIT to indicate that // a verification block should be inserted #define SEH_VERIFICATION_EXCEPTION 0xe0564552 // VER /***************************************************************************** * Forward declarations */ struct InfoHdr; // defined in GCInfo.h struct escapeMapping_t; // defined in flowgraph.cpp class emitter; // defined in emit.h struct ShadowParamVarInfo; // defined in GSChecks.cpp struct InitVarDscInfo; // defined in register_arg_convention.h class FgStack; // defined in flowgraph.cpp #if FEATURE_ANYCSE class CSE_DataFlow; // defined in OptCSE.cpp #endif #ifdef DEBUG struct IndentStack; #endif class Lowering; // defined in lower.h // The following are defined in this file, Compiler.h class Compiler; /***************************************************************************** * Unwind info */ #include "unwind.h" /*****************************************************************************/ // // Declare global operator new overloads that use the compiler's arena allocator // // I wanted to make the second argument optional, with default = CMK_Unknown, but that // caused these to be ambiguous with the global placement new operators. void* __cdecl operator new(size_t n, Compiler* context, CompMemKind cmk); void* __cdecl operator new[](size_t n, Compiler* context, CompMemKind cmk); void* __cdecl operator new(size_t n, void* p, const jitstd::placement_t& syntax_difference); // Requires the definitions of "operator new" so including "LoopCloning.h" after the definitions. #include "loopcloning.h" /*****************************************************************************/ /* This is included here and not earlier as it needs the definition of "CSE" * which is defined in the section above */ /*****************************************************************************/ unsigned genLog2(unsigned value); unsigned genLog2(unsigned __int64 value); var_types genActualType(var_types type); var_types genUnsignedType(var_types type); var_types genSignedType(var_types type); unsigned ReinterpretHexAsDecimal(unsigned); /*****************************************************************************/ const unsigned FLG_CCTOR = (CORINFO_FLG_CONSTRUCTOR | CORINFO_FLG_STATIC); #ifdef DEBUG const int BAD_STK_OFFS = 0xBAADF00D; // for LclVarDsc::lvStkOffs #endif //------------------------------------------------------------------------ // HFA info shared by LclVarDsc and fgArgTabEntry //------------------------------------------------------------------------ #ifdef FEATURE_HFA enum HfaElemKind : unsigned int { HFA_ELEM_NONE, HFA_ELEM_FLOAT, HFA_ELEM_DOUBLE, HFA_ELEM_SIMD16 }; inline bool IsHfa(HfaElemKind kind) { return kind != HFA_ELEM_NONE; } inline var_types HfaTypeFromElemKind(HfaElemKind kind) { switch (kind) { case HFA_ELEM_FLOAT: return TYP_FLOAT; case HFA_ELEM_DOUBLE: return TYP_DOUBLE; #ifdef FEATURE_SIMD case HFA_ELEM_SIMD16: return TYP_SIMD16; #endif case HFA_ELEM_NONE: return TYP_UNDEF; default: assert(!"Invalid HfaElemKind"); return TYP_UNDEF; } } inline HfaElemKind HfaElemKindFromType(var_types type) { switch (type) { case TYP_FLOAT: return HFA_ELEM_FLOAT; case TYP_DOUBLE: return HFA_ELEM_DOUBLE; #ifdef FEATURE_SIMD case TYP_SIMD16: return HFA_ELEM_SIMD16; #endif case TYP_UNDEF: return HFA_ELEM_NONE; default: assert(!"Invalid HFA Type"); return HFA_ELEM_NONE; } } #endif // FEATURE_HFA // The following holds the Local var info (scope information) typedef const char* VarName; // Actual ASCII string struct VarScopeDsc { IL_OFFSET vsdLifeBeg; // instr offset of beg of life IL_OFFSET vsdLifeEnd; // instr offset of end of life unsigned vsdVarNum; // (remapped) LclVarDsc number #ifdef DEBUG VarName vsdName; // name of the var #endif unsigned vsdLVnum; // 'which' in eeGetLVinfo(). // Also, it is the index of this entry in the info.compVarScopes array, // which is useful since the array is also accessed via the // compEnterScopeList and compExitScopeList sorted arrays. }; // This is the location of a SSA definition. struct DefLoc { BasicBlock* m_blk; GenTree* m_tree; DefLoc() : m_blk(nullptr), m_tree(nullptr) { } DefLoc(BasicBlock* block, GenTree* tree) : m_blk(block), m_tree(tree) { } }; // This class stores information associated with a LclVar SSA definition. class LclSsaVarDsc { public: LclSsaVarDsc() { } LclSsaVarDsc(BasicBlock* block, GenTree* tree) : m_defLoc(block, tree) { } ValueNumPair m_vnPair; DefLoc m_defLoc; }; // This class stores information associated with a memory SSA definition. class SsaMemDef { public: ValueNumPair m_vnPair; }; //------------------------------------------------------------------------ // SsaDefArray: A resizable array of SSA definitions. // // Unlike an ordinary resizable array implementation, this allows only element // addition (by calling AllocSsaNum) and has special handling for RESERVED_SSA_NUM // (basically it's a 1-based array). The array doesn't impose any particular // requirements on the elements it stores and AllocSsaNum forwards its arguments // to the array element constructor, this way the array supports both LclSsaVarDsc // and SsaMemDef elements. // template class SsaDefArray { T* m_array; unsigned m_arraySize; unsigned m_count; static_assert_no_msg(SsaConfig::RESERVED_SSA_NUM == 0); static_assert_no_msg(SsaConfig::FIRST_SSA_NUM == 1); // Get the minimum valid SSA number. unsigned GetMinSsaNum() const { return SsaConfig::FIRST_SSA_NUM; } // Increase (double) the size of the array. void GrowArray(CompAllocator alloc) { unsigned oldSize = m_arraySize; unsigned newSize = max(2, oldSize * 2); T* newArray = alloc.allocate(newSize); for (unsigned i = 0; i < oldSize; i++) { newArray[i] = m_array[i]; } m_array = newArray; m_arraySize = newSize; } public: // Construct an empty SsaDefArray. SsaDefArray() : m_array(nullptr), m_arraySize(0), m_count(0) { } // Reset the array (used only if the SSA form is reconstructed). void Reset() { m_count = 0; } // Allocate a new SSA number (starting with SsaConfig::FIRST_SSA_NUM). template unsigned AllocSsaNum(CompAllocator alloc, Args&&... args) { if (m_count == m_arraySize) { GrowArray(alloc); } unsigned ssaNum = GetMinSsaNum() + m_count; m_array[m_count++] = T(jitstd::forward(args)...); // Ensure that the first SSA number we allocate is SsaConfig::FIRST_SSA_NUM assert((ssaNum == SsaConfig::FIRST_SSA_NUM) || (m_count > 1)); return ssaNum; } // Get the number of SSA definitions in the array. unsigned GetCount() const { return m_count; } // Get a pointer to the SSA definition at the specified index. T* GetSsaDefByIndex(unsigned index) { assert(index < m_count); return &m_array[index]; } // Check if the specified SSA number is valid. bool IsValidSsaNum(unsigned ssaNum) const { return (GetMinSsaNum() <= ssaNum) && (ssaNum < (GetMinSsaNum() + m_count)); } // Get a pointer to the SSA definition associated with the specified SSA number. T* GetSsaDef(unsigned ssaNum) { assert(ssaNum != SsaConfig::RESERVED_SSA_NUM); return GetSsaDefByIndex(ssaNum - GetMinSsaNum()); } }; enum RefCountState { RCS_INVALID, // not valid to get/set ref counts RCS_EARLY, // early counts for struct promotion and struct passing RCS_NORMAL, // normal ref counts (from lvaMarkRefs onward) }; class LclVarDsc { public: // The constructor. Most things can just be zero'ed. // // Initialize the ArgRegs to REG_STK. // Morph will update if this local is passed in a register. LclVarDsc() : _lvArgReg(REG_STK) , #if FEATURE_MULTIREG_ARGS _lvOtherArgReg(REG_STK) , #endif // FEATURE_MULTIREG_ARGS #if ASSERTION_PROP lvRefBlks(BlockSetOps::UninitVal()) , #endif // ASSERTION_PROP lvPerSsaData() { } // note this only packs because var_types is a typedef of unsigned char var_types lvType : 5; // TYP_INT/LONG/FLOAT/DOUBLE/REF unsigned char lvIsParam : 1; // is this a parameter? unsigned char lvIsRegArg : 1; // is this a register argument? unsigned char lvFramePointerBased : 1; // 0 = off of REG_SPBASE (e.g., ESP), 1 = off of REG_FPBASE (e.g., EBP) unsigned char lvStructGcCount : 3; // if struct, how many GC pointer (stop counting at 7). The only use of values >1 // is to help determine whether to use block init in the prolog. unsigned char lvOnFrame : 1; // (part of) the variable lives on the frame unsigned char lvRegister : 1; // assigned to live in a register? For RyuJIT backend, this is only set if the // variable is in the same register for the entire function. unsigned char lvTracked : 1; // is this a tracked variable? bool lvTrackedNonStruct() { return lvTracked && lvType != TYP_STRUCT; } unsigned char lvPinned : 1; // is this a pinned variable? unsigned char lvMustInit : 1; // must be initialized unsigned char lvAddrExposed : 1; // The address of this variable is "exposed" -- passed as an argument, stored in a // global location, etc. // We cannot reason reliably about the value of the variable. unsigned char lvDoNotEnregister : 1; // Do not enregister this variable. unsigned char lvFieldAccessed : 1; // The var is a struct local, and a field of the variable is accessed. Affects // struct promotion. unsigned char lvInSsa : 1; // The variable is in SSA form (set by SsaBuilder) #ifdef DEBUG // These further document the reasons for setting "lvDoNotEnregister". (Note that "lvAddrExposed" is one of the // reasons; // also, lvType == TYP_STRUCT prevents enregistration. At least one of the reasons should be true. unsigned char lvVMNeedsStackAddr : 1; // The VM may have access to a stack-relative address of the variable, and // read/write its value. unsigned char lvLiveInOutOfHndlr : 1; // The variable was live in or out of an exception handler, and this required // the variable to be // in the stack (at least at those boundaries.) unsigned char lvLclFieldExpr : 1; // The variable is not a struct, but was accessed like one (e.g., reading a // particular byte from an int). unsigned char lvLclBlockOpAddr : 1; // The variable was written to via a block operation that took its address. unsigned char lvLiveAcrossUCall : 1; // The variable is live across an unmanaged call. #endif unsigned char lvIsCSE : 1; // Indicates if this LclVar is a CSE variable. unsigned char lvHasLdAddrOp : 1; // has ldloca or ldarga opcode on this local. unsigned char lvStackByref : 1; // This is a compiler temporary of TYP_BYREF that is known to point into our local // stack frame. unsigned char lvHasILStoreOp : 1; // there is at least one STLOC or STARG on this local unsigned char lvHasMultipleILStoreOp : 1; // there is more than one STLOC on this local unsigned char lvIsTemp : 1; // Short-lifetime compiler temp #if defined(_TARGET_AMD64_) || defined(_TARGET_ARM64_) unsigned char lvIsImplicitByRef : 1; // Set if the argument is an implicit byref. #endif // defined(_TARGET_AMD64_) || defined(_TARGET_ARM64_) #if OPT_BOOL_OPS unsigned char lvIsBoolean : 1; // set if variable is boolean #endif unsigned char lvSingleDef : 1; // variable has a single def // before lvaMarkLocalVars: identifies ref type locals that can get type updates // after lvaMarkLocalVars: identifies locals that are suitable for optAddCopies #if ASSERTION_PROP unsigned char lvDisqualify : 1; // variable is no longer OK for add copy optimization unsigned char lvVolatileHint : 1; // hint for AssertionProp #endif #ifndef _TARGET_64BIT_ unsigned char lvStructDoubleAlign : 1; // Must we double align this struct? #endif // !_TARGET_64BIT_ #ifdef _TARGET_64BIT_ unsigned char lvQuirkToLong : 1; // Quirk to allocate this LclVar as a 64-bit long #endif #ifdef DEBUG unsigned char lvKeepType : 1; // Don't change the type of this variable unsigned char lvNoLclFldStress : 1; // Can't apply local field stress on this one #endif unsigned char lvIsPtr : 1; // Might this be used in an address computation? (used by buffer overflow security // checks) unsigned char lvIsUnsafeBuffer : 1; // Does this contain an unsafe buffer requiring buffer overflow security checks? unsigned char lvPromoted : 1; // True when this local is a promoted struct, a normed struct, or a "split" long on a // 32-bit target. For implicit byref parameters, this gets hijacked between // fgRetypeImplicitByRefArgs and fgMarkDemotedImplicitByRefArgs to indicate whether // references to the arg are being rewritten as references to a promoted shadow local. unsigned char lvIsStructField : 1; // Is this local var a field of a promoted struct local? unsigned char lvOverlappingFields : 1; // True when we have a struct with possibly overlapping fields unsigned char lvContainsHoles : 1; // True when we have a promoted struct that contains holes unsigned char lvCustomLayout : 1; // True when this struct has "CustomLayout" unsigned char lvIsMultiRegArg : 1; // true if this is a multireg LclVar struct used in an argument context unsigned char lvIsMultiRegRet : 1; // true if this is a multireg LclVar struct assigned from a multireg call #ifdef FEATURE_HFA HfaElemKind _lvHfaElemKind : 2; // What kind of an HFA this is (HFA_ELEM_NONE if it is not an HFA). #endif // FEATURE_HFA #ifdef DEBUG // TODO-Cleanup: See the note on lvSize() - this flag is only in use by asserts that are checking for struct // types, and is needed because of cases where TYP_STRUCT is bashed to an integral type. // Consider cleaning this up so this workaround is not required. unsigned char lvUnusedStruct : 1; // All references to this promoted struct are through its field locals. // I.e. there is no longer any reference to the struct directly. // In this case we can simply remove this struct local. #endif unsigned char lvLRACandidate : 1; // Tracked for linear scan register allocation purposes #ifdef FEATURE_SIMD // Note that both SIMD vector args and locals are marked as lvSIMDType = true, but the // type of an arg node is TYP_BYREF and a local node is TYP_SIMD*. unsigned char lvSIMDType : 1; // This is a SIMD struct unsigned char lvUsedInSIMDIntrinsic : 1; // This tells lclvar is used for simd intrinsic var_types lvBaseType : 5; // Note: this only packs because var_types is a typedef of unsigned char #endif // FEATURE_SIMD unsigned char lvRegStruct : 1; // This is a reg-sized non-field-addressed struct. unsigned char lvClassIsExact : 1; // lvClassHandle is the exact type #ifdef DEBUG unsigned char lvClassInfoUpdated : 1; // true if this var has updated class handle or exactness #endif unsigned char lvImplicitlyReferenced : 1; // true if there are non-IR references to this local (prolog, epilog, gc, // eh) union { unsigned lvFieldLclStart; // The index of the local var representing the first field in the promoted struct // local. For implicit byref parameters, this gets hijacked between // fgRetypeImplicitByRefArgs and fgMarkDemotedImplicitByRefArgs to point to the // struct local created to model the parameter's struct promotion, if any. unsigned lvParentLcl; // The index of the local var representing the parent (i.e. the promoted struct local). // Valid on promoted struct local fields. }; unsigned char lvFieldCnt; // Number of fields in the promoted VarDsc. unsigned char lvFldOffset; unsigned char lvFldOrdinal; #if FEATURE_MULTIREG_ARGS regNumber lvRegNumForSlot(unsigned slotNum) { if (slotNum == 0) { return lvArgReg; } else if (slotNum == 1) { return lvOtherArgReg; } else { assert(false && "Invalid slotNum!"); } unreached(); } #endif // FEATURE_MULTIREG_ARGS bool lvIsHfa() const { #ifdef FEATURE_HFA return IsHfa(_lvHfaElemKind); #else return false; #endif } bool lvIsHfaRegArg() const { #ifdef FEATURE_HFA return lvIsRegArg && lvIsHfa(); #else return false; #endif } //------------------------------------------------------------------------------ // lvHfaSlots: Get the number of slots used by an HFA local // // Return Value: // On Arm64 - Returns 1-4 indicating the number of register slots used by the HFA // On Arm32 - Returns the total number of single FP register slots used by the HFA, max is 8 // unsigned lvHfaSlots() const { assert(lvIsHfa()); assert(varTypeIsStruct(lvType)); unsigned slots = 0; #ifdef _TARGET_ARM_ slots = lvExactSize / sizeof(float); assert(slots <= 8); #elif defined(_TARGET_ARM64_) switch (_lvHfaElemKind) { case HFA_ELEM_NONE: assert(!"lvHfaSlots called for non-HFA"); break; case HFA_ELEM_FLOAT: assert((lvExactSize % 4) == 0); slots = lvExactSize >> 2; break; case HFA_ELEM_DOUBLE: assert((lvExactSize % 8) == 0); slots = lvExactSize >> 3; break; case HFA_ELEM_SIMD16: assert((lvExactSize % 16) == 0); slots = lvExactSize >> 4; break; default: unreached(); } assert(slots <= 4); #endif // _TARGET_ARM64_ return slots; } // lvIsMultiRegArgOrRet() // returns true if this is a multireg LclVar struct used in an argument context // or if this is a multireg LclVar struct assigned from a multireg call bool lvIsMultiRegArgOrRet() { return lvIsMultiRegArg || lvIsMultiRegRet; } private: regNumberSmall _lvRegNum; // Used to store the register this variable is in (or, the low register of a // register pair). It is set during codegen any time the // variable is enregistered (lvRegister is only set // to non-zero if the variable gets the same register assignment for its entire // lifetime). #if !defined(_TARGET_64BIT_) regNumberSmall _lvOtherReg; // Used for "upper half" of long var. #endif // !defined(_TARGET_64BIT_) regNumberSmall _lvArgReg; // The (first) register in which this argument is passed. #if FEATURE_MULTIREG_ARGS regNumberSmall _lvOtherArgReg; // Used for the second part of the struct passed in a register. // Note this is defined but not used by ARM32 #endif // FEATURE_MULTIREG_ARGS regNumberSmall _lvArgInitReg; // the register into which the argument is moved at entry public: // The register number is stored in a small format (8 bits), but the getters return and the setters take // a full-size (unsigned) format, to localize the casts here. ///////////////////// __declspec(property(get = GetRegNum, put = SetRegNum)) regNumber lvRegNum; regNumber GetRegNum() const { return (regNumber)_lvRegNum; } void SetRegNum(regNumber reg) { _lvRegNum = (regNumberSmall)reg; assert(_lvRegNum == reg); } ///////////////////// #if defined(_TARGET_64BIT_) __declspec(property(get = GetOtherReg, put = SetOtherReg)) regNumber lvOtherReg; regNumber GetOtherReg() const { assert(!"shouldn't get here"); // can't use "unreached();" because it's NORETURN, which causes C4072 // "unreachable code" warnings return REG_NA; } void SetOtherReg(regNumber reg) { assert(!"shouldn't get here"); // can't use "unreached();" because it's NORETURN, which causes C4072 // "unreachable code" warnings } #else // !_TARGET_64BIT_ __declspec(property(get = GetOtherReg, put = SetOtherReg)) regNumber lvOtherReg; regNumber GetOtherReg() const { return (regNumber)_lvOtherReg; } void SetOtherReg(regNumber reg) { _lvOtherReg = (regNumberSmall)reg; assert(_lvOtherReg == reg); } #endif // !_TARGET_64BIT_ ///////////////////// __declspec(property(get = GetArgReg, put = SetArgReg)) regNumber lvArgReg; regNumber GetArgReg() const { return (regNumber)_lvArgReg; } void SetArgReg(regNumber reg) { _lvArgReg = (regNumberSmall)reg; assert(_lvArgReg == reg); } #if FEATURE_MULTIREG_ARGS __declspec(property(get = GetOtherArgReg, put = SetOtherArgReg)) regNumber lvOtherArgReg; regNumber GetOtherArgReg() const { return (regNumber)_lvOtherArgReg; } void SetOtherArgReg(regNumber reg) { _lvOtherArgReg = (regNumberSmall)reg; assert(_lvOtherArgReg == reg); } #endif // FEATURE_MULTIREG_ARGS #ifdef FEATURE_SIMD // Is this is a SIMD struct? bool lvIsSIMDType() const { return lvSIMDType; } // Is this is a SIMD struct which is used for SIMD intrinsic? bool lvIsUsedInSIMDIntrinsic() const { return lvUsedInSIMDIntrinsic; } #else // If feature_simd not enabled, return false bool lvIsSIMDType() const { return false; } bool lvIsUsedInSIMDIntrinsic() const { return false; } #endif ///////////////////// __declspec(property(get = GetArgInitReg, put = SetArgInitReg)) regNumber lvArgInitReg; regNumber GetArgInitReg() const { return (regNumber)_lvArgInitReg; } void SetArgInitReg(regNumber reg) { _lvArgInitReg = (regNumberSmall)reg; assert(_lvArgInitReg == reg); } ///////////////////// bool lvIsRegCandidate() const { return lvLRACandidate != 0; } bool lvIsInReg() const { return lvIsRegCandidate() && (lvRegNum != REG_STK); } regMaskTP lvRegMask() const { regMaskTP regMask = RBM_NONE; if (varTypeIsFloating(TypeGet())) { if (lvRegNum != REG_STK) { regMask = genRegMaskFloat(lvRegNum, TypeGet()); } } else { if (lvRegNum != REG_STK) { regMask = genRegMask(lvRegNum); } } return regMask; } unsigned short lvVarIndex; // variable tracking index private: unsigned short m_lvRefCnt; // unweighted (real) reference count. For implicit by reference // parameters, this gets hijacked from fgResetImplicitByRefRefCount // through fgMarkDemotedImplicitByRefArgs, to provide a static // appearance count (computed during address-exposed analysis) // that fgMakeOutgoingStructArgCopy consults during global morph // to determine if eliding its copy is legal. BasicBlock::weight_t m_lvRefCntWtd; // weighted reference count public: unsigned short lvRefCnt(RefCountState state = RCS_NORMAL) const; void incLvRefCnt(unsigned short delta, RefCountState state = RCS_NORMAL); void setLvRefCnt(unsigned short newValue, RefCountState state = RCS_NORMAL); BasicBlock::weight_t lvRefCntWtd(RefCountState state = RCS_NORMAL) const; void incLvRefCntWtd(BasicBlock::weight_t delta, RefCountState state = RCS_NORMAL); void setLvRefCntWtd(BasicBlock::weight_t newValue, RefCountState state = RCS_NORMAL); int lvStkOffs; // stack offset of home unsigned lvExactSize; // (exact) size of the type in bytes // Is this a promoted struct? // This method returns true only for structs (including SIMD structs), not for // locals that are split on a 32-bit target. // It is only necessary to use this: // 1) if only structs are wanted, and // 2) if Lowering has already been done. // Otherwise lvPromoted is valid. bool lvPromotedStruct() { #if !defined(_TARGET_64BIT_) return (lvPromoted && !varTypeIsLong(lvType)); #else // defined(_TARGET_64BIT_) return lvPromoted; #endif // defined(_TARGET_64BIT_) } unsigned lvSize() const // Size needed for storage representation. Only used for structs or TYP_BLK. { // TODO-Review: Sometimes we get called on ARM with HFA struct variables that have been promoted, // where the struct itself is no longer used because all access is via its member fields. // When that happens, the struct is marked as unused and its type has been changed to // TYP_INT (to keep the GC tracking code from looking at it). // See Compiler::raAssignVars() for details. For example: // N002 ( 4, 3) [00EA067C] ------------- return struct $346 // N001 ( 3, 2) [00EA0628] ------------- lclVar struct(U) V03 loc2 // float V03.f1 (offs=0x00) -> V12 tmp7 // f8 (last use) (last use) $345 // Here, the "struct(U)" shows that the "V03 loc2" variable is unused. Not shown is that V03 // is now TYP_INT in the local variable table. It's not really unused, because it's in the tree. assert(varTypeIsStruct(lvType) || (lvType == TYP_BLK) || (lvPromoted && lvUnusedStruct)); #if defined(FEATURE_SIMD) && !defined(_TARGET_64BIT_) // For 32-bit architectures, we make local variable SIMD12 types 16 bytes instead of just 12. We can't do // this for arguments, which must be passed according the defined ABI. We don't want to do this for // dependently promoted struct fields, but we don't know that here. See lvaMapSimd12ToSimd16(). // (Note that for 64-bits, we are already rounding up to 16.) if ((lvType == TYP_SIMD12) && !lvIsParam) { assert(lvExactSize == 12); return 16; } #endif // defined(FEATURE_SIMD) && !defined(_TARGET_64BIT_) return roundUp(lvExactSize, TARGET_POINTER_SIZE); } size_t lvArgStackSize() const; unsigned lvSlotNum; // original slot # (if remapped) typeInfo lvVerTypeInfo; // type info needed for verification CORINFO_CLASS_HANDLE lvClassHnd; // class handle for the local, or null if not known CORINFO_FIELD_HANDLE lvFieldHnd; // field handle for promoted struct fields BYTE* lvGcLayout; // GC layout info for structs #if ASSERTION_PROP BlockSet lvRefBlks; // Set of blocks that contain refs GenTreeStmt* lvDefStmt; // Pointer to the statement with the single definition void lvaDisqualifyVar(); // Call to disqualify a local variable from use in optAddCopies #endif var_types TypeGet() const { return (var_types)lvType; } bool lvStackAligned() const { assert(lvIsStructField); return ((lvFldOffset % TARGET_POINTER_SIZE) == 0); } bool lvNormalizeOnLoad() const { return varTypeIsSmall(TypeGet()) && // lvIsStructField is treated the same as the aliased local, see fgDoNormalizeOnStore. (lvIsParam || lvAddrExposed || lvIsStructField); } bool lvNormalizeOnStore() { return varTypeIsSmall(TypeGet()) && // lvIsStructField is treated the same as the aliased local, see fgDoNormalizeOnStore. !(lvIsParam || lvAddrExposed || lvIsStructField); } void incRefCnts(BasicBlock::weight_t weight, Compiler* pComp, RefCountState state = RCS_NORMAL, bool propagate = true); bool IsFloatRegType() const { return isFloatRegType(lvType) || lvIsHfaRegArg(); } var_types GetHfaType() const { #ifdef FEATURE_HFA assert(lvIsHfa()); return HfaTypeFromElemKind(_lvHfaElemKind); #else return TYP_UNDEF; #endif // FEATURE_HFA } void SetHfaType(var_types type) { #ifdef FEATURE_HFA _lvHfaElemKind = HfaElemKindFromType(type); #endif // FEATURE_HFA } var_types lvaArgType(); SsaDefArray lvPerSsaData; // Returns the address of the per-Ssa data for the given ssaNum (which is required // not to be the SsaConfig::RESERVED_SSA_NUM, which indicates that the variable is // not an SSA variable). LclSsaVarDsc* GetPerSsaData(unsigned ssaNum) { return lvPerSsaData.GetSsaDef(ssaNum); } #ifdef DEBUG public: const char* lvReason; void PrintVarReg() const { printf("%s", getRegName(lvRegNum)); } #endif // DEBUG }; // class LclVarDsc /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX TempsInfo XX XX XX XX The temporary lclVars allocated by the compiler for code generation XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ /***************************************************************************** * * The following keeps track of temporaries allocated in the stack frame * during code-generation (after register allocation). These spill-temps are * only used if we run out of registers while evaluating a tree. * * These are different from the more common temps allocated by lvaGrabTemp(). */ class TempDsc { public: TempDsc* tdNext; private: int tdOffs; #ifdef DEBUG static const int BAD_TEMP_OFFSET = 0xDDDDDDDD; // used as a sentinel "bad value" for tdOffs in DEBUG #endif // DEBUG int tdNum; BYTE tdSize; var_types tdType; public: TempDsc(int _tdNum, unsigned _tdSize, var_types _tdType) : tdNum(_tdNum), tdSize((BYTE)_tdSize), tdType(_tdType) { #ifdef DEBUG assert(tdNum < 0); // temps must have a negative number (so they have a different number from all local variables) tdOffs = BAD_TEMP_OFFSET; #endif // DEBUG if (tdNum != _tdNum) { IMPL_LIMITATION("too many spill temps"); } } #ifdef DEBUG bool tdLegalOffset() const { return tdOffs != BAD_TEMP_OFFSET; } #endif // DEBUG int tdTempOffs() const { assert(tdLegalOffset()); return tdOffs; } void tdSetTempOffs(int offs) { tdOffs = offs; assert(tdLegalOffset()); } void tdAdjustTempOffs(int offs) { tdOffs += offs; assert(tdLegalOffset()); } int tdTempNum() const { assert(tdNum < 0); return tdNum; } unsigned tdTempSize() const { return tdSize; } var_types tdTempType() const { return tdType; } }; // interface to hide linearscan implementation from rest of compiler class LinearScanInterface { public: virtual void doLinearScan() = 0; virtual void recordVarLocationsAtStartOfBB(BasicBlock* bb) = 0; virtual bool willEnregisterLocalVars() const = 0; }; LinearScanInterface* getLinearScanAllocator(Compiler* comp); // Information about arrays: their element type and size, and the offset of the first element. // We label GT_IND's that are array indices with GTF_IND_ARR_INDEX, and, for such nodes, // associate an array info via the map retrieved by GetArrayInfoMap(). This information is used, // for example, in value numbering of array index expressions. struct ArrayInfo { var_types m_elemType; CORINFO_CLASS_HANDLE m_elemStructType; unsigned m_elemSize; unsigned m_elemOffset; ArrayInfo() : m_elemType(TYP_UNDEF), m_elemStructType(nullptr), m_elemSize(0), m_elemOffset(0) { } ArrayInfo(var_types elemType, unsigned elemSize, unsigned elemOffset, CORINFO_CLASS_HANDLE elemStructType) : m_elemType(elemType), m_elemStructType(elemStructType), m_elemSize(elemSize), m_elemOffset(elemOffset) { } }; // This enumeration names the phases into which we divide compilation. The phases should completely // partition a compilation. enum Phases { #define CompPhaseNameMacro(enum_nm, string_nm, short_nm, hasChildren, parent, measureIR) enum_nm, #include "compphases.h" PHASE_NUMBER_OF }; extern const char* PhaseNames[]; extern const char* PhaseEnums[]; extern const LPCWSTR PhaseShortNames[]; // The following enum provides a simple 1:1 mapping to CLR API's enum API_ICorJitInfo_Names { #define DEF_CLR_API(name) API_##name, #include "ICorJitInfo_API_names.h" API_COUNT }; //--------------------------------------------------------------- // Compilation time. // // A "CompTimeInfo" is a structure for tracking the compilation time of one or more methods. // We divide a compilation into a sequence of contiguous phases, and track the total (per-thread) cycles // of the compilation, as well as the cycles for each phase. We also track the number of bytecodes. // If there is a failure in reading a timer at any point, the "CompTimeInfo" becomes invalid, as indicated // by "m_timerFailure" being true. // If FEATURE_JIT_METHOD_PERF is not set, we define a minimal form of this, enough to let other code compile. struct CompTimeInfo { #ifdef FEATURE_JIT_METHOD_PERF // The string names of the phases. static const char* PhaseNames[]; static bool PhaseHasChildren[]; static int PhaseParent[]; static bool PhaseReportsIRSize[]; unsigned m_byteCodeBytes; unsigned __int64 m_totalCycles; unsigned __int64 m_invokesByPhase[PHASE_NUMBER_OF]; unsigned __int64 m_cyclesByPhase[PHASE_NUMBER_OF]; #if MEASURE_CLRAPI_CALLS unsigned __int64 m_CLRinvokesByPhase[PHASE_NUMBER_OF]; unsigned __int64 m_CLRcyclesByPhase[PHASE_NUMBER_OF]; #endif unsigned m_nodeCountAfterPhase[PHASE_NUMBER_OF]; // For better documentation, we call EndPhase on // non-leaf phases. We should also call EndPhase on the // last leaf subphase; obviously, the elapsed cycles between the EndPhase // for the last leaf subphase and the EndPhase for an ancestor should be very small. // We add all such "redundant end phase" intervals to this variable below; we print // it out in a report, so we can verify that it is, indeed, very small. If it ever // isn't, this means that we're doing something significant between the end of the last // declared subphase and the end of its parent. unsigned __int64 m_parentPhaseEndSlop; bool m_timerFailure; #if MEASURE_CLRAPI_CALLS // The following measures the time spent inside each individual CLR API call. unsigned m_allClrAPIcalls; unsigned m_perClrAPIcalls[API_ICorJitInfo_Names::API_COUNT]; unsigned __int64 m_allClrAPIcycles; unsigned __int64 m_perClrAPIcycles[API_ICorJitInfo_Names::API_COUNT]; unsigned __int32 m_maxClrAPIcycles[API_ICorJitInfo_Names::API_COUNT]; #endif // MEASURE_CLRAPI_CALLS CompTimeInfo(unsigned byteCodeBytes); #endif }; #ifdef FEATURE_JIT_METHOD_PERF #if MEASURE_CLRAPI_CALLS struct WrapICorJitInfo; #endif // This class summarizes the JIT time information over the course of a run: the number of methods compiled, // and the total and maximum timings. (These are instances of the "CompTimeInfo" type described above). // The operation of adding a single method's timing to the summary may be performed concurrently by several // threads, so it is protected by a lock. // This class is intended to be used as a singleton type, with only a single instance. class CompTimeSummaryInfo { // This lock protects the fields of all CompTimeSummaryInfo(s) (of which we expect there to be one). static CritSecObject s_compTimeSummaryLock; int m_numMethods; int m_totMethods; CompTimeInfo m_total; CompTimeInfo m_maximum; int m_numFilteredMethods; CompTimeInfo m_filtered; // This can use what ever data you want to determine if the value to be added // belongs in the filtered section (it's always included in the unfiltered section) bool IncludedInFilteredData(CompTimeInfo& info); public: // This is the unique CompTimeSummaryInfo object for this instance of the runtime. static CompTimeSummaryInfo s_compTimeSummary; CompTimeSummaryInfo() : m_numMethods(0), m_totMethods(0), m_total(0), m_maximum(0), m_numFilteredMethods(0), m_filtered(0) { } // Assumes that "info" is a completed CompTimeInfo for a compilation; adds it to the summary. // This is thread safe. void AddInfo(CompTimeInfo& info, bool includePhases); // Print the summary information to "f". // This is not thread-safe; assumed to be called by only one thread. void Print(FILE* f); }; // A JitTimer encapsulates a CompTimeInfo for a single compilation. It also tracks the start of compilation, // and when the current phase started. This is intended to be part of a Compilation object. This is // disabled (FEATURE_JIT_METHOD_PERF not defined) when FEATURE_CORECLR is set, or on non-windows platforms. // class JitTimer { unsigned __int64 m_start; // Start of the compilation. unsigned __int64 m_curPhaseStart; // Start of the current phase. #if MEASURE_CLRAPI_CALLS unsigned __int64 m_CLRcallStart; // Start of the current CLR API call (if any). unsigned __int64 m_CLRcallInvokes; // CLR API invokes under current outer so far unsigned __int64 m_CLRcallCycles; // CLR API cycles under current outer so far. int m_CLRcallAPInum; // The enum/index of the current CLR API call (or -1). static double s_cyclesPerSec; // Cached for speedier measurements #endif #ifdef DEBUG Phases m_lastPhase; // The last phase that was completed (or (Phases)-1 to start). #endif CompTimeInfo m_info; // The CompTimeInfo for this compilation. static CritSecObject s_csvLock; // Lock to protect the time log file. void PrintCsvMethodStats(Compiler* comp); private: void* operator new(size_t); void* operator new[](size_t); void operator delete(void*); void operator delete[](void*); public: // Initialized the timer instance JitTimer(unsigned byteCodeSize); static JitTimer* Create(Compiler* comp, unsigned byteCodeSize) { return ::new (comp, CMK_Unknown) JitTimer(byteCodeSize); } static void PrintCsvHeader(); // Ends the current phase (argument is for a redundant check). void EndPhase(Compiler* compiler, Phases phase); #if MEASURE_CLRAPI_CALLS // Start and end a timed CLR API call. void CLRApiCallEnter(unsigned apix); void CLRApiCallLeave(unsigned apix); #endif // MEASURE_CLRAPI_CALLS // Completes the timing of the current method, which is assumed to have "byteCodeBytes" bytes of bytecode, // and adds it to "sum". void Terminate(Compiler* comp, CompTimeSummaryInfo& sum, bool includePhases); // Attempts to query the cycle counter of the current thread. If successful, returns "true" and sets // *cycles to the cycle counter value. Otherwise, returns false and sets the "m_timerFailure" flag of // "m_info" to true. bool GetThreadCycles(unsigned __int64* cycles) { bool res = CycleTimer::GetThreadCyclesS(cycles); if (!res) { m_info.m_timerFailure = true; } return res; } }; #endif // FEATURE_JIT_METHOD_PERF //------------------- Function/Funclet info ------------------------------- enum FuncKind : BYTE { FUNC_ROOT, // The main/root function (always id==0) FUNC_HANDLER, // a funclet associated with an EH handler (finally, fault, catch, filter handler) FUNC_FILTER, // a funclet associated with an EH filter FUNC_COUNT }; class emitLocation; struct FuncInfoDsc { FuncKind funKind; BYTE funFlags; // Currently unused, just here for padding unsigned short funEHIndex; // index, into the ebd table, of innermost EH clause corresponding to this // funclet. It is only valid if funKind field indicates this is a // EH-related funclet: FUNC_HANDLER or FUNC_FILTER #if defined(_TARGET_AMD64_) // TODO-AMD64-Throughput: make the AMD64 info more like the ARM info to avoid having this large static array. emitLocation* startLoc; emitLocation* endLoc; emitLocation* coldStartLoc; // locations for the cold section, if there is one. emitLocation* coldEndLoc; UNWIND_INFO unwindHeader; // Maximum of 255 UNWIND_CODE 'nodes' and then the unwind header. If there are an odd // number of codes, the VM or Zapper will 4-byte align the whole thing. BYTE unwindCodes[offsetof(UNWIND_INFO, UnwindCode) + (0xFF * sizeof(UNWIND_CODE))]; unsigned unwindCodeSlot; #elif defined(_TARGET_X86_) #if defined(_TARGET_UNIX_) emitLocation* startLoc; emitLocation* endLoc; emitLocation* coldStartLoc; // locations for the cold section, if there is one. emitLocation* coldEndLoc; #endif // _TARGET_UNIX_ #elif defined(_TARGET_ARMARCH_) UnwindInfo uwi; // Unwind information for this function/funclet's hot section UnwindInfo* uwiCold; // Unwind information for this function/funclet's cold section // Note: we only have a pointer here instead of the actual object, // to save memory in the JIT case (compared to the NGEN case), // where we don't have any cold section. // Note 2: we currently don't support hot/cold splitting in functions // with EH, so uwiCold will be NULL for all funclets. #if defined(_TARGET_UNIX_) emitLocation* startLoc; emitLocation* endLoc; emitLocation* coldStartLoc; // locations for the cold section, if there is one. emitLocation* coldEndLoc; #endif // _TARGET_UNIX_ #endif // _TARGET_ARMARCH_ #if defined(_TARGET_UNIX_) jitstd::vector* cfiCodes; #endif // _TARGET_UNIX_ // Eventually we may want to move rsModifiedRegsMask, lvaOutgoingArgSize, and anything else // that isn't shared between the main function body and funclets. }; struct fgArgTabEntry { GenTree* node; // Initially points at the Op1 field of 'parent', but if the argument is replaced with an GT_ASG or // placeholder it will point at the actual argument in the gtCallLateArgs list. GenTree* parent; // Points at the GT_LIST node in the gtCallArgs for this argument unsigned argNum; // The original argument number, also specifies the required argument evaluation order from the IL private: regNumberSmall regNums[MAX_ARG_REG_COUNT]; // The registers to use when passing this argument, set to REG_STK for // arguments passed on the stack public: unsigned numRegs; // Count of number of registers that this argument uses. // Note that on ARM, if we have a double hfa, this reflects the number // of DOUBLE registers. // A slot is a pointer sized region in the OutArg area. unsigned slotNum; // When an argument is passed in the OutArg area this is the slot number in the OutArg area unsigned numSlots; // Count of number of slots that this argument uses unsigned alignment; // 1 or 2 (slots/registers) private: unsigned _lateArgInx; // index into gtCallLateArgs list; UINT_MAX if this is not a late arg. public: unsigned tmpNum; // the LclVar number if we had to force evaluation of this arg var_types argType; // The type used to pass this argument. This is generally the original argument type, but when a // struct is passed as a scalar type, this is that type. // Note that if a struct is passed by reference, this will still be the struct type. bool needTmp : 1; // True when we force this argument's evaluation into a temp LclVar bool needPlace : 1; // True when we must replace this argument with a placeholder node bool isTmp : 1; // True when we setup a temp LclVar for this argument due to size issues with the struct bool processed : 1; // True when we have decided the evaluation order for this argument in the gtCallLateArgs bool isBackFilled : 1; // True when the argument fills a register slot skipped due to alignment requirements of // previous arguments. bool isNonStandard : 1; // True if it is an arg that is passed in a reg other than a standard arg reg, or is forced // to be on the stack despite its arg list position. bool isStruct : 1; // True if this is a struct arg bool _isVararg : 1; // True if the argument is in a vararg context. bool passedByRef : 1; // True iff the argument is passed by reference. #ifdef FEATURE_ARG_SPLIT bool _isSplit : 1; // True when this argument is split between the registers and OutArg area #endif // FEATURE_ARG_SPLIT #ifdef FEATURE_HFA HfaElemKind _hfaElemKind : 2; // What kind of an HFA this is (HFA_ELEM_NONE if it is not an HFA). #endif bool isLateArg() { bool isLate = (_lateArgInx != UINT_MAX); return isLate; } __declspec(property(get = getLateArgInx, put = setLateArgInx)) unsigned lateArgInx; unsigned getLateArgInx() { assert(isLateArg()); return _lateArgInx; } void setLateArgInx(unsigned inx) { _lateArgInx = inx; } __declspec(property(get = getRegNum)) regNumber regNum; regNumber getRegNum() { return (regNumber)regNums[0]; } __declspec(property(get = getOtherRegNum)) regNumber otherRegNum; regNumber getOtherRegNum() { return (regNumber)regNums[1]; } #if defined(UNIX_AMD64_ABI) SYSTEMV_AMD64_CORINFO_STRUCT_REG_PASSING_DESCRIPTOR structDesc; #endif void setRegNum(unsigned int i, regNumber regNum) { assert(i < MAX_ARG_REG_COUNT); regNums[i] = (regNumberSmall)regNum; } regNumber getRegNum(unsigned int i) { assert(i < MAX_ARG_REG_COUNT); return (regNumber)regNums[i]; } __declspec(property(get = getIsSplit, put = setIsSplit)) bool isSplit; bool getIsSplit() { #ifdef FEATURE_ARG_SPLIT return _isSplit; #else // FEATURE_ARG_SPLIT return false; #endif } void setIsSplit(bool value) { #ifdef FEATURE_ARG_SPLIT _isSplit = value; #endif } __declspec(property(get = getIsVararg, put = setIsVararg)) bool isVararg; bool getIsVararg() { #ifdef FEATURE_VARARG return _isVararg; #else return false; #endif } void setIsVararg(bool value) { #ifdef FEATURE_VARARG _isVararg = value; #endif // FEATURE_VARARG } __declspec(property(get = getIsHfaArg)) bool isHfaArg; bool getIsHfaArg() { #ifdef FEATURE_HFA return IsHfa(_hfaElemKind); #else return false; #endif } __declspec(property(get = getIsHfaRegArg)) bool isHfaRegArg; bool getIsHfaRegArg() { #ifdef FEATURE_HFA return IsHfa(_hfaElemKind) && isPassedInRegisters(); #else return false; #endif } __declspec(property(get = GetHfaType)) var_types hfaType; var_types GetHfaType() { #ifdef FEATURE_HFA return HfaTypeFromElemKind(_hfaElemKind); #else return TYP_UNDEF; #endif // FEATURE_HFA } void SetHfaType(var_types type, unsigned hfaSlots) { #ifdef FEATURE_HFA if (type != TYP_UNDEF) { // We must already have set the passing mode. assert(numRegs != 0 || numSlots != 0); // We originally set numRegs according to the size of the struct, but if the size of the // hfaType is not the same as the pointer size, we need to correct it. // Note that hfaSlots is the number of registers we will use. For ARM, that is twice // the number of "double registers". unsigned numHfaRegs = hfaSlots; #ifdef _TARGET_ARM_ if (type == TYP_DOUBLE) { // Must be an even number of registers. assert((numRegs & 1) == 0); numHfaRegs = hfaSlots / 2; } #endif // _TARGET_ARM_ if (!isHfaArg) { // We haven't previously set this; do so now. _hfaElemKind = HfaElemKindFromType(type); if (isPassedInRegisters()) { numRegs = numHfaRegs; } } else { // We've already set this; ensure that it's consistent. if (isPassedInRegisters()) { assert(numRegs == numHfaRegs); } assert(type == HfaTypeFromElemKind(_hfaElemKind)); } } #endif // FEATURE_HFA } #ifdef _TARGET_ARM_ void SetIsBackFilled(bool backFilled) { isBackFilled = backFilled; } bool IsBackFilled() const { return isBackFilled; } #else // !_TARGET_ARM_ void SetIsBackFilled(bool backFilled) { } bool IsBackFilled() const { return false; } #endif // !_TARGET_ARM_ bool isPassedInRegisters() { return !isSplit && (numRegs != 0); } bool isPassedInFloatRegisters() { #ifdef _TARGET_X86 return false; #else return isValidFloatArgReg(regNum); #endif } bool isSingleRegOrSlot() { return !isSplit && ((numRegs == 1) || (numSlots == 1)); } // Returns the number of "slots" used, where for this purpose a // register counts as a slot. unsigned getSlotCount() { if (isBackFilled) { assert(isPassedInRegisters()); assert(numRegs == 1); } else if (regNum == REG_STK) { assert(!isPassedInRegisters()); assert(numRegs == 0); } else { assert(numRegs > 0); } return numSlots + numRegs; } // Returns the size as a multiple of pointer-size. // For targets without HFAs, this is the same as getSlotCount(). unsigned getSize() { unsigned size = getSlotCount(); #ifdef FEATURE_HFA if (isHfaRegArg) { #ifdef _TARGET_ARM_ // We counted the number of regs, but if they are DOUBLE hfa regs we have to double the size. if (hfaType == TYP_DOUBLE) { assert(!isSplit); size <<= 1; } #elif defined(_TARGET_ARM64_) // We counted the number of regs, but if they are FLOAT hfa regs we have to halve the size, // or if they are SIMD16 vector hfa regs we have to double the size. if (hfaType == TYP_FLOAT) { // Round up in case of odd HFA count. size = (size + 1) >> 1; } #ifdef FEATURE_SIMD else if (hfaType == TYP_SIMD16) { size <<= 1; } #endif // FEATURE_SIMD #endif // _TARGET_ARM64_ } #endif // FEATURE_HFA return size; } // Set the register numbers for a multireg argument. // There's nothing to do on x64/Ux because the structDesc has already been used to set the // register numbers. void SetMultiRegNums() { #if FEATURE_MULTIREG_ARGS && !defined(UNIX_AMD64_ABI) if (numRegs == 1) { return; } regNumber argReg = getRegNum(0); #ifdef _TARGET_ARM_ unsigned int regSize = (hfaType == TYP_DOUBLE) ? 2 : 1; #else unsigned int regSize = 1; #endif for (unsigned int regIndex = 1; regIndex < numRegs; regIndex++) { argReg = (regNumber)(argReg + regSize); setRegNum(regIndex, argReg); } #endif // FEATURE_MULTIREG_ARGS && !defined(UNIX_AMD64_ABI) } // Check that the value of 'isStruct' is consistent. // A struct arg must be one of the following: // - A node of struct type, // - A GT_FIELD_LIST, or // - A node of a scalar type, passed in a single register or slot // (or two slots in the case of a struct pass on the stack as TYP_DOUBLE). // void checkIsStruct() { if (isStruct) { if (!varTypeIsStruct(node) && !node->OperIs(GT_FIELD_LIST)) { // This is the case where we are passing a struct as a primitive type. // On most targets, this is always a single register or slot. // However, on ARM this could be two slots if it is TYP_DOUBLE. bool isPassedAsPrimitiveType = ((numRegs == 1) || ((numRegs == 0) && (numSlots == 1))); #ifdef _TARGET_ARM_ if (!isPassedAsPrimitiveType) { if (node->TypeGet() == TYP_DOUBLE && numRegs == 0 && (numSlots == 2)) { isPassedAsPrimitiveType = true; } } #endif // _TARGET_ARM_ assert(isPassedAsPrimitiveType); } } else { assert(!varTypeIsStruct(node)); } } #ifdef DEBUG void Dump(); #endif }; //------------------------------------------------------------------------- // // The class fgArgInfo is used to handle the arguments // when morphing a GT_CALL node. // class fgArgInfo { Compiler* compiler; // Back pointer to the compiler instance so that we can allocate memory GenTreeCall* callTree; // Back pointer to the GT_CALL node for this fgArgInfo unsigned argCount; // Updatable arg count value unsigned nextSlotNum; // Updatable slot count value unsigned stkLevel; // Stack depth when we make this call (for x86) #if defined(UNIX_X86_ABI) bool alignmentDone; // Updateable flag, set to 'true' after we've done any required alignment. unsigned stkSizeBytes; // Size of stack used by this call, in bytes. Calculated during fgMorphArgs(). unsigned padStkAlign; // Stack alignment in bytes required before arguments are pushed for this call. // Computed dynamically during codegen, based on stkSizeBytes and the current // stack level (genStackLevel) when the first stack adjustment is made for // this call. #endif #if FEATURE_FIXED_OUT_ARGS unsigned outArgSize; // Size of the out arg area for the call, will be at least MIN_ARG_AREA_FOR_CALL #endif unsigned argTableSize; // size of argTable array (equal to the argCount when done with fgMorphArgs) bool hasRegArgs; // true if we have one or more register arguments bool hasStackArgs; // true if we have one or more stack arguments bool argsComplete; // marker for state bool argsSorted; // marker for state bool needsTemps; // one or more arguments must be copied to a temp by EvalArgsToTemps fgArgTabEntry** argTable; // variable sized array of per argument descrption: (i.e. argTable[argTableSize]) private: void AddArg(fgArgTabEntry* curArgTabEntry); public: fgArgInfo(Compiler* comp, GenTreeCall* call, unsigned argCount); fgArgInfo(GenTreeCall* newCall, GenTreeCall* oldCall); fgArgTabEntry* AddRegArg(unsigned argNum, GenTree* node, GenTree* parent, regNumber regNum, unsigned numRegs, unsigned alignment, bool isStruct, bool isVararg = false); #ifdef UNIX_AMD64_ABI fgArgTabEntry* AddRegArg(unsigned argNum, GenTree* node, GenTree* parent, regNumber regNum, unsigned numRegs, unsigned alignment, const bool isStruct, const bool isVararg, const regNumber otherRegNum, const SYSTEMV_AMD64_CORINFO_STRUCT_REG_PASSING_DESCRIPTOR* const structDescPtr = nullptr); #endif // UNIX_AMD64_ABI fgArgTabEntry* AddStkArg(unsigned argNum, GenTree* node, GenTree* parent, unsigned numSlots, unsigned alignment, bool isStruct, bool isVararg = false); void RemorphReset(); void UpdateRegArg(fgArgTabEntry* argEntry, GenTree* node, bool reMorphing); void UpdateStkArg(fgArgTabEntry* argEntry, GenTree* node, bool reMorphing); void SplitArg(unsigned argNum, unsigned numRegs, unsigned numSlots); void EvalToTmp(fgArgTabEntry* curArgTabEntry, unsigned tmpNum, GenTree* newNode); void ArgsComplete(); void SortArgs(); void EvalArgsToTemps(); unsigned ArgCount() { return argCount; } fgArgTabEntry** ArgTable() { return argTable; } unsigned GetNextSlotNum() { return nextSlotNum; } bool HasRegArgs() { return hasRegArgs; } bool NeedsTemps() { return needsTemps; } bool HasStackArgs() { return hasStackArgs; } bool AreArgsComplete() const { return argsComplete; } #if FEATURE_FIXED_OUT_ARGS unsigned GetOutArgSize() const { return outArgSize; } void SetOutArgSize(unsigned newVal) { outArgSize = newVal; } #endif // FEATURE_FIXED_OUT_ARGS #if defined(UNIX_X86_ABI) void ComputeStackAlignment(unsigned curStackLevelInBytes) { padStkAlign = AlignmentPad(curStackLevelInBytes, STACK_ALIGN); } unsigned GetStkAlign() { return padStkAlign; } void SetStkSizeBytes(unsigned newStkSizeBytes) { stkSizeBytes = newStkSizeBytes; } unsigned GetStkSizeBytes() const { return stkSizeBytes; } bool IsStkAlignmentDone() const { return alignmentDone; } void SetStkAlignmentDone() { alignmentDone = true; } #endif // defined(UNIX_X86_ABI) // Get the fgArgTabEntry for the arg at position argNum. fgArgTabEntry* GetArgEntry(unsigned argNum, bool reMorphing = true) { fgArgTabEntry* curArgTabEntry = nullptr; if (!reMorphing) { // The arg table has not yet been sorted. curArgTabEntry = argTable[argNum]; assert(curArgTabEntry->argNum == argNum); return curArgTabEntry; } for (unsigned i = 0; i < argCount; i++) { curArgTabEntry = argTable[i]; if (curArgTabEntry->argNum == argNum) { return curArgTabEntry; } } noway_assert(!"GetArgEntry: argNum not found"); return nullptr; } // Get the node for the arg at position argIndex. // Caller must ensure that this index is a valid arg index. GenTree* GetArgNode(unsigned argIndex) { return GetArgEntry(argIndex)->node; } void Dump(Compiler* compiler); }; #ifdef DEBUG // XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX // We have the ability to mark source expressions with "Test Labels." // These drive assertions within the JIT, or internal JIT testing. For example, we could label expressions // that should be CSE defs, and other expressions that should uses of those defs, with a shared label. enum TestLabel // This must be kept identical to System.Runtime.CompilerServices.JitTestLabel.TestLabel. { TL_SsaName, TL_VN, // Defines a "VN equivalence class". (For full VN, including exceptions thrown). TL_VNNorm, // Like above, but uses the non-exceptional value of the expression. TL_CSE_Def, // This must be identified in the JIT as a CSE def TL_CSE_Use, // This must be identified in the JIT as a CSE use TL_LoopHoist, // Expression must (or must not) be hoisted out of the loop. }; struct TestLabelAndNum { TestLabel m_tl; ssize_t m_num; TestLabelAndNum() : m_tl(TestLabel(0)), m_num(0) { } }; typedef JitHashTable, TestLabelAndNum> NodeToTestDataMap; // XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX #endif // DEBUG /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX The big guy. The sections are currently organized as : XX XX XX XX o GenTree and BasicBlock XX XX o LclVarsInfo XX XX o Importer XX XX o FlowGraph XX XX o Optimizer XX XX o RegAlloc XX XX o EEInterface XX XX o TempsInfo XX XX o RegSet XX XX o GCInfo XX XX o Instruction XX XX o ScopeInfo XX XX o PrologScopeInfo XX XX o CodeGenerator XX XX o UnwindInfo XX XX o Compiler XX XX o typeInfo XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ struct HWIntrinsicInfo; class Compiler { friend class emitter; friend class UnwindInfo; friend class UnwindFragmentInfo; friend class UnwindEpilogInfo; friend class JitTimer; friend class LinearScan; friend class fgArgInfo; friend class Rationalizer; friend class Phase; friend class Lowering; friend class CSE_DataFlow; friend class CSE_Heuristic; friend class CodeGenInterface; friend class CodeGen; friend class LclVarDsc; friend class TempDsc; friend class LIR; friend class ObjectAllocator; friend class LocalAddressVisitor; friend struct GenTree; #ifdef FEATURE_HW_INTRINSICS friend struct HWIntrinsicInfo; #endif // FEATURE_HW_INTRINSICS #ifndef _TARGET_64BIT_ friend class DecomposeLongs; #endif // !_TARGET_64BIT_ /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX Misc structs definitions XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: hashBvGlobalData hbvGlobalData; // Used by the hashBv bitvector package. #ifdef DEBUG bool verbose; bool dumpIR; bool dumpIRNodes; bool dumpIRTypes; bool dumpIRKinds; bool dumpIRLocals; bool dumpIRRegs; bool dumpIRSsa; bool dumpIRValnums; bool dumpIRCosts; bool dumpIRFlags; bool dumpIRNoLists; bool dumpIRNoLeafs; bool dumpIRNoStmts; bool dumpIRTrees; bool dumpIRLinear; bool dumpIRDataflow; bool dumpIRBlockHeaders; bool dumpIRExit; LPCWSTR dumpIRPhase; LPCWSTR dumpIRFormat; bool verboseTrees; bool shouldUseVerboseTrees(); bool asciiTrees; // If true, dump trees using only ASCII characters bool shouldDumpASCIITrees(); bool verboseSsa; // If true, produce especially verbose dump output in SSA construction. bool shouldUseVerboseSsa(); bool treesBeforeAfterMorph; // If true, print trees before/after morphing (paired by an intra-compilation id: int morphNum; // This counts the the trees that have been morphed, allowing us to label each uniquely. const char* VarNameToStr(VarName name) { return name; } DWORD expensiveDebugCheckLevel; #endif #if FEATURE_MULTIREG_RET GenTree* impAssignMultiRegTypeToVar(GenTree* op, CORINFO_CLASS_HANDLE hClass); #endif // FEATURE_MULTIREG_RET GenTree* impAssignSmallStructTypeToVar(GenTree* op, CORINFO_CLASS_HANDLE hClass); #ifdef ARM_SOFTFP bool isSingleFloat32Struct(CORINFO_CLASS_HANDLE hClass); #endif // ARM_SOFTFP //------------------------------------------------------------------------- // Functions to handle homogeneous floating-point aggregates (HFAs) in ARM. // HFAs are one to four element structs where each element is the same // type, either all float or all double. They are treated specially // in the ARM Procedure Call Standard, specifically, they are passed in // floating-point registers instead of the general purpose registers. // bool IsHfa(CORINFO_CLASS_HANDLE hClass); bool IsHfa(GenTree* tree); var_types GetHfaType(GenTree* tree); unsigned GetHfaCount(GenTree* tree); var_types GetHfaType(CORINFO_CLASS_HANDLE hClass); unsigned GetHfaCount(CORINFO_CLASS_HANDLE hClass); bool IsMultiRegReturnedType(CORINFO_CLASS_HANDLE hClass); //------------------------------------------------------------------------- // The following is used for validating format of EH table // struct EHNodeDsc; typedef struct EHNodeDsc* pEHNodeDsc; EHNodeDsc* ehnTree; // root of the tree comprising the EHnodes. EHNodeDsc* ehnNext; // root of the tree comprising the EHnodes. struct EHNodeDsc { enum EHBlockType { TryNode, FilterNode, HandlerNode, FinallyNode, FaultNode }; EHBlockType ehnBlockType; // kind of EH block IL_OFFSET ehnStartOffset; // IL offset of start of the EH block IL_OFFSET ehnEndOffset; // IL offset past end of the EH block. (TODO: looks like verInsertEhNode() sets this to // the last IL offset, not "one past the last one", i.e., the range Start to End is // inclusive). pEHNodeDsc ehnNext; // next (non-nested) block in sequential order pEHNodeDsc ehnChild; // leftmost nested block union { pEHNodeDsc ehnTryNode; // for filters and handlers, the corresponding try node pEHNodeDsc ehnHandlerNode; // for a try node, the corresponding handler node }; pEHNodeDsc ehnFilterNode; // if this is a try node and has a filter, otherwise 0 pEHNodeDsc ehnEquivalent; // if blockType=tryNode, start offset and end offset is same, void ehnSetTryNodeType() { ehnBlockType = TryNode; } void ehnSetFilterNodeType() { ehnBlockType = FilterNode; } void ehnSetHandlerNodeType() { ehnBlockType = HandlerNode; } void ehnSetFinallyNodeType() { ehnBlockType = FinallyNode; } void ehnSetFaultNodeType() { ehnBlockType = FaultNode; } BOOL ehnIsTryBlock() { return ehnBlockType == TryNode; } BOOL ehnIsFilterBlock() { return ehnBlockType == FilterNode; } BOOL ehnIsHandlerBlock() { return ehnBlockType == HandlerNode; } BOOL ehnIsFinallyBlock() { return ehnBlockType == FinallyNode; } BOOL ehnIsFaultBlock() { return ehnBlockType == FaultNode; } // returns true if there is any overlap between the two nodes static BOOL ehnIsOverlap(pEHNodeDsc node1, pEHNodeDsc node2) { if (node1->ehnStartOffset < node2->ehnStartOffset) { return (node1->ehnEndOffset >= node2->ehnStartOffset); } else { return (node1->ehnStartOffset <= node2->ehnEndOffset); } } // fails with BADCODE if inner is not completely nested inside outer static BOOL ehnIsNested(pEHNodeDsc inner, pEHNodeDsc outer) { return ((inner->ehnStartOffset >= outer->ehnStartOffset) && (inner->ehnEndOffset <= outer->ehnEndOffset)); } }; //------------------------------------------------------------------------- // Exception handling functions // #if !FEATURE_EH_FUNCLETS bool ehNeedsShadowSPslots() { return (info.compXcptnsCount || opts.compDbgEnC); } // 0 for methods with no EH // 1 for methods with non-nested EH, or where only the try blocks are nested // 2 for a method with a catch within a catch // etc. unsigned ehMaxHndNestingCount; #endif // !FEATURE_EH_FUNCLETS static bool jitIsBetween(unsigned value, unsigned start, unsigned end); static bool jitIsBetweenInclusive(unsigned value, unsigned start, unsigned end); bool bbInCatchHandlerILRange(BasicBlock* blk); bool bbInFilterILRange(BasicBlock* blk); bool bbInTryRegions(unsigned regionIndex, BasicBlock* blk); bool bbInExnFlowRegions(unsigned regionIndex, BasicBlock* blk); bool bbInHandlerRegions(unsigned regionIndex, BasicBlock* blk); bool bbInCatchHandlerRegions(BasicBlock* tryBlk, BasicBlock* hndBlk); unsigned short bbFindInnermostCommonTryRegion(BasicBlock* bbOne, BasicBlock* bbTwo); unsigned short bbFindInnermostTryRegionContainingHandlerRegion(unsigned handlerIndex); unsigned short bbFindInnermostHandlerRegionContainingTryRegion(unsigned tryIndex); // Returns true if "block" is the start of a try region. bool bbIsTryBeg(BasicBlock* block); // Returns true if "block" is the start of a handler or filter region. bool bbIsHandlerBeg(BasicBlock* block); // Returns true iff "block" is where control flows if an exception is raised in the // try region, and sets "*regionIndex" to the index of the try for the handler. // Differs from "IsHandlerBeg" in the case of filters, where this is true for the first // block of the filter, but not for the filter's handler. bool bbIsExFlowBlock(BasicBlock* block, unsigned* regionIndex); bool ehHasCallableHandlers(); // Return the EH descriptor for the given region index. EHblkDsc* ehGetDsc(unsigned regionIndex); // Return the EH index given a region descriptor. unsigned ehGetIndex(EHblkDsc* ehDsc); // Return the EH descriptor index of the enclosing try, for the given region index. unsigned ehGetEnclosingTryIndex(unsigned regionIndex); // Return the EH descriptor index of the enclosing handler, for the given region index. unsigned ehGetEnclosingHndIndex(unsigned regionIndex); // Return the EH descriptor for the most nested 'try' region this BasicBlock is a member of (or nullptr if this // block is not in a 'try' region). EHblkDsc* ehGetBlockTryDsc(BasicBlock* block); // Return the EH descriptor for the most nested filter or handler region this BasicBlock is a member of (or nullptr // if this block is not in a filter or handler region). EHblkDsc* ehGetBlockHndDsc(BasicBlock* block); // Return the EH descriptor for the most nested region that may handle exceptions raised in this BasicBlock (or // nullptr if this block's exceptions propagate to caller). EHblkDsc* ehGetBlockExnFlowDsc(BasicBlock* block); EHblkDsc* ehIsBlockTryLast(BasicBlock* block); EHblkDsc* ehIsBlockHndLast(BasicBlock* block); bool ehIsBlockEHLast(BasicBlock* block); bool ehBlockHasExnFlowDsc(BasicBlock* block); // Return the region index of the most nested EH region this block is in. unsigned ehGetMostNestedRegionIndex(BasicBlock* block, bool* inTryRegion); // Find the true enclosing try index, ignoring 'mutual protect' try. Uses IL ranges to check. unsigned ehTrueEnclosingTryIndexIL(unsigned regionIndex); // Return the index of the most nested enclosing region for a particular EH region. Returns NO_ENCLOSING_INDEX // if there is no enclosing region. If the returned index is not NO_ENCLOSING_INDEX, then '*inTryRegion' // is set to 'true' if the enclosing region is a 'try', or 'false' if the enclosing region is a handler. // (It can never be a filter.) unsigned ehGetEnclosingRegionIndex(unsigned regionIndex, bool* inTryRegion); // A block has been deleted. Update the EH table appropriately. void ehUpdateForDeletedBlock(BasicBlock* block); // Determine whether a block can be deleted while preserving the EH normalization rules. bool ehCanDeleteEmptyBlock(BasicBlock* block); // Update the 'last' pointers in the EH table to reflect new or deleted blocks in an EH region. void ehUpdateLastBlocks(BasicBlock* oldLast, BasicBlock* newLast); // For a finally handler, find the region index that the BBJ_CALLFINALLY lives in that calls the handler, // or NO_ENCLOSING_INDEX if the BBJ_CALLFINALLY lives in the main function body. Normally, the index // is the same index as the handler (and the BBJ_CALLFINALLY lives in the 'try' region), but for AMD64 the // BBJ_CALLFINALLY lives in the enclosing try or handler region, whichever is more nested, or the main function // body. If the returned index is not NO_ENCLOSING_INDEX, then '*inTryRegion' is set to 'true' if the // BBJ_CALLFINALLY lives in the returned index's 'try' region, or 'false' if lives in the handler region. (It never // lives in a filter.) unsigned ehGetCallFinallyRegionIndex(unsigned finallyIndex, bool* inTryRegion); // Find the range of basic blocks in which all BBJ_CALLFINALLY will be found that target the 'finallyIndex' region's // handler. Set begBlk to the first block, and endBlk to the block after the last block of the range // (nullptr if the last block is the last block in the program). // Precondition: 'finallyIndex' is the EH region of a try/finally clause. void ehGetCallFinallyBlockRange(unsigned finallyIndex, BasicBlock** begBlk, BasicBlock** endBlk); #ifdef DEBUG // Given a BBJ_CALLFINALLY block and the EH region index of the finally it is calling, return // 'true' if the BBJ_CALLFINALLY is in the correct EH region. bool ehCallFinallyInCorrectRegion(BasicBlock* blockCallFinally, unsigned finallyIndex); #endif // DEBUG #if FEATURE_EH_FUNCLETS // Do we need a PSPSym in the main function? For codegen purposes, we only need one // if there is a filter that protects a region with a nested EH clause (such as a // try/catch nested in the 'try' body of a try/filter/filter-handler). See // genFuncletProlog() for more details. However, the VM seems to use it for more // purposes, maybe including debugging. Until we are sure otherwise, always create // a PSPSym for functions with any EH. bool ehNeedsPSPSym() const { #ifdef _TARGET_X86_ return false; #else // _TARGET_X86_ return compHndBBtabCount > 0; #endif // _TARGET_X86_ } bool ehAnyFunclets(); // Are there any funclets in this function? unsigned ehFuncletCount(); // Return the count of funclets in the function unsigned bbThrowIndex(BasicBlock* blk); // Get the index to use as the cache key for sharing throw blocks #else // !FEATURE_EH_FUNCLETS bool ehAnyFunclets() { return false; } unsigned ehFuncletCount() { return 0; } unsigned bbThrowIndex(BasicBlock* blk) { return blk->bbTryIndex; } // Get the index to use as the cache key for sharing throw blocks #endif // !FEATURE_EH_FUNCLETS // Returns a flowList representing the "EH predecessors" of "blk". These are the normal predecessors of // "blk", plus one special case: if "blk" is the first block of a handler, considers the predecessor(s) of the first // first block of the corresponding try region to be "EH predecessors". (If there is a single such predecessor, // for example, we want to consider that the immediate dominator of the catch clause start block, so it's // convenient to also consider it a predecessor.) flowList* BlockPredsWithEH(BasicBlock* blk); // This table is useful for memoization of the method above. typedef JitHashTable, flowList*> BlockToFlowListMap; BlockToFlowListMap* m_blockToEHPreds; BlockToFlowListMap* GetBlockToEHPreds() { if (m_blockToEHPreds == nullptr) { m_blockToEHPreds = new (getAllocator()) BlockToFlowListMap(getAllocator()); } return m_blockToEHPreds; } void* ehEmitCookie(BasicBlock* block); UNATIVE_OFFSET ehCodeOffset(BasicBlock* block); EHblkDsc* ehInitHndRange(BasicBlock* src, IL_OFFSET* hndBeg, IL_OFFSET* hndEnd, bool* inFilter); EHblkDsc* ehInitTryRange(BasicBlock* src, IL_OFFSET* tryBeg, IL_OFFSET* tryEnd); EHblkDsc* ehInitHndBlockRange(BasicBlock* blk, BasicBlock** hndBeg, BasicBlock** hndLast, bool* inFilter); EHblkDsc* ehInitTryBlockRange(BasicBlock* blk, BasicBlock** tryBeg, BasicBlock** tryLast); void fgSetTryEnd(EHblkDsc* handlerTab, BasicBlock* newTryLast); void fgSetHndEnd(EHblkDsc* handlerTab, BasicBlock* newHndLast); void fgSkipRmvdBlocks(EHblkDsc* handlerTab); void fgAllocEHTable(); void fgRemoveEHTableEntry(unsigned XTnum); #if FEATURE_EH_FUNCLETS EHblkDsc* fgAddEHTableEntry(unsigned XTnum); #endif // FEATURE_EH_FUNCLETS #if !FEATURE_EH void fgRemoveEH(); #endif // !FEATURE_EH void fgSortEHTable(); // Causes the EH table to obey some well-formedness conditions, by inserting // empty BB's when necessary: // * No block is both the first block of a handler and the first block of a try. // * No block is the first block of multiple 'try' regions. // * No block is the last block of multiple EH regions. void fgNormalizeEH(); bool fgNormalizeEHCase1(); bool fgNormalizeEHCase2(); bool fgNormalizeEHCase3(); #ifdef DEBUG void dispIncomingEHClause(unsigned num, const CORINFO_EH_CLAUSE& clause); void dispOutgoingEHClause(unsigned num, const CORINFO_EH_CLAUSE& clause); void fgVerifyHandlerTab(); void fgDispHandlerTab(); #endif // DEBUG bool fgNeedToSortEHTable; void verInitEHTree(unsigned numEHClauses); void verInsertEhNode(CORINFO_EH_CLAUSE* clause, EHblkDsc* handlerTab); void verInsertEhNodeInTree(EHNodeDsc** ppRoot, EHNodeDsc* node); void verInsertEhNodeParent(EHNodeDsc** ppRoot, EHNodeDsc* node); void verCheckNestingLevel(EHNodeDsc* initRoot); /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX GenTree and BasicBlock XX XX XX XX Functions to allocate and display the GenTrees and BasicBlocks XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ // Functions to create nodes GenTreeStmt* gtNewStmt(GenTree* expr = nullptr, IL_OFFSETX offset = BAD_IL_OFFSET); // For unary opers. GenTree* gtNewOperNode(genTreeOps oper, var_types type, GenTree* op1, bool doSimplifications = TRUE); // For binary opers. GenTree* gtNewOperNode(genTreeOps oper, var_types type, GenTree* op1, GenTree* op2); GenTree* gtNewQmarkNode(var_types type, GenTree* cond, GenTree* colon); GenTree* gtNewLargeOperNode(genTreeOps oper, var_types type = TYP_I_IMPL, GenTree* op1 = nullptr, GenTree* op2 = nullptr); GenTreeIntCon* gtNewIconNode(ssize_t value, var_types type = TYP_INT); GenTree* gtNewPhysRegNode(regNumber reg, var_types type); GenTree* gtNewJmpTableNode(); GenTree* gtNewIndOfIconHandleNode(var_types indType, size_t value, unsigned iconFlags, bool isInvariant); GenTree* gtNewIconHandleNode(size_t value, unsigned flags, FieldSeqNode* fields = nullptr); unsigned gtTokenToIconFlags(unsigned token); GenTree* gtNewIconEmbHndNode(void* value, void* pValue, unsigned flags, void* compileTimeHandle); GenTree* gtNewIconEmbScpHndNode(CORINFO_MODULE_HANDLE scpHnd); GenTree* gtNewIconEmbClsHndNode(CORINFO_CLASS_HANDLE clsHnd); GenTree* gtNewIconEmbMethHndNode(CORINFO_METHOD_HANDLE methHnd); GenTree* gtNewIconEmbFldHndNode(CORINFO_FIELD_HANDLE fldHnd); GenTree* gtNewStringLiteralNode(InfoAccessType iat, void* pValue); GenTree* gtNewLconNode(__int64 value); GenTree* gtNewDconNode(double value, var_types type = TYP_DOUBLE); GenTree* gtNewSconNode(int CPX, CORINFO_MODULE_HANDLE scpHandle); GenTree* gtNewZeroConNode(var_types type); GenTree* gtNewOneConNode(var_types type); #ifdef FEATURE_SIMD GenTree* gtNewSIMDVectorZero(var_types simdType, var_types baseType, unsigned size); GenTree* gtNewSIMDVectorOne(var_types simdType, var_types baseType, unsigned size); #endif GenTree* gtNewBlkOpNode(GenTree* dst, GenTree* srcOrFillVal, unsigned size, bool isVolatile, bool isCopyBlock); GenTree* gtNewPutArgReg(var_types type, GenTree* arg, regNumber argReg); GenTree* gtNewBitCastNode(var_types type, GenTree* arg); protected: void gtBlockOpInit(GenTree* result, GenTree* dst, GenTree* srcOrFillVal, bool isVolatile); public: GenTree* gtNewObjNode(CORINFO_CLASS_HANDLE structHnd, GenTree* addr); void gtSetObjGcInfo(GenTreeObj* objNode); GenTree* gtNewStructVal(CORINFO_CLASS_HANDLE structHnd, GenTree* addr); GenTree* gtNewBlockVal(GenTree* addr, unsigned size); GenTree* gtNewCpObjNode(GenTree* dst, GenTree* src, CORINFO_CLASS_HANDLE structHnd, bool isVolatile); GenTreeArgList* gtNewListNode(GenTree* op1, GenTreeArgList* op2); GenTreeCall* gtNewCallNode(gtCallTypes callType, CORINFO_METHOD_HANDLE handle, var_types type, GenTreeArgList* args, IL_OFFSETX ilOffset = BAD_IL_OFFSET); GenTreeCall* gtNewIndCallNode(GenTree* addr, var_types type, GenTreeArgList* args, IL_OFFSETX ilOffset = BAD_IL_OFFSET); GenTreeCall* gtNewHelperCallNode(unsigned helper, var_types type, GenTreeArgList* args = nullptr); GenTree* gtNewLclvNode(unsigned lnum, var_types type DEBUGARG(IL_OFFSETX ILoffs = BAD_IL_OFFSET)); GenTree* gtNewLclLNode(unsigned lnum, var_types type DEBUGARG(IL_OFFSETX ILoffs = BAD_IL_OFFSET)); #ifdef FEATURE_SIMD GenTreeSIMD* gtNewSIMDNode( var_types type, GenTree* op1, SIMDIntrinsicID simdIntrinsicID, var_types baseType, unsigned size); GenTreeSIMD* gtNewSIMDNode( var_types type, GenTree* op1, GenTree* op2, SIMDIntrinsicID simdIntrinsicID, var_types baseType, unsigned size); void SetOpLclRelatedToSIMDIntrinsic(GenTree* op); #endif #ifdef FEATURE_HW_INTRINSICS GenTreeHWIntrinsic* gtNewSimdHWIntrinsicNode(var_types type, NamedIntrinsic hwIntrinsicID, var_types baseType, unsigned size); GenTreeHWIntrinsic* gtNewSimdHWIntrinsicNode( var_types type, GenTree* op1, NamedIntrinsic hwIntrinsicID, var_types baseType, unsigned size); GenTreeHWIntrinsic* gtNewSimdHWIntrinsicNode( var_types type, GenTree* op1, GenTree* op2, NamedIntrinsic hwIntrinsicID, var_types baseType, unsigned size); GenTreeHWIntrinsic* gtNewSimdHWIntrinsicNode(var_types type, GenTree* op1, GenTree* op2, GenTree* op3, NamedIntrinsic hwIntrinsicID, var_types baseType, unsigned size); GenTreeHWIntrinsic* gtNewSimdHWIntrinsicNode(var_types type, GenTree* op1, GenTree* op2, GenTree* op3, GenTree* op4, NamedIntrinsic hwIntrinsicID, var_types baseType, unsigned size); GenTreeHWIntrinsic* gtNewScalarHWIntrinsicNode(var_types type, GenTree* op1, NamedIntrinsic hwIntrinsicID); GenTreeHWIntrinsic* gtNewScalarHWIntrinsicNode(var_types type, GenTree* op1, GenTree* op2, NamedIntrinsic hwIntrinsicID); GenTreeHWIntrinsic* gtNewScalarHWIntrinsicNode( var_types type, GenTree* op1, GenTree* op2, GenTree* op3, NamedIntrinsic hwIntrinsicID); GenTree* gtNewMustThrowException(unsigned helper, var_types type, CORINFO_CLASS_HANDLE clsHnd); CORINFO_CLASS_HANDLE gtGetStructHandleForHWSIMD(var_types simdType, var_types simdBaseType); #endif // FEATURE_HW_INTRINSICS GenTreeLclFld* gtNewLclFldNode(unsigned lnum, var_types type, unsigned offset); GenTree* gtNewInlineCandidateReturnExpr(GenTree* inlineCandidate, var_types type); GenTree* gtNewFieldRef(var_types typ, CORINFO_FIELD_HANDLE fldHnd, GenTree* obj = nullptr, DWORD offset = 0); GenTree* gtNewIndexRef(var_types typ, GenTree* arrayOp, GenTree* indexOp); GenTreeArrLen* gtNewArrLen(var_types typ, GenTree* arrayOp, int lenOffset); GenTree* gtNewIndir(var_types typ, GenTree* addr); GenTreeArgList* gtNewArgList(GenTree* op); GenTreeArgList* gtNewArgList(GenTree* op1, GenTree* op2); GenTreeArgList* gtNewArgList(GenTree* op1, GenTree* op2, GenTree* op3); GenTreeArgList* gtNewArgList(GenTree* op1, GenTree* op2, GenTree* op3, GenTree* op4); static fgArgTabEntry* gtArgEntryByArgNum(GenTreeCall* call, unsigned argNum); static fgArgTabEntry* gtArgEntryByNode(GenTreeCall* call, GenTree* node); fgArgTabEntry* gtArgEntryByLateArgIndex(GenTreeCall* call, unsigned lateArgInx); static GenTree* gtArgNodeByLateArgInx(GenTreeCall* call, unsigned lateArgInx); bool gtArgIsThisPtr(fgArgTabEntry* argEntry); GenTree* gtNewAssignNode(GenTree* dst, GenTree* src); GenTree* gtNewTempAssign(unsigned tmp, GenTree* val, GenTreeStmt** pAfterStmt = nullptr, IL_OFFSETX ilOffset = BAD_IL_OFFSET, BasicBlock* block = nullptr); GenTree* gtNewRefCOMfield(GenTree* objPtr, CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_ACCESS_FLAGS access, CORINFO_FIELD_INFO* pFieldInfo, var_types lclTyp, CORINFO_CLASS_HANDLE structType, GenTree* assg); GenTree* gtNewNothingNode(); GenTree* gtNewArgPlaceHolderNode(var_types type, CORINFO_CLASS_HANDLE clsHnd); GenTree* gtUnusedValNode(GenTree* expr); GenTreeCast* gtNewCastNode(var_types typ, GenTree* op1, bool fromUnsigned, var_types castType); GenTreeCast* gtNewCastNodeL(var_types typ, GenTree* op1, bool fromUnsigned, var_types castType); GenTreeAllocObj* gtNewAllocObjNode( unsigned int helper, bool helperHasSideEffects, CORINFO_CLASS_HANDLE clsHnd, var_types type, GenTree* op1); GenTreeAllocObj* gtNewAllocObjNode(CORINFO_RESOLVED_TOKEN* pResolvedToken, BOOL useParent); GenTree* gtNewRuntimeLookup(CORINFO_GENERIC_HANDLE hnd, CorInfoGenericHandleType hndTyp, GenTree* lookupTree); //------------------------------------------------------------------------ // Other GenTree functions GenTree* gtClone(GenTree* tree, bool complexOK = false); // If `tree` is a lclVar with lclNum `varNum`, return an IntCns with value `varVal`; otherwise, // create a copy of `tree`, adding specified flags, replacing uses of lclVar `deepVarNum` with // IntCnses with value `deepVarVal`. GenTree* gtCloneExpr( GenTree* tree, unsigned addFlags, unsigned varNum, int varVal, unsigned deepVarNum, int deepVarVal); // Create a copy of `tree`, optionally adding specifed flags, and optionally mapping uses of local // `varNum` to int constants with value `varVal`. GenTree* gtCloneExpr(GenTree* tree, unsigned addFlags = 0, unsigned varNum = BAD_VAR_NUM, int varVal = 0) { return gtCloneExpr(tree, addFlags, varNum, varVal, varNum, varVal); } // Internal helper for cloning a call GenTreeCall* gtCloneExprCallHelper(GenTreeCall* call, unsigned addFlags = 0, unsigned deepVarNum = BAD_VAR_NUM, int deepVarVal = 0); // Create copy of an inline or guarded devirtualization candidate tree. GenTreeCall* gtCloneCandidateCall(GenTreeCall* call); GenTree* gtReplaceTree(GenTreeStmt* stmt, GenTree* tree, GenTree* replacementTree); void gtUpdateSideEffects(GenTreeStmt* stmt, GenTree* tree); void gtUpdateTreeAncestorsSideEffects(GenTree* tree); void gtUpdateStmtSideEffects(GenTreeStmt* stmt); void gtUpdateNodeSideEffects(GenTree* tree); void gtUpdateNodeOperSideEffects(GenTree* tree); // Returns "true" iff the complexity (not formally defined, but first interpretation // is #of nodes in subtree) of "tree" is greater than "limit". // (This is somewhat redundant with the "gtCostEx/gtCostSz" fields, but can be used // before they have been set.) bool gtComplexityExceeds(GenTree** tree, unsigned limit); bool gtCompareTree(GenTree* op1, GenTree* op2); GenTree* gtReverseCond(GenTree* tree); bool gtHasRef(GenTree* tree, ssize_t lclNum, bool defOnly); bool gtHasLocalsWithAddrOp(GenTree* tree); unsigned gtSetListOrder(GenTree* list, bool regs, bool isListCallArgs); void gtWalkOp(GenTree** op1, GenTree** op2, GenTree* base, bool constOnly); #ifdef DEBUG unsigned gtHashValue(GenTree* tree); GenTree* gtWalkOpEffectiveVal(GenTree* op); #endif void gtPrepareCost(GenTree* tree); bool gtIsLikelyRegVar(GenTree* tree); // Returns true iff the secondNode can be swapped with firstNode. bool gtCanSwapOrder(GenTree* firstNode, GenTree* secondNode); // Given an address expression, compute its costs and addressing mode opportunities, // and mark addressing mode candidates as GTF_DONT_CSE. // TODO-Throughput - Consider actually instantiating these early, to avoid // having to re-run the algorithm that looks for them (might also improve CQ). bool gtMarkAddrMode(GenTree* addr, int* costEx, int* costSz, var_types type); unsigned gtSetEvalOrder(GenTree* tree); void gtSetStmtInfo(GenTreeStmt* stmt); // Returns "true" iff "node" has any of the side effects in "flags". bool gtNodeHasSideEffects(GenTree* node, unsigned flags); // Returns "true" iff "tree" or its (transitive) children have any of the side effects in "flags". bool gtTreeHasSideEffects(GenTree* tree, unsigned flags); // Appends 'expr' in front of 'list' // 'list' will typically start off as 'nullptr' // when 'list' is non-null a GT_COMMA node is used to insert 'expr' GenTree* gtBuildCommaList(GenTree* list, GenTree* expr); void gtExtractSideEffList(GenTree* expr, GenTree** pList, unsigned flags = GTF_SIDE_EFFECT, bool ignoreRoot = false); GenTree* gtGetThisArg(GenTreeCall* call); // Static fields of struct types (and sometimes the types that those are reduced to) are represented by having the // static field contain an object pointer to the boxed struct. This simplifies the GC implementation...but // complicates the JIT somewhat. This predicate returns "true" iff a node with type "fieldNodeType", representing // the given "fldHnd", is such an object pointer. bool gtIsStaticFieldPtrToBoxedStruct(var_types fieldNodeType, CORINFO_FIELD_HANDLE fldHnd); // Return true if call is a recursive call; return false otherwise. // Note when inlining, this looks for calls back to the root method. bool gtIsRecursiveCall(GenTreeCall* call) { return gtIsRecursiveCall(call->gtCallMethHnd); } bool gtIsRecursiveCall(CORINFO_METHOD_HANDLE callMethodHandle) { return (callMethodHandle == impInlineRoot()->info.compMethodHnd); } //------------------------------------------------------------------------- GenTree* gtFoldExpr(GenTree* tree); GenTree* #ifdef __clang__ // TODO-Amd64-Unix: Remove this when the clang optimizer is fixed and/or the method implementation is // refactored in a simpler code. This is a workaround for a bug in the clang-3.5 optimizer. The issue is that in // release build the optimizer is mistyping (or just wrongly decides to use 32 bit operation for a corner case // of MIN_LONG) the args of the (ltemp / lval2) to int (it does a 32 bit div operation instead of 64 bit) - see // the implementation of the method in gentree.cpp. For the case of lval1 and lval2 equal to MIN_LONG // (0x8000000000000000) this results in raising a SIGFPE. The method implementation is rather complex. Disable // optimizations for now. __attribute__((optnone)) #endif // __clang__ gtFoldExprConst(GenTree* tree); GenTree* gtFoldExprSpecial(GenTree* tree); GenTree* gtFoldExprCompare(GenTree* tree); GenTree* gtCreateHandleCompare(genTreeOps oper, GenTree* op1, GenTree* op2, CorInfoInlineTypeCheck typeCheckInliningResult); GenTree* gtFoldExprCall(GenTreeCall* call); GenTree* gtFoldTypeCompare(GenTree* tree); GenTree* gtFoldTypeEqualityCall(CorInfoIntrinsics methodID, GenTree* op1, GenTree* op2); // Options to control behavior of gtTryRemoveBoxUpstreamEffects enum BoxRemovalOptions { BR_REMOVE_AND_NARROW, // remove effects, minimize remaining work, return possibly narrowed source tree BR_REMOVE_AND_NARROW_WANT_TYPE_HANDLE, // remove effects and minimize remaining work, return type handle tree BR_REMOVE_BUT_NOT_NARROW, // remove effects, return original source tree BR_DONT_REMOVE, // check if removal is possible, return copy source tree BR_DONT_REMOVE_WANT_TYPE_HANDLE, // check if removal is possible, return type handle tree BR_MAKE_LOCAL_COPY // revise box to copy to temp local and return local's address }; GenTree* gtTryRemoveBoxUpstreamEffects(GenTree* tree, BoxRemovalOptions options = BR_REMOVE_AND_NARROW); GenTree* gtOptimizeEnumHasFlag(GenTree* thisOp, GenTree* flagOp); //------------------------------------------------------------------------- // Get the handle, if any. CORINFO_CLASS_HANDLE gtGetStructHandleIfPresent(GenTree* tree); // Get the handle, and assert if not found. CORINFO_CLASS_HANDLE gtGetStructHandle(GenTree* tree); // Get the handle for a ref type. CORINFO_CLASS_HANDLE gtGetClassHandle(GenTree* tree, bool* pIsExact, bool* pIsNonNull); // Get the class handle for an helper call CORINFO_CLASS_HANDLE gtGetHelperCallClassHandle(GenTreeCall* call, bool* pIsExact, bool* pIsNonNull); // Get the element handle for an array of ref type. CORINFO_CLASS_HANDLE gtGetArrayElementClassHandle(GenTree* array); // Get a class handle from a helper call argument CORINFO_CLASS_HANDLE gtGetHelperArgClassHandle(GenTree* array, unsigned* runtimeLookupCount = nullptr, GenTree** handleTree = nullptr); // Get the class handle for a field CORINFO_CLASS_HANDLE gtGetFieldClassHandle(CORINFO_FIELD_HANDLE fieldHnd, bool* pIsExact, bool* pIsNonNull); // Check if this tree is a gc static base helper call bool gtIsStaticGCBaseHelperCall(GenTree* tree); //------------------------------------------------------------------------- // Functions to display the trees #ifdef DEBUG void gtDispNode(GenTree* tree, IndentStack* indentStack, __in_z const char* msg, bool isLIR); void gtDispConst(GenTree* tree); void gtDispLeaf(GenTree* tree, IndentStack* indentStack); void gtDispNodeName(GenTree* tree); void gtDispRegVal(GenTree* tree); void gtDispZeroFieldSeq(GenTree* tree); void gtDispVN(GenTree* tree); void gtDispCommonEndLine(GenTree* tree); enum IndentInfo { IINone, IIArc, IIArcTop, IIArcBottom, IIEmbedded, IIError, IndentInfoCount }; void gtDispChild(GenTree* child, IndentStack* indentStack, IndentInfo arcType, __in_opt const char* msg = nullptr, bool topOnly = false); void gtDispTree(GenTree* tree, IndentStack* indentStack = nullptr, __in_opt const char* msg = nullptr, bool topOnly = false, bool isLIR = false); void gtGetLclVarNameInfo(unsigned lclNum, const char** ilKindOut, const char** ilNameOut, unsigned* ilNumOut); int gtGetLclVarName(unsigned lclNum, char* buf, unsigned buf_remaining); char* gtGetLclVarName(unsigned lclNum); void gtDispLclVar(unsigned varNum, bool padForBiggestDisp = true); void gtDispTreeList(GenTree* tree, IndentStack* indentStack = nullptr); void gtGetArgMsg(GenTreeCall* call, GenTree* arg, unsigned argNum, int listCount, char* bufp, unsigned bufLength); void gtGetLateArgMsg(GenTreeCall* call, GenTree* arg, int argNum, int listCount, char* bufp, unsigned bufLength); void gtDispArgList(GenTreeCall* call, IndentStack* indentStack); void gtDispFieldSeq(FieldSeqNode* pfsn); void gtDispRange(LIR::ReadOnlyRange const& range); void gtDispTreeRange(LIR::Range& containingRange, GenTree* tree); void gtDispLIRNode(GenTree* node, const char* prefixMsg = nullptr); #endif // For tree walks enum fgWalkResult { WALK_CONTINUE, WALK_SKIP_SUBTREES, WALK_ABORT }; struct fgWalkData; typedef fgWalkResult(fgWalkPreFn)(GenTree** pTree, fgWalkData* data); typedef fgWalkResult(fgWalkPostFn)(GenTree** pTree, fgWalkData* data); #ifdef DEBUG static fgWalkPreFn gtAssertColonCond; #endif static fgWalkPreFn gtMarkColonCond; static fgWalkPreFn gtClearColonCond; struct FindLinkData { GenTree* nodeToFind; GenTree** result; GenTree* parent; }; FindLinkData gtFindLink(GenTreeStmt* stmt, GenTree* node); bool gtHasCatchArg(GenTree* tree); typedef ArrayStack GenTreeStack; static bool gtHasCallOnStack(GenTreeStack* parentStack); //========================================================================= // BasicBlock functions #ifdef DEBUG // This is a debug flag we will use to assert when creating block during codegen // as this interferes with procedure splitting. If you know what you're doing, set // it to true before creating the block. (DEBUG only) bool fgSafeBasicBlockCreation; #endif BasicBlock* bbNewBasicBlock(BBjumpKinds jumpKind); /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX LclVarsInfo XX XX XX XX The variables to be used by the code generator. XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ // // For both PROMOTION_TYPE_NONE and PROMOTION_TYPE_DEPENDENT the struct will // be placed in the stack frame and it's fields must be laid out sequentially. // // For PROMOTION_TYPE_INDEPENDENT each of the struct's fields is replaced by // a local variable that can be enregistered or placed in the stack frame. // The fields do not need to be laid out sequentially // enum lvaPromotionType { PROMOTION_TYPE_NONE, // The struct local is not promoted PROMOTION_TYPE_INDEPENDENT, // The struct local is promoted, // and its field locals are independent of its parent struct local. PROMOTION_TYPE_DEPENDENT // The struct local is promoted, // but its field locals depend on its parent struct local. }; static int __cdecl RefCntCmp(const void* op1, const void* op2); static int __cdecl WtdRefCntCmp(const void* op1, const void* op2); /*****************************************************************************/ enum FrameLayoutState { NO_FRAME_LAYOUT, INITIAL_FRAME_LAYOUT, PRE_REGALLOC_FRAME_LAYOUT, REGALLOC_FRAME_LAYOUT, TENTATIVE_FRAME_LAYOUT, FINAL_FRAME_LAYOUT }; public: RefCountState lvaRefCountState; // Current local ref count state bool lvaLocalVarRefCounted() const { return lvaRefCountState == RCS_NORMAL; } bool lvaTrackedFixed; // true: We cannot add new 'tracked' variable unsigned lvaCount; // total number of locals unsigned lvaRefCount; // total number of references to locals LclVarDsc* lvaTable; // variable descriptor table unsigned lvaTableCnt; // lvaTable size (>= lvaCount) LclVarDsc** lvaRefSorted; // table sorted by refcount unsigned short lvaTrackedCount; // actual # of locals being tracked unsigned lvaTrackedCountInSizeTUnits; // min # of size_t's sufficient to hold a bit for all the locals being tracked #ifdef DEBUG VARSET_TP lvaTrackedVars; // set of tracked variables #endif #ifndef _TARGET_64BIT_ VARSET_TP lvaLongVars; // set of long (64-bit) variables #endif VARSET_TP lvaFloatVars; // set of floating-point (32-bit and 64-bit) variables unsigned lvaCurEpoch; // VarSets are relative to a specific set of tracked var indices. // It that changes, this changes. VarSets from different epochs // cannot be meaningfully combined. unsigned GetCurLVEpoch() { return lvaCurEpoch; } // reverse map of tracked number to var number unsigned* lvaTrackedToVarNum; #if DOUBLE_ALIGN #ifdef DEBUG // # of procs compiled a with double-aligned stack static unsigned s_lvaDoubleAlignedProcsCount; #endif #endif // Getters and setters for address-exposed and do-not-enregister local var properties. bool lvaVarAddrExposed(unsigned varNum); void lvaSetVarAddrExposed(unsigned varNum); bool lvaVarDoNotEnregister(unsigned varNum); #ifdef DEBUG // Reasons why we can't enregister. Some of these correspond to debug properties of local vars. enum DoNotEnregisterReason { DNER_AddrExposed, DNER_IsStruct, DNER_LocalField, DNER_VMNeedsStackAddr, DNER_LiveInOutOfHandler, DNER_LiveAcrossUnmanagedCall, DNER_BlockOp, // Is read or written via a block operation that explicitly takes the address. DNER_IsStructArg, // Is a struct passed as an argument in a way that requires a stack location. DNER_DepField, // It is a field of a dependently promoted struct DNER_NoRegVars, // opts.compFlags & CLFLG_REGVAR is not set DNER_MinOptsGC, // It is a GC Ref and we are compiling MinOpts #if !defined(_TARGET_64BIT_) DNER_LongParamField, // It is a decomposed field of a long parameter. #endif #ifdef JIT32_GCENCODER DNER_PinningRef, #endif }; #endif void lvaSetVarDoNotEnregister(unsigned varNum DEBUGARG(DoNotEnregisterReason reason)); unsigned lvaVarargsHandleArg; #ifdef _TARGET_X86_ unsigned lvaVarargsBaseOfStkArgs; // Pointer (computed based on incoming varargs handle) to the start of the stack // arguments #endif // _TARGET_X86_ unsigned lvaInlinedPInvokeFrameVar; // variable representing the InlinedCallFrame unsigned lvaReversePInvokeFrameVar; // variable representing the reverse PInvoke frame #if FEATURE_FIXED_OUT_ARGS unsigned lvaPInvokeFrameRegSaveVar; // variable representing the RegSave for PInvoke inlining. #endif unsigned lvaMonAcquired; // boolean variable introduced into in synchronized methods // that tracks whether the lock has been taken unsigned lvaArg0Var; // The lclNum of arg0. Normally this will be info.compThisArg. // However, if there is a "ldarga 0" or "starg 0" in the IL, // we will redirect all "ldarg(a) 0" and "starg 0" to this temp. unsigned lvaInlineeReturnSpillTemp; // The temp to spill the non-VOID return expression // in case there are multiple BBJ_RETURN blocks in the inlinee // or if the inlinee has GC ref locals. #if FEATURE_FIXED_OUT_ARGS unsigned lvaOutgoingArgSpaceVar; // dummy TYP_LCLBLK var for fixed outgoing argument space PhasedVar lvaOutgoingArgSpaceSize; // size of fixed outgoing argument space #endif // FEATURE_FIXED_OUT_ARGS #ifdef _TARGET_ARM_ // On architectures whose ABIs allow structs to be passed in registers, struct promotion will sometimes // require us to "rematerialize" a struct from it's separate constituent field variables. Packing several sub-word // field variables into an argument register is a hard problem. It's easier to reserve a word of memory into which // such field can be copied, after which the assembled memory word can be read into the register. We will allocate // this variable to be this scratch word whenever struct promotion occurs. unsigned lvaPromotedStructAssemblyScratchVar; #endif // _TARGET_ARM_ #if defined(DEBUG) && defined(_TARGET_XARCH_) unsigned lvaReturnSpCheck; // Stores SP to confirm it is not corrupted on return. #endif // defined(DEBUG) && defined(_TARGET_XARCH_) #if defined(DEBUG) && defined(_TARGET_X86_) unsigned lvaCallSpCheck; // Stores SP to confirm it is not corrupted after every call. #endif // defined(DEBUG) && defined(_TARGET_X86_) unsigned lvaGenericsContextUseCount; bool lvaKeepAliveAndReportThis(); // Synchronized instance method of a reference type, or // CORINFO_GENERICS_CTXT_FROM_THIS? bool lvaReportParamTypeArg(); // Exceptions and CORINFO_GENERICS_CTXT_FROM_PARAMTYPEARG? //------------------------------------------------------------------------- // All these frame offsets are inter-related and must be kept in sync #if !FEATURE_EH_FUNCLETS // This is used for the callable handlers unsigned lvaShadowSPslotsVar; // TYP_BLK variable for all the shadow SP slots #endif // FEATURE_EH_FUNCLETS int lvaCachedGenericContextArgOffs; int lvaCachedGenericContextArgOffset(); // For CORINFO_CALLCONV_PARAMTYPE and if generic context is passed as // THIS pointer #ifdef JIT32_GCENCODER unsigned lvaLocAllocSPvar; // variable which stores the value of ESP after the the last alloca/localloc #endif // JIT32_GCENCODER unsigned lvaNewObjArrayArgs; // variable with arguments for new MD array helper // TODO-Review: Prior to reg predict we reserve 24 bytes for Spill temps. // after the reg predict we will use a computed maxTmpSize // which is based upon the number of spill temps predicted by reg predict // All this is necessary because if we under-estimate the size of the spill // temps we could fail when encoding instructions that reference stack offsets for ARM. // // Pre codegen max spill temp size. static const unsigned MAX_SPILL_TEMP_SIZE = 24; //------------------------------------------------------------------------- unsigned lvaGetMaxSpillTempSize(); #ifdef _TARGET_ARM_ bool lvaIsPreSpilled(unsigned lclNum, regMaskTP preSpillMask); #endif // _TARGET_ARM_ void lvaAssignFrameOffsets(FrameLayoutState curState); void lvaFixVirtualFrameOffsets(); void lvaUpdateArgsWithInitialReg(); void lvaAssignVirtualFrameOffsetsToArgs(); #ifdef UNIX_AMD64_ABI int lvaAssignVirtualFrameOffsetToArg(unsigned lclNum, unsigned argSize, int argOffs, int* callerArgOffset); #else // !UNIX_AMD64_ABI int lvaAssignVirtualFrameOffsetToArg(unsigned lclNum, unsigned argSize, int argOffs); #endif // !UNIX_AMD64_ABI void lvaAssignVirtualFrameOffsetsToLocals(); int lvaAllocLocalAndSetVirtualOffset(unsigned lclNum, unsigned size, int stkOffs); #ifdef _TARGET_AMD64_ // Returns true if compCalleeRegsPushed (including RBP if used as frame pointer) is even. bool lvaIsCalleeSavedIntRegCountEven(); #endif void lvaAlignFrame(); void lvaAssignFrameOffsetsToPromotedStructs(); int lvaAllocateTemps(int stkOffs, bool mustDoubleAlign); #ifdef DEBUG void lvaDumpRegLocation(unsigned lclNum); void lvaDumpFrameLocation(unsigned lclNum); void lvaDumpEntry(unsigned lclNum, FrameLayoutState curState, size_t refCntWtdWidth = 6); void lvaTableDump(FrameLayoutState curState = NO_FRAME_LAYOUT); // NO_FRAME_LAYOUT means use the current frame // layout state defined by lvaDoneFrameLayout #endif // Limit frames size to 1GB. The maximum is 2GB in theory - make it intentionally smaller // to avoid bugs from borderline cases. #define MAX_FrameSize 0x3FFFFFFF void lvaIncrementFrameSize(unsigned size); unsigned lvaFrameSize(FrameLayoutState curState); // Returns the caller-SP-relative offset for the SP/FP relative offset determined by FP based. int lvaToCallerSPRelativeOffset(int offs, bool isFpBased) const; // Returns the caller-SP-relative offset for the local variable "varNum." int lvaGetCallerSPRelativeOffset(unsigned varNum); // Returns the SP-relative offset for the local variable "varNum". Illegal to ask this for functions with localloc. int lvaGetSPRelativeOffset(unsigned varNum); int lvaToInitialSPRelativeOffset(unsigned offset, bool isFpBased); int lvaGetInitialSPRelativeOffset(unsigned varNum); //------------------------ For splitting types ---------------------------- void lvaInitTypeRef(); void lvaInitArgs(InitVarDscInfo* varDscInfo); void lvaInitThisPtr(InitVarDscInfo* varDscInfo); void lvaInitRetBuffArg(InitVarDscInfo* varDscInfo); void lvaInitUserArgs(InitVarDscInfo* varDscInfo); void lvaInitGenericsCtxt(InitVarDscInfo* varDscInfo); void lvaInitVarArgsHandle(InitVarDscInfo* varDscInfo); void lvaInitVarDsc(LclVarDsc* varDsc, unsigned varNum, CorInfoType corInfoType, CORINFO_CLASS_HANDLE typeHnd, CORINFO_ARG_LIST_HANDLE varList, CORINFO_SIG_INFO* varSig); static unsigned lvaTypeRefMask(var_types type); var_types lvaGetActualType(unsigned lclNum); var_types lvaGetRealType(unsigned lclNum); //------------------------------------------------------------------------- void lvaInit(); LclVarDsc* lvaGetDesc(unsigned lclNum) { assert(lclNum < lvaCount); return &lvaTable[lclNum]; } LclVarDsc* lvaGetDesc(GenTreeLclVarCommon* lclVar) { assert(lclVar->GetLclNum() < lvaCount); return &lvaTable[lclVar->GetLclNum()]; } unsigned lvaLclSize(unsigned varNum); unsigned lvaLclExactSize(unsigned varNum); bool lvaHaveManyLocals() const; unsigned lvaGrabTemp(bool shortLifetime DEBUGARG(const char* reason)); unsigned lvaGrabTemps(unsigned cnt DEBUGARG(const char* reason)); unsigned lvaGrabTempWithImplicitUse(bool shortLifetime DEBUGARG(const char* reason)); void lvaSortOnly(); void lvaSortByRefCount(); void lvaDumpRefCounts(); void lvaMarkLocalVars(); // Local variable ref-counting void lvaComputeRefCounts(bool isRecompute, bool setSlotNumbers); void lvaMarkLocalVars(BasicBlock* block, bool isRecompute); void lvaAllocOutgoingArgSpaceVar(); // Set up lvaOutgoingArgSpaceVar VARSET_VALRET_TP lvaStmtLclMask(GenTreeStmt* stmt); #ifdef DEBUG struct lvaStressLclFldArgs { Compiler* m_pCompiler; bool m_bFirstPass; }; static fgWalkPreFn lvaStressLclFldCB; void lvaStressLclFld(); void lvaDispVarSet(VARSET_VALARG_TP set, VARSET_VALARG_TP allVars); void lvaDispVarSet(VARSET_VALARG_TP set); #endif #ifdef _TARGET_ARM_ int lvaFrameAddress(int varNum, bool mustBeFPBased, regNumber* pBaseReg, int addrModeOffset, bool isFloatUsage); #else int lvaFrameAddress(int varNum, bool* pFPbased); #endif bool lvaIsParameter(unsigned varNum); bool lvaIsRegArgument(unsigned varNum); BOOL lvaIsOriginalThisArg(unsigned varNum); // Is this varNum the original this argument? BOOL lvaIsOriginalThisReadOnly(); // return TRUE if there is no place in the code // that writes to arg0 // For x64 this is 3, 5, 6, 7, >8 byte structs that are passed by reference. // For ARM64, this is structs larger than 16 bytes that are passed by reference. bool lvaIsImplicitByRefLocal(unsigned varNum) { #if defined(_TARGET_AMD64_) || defined(_TARGET_ARM64_) LclVarDsc* varDsc = lvaGetDesc(varNum); if (varDsc->lvIsImplicitByRef) { assert(varDsc->lvIsParam); assert(varTypeIsStruct(varDsc) || (varDsc->lvType == TYP_BYREF)); return true; } #endif // defined(_TARGET_AMD64_) || defined(_TARGET_ARM64_) return false; } // Returns true if this local var is a multireg struct bool lvaIsMultiregStruct(LclVarDsc* varDsc, bool isVararg); // If the local is a TYP_STRUCT, get/set a class handle describing it CORINFO_CLASS_HANDLE lvaGetStruct(unsigned varNum); void lvaSetStruct(unsigned varNum, CORINFO_CLASS_HANDLE typeHnd, bool unsafeValueClsCheck, bool setTypeInfo = true); void lvaSetStructUsedAsVarArg(unsigned varNum); // If the local is TYP_REF, set or update the associated class information. void lvaSetClass(unsigned varNum, CORINFO_CLASS_HANDLE clsHnd, bool isExact = false); void lvaSetClass(unsigned varNum, GenTree* tree, CORINFO_CLASS_HANDLE stackHandle = nullptr); void lvaUpdateClass(unsigned varNum, CORINFO_CLASS_HANDLE clsHnd, bool isExact = false); void lvaUpdateClass(unsigned varNum, GenTree* tree, CORINFO_CLASS_HANDLE stackHandle = nullptr); #define MAX_NumOfFieldsInPromotableStruct 4 // Maximum number of fields in promotable struct // Info about struct type fields. struct lvaStructFieldInfo { CORINFO_FIELD_HANDLE fldHnd; unsigned char fldOffset; unsigned char fldOrdinal; var_types fldType; unsigned fldSize; CORINFO_CLASS_HANDLE fldTypeHnd; lvaStructFieldInfo() : fldHnd(nullptr), fldOffset(0), fldOrdinal(0), fldType(TYP_UNDEF), fldSize(0), fldTypeHnd(nullptr) { } }; // Info about a struct type, instances of which may be candidates for promotion. struct lvaStructPromotionInfo { CORINFO_CLASS_HANDLE typeHnd; bool canPromote; bool containsHoles; bool customLayout; bool fieldsSorted; unsigned char fieldCnt; lvaStructFieldInfo fields[MAX_NumOfFieldsInPromotableStruct]; lvaStructPromotionInfo(CORINFO_CLASS_HANDLE typeHnd = nullptr) : typeHnd(typeHnd) , canPromote(false) , containsHoles(false) , customLayout(false) , fieldsSorted(false) , fieldCnt(0) { } }; static int __cdecl lvaFieldOffsetCmp(const void* field1, const void* field2); // This class is responsible for checking validity and profitability of struct promotion. // If it is both legal and profitable, then TryPromoteStructVar promotes the struct and initializes // nessesary information for fgMorphStructField to use. class StructPromotionHelper { public: StructPromotionHelper(Compiler* compiler); bool CanPromoteStructType(CORINFO_CLASS_HANDLE typeHnd); bool TryPromoteStructVar(unsigned lclNum); #ifdef DEBUG void CheckRetypedAsScalar(CORINFO_FIELD_HANDLE fieldHnd, var_types requestedType); #endif // DEBUG #ifdef _TARGET_ARM_ bool GetRequiresScratchVar(); #endif // _TARGET_ARM_ private: bool CanPromoteStructVar(unsigned lclNum); bool ShouldPromoteStructVar(unsigned lclNum); void PromoteStructVar(unsigned lclNum); void SortStructFields(); lvaStructFieldInfo GetFieldInfo(CORINFO_FIELD_HANDLE fieldHnd, BYTE ordinal); bool TryPromoteStructField(lvaStructFieldInfo& outerFieldInfo); private: Compiler* compiler; lvaStructPromotionInfo structPromotionInfo; #ifdef _TARGET_ARM_ bool requiresScratchVar; #endif // _TARGET_ARM_ #ifdef DEBUG typedef JitHashTable, var_types> RetypedAsScalarFieldsMap; RetypedAsScalarFieldsMap retypedFieldsMap; #endif // DEBUG }; StructPromotionHelper* structPromotionHelper; #if !defined(_TARGET_64BIT_) void lvaPromoteLongVars(); #endif // !defined(_TARGET_64BIT_) unsigned lvaGetFieldLocal(const LclVarDsc* varDsc, unsigned int fldOffset); lvaPromotionType lvaGetPromotionType(const LclVarDsc* varDsc); lvaPromotionType lvaGetPromotionType(unsigned varNum); lvaPromotionType lvaGetParentPromotionType(const LclVarDsc* varDsc); lvaPromotionType lvaGetParentPromotionType(unsigned varNum); bool lvaIsFieldOfDependentlyPromotedStruct(const LclVarDsc* varDsc); bool lvaIsGCTracked(const LclVarDsc* varDsc); #if defined(FEATURE_SIMD) bool lvaMapSimd12ToSimd16(const LclVarDsc* varDsc) { assert(varDsc->lvType == TYP_SIMD12); assert(varDsc->lvExactSize == 12); #if defined(_TARGET_64BIT_) assert(varDsc->lvSize() == 16); #endif // defined(_TARGET_64BIT_) // We make local variable SIMD12 types 16 bytes instead of just 12. lvSize() // already does this calculation. However, we also need to prevent mapping types if the var is a // dependently promoted struct field, which must remain its exact size within its parent struct. // However, we don't know this until late, so we may have already pretended the field is bigger // before that. if ((varDsc->lvSize() == 16) && !lvaIsFieldOfDependentlyPromotedStruct(varDsc)) { return true; } else { return false; } } #endif // defined(FEATURE_SIMD) BYTE* lvaGetGcLayout(unsigned varNum); bool lvaTypeIsGC(unsigned varNum); unsigned lvaGSSecurityCookie; // LclVar number bool lvaTempsHaveLargerOffsetThanVars(); // Returns "true" iff local variable "lclNum" is in SSA form. bool lvaInSsa(unsigned lclNum) { assert(lclNum < lvaCount); return lvaTable[lclNum].lvInSsa; } unsigned lvaSecurityObject; // variable representing the security object on the stack unsigned lvaStubArgumentVar; // variable representing the secret stub argument coming in EAX #if FEATURE_EH_FUNCLETS unsigned lvaPSPSym; // variable representing the PSPSym #endif InlineInfo* impInlineInfo; InlineStrategy* m_inlineStrategy; // The Compiler* that is the root of the inlining tree of which "this" is a member. Compiler* impInlineRoot(); #if defined(DEBUG) || defined(INLINE_DATA) unsigned __int64 getInlineCycleCount() { return m_compCycles; } #endif // defined(DEBUG) || defined(INLINE_DATA) bool fgNoStructPromotion; // Set to TRUE to turn off struct promotion for this method. bool fgNoStructParamPromotion; // Set to TRUE to turn off struct promotion for parameters this method. //========================================================================= // PROTECTED //========================================================================= protected: //---------------- Local variable ref-counting ---------------------------- void lvaMarkLclRefs(GenTree* tree, BasicBlock* block, GenTreeStmt* stmt, bool isRecompute); bool IsDominatedByExceptionalEntry(BasicBlock* block); void SetVolatileHint(LclVarDsc* varDsc); // Keeps the mapping from SSA #'s to VN's for the implicit memory variables. SsaDefArray lvMemoryPerSsaData; public: // Returns the address of the per-Ssa data for memory at the given ssaNum (which is required // not to be the SsaConfig::RESERVED_SSA_NUM, which indicates that the variable is // not an SSA variable). SsaMemDef* GetMemoryPerSsaData(unsigned ssaNum) { return lvMemoryPerSsaData.GetSsaDef(ssaNum); } /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX Importer XX XX XX XX Imports the given method and converts it to semantic trees XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: void impInit(); void impImport(BasicBlock* method); CORINFO_CLASS_HANDLE impGetRefAnyClass(); CORINFO_CLASS_HANDLE impGetRuntimeArgumentHandle(); CORINFO_CLASS_HANDLE impGetTypeHandleClass(); CORINFO_CLASS_HANDLE impGetStringClass(); CORINFO_CLASS_HANDLE impGetObjectClass(); // Returns underlying type of handles returned by ldtoken instruction var_types GetRuntimeHandleUnderlyingType() { // RuntimeTypeHandle is backed by raw pointer on CoreRT and by object reference on other runtimes return IsTargetAbi(CORINFO_CORERT_ABI) ? TYP_I_IMPL : TYP_REF; } void impDevirtualizeCall(GenTreeCall* call, CORINFO_METHOD_HANDLE* method, unsigned* methodFlags, CORINFO_CONTEXT_HANDLE* contextHandle, CORINFO_CONTEXT_HANDLE* exactContextHandle, bool isLateDevirtualization, bool isExplicitTailCall); //========================================================================= // PROTECTED //========================================================================= protected: //-------------------- Stack manipulation --------------------------------- unsigned impStkSize; // Size of the full stack #define SMALL_STACK_SIZE 16 // number of elements in impSmallStack struct SavedStack // used to save/restore stack contents. { unsigned ssDepth; // number of values on stack StackEntry* ssTrees; // saved tree values }; bool impIsPrimitive(CorInfoType type); bool impILConsumesAddr(const BYTE* codeAddr, CORINFO_METHOD_HANDLE fncHandle, CORINFO_MODULE_HANDLE scpHandle); void impResolveToken(const BYTE* addr, CORINFO_RESOLVED_TOKEN* pResolvedToken, CorInfoTokenKind kind); void impPushOnStack(GenTree* tree, typeInfo ti); void impPushNullObjRefOnStack(); StackEntry impPopStack(); StackEntry& impStackTop(unsigned n = 0); unsigned impStackHeight(); void impSaveStackState(SavedStack* savePtr, bool copy); void impRestoreStackState(SavedStack* savePtr); GenTree* impImportLdvirtftn(GenTree* thisPtr, CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_CALL_INFO* pCallInfo); int impBoxPatternMatch(CORINFO_RESOLVED_TOKEN* pResolvedToken, const BYTE* codeAddr, const BYTE* codeEndp); void impImportAndPushBox(CORINFO_RESOLVED_TOKEN* pResolvedToken); void impImportNewObjArray(CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_CALL_INFO* pCallInfo); bool impCanPInvokeInline(); bool impCanPInvokeInlineCallSite(BasicBlock* block); void impCheckForPInvokeCall( GenTreeCall* call, CORINFO_METHOD_HANDLE methHnd, CORINFO_SIG_INFO* sig, unsigned mflags, BasicBlock* block); GenTreeCall* impImportIndirectCall(CORINFO_SIG_INFO* sig, IL_OFFSETX ilOffset = BAD_IL_OFFSET); void impPopArgsForUnmanagedCall(GenTree* call, CORINFO_SIG_INFO* sig); void impInsertHelperCall(CORINFO_HELPER_DESC* helperCall); void impHandleAccessAllowed(CorInfoIsAccessAllowedResult result, CORINFO_HELPER_DESC* helperCall); void impHandleAccessAllowedInternal(CorInfoIsAccessAllowedResult result, CORINFO_HELPER_DESC* helperCall); var_types impImportCall(OPCODE opcode, CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_RESOLVED_TOKEN* pConstrainedResolvedToken, // Is this a "constrained." call on a // type parameter? GenTree* newobjThis, int prefixFlags, CORINFO_CALL_INFO* callInfo, IL_OFFSET rawILOffset); CORINFO_CLASS_HANDLE impGetSpecialIntrinsicExactReturnType(CORINFO_METHOD_HANDLE specialIntrinsicHandle); bool impMethodInfo_hasRetBuffArg(CORINFO_METHOD_INFO* methInfo); GenTree* impFixupCallStructReturn(GenTreeCall* call, CORINFO_CLASS_HANDLE retClsHnd); GenTree* impFixupStructReturnType(GenTree* op, CORINFO_CLASS_HANDLE retClsHnd); #ifdef DEBUG var_types impImportJitTestLabelMark(int numArgs); #endif // DEBUG GenTree* impInitClass(CORINFO_RESOLVED_TOKEN* pResolvedToken); GenTree* impImportStaticReadOnlyField(void* fldAddr, var_types lclTyp); GenTree* impImportStaticFieldAccess(CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_ACCESS_FLAGS access, CORINFO_FIELD_INFO* pFieldInfo, var_types lclTyp); static void impBashVarAddrsToI(GenTree* tree1, GenTree* tree2 = nullptr); GenTree* impImplicitIorI4Cast(GenTree* tree, var_types dstTyp); GenTree* impImplicitR4orR8Cast(GenTree* tree, var_types dstTyp); void impImportLeave(BasicBlock* block); void impResetLeaveBlock(BasicBlock* block, unsigned jmpAddr); GenTree* impIntrinsic(GenTree* newobjThis, CORINFO_CLASS_HANDLE clsHnd, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, unsigned methodFlags, int memberRef, bool readonlyCall, bool tailCall, CORINFO_RESOLVED_TOKEN* pContstrainedResolvedToken, CORINFO_THIS_TRANSFORM constraintCallThisTransform, CorInfoIntrinsics* pIntrinsicID, bool* isSpecialIntrinsic = nullptr); GenTree* impMathIntrinsic(CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, var_types callType, CorInfoIntrinsics intrinsicID, bool tailCall); NamedIntrinsic lookupNamedIntrinsic(CORINFO_METHOD_HANDLE method); #ifdef FEATURE_HW_INTRINSICS GenTree* impHWIntrinsic(NamedIntrinsic intrinsic, CORINFO_CLASS_HANDLE clsHnd, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impUnsupportedHWIntrinsic(unsigned helper, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); protected: bool compSupportsHWIntrinsic(InstructionSet isa); #ifdef _TARGET_XARCH_ GenTree* impBaseIntrinsic(NamedIntrinsic intrinsic, CORINFO_CLASS_HANDLE clsHnd, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impSSEIntrinsic(NamedIntrinsic intrinsic, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impSSE2Intrinsic(NamedIntrinsic intrinsic, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impSSE42Intrinsic(NamedIntrinsic intrinsic, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impAvxOrAvx2Intrinsic(NamedIntrinsic intrinsic, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impAESIntrinsic(NamedIntrinsic intrinsic, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impBMI1OrBMI2Intrinsic(NamedIntrinsic intrinsic, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impFMAIntrinsic(NamedIntrinsic intrinsic, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impLZCNTIntrinsic(NamedIntrinsic intrinsic, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impPCLMULQDQIntrinsic(NamedIntrinsic intrinsic, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); GenTree* impPOPCNTIntrinsic(NamedIntrinsic intrinsic, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, bool mustExpand); protected: GenTree* getArgForHWIntrinsic(var_types argType, CORINFO_CLASS_HANDLE argClass); GenTree* impNonConstFallback(NamedIntrinsic intrinsic, var_types simdType, var_types baseType); GenTree* addRangeCheckIfNeeded(NamedIntrinsic intrinsic, GenTree* lastOp, bool mustExpand); #endif // _TARGET_XARCH_ #ifdef _TARGET_ARM64_ InstructionSet lookupHWIntrinsicISA(const char* className); NamedIntrinsic lookupHWIntrinsic(const char* className, const char* methodName); GenTree* addRangeCheckIfNeeded(GenTree* lastOp, unsigned int max, bool mustExpand); #endif // _TARGET_ARM64_ #endif // FEATURE_HW_INTRINSICS GenTree* impArrayAccessIntrinsic(CORINFO_CLASS_HANDLE clsHnd, CORINFO_SIG_INFO* sig, int memberRef, bool readonlyCall, CorInfoIntrinsics intrinsicID); GenTree* impInitializeArrayIntrinsic(CORINFO_SIG_INFO* sig); GenTree* impMethodPointer(CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_CALL_INFO* pCallInfo); GenTree* impTransformThis(GenTree* thisPtr, CORINFO_RESOLVED_TOKEN* pConstrainedResolvedToken, CORINFO_THIS_TRANSFORM transform); //----------------- Manipulating the trees and stmts ---------------------- GenTreeStmt* impStmtList; // Statements for the BB being imported. GenTreeStmt* impLastStmt; // The last statement for the current BB. public: enum { CHECK_SPILL_ALL = -1, CHECK_SPILL_NONE = -2 }; void impBeginTreeList(); void impEndTreeList(BasicBlock* block, GenTreeStmt* firstStmt, GenTreeStmt* lastStmt); void impEndTreeList(BasicBlock* block); void impAppendStmtCheck(GenTreeStmt* stmt, unsigned chkLevel); void impAppendStmt(GenTreeStmt* stmt, unsigned chkLevel); void impAppendStmt(GenTreeStmt* stmt); void impInsertStmtBefore(GenTreeStmt* stmt, GenTreeStmt* stmtBefore); GenTreeStmt* impAppendTree(GenTree* tree, unsigned chkLevel, IL_OFFSETX offset); void impInsertTreeBefore(GenTree* tree, IL_OFFSETX offset, GenTreeStmt* stmtBefore); void impAssignTempGen(unsigned tmp, GenTree* val, unsigned curLevel, GenTreeStmt** pAfterStmt = nullptr, IL_OFFSETX ilOffset = BAD_IL_OFFSET, BasicBlock* block = nullptr); void impAssignTempGen(unsigned tmpNum, GenTree* val, CORINFO_CLASS_HANDLE structHnd, unsigned curLevel, GenTreeStmt** pAfterStmt = nullptr, IL_OFFSETX ilOffset = BAD_IL_OFFSET, BasicBlock* block = nullptr); GenTreeStmt* impExtractLastStmt(); GenTree* impCloneExpr(GenTree* tree, GenTree** clone, CORINFO_CLASS_HANDLE structHnd, unsigned curLevel, GenTreeStmt** pAfterStmt DEBUGARG(const char* reason)); GenTree* impAssignStruct(GenTree* dest, GenTree* src, CORINFO_CLASS_HANDLE structHnd, unsigned curLevel, GenTreeStmt** pAfterStmt = nullptr, IL_OFFSETX ilOffset = BAD_IL_OFFSET, BasicBlock* block = nullptr); GenTree* impAssignStructPtr(GenTree* dest, GenTree* src, CORINFO_CLASS_HANDLE structHnd, unsigned curLevel, GenTreeStmt** pAfterStmt = nullptr, IL_OFFSETX ilOffset = BAD_IL_OFFSET, BasicBlock* block = nullptr); GenTree* impGetStructAddr(GenTree* structVal, CORINFO_CLASS_HANDLE structHnd, unsigned curLevel, bool willDeref); var_types impNormStructType(CORINFO_CLASS_HANDLE structHnd, BYTE* gcLayout = nullptr, unsigned* numGCVars = nullptr, var_types* simdBaseType = nullptr); GenTree* impNormStructVal(GenTree* structVal, CORINFO_CLASS_HANDLE structHnd, unsigned curLevel, bool forceNormalization = false); GenTree* impTokenToHandle(CORINFO_RESOLVED_TOKEN* pResolvedToken, BOOL* pRuntimeLookup = nullptr, BOOL mustRestoreHandle = FALSE, BOOL importParent = FALSE); GenTree* impParentClassTokenToHandle(CORINFO_RESOLVED_TOKEN* pResolvedToken, BOOL* pRuntimeLookup = nullptr, BOOL mustRestoreHandle = FALSE) { return impTokenToHandle(pResolvedToken, pRuntimeLookup, mustRestoreHandle, TRUE); } GenTree* impLookupToTree(CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_LOOKUP* pLookup, unsigned flags, void* compileTimeHandle); GenTree* getRuntimeContextTree(CORINFO_RUNTIME_LOOKUP_KIND kind); GenTree* impRuntimeLookupToTree(CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_LOOKUP* pLookup, void* compileTimeHandle); GenTree* impReadyToRunLookupToTree(CORINFO_CONST_LOOKUP* pLookup, unsigned flags, void* compileTimeHandle); GenTreeCall* impReadyToRunHelperToTree(CORINFO_RESOLVED_TOKEN* pResolvedToken, CorInfoHelpFunc helper, var_types type, GenTreeArgList* arg = nullptr, CORINFO_LOOKUP_KIND* pGenericLookupKind = nullptr); GenTree* impCastClassOrIsInstToTree(GenTree* op1, GenTree* op2, CORINFO_RESOLVED_TOKEN* pResolvedToken, bool isCastClass); GenTree* impOptimizeCastClassOrIsInst(GenTree* op1, CORINFO_RESOLVED_TOKEN* pResolvedToken, bool isCastClass); bool VarTypeIsMultiByteAndCanEnreg( var_types type, CORINFO_CLASS_HANDLE typeClass, unsigned* typeSize, bool forReturn, bool isVarArg); bool IsIntrinsicImplementedByUserCall(CorInfoIntrinsics intrinsicId); bool IsTargetIntrinsic(CorInfoIntrinsics intrinsicId); bool IsMathIntrinsic(CorInfoIntrinsics intrinsicId); bool IsMathIntrinsic(GenTree* tree); private: //----------------- Importing the method ---------------------------------- CORINFO_CONTEXT_HANDLE impTokenLookupContextHandle; // The context used for looking up tokens. #ifdef DEBUG unsigned impCurOpcOffs; const char* impCurOpcName; bool impNestedStackSpill; // For displaying instrs with generated native code (-n:B) GenTreeStmt* impLastILoffsStmt; // oldest stmt added for which we did not gtStmtLastILoffs void impNoteLastILoffs(); #endif /* IL offset of the stmt currently being imported. It gets set to BAD_IL_OFFSET after it has been set in the appended trees. Then it gets updated at IL offsets for which we have to report mapping info. It also includes flag bits, so use jitGetILoffs() to get the actual IL offset value. */ IL_OFFSETX impCurStmtOffs; void impCurStmtOffsSet(IL_OFFSET offs); void impNoteBranchOffs(); unsigned impInitBlockLineInfo(); GenTree* impCheckForNullPointer(GenTree* obj); bool impIsThis(GenTree* obj); bool impIsLDFTN_TOKEN(const BYTE* delegateCreateStart, const BYTE* newobjCodeAddr); bool impIsDUP_LDVIRTFTN_TOKEN(const BYTE* delegateCreateStart, const BYTE* newobjCodeAddr); bool impIsAnySTLOC(OPCODE opcode) { return ((opcode == CEE_STLOC) || (opcode == CEE_STLOC_S) || ((opcode >= CEE_STLOC_0) && (opcode <= CEE_STLOC_3))); } GenTreeArgList* impPopList(unsigned count, CORINFO_SIG_INFO* sig, GenTreeArgList* prefixTree = nullptr); GenTreeArgList* impPopRevList(unsigned count, CORINFO_SIG_INFO* sig, unsigned skipReverseCount = 0); /* * Get current IL offset with stack-empty info incoporated */ IL_OFFSETX impCurILOffset(IL_OFFSET offs, bool callInstruction = false); //---------------- Spilling the importer stack ---------------------------- // The maximum number of bytes of IL processed without clean stack state. // It allows to limit the maximum tree size and depth. static const unsigned MAX_TREE_SIZE = 200; bool impCanSpillNow(OPCODE prevOpcode); struct PendingDsc { PendingDsc* pdNext; BasicBlock* pdBB; SavedStack pdSavedStack; ThisInitState pdThisPtrInit; }; PendingDsc* impPendingList; // list of BBs currently waiting to be imported. PendingDsc* impPendingFree; // Freed up dscs that can be reused // We keep a byte-per-block map (dynamically extended) in the top-level Compiler object of a compilation. JitExpandArray impPendingBlockMembers; // Return the byte for "b" (allocating/extending impPendingBlockMembers if necessary.) // Operates on the map in the top-level ancestor. BYTE impGetPendingBlockMember(BasicBlock* blk) { return impInlineRoot()->impPendingBlockMembers.Get(blk->bbInd()); } // Set the byte for "b" to "val" (allocating/extending impPendingBlockMembers if necessary.) // Operates on the map in the top-level ancestor. void impSetPendingBlockMember(BasicBlock* blk, BYTE val) { impInlineRoot()->impPendingBlockMembers.Set(blk->bbInd(), val); } bool impCanReimport; bool impSpillStackEntry(unsigned level, unsigned varNum #ifdef DEBUG , bool bAssertOnRecursion, const char* reason #endif ); void impSpillStackEnsure(bool spillLeaves = false); void impEvalSideEffects(); void impSpillSpecialSideEff(); void impSpillSideEffects(bool spillGlobEffects, unsigned chkLevel DEBUGARG(const char* reason)); void impSpillValueClasses(); void impSpillEvalStack(); static fgWalkPreFn impFindValueClasses; void impSpillLclRefs(ssize_t lclNum); BasicBlock* impPushCatchArgOnStack(BasicBlock* hndBlk, CORINFO_CLASS_HANDLE clsHnd, bool isSingleBlockFilter); void impImportBlockCode(BasicBlock* block); void impReimportMarkBlock(BasicBlock* block); void impReimportMarkSuccessors(BasicBlock* block); void impVerifyEHBlock(BasicBlock* block, bool isTryStart); void impImportBlockPending(BasicBlock* block); // Similar to impImportBlockPending, but assumes that block has already been imported once and is being // reimported for some reason. It specifically does *not* look at verCurrentState to set the EntryState // for the block, but instead, just re-uses the block's existing EntryState. void impReimportBlockPending(BasicBlock* block); var_types impGetByRefResultType(genTreeOps oper, bool fUnsigned, GenTree** pOp1, GenTree** pOp2); void impImportBlock(BasicBlock* block); // Assumes that "block" is a basic block that completes with a non-empty stack. We will assign the values // on the stack to local variables (the "spill temp" variables). The successor blocks will assume that // its incoming stack contents are in those locals. This requires "block" and its successors to agree on // the variables that will be used -- and for all the predecessors of those successors, and the // successors of those predecessors, etc. Call such a set of blocks closed under alternating // successor/predecessor edges a "spill clique." A block is a "predecessor" or "successor" member of the // clique (or, conceivably, both). Each block has a specified sequence of incoming and outgoing spill // temps. If "block" already has its outgoing spill temps assigned (they are always a contiguous series // of local variable numbers, so we represent them with the base local variable number), returns that. // Otherwise, picks a set of spill temps, and propagates this choice to all blocks in the spill clique of // which "block" is a member (asserting, in debug mode, that no block in this clique had its spill temps // chosen already. More precisely, that the incoming or outgoing spill temps are not chosen, depending // on which kind of member of the clique the block is). unsigned impGetSpillTmpBase(BasicBlock* block); // Assumes that "block" is a basic block that completes with a non-empty stack. We have previously // assigned the values on the stack to local variables (the "spill temp" variables). The successor blocks // will assume that its incoming stack contents are in those locals. This requires "block" and its // successors to agree on the variables and their types that will be used. The CLI spec allows implicit // conversions between 'int' and 'native int' or 'float' and 'double' stack types. So one predecessor can // push an int and another can push a native int. For 64-bit we have chosen to implement this by typing // the "spill temp" as native int, and then importing (or re-importing as needed) so that all the // predecessors in the "spill clique" push a native int (sign-extending if needed), and all the // successors receive a native int. Similarly float and double are unified to double. // This routine is called after a type-mismatch is detected, and it will walk the spill clique to mark // blocks for re-importation as appropriate (both successors, so they get the right incoming type, and // predecessors, so they insert an upcast if needed). void impReimportSpillClique(BasicBlock* block); // When we compute a "spill clique" (see above) these byte-maps are allocated to have a byte per basic // block, and represent the predecessor and successor members of the clique currently being computed. // *** Access to these will need to be locked in a parallel compiler. JitExpandArray impSpillCliquePredMembers; JitExpandArray impSpillCliqueSuccMembers; enum SpillCliqueDir { SpillCliquePred, SpillCliqueSucc }; // Abstract class for receiving a callback while walking a spill clique class SpillCliqueWalker { public: virtual void Visit(SpillCliqueDir predOrSucc, BasicBlock* blk) = 0; }; // This class is used for setting the bbStkTempsIn and bbStkTempsOut on the blocks within a spill clique class SetSpillTempsBase : public SpillCliqueWalker { unsigned m_baseTmp; public: SetSpillTempsBase(unsigned baseTmp) : m_baseTmp(baseTmp) { } virtual void Visit(SpillCliqueDir predOrSucc, BasicBlock* blk); }; // This class is used for implementing impReimportSpillClique part on each block within the spill clique class ReimportSpillClique : public SpillCliqueWalker { Compiler* m_pComp; public: ReimportSpillClique(Compiler* pComp) : m_pComp(pComp) { } virtual void Visit(SpillCliqueDir predOrSucc, BasicBlock* blk); }; // This is the heart of the algorithm for walking spill cliques. It invokes callback->Visit for each // predecessor or successor within the spill clique void impWalkSpillCliqueFromPred(BasicBlock* pred, SpillCliqueWalker* callback); // For a BasicBlock that has already been imported, the EntryState has an array of GenTrees for the // incoming locals. This walks that list an resets the types of the GenTrees to match the types of // the VarDscs. They get out of sync when we have int/native int issues (see impReimportSpillClique). void impRetypeEntryStateTemps(BasicBlock* blk); BYTE impSpillCliqueGetMember(SpillCliqueDir predOrSucc, BasicBlock* blk); void impSpillCliqueSetMember(SpillCliqueDir predOrSucc, BasicBlock* blk, BYTE val); void impPushVar(GenTree* op, typeInfo tiRetVal); void impLoadVar(unsigned lclNum, IL_OFFSET offset, typeInfo tiRetVal); void impLoadVar(unsigned lclNum, IL_OFFSET offset) { impLoadVar(lclNum, offset, lvaTable[lclNum].lvVerTypeInfo); } void impLoadArg(unsigned ilArgNum, IL_OFFSET offset); void impLoadLoc(unsigned ilLclNum, IL_OFFSET offset); bool impReturnInstruction(BasicBlock* block, int prefixFlags, OPCODE& opcode); #ifdef _TARGET_ARM_ void impMarkLclDstNotPromotable(unsigned tmpNum, GenTree* op, CORINFO_CLASS_HANDLE hClass); #endif // A free list of linked list nodes used to represent to-do stacks of basic blocks. struct BlockListNode { BasicBlock* m_blk; BlockListNode* m_next; BlockListNode(BasicBlock* blk, BlockListNode* next = nullptr) : m_blk(blk), m_next(next) { } void* operator new(size_t sz, Compiler* comp); }; BlockListNode* impBlockListNodeFreeList; void FreeBlockListNode(BlockListNode* node); bool impIsValueType(typeInfo* pTypeInfo); var_types mangleVarArgsType(var_types type); #if FEATURE_VARARG regNumber getCallArgIntRegister(regNumber floatReg); regNumber getCallArgFloatRegister(regNumber intReg); #endif // FEATURE_VARARG #if defined(DEBUG) static unsigned jitTotalMethodCompiled; #endif #ifdef DEBUG static LONG jitNestingLevel; #endif // DEBUG static BOOL impIsAddressInLocal(GenTree* tree, GenTree** lclVarTreeOut); void impMakeDiscretionaryInlineObservations(InlineInfo* pInlineInfo, InlineResult* inlineResult); // STATIC inlining decision based on the IL code. void impCanInlineIL(CORINFO_METHOD_HANDLE fncHandle, CORINFO_METHOD_INFO* methInfo, bool forceInline, InlineResult* inlineResult); void impCheckCanInline(GenTreeCall* call, CORINFO_METHOD_HANDLE fncHandle, unsigned methAttr, CORINFO_CONTEXT_HANDLE exactContextHnd, InlineCandidateInfo** ppInlineCandidateInfo, InlineResult* inlineResult); void impInlineRecordArgInfo(InlineInfo* pInlineInfo, GenTree* curArgVal, unsigned argNum, InlineResult* inlineResult); void impInlineInitVars(InlineInfo* pInlineInfo); unsigned impInlineFetchLocal(unsigned lclNum DEBUGARG(const char* reason)); GenTree* impInlineFetchArg(unsigned lclNum, InlArgInfo* inlArgInfo, InlLclVarInfo* lclTypeInfo); BOOL impInlineIsThis(GenTree* tree, InlArgInfo* inlArgInfo); BOOL impInlineIsGuaranteedThisDerefBeforeAnySideEffects(GenTree* additionalTreesToBeEvaluatedBefore, GenTree* variableBeingDereferenced, InlArgInfo* inlArgInfo); void impMarkInlineCandidate(GenTree* call, CORINFO_CONTEXT_HANDLE exactContextHnd, bool exactContextNeedsRuntimeLookup, CORINFO_CALL_INFO* callInfo); void impMarkInlineCandidateHelper(GenTreeCall* call, CORINFO_CONTEXT_HANDLE exactContextHnd, bool exactContextNeedsRuntimeLookup, CORINFO_CALL_INFO* callInfo); bool impTailCallRetTypeCompatible(var_types callerRetType, CORINFO_CLASS_HANDLE callerRetTypeClass, var_types calleeRetType, CORINFO_CLASS_HANDLE calleeRetTypeClass); bool impIsTailCallILPattern(bool tailPrefixed, OPCODE curOpcode, const BYTE* codeAddrOfNextOpcode, const BYTE* codeEnd, bool isRecursive, bool* IsCallPopRet = nullptr); bool impIsImplicitTailCallCandidate( OPCODE curOpcode, const BYTE* codeAddrOfNextOpcode, const BYTE* codeEnd, int prefixFlags, bool isRecursive); CORINFO_RESOLVED_TOKEN* impAllocateToken(CORINFO_RESOLVED_TOKEN token); /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX FlowGraph XX XX XX XX Info about the basic-blocks, their contents and the flow analysis XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: BasicBlock* fgFirstBB; // Beginning of the basic block list BasicBlock* fgLastBB; // End of the basic block list BasicBlock* fgFirstColdBlock; // First block to be placed in the cold section #if FEATURE_EH_FUNCLETS BasicBlock* fgFirstFuncletBB; // First block of outlined funclets (to allow block insertion before the funclets) #endif BasicBlock* fgFirstBBScratch; // Block inserted for initialization stuff. Is nullptr if no such block has been // created. BasicBlockList* fgReturnBlocks; // list of BBJ_RETURN blocks unsigned fgEdgeCount; // # of control flow edges between the BBs unsigned fgBBcount; // # of BBs in the method #ifdef DEBUG unsigned fgBBcountAtCodegen; // # of BBs in the method at the start of codegen #endif unsigned fgBBNumMax; // The max bbNum that has been assigned to basic blocks unsigned fgDomBBcount; // # of BBs for which we have dominator and reachability information BasicBlock** fgBBInvPostOrder; // The flow graph stored in an array sorted in topological order, needed to compute // dominance. Indexed by block number. Size: fgBBNumMax + 1. // After the dominance tree is computed, we cache a DFS preorder number and DFS postorder number to compute // dominance queries in O(1). fgDomTreePreOrder and fgDomTreePostOrder are arrays giving the block's preorder and // postorder number, respectively. The arrays are indexed by basic block number. (Note that blocks are numbered // starting from one. Thus, we always waste element zero. This makes debugging easier and makes the code less likely // to suffer from bugs stemming from forgetting to add or subtract one from the block number to form an array // index). The arrays are of size fgBBNumMax + 1. unsigned* fgDomTreePreOrder; unsigned* fgDomTreePostOrder; bool fgBBVarSetsInited; // Allocate array like T* a = new T[fgBBNumMax + 1]; // Using helper so we don't keep forgetting +1. template T* fgAllocateTypeForEachBlk(CompMemKind cmk = CMK_Unknown) { return getAllocator(cmk).allocate(fgBBNumMax + 1); } // BlockSets are relative to a specific set of BasicBlock numbers. If that changes // (if the blocks are renumbered), this changes. BlockSets from different epochs // cannot be meaningfully combined. Note that new blocks can be created with higher // block numbers without changing the basic block epoch. These blocks *cannot* // participate in a block set until the blocks are all renumbered, causing the epoch // to change. This is useful if continuing to use previous block sets is valuable. // If the epoch is zero, then it is uninitialized, and block sets can't be used. unsigned fgCurBBEpoch; unsigned GetCurBasicBlockEpoch() { return fgCurBBEpoch; } // The number of basic blocks in the current epoch. When the blocks are renumbered, // this is fgBBcount. As blocks are added, fgBBcount increases, fgCurBBEpochSize remains // the same, until a new BasicBlock epoch is created, such as when the blocks are all renumbered. unsigned fgCurBBEpochSize; // The number of "size_t" elements required to hold a bitset large enough for fgCurBBEpochSize // bits. This is precomputed to avoid doing math every time BasicBlockBitSetTraits::GetArrSize() is called. unsigned fgBBSetCountInSizeTUnits; void NewBasicBlockEpoch() { INDEBUG(unsigned oldEpochArrSize = fgBBSetCountInSizeTUnits); // We have a new epoch. Compute and cache the size needed for new BlockSets. fgCurBBEpoch++; fgCurBBEpochSize = fgBBNumMax + 1; fgBBSetCountInSizeTUnits = roundUp(fgCurBBEpochSize, (unsigned)(sizeof(size_t) * 8)) / unsigned(sizeof(size_t) * 8); #ifdef DEBUG // All BlockSet objects are now invalid! fgReachabilitySetsValid = false; // the bbReach sets are now invalid! fgEnterBlksSetValid = false; // the fgEnterBlks set is now invalid! if (verbose) { unsigned epochArrSize = BasicBlockBitSetTraits::GetArrSize(this, sizeof(size_t)); printf("\nNew BlockSet epoch %d, # of blocks (including unused BB00): %u, bitset array size: %u (%s)", fgCurBBEpoch, fgCurBBEpochSize, epochArrSize, (epochArrSize <= 1) ? "short" : "long"); if ((fgCurBBEpoch != 1) && ((oldEpochArrSize <= 1) != (epochArrSize <= 1))) { // If we're not just establishing the first epoch, and the epoch array size has changed such that we're // going to change our bitset representation from short (just a size_t bitset) to long (a pointer to an // array of size_t bitsets), then print that out. printf("; NOTE: BlockSet size was previously %s!", (oldEpochArrSize <= 1) ? "short" : "long"); } printf("\n"); } #endif // DEBUG } void EnsureBasicBlockEpoch() { if (fgCurBBEpochSize != fgBBNumMax + 1) { NewBasicBlockEpoch(); } } BasicBlock* fgNewBasicBlock(BBjumpKinds jumpKind); void fgEnsureFirstBBisScratch(); bool fgFirstBBisScratch(); bool fgBBisScratch(BasicBlock* block); void fgExtendEHRegionBefore(BasicBlock* block); void fgExtendEHRegionAfter(BasicBlock* block); BasicBlock* fgNewBBbefore(BBjumpKinds jumpKind, BasicBlock* block, bool extendRegion); BasicBlock* fgNewBBafter(BBjumpKinds jumpKind, BasicBlock* block, bool extendRegion); BasicBlock* fgNewBBinRegion(BBjumpKinds jumpKind, unsigned tryIndex, unsigned hndIndex, BasicBlock* nearBlk, bool putInFilter = false, bool runRarely = false, bool insertAtEnd = false); BasicBlock* fgNewBBinRegion(BBjumpKinds jumpKind, BasicBlock* srcBlk, bool runRarely = false, bool insertAtEnd = false); BasicBlock* fgNewBBinRegion(BBjumpKinds jumpKind); BasicBlock* fgNewBBinRegionWorker(BBjumpKinds jumpKind, BasicBlock* afterBlk, unsigned xcptnIndex, bool putInTryRegion); void fgInsertBBbefore(BasicBlock* insertBeforeBlk, BasicBlock* newBlk); void fgInsertBBafter(BasicBlock* insertAfterBlk, BasicBlock* newBlk); void fgUnlinkBlock(BasicBlock* block); unsigned fgMeasureIR(); bool fgModified; // True if the flow graph has been modified recently bool fgComputePredsDone; // Have we computed the bbPreds list bool fgCheapPredsValid; // Is the bbCheapPreds list valid? bool fgDomsComputed; // Have we computed the dominator sets? bool fgOptimizedFinally; // Did we optimize any try-finallys? bool fgHasSwitch; // any BBJ_SWITCH jumps? BlockSet fgEnterBlks; // Set of blocks which have a special transfer of control; the "entry" blocks plus EH handler // begin blocks. #ifdef DEBUG bool fgReachabilitySetsValid; // Are the bbReach sets valid? bool fgEnterBlksSetValid; // Is the fgEnterBlks set valid? #endif // DEBUG bool fgRemoveRestOfBlock; // true if we know that we will throw bool fgStmtRemoved; // true if we remove statements -> need new DFA // There are two modes for ordering of the trees. // - In FGOrderTree, the dominant ordering is the tree order, and the nodes contained in // each tree and sub-tree are contiguous, and can be traversed (in gtNext/gtPrev order) // by traversing the tree according to the order of the operands. // - In FGOrderLinear, the dominant ordering is the linear order. enum FlowGraphOrder { FGOrderTree, FGOrderLinear }; FlowGraphOrder fgOrder; // The following are boolean flags that keep track of the state of internal data structures bool fgStmtListThreaded; // true if the node list is now threaded bool fgCanRelocateEHRegions; // true if we are allowed to relocate the EH regions bool fgEdgeWeightsComputed; // true after we have called fgComputeEdgeWeights bool fgHaveValidEdgeWeights; // true if we were successful in computing all of the edge weights bool fgSlopUsedInEdgeWeights; // true if their was some slop used when computing the edge weights bool fgRangeUsedInEdgeWeights; // true if some of the edgeWeight are expressed in Min..Max form bool fgNeedsUpdateFlowGraph; // true if we need to run fgUpdateFlowGraph BasicBlock::weight_t fgCalledCount; // count of the number of times this method was called // This is derived from the profile data // or is BB_UNITY_WEIGHT when we don't have profile data #if FEATURE_EH_FUNCLETS bool fgFuncletsCreated; // true if the funclet creation phase has been run #endif // FEATURE_EH_FUNCLETS bool fgGlobalMorph; // indicates if we are during the global morphing phase // since fgMorphTree can be called from several places bool impBoxTempInUse; // the temp below is valid and available unsigned impBoxTemp; // a temporary that is used for boxing #ifdef DEBUG bool jitFallbackCompile; // Are we doing a fallback compile? That is, have we executed a NO_WAY assert, // and we are trying to compile again in a "safer", minopts mode? #endif #if defined(DEBUG) unsigned impInlinedCodeSize; #endif //------------------------------------------------------------------------- void fgInit(); void fgImport(); void fgTransformIndirectCalls(); void fgInline(); void fgRemoveEmptyTry(); void fgRemoveEmptyFinally(); void fgMergeFinallyChains(); void fgCloneFinally(); void fgCleanupContinuation(BasicBlock* continuation); void fgUpdateFinallyTargetFlags(); void fgClearAllFinallyTargetBits(); void fgAddFinallyTargetFlags(); #if FEATURE_EH_FUNCLETS && defined(_TARGET_ARM_) // Sometimes we need to defer updating the BBF_FINALLY_TARGET bit. fgNeedToAddFinallyTargetBits signals // when this is necessary. bool fgNeedToAddFinallyTargetBits; #endif // FEATURE_EH_FUNCLETS && defined(_TARGET_ARM_) bool fgRetargetBranchesToCanonicalCallFinally(BasicBlock* block, BasicBlock* handler, BlockToBlockMap& continuationMap); GenTree* fgGetCritSectOfStaticMethod(); #if FEATURE_EH_FUNCLETS void fgAddSyncMethodEnterExit(); GenTree* fgCreateMonitorTree(unsigned lvaMonitorBool, unsigned lvaThisVar, BasicBlock* block, bool enter); void fgConvertSyncReturnToLeave(BasicBlock* block); #endif // FEATURE_EH_FUNCLETS void fgAddReversePInvokeEnterExit(); bool fgMoreThanOneReturnBlock(); // The number of separate return points in the method. unsigned fgReturnCount; void fgAddInternal(); bool fgFoldConditional(BasicBlock* block); void fgMorphStmts(BasicBlock* block, bool* lnot, bool* loadw); void fgMorphBlocks(); bool fgMorphBlockStmt(BasicBlock* block, GenTreeStmt* stmt DEBUGARG(const char* msg)); void fgSetOptions(); #ifdef DEBUG static fgWalkPreFn fgAssertNoQmark; void fgPreExpandQmarkChecks(GenTree* expr); void fgPostExpandQmarkChecks(); static void fgCheckQmarkAllowedForm(GenTree* tree); #endif IL_OFFSET fgFindBlockILOffset(BasicBlock* block); BasicBlock* fgSplitBlockAtBeginning(BasicBlock* curr); BasicBlock* fgSplitBlockAtEnd(BasicBlock* curr); BasicBlock* fgSplitBlockAfterStatement(BasicBlock* curr, GenTreeStmt* stmt); BasicBlock* fgSplitBlockAfterNode(BasicBlock* curr, GenTree* node); // for LIR BasicBlock* fgSplitEdge(BasicBlock* curr, BasicBlock* succ); GenTreeStmt* fgNewStmtFromTree(GenTree* tree, BasicBlock* block, IL_OFFSETX offs); GenTreeStmt* fgNewStmtFromTree(GenTree* tree); GenTreeStmt* fgNewStmtFromTree(GenTree* tree, BasicBlock* block); GenTreeStmt* fgNewStmtFromTree(GenTree* tree, IL_OFFSETX offs); GenTree* fgGetTopLevelQmark(GenTree* expr, GenTree** ppDst = nullptr); void fgExpandQmarkForCastInstOf(BasicBlock* block, GenTreeStmt* stmt); void fgExpandQmarkStmt(BasicBlock* block, GenTreeStmt* stmt); void fgExpandQmarkNodes(); void fgMorph(); // Do "simple lowering." This functionality is (conceptually) part of "general" // lowering that is distributed between fgMorph and the lowering phase of LSRA. void fgSimpleLowering(); GenTree* fgInitThisClass(); GenTreeCall* fgGetStaticsCCtorHelper(CORINFO_CLASS_HANDLE cls, CorInfoHelpFunc helper); GenTreeCall* fgGetSharedCCtor(CORINFO_CLASS_HANDLE cls); bool backendRequiresLocalVarLifetimes() { return !opts.MinOpts() || m_pLinearScan->willEnregisterLocalVars(); } void fgLocalVarLiveness(); void fgLocalVarLivenessInit(); void fgPerNodeLocalVarLiveness(GenTree* node); void fgPerBlockLocalVarLiveness(); VARSET_VALRET_TP fgGetHandlerLiveVars(BasicBlock* block); void fgLiveVarAnalysis(bool updateInternalOnly = false); void fgComputeLifeCall(VARSET_TP& life, GenTreeCall* call); void fgComputeLifeTrackedLocalUse(VARSET_TP& life, LclVarDsc& varDsc, GenTreeLclVarCommon* node); bool fgComputeLifeTrackedLocalDef(VARSET_TP& life, VARSET_VALARG_TP keepAliveVars, LclVarDsc& varDsc, GenTreeLclVarCommon* node); void fgComputeLifeUntrackedLocal(VARSET_TP& life, VARSET_VALARG_TP keepAliveVars, LclVarDsc& varDsc, GenTreeLclVarCommon* lclVarNode); bool fgComputeLifeLocal(VARSET_TP& life, VARSET_VALARG_TP keepAliveVars, GenTree* lclVarNode); void fgComputeLife(VARSET_TP& life, GenTree* startNode, GenTree* endNode, VARSET_VALARG_TP volatileVars, bool* pStmtInfoDirty DEBUGARG(bool* treeModf)); void fgComputeLifeLIR(VARSET_TP& life, BasicBlock* block, VARSET_VALARG_TP volatileVars); bool fgRemoveDeadStore(GenTree** pTree, LclVarDsc* varDsc, VARSET_VALARG_TP life, bool* doAgain, bool* pStmtInfoDirty DEBUGARG(bool* treeModf)); // For updating liveset during traversal AFTER fgComputeLife has completed VARSET_VALRET_TP fgGetVarBits(GenTree* tree); VARSET_VALRET_TP fgUpdateLiveSet(VARSET_VALARG_TP liveSet, GenTree* tree); // Returns the set of live variables after endTree, // assuming that liveSet is the set of live variables BEFORE tree. // Requires that fgComputeLife has completed, and that tree is in the same // statement as endTree, and that it comes before endTree in execution order VARSET_VALRET_TP fgUpdateLiveSet(VARSET_VALARG_TP liveSet, GenTree* tree, GenTree* endTree) { VARSET_TP newLiveSet(VarSetOps::MakeCopy(this, liveSet)); while (tree != nullptr && tree != endTree->gtNext) { VarSetOps::AssignNoCopy(this, newLiveSet, fgUpdateLiveSet(newLiveSet, tree)); tree = tree->gtNext; } assert(tree == endTree->gtNext); return newLiveSet; } void fgInterBlockLocalVarLiveness(); // The presence of a partial definition presents some difficulties for SSA: this is both a use of some SSA name // of "x", and a def of a new SSA name for "x". The tree only has one local variable for "x", so it has to choose // whether to treat that as the use or def. It chooses the "use", and thus the old SSA name. This map allows us // to record/recover the "def" SSA number, given the lcl var node for "x" in such a tree. typedef JitHashTable, unsigned> NodeToUnsignedMap; NodeToUnsignedMap* m_opAsgnVarDefSsaNums; NodeToUnsignedMap* GetOpAsgnVarDefSsaNums() { if (m_opAsgnVarDefSsaNums == nullptr) { m_opAsgnVarDefSsaNums = new (getAllocator()) NodeToUnsignedMap(getAllocator()); } return m_opAsgnVarDefSsaNums; } // Requires value numbering phase to have completed. Returns the value number ("gtVN") of the // "tree," EXCEPT in the case of GTF_VAR_USEASG, because the tree node's gtVN member is the // "use" VN. Performs a lookup into the map of (use asg tree -> def VN.) to return the "def's" // VN. inline ValueNum GetUseAsgDefVNOrTreeVN(GenTree* tree); // Requires that "lcl" has the GTF_VAR_DEF flag set. Returns the SSA number of "lcl". // Except: assumes that lcl is a def, and if it is // a partial def (GTF_VAR_USEASG), looks up and returns the SSA number for the "def", // rather than the "use" SSA number recorded in the tree "lcl". inline unsigned GetSsaNumForLocalVarDef(GenTree* lcl); // Performs SSA conversion. void fgSsaBuild(); // Reset any data structures to the state expected by "fgSsaBuild", so it can be run again. void fgResetForSsa(); unsigned fgSsaPassesCompleted; // Number of times fgSsaBuild has been run. // Returns "true" if a struct temp of the given type requires needs zero init in this block inline bool fgStructTempNeedsExplicitZeroInit(LclVarDsc* varDsc, BasicBlock* block); // The value numbers for this compilation. ValueNumStore* vnStore; public: ValueNumStore* GetValueNumStore() { return vnStore; } // Do value numbering (assign a value number to each // tree node). void fgValueNumber(); // Computes new GcHeap VN via the assignment H[elemTypeEq][arrVN][inx][fldSeq] = rhsVN. // Assumes that "elemTypeEq" is the (equivalence class rep) of the array element type. // The 'indType' is the indirection type of the lhs of the assignment and will typically // match the element type of the array or fldSeq. When this type doesn't match // or if the fldSeq is 'NotAField' we invalidate the array contents H[elemTypeEq][arrVN] // ValueNum fgValueNumberArrIndexAssign(CORINFO_CLASS_HANDLE elemTypeEq, ValueNum arrVN, ValueNum inxVN, FieldSeqNode* fldSeq, ValueNum rhsVN, var_types indType); // Requires that "tree" is a GT_IND marked as an array index, and that its address argument // has been parsed to yield the other input arguments. If evaluation of the address // can raise exceptions, those should be captured in the exception set "excVN." // Assumes that "elemTypeEq" is the (equivalence class rep) of the array element type. // Marks "tree" with the VN for H[elemTypeEq][arrVN][inx][fldSeq] (for the liberal VN; a new unique // VN for the conservative VN.) Also marks the tree's argument as the address of an array element. // The type tree->TypeGet() will typically match the element type of the array or fldSeq. // When this type doesn't match or if the fldSeq is 'NotAField' we return a new unique VN // ValueNum fgValueNumberArrIndexVal(GenTree* tree, CORINFO_CLASS_HANDLE elemTypeEq, ValueNum arrVN, ValueNum inxVN, ValueNum excVN, FieldSeqNode* fldSeq); // Requires "funcApp" to be a VNF_PtrToArrElem, and "addrXvn" to represent the exception set thrown // by evaluating the array index expression "tree". Returns the value number resulting from // dereferencing the array in the current GcHeap state. If "tree" is non-null, it must be the // "GT_IND" that does the dereference, and it is given the returned value number. ValueNum fgValueNumberArrIndexVal(GenTree* tree, struct VNFuncApp* funcApp, ValueNum addrXvn); // Compute the value number for a byref-exposed load of the given type via the given pointerVN. ValueNum fgValueNumberByrefExposedLoad(var_types type, ValueNum pointerVN); unsigned fgVNPassesCompleted; // Number of times fgValueNumber has been run. // Utility functions for fgValueNumber. // Perform value-numbering for the trees in "blk". void fgValueNumberBlock(BasicBlock* blk); // Requires that "entryBlock" is the entry block of loop "loopNum", and that "loopNum" is the // innermost loop of which "entryBlock" is the entry. Returns the value number that should be // assumed for the memoryKind at the start "entryBlk". ValueNum fgMemoryVNForLoopSideEffects(MemoryKind memoryKind, BasicBlock* entryBlock, unsigned loopNum); // Called when an operation (performed by "tree", described by "msg") may cause the GcHeap to be mutated. // As GcHeap is a subset of ByrefExposed, this will also annotate the ByrefExposed mutation. void fgMutateGcHeap(GenTree* tree DEBUGARG(const char* msg)); // Called when an operation (performed by "tree", described by "msg") may cause an address-exposed local to be // mutated. void fgMutateAddressExposedLocal(GenTree* tree DEBUGARG(const char* msg)); // For a GC heap store at curTree, record the new curMemoryVN's and update curTree's MemorySsaMap. // As GcHeap is a subset of ByrefExposed, this will also record the ByrefExposed store. void recordGcHeapStore(GenTree* curTree, ValueNum gcHeapVN DEBUGARG(const char* msg)); // For a store to an address-exposed local at curTree, record the new curMemoryVN and update curTree's MemorySsaMap. void recordAddressExposedLocalStore(GenTree* curTree, ValueNum memoryVN DEBUGARG(const char* msg)); // Tree caused an update in the current memory VN. If "tree" has an associated heap SSA #, record that // value in that SSA #. void fgValueNumberRecordMemorySsa(MemoryKind memoryKind, GenTree* tree); // The input 'tree' is a leaf node that is a constant // Assign the proper value number to the tree void fgValueNumberTreeConst(GenTree* tree); // Assumes that all inputs to "tree" have had value numbers assigned; assigns a VN to tree. // (With some exceptions: the VN of the lhs of an assignment is assigned as part of the // assignment.) void fgValueNumberTree(GenTree* tree); // Does value-numbering for a block assignment. void fgValueNumberBlockAssignment(GenTree* tree); // Does value-numbering for a cast tree. void fgValueNumberCastTree(GenTree* tree); // Does value-numbering for an intrinsic tree. void fgValueNumberIntrinsic(GenTree* tree); // Does value-numbering for a call. We interpret some helper calls. void fgValueNumberCall(GenTreeCall* call); // The VN of some nodes in "args" may have changed -- reassign VNs to the arg list nodes. void fgUpdateArgListVNs(GenTreeArgList* args); // Does value-numbering for a helper "call" that has a VN function symbol "vnf". void fgValueNumberHelperCallFunc(GenTreeCall* call, VNFunc vnf, ValueNumPair vnpExc); // Requires "helpCall" to be a helper call. Assigns it a value number; // we understand the semantics of some of the calls. Returns "true" if // the call may modify the heap (we assume arbitrary memory side effects if so). bool fgValueNumberHelperCall(GenTreeCall* helpCall); // Requires that "helpFunc" is one of the pure Jit Helper methods. // Returns the corresponding VNFunc to use for value numbering VNFunc fgValueNumberJitHelperMethodVNFunc(CorInfoHelpFunc helpFunc); // Adds the exception set for the current tree node which has a memory indirection operation void fgValueNumberAddExceptionSetForIndirection(GenTree* tree, GenTree* baseAddr); // Adds the exception sets for the current tree node which is performing a division or modulus operation void fgValueNumberAddExceptionSetForDivision(GenTree* tree); // Adds the exception set for the current tree node which is performing a overflow checking operation void fgValueNumberAddExceptionSetForOverflow(GenTree* tree); // Adds the exception set for the current tree node which is performing a ckfinite operation void fgValueNumberAddExceptionSetForCkFinite(GenTree* tree); // Adds the exception sets for the current tree node void fgValueNumberAddExceptionSet(GenTree* tree); // These are the current value number for the memory implicit variables while // doing value numbering. These are the value numbers under the "liberal" interpretation // of memory values; the "conservative" interpretation needs no VN, since every access of // memory yields an unknown value. ValueNum fgCurMemoryVN[MemoryKindCount]; // Return a "pseudo"-class handle for an array element type. If "elemType" is TYP_STRUCT, // requires "elemStructType" to be non-null (and to have a low-order zero). Otherwise, low order bit // is 1, and the rest is an encoding of "elemTyp". static CORINFO_CLASS_HANDLE EncodeElemType(var_types elemTyp, CORINFO_CLASS_HANDLE elemStructType) { if (elemStructType != nullptr) { assert(varTypeIsStruct(elemTyp) || elemTyp == TYP_REF || elemTyp == TYP_BYREF || varTypeIsIntegral(elemTyp)); assert((size_t(elemStructType) & 0x1) == 0x0); // Make sure the encoding below is valid. return elemStructType; } else { assert(elemTyp != TYP_STRUCT); elemTyp = varTypeUnsignedToSigned(elemTyp); return CORINFO_CLASS_HANDLE(size_t(elemTyp) << 1 | 0x1); } } // If "clsHnd" is the result of an "EncodePrim" call, returns true and sets "*pPrimType" to the // var_types it represents. Otherwise, returns TYP_STRUCT (on the assumption that "clsHnd" is // the struct type of the element). static var_types DecodeElemType(CORINFO_CLASS_HANDLE clsHnd) { size_t clsHndVal = size_t(clsHnd); if (clsHndVal & 0x1) { return var_types(clsHndVal >> 1); } else { return TYP_STRUCT; } } // Convert a BYTE which represents the VM's CorInfoGCtype to the JIT's var_types var_types getJitGCType(BYTE gcType); enum structPassingKind { SPK_Unknown, // Invalid value, never returned SPK_PrimitiveType, // The struct is passed/returned using a primitive type. SPK_EnclosingType, // Like SPK_Primitive type, but used for return types that // require a primitive type temp that is larger than the struct size. // Currently used for structs of size 3, 5, 6, or 7 bytes. SPK_ByValue, // The struct is passed/returned by value (using the ABI rules) // for ARM64 and UNIX_X64 in multiple registers. (when all of the // parameters registers are used, then the stack will be used) // for X86 passed on the stack, for ARM32 passed in registers // or the stack or split between registers and the stack. SPK_ByValueAsHfa, // The struct is passed/returned as an HFA in multiple registers. SPK_ByReference }; // The struct is passed/returned by reference to a copy/buffer. // Get the "primitive" type that is is used when we are given a struct of size 'structSize'. // For pointer sized structs the 'clsHnd' is used to determine if the struct contains GC ref. // A "primitive" type is one of the scalar types: byte, short, int, long, ref, float, double // If we can't or shouldn't use a "primitive" type then TYP_UNKNOWN is returned. // // isVarArg is passed for use on Windows Arm64 to change the decision returned regarding // hfa types. // var_types getPrimitiveTypeForStruct(unsigned structSize, CORINFO_CLASS_HANDLE clsHnd, bool isVarArg); // Get the type that is used to pass values of the given struct type. // isVarArg is passed for use on Windows Arm64 to change the decision returned regarding // hfa types. // var_types getArgTypeForStruct(CORINFO_CLASS_HANDLE clsHnd, structPassingKind* wbPassStruct, bool isVarArg, unsigned structSize); // Get the type that is used to return values of the given struct type. // If the size is unknown, pass 0 and it will be determined from 'clsHnd'. var_types getReturnTypeForStruct(CORINFO_CLASS_HANDLE clsHnd, structPassingKind* wbPassStruct = nullptr, unsigned structSize = 0); #ifdef DEBUG // Print a representation of "vnp" or "vn" on standard output. // If "level" is non-zero, we also print out a partial expansion of the value. void vnpPrint(ValueNumPair vnp, unsigned level); void vnPrint(ValueNum vn, unsigned level); #endif bool fgDominate(BasicBlock* b1, BasicBlock* b2); // Return true if b1 dominates b2 // Dominator computation member functions // Not exposed outside Compiler protected: bool fgReachable(BasicBlock* b1, BasicBlock* b2); // Returns true if block b1 can reach block b2 void fgComputeDoms(); // Computes the immediate dominators for each basic block in the // flow graph. We first assume the fields bbIDom on each // basic block are invalid. This computation is needed later // by fgBuildDomTree to build the dominance tree structure. // Based on: A Simple, Fast Dominance Algorithm // by Keith D. Cooper, Timothy J. Harvey, and Ken Kennedy void fgCompDominatedByExceptionalEntryBlocks(); BlockSet_ValRet_T fgGetDominatorSet(BasicBlock* block); // Returns a set of blocks that dominate the given block. // Note: this is relatively slow compared to calling fgDominate(), // especially if dealing with a single block versus block check. void fgComputeReachabilitySets(); // Compute bbReach sets. (Also sets BBF_GC_SAFE_POINT flag on blocks.) void fgComputeEnterBlocksSet(); // Compute the set of entry blocks, 'fgEnterBlks'. bool fgRemoveUnreachableBlocks(); // Remove blocks determined to be unreachable by the bbReach sets. void fgComputeReachability(); // Perform flow graph node reachability analysis. BasicBlock* fgIntersectDom(BasicBlock* a, BasicBlock* b); // Intersect two immediate dominator sets. void fgDfsInvPostOrder(); // In order to compute dominance using fgIntersectDom, the flow graph nodes must be // processed in topological sort, this function takes care of that. void fgDfsInvPostOrderHelper(BasicBlock* block, BlockSet& visited, unsigned* count); BlockSet_ValRet_T fgDomFindStartNodes(); // Computes which basic blocks don't have incoming edges in the flow graph. // Returns this as a set. BlockSet_ValRet_T fgDomTreeEntryNodes(BasicBlockList** domTree); // Computes which nodes in the dominance forest are // root nodes. Returns this as a set. #ifdef DEBUG void fgDispDomTree(BasicBlockList** domTree); // Helper that prints out the Dominator Tree in debug builds. #endif // DEBUG void fgBuildDomTree(); // Once we compute all the immediate dominator sets for each node in the flow graph // (performed by fgComputeDoms), this procedure builds the dominance tree represented // adjacency lists. // In order to speed up the queries of the form 'Does A dominates B', we can perform a DFS preorder and postorder // traversal of the dominance tree and the dominance query will become A dominates B iif preOrder(A) <= preOrder(B) // && postOrder(A) >= postOrder(B) making the computation O(1). void fgTraverseDomTree(unsigned bbNum, BasicBlockList** domTree, unsigned* preNum, unsigned* postNum); // When the flow graph changes, we need to update the block numbers, predecessor lists, reachability sets, and // dominators. void fgUpdateChangedFlowGraph(); public: // Compute the predecessors of the blocks in the control flow graph. void fgComputePreds(); // Remove all predecessor information. void fgRemovePreds(); // Compute the cheap flow graph predecessors lists. This is used in some early phases // before the full predecessors lists are computed. void fgComputeCheapPreds(); private: void fgAddCheapPred(BasicBlock* block, BasicBlock* blockPred); void fgRemoveCheapPred(BasicBlock* block, BasicBlock* blockPred); public: enum GCPollType { GCPOLL_NONE, GCPOLL_CALL, GCPOLL_INLINE }; // Initialize the per-block variable sets (used for liveness analysis). void fgInitBlockVarSets(); // true if we've gone through and created GC Poll calls. bool fgGCPollsCreated; void fgMarkGCPollBlocks(); void fgCreateGCPolls(); bool fgCreateGCPoll(GCPollType pollType, BasicBlock* block); // Requires that "block" is a block that returns from // a finally. Returns the number of successors (jump targets of // of blocks in the covered "try" that did a "LEAVE".) unsigned fgNSuccsOfFinallyRet(BasicBlock* block); // Requires that "block" is a block that returns (in the sense of BBJ_EHFINALLYRET) from // a finally. Returns its "i"th successor (jump targets of // of blocks in the covered "try" that did a "LEAVE".) // Requires that "i" < fgNSuccsOfFinallyRet(block). BasicBlock* fgSuccOfFinallyRet(BasicBlock* block, unsigned i); private: // Factor out common portions of the impls of the methods above. void fgSuccOfFinallyRetWork(BasicBlock* block, unsigned i, BasicBlock** bres, unsigned* nres); public: // For many purposes, it is desirable to be able to enumerate the *distinct* targets of a switch statement, // skipping duplicate targets. (E.g., in flow analyses that are only interested in the set of possible targets.) // SwitchUniqueSuccSet contains the non-duplicated switch targets. // (Code that modifies the jump table of a switch has an obligation to call Compiler::UpdateSwitchTableTarget, // which in turn will call the "UpdateTarget" method of this type if a SwitchUniqueSuccSet has already // been computed for the switch block. If a switch block is deleted or is transformed into a non-switch, // we leave the entry associated with the block, but it will no longer be accessed.) struct SwitchUniqueSuccSet { unsigned numDistinctSuccs; // Number of distinct targets of the switch. BasicBlock** nonDuplicates; // Array of "numDistinctSuccs", containing all the distinct switch target // successors. // The switch block "switchBlk" just had an entry with value "from" modified to the value "to". // Update "this" as necessary: if "from" is no longer an element of the jump table of "switchBlk", // remove it from "this", and ensure that "to" is a member. Use "alloc" to do any required allocation. void UpdateTarget(CompAllocator alloc, BasicBlock* switchBlk, BasicBlock* from, BasicBlock* to); }; typedef JitHashTable, SwitchUniqueSuccSet> BlockToSwitchDescMap; private: // Maps BasicBlock*'s that end in switch statements to SwitchUniqueSuccSets that allow // iteration over only the distinct successors. BlockToSwitchDescMap* m_switchDescMap; public: BlockToSwitchDescMap* GetSwitchDescMap(bool createIfNull = true) { if ((m_switchDescMap == nullptr) && createIfNull) { m_switchDescMap = new (getAllocator()) BlockToSwitchDescMap(getAllocator()); } return m_switchDescMap; } // Invalidate the map of unique switch block successors. For example, since the hash key of the map // depends on block numbers, we must invalidate the map when the blocks are renumbered, to ensure that // we don't accidentally look up and return the wrong switch data. void InvalidateUniqueSwitchSuccMap() { m_switchDescMap = nullptr; } // Requires "switchBlock" to be a block that ends in a switch. Returns // the corresponding SwitchUniqueSuccSet. SwitchUniqueSuccSet GetDescriptorForSwitch(BasicBlock* switchBlk); // The switch block "switchBlk" just had an entry with value "from" modified to the value "to". // Update "this" as necessary: if "from" is no longer an element of the jump table of "switchBlk", // remove it from "this", and ensure that "to" is a member. void UpdateSwitchTableTarget(BasicBlock* switchBlk, BasicBlock* from, BasicBlock* to); // Remove the "SwitchUniqueSuccSet" of "switchBlk" in the BlockToSwitchDescMap. void fgInvalidateSwitchDescMapEntry(BasicBlock* switchBlk); BasicBlock* fgFirstBlockOfHandler(BasicBlock* block); flowList* fgGetPredForBlock(BasicBlock* block, BasicBlock* blockPred); flowList* fgGetPredForBlock(BasicBlock* block, BasicBlock* blockPred, flowList*** ptrToPred); flowList* fgSpliceOutPred(BasicBlock* block, BasicBlock* blockPred); flowList* fgRemoveRefPred(BasicBlock* block, BasicBlock* blockPred); flowList* fgRemoveAllRefPreds(BasicBlock* block, BasicBlock* blockPred); flowList* fgRemoveAllRefPreds(BasicBlock* block, flowList** ptrToPred); void fgRemoveBlockAsPred(BasicBlock* block); void fgChangeSwitchBlock(BasicBlock* oldSwitchBlock, BasicBlock* newSwitchBlock); void fgReplaceSwitchJumpTarget(BasicBlock* blockSwitch, BasicBlock* newTarget, BasicBlock* oldTarget); void fgReplaceJumpTarget(BasicBlock* block, BasicBlock* newTarget, BasicBlock* oldTarget); void fgReplacePred(BasicBlock* block, BasicBlock* oldPred, BasicBlock* newPred); flowList* fgAddRefPred(BasicBlock* block, BasicBlock* blockPred, flowList* oldEdge = nullptr, bool initializingPreds = false); // Only set to 'true' when we are computing preds in // fgComputePreds() void fgFindBasicBlocks(); bool fgIsBetterFallThrough(BasicBlock* bCur, BasicBlock* bAlt); bool fgCheckEHCanInsertAfterBlock(BasicBlock* blk, unsigned regionIndex, bool putInTryRegion); BasicBlock* fgFindInsertPoint(unsigned regionIndex, bool putInTryRegion, BasicBlock* startBlk, BasicBlock* endBlk, BasicBlock* nearBlk, BasicBlock* jumpBlk, bool runRarely); unsigned fgGetNestingLevel(BasicBlock* block, unsigned* pFinallyNesting = nullptr); void fgRemoveEmptyBlocks(); void fgRemoveStmt(BasicBlock* block, GenTreeStmt* stmt); bool fgCheckRemoveStmt(BasicBlock* block, GenTreeStmt* stmt); void fgCreateLoopPreHeader(unsigned lnum); void fgUnreachableBlock(BasicBlock* block); void fgRemoveConditionalJump(BasicBlock* block); BasicBlock* fgLastBBInMainFunction(); BasicBlock* fgEndBBAfterMainFunction(); void fgUnlinkRange(BasicBlock* bBeg, BasicBlock* bEnd); void fgRemoveBlock(BasicBlock* block, bool unreachable); bool fgCanCompactBlocks(BasicBlock* block, BasicBlock* bNext); void fgCompactBlocks(BasicBlock* block, BasicBlock* bNext); void fgUpdateLoopsAfterCompacting(BasicBlock* block, BasicBlock* bNext); BasicBlock* fgConnectFallThrough(BasicBlock* bSrc, BasicBlock* bDst); bool fgRenumberBlocks(); bool fgExpandRarelyRunBlocks(); bool fgEhAllowsMoveBlock(BasicBlock* bBefore, BasicBlock* bAfter); void fgMoveBlocksAfter(BasicBlock* bStart, BasicBlock* bEnd, BasicBlock* insertAfterBlk); enum FG_RELOCATE_TYPE { FG_RELOCATE_TRY, // relocate the 'try' region FG_RELOCATE_HANDLER // relocate the handler region (including the filter if necessary) }; BasicBlock* fgRelocateEHRange(unsigned regionIndex, FG_RELOCATE_TYPE relocateType); #if FEATURE_EH_FUNCLETS #if defined(_TARGET_ARM_) void fgClearFinallyTargetBit(BasicBlock* block); #endif // defined(_TARGET_ARM_) bool fgIsIntraHandlerPred(BasicBlock* predBlock, BasicBlock* block); bool fgAnyIntraHandlerPreds(BasicBlock* block); void fgInsertFuncletPrologBlock(BasicBlock* block); void fgCreateFuncletPrologBlocks(); void fgCreateFunclets(); #else // !FEATURE_EH_FUNCLETS bool fgRelocateEHRegions(); #endif // !FEATURE_EH_FUNCLETS bool fgOptimizeUncondBranchToSimpleCond(BasicBlock* block, BasicBlock* target); bool fgBlockEndFavorsTailDuplication(BasicBlock* block); bool fgBlockIsGoodTailDuplicationCandidate(BasicBlock* block); bool fgOptimizeEmptyBlock(BasicBlock* block); bool fgOptimizeBranchToEmptyUnconditional(BasicBlock* block, BasicBlock* bDest); bool fgOptimizeBranch(BasicBlock* bJump); bool fgOptimizeSwitchBranches(BasicBlock* block); bool fgOptimizeBranchToNext(BasicBlock* block, BasicBlock* bNext, BasicBlock* bPrev); bool fgOptimizeSwitchJumps(); #ifdef DEBUG void fgPrintEdgeWeights(); #endif void fgComputeBlockAndEdgeWeights(); BasicBlock::weight_t fgComputeMissingBlockWeights(); void fgComputeCalledCount(BasicBlock::weight_t returnWeight); void fgComputeEdgeWeights(); void fgReorderBlocks(); void fgDetermineFirstColdBlock(); bool fgIsForwardBranch(BasicBlock* bJump, BasicBlock* bSrc = nullptr); bool fgUpdateFlowGraph(bool doTailDup = false); void fgFindOperOrder(); // method that returns if you should split here typedef bool(fgSplitPredicate)(GenTree* tree, GenTree* parent, fgWalkData* data); void fgSetBlockOrder(); void fgRemoveReturnBlock(BasicBlock* block); /* Helper code that has been factored out */ inline void fgConvertBBToThrowBB(BasicBlock* block); bool fgCastNeeded(GenTree* tree, var_types toType); GenTree* fgDoNormalizeOnStore(GenTree* tree); GenTree* fgMakeTmpArgNode(fgArgTabEntry* curArgTabEntry); // The following check for loops that don't execute calls bool fgLoopCallMarked; void fgLoopCallTest(BasicBlock* srcBB, BasicBlock* dstBB); void fgLoopCallMark(); void fgMarkLoopHead(BasicBlock* block); unsigned fgGetCodeEstimate(BasicBlock* block); #if DUMP_FLOWGRAPHS const char* fgProcessEscapes(const char* nameIn, escapeMapping_t* map); FILE* fgOpenFlowGraphFile(bool* wbDontClose, Phases phase, LPCWSTR type); bool fgDumpFlowGraph(Phases phase); #endif // DUMP_FLOWGRAPHS #ifdef DEBUG void fgDispDoms(); void fgDispReach(); void fgDispBBLiveness(BasicBlock* block); void fgDispBBLiveness(); void fgTableDispBasicBlock(BasicBlock* block, int ibcColWidth = 0); void fgDispBasicBlocks(BasicBlock* firstBlock, BasicBlock* lastBlock, bool dumpTrees); void fgDispBasicBlocks(bool dumpTrees = false); void fgDumpStmtTree(GenTreeStmt* stmt, unsigned bbNum); void fgDumpBlock(BasicBlock* block); void fgDumpTrees(BasicBlock* firstBlock, BasicBlock* lastBlock); static fgWalkPreFn fgStress64RsltMulCB; void fgStress64RsltMul(); void fgDebugCheckUpdate(); void fgDebugCheckBBlist(bool checkBBNum = false, bool checkBBRefs = true); void fgDebugCheckBlockLinks(); void fgDebugCheckLinks(bool morphTrees = false); void fgDebugCheckStmtsList(BasicBlock* block, bool morphTrees); void fgDebugCheckNodeLinks(BasicBlock* block, GenTreeStmt* stmt); void fgDebugCheckNodesUniqueness(); void fgDebugCheckFlags(GenTree* tree); void fgDebugCheckFlagsHelper(GenTree* tree, unsigned treeFlags, unsigned chkFlags); void fgDebugCheckTryFinallyExits(); #endif static GenTree* fgGetFirstNode(GenTree* tree); //--------------------- Walking the trees in the IR ----------------------- struct fgWalkData { Compiler* compiler; fgWalkPreFn* wtprVisitorFn; fgWalkPostFn* wtpoVisitorFn; void* pCallbackData; // user-provided data bool wtprLclsOnly; // whether to only visit lclvar nodes GenTree* parent; // parent of current node, provided to callback GenTreeStack* parentStack; // stack of parent nodes, if asked for #ifdef DEBUG bool printModified; // callback can use this #endif }; fgWalkResult fgWalkTreePre(GenTree** pTree, fgWalkPreFn* visitor, void* pCallBackData = nullptr, bool lclVarsOnly = false, bool computeStack = false); fgWalkResult fgWalkTree(GenTree** pTree, fgWalkPreFn* preVisitor, fgWalkPostFn* postVisitor, void* pCallBackData = nullptr); void fgWalkAllTreesPre(fgWalkPreFn* visitor, void* pCallBackData); //----- Postorder fgWalkResult fgWalkTreePost(GenTree** pTree, fgWalkPostFn* visitor, void* pCallBackData = nullptr, bool computeStack = false); // An fgWalkPreFn that looks for expressions that have inline throws in // minopts mode. Basically it looks for tress with gtOverflowEx() or // GTF_IND_RNGCHK. It returns WALK_ABORT if one is found. It // returns WALK_SKIP_SUBTREES if GTF_EXCEPT is not set (assumes flags // properly propagated to parent trees). It returns WALK_CONTINUE // otherwise. static fgWalkResult fgChkThrowCB(GenTree** pTree, Compiler::fgWalkData* data); static fgWalkResult fgChkLocAllocCB(GenTree** pTree, Compiler::fgWalkData* data); static fgWalkResult fgChkQmarkCB(GenTree** pTree, Compiler::fgWalkData* data); /************************************************************************** * PROTECTED *************************************************************************/ protected: friend class SsaBuilder; friend struct ValueNumberState; //--------------------- Detect the basic blocks --------------------------- BasicBlock** fgBBs; // Table of pointers to the BBs void fgInitBBLookup(); BasicBlock* fgLookupBB(unsigned addr); bool fgHasBackwardJump; bool fgCanSwitchToOptimized(); void fgSwitchToOptimized(); bool fgMayExplicitTailCall(); void fgFindJumpTargets(const BYTE* codeAddr, IL_OFFSET codeSize, FixedBitVect* jumpTarget); void fgMarkBackwardJump(BasicBlock* startBlock, BasicBlock* endBlock); void fgLinkBasicBlocks(); unsigned fgMakeBasicBlocks(const BYTE* codeAddr, IL_OFFSET codeSize, FixedBitVect* jumpTarget); void fgCheckBasicBlockControlFlow(); void fgControlFlowPermitted(BasicBlock* blkSrc, BasicBlock* blkDest, BOOL IsLeave = false /* is the src a leave block */); bool fgFlowToFirstBlockOfInnerTry(BasicBlock* blkSrc, BasicBlock* blkDest, bool sibling); void fgObserveInlineConstants(OPCODE opcode, const FgStack& stack, bool isInlining); void fgAdjustForAddressExposedOrWrittenThis(); bool fgProfileData_ILSizeMismatch; ICorJitInfo::BlockCounts* fgBlockCounts; UINT32 fgBlockCountsCount; UINT32 fgNumProfileRuns; unsigned fgStressBBProf() { #ifdef DEBUG unsigned result = JitConfig.JitStressBBProf(); if (result == 0) { if (compStressCompile(STRESS_BB_PROFILE, 15)) { result = 1; } } return result; #else return 0; #endif } bool fgHaveProfileData(); bool fgGetProfileWeightForBasicBlock(IL_OFFSET offset, unsigned* weight); void fgInstrumentMethod(); public: // fgIsUsingProfileWeights - returns true if we have real profile data for this method // or if we have some fake profile data for the stress mode bool fgIsUsingProfileWeights() { return (fgHaveProfileData() || fgStressBBProf()); } // fgProfileRunsCount - returns total number of scenario runs for the profile data // or BB_UNITY_WEIGHT when we aren't using profile data. unsigned fgProfileRunsCount() { return fgIsUsingProfileWeights() ? fgNumProfileRuns : BB_UNITY_WEIGHT; } //-------- Insert a statement at the start or end of a basic block -------- #ifdef DEBUG public: static bool fgBlockContainsStatementBounded(BasicBlock* block, GenTreeStmt* stmt, bool answerOnBoundExceeded = true); #endif public: GenTreeStmt* fgInsertStmtAtEnd(BasicBlock* block, GenTree* node); public: // Used by linear scan register allocation GenTreeStmt* fgInsertStmtNearEnd(BasicBlock* block, GenTree* node); private: GenTreeStmt* fgInsertStmtAtBeg(BasicBlock* block, GenTree* node); GenTreeStmt* fgInsertStmtAfter(BasicBlock* block, GenTreeStmt* insertionPoint, GenTreeStmt* stmt); public: // Used by linear scan register allocation GenTreeStmt* fgInsertStmtBefore(BasicBlock* block, GenTreeStmt* insertionPoint, GenTreeStmt* stmt); private: GenTreeStmt* fgInsertStmtListAfter(BasicBlock* block, GenTreeStmt* stmtAfter, GenTreeStmt* stmtList); // Create a new temporary variable to hold the result of *ppTree, // and transform the graph accordingly. GenTree* fgInsertCommaFormTemp(GenTree** ppTree, CORINFO_CLASS_HANDLE structType = nullptr); GenTree* fgMakeMultiUse(GenTree** ppTree); private: // Recognize a bitwise rotation pattern and convert into a GT_ROL or a GT_ROR node. GenTree* fgRecognizeAndMorphBitwiseRotation(GenTree* tree); bool fgOperIsBitwiseRotationRoot(genTreeOps oper); //-------- Determine the order in which the trees will be evaluated ------- unsigned fgTreeSeqNum; GenTree* fgTreeSeqLst; GenTree* fgTreeSeqBeg; GenTree* fgSetTreeSeq(GenTree* tree, GenTree* prev = nullptr, bool isLIR = false); void fgSetTreeSeqHelper(GenTree* tree, bool isLIR); void fgSetTreeSeqFinish(GenTree* tree, bool isLIR); void fgSetStmtSeq(GenTreeStmt* stmt); void fgSetBlockOrder(BasicBlock* block); //------------------------- Morphing -------------------------------------- unsigned fgPtrArgCntMax; public: //------------------------------------------------------------------------ // fgGetPtrArgCntMax: Return the maximum number of pointer-sized stack arguments that calls inside this method // can push on the stack. This value is calculated during morph. // // Return Value: // Returns fgPtrArgCntMax, that is a private field. // unsigned fgGetPtrArgCntMax() const { return fgPtrArgCntMax; } //------------------------------------------------------------------------ // fgSetPtrArgCntMax: Set the maximum number of pointer-sized stack arguments that calls inside this method // can push on the stack. This function is used during StackLevelSetter to fix incorrect morph calculations. // void fgSetPtrArgCntMax(unsigned argCntMax) { fgPtrArgCntMax = argCntMax; } bool compCanEncodePtrArgCntMax(); private: hashBv* fgOutgoingArgTemps; hashBv* fgCurrentlyInUseArgTemps; void fgSetRngChkTarget(GenTree* tree, bool delay = true); BasicBlock* fgSetRngChkTargetInner(SpecialCodeKind kind, bool delay); #if REARRANGE_ADDS void fgMoveOpsLeft(GenTree* tree); #endif bool fgIsCommaThrow(GenTree* tree, bool forFolding = false); bool fgIsThrow(GenTree* tree); bool fgInDifferentRegions(BasicBlock* blk1, BasicBlock* blk2); bool fgIsBlockCold(BasicBlock* block); GenTree* fgMorphCastIntoHelper(GenTree* tree, int helper, GenTree* oper); GenTree* fgMorphIntoHelperCall(GenTree* tree, int helper, GenTreeArgList* args, bool morphArgs = true); GenTree* fgMorphStackArgForVarArgs(unsigned lclNum, var_types varType, unsigned lclOffs); // A "MorphAddrContext" carries information from the surrounding context. If we are evaluating a byref address, // it is useful to know whether the address will be immediately dereferenced, or whether the address value will // be used, perhaps by passing it as an argument to a called method. This affects how null checking is done: // for sufficiently small offsets, we can rely on OS page protection to implicitly null-check addresses that we // know will be dereferenced. To know that reliance on implicit null checking is sound, we must further know that // all offsets between the top-level indirection and the bottom are constant, and that their sum is sufficiently // small; hence the other fields of MorphAddrContext. enum MorphAddrContextKind { MACK_Ind, MACK_Addr, }; struct MorphAddrContext { MorphAddrContextKind m_kind; bool m_allConstantOffsets; // Valid only for "m_kind == MACK_Ind". True iff all offsets between // top-level indirection and here have been constants. size_t m_totalOffset; // Valid only for "m_kind == MACK_Ind", and if "m_allConstantOffsets" is true. // In that case, is the sum of those constant offsets. MorphAddrContext(MorphAddrContextKind kind) : m_kind(kind), m_allConstantOffsets(true), m_totalOffset(0) { } }; // A MACK_CopyBlock context is immutable, so we can just make one of these and share it. static MorphAddrContext s_CopyBlockMAC; #ifdef FEATURE_SIMD GenTree* getSIMDStructFromField(GenTree* tree, var_types* baseTypeOut, unsigned* indexOut, unsigned* simdSizeOut, bool ignoreUsedInSIMDIntrinsic = false); GenTree* fgMorphFieldAssignToSIMDIntrinsicSet(GenTree* tree); GenTree* fgMorphFieldToSIMDIntrinsicGet(GenTree* tree); bool fgMorphCombineSIMDFieldAssignments(BasicBlock* block, GenTreeStmt* stmt); void impMarkContiguousSIMDFieldAssignments(GenTreeStmt* stmt); // fgPreviousCandidateSIMDFieldAsgStmt is only used for tracking previous simd field assignment // in function: Complier::impMarkContiguousSIMDFieldAssignments. GenTreeStmt* fgPreviousCandidateSIMDFieldAsgStmt; #endif // FEATURE_SIMD GenTree* fgMorphArrayIndex(GenTree* tree); GenTree* fgMorphCast(GenTree* tree); GenTree* fgUnwrapProxy(GenTree* objRef); GenTreeFieldList* fgMorphLclArgToFieldlist(GenTreeLclVarCommon* lcl); void fgInitArgInfo(GenTreeCall* call); GenTreeCall* fgMorphArgs(GenTreeCall* call); GenTreeArgList* fgMorphArgList(GenTreeArgList* args, MorphAddrContext* mac); void fgMakeOutgoingStructArgCopy(GenTreeCall* call, GenTree* args, unsigned argIndex, CORINFO_CLASS_HANDLE copyBlkClass); void fgFixupStructReturn(GenTree* call); GenTree* fgMorphLocalVar(GenTree* tree, bool forceRemorph); public: bool fgAddrCouldBeNull(GenTree* addr); private: GenTree* fgMorphField(GenTree* tree, MorphAddrContext* mac); bool fgCanFastTailCall(GenTreeCall* call); bool fgCheckStmtAfterTailCall(); void fgMorphTailCall(GenTreeCall* call, void* pfnCopyArgs); GenTree* fgGetStubAddrArg(GenTreeCall* call); void fgMorphRecursiveFastTailCallIntoLoop(BasicBlock* block, GenTreeCall* recursiveTailCall); GenTreeStmt* fgAssignRecursiveCallArgToCallerParam(GenTree* arg, fgArgTabEntry* argTabEntry, BasicBlock* block, IL_OFFSETX callILOffset, GenTreeStmt* tmpAssignmentInsertionPoint, GenTreeStmt* paramAssignmentInsertionPoint); static int fgEstimateCallStackSize(GenTreeCall* call); GenTree* fgMorphCall(GenTreeCall* call); void fgMorphCallInline(GenTreeCall* call, InlineResult* result); void fgMorphCallInlineHelper(GenTreeCall* call, InlineResult* result); #if DEBUG void fgNoteNonInlineCandidate(GenTreeStmt* stmt, GenTreeCall* call); static fgWalkPreFn fgFindNonInlineCandidate; #endif GenTree* fgOptimizeDelegateConstructor(GenTreeCall* call, CORINFO_CONTEXT_HANDLE* ExactContextHnd, CORINFO_RESOLVED_TOKEN* ldftnToken); GenTree* fgMorphLeaf(GenTree* tree); void fgAssignSetVarDef(GenTree* tree); GenTree* fgMorphOneAsgBlockOp(GenTree* tree); GenTree* fgMorphInitBlock(GenTree* tree); GenTree* fgMorphPromoteLocalInitBlock(GenTreeLclVar* destLclNode, GenTree* initVal, unsigned blockSize); GenTree* fgMorphBlkToInd(GenTreeBlk* tree, var_types type); GenTree* fgMorphGetStructAddr(GenTree** pTree, CORINFO_CLASS_HANDLE clsHnd, bool isRValue = false); GenTree* fgMorphBlkNode(GenTree* tree, bool isDest); GenTree* fgMorphBlockOperand(GenTree* tree, var_types asgType, unsigned blockWidth, bool isDest); void fgMorphUnsafeBlk(GenTreeObj* obj); GenTree* fgMorphCopyBlock(GenTree* tree); GenTree* fgMorphForRegisterFP(GenTree* tree); GenTree* fgMorphSmpOp(GenTree* tree, MorphAddrContext* mac = nullptr); GenTree* fgMorphModToSubMulDiv(GenTreeOp* tree); GenTree* fgMorphSmpOpOptional(GenTreeOp* tree); GenTree* fgMorphRecognizeBoxNullable(GenTree* compare); GenTree* fgMorphToEmulatedFP(GenTree* tree); GenTree* fgMorphConst(GenTree* tree); public: GenTree* fgMorphTree(GenTree* tree, MorphAddrContext* mac = nullptr); private: #if LOCAL_ASSERTION_PROP void fgKillDependentAssertionsSingle(unsigned lclNum DEBUGARG(GenTree* tree)); void fgKillDependentAssertions(unsigned lclNum DEBUGARG(GenTree* tree)); #endif void fgMorphTreeDone(GenTree* tree, GenTree* oldTree = nullptr DEBUGARG(int morphNum = 0)); GenTreeStmt* fgMorphStmt; unsigned fgGetBigOffsetMorphingTemp(var_types type); // We cache one temp per type to be // used when morphing big offset. //----------------------- Liveness analysis ------------------------------- VARSET_TP fgCurUseSet; // vars used by block (before an assignment) VARSET_TP fgCurDefSet; // vars assigned by block (before a use) MemoryKindSet fgCurMemoryUse; // True iff the current basic block uses memory. MemoryKindSet fgCurMemoryDef; // True iff the current basic block modifies memory. MemoryKindSet fgCurMemoryHavoc; // True if the current basic block is known to set memory to a "havoc" value. bool byrefStatesMatchGcHeapStates; // True iff GcHeap and ByrefExposed memory have all the same def points. void fgMarkUseDef(GenTreeLclVarCommon* tree); void fgBeginScopeLife(VARSET_TP* inScope, VarScopeDsc* var); void fgEndScopeLife(VARSET_TP* inScope, VarScopeDsc* var); void fgMarkInScope(BasicBlock* block, VARSET_VALARG_TP inScope); void fgUnmarkInScope(BasicBlock* block, VARSET_VALARG_TP unmarkScope); void fgExtendDbgScopes(); void fgExtendDbgLifetimes(); #ifdef DEBUG void fgDispDebugScopes(); #endif // DEBUG //------------------------------------------------------------------------- // // The following keeps track of any code we've added for things like array // range checking or explicit calls to enable GC, and so on. // public: struct AddCodeDsc { AddCodeDsc* acdNext; BasicBlock* acdDstBlk; // block to which we jump unsigned acdData; SpecialCodeKind acdKind; // what kind of a special block is this? #if !FEATURE_FIXED_OUT_ARGS bool acdStkLvlInit; // has acdStkLvl value been already set? unsigned acdStkLvl; #endif // !FEATURE_FIXED_OUT_ARGS }; private: static unsigned acdHelper(SpecialCodeKind codeKind); AddCodeDsc* fgAddCodeList; bool fgAddCodeModf; bool fgRngChkThrowAdded; AddCodeDsc* fgExcptnTargetCache[SCK_COUNT]; BasicBlock* fgRngChkTarget(BasicBlock* block, SpecialCodeKind kind); BasicBlock* fgAddCodeRef(BasicBlock* srcBlk, unsigned refData, SpecialCodeKind kind); public: AddCodeDsc* fgFindExcptnTarget(SpecialCodeKind kind, unsigned refData); bool fgUseThrowHelperBlocks(); AddCodeDsc* fgGetAdditionalCodeDescriptors() { return fgAddCodeList; } private: bool fgIsCodeAdded(); bool fgIsThrowHlpBlk(BasicBlock* block); #if !FEATURE_FIXED_OUT_ARGS unsigned fgThrowHlpBlkStkLevel(BasicBlock* block); #endif // !FEATURE_FIXED_OUT_ARGS unsigned fgBigOffsetMorphingTemps[TYP_COUNT]; unsigned fgCheckInlineDepthAndRecursion(InlineInfo* inlineInfo); void fgInvokeInlineeCompiler(GenTreeCall* call, InlineResult* result); void fgInsertInlineeBlocks(InlineInfo* pInlineInfo); GenTreeStmt* fgInlinePrependStatements(InlineInfo* inlineInfo); void fgInlineAppendStatements(InlineInfo* inlineInfo, BasicBlock* block, GenTreeStmt* stmt); #if FEATURE_MULTIREG_RET GenTree* fgGetStructAsStructPtr(GenTree* tree); GenTree* fgAssignStructInlineeToVar(GenTree* child, CORINFO_CLASS_HANDLE retClsHnd); void fgAttachStructInlineeToAsg(GenTree* tree, GenTree* child, CORINFO_CLASS_HANDLE retClsHnd); #endif // FEATURE_MULTIREG_RET static fgWalkPreFn fgUpdateInlineReturnExpressionPlaceHolder; static fgWalkPostFn fgLateDevirtualization; #ifdef DEBUG static fgWalkPreFn fgDebugCheckInlineCandidates; void CheckNoTransformableIndirectCallsRemain(); static fgWalkPreFn fgDebugCheckForTransformableIndirectCalls; #endif void fgPromoteStructs(); void fgMorphStructField(GenTree* tree, GenTree* parent); void fgMorphLocalField(GenTree* tree, GenTree* parent); // Reset the refCount for implicit byrefs. void fgResetImplicitByRefRefCount(); // Change implicit byrefs' types from struct to pointer, and for any that were // promoted, create new promoted struct temps. void fgRetypeImplicitByRefArgs(); // Rewrite appearances of implicit byrefs (manifest the implied additional level of indirection). bool fgMorphImplicitByRefArgs(GenTree* tree); GenTree* fgMorphImplicitByRefArgs(GenTree* tree, bool isAddr); // Clear up annotations for any struct promotion temps created for implicit byrefs. void fgMarkDemotedImplicitByRefArgs(); void fgMarkAddressExposedLocals(); static fgWalkPreFn fgUpdateSideEffectsPre; static fgWalkPostFn fgUpdateSideEffectsPost; // The given local variable, required to be a struct variable, is being assigned via // a "lclField", to make it masquerade as an integral type in the ABI. Make sure that // the variable is not enregistered, and is therefore not promoted independently. void fgLclFldAssign(unsigned lclNum); static fgWalkPreFn gtHasLocalsWithAddrOpCB; enum TypeProducerKind { TPK_Unknown = 0, // May not be a RuntimeType TPK_Handle = 1, // RuntimeType via handle TPK_GetType = 2, // RuntimeType via Object.get_Type() TPK_Null = 3, // Tree value is null TPK_Other = 4 // RuntimeType via other means }; TypeProducerKind gtGetTypeProducerKind(GenTree* tree); bool gtIsTypeHandleToRuntimeTypeHelper(GenTreeCall* call); bool gtIsTypeHandleToRuntimeTypeHandleHelper(GenTreeCall* call, CorInfoHelpFunc* pHelper = nullptr); bool gtIsActiveCSE_Candidate(GenTree* tree); #ifdef DEBUG bool fgPrintInlinedMethods; #endif bool fgIsBigOffset(size_t offset); bool fgNeedReturnSpillTemp(); /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX Optimizer XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: void optInit(); void optRemoveRangeCheck(GenTree* tree, GenTreeStmt* stmt); bool optIsRangeCheckRemovable(GenTree* tree); protected: static fgWalkPreFn optValidRangeCheckIndex; /************************************************************************** * *************************************************************************/ protected: // Do hoisting for all loops. void optHoistLoopCode(); // To represent sets of VN's that have already been hoisted in outer loops. typedef JitHashTable, bool> VNToBoolMap; typedef VNToBoolMap VNSet; struct LoopHoistContext { private: // The set of variables hoisted in the current loop (or nullptr if there are none). VNSet* m_pHoistedInCurLoop; public: // Value numbers of expressions that have been hoisted in parent loops in the loop nest. VNSet m_hoistedInParentLoops; // Value numbers of expressions that have been hoisted in the current (or most recent) loop in the nest. // Previous decisions on loop-invariance of value numbers in the current loop. VNToBoolMap m_curLoopVnInvariantCache; VNSet* GetHoistedInCurLoop(Compiler* comp) { if (m_pHoistedInCurLoop == nullptr) { m_pHoistedInCurLoop = new (comp->getAllocatorLoopHoist()) VNSet(comp->getAllocatorLoopHoist()); } return m_pHoistedInCurLoop; } VNSet* ExtractHoistedInCurLoop() { VNSet* res = m_pHoistedInCurLoop; m_pHoistedInCurLoop = nullptr; return res; } LoopHoistContext(Compiler* comp) : m_pHoistedInCurLoop(nullptr) , m_hoistedInParentLoops(comp->getAllocatorLoopHoist()) , m_curLoopVnInvariantCache(comp->getAllocatorLoopHoist()) { } }; // Do hoisting for loop "lnum" (an index into the optLoopTable), and all loops nested within it. // Tracks the expressions that have been hoisted by containing loops by temporary recording their // value numbers in "m_hoistedInParentLoops". This set is not modified by the call. void optHoistLoopNest(unsigned lnum, LoopHoistContext* hoistCtxt); // Do hoisting for a particular loop ("lnum" is an index into the optLoopTable.) // Assumes that expressions have been hoisted in containing loops if their value numbers are in // "m_hoistedInParentLoops". // void optHoistThisLoop(unsigned lnum, LoopHoistContext* hoistCtxt); // Hoist all expressions in "blk" that are invariant in loop "lnum" (an index into the optLoopTable) // outside of that loop. Exempt expressions whose value number is in "m_hoistedInParentLoops"; add VN's of hoisted // expressions to "hoistInLoop". void optHoistLoopExprsForBlock(BasicBlock* blk, unsigned lnum, LoopHoistContext* hoistCtxt); // Return true if the tree looks profitable to hoist out of loop 'lnum'. bool optIsProfitableToHoistableTree(GenTree* tree, unsigned lnum); // Hoist all proper sub-expressions of "tree" (which occurs in "stmt", which occurs in "blk") // that are invariant in loop "lnum" (an index into the optLoopTable) // outside of that loop. Exempt expressions whose value number is in "hoistedInParents"; add VN's of hoisted // expressions to "hoistInLoop". // Returns "true" iff "tree" is loop-invariant (wrt "lnum"). // Assumes that the value of "*firstBlockAndBeforeSideEffect" indicates that we're in the first block, and before // any possible globally visible side effects. Assume is called in evaluation order, and updates this. bool optHoistLoopExprsForTree(GenTree* tree, unsigned lnum, LoopHoistContext* hoistCtxt, bool* firstBlockAndBeforeSideEffect, bool* pHoistable, bool* pCctorDependent); // Performs the hoisting 'tree' into the PreHeader for loop 'lnum' void optHoistCandidate(GenTree* tree, unsigned lnum, LoopHoistContext* hoistCtxt); // Returns true iff the ValueNum "vn" represents a value that is loop-invariant in "lnum". // Constants and init values are always loop invariant. // VNPhi's connect VN's to the SSA definition, so we can know if the SSA def occurs in the loop. bool optVNIsLoopInvariant(ValueNum vn, unsigned lnum, VNToBoolMap* recordedVNs); // Returns "true" iff "tree" is valid at the head of loop "lnum", in the context of the hoist substitution // "subst". If "tree" is a local SSA var, it is valid if its SSA definition occurs outside of the loop, or // if it is in the domain of "subst" (meaning that it's definition has been previously hoisted, with a "standin" // local.) If tree is a constant, it is valid. Otherwise, if it is an operator, it is valid iff its children are. bool optTreeIsValidAtLoopHead(GenTree* tree, unsigned lnum); // If "blk" is the entry block of a natural loop, returns true and sets "*pLnum" to the index of the loop // in the loop table. bool optBlockIsLoopEntry(BasicBlock* blk, unsigned* pLnum); // Records the set of "side effects" of all loops: fields (object instance and static) // written to, and SZ-array element type equivalence classes updated. void optComputeLoopSideEffects(); private: // Requires "lnum" to be the index of an outermost loop in the loop table. Traverses the body of that loop, // including all nested loops, and records the set of "side effects" of the loop: fields (object instance and // static) written to, and SZ-array element type equivalence classes updated. void optComputeLoopNestSideEffects(unsigned lnum); // Add the side effects of "blk" (which is required to be within a loop) to all loops of which it is a part. void optComputeLoopSideEffectsOfBlock(BasicBlock* blk); // Hoist the expression "expr" out of loop "lnum". void optPerformHoistExpr(GenTree* expr, unsigned lnum); public: void optOptimizeBools(); private: GenTree* optIsBoolCond(GenTree* condBranch, GenTree** compPtr, bool* boolPtr); #ifdef DEBUG void optOptimizeBoolsGcStress(BasicBlock* condBlock); #endif public: void optOptimizeLayout(); // Optimize the BasicBlock layout of the method void optOptimizeLoops(); // for "while-do" loops duplicates simple loop conditions and transforms // the loop into a "do-while" loop // Also finds all natural loops and records them in the loop table // Optionally clone loops in the loop table. void optCloneLoops(); // Clone loop "loopInd" in the loop table. void optCloneLoop(unsigned loopInd, LoopCloneContext* context); // Ensure that loop "loopInd" has a unique head block. (If the existing entry has // non-loop predecessors other than the head entry, create a new, empty block that goes (only) to the entry, // and redirects the preds of the entry to this new block.) Sets the weight of the newly created block to // "ambientWeight". void optEnsureUniqueHead(unsigned loopInd, unsigned ambientWeight); void optUnrollLoops(); // Unrolls loops (needs to have cost info) protected: // This enumeration describes what is killed by a call. enum callInterf { CALLINT_NONE, // no interference (most helpers) CALLINT_REF_INDIRS, // kills GC ref indirections (SETFIELD OBJ) CALLINT_SCL_INDIRS, // kills non GC ref indirections (SETFIELD non-OBJ) CALLINT_ALL_INDIRS, // kills both GC ref and non GC ref indirections (SETFIELD STRUCT) CALLINT_ALL, // kills everything (normal method call) }; public: // A "LoopDsc" describes a ("natural") loop. We (currently) require the body of a loop to be a contiguous (in // bbNext order) sequence of basic blocks. (At times, we may require the blocks in a loop to be "properly numbered" // in bbNext order; we use comparisons on the bbNum to decide order.) // The blocks that define the body are // first <= top <= entry <= bottom . // The "head" of the loop is a block outside the loop that has "entry" as a successor. We only support loops with a // single 'head' block. The meanings of these blocks are given in the definitions below. Also see the picture at // Compiler::optFindNaturalLoops(). struct LoopDsc { BasicBlock* lpHead; // HEAD of the loop (not part of the looping of the loop) -- has ENTRY as a successor. BasicBlock* lpFirst; // FIRST block (in bbNext order) reachable within this loop. (May be part of a nested // loop, but not the outer loop.) BasicBlock* lpTop; // loop TOP (the back edge from lpBottom reaches here) (in most cases FIRST and TOP are the // same) BasicBlock* lpEntry; // the ENTRY in the loop (in most cases TOP or BOTTOM) BasicBlock* lpBottom; // loop BOTTOM (from here we have a back edge to the TOP) BasicBlock* lpExit; // if a single exit loop this is the EXIT (in most cases BOTTOM) callInterf lpAsgCall; // "callInterf" for calls in the loop ALLVARSET_TP lpAsgVars; // set of vars assigned within the loop (all vars, not just tracked) varRefKinds lpAsgInds : 8; // set of inds modified within the loop unsigned short lpFlags; // Mask of the LPFLG_* constants unsigned char lpExitCnt; // number of exits from the loop unsigned char lpParent; // The index of the most-nested loop that completely contains this one, // or else BasicBlock::NOT_IN_LOOP if no such loop exists. unsigned char lpChild; // The index of a nested loop, or else BasicBlock::NOT_IN_LOOP if no child exists. // (Actually, an "immediately" nested loop -- // no other child of this loop is a parent of lpChild.) unsigned char lpSibling; // The index of another loop that is an immediate child of lpParent, // or else BasicBlock::NOT_IN_LOOP. One can enumerate all the children of a loop // by following "lpChild" then "lpSibling" links. #define LPFLG_DO_WHILE 0x0001 // it's a do-while loop (i.e ENTRY is at the TOP) #define LPFLG_ONE_EXIT 0x0002 // the loop has only one exit #define LPFLG_ITER 0x0004 // for (i = icon or lclVar; test_condition(); i++) #define LPFLG_HOISTABLE 0x0008 // the loop is in a form that is suitable for hoisting expressions #define LPFLG_CONST 0x0010 // for (i=icon;i, bool> FieldHandleSet; FieldHandleSet* lpFieldsModified; // This has entries (mappings to "true") for all static field and object // instance fields modified // in the loop. typedef JitHashTable, bool> ClassHandleSet; ClassHandleSet* lpArrayElemTypesModified; // Bits set indicate the set of sz array element types such that // arrays of that type are modified // in the loop. // Adds the variable liveness information for 'blk' to 'this' LoopDsc void AddVariableLiveness(Compiler* comp, BasicBlock* blk); inline void AddModifiedField(Compiler* comp, CORINFO_FIELD_HANDLE fldHnd); // This doesn't *always* take a class handle -- it can also take primitive types, encoded as class handles // (shifted left, with a low-order bit set to distinguish.) // Use the {Encode/Decode}ElemType methods to construct/destruct these. inline void AddModifiedElemType(Compiler* comp, CORINFO_CLASS_HANDLE structHnd); /* The following values are set only for iterator loops, i.e. has the flag LPFLG_ITER set */ GenTree* lpIterTree; // The "i = i const" tree unsigned lpIterVar(); // iterator variable # int lpIterConst(); // the constant with which the iterator is incremented genTreeOps lpIterOper(); // the type of the operation on the iterator (ASG_ADD, ASG_SUB, etc.) void VERIFY_lpIterTree(); var_types lpIterOperType(); // For overflow instructions union { int lpConstInit; // initial constant value of iterator : Valid if LPFLG_CONST_INIT unsigned lpVarInit; // initial local var number to which we initialize the iterator : Valid if // LPFLG_VAR_INIT }; /* The following is for LPFLG_ITER loops only (i.e. the loop condition is "i RELOP const or var" */ GenTree* lpTestTree; // pointer to the node containing the loop test genTreeOps lpTestOper(); // the type of the comparison between the iterator and the limit (GT_LE, GT_GE, etc.) void VERIFY_lpTestTree(); bool lpIsReversed(); // true if the iterator node is the second operand in the loop condition GenTree* lpIterator(); // the iterator node in the loop test GenTree* lpLimit(); // the limit node in the loop test int lpConstLimit(); // limit constant value of iterator - loop condition is "i RELOP const" : Valid if // LPFLG_CONST_LIMIT unsigned lpVarLimit(); // the lclVar # in the loop condition ( "i RELOP lclVar" ) : Valid if // LPFLG_VAR_LIMIT bool lpArrLenLimit(Compiler* comp, ArrIndex* index); // The array length in the loop condition ( "i RELOP // arr.len" or "i RELOP arr[i][j].len" ) : Valid if // LPFLG_ARRLEN_LIMIT // Returns "true" iff "*this" contains the blk. bool lpContains(BasicBlock* blk) { return lpFirst->bbNum <= blk->bbNum && blk->bbNum <= lpBottom->bbNum; } // Returns "true" iff "*this" (properly) contains the range [first, bottom] (allowing firsts // to be equal, but requiring bottoms to be different.) bool lpContains(BasicBlock* first, BasicBlock* bottom) { return lpFirst->bbNum <= first->bbNum && bottom->bbNum < lpBottom->bbNum; } // Returns "true" iff "*this" (properly) contains "lp2" (allowing firsts to be equal, but requiring // bottoms to be different.) bool lpContains(const LoopDsc& lp2) { return lpContains(lp2.lpFirst, lp2.lpBottom); } // Returns "true" iff "*this" is (properly) contained by the range [first, bottom] // (allowing firsts to be equal, but requiring bottoms to be different.) bool lpContainedBy(BasicBlock* first, BasicBlock* bottom) { return first->bbNum <= lpFirst->bbNum && lpBottom->bbNum < bottom->bbNum; } // Returns "true" iff "*this" is (properly) contained by "lp2" // (allowing firsts to be equal, but requiring bottoms to be different.) bool lpContainedBy(const LoopDsc& lp2) { return lpContains(lp2.lpFirst, lp2.lpBottom); } // Returns "true" iff "*this" is disjoint from the range [top, bottom]. bool lpDisjoint(BasicBlock* first, BasicBlock* bottom) { return bottom->bbNum < lpFirst->bbNum || lpBottom->bbNum < first->bbNum; } // Returns "true" iff "*this" is disjoint from "lp2". bool lpDisjoint(const LoopDsc& lp2) { return lpDisjoint(lp2.lpFirst, lp2.lpBottom); } // Returns "true" iff the loop is well-formed (see code for defn). bool lpWellFormed() { return lpFirst->bbNum <= lpTop->bbNum && lpTop->bbNum <= lpEntry->bbNum && lpEntry->bbNum <= lpBottom->bbNum && (lpHead->bbNum < lpTop->bbNum || lpHead->bbNum > lpBottom->bbNum); } }; protected: bool fgMightHaveLoop(); // returns true if there are any backedges bool fgHasLoops; // True if this method has any loops, set in fgComputeReachability public: LoopDsc* optLoopTable; // loop descriptor table unsigned char optLoopCount; // number of tracked loops bool optRecordLoop(BasicBlock* head, BasicBlock* first, BasicBlock* top, BasicBlock* entry, BasicBlock* bottom, BasicBlock* exit, unsigned char exitCnt); protected: unsigned optCallCount; // number of calls made in the method unsigned optIndirectCallCount; // number of virtual, interface and indirect calls made in the method unsigned optNativeCallCount; // number of Pinvoke/Native calls made in the method unsigned optLoopsCloned; // number of loops cloned in the current method. #ifdef DEBUG unsigned optFindLoopNumberFromBeginBlock(BasicBlock* begBlk); void optPrintLoopInfo(unsigned loopNum, BasicBlock* lpHead, BasicBlock* lpFirst, BasicBlock* lpTop, BasicBlock* lpEntry, BasicBlock* lpBottom, unsigned char lpExitCnt, BasicBlock* lpExit, unsigned parentLoop = BasicBlock::NOT_IN_LOOP); void optPrintLoopInfo(unsigned lnum); void optPrintLoopRecording(unsigned lnum); void optCheckPreds(); #endif void optSetBlockWeights(); void optMarkLoopBlocks(BasicBlock* begBlk, BasicBlock* endBlk, bool excludeEndBlk); void optUnmarkLoopBlocks(BasicBlock* begBlk, BasicBlock* endBlk); void optUpdateLoopsBeforeRemoveBlock(BasicBlock* block, bool skipUnmarkLoop = false); bool optIsLoopTestEvalIntoTemp(GenTreeStmt* testStmt, GenTreeStmt** newTestStmt); unsigned optIsLoopIncrTree(GenTree* incr); bool optCheckIterInLoopTest(unsigned loopInd, GenTree* test, BasicBlock* from, BasicBlock* to, unsigned iterVar); bool optComputeIterInfo(GenTree* incr, BasicBlock* from, BasicBlock* to, unsigned* pIterVar); bool optPopulateInitInfo(unsigned loopInd, GenTree* init, unsigned iterVar); bool optExtractInitTestIncr( BasicBlock* head, BasicBlock* bottom, BasicBlock* exit, GenTree** ppInit, GenTree** ppTest, GenTree** ppIncr); void optFindNaturalLoops(); // Ensures that all the loops in the loop nest rooted at "loopInd" (an index into the loop table) are 'canonical' -- // each loop has a unique "top." Returns "true" iff the flowgraph has been modified. bool optCanonicalizeLoopNest(unsigned char loopInd); // Ensures that the loop "loopInd" (an index into the loop table) is 'canonical' -- it has a unique "top," // unshared with any other loop. Returns "true" iff the flowgraph has been modified bool optCanonicalizeLoop(unsigned char loopInd); // Requires "l1" to be a valid loop table index, and not "BasicBlock::NOT_IN_LOOP". Requires "l2" to be // a valid loop table index, or else "BasicBlock::NOT_IN_LOOP". Returns true // iff "l2" is not NOT_IN_LOOP, and "l1" contains "l2". bool optLoopContains(unsigned l1, unsigned l2); // Requires "loopInd" to be a valid index into the loop table. // Updates the loop table by changing loop "loopInd", whose head is required // to be "from", to be "to". Also performs this transformation for any // loop nested in "loopInd" that shares the same head as "loopInd". void optUpdateLoopHead(unsigned loopInd, BasicBlock* from, BasicBlock* to); // Updates the successors of "blk": if "blk2" is a successor of "blk", and there is a mapping for "blk2->blk3" in // "redirectMap", change "blk" so that "blk3" is this successor. Note that the predecessor lists are not updated. void optRedirectBlock(BasicBlock* blk, BlockToBlockMap* redirectMap); // Marks the containsCall information to "lnum" and any parent loops. void AddContainsCallAllContainingLoops(unsigned lnum); // Adds the variable liveness information from 'blk' to "lnum" and any parent loops. void AddVariableLivenessAllContainingLoops(unsigned lnum, BasicBlock* blk); // Adds "fldHnd" to the set of modified fields of "lnum" and any parent loops. void AddModifiedFieldAllContainingLoops(unsigned lnum, CORINFO_FIELD_HANDLE fldHnd); // Adds "elemType" to the set of modified array element types of "lnum" and any parent loops. void AddModifiedElemTypeAllContainingLoops(unsigned lnum, CORINFO_CLASS_HANDLE elemType); // Requires that "from" and "to" have the same "bbJumpKind" (perhaps because "to" is a clone // of "from".) Copies the jump destination from "from" to "to". void optCopyBlkDest(BasicBlock* from, BasicBlock* to); // The depth of the loop described by "lnum" (an index into the loop table.) (0 == top level) unsigned optLoopDepth(unsigned lnum) { unsigned par = optLoopTable[lnum].lpParent; if (par == BasicBlock::NOT_IN_LOOP) { return 0; } else { return 1 + optLoopDepth(par); } } void fgOptWhileLoop(BasicBlock* block); bool optComputeLoopRep(int constInit, int constLimit, int iterInc, genTreeOps iterOper, var_types iterType, genTreeOps testOper, bool unsignedTest, bool dupCond, unsigned* iterCount); private: static fgWalkPreFn optIsVarAssgCB; protected: bool optIsVarAssigned(BasicBlock* beg, BasicBlock* end, GenTree* skip, unsigned var); bool optIsVarAssgLoop(unsigned lnum, unsigned var); int optIsSetAssgLoop(unsigned lnum, ALLVARSET_VALARG_TP vars, varRefKinds inds = VR_NONE); bool optNarrowTree(GenTree* tree, var_types srct, var_types dstt, ValueNumPair vnpNarrow, bool doit); /************************************************************************** * Optimization conditions *************************************************************************/ bool optFastCodeOrBlendedLoop(BasicBlock::weight_t bbWeight); bool optPentium4(void); bool optAvoidIncDec(BasicBlock::weight_t bbWeight); bool optAvoidIntMult(void); #if FEATURE_ANYCSE protected: // The following is the upper limit on how many expressions we'll keep track // of for the CSE analysis. // static const unsigned MAX_CSE_CNT = EXPSET_SZ; static const int MIN_CSE_COST = 2; // Keeps tracked cse indices BitVecTraits* cseTraits; EXPSET_TP cseFull; /* Generic list of nodes - used by the CSE logic */ struct treeLst { treeLst* tlNext; GenTree* tlTree; }; struct treeStmtLst { treeStmtLst* tslNext; GenTree* tslTree; // tree node GenTreeStmt* tslStmt; // statement containing the tree BasicBlock* tslBlock; // block containing the statement }; // The following logic keeps track of expressions via a simple hash table. struct CSEdsc { CSEdsc* csdNextInBucket; // used by the hash table unsigned csdHashKey; // the orginal hashkey unsigned csdIndex; // 1..optCSECandidateCount char csdLiveAcrossCall; // 0 or 1 unsigned short csdDefCount; // definition count unsigned short csdUseCount; // use count (excluding the implicit uses at defs) unsigned csdDefWtCnt; // weighted def count unsigned csdUseWtCnt; // weighted use count (excluding the implicit uses at defs) GenTree* csdTree; // treenode containing the 1st occurance GenTreeStmt* csdStmt; // stmt containing the 1st occurance BasicBlock* csdBlock; // block containing the 1st occurance treeStmtLst* csdTreeList; // list of matching tree nodes: head treeStmtLst* csdTreeLast; // list of matching tree nodes: tail ValueNum defExcSetPromise; // The exception set that is now required for all defs of this CSE. // This will be set to NoVN if we decide to abandon this CSE ValueNum defExcSetCurrent; // The set of exceptions we currently can use for CSE uses. ValueNum defConservNormVN; // if all def occurrences share the same conservative normal value // number, this will reflect it; otherwise, NoVN. }; static const size_t s_optCSEhashSize; CSEdsc** optCSEhash; CSEdsc** optCSEtab; typedef JitHashTable, GenTree*> NodeToNodeMap; NodeToNodeMap* optCseCheckedBoundMap; // Maps bound nodes to ancestor compares that should be // re-numbered with the bound to improve range check elimination // Given a compare, look for a cse candidate checked bound feeding it and add a map entry if found. void optCseUpdateCheckedBoundMap(GenTree* compare); void optCSEstop(); CSEdsc* optCSEfindDsc(unsigned index); bool optUnmarkCSE(GenTree* tree); // user defined callback data for the tree walk function optCSE_MaskHelper() struct optCSE_MaskData { EXPSET_TP CSE_defMask; EXPSET_TP CSE_useMask; }; // Treewalk helper for optCSE_DefMask and optCSE_UseMask static fgWalkPreFn optCSE_MaskHelper; // This function walks all the node for an given tree // and return the mask of CSE definitions and uses for the tree // void optCSE_GetMaskData(GenTree* tree, optCSE_MaskData* pMaskData); // Given a binary tree node return true if it is safe to swap the order of evaluation for op1 and op2. bool optCSE_canSwap(GenTree* firstNode, GenTree* secondNode); bool optCSE_canSwap(GenTree* tree); static int __cdecl optCSEcostCmpEx(const void* op1, const void* op2); static int __cdecl optCSEcostCmpSz(const void* op1, const void* op2); void optCleanupCSEs(); #ifdef DEBUG void optEnsureClearCSEInfo(); #endif // DEBUG #endif // FEATURE_ANYCSE #if FEATURE_VALNUM_CSE /************************************************************************** * Value Number based CSEs *************************************************************************/ public: void optOptimizeValnumCSEs(); protected: void optValnumCSE_Init(); unsigned optValnumCSE_Index(GenTree* tree, GenTreeStmt* stmt); unsigned optValnumCSE_Locate(); void optValnumCSE_InitDataFlow(); void optValnumCSE_DataFlow(); void optValnumCSE_Availablity(); void optValnumCSE_Heuristic(); #endif // FEATURE_VALNUM_CSE #if FEATURE_ANYCSE bool optDoCSE; // True when we have found a duplicate CSE tree bool optValnumCSE_phase; // True when we are executing the optValnumCSE_phase unsigned optCSECandidateTotal; // Grand total of CSE candidates for both Lexical and ValNum unsigned optCSECandidateCount; // Count of CSE's candidates, reset for Lexical and ValNum CSE's unsigned optCSEstart; // The first local variable number that is a CSE unsigned optCSEcount; // The total count of CSE's introduced. unsigned optCSEweight; // The weight of the current block when we are // scanning for CSE expressions bool optIsCSEcandidate(GenTree* tree); // lclNumIsTrueCSE returns true if the LclVar was introduced by the CSE phase of the compiler // bool lclNumIsTrueCSE(unsigned lclNum) const { return ((optCSEcount > 0) && (lclNum >= optCSEstart) && (lclNum < optCSEstart + optCSEcount)); } // lclNumIsCSE returns true if the LclVar should be treated like a CSE with regards to constant prop. // bool lclNumIsCSE(unsigned lclNum) const { return lvaTable[lclNum].lvIsCSE; } #ifdef DEBUG bool optConfigDisableCSE(); bool optConfigDisableCSE2(); #endif void optOptimizeCSEs(); #endif // FEATURE_ANYCSE struct isVarAssgDsc { GenTree* ivaSkip; #ifdef DEBUG void* ivaSelf; #endif unsigned ivaVar; // Variable we are interested in, or -1 ALLVARSET_TP ivaMaskVal; // Set of variables assigned to. This is a set of all vars, not tracked vars. bool ivaMaskIncomplete; // Variables not representable in ivaMaskVal were assigned to. varRefKinds ivaMaskInd; // What kind of indirect assignments are there? callInterf ivaMaskCall; // What kind of calls are there? }; static callInterf optCallInterf(GenTreeCall* call); public: // VN based copy propagation. typedef ArrayStack GenTreePtrStack; typedef JitHashTable, GenTreePtrStack*> LclNumToGenTreePtrStack; // Kill set to track variables with intervening definitions. VARSET_TP optCopyPropKillSet; // Copy propagation functions. void optCopyProp(BasicBlock* block, GenTreeStmt* stmt, GenTree* tree, LclNumToGenTreePtrStack* curSsaName); void optBlockCopyPropPopStacks(BasicBlock* block, LclNumToGenTreePtrStack* curSsaName); void optBlockCopyProp(BasicBlock* block, LclNumToGenTreePtrStack* curSsaName); bool optIsSsaLocal(GenTree* tree); int optCopyProp_LclVarScore(LclVarDsc* lclVarDsc, LclVarDsc* copyVarDsc, bool preferOp2); void optVnCopyProp(); INDEBUG(void optDumpCopyPropStack(LclNumToGenTreePtrStack* curSsaName)); /************************************************************************** * Early value propagation *************************************************************************/ struct SSAName { unsigned m_lvNum; unsigned m_ssaNum; SSAName(unsigned lvNum, unsigned ssaNum) : m_lvNum(lvNum), m_ssaNum(ssaNum) { } static unsigned GetHashCode(SSAName ssaNm) { return (ssaNm.m_lvNum << 16) | (ssaNm.m_ssaNum); } static bool Equals(SSAName ssaNm1, SSAName ssaNm2) { return (ssaNm1.m_lvNum == ssaNm2.m_lvNum) && (ssaNm1.m_ssaNum == ssaNm2.m_ssaNum); } }; #define OMF_HAS_NEWARRAY 0x00000001 // Method contains 'new' of an array #define OMF_HAS_NEWOBJ 0x00000002 // Method contains 'new' of an object type. #define OMF_HAS_ARRAYREF 0x00000004 // Method contains array element loads or stores. #define OMF_HAS_VTABLEREF 0x00000008 // Method contains method table reference. #define OMF_HAS_NULLCHECK 0x00000010 // Method contains null check. #define OMF_HAS_FATPOINTER 0x00000020 // Method contains call, that needs fat pointer transformation. #define OMF_HAS_OBJSTACKALLOC 0x00000040 // Method contains an object allocated on the stack. #define OMF_HAS_GUARDEDDEVIRT 0x00000080 // Method contains guarded devirtualization candidate bool doesMethodHaveFatPointer() { return (optMethodFlags & OMF_HAS_FATPOINTER) != 0; } void setMethodHasFatPointer() { optMethodFlags |= OMF_HAS_FATPOINTER; } void clearMethodHasFatPointer() { optMethodFlags &= ~OMF_HAS_FATPOINTER; } void addFatPointerCandidate(GenTreeCall* call); bool doesMethodHaveGuardedDevirtualization() { return (optMethodFlags & OMF_HAS_GUARDEDDEVIRT) != 0; } void setMethodHasGuardedDevirtualization() { optMethodFlags |= OMF_HAS_GUARDEDDEVIRT; } void clearMethodHasGuardedDevirtualization() { optMethodFlags &= ~OMF_HAS_GUARDEDDEVIRT; } void addGuardedDevirtualizationCandidate(GenTreeCall* call, CORINFO_METHOD_HANDLE methodHandle, CORINFO_CLASS_HANDLE classHandle, unsigned methodAttr, unsigned classAttr); unsigned optMethodFlags; // Recursion bound controls how far we can go backwards tracking for a SSA value. // No throughput diff was found with backward walk bound between 3-8. static const int optEarlyPropRecurBound = 5; enum class optPropKind { OPK_INVALID, OPK_ARRAYLEN, OPK_OBJ_GETTYPE, OPK_NULLCHECK }; bool gtIsVtableRef(GenTree* tree); GenTree* getArrayLengthFromAllocation(GenTree* tree); GenTree* getObjectHandleNodeFromAllocation(GenTree* tree); GenTree* optPropGetValueRec(unsigned lclNum, unsigned ssaNum, optPropKind valueKind, int walkDepth); GenTree* optPropGetValue(unsigned lclNum, unsigned ssaNum, optPropKind valueKind); GenTree* optEarlyPropRewriteTree(GenTree* tree); bool optDoEarlyPropForBlock(BasicBlock* block); bool optDoEarlyPropForFunc(); void optEarlyProp(); void optFoldNullCheck(GenTree* tree); bool optCanMoveNullCheckPastTree(GenTree* tree, bool isInsideTry); #if ASSERTION_PROP /************************************************************************** * Value/Assertion propagation *************************************************************************/ public: // Data structures for assertion prop BitVecTraits* apTraits; ASSERT_TP apFull; enum optAssertionKind { OAK_INVALID, OAK_EQUAL, OAK_NOT_EQUAL, OAK_SUBRANGE, OAK_NO_THROW, OAK_COUNT }; enum optOp1Kind { O1K_INVALID, O1K_LCLVAR, O1K_ARR_BND, O1K_BOUND_OPER_BND, O1K_BOUND_LOOP_BND, O1K_CONSTANT_LOOP_BND, O1K_EXACT_TYPE, O1K_SUBTYPE, O1K_VALUE_NUMBER, O1K_COUNT }; enum optOp2Kind { O2K_INVALID, O2K_LCLVAR_COPY, O2K_IND_CNS_INT, O2K_CONST_INT, O2K_CONST_LONG, O2K_CONST_DOUBLE, O2K_ARR_LEN, O2K_SUBRANGE, O2K_COUNT }; struct AssertionDsc { optAssertionKind assertionKind; struct SsaVar { unsigned lclNum; // assigned to or property of this local var number unsigned ssaNum; }; struct ArrBnd { ValueNum vnIdx; ValueNum vnLen; }; struct AssertionDscOp1 { optOp1Kind kind; // a normal LclVar, or Exact-type or Subtype ValueNum vn; union { SsaVar lcl; ArrBnd bnd; }; } op1; struct AssertionDscOp2 { optOp2Kind kind; // a const or copy assignment ValueNum vn; struct IntVal { ssize_t iconVal; // integer unsigned iconFlags; // gtFlags }; struct Range // integer subrange { ssize_t loBound; ssize_t hiBound; }; union { SsaVar lcl; IntVal u1; __int64 lconVal; double dconVal; Range u2; }; } op2; bool IsCheckedBoundArithBound() { return ((assertionKind == OAK_EQUAL || assertionKind == OAK_NOT_EQUAL) && op1.kind == O1K_BOUND_OPER_BND); } bool IsCheckedBoundBound() { return ((assertionKind == OAK_EQUAL || assertionKind == OAK_NOT_EQUAL) && op1.kind == O1K_BOUND_LOOP_BND); } bool IsConstantBound() { return ((assertionKind == OAK_EQUAL || assertionKind == OAK_NOT_EQUAL) && op1.kind == O1K_CONSTANT_LOOP_BND); } bool IsBoundsCheckNoThrow() { return ((assertionKind == OAK_NO_THROW) && (op1.kind == O1K_ARR_BND)); } bool IsCopyAssertion() { return ((assertionKind == OAK_EQUAL) && (op1.kind == O1K_LCLVAR) && (op2.kind == O2K_LCLVAR_COPY)); } static bool SameKind(AssertionDsc* a1, AssertionDsc* a2) { return a1->assertionKind == a2->assertionKind && a1->op1.kind == a2->op1.kind && a1->op2.kind == a2->op2.kind; } static bool ComplementaryKind(optAssertionKind kind, optAssertionKind kind2) { if (kind == OAK_EQUAL) { return kind2 == OAK_NOT_EQUAL; } else if (kind == OAK_NOT_EQUAL) { return kind2 == OAK_EQUAL; } return false; } static ssize_t GetLowerBoundForIntegralType(var_types type) { switch (type) { case TYP_BYTE: return SCHAR_MIN; case TYP_SHORT: return SHRT_MIN; case TYP_INT: return INT_MIN; case TYP_BOOL: case TYP_UBYTE: case TYP_USHORT: case TYP_UINT: return 0; default: unreached(); } } static ssize_t GetUpperBoundForIntegralType(var_types type) { switch (type) { case TYP_BOOL: return 1; case TYP_BYTE: return SCHAR_MAX; case TYP_SHORT: return SHRT_MAX; case TYP_INT: return INT_MAX; case TYP_UBYTE: return UCHAR_MAX; case TYP_USHORT: return USHRT_MAX; case TYP_UINT: return UINT_MAX; default: unreached(); } } bool HasSameOp1(AssertionDsc* that, bool vnBased) { if (op1.kind != that->op1.kind) { return false; } else if (op1.kind == O1K_ARR_BND) { assert(vnBased); return (op1.bnd.vnIdx == that->op1.bnd.vnIdx) && (op1.bnd.vnLen == that->op1.bnd.vnLen); } else { return ((vnBased && (op1.vn == that->op1.vn)) || (!vnBased && (op1.lcl.lclNum == that->op1.lcl.lclNum))); } } bool HasSameOp2(AssertionDsc* that, bool vnBased) { if (op2.kind != that->op2.kind) { return false; } switch (op2.kind) { case O2K_IND_CNS_INT: case O2K_CONST_INT: return ((op2.u1.iconVal == that->op2.u1.iconVal) && (op2.u1.iconFlags == that->op2.u1.iconFlags)); case O2K_CONST_LONG: return (op2.lconVal == that->op2.lconVal); case O2K_CONST_DOUBLE: // exact match because of positive and negative zero. return (memcmp(&op2.dconVal, &that->op2.dconVal, sizeof(double)) == 0); case O2K_LCLVAR_COPY: case O2K_ARR_LEN: return (op2.lcl.lclNum == that->op2.lcl.lclNum) && (!vnBased || op2.lcl.ssaNum == that->op2.lcl.ssaNum); case O2K_SUBRANGE: return ((op2.u2.loBound == that->op2.u2.loBound) && (op2.u2.hiBound == that->op2.u2.hiBound)); case O2K_INVALID: // we will return false break; default: assert(!"Unexpected value for op2.kind in AssertionDsc."); break; } return false; } bool Complementary(AssertionDsc* that, bool vnBased) { return ComplementaryKind(assertionKind, that->assertionKind) && HasSameOp1(that, vnBased) && HasSameOp2(that, vnBased); } bool Equals(AssertionDsc* that, bool vnBased) { if (assertionKind != that->assertionKind) { return false; } else if (assertionKind == OAK_NO_THROW) { assert(op2.kind == O2K_INVALID); return HasSameOp1(that, vnBased); } else { return HasSameOp1(that, vnBased) && HasSameOp2(that, vnBased); } } }; protected: static fgWalkPreFn optAddCopiesCallback; static fgWalkPreFn optVNAssertionPropCurStmtVisitor; unsigned optAddCopyLclNum; GenTree* optAddCopyAsgnNode; bool optLocalAssertionProp; // indicates that we are performing local assertion prop bool optAssertionPropagated; // set to true if we modified the trees bool optAssertionPropagatedCurrentStmt; #ifdef DEBUG GenTree* optAssertionPropCurrentTree; #endif AssertionIndex* optComplementaryAssertionMap; JitExpandArray* optAssertionDep; // table that holds dependent assertions (assertions // using the value of a local var) for each local var AssertionDsc* optAssertionTabPrivate; // table that holds info about value assignments AssertionIndex optAssertionCount; // total number of assertions in the assertion table AssertionIndex optMaxAssertionCount; public: void optVnNonNullPropCurStmt(BasicBlock* block, GenTreeStmt* stmt, GenTree* tree); fgWalkResult optVNConstantPropCurStmt(BasicBlock* block, GenTreeStmt* stmt, GenTree* tree); GenTree* optVNConstantPropOnJTrue(BasicBlock* block, GenTree* test); GenTree* optVNConstantPropOnTree(BasicBlock* block, GenTree* tree); GenTree* optExtractSideEffListFromConst(GenTree* tree); AssertionIndex GetAssertionCount() { return optAssertionCount; } ASSERT_TP* bbJtrueAssertionOut; typedef JitHashTable, ASSERT_TP> ValueNumToAssertsMap; ValueNumToAssertsMap* optValueNumToAsserts; // Assertion prop helpers. ASSERT_TP& GetAssertionDep(unsigned lclNum); AssertionDsc* optGetAssertion(AssertionIndex assertIndex); void optAssertionInit(bool isLocalProp); void optAssertionTraitsInit(AssertionIndex assertionCount); #if LOCAL_ASSERTION_PROP void optAssertionReset(AssertionIndex limit); void optAssertionRemove(AssertionIndex index); #endif // Assertion prop data flow functions. void optAssertionPropMain(); GenTreeStmt* optVNAssertionPropCurStmt(BasicBlock* block, GenTreeStmt* stmt); bool optIsTreeKnownIntValue(bool vnBased, GenTree* tree, ssize_t* pConstant, unsigned* pIconFlags); ASSERT_TP* optInitAssertionDataflowFlags(); ASSERT_TP* optComputeAssertionGen(); // Assertion Gen functions. void optAssertionGen(GenTree* tree); AssertionIndex optAssertionGenPhiDefn(GenTree* tree); AssertionInfo optCreateJTrueBoundsAssertion(GenTree* tree); AssertionInfo optAssertionGenJtrue(GenTree* tree); AssertionIndex optCreateJtrueAssertions(GenTree* op1, GenTree* op2, Compiler::optAssertionKind assertionKind); AssertionIndex optFindComplementary(AssertionIndex assertionIndex); void optMapComplementary(AssertionIndex assertionIndex, AssertionIndex index); // Assertion creation functions. AssertionIndex optCreateAssertion(GenTree* op1, GenTree* op2, optAssertionKind assertionKind); AssertionIndex optCreateAssertion(GenTree* op1, GenTree* op2, optAssertionKind assertionKind, AssertionDsc* assertion); void optCreateComplementaryAssertion(AssertionIndex assertionIndex, GenTree* op1, GenTree* op2); bool optAssertionVnInvolvesNan(AssertionDsc* assertion); AssertionIndex optAddAssertion(AssertionDsc* assertion); void optAddVnAssertionMapping(ValueNum vn, AssertionIndex index); #ifdef DEBUG void optPrintVnAssertionMapping(); #endif ASSERT_TP optGetVnMappedAssertions(ValueNum vn); // Used for respective assertion propagations. AssertionIndex optAssertionIsSubrange(GenTree* tree, var_types fromType, var_types toType, ASSERT_VALARG_TP assertions); AssertionIndex optAssertionIsSubtype(GenTree* tree, GenTree* methodTableArg, ASSERT_VALARG_TP assertions); AssertionIndex optAssertionIsNonNullInternal(GenTree* op, ASSERT_VALARG_TP assertions); bool optAssertionIsNonNull(GenTree* op, ASSERT_VALARG_TP assertions DEBUGARG(bool* pVnBased) DEBUGARG(AssertionIndex* pIndex)); // Used for Relop propagation. AssertionIndex optGlobalAssertionIsEqualOrNotEqual(ASSERT_VALARG_TP assertions, GenTree* op1, GenTree* op2); AssertionIndex optGlobalAssertionIsEqualOrNotEqualZero(ASSERT_VALARG_TP assertions, GenTree* op1); AssertionIndex optLocalAssertionIsEqualOrNotEqual( optOp1Kind op1Kind, unsigned lclNum, optOp2Kind op2Kind, ssize_t cnsVal, ASSERT_VALARG_TP assertions); // Assertion prop for lcl var functions. bool optAssertionProp_LclVarTypeCheck(GenTree* tree, LclVarDsc* lclVarDsc, LclVarDsc* copyVarDsc); GenTree* optCopyAssertionProp(AssertionDsc* curAssertion, GenTree* tree, GenTreeStmt* stmt DEBUGARG(AssertionIndex index)); GenTree* optConstantAssertionProp(AssertionDsc* curAssertion, GenTree* tree, GenTreeStmt* stmt DEBUGARG(AssertionIndex index)); // Assertion propagation functions. GenTree* optAssertionProp(ASSERT_VALARG_TP assertions, GenTree* tree, GenTreeStmt* stmt); GenTree* optAssertionProp_LclVar(ASSERT_VALARG_TP assertions, GenTree* tree, GenTreeStmt* stmt); GenTree* optAssertionProp_Ind(ASSERT_VALARG_TP assertions, GenTree* tree, GenTreeStmt* stmt); GenTree* optAssertionProp_Cast(ASSERT_VALARG_TP assertions, GenTree* tree, GenTreeStmt* stmt); GenTree* optAssertionProp_Call(ASSERT_VALARG_TP assertions, GenTreeCall* call, GenTreeStmt* stmt); GenTree* optAssertionProp_RelOp(ASSERT_VALARG_TP assertions, GenTree* tree, GenTreeStmt* stmt); GenTree* optAssertionProp_Comma(ASSERT_VALARG_TP assertions, GenTree* tree, GenTreeStmt* stmt); GenTree* optAssertionProp_BndsChk(ASSERT_VALARG_TP assertions, GenTree* tree); GenTree* optAssertionPropGlobal_RelOp(ASSERT_VALARG_TP assertions, GenTree* tree, GenTreeStmt* stmt); GenTree* optAssertionPropLocal_RelOp(ASSERT_VALARG_TP assertions, GenTree* tree, GenTreeStmt* stmt); GenTree* optAssertionProp_Update(GenTree* newTree, GenTree* tree, GenTreeStmt* stmt); GenTree* optNonNullAssertionProp_Call(ASSERT_VALARG_TP assertions, GenTreeCall* call); // Implied assertion functions. void optImpliedAssertions(AssertionIndex assertionIndex, ASSERT_TP& activeAssertions); void optImpliedByTypeOfAssertions(ASSERT_TP& activeAssertions); void optImpliedByCopyAssertion(AssertionDsc* copyAssertion, AssertionDsc* depAssertion, ASSERT_TP& result); void optImpliedByConstAssertion(AssertionDsc* curAssertion, ASSERT_TP& result); #ifdef DEBUG void optPrintAssertion(AssertionDsc* newAssertion, AssertionIndex assertionIndex = 0); void optDebugCheckAssertion(AssertionDsc* assertion); void optDebugCheckAssertions(AssertionIndex AssertionIndex); #endif void optAddCopies(); #endif // ASSERTION_PROP /************************************************************************** * Range checks *************************************************************************/ public: struct LoopCloneVisitorInfo { LoopCloneContext* context; unsigned loopNum; GenTreeStmt* stmt; LoopCloneVisitorInfo(LoopCloneContext* context, unsigned loopNum, GenTreeStmt* stmt) : context(context), loopNum(loopNum), stmt(nullptr) { } }; bool optIsStackLocalInvariant(unsigned loopNum, unsigned lclNum); bool optExtractArrIndex(GenTree* tree, ArrIndex* result, unsigned lhsNum); bool optReconstructArrIndex(GenTree* tree, ArrIndex* result, unsigned lhsNum); bool optIdentifyLoopOptInfo(unsigned loopNum, LoopCloneContext* context); static fgWalkPreFn optCanOptimizeByLoopCloningVisitor; fgWalkResult optCanOptimizeByLoopCloning(GenTree* tree, LoopCloneVisitorInfo* info); void optObtainLoopCloningOpts(LoopCloneContext* context); bool optIsLoopClonable(unsigned loopInd); bool optCanCloneLoops(); #ifdef DEBUG void optDebugLogLoopCloning(BasicBlock* block, GenTreeStmt* insertBefore); #endif void optPerformStaticOptimizations(unsigned loopNum, LoopCloneContext* context DEBUGARG(bool fastPath)); bool optComputeDerefConditions(unsigned loopNum, LoopCloneContext* context); bool optDeriveLoopCloningConditions(unsigned loopNum, LoopCloneContext* context); BasicBlock* optInsertLoopChoiceConditions(LoopCloneContext* context, unsigned loopNum, BasicBlock* head, BasicBlock* slow); protected: ssize_t optGetArrayRefScaleAndIndex(GenTree* mul, GenTree** pIndex DEBUGARG(bool bRngChk)); bool optReachWithoutCall(BasicBlock* srcBB, BasicBlock* dstBB); protected: bool optLoopsMarked; /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX RegAlloc XX XX XX XX Does the register allocation and puts the remaining lclVars on the stack XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: regNumber raUpdateRegStateForArg(RegState* regState, LclVarDsc* argDsc); void raMarkStkVars(); protected: // Some things are used by both LSRA and regpredict allocators. FrameType rpFrameType; bool rpMustCreateEBPCalled; // Set to true after we have called rpMustCreateEBPFrame once bool rpMustCreateEBPFrame(INDEBUG(const char** wbReason)); private: Lowering* m_pLowering; // Lowering; needed to Lower IR that's added or modified after Lowering. LinearScanInterface* m_pLinearScan; // Linear Scan allocator /* raIsVarargsStackArg is called by raMaskStkVars and by lvaSortByRefCount. It identifies the special case where a varargs function has a parameter passed on the stack, other than the special varargs handle. Such parameters require special treatment, because they cannot be tracked by the GC (their offsets in the stack are not known at compile time). */ bool raIsVarargsStackArg(unsigned lclNum) { #ifdef _TARGET_X86_ LclVarDsc* varDsc = &lvaTable[lclNum]; assert(varDsc->lvIsParam); return (info.compIsVarArgs && !varDsc->lvIsRegArg && (lclNum != lvaVarargsHandleArg)); #else // _TARGET_X86_ return false; #endif // _TARGET_X86_ } /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX EEInterface XX XX XX XX Get to the class and method info from the Execution Engine given XX XX tokens for the class and method XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: // Get handles void eeGetCallInfo(CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_RESOLVED_TOKEN* pConstrainedToken, CORINFO_CALLINFO_FLAGS flags, CORINFO_CALL_INFO* pResult); inline CORINFO_CALLINFO_FLAGS addVerifyFlag(CORINFO_CALLINFO_FLAGS flags); void eeGetFieldInfo(CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_ACCESS_FLAGS flags, CORINFO_FIELD_INFO* pResult); // Get the flags BOOL eeIsValueClass(CORINFO_CLASS_HANDLE clsHnd); #if defined(DEBUG) || defined(FEATURE_JIT_METHOD_PERF) || defined(FEATURE_SIMD) || defined(TRACK_LSRA_STATS) bool IsSuperPMIException(unsigned code) { // Copied from NDP\clr\src\ToolBox\SuperPMI\SuperPMI-Shared\ErrorHandling.h const unsigned EXCEPTIONCODE_DebugBreakorAV = 0xe0421000; const unsigned EXCEPTIONCODE_MC = 0xe0422000; const unsigned EXCEPTIONCODE_LWM = 0xe0423000; const unsigned EXCEPTIONCODE_SASM = 0xe0424000; const unsigned EXCEPTIONCODE_SSYM = 0xe0425000; const unsigned EXCEPTIONCODE_CALLUTILS = 0xe0426000; const unsigned EXCEPTIONCODE_TYPEUTILS = 0xe0427000; const unsigned EXCEPTIONCODE_ASSERT = 0xe0440000; switch (code) { case EXCEPTIONCODE_DebugBreakorAV: case EXCEPTIONCODE_MC: case EXCEPTIONCODE_LWM: case EXCEPTIONCODE_SASM: case EXCEPTIONCODE_SSYM: case EXCEPTIONCODE_CALLUTILS: case EXCEPTIONCODE_TYPEUTILS: case EXCEPTIONCODE_ASSERT: return true; default: return false; } } const char* eeGetMethodName(CORINFO_METHOD_HANDLE hnd, const char** className); const char* eeGetMethodFullName(CORINFO_METHOD_HANDLE hnd); bool eeIsNativeMethod(CORINFO_METHOD_HANDLE method); CORINFO_METHOD_HANDLE eeGetMethodHandleForNative(CORINFO_METHOD_HANDLE method); #endif var_types eeGetArgType(CORINFO_ARG_LIST_HANDLE list, CORINFO_SIG_INFO* sig); var_types eeGetArgType(CORINFO_ARG_LIST_HANDLE list, CORINFO_SIG_INFO* sig, bool* isPinned); unsigned eeGetArgSize(CORINFO_ARG_LIST_HANDLE list, CORINFO_SIG_INFO* sig); // VOM info, method sigs void eeGetSig(unsigned sigTok, CORINFO_MODULE_HANDLE scope, CORINFO_CONTEXT_HANDLE context, CORINFO_SIG_INFO* retSig); void eeGetCallSiteSig(unsigned sigTok, CORINFO_MODULE_HANDLE scope, CORINFO_CONTEXT_HANDLE context, CORINFO_SIG_INFO* retSig); void eeGetMethodSig(CORINFO_METHOD_HANDLE methHnd, CORINFO_SIG_INFO* retSig, CORINFO_CLASS_HANDLE owner = nullptr); // Method entry-points, instrs CORINFO_METHOD_HANDLE eeMarkNativeTarget(CORINFO_METHOD_HANDLE method); CORINFO_EE_INFO eeInfo; bool eeInfoInitialized; CORINFO_EE_INFO* eeGetEEInfo(); // Gets the offset of a SDArray's first element unsigned eeGetArrayDataOffset(var_types type); // Gets the offset of a MDArray's first element unsigned eeGetMDArrayDataOffset(var_types type, unsigned rank); GenTree* eeGetPInvokeCookie(CORINFO_SIG_INFO* szMetaSig); // Returns the page size for the target machine as reported by the EE. target_size_t eeGetPageSize() { return (target_size_t)eeGetEEInfo()->osPageSize; } // Returns the frame size at which we will generate a loop to probe the stack. target_size_t getVeryLargeFrameSize() { #ifdef _TARGET_ARM_ // The looping probe code is 40 bytes, whereas the straight-line probing for // the (0x2000..0x3000) case is 44, so use looping for anything 0x2000 bytes // or greater, to generate smaller code. return 2 * eeGetPageSize(); #else return 3 * eeGetPageSize(); #endif } //------------------------------------------------------------------------ // VirtualStubParam: virtual stub dispatch extra parameter (slot address). // // It represents Abi and target specific registers for the parameter. // class VirtualStubParamInfo { public: VirtualStubParamInfo(bool isCoreRTABI) { #if defined(_TARGET_X86_) reg = REG_EAX; regMask = RBM_EAX; #elif defined(_TARGET_AMD64_) if (isCoreRTABI) { reg = REG_R10; regMask = RBM_R10; } else { reg = REG_R11; regMask = RBM_R11; } #elif defined(_TARGET_ARM_) if (isCoreRTABI) { reg = REG_R12; regMask = RBM_R12; } else { reg = REG_R4; regMask = RBM_R4; } #elif defined(_TARGET_ARM64_) reg = REG_R11; regMask = RBM_R11; #else #error Unsupported or unset target architecture #endif } regNumber GetReg() const { return reg; } _regMask_enum GetRegMask() const { return regMask; } private: regNumber reg; _regMask_enum regMask; }; VirtualStubParamInfo* virtualStubParamInfo; bool IsTargetAbi(CORINFO_RUNTIME_ABI abi) { return eeGetEEInfo()->targetAbi == abi; } bool generateCFIUnwindCodes() { #if defined(_TARGET_UNIX_) return IsTargetAbi(CORINFO_CORERT_ABI); #else return false; #endif } // Debugging support - Line number info void eeGetStmtOffsets(); unsigned eeBoundariesCount; struct boundariesDsc { UNATIVE_OFFSET nativeIP; IL_OFFSET ilOffset; unsigned sourceReason; } * eeBoundaries; // Boundaries to report to EE void eeSetLIcount(unsigned count); void eeSetLIinfo(unsigned which, UNATIVE_OFFSET offs, unsigned srcIP, bool stkEmpty, bool callInstruction); void eeSetLIdone(); #ifdef DEBUG static void eeDispILOffs(IL_OFFSET offs); static void eeDispLineInfo(const boundariesDsc* line); void eeDispLineInfos(); #endif // DEBUG // Debugging support - Local var info void eeGetVars(); unsigned eeVarsCount; struct VarResultInfo { UNATIVE_OFFSET startOffset; UNATIVE_OFFSET endOffset; DWORD varNumber; CodeGenInterface::siVarLoc loc; } * eeVars; void eeSetLVcount(unsigned count); void eeSetLVinfo(unsigned which, UNATIVE_OFFSET startOffs, UNATIVE_OFFSET length, unsigned varNum, const CodeGenInterface::siVarLoc& loc); void eeSetLVdone(); #ifdef DEBUG void eeDispVar(ICorDebugInfo::NativeVarInfo* var); void eeDispVars(CORINFO_METHOD_HANDLE ftn, ULONG32 cVars, ICorDebugInfo::NativeVarInfo* vars); #endif // DEBUG // ICorJitInfo wrappers void eeReserveUnwindInfo(BOOL isFunclet, BOOL isColdCode, ULONG unwindSize); void eeAllocUnwindInfo(BYTE* pHotCode, BYTE* pColdCode, ULONG startOffset, ULONG endOffset, ULONG unwindSize, BYTE* pUnwindBlock, CorJitFuncKind funcKind); void eeSetEHcount(unsigned cEH); void eeSetEHinfo(unsigned EHnumber, const CORINFO_EH_CLAUSE* clause); WORD eeGetRelocTypeHint(void* target); // ICorStaticInfo wrapper functions bool eeTryResolveToken(CORINFO_RESOLVED_TOKEN* resolvedToken); #if defined(UNIX_AMD64_ABI) #ifdef DEBUG static void dumpSystemVClassificationType(SystemVClassificationType ct); #endif // DEBUG void eeGetSystemVAmd64PassStructInRegisterDescriptor( /*IN*/ CORINFO_CLASS_HANDLE structHnd, /*OUT*/ SYSTEMV_AMD64_CORINFO_STRUCT_REG_PASSING_DESCRIPTOR* structPassInRegDescPtr); #endif // UNIX_AMD64_ABI template bool eeRunWithErrorTrap(void (*function)(ParamType*), ParamType* param) { return eeRunWithErrorTrapImp(reinterpret_cast(function), reinterpret_cast(param)); } bool eeRunWithErrorTrapImp(void (*function)(void*), void* param); // Utility functions const char* eeGetFieldName(CORINFO_FIELD_HANDLE fieldHnd, const char** classNamePtr = nullptr); #if defined(DEBUG) const wchar_t* eeGetCPString(size_t stringHandle); #endif const char* eeGetClassName(CORINFO_CLASS_HANDLE clsHnd); static CORINFO_METHOD_HANDLE eeFindHelper(unsigned helper); static CorInfoHelpFunc eeGetHelperNum(CORINFO_METHOD_HANDLE method); static fgWalkPreFn CountSharedStaticHelper; static bool IsSharedStaticHelper(GenTree* tree); static bool IsTreeAlwaysHoistable(GenTree* tree); static bool IsGcSafePoint(GenTree* tree); static CORINFO_FIELD_HANDLE eeFindJitDataOffs(unsigned jitDataOffs); // returns true/false if 'field' is a Jit Data offset static bool eeIsJitDataOffs(CORINFO_FIELD_HANDLE field); // returns a number < 0 if 'field' is not a Jit Data offset, otherwise the data offset (limited to 2GB) static int eeGetJitDataOffs(CORINFO_FIELD_HANDLE field); /*****************************************************************************/ /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX CodeGenerator XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: CodeGenInterface* codeGen; // The following holds information about instr offsets in terms of generated code. struct IPmappingDsc { IPmappingDsc* ipmdNext; // next line# record IL_OFFSETX ipmdILoffsx; // the instr offset emitLocation ipmdNativeLoc; // the emitter location of the native code corresponding to the IL offset bool ipmdIsLabel; // Can this code be a branch label? }; // Record the instr offset mapping to the generated code IPmappingDsc* genIPmappingList; IPmappingDsc* genIPmappingLast; // Managed RetVal - A side hash table meant to record the mapping from a // GT_CALL node to its IL offset. This info is used to emit sequence points // that can be used by debugger to determine the native offset at which the // managed RetVal will be available. // // In fact we can store IL offset in a GT_CALL node. This was ruled out in // favor of a side table for two reasons: 1) We need IL offset for only those // GT_CALL nodes (created during importation) that correspond to an IL call and // whose return type is other than TYP_VOID. 2) GT_CALL node is a frequently used // structure and IL offset is needed only when generating debuggable code. Therefore // it is desirable to avoid memory size penalty in retail scenarios. typedef JitHashTable, IL_OFFSETX> CallSiteILOffsetTable; CallSiteILOffsetTable* genCallSite2ILOffsetMap; unsigned genReturnLocal; // Local number for the return value when applicable. BasicBlock* genReturnBB; // jumped to when not optimizing for speed. // The following properties are part of CodeGenContext. Getters are provided here for // convenience and backward compatibility, but the properties can only be set by invoking // the setter on CodeGenContext directly. __declspec(property(get = getEmitter)) emitter* genEmitter; emitter* getEmitter() const { return codeGen->getEmitter(); } bool isFramePointerUsed() const { return codeGen->isFramePointerUsed(); } __declspec(property(get = getInterruptible, put = setInterruptible)) bool genInterruptible; bool getInterruptible() { return codeGen->genInterruptible; } void setInterruptible(bool value) { codeGen->setInterruptible(value); } #ifdef _TARGET_ARMARCH_ __declspec(property(get = getHasTailCalls, put = setHasTailCalls)) bool hasTailCalls; bool getHasTailCalls() { return codeGen->hasTailCalls; } void setHasTailCalls(bool value) { codeGen->setHasTailCalls(value); } #endif // _TARGET_ARMARCH_ #if DOUBLE_ALIGN const bool genDoubleAlign() { return codeGen->doDoubleAlign(); } DWORD getCanDoubleAlign(); bool shouldDoubleAlign(unsigned refCntStk, unsigned refCntReg, unsigned refCntWtdReg, unsigned refCntStkParam, unsigned refCntWtdStkDbl); #endif // DOUBLE_ALIGN __declspec(property(get = getFullPtrRegMap, put = setFullPtrRegMap)) bool genFullPtrRegMap; bool getFullPtrRegMap() { return codeGen->genFullPtrRegMap; } void setFullPtrRegMap(bool value) { codeGen->setFullPtrRegMap(value); } // Things that MAY belong either in CodeGen or CodeGenContext #if FEATURE_EH_FUNCLETS FuncInfoDsc* compFuncInfos; unsigned short compCurrFuncIdx; unsigned short compFuncInfoCount; unsigned short compFuncCount() { assert(fgFuncletsCreated); return compFuncInfoCount; } #else // !FEATURE_EH_FUNCLETS // This is a no-op when there are no funclets! void genUpdateCurrentFunclet(BasicBlock* block) { return; } FuncInfoDsc compFuncInfoRoot; static const unsigned compCurrFuncIdx = 0; unsigned short compFuncCount() { return 1; } #endif // !FEATURE_EH_FUNCLETS FuncInfoDsc* funCurrentFunc(); void funSetCurrentFunc(unsigned funcIdx); FuncInfoDsc* funGetFunc(unsigned funcIdx); unsigned int funGetFuncIdx(BasicBlock* block); // LIVENESS VARSET_TP compCurLife; // current live variables GenTree* compCurLifeTree; // node after which compCurLife has been computed template void compChangeLife(VARSET_VALARG_TP newLife); template inline void compUpdateLife(VARSET_VALARG_TP newLife); // Gets a register mask that represent the kill set for a helper call since // not all JIT Helper calls follow the standard ABI on the target architecture. regMaskTP compHelperCallKillSet(CorInfoHelpFunc helper); #ifdef _TARGET_ARM_ // Requires that "varDsc" be a promoted struct local variable being passed as an argument, beginning at // "firstArgRegNum", which is assumed to have already been aligned to the register alignment restriction of the // struct type. Adds bits to "*pArgSkippedRegMask" for any argument registers *not* used in passing "varDsc" -- // i.e., internal "holes" caused by internal alignment constraints. For example, if the struct contained an int and // a double, and we at R0 (on ARM), then R1 would be skipped, and the bit for R1 would be added to the mask. void fgAddSkippedRegsInPromotedStructArg(LclVarDsc* varDsc, unsigned firstArgRegNum, regMaskTP* pArgSkippedRegMask); #endif // _TARGET_ARM_ // If "tree" is a indirection (GT_IND, or GT_OBJ) whose arg is an ADDR, whose arg is a LCL_VAR, return that LCL_VAR // node, else NULL. static GenTree* fgIsIndirOfAddrOfLocal(GenTree* tree); // This map is indexed by GT_OBJ nodes that are address of promoted struct variables, which // have been annotated with the GTF_VAR_DEATH flag. If such a node is *not* mapped in this // table, one may assume that all the (tracked) field vars die at this GT_OBJ. Otherwise, // the node maps to a pointer to a VARSET_TP, containing set bits for each of the tracked field // vars of the promoted struct local that go dead at the given node (the set bits are the bits // for the tracked var indices of the field vars, as in a live var set). // // The map is allocated on demand so all map operations should use one of the following three // wrapper methods. NodeToVarsetPtrMap* m_promotedStructDeathVars; NodeToVarsetPtrMap* GetPromotedStructDeathVars() { if (m_promotedStructDeathVars == nullptr) { m_promotedStructDeathVars = new (getAllocator()) NodeToVarsetPtrMap(getAllocator()); } return m_promotedStructDeathVars; } void ClearPromotedStructDeathVars() { if (m_promotedStructDeathVars != nullptr) { m_promotedStructDeathVars->RemoveAll(); } } bool LookupPromotedStructDeathVars(GenTree* tree, VARSET_TP** bits) { bits = nullptr; bool result = false; if (m_promotedStructDeathVars != nullptr) { result = m_promotedStructDeathVars->Lookup(tree, bits); } return result; } /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX UnwindInfo XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ #if !defined(__GNUC__) #pragma region Unwind information #endif public: // // Infrastructure functions: start/stop/reserve/emit. // void unwindBegProlog(); void unwindEndProlog(); void unwindBegEpilog(); void unwindEndEpilog(); void unwindReserve(); void unwindEmit(void* pHotCode, void* pColdCode); // // Specific unwind information functions: called by code generation to indicate a particular // prolog or epilog unwindable instruction has been generated. // void unwindPush(regNumber reg); void unwindAllocStack(unsigned size); void unwindSetFrameReg(regNumber reg, unsigned offset); void unwindSaveReg(regNumber reg, unsigned offset); #if defined(_TARGET_ARM_) void unwindPushMaskInt(regMaskTP mask); void unwindPushMaskFloat(regMaskTP mask); void unwindPopMaskInt(regMaskTP mask); void unwindPopMaskFloat(regMaskTP mask); void unwindBranch16(); // The epilog terminates with a 16-bit branch (e.g., "bx lr") void unwindNop(unsigned codeSizeInBytes); // Generate unwind NOP code. 'codeSizeInBytes' is 2 or 4 bytes. Only // called via unwindPadding(). void unwindPadding(); // Generate a sequence of unwind NOP codes representing instructions between the last // instruction and the current location. #endif // _TARGET_ARM_ #if defined(_TARGET_ARM64_) void unwindNop(); void unwindPadding(); // Generate a sequence of unwind NOP codes representing instructions between the last // instruction and the current location. void unwindSaveReg(regNumber reg, int offset); // str reg, [sp, #offset] void unwindSaveRegPreindexed(regNumber reg, int offset); // str reg, [sp, #offset]! void unwindSaveRegPair(regNumber reg1, regNumber reg2, int offset); // stp reg1, reg2, [sp, #offset] void unwindSaveRegPairPreindexed(regNumber reg1, regNumber reg2, int offset); // stp reg1, reg2, [sp, #offset]! void unwindSaveNext(); // unwind code: save_next void unwindReturn(regNumber reg); // ret lr #endif // defined(_TARGET_ARM64_) // // Private "helper" functions for the unwind implementation. // private: #if FEATURE_EH_FUNCLETS void unwindGetFuncLocations(FuncInfoDsc* func, bool getHotSectionData, /* OUT */ emitLocation** ppStartLoc, /* OUT */ emitLocation** ppEndLoc); #endif // FEATURE_EH_FUNCLETS void unwindReserveFunc(FuncInfoDsc* func); void unwindEmitFunc(FuncInfoDsc* func, void* pHotCode, void* pColdCode); #if defined(_TARGET_AMD64_) || (defined(_TARGET_X86_) && FEATURE_EH_FUNCLETS) void unwindReserveFuncHelper(FuncInfoDsc* func, bool isHotCode); void unwindEmitFuncHelper(FuncInfoDsc* func, void* pHotCode, void* pColdCode, bool isHotCode); #endif // _TARGET_AMD64_ || (_TARGET_X86_ && FEATURE_EH_FUNCLETS) UNATIVE_OFFSET unwindGetCurrentOffset(FuncInfoDsc* func); #if defined(_TARGET_AMD64_) void unwindBegPrologWindows(); void unwindPushWindows(regNumber reg); void unwindAllocStackWindows(unsigned size); void unwindSetFrameRegWindows(regNumber reg, unsigned offset); void unwindSaveRegWindows(regNumber reg, unsigned offset); #ifdef UNIX_AMD64_ABI void unwindSaveRegCFI(regNumber reg, unsigned offset); #endif // UNIX_AMD64_ABI #elif defined(_TARGET_ARM_) void unwindPushPopMaskInt(regMaskTP mask, bool useOpsize16); void unwindPushPopMaskFloat(regMaskTP mask); #endif // _TARGET_ARM_ #if defined(_TARGET_UNIX_) short mapRegNumToDwarfReg(regNumber reg); void createCfiCode(FuncInfoDsc* func, UNATIVE_OFFSET codeOffset, UCHAR opcode, short dwarfReg, INT offset = 0); void unwindPushPopCFI(regNumber reg); void unwindBegPrologCFI(); void unwindPushPopMaskCFI(regMaskTP regMask, bool isFloat); void unwindAllocStackCFI(unsigned size); void unwindSetFrameRegCFI(regNumber reg, unsigned offset); void unwindEmitFuncCFI(FuncInfoDsc* func, void* pHotCode, void* pColdCode); #ifdef DEBUG void DumpCfiInfo(bool isHotCode, UNATIVE_OFFSET startOffset, UNATIVE_OFFSET endOffset, DWORD cfiCodeBytes, const CFI_CODE* const pCfiCode); #endif #endif // _TARGET_UNIX_ #if !defined(__GNUC__) #pragma endregion // Note: region is NOT under !defined(__GNUC__) #endif /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX SIMD XX XX XX XX Info about SIMD types, methods and the SIMD assembly (i.e. the assembly XX XX that contains the distinguished, well-known SIMD type definitions). XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ // Get highest available level for SIMD codegen SIMDLevel getSIMDSupportLevel() { #if defined(_TARGET_XARCH_) if (compSupports(InstructionSet_AVX2)) { return SIMD_AVX2_Supported; } if (compSupports(InstructionSet_SSE42)) { return SIMD_SSE4_Supported; } // min bar is SSE2 return SIMD_SSE2_Supported; #else assert(!"Available instruction set(s) for SIMD codegen is not defined for target arch"); unreached(); return SIMD_Not_Supported; #endif } #ifdef FEATURE_SIMD // Should we support SIMD intrinsics? bool featureSIMD; // Should we recognize SIMD types? // We always do this on ARM64 to support HVA types. bool supportSIMDTypes() { #ifdef _TARGET_ARM64_ return true; #else return featureSIMD; #endif } // Have we identified any SIMD types? // This is currently used by struct promotion to avoid getting type information for a struct // field to see if it is a SIMD type, if we haven't seen any SIMD types or operations in // the method. bool _usesSIMDTypes; bool usesSIMDTypes() { return _usesSIMDTypes; } void setUsesSIMDTypes(bool value) { _usesSIMDTypes = value; } // This is a temp lclVar allocated on the stack as TYP_SIMD. It is used to implement intrinsics // that require indexed access to the individual fields of the vector, which is not well supported // by the hardware. It is allocated when/if such situations are encountered during Lowering. unsigned lvaSIMDInitTempVarNum; struct SIMDHandlesCache { // SIMD Types CORINFO_CLASS_HANDLE SIMDFloatHandle; CORINFO_CLASS_HANDLE SIMDDoubleHandle; CORINFO_CLASS_HANDLE SIMDIntHandle; CORINFO_CLASS_HANDLE SIMDUShortHandle; CORINFO_CLASS_HANDLE SIMDUByteHandle; CORINFO_CLASS_HANDLE SIMDShortHandle; CORINFO_CLASS_HANDLE SIMDByteHandle; CORINFO_CLASS_HANDLE SIMDLongHandle; CORINFO_CLASS_HANDLE SIMDUIntHandle; CORINFO_CLASS_HANDLE SIMDULongHandle; CORINFO_CLASS_HANDLE SIMDVector2Handle; CORINFO_CLASS_HANDLE SIMDVector3Handle; CORINFO_CLASS_HANDLE SIMDVector4Handle; CORINFO_CLASS_HANDLE SIMDVectorHandle; #ifdef FEATURE_HW_INTRINSICS #if defined(_TARGET_ARM64_) CORINFO_CLASS_HANDLE Vector64FloatHandle; CORINFO_CLASS_HANDLE Vector64IntHandle; CORINFO_CLASS_HANDLE Vector64UShortHandle; CORINFO_CLASS_HANDLE Vector64UByteHandle; CORINFO_CLASS_HANDLE Vector64ShortHandle; CORINFO_CLASS_HANDLE Vector64ByteHandle; CORINFO_CLASS_HANDLE Vector64UIntHandle; #endif // defined(_TARGET_ARM64_) CORINFO_CLASS_HANDLE Vector128FloatHandle; CORINFO_CLASS_HANDLE Vector128DoubleHandle; CORINFO_CLASS_HANDLE Vector128IntHandle; CORINFO_CLASS_HANDLE Vector128UShortHandle; CORINFO_CLASS_HANDLE Vector128UByteHandle; CORINFO_CLASS_HANDLE Vector128ShortHandle; CORINFO_CLASS_HANDLE Vector128ByteHandle; CORINFO_CLASS_HANDLE Vector128LongHandle; CORINFO_CLASS_HANDLE Vector128UIntHandle; CORINFO_CLASS_HANDLE Vector128ULongHandle; #if defined(_TARGET_XARCH_) CORINFO_CLASS_HANDLE Vector256FloatHandle; CORINFO_CLASS_HANDLE Vector256DoubleHandle; CORINFO_CLASS_HANDLE Vector256IntHandle; CORINFO_CLASS_HANDLE Vector256UShortHandle; CORINFO_CLASS_HANDLE Vector256UByteHandle; CORINFO_CLASS_HANDLE Vector256ShortHandle; CORINFO_CLASS_HANDLE Vector256ByteHandle; CORINFO_CLASS_HANDLE Vector256LongHandle; CORINFO_CLASS_HANDLE Vector256UIntHandle; CORINFO_CLASS_HANDLE Vector256ULongHandle; #endif // defined(_TARGET_XARCH_) #endif // FEATURE_HW_INTRINSICS SIMDHandlesCache() { memset(this, 0, sizeof(*this)); } }; SIMDHandlesCache* m_simdHandleCache; // Get an appropriate "zero" for the given type and class handle. GenTree* gtGetSIMDZero(var_types simdType, var_types baseType, CORINFO_CLASS_HANDLE simdHandle); // Get the handle for a SIMD type. CORINFO_CLASS_HANDLE gtGetStructHandleForSIMD(var_types simdType, var_types simdBaseType) { if (m_simdHandleCache == nullptr) { // This may happen if the JIT generates SIMD node on its own, without importing them. // Otherwise getBaseTypeAndSizeOfSIMDType should have created the cache. return NO_CLASS_HANDLE; } if (simdBaseType == TYP_FLOAT) { switch (simdType) { case TYP_SIMD8: return m_simdHandleCache->SIMDVector2Handle; case TYP_SIMD12: return m_simdHandleCache->SIMDVector3Handle; case TYP_SIMD16: if ((getSIMDVectorType() == TYP_SIMD32) || (m_simdHandleCache->SIMDVector4Handle != NO_CLASS_HANDLE)) { return m_simdHandleCache->SIMDVector4Handle; } break; case TYP_SIMD32: break; default: unreached(); } } assert(emitTypeSize(simdType) <= maxSIMDStructBytes()); switch (simdBaseType) { case TYP_FLOAT: return m_simdHandleCache->SIMDFloatHandle; case TYP_DOUBLE: return m_simdHandleCache->SIMDDoubleHandle; case TYP_INT: return m_simdHandleCache->SIMDIntHandle; case TYP_USHORT: return m_simdHandleCache->SIMDUShortHandle; case TYP_UBYTE: return m_simdHandleCache->SIMDUByteHandle; case TYP_SHORT: return m_simdHandleCache->SIMDShortHandle; case TYP_BYTE: return m_simdHandleCache->SIMDByteHandle; case TYP_LONG: return m_simdHandleCache->SIMDLongHandle; case TYP_UINT: return m_simdHandleCache->SIMDUIntHandle; case TYP_ULONG: return m_simdHandleCache->SIMDULongHandle; default: assert(!"Didn't find a class handle for simdType"); } return NO_CLASS_HANDLE; } // Returns true if this is a SIMD type that should be considered an opaque // vector type (i.e. do not analyze or promote its fields). // Note that all but the fixed vector types are opaque, even though they may // actually be declared as having fields. bool isOpaqueSIMDType(CORINFO_CLASS_HANDLE structHandle) { return ((m_simdHandleCache != nullptr) && (structHandle != m_simdHandleCache->SIMDVector2Handle) && (structHandle != m_simdHandleCache->SIMDVector3Handle) && (structHandle != m_simdHandleCache->SIMDVector4Handle)); } // Returns true if the tree corresponds to a TYP_SIMD lcl var. // Note that both SIMD vector args and locals are mared as lvSIMDType = true, but // type of an arg node is TYP_BYREF and a local node is TYP_SIMD or TYP_STRUCT. bool isSIMDTypeLocal(GenTree* tree) { return tree->OperIsLocal() && lvaTable[tree->AsLclVarCommon()->gtLclNum].lvSIMDType; } // Returns true if the lclVar is an opaque SIMD type. bool isOpaqueSIMDLclVar(LclVarDsc* varDsc) { if (!varDsc->lvSIMDType) { return false; } return isOpaqueSIMDType(varDsc->lvVerTypeInfo.GetClassHandle()); } // Returns true if the type of the tree is a byref of TYP_SIMD bool isAddrOfSIMDType(GenTree* tree) { if (tree->TypeGet() == TYP_BYREF || tree->TypeGet() == TYP_I_IMPL) { switch (tree->OperGet()) { case GT_ADDR: return varTypeIsSIMD(tree->gtGetOp1()); case GT_LCL_VAR_ADDR: return lvaTable[tree->AsLclVarCommon()->gtLclNum].lvSIMDType; default: return isSIMDTypeLocal(tree); } } return false; } static bool isRelOpSIMDIntrinsic(SIMDIntrinsicID intrinsicId) { return (intrinsicId == SIMDIntrinsicEqual || intrinsicId == SIMDIntrinsicLessThan || intrinsicId == SIMDIntrinsicLessThanOrEqual || intrinsicId == SIMDIntrinsicGreaterThan || intrinsicId == SIMDIntrinsicGreaterThanOrEqual); } // Returns base type of a TYP_SIMD local. // Returns TYP_UNKNOWN if the local is not TYP_SIMD. var_types getBaseTypeOfSIMDLocal(GenTree* tree) { if (isSIMDTypeLocal(tree)) { return lvaTable[tree->AsLclVarCommon()->gtLclNum].lvBaseType; } return TYP_UNKNOWN; } bool isSIMDClass(CORINFO_CLASS_HANDLE clsHnd) { return info.compCompHnd->isInSIMDModule(clsHnd); } bool isIntrinsicType(CORINFO_CLASS_HANDLE clsHnd) { return (info.compCompHnd->getClassAttribs(clsHnd) & CORINFO_FLG_INTRINSIC_TYPE) != 0; } const char* getClassNameFromMetadata(CORINFO_CLASS_HANDLE cls, const char** namespaceName) { return info.compCompHnd->getClassNameFromMetadata(cls, namespaceName); } CORINFO_CLASS_HANDLE getTypeInstantiationArgument(CORINFO_CLASS_HANDLE cls, unsigned index) { return info.compCompHnd->getTypeInstantiationArgument(cls, index); } bool isSIMDClass(typeInfo* pTypeInfo) { return pTypeInfo->IsStruct() && isSIMDClass(pTypeInfo->GetClassHandleForValueClass()); } bool isHWSIMDClass(CORINFO_CLASS_HANDLE clsHnd) { #ifdef FEATURE_HW_INTRINSICS if (isIntrinsicType(clsHnd)) { const char* namespaceName = nullptr; (void)getClassNameFromMetadata(clsHnd, &namespaceName); return strcmp(namespaceName, "System.Runtime.Intrinsics") == 0; } #endif // FEATURE_HW_INTRINSICS return false; } bool isHWSIMDClass(typeInfo* pTypeInfo) { #ifdef FEATURE_HW_INTRINSICS return pTypeInfo->IsStruct() && isHWSIMDClass(pTypeInfo->GetClassHandleForValueClass()); #else return false; #endif } bool isSIMDorHWSIMDClass(CORINFO_CLASS_HANDLE clsHnd) { return isSIMDClass(clsHnd) || isHWSIMDClass(clsHnd); } bool isSIMDorHWSIMDClass(typeInfo* pTypeInfo) { return isSIMDClass(pTypeInfo) || isHWSIMDClass(pTypeInfo); } // Get the base (element) type and size in bytes for a SIMD type. Returns TYP_UNKNOWN // if it is not a SIMD type or is an unsupported base type. var_types getBaseTypeAndSizeOfSIMDType(CORINFO_CLASS_HANDLE typeHnd, unsigned* sizeBytes = nullptr); var_types getBaseTypeOfSIMDType(CORINFO_CLASS_HANDLE typeHnd) { return getBaseTypeAndSizeOfSIMDType(typeHnd, nullptr); } // Get SIMD Intrinsic info given the method handle. // Also sets typeHnd, argCount, baseType and sizeBytes out params. const SIMDIntrinsicInfo* getSIMDIntrinsicInfo(CORINFO_CLASS_HANDLE* typeHnd, CORINFO_METHOD_HANDLE methodHnd, CORINFO_SIG_INFO* sig, bool isNewObj, unsigned* argCount, var_types* baseType, unsigned* sizeBytes); // Pops and returns GenTree node from importers type stack. // Normalizes TYP_STRUCT value in case of GT_CALL, GT_RET_EXPR and arg nodes. GenTree* impSIMDPopStack(var_types type, bool expectAddr = false, CORINFO_CLASS_HANDLE structType = nullptr); // Create a GT_SIMD tree for a Get property of SIMD vector with a fixed index. GenTreeSIMD* impSIMDGetFixed(var_types simdType, var_types baseType, unsigned simdSize, int index); // Creates a GT_SIMD tree for Select operation GenTree* impSIMDSelect(CORINFO_CLASS_HANDLE typeHnd, var_types baseType, unsigned simdVectorSize, GenTree* op1, GenTree* op2, GenTree* op3); // Creates a GT_SIMD tree for Min/Max operation GenTree* impSIMDMinMax(SIMDIntrinsicID intrinsicId, CORINFO_CLASS_HANDLE typeHnd, var_types baseType, unsigned simdVectorSize, GenTree* op1, GenTree* op2); // Transforms operands and returns the SIMD intrinsic to be applied on // transformed operands to obtain given relop result. SIMDIntrinsicID impSIMDRelOp(SIMDIntrinsicID relOpIntrinsicId, CORINFO_CLASS_HANDLE typeHnd, unsigned simdVectorSize, var_types* baseType, GenTree** op1, GenTree** op2); // Creates a GT_SIMD tree for Abs intrinsic. GenTree* impSIMDAbs(CORINFO_CLASS_HANDLE typeHnd, var_types baseType, unsigned simdVectorSize, GenTree* op1); #if defined(_TARGET_XARCH_) // Transforms operands and returns the SIMD intrinsic to be applied on // transformed operands to obtain == comparison result. SIMDIntrinsicID impSIMDLongRelOpEqual(CORINFO_CLASS_HANDLE typeHnd, unsigned simdVectorSize, GenTree** op1, GenTree** op2); // Transforms operands and returns the SIMD intrinsic to be applied on // transformed operands to obtain > comparison result. SIMDIntrinsicID impSIMDLongRelOpGreaterThan(CORINFO_CLASS_HANDLE typeHnd, unsigned simdVectorSize, GenTree** op1, GenTree** op2); // Transforms operands and returns the SIMD intrinsic to be applied on // transformed operands to obtain >= comparison result. SIMDIntrinsicID impSIMDLongRelOpGreaterThanOrEqual(CORINFO_CLASS_HANDLE typeHnd, unsigned simdVectorSize, GenTree** op1, GenTree** op2); // Transforms operands and returns the SIMD intrinsic to be applied on // transformed operands to obtain >= comparison result in case of int32 // and small int base type vectors. SIMDIntrinsicID impSIMDIntegralRelOpGreaterThanOrEqual( CORINFO_CLASS_HANDLE typeHnd, unsigned simdVectorSize, var_types baseType, GenTree** op1, GenTree** op2); #endif // defined(_TARGET_XARCH_) void setLclRelatedToSIMDIntrinsic(GenTree* tree); bool areFieldsContiguous(GenTree* op1, GenTree* op2); bool areArrayElementsContiguous(GenTree* op1, GenTree* op2); bool areArgumentsContiguous(GenTree* op1, GenTree* op2); GenTree* createAddressNodeForSIMDInit(GenTree* tree, unsigned simdSize); // check methodHnd to see if it is a SIMD method that is expanded as an intrinsic in the JIT. GenTree* impSIMDIntrinsic(OPCODE opcode, GenTree* newobjThis, CORINFO_CLASS_HANDLE clsHnd, CORINFO_METHOD_HANDLE method, CORINFO_SIG_INFO* sig, unsigned methodFlags, int memberRef); GenTree* getOp1ForConstructor(OPCODE opcode, GenTree* newobjThis, CORINFO_CLASS_HANDLE clsHnd); // Whether SIMD vector occupies part of SIMD register. // SSE2: vector2f/3f are considered sub register SIMD types. // AVX: vector2f, 3f and 4f are all considered sub register SIMD types. bool isSubRegisterSIMDType(CORINFO_CLASS_HANDLE typeHnd) { unsigned sizeBytes = 0; var_types baseType = getBaseTypeAndSizeOfSIMDType(typeHnd, &sizeBytes); return (baseType == TYP_FLOAT) && (sizeBytes < getSIMDVectorRegisterByteLength()); } bool isSubRegisterSIMDType(GenTreeSIMD* simdNode) { return (simdNode->gtSIMDSize < getSIMDVectorRegisterByteLength()); } // Get the type for the hardware SIMD vector. // This is the maximum SIMD type supported for this target. var_types getSIMDVectorType() { #if defined(_TARGET_XARCH_) if (getSIMDSupportLevel() == SIMD_AVX2_Supported) { return TYP_SIMD32; } else { assert(getSIMDSupportLevel() >= SIMD_SSE2_Supported); return TYP_SIMD16; } #elif defined(_TARGET_ARM64_) return TYP_SIMD16; #else assert(!"getSIMDVectorType() unimplemented on target arch"); unreached(); #endif } // Get the size of the SIMD type in bytes int getSIMDTypeSizeInBytes(CORINFO_CLASS_HANDLE typeHnd) { unsigned sizeBytes = 0; (void)getBaseTypeAndSizeOfSIMDType(typeHnd, &sizeBytes); return sizeBytes; } // Get the the number of elements of basetype of SIMD vector given by its size and baseType static int getSIMDVectorLength(unsigned simdSize, var_types baseType); // Get the the number of elements of basetype of SIMD vector given by its type handle int getSIMDVectorLength(CORINFO_CLASS_HANDLE typeHnd); // Get preferred alignment of SIMD type. int getSIMDTypeAlignment(var_types simdType); // Get the number of bytes in a System.Numeric.Vector for the current compilation. // Note - cannot be used for System.Runtime.Intrinsic unsigned getSIMDVectorRegisterByteLength() { #if defined(_TARGET_XARCH_) if (getSIMDSupportLevel() == SIMD_AVX2_Supported) { return YMM_REGSIZE_BYTES; } else { assert(getSIMDSupportLevel() >= SIMD_SSE2_Supported); return XMM_REGSIZE_BYTES; } #elif defined(_TARGET_ARM64_) return FP_REGSIZE_BYTES; #else assert(!"getSIMDVectorRegisterByteLength() unimplemented on target arch"); unreached(); #endif } // The minimum and maximum possible number of bytes in a SIMD vector. // maxSIMDStructBytes // The minimum SIMD size supported by System.Numeric.Vectors or System.Runtime.Intrinsic // SSE: 16-byte Vector and Vector128 // AVX: 32-byte Vector256 (Vector is 16-byte) // AVX2: 32-byte Vector and Vector256 unsigned int maxSIMDStructBytes() { #if defined(FEATURE_HW_INTRINSICS) && defined(_TARGET_XARCH_) if (compSupports(InstructionSet_AVX)) { return YMM_REGSIZE_BYTES; } else { assert(getSIMDSupportLevel() >= SIMD_SSE2_Supported); return XMM_REGSIZE_BYTES; } #else return getSIMDVectorRegisterByteLength(); #endif } unsigned int minSIMDStructBytes() { return emitTypeSize(TYP_SIMD8); } // Returns the codegen type for a given SIMD size. var_types getSIMDTypeForSize(unsigned size) { var_types simdType = TYP_UNDEF; if (size == 8) { simdType = TYP_SIMD8; } else if (size == 12) { simdType = TYP_SIMD12; } else if (size == 16) { simdType = TYP_SIMD16; } else if (size == 32) { simdType = TYP_SIMD32; } else { noway_assert(!"Unexpected size for SIMD type"); } return simdType; } unsigned getSIMDInitTempVarNum() { if (lvaSIMDInitTempVarNum == BAD_VAR_NUM) { lvaSIMDInitTempVarNum = lvaGrabTempWithImplicitUse(false DEBUGARG("SIMDInitTempVar")); lvaTable[lvaSIMDInitTempVarNum].lvType = getSIMDVectorType(); } return lvaSIMDInitTempVarNum; } #else // !FEATURE_SIMD bool isOpaqueSIMDLclVar(LclVarDsc* varDsc) { return false; } #endif // FEATURE_SIMD public: //------------------------------------------------------------------------ // largestEnregisterableStruct: The size in bytes of the largest struct that can be enregistered. // // Notes: It is not guaranteed that the struct of this size or smaller WILL be a // candidate for enregistration. unsigned largestEnregisterableStructSize() { #ifdef FEATURE_SIMD unsigned vectorRegSize = getSIMDVectorRegisterByteLength(); if (vectorRegSize > TARGET_POINTER_SIZE) { return vectorRegSize; } else #endif // FEATURE_SIMD { return TARGET_POINTER_SIZE; } } private: // These routines need not be enclosed under FEATURE_SIMD since lvIsSIMDType() // is defined for both FEATURE_SIMD and !FEATURE_SIMD apropriately. The use // of this routines also avoids the need of #ifdef FEATURE_SIMD specific code. // Is this var is of type simd struct? bool lclVarIsSIMDType(unsigned varNum) { LclVarDsc* varDsc = lvaTable + varNum; return varDsc->lvIsSIMDType(); } // Is this Local node a SIMD local? bool lclVarIsSIMDType(GenTreeLclVarCommon* lclVarTree) { return lclVarIsSIMDType(lclVarTree->gtLclNum); } // Returns true if the TYP_SIMD locals on stack are aligned at their // preferred byte boundary specified by getSIMDTypeAlignment(). // // As per the Intel manual, the preferred alignment for AVX vectors is 32-bytes. On Amd64, // RSP/EBP is aligned at 16-bytes, therefore to align SIMD types at 32-bytes we need even // RSP/EBP to be 32-byte aligned. It is not clear whether additional stack space used in // aligning stack is worth the benefit and for now will use 16-byte alignment for AVX // 256-bit vectors with unaligned load/stores to/from memory. On x86, the stack frame // is aligned to 4 bytes. We need to extend existing support for double (8-byte) alignment // to 16 or 32 byte alignment for frames with local SIMD vars, if that is determined to be // profitable. // bool isSIMDTypeLocalAligned(unsigned varNum) { #if defined(FEATURE_SIMD) && ALIGN_SIMD_TYPES if (lclVarIsSIMDType(varNum) && lvaTable[varNum].lvType != TYP_BYREF) { bool ebpBased; int off = lvaFrameAddress(varNum, &ebpBased); // TODO-Cleanup: Can't this use the lvExactSize on the varDsc? int alignment = getSIMDTypeAlignment(lvaTable[varNum].lvType); bool isAligned = (alignment <= STACK_ALIGN) && ((off % alignment) == 0); return isAligned; } #endif // FEATURE_SIMD return false; } bool compSupports(InstructionSet isa) const { #if defined(_TARGET_XARCH_) || defined(_TARGET_ARM64_) return (opts.compSupportsISA & (1ULL << isa)) != 0; #else return false; #endif } bool canUseVexEncoding() const { #ifdef _TARGET_XARCH_ return compSupports(InstructionSet_AVX); #else return false; #endif } /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX Compiler XX XX XX XX Generic info about the compilation and the method being compiled. XX XX It is responsible for driving the other phases. XX XX It is also responsible for all the memory management. XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: Compiler* InlineeCompiler; // The Compiler instance for the inlinee InlineResult* compInlineResult; // The result of importing the inlinee method. bool compDoAggressiveInlining; // If true, mark every method as CORINFO_FLG_FORCEINLINE bool compJmpOpUsed; // Does the method do a JMP bool compLongUsed; // Does the method use TYP_LONG bool compFloatingPointUsed; // Does the method use TYP_FLOAT or TYP_DOUBLE bool compTailCallUsed; // Does the method do a tailcall bool compLocallocUsed; // Does the method use localloc. bool compLocallocOptimized; // Does the method have an optimized localloc bool compQmarkUsed; // Does the method use GT_QMARK/GT_COLON bool compQmarkRationalized; // Is it allowed to use a GT_QMARK/GT_COLON node. bool compUnsafeCastUsed; // Does the method use LDIND/STIND to cast between scalar/refernce types // NOTE: These values are only reliable after // the importing is completely finished. #ifdef DEBUG // State information - which phases have completed? // These are kept together for easy discoverability bool bRangeAllowStress; bool compCodeGenDone; int64_t compNumStatementLinksTraversed; // # of links traversed while doing debug checks bool fgNormalizeEHDone; // Has the flowgraph EH normalization phase been done? size_t compSizeEstimate; // The estimated size of the method as per `gtSetEvalOrder`. size_t compCycleEstimate; // The estimated cycle count of the method as per `gtSetEvalOrder` #endif // DEBUG bool fgLocalVarLivenessDone; // Note that this one is used outside of debug. bool fgLocalVarLivenessChanged; bool compLSRADone; bool compRationalIRForm; bool compUsesThrowHelper; // There is a call to a THOROW_HELPER for the compiled method. bool compGeneratingProlog; bool compGeneratingEpilog; bool compNeedsGSSecurityCookie; // There is an unsafe buffer (or localloc) on the stack. // Insert cookie on frame and code to check the cookie, like VC++ -GS. bool compGSReorderStackLayout; // There is an unsafe buffer on the stack, reorder locals and make local // copies of susceptible parameters to avoid buffer overrun attacks through locals/params bool getNeedsGSSecurityCookie() const { return compNeedsGSSecurityCookie; } void setNeedsGSSecurityCookie() { compNeedsGSSecurityCookie = true; } FrameLayoutState lvaDoneFrameLayout; // The highest frame layout state that we've completed. During // frame layout calculations, this is the level we are currently // computing. //---------------------------- JITing options ----------------------------- enum codeOptimize { BLENDED_CODE, SMALL_CODE, FAST_CODE, COUNT_OPT_CODE }; struct Options { JitFlags* jitFlags; // all flags passed from the EE unsigned compFlags; // method attributes codeOptimize compCodeOpt; // what type of code optimizations bool compUseFCOMI; bool compUseCMOV; #if defined(_TARGET_XARCH_) || defined(_TARGET_ARM64_) uint64_t compSupportsISA; void setSupportedISA(InstructionSet isa) { compSupportsISA |= 1ULL << isa; } #endif // optimize maximally and/or favor speed over size? #define DEFAULT_MIN_OPTS_CODE_SIZE 60000 #define DEFAULT_MIN_OPTS_INSTR_COUNT 20000 #define DEFAULT_MIN_OPTS_BB_COUNT 2000 #define DEFAULT_MIN_OPTS_LV_NUM_COUNT 2000 #define DEFAULT_MIN_OPTS_LV_REF_COUNT 8000 // Maximun number of locals before turning off the inlining #define MAX_LV_NUM_COUNT_FOR_INLINING 512 bool compMinOpts; unsigned instrCount; unsigned lvRefCount; bool compMinOptsIsSet; #ifdef DEBUG bool compMinOptsIsUsed; bool MinOpts() { assert(compMinOptsIsSet); compMinOptsIsUsed = true; return compMinOpts; } bool IsMinOptsSet() { return compMinOptsIsSet; } #else // !DEBUG bool MinOpts() { return compMinOpts; } bool IsMinOptsSet() { return compMinOptsIsSet; } #endif // !DEBUG bool OptimizationDisabled() { return MinOpts() || compDbgCode; } bool OptimizationEnabled() { return !OptimizationDisabled(); } void SetMinOpts(bool val) { assert(!compMinOptsIsUsed); assert(!compMinOptsIsSet || (compMinOpts == val)); compMinOpts = val; compMinOptsIsSet = true; } // true if the CLFLG_* for an optimization is set. bool OptEnabled(unsigned optFlag) { return !!(compFlags & optFlag); } #ifdef FEATURE_READYTORUN_COMPILER bool IsReadyToRun() { return jitFlags->IsSet(JitFlags::JIT_FLAG_READYTORUN); } #else bool IsReadyToRun() { return false; } #endif // true if we should use the PINVOKE_{BEGIN,END} helpers instead of generating // PInvoke transitions inline (e.g. when targeting CoreRT). bool ShouldUsePInvokeHelpers() { return jitFlags->IsSet(JitFlags::JIT_FLAG_USE_PINVOKE_HELPERS); } // true if we should use insert the REVERSE_PINVOKE_{ENTER,EXIT} helpers in the method // prolog/epilog bool IsReversePInvoke() { return jitFlags->IsSet(JitFlags::JIT_FLAG_REVERSE_PINVOKE); } // true if we must generate code compatible with JIT32 quirks bool IsJit32Compat() { #if defined(_TARGET_X86_) return jitFlags->IsSet(JitFlags::JIT_FLAG_DESKTOP_QUIRKS); #else return false; #endif } // true if we must generate code compatible with Jit64 quirks bool IsJit64Compat() { #if defined(_TARGET_AMD64_) return jitFlags->IsSet(JitFlags::JIT_FLAG_DESKTOP_QUIRKS); #elif !defined(FEATURE_CORECLR) return true; #else return false; #endif } bool compScopeInfo; // Generate the LocalVar info ? bool compDbgCode; // Generate debugger-friendly code? bool compDbgInfo; // Gather debugging info? bool compDbgEnC; #ifdef PROFILING_SUPPORTED bool compNoPInvokeInlineCB; #else static const bool compNoPInvokeInlineCB; #endif #ifdef DEBUG bool compGcChecks; // Check arguments and return values to ensure they are sane #endif #if defined(DEBUG) && defined(_TARGET_XARCH_) bool compStackCheckOnRet; // Check stack pointer on return to ensure it is correct. #endif // defined(DEBUG) && defined(_TARGET_XARCH_) #if defined(DEBUG) && defined(_TARGET_X86_) bool compStackCheckOnCall; // Check stack pointer after call to ensure it is correct. Only for x86. #endif // defined(DEBUG) && defined(_TARGET_X86_) bool compNeedSecurityCheck; // This flag really means where or not a security object needs // to be allocated on the stack. // It will be set to true in the following cases: // 1. When the method being compiled has a declarative security // (i.e. when CORINFO_FLG_NOSECURITYWRAP is reset for the current method). // This is also the case when we inject a prolog and epilog in the method. // (or) // 2. When the method being compiled has imperative security (i.e. the method // calls into another method that has CORINFO_FLG_SECURITYCHECK flag set). // (or) // 3. When opts.compDbgEnC is true. (See also Compiler::compCompile). // // When this flag is set, jit will allocate a gc-reference local variable (lvaSecurityObject), // which gets reported as a GC root to stackwalker. // (See also ICodeManager::GetAddrOfSecurityObject.) bool compReloc; // Generate relocs for pointers in code, true for all ngen/prejit codegen #ifdef DEBUG #if defined(_TARGET_XARCH_) bool compEnablePCRelAddr; // Whether absolute addr be encoded as PC-rel offset by RyuJIT where possible #endif #endif // DEBUG #ifdef UNIX_AMD64_ABI // This flag is indicating if there is a need to align the frame. // On AMD64-Windows, if there are calls, 4 slots for the outgoing ars are allocated, except for // FastTailCall. This slots makes the frame size non-zero, so alignment logic will be called. // On AMD64-Unix, there are no such slots. There is a possibility to have calls in the method with frame size of // 0. The frame alignment logic won't kick in. This flags takes care of the AMD64-Unix case by remembering that // there are calls and making sure the frame alignment logic is executed. bool compNeedToAlignFrame; #endif // UNIX_AMD64_ABI bool compProcedureSplitting; // Separate cold code from hot code bool genFPorder; // Preserve FP order (operations are non-commutative) bool genFPopt; // Can we do frame-pointer-omission optimization? bool altJit; // True if we are an altjit and are compiling this method #ifdef OPT_CONFIG bool optRepeat; // Repeat optimizer phases k times #endif #ifdef DEBUG bool compProcedureSplittingEH; // Separate cold code from hot code for functions with EH bool dspCode; // Display native code generated bool dspEHTable; // Display the EH table reported to the VM bool dspDebugInfo; // Display the Debug info reported to the VM bool dspInstrs; // Display the IL instructions intermixed with the native code output bool dspEmit; // Display emitter output bool dspLines; // Display source-code lines intermixed with native code output bool dmpHex; // Display raw bytes in hex of native code output bool varNames; // Display variables names in native code output bool disAsm; // Display native code as it is generated bool disAsmSpilled; // Display native code when any register spilling occurs bool disDiffable; // Makes the Disassembly code 'diff-able' bool disAsm2; // Display native code after it is generated using external disassembler bool dspOrder; // Display names of each of the methods that we ngen/jit bool dspUnwind; // Display the unwind info output bool dspDiffable; // Makes the Jit Dump 'diff-able' (currently uses same COMPlus_* flag as disDiffable) bool compLongAddress; // Force using large pseudo instructions for long address // (IF_LARGEJMP/IF_LARGEADR/IF_LARGLDC) bool dspGCtbls; // Display the GC tables #endif #ifdef LATE_DISASM bool doLateDisasm; // Run the late disassembler #endif // LATE_DISASM #if DUMP_GC_TABLES && !defined(DEBUG) && defined(JIT32_GCENCODER) // Only the JIT32_GCENCODER implements GC dumping in non-DEBUG code. #pragma message("NOTE: this non-debug build has GC ptr table dumping always enabled!") static const bool dspGCtbls = true; #endif #ifdef PROFILING_SUPPORTED // Whether to emit Enter/Leave/TailCall hooks using a dummy stub (DummyProfilerELTStub()). // This option helps make the JIT behave as if it is running under a profiler. bool compJitELTHookEnabled; #endif // PROFILING_SUPPORTED #if FEATURE_TAILCALL_OPT // Whether opportunistic or implicit tail call optimization is enabled. bool compTailCallOpt; // Whether optimization of transforming a recursive tail call into a loop is enabled. bool compTailCallLoopOpt; #endif #if defined(_TARGET_ARM64_) // Decision about whether to save FP/LR registers with callee-saved registers (see // COMPlus_JitSaveFpLrWithCalleSavedRegisters). int compJitSaveFpLrWithCalleeSavedRegisters; #endif // defined(_TARGET_ARM64_) #ifdef ARM_SOFTFP static const bool compUseSoftFP = true; #else // !ARM_SOFTFP static const bool compUseSoftFP = false; #endif GCPollType compGCPollType; } opts; #ifdef ALT_JIT static bool s_pAltJitExcludeAssembliesListInitialized; static AssemblyNamesList2* s_pAltJitExcludeAssembliesList; #endif // ALT_JIT #ifdef DEBUG static bool s_pJitDisasmIncludeAssembliesListInitialized; static AssemblyNamesList2* s_pJitDisasmIncludeAssembliesList; static bool s_pJitFunctionFileInitialized; static MethodSet* s_pJitMethodSet; #endif // DEBUG #ifdef DEBUG // silence warning of cast to greater size. It is easier to silence than construct code the compiler is happy with, and // it is safe in this case #pragma warning(push) #pragma warning(disable : 4312) template T dspPtr(T p) { return (p == ZERO) ? ZERO : (opts.dspDiffable ? T(0xD1FFAB1E) : p); } template T dspOffset(T o) { return (o == ZERO) ? ZERO : (opts.dspDiffable ? T(0xD1FFAB1E) : o); } #pragma warning(pop) static int dspTreeID(GenTree* tree) { return tree->gtTreeID; } static void printTreeID(GenTree* tree) { if (tree == nullptr) { printf("[------]"); } else { printf("[%06d]", dspTreeID(tree)); } } #endif // DEBUG // clang-format off #define STRESS_MODES \ \ STRESS_MODE(NONE) \ \ /* "Variations" stress areas which we try to mix up with each other. */ \ /* These should not be exhaustively used as they might */ \ /* hide/trivialize other areas */ \ \ STRESS_MODE(REGS) \ STRESS_MODE(DBL_ALN) \ STRESS_MODE(LCL_FLDS) \ STRESS_MODE(UNROLL_LOOPS) \ STRESS_MODE(MAKE_CSE) \ STRESS_MODE(LEGACY_INLINE) \ STRESS_MODE(CLONE_EXPR) \ STRESS_MODE(USE_FCOMI) \ STRESS_MODE(USE_CMOV) \ STRESS_MODE(FOLD) \ STRESS_MODE(BB_PROFILE) \ STRESS_MODE(OPT_BOOLS_GC) \ STRESS_MODE(REMORPH_TREES) \ STRESS_MODE(64RSLT_MUL) \ STRESS_MODE(DO_WHILE_LOOPS) \ STRESS_MODE(MIN_OPTS) \ STRESS_MODE(REVERSE_FLAG) /* Will set GTF_REVERSE_OPS whenever we can */ \ STRESS_MODE(REVERSE_COMMA) /* Will reverse commas created with gtNewCommaNode */ \ STRESS_MODE(TAILCALL) /* Will make the call as a tailcall whenever legal */ \ STRESS_MODE(CATCH_ARG) /* Will spill catch arg */ \ STRESS_MODE(UNSAFE_BUFFER_CHECKS) \ STRESS_MODE(NULL_OBJECT_CHECK) \ STRESS_MODE(PINVOKE_RESTORE_ESP) \ STRESS_MODE(RANDOM_INLINE) \ STRESS_MODE(SWITCH_CMP_BR_EXPANSION) \ STRESS_MODE(GENERIC_VARN) \ \ /* After COUNT_VARN, stress level 2 does all of these all the time */ \ \ STRESS_MODE(COUNT_VARN) \ \ /* "Check" stress areas that can be exhaustively used if we */ \ /* dont care about performance at all */ \ \ STRESS_MODE(FORCE_INLINE) /* Treat every method as AggressiveInlining */ \ STRESS_MODE(CHK_FLOW_UPDATE) \ STRESS_MODE(EMITTER) \ STRESS_MODE(CHK_REIMPORT) \ STRESS_MODE(FLATFP) \ STRESS_MODE(GENERIC_CHECK) \ STRESS_MODE(COUNT) enum compStressArea { #define STRESS_MODE(mode) STRESS_##mode, STRESS_MODES #undef STRESS_MODE }; // clang-format on #ifdef DEBUG static const LPCWSTR s_compStressModeNames[STRESS_COUNT + 1]; BYTE compActiveStressModes[STRESS_COUNT]; #endif // DEBUG #define MAX_STRESS_WEIGHT 100 bool compStressCompile(compStressArea stressArea, unsigned weightPercentage); #ifdef DEBUG bool compInlineStress() { return compStressCompile(STRESS_LEGACY_INLINE, 50); } bool compRandomInlineStress() { return compStressCompile(STRESS_RANDOM_INLINE, 50); } #endif // DEBUG bool compTailCallStress() { #ifdef DEBUG return (JitConfig.TailcallStress() != 0 || compStressCompile(STRESS_TAILCALL, 5)); #else return false; #endif } codeOptimize compCodeOpt() { #if 0 // Switching between size & speed has measurable throughput impact // (3.5% on NGen mscorlib when measured). It used to be enabled for // DEBUG, but should generate identical code between CHK & RET builds, // so that's not acceptable. // TODO-Throughput: Figure out what to do about size vs. speed & throughput. // Investigate the cause of the throughput regression. return opts.compCodeOpt; #else return BLENDED_CODE; #endif } //--------------------- Info about the procedure -------------------------- struct Info { COMP_HANDLE compCompHnd; CORINFO_MODULE_HANDLE compScopeHnd; CORINFO_CLASS_HANDLE compClassHnd; CORINFO_METHOD_HANDLE compMethodHnd; CORINFO_METHOD_INFO* compMethodInfo; BOOL hasCircularClassConstraints; BOOL hasCircularMethodConstraints; #if defined(DEBUG) || defined(LATE_DISASM) const char* compMethodName; const char* compClassName; const char* compFullName; #endif // defined(DEBUG) || defined(LATE_DISASM) #if defined(DEBUG) || defined(INLINE_DATA) // Method hash is logcally const, but computed // on first demand. mutable unsigned compMethodHashPrivate; unsigned compMethodHash() const; #endif // defined(DEBUG) || defined(INLINE_DATA) #ifdef PSEUDORANDOM_NOP_INSERTION // things for pseudorandom nop insertion unsigned compChecksum; CLRRandom compRNG; #endif // The following holds the FLG_xxxx flags for the method we're compiling. unsigned compFlags; // The following holds the class attributes for the method we're compiling. unsigned compClassAttr; const BYTE* compCode; IL_OFFSET compILCodeSize; // The IL code size IL_OFFSET compILImportSize; // Estimated amount of IL actually imported UNATIVE_OFFSET compNativeCodeSize; // The native code size, after instructions are issued. This // is less than (compTotalHotCodeSize + compTotalColdCodeSize) only if: // (1) the code is not hot/cold split, and we issued less code than we expected, or // (2) the code is hot/cold split, and we issued less code than we expected // in the cold section (the hot section will always be padded out to compTotalHotCodeSize). bool compIsStatic : 1; // Is the method static (no 'this' pointer)? bool compIsVarArgs : 1; // Does the method have varargs parameters? bool compIsContextful : 1; // contextful method bool compInitMem : 1; // Is the CORINFO_OPT_INIT_LOCALS bit set in the method info options? bool compUnwrapContextful : 1; // JIT should unwrap proxies when possible bool compProfilerCallback : 1; // JIT inserted a profiler Enter callback bool compPublishStubParam : 1; // EAX captured in prolog will be available through an instrinsic bool compRetBuffDefStack : 1; // The ret buff argument definitely points into the stack. var_types compRetType; // Return type of the method as declared in IL var_types compRetNativeType; // Normalized return type as per target arch ABI unsigned compILargsCount; // Number of arguments (incl. implicit but not hidden) unsigned compArgsCount; // Number of arguments (incl. implicit and hidden) #if FEATURE_FASTTAILCALL size_t compArgStackSize; // Incoming argument stack size in bytes bool compHasMultiSlotArgs; // Caller has >8 byte sized struct parameter #endif // FEATURE_FASTTAILCALL unsigned compRetBuffArg; // position of hidden return param var (0, 1) (BAD_VAR_NUM means not present); int compTypeCtxtArg; // position of hidden param for type context for generic code (CORINFO_CALLCONV_PARAMTYPE) unsigned compThisArg; // position of implicit this pointer param (not to be confused with lvaArg0Var) unsigned compILlocalsCount; // Number of vars : args + locals (incl. implicit but not hidden) unsigned compLocalsCount; // Number of vars : args + locals (incl. implicit and hidden) unsigned compMaxStack; UNATIVE_OFFSET compTotalHotCodeSize; // Total number of bytes of Hot Code in the method UNATIVE_OFFSET compTotalColdCodeSize; // Total number of bytes of Cold Code in the method unsigned compCallUnmanaged; // count of unmanaged calls unsigned compLvFrameListRoot; // lclNum for the Frame root unsigned compXcptnsCount; // Number of exception-handling clauses read in the method's IL. // You should generally use compHndBBtabCount instead: it is the // current number of EH clauses (after additions like synchronized // methods and funclets, and removals like unreachable code deletion). bool compMatchedVM; // true if the VM is "matched": either the JIT is a cross-compiler // and the VM expects that, or the JIT is a "self-host" compiler // (e.g., x86 hosted targeting x86) and the VM expects that. /* The following holds IL scope information about local variables. */ unsigned compVarScopesCount; VarScopeDsc* compVarScopes; /* The following holds information about instr offsets for * which we need to report IP-mappings */ IL_OFFSET* compStmtOffsets; // sorted unsigned compStmtOffsetsCount; ICorDebugInfo::BoundaryTypes compStmtOffsetsImplicit; #define CPU_X86 0x0100 // The generic X86 CPU #define CPU_X86_PENTIUM_4 0x0110 #define CPU_X64 0x0200 // The generic x64 CPU #define CPU_AMD_X64 0x0210 // AMD x64 CPU #define CPU_INTEL_X64 0x0240 // Intel x64 CPU #define CPU_ARM 0x0300 // The generic ARM CPU #define CPU_ARM64 0x0400 // The generic ARM64 CPU unsigned genCPU; // What CPU are we running on } info; // Returns true if the method being compiled returns a non-void and non-struct value. // Note that lvaInitTypeRef() normalizes compRetNativeType for struct returns in a // single register as per target arch ABI (e.g on Amd64 Windows structs of size 1, 2, // 4 or 8 gets normalized to TYP_BYTE/TYP_SHORT/TYP_INT/TYP_LONG; On Arm HFA structs). // Methods returning such structs are considered to return non-struct return value and // this method returns true in that case. bool compMethodReturnsNativeScalarType() { return (info.compRetType != TYP_VOID) && !varTypeIsStruct(info.compRetNativeType); } // Returns true if the method being compiled returns RetBuf addr as its return value bool compMethodReturnsRetBufAddr() { // There are cases where implicit RetBuf argument should be explicitly returned in a register. // In such cases the return type is changed to TYP_BYREF and appropriate IR is generated. // These cases are: // 1. Profiler Leave calllback expects the address of retbuf as return value for // methods with hidden RetBuf argument. impReturnInstruction() when profiler // callbacks are needed creates GT_RETURN(TYP_BYREF, op1 = Addr of RetBuf) for // methods with hidden RetBufArg. // // 2. As per the System V ABI, the address of RetBuf needs to be returned by // methods with hidden RetBufArg in RAX. In such case GT_RETURN is of TYP_BYREF, // returning the address of RetBuf. // // 3. Windows 64-bit native calling convention also requires the address of RetBuff // to be returned in RAX. CLANG_FORMAT_COMMENT_ANCHOR; #ifdef _TARGET_AMD64_ return (info.compRetBuffArg != BAD_VAR_NUM); #else // !_TARGET_AMD64_ return (compIsProfilerHookNeeded()) && (info.compRetBuffArg != BAD_VAR_NUM); #endif // !_TARGET_AMD64_ } // Returns true if the method returns a value in more than one return register // TODO-ARM-Bug: Deal with multi-register genReturnLocaled structs? // TODO-ARM64: Does this apply for ARM64 too? bool compMethodReturnsMultiRegRetType() { #if FEATURE_MULTIREG_RET #if defined(_TARGET_X86_) // On x86 only 64-bit longs are returned in multiple registers return varTypeIsLong(info.compRetNativeType); #else // targets: X64-UNIX, ARM64 or ARM32 // On all other targets that support multireg return values: // Methods returning a struct in multiple registers have a return value of TYP_STRUCT. // Such method's compRetNativeType is TYP_STRUCT without a hidden RetBufArg return varTypeIsStruct(info.compRetNativeType) && (info.compRetBuffArg == BAD_VAR_NUM); #endif // TARGET_XXX #else // not FEATURE_MULTIREG_RET // For this architecture there are no multireg returns return false; #endif // FEATURE_MULTIREG_RET } #if FEATURE_MULTIREG_ARGS // Given a GenTree node of TYP_STRUCT that represents a pass by value argument // return the gcPtr layout for the pointers sized fields void getStructGcPtrsFromOp(GenTree* op, BYTE* gcPtrsOut); #endif // FEATURE_MULTIREG_ARGS // Returns true if the method being compiled returns a value bool compMethodHasRetVal() { return compMethodReturnsNativeScalarType() || compMethodReturnsRetBufAddr() || compMethodReturnsMultiRegRetType(); } #if defined(DEBUG) void compDispLocalVars(); #endif // DEBUG //-------------------------- Global Compiler Data ------------------------------------ #ifdef DEBUG static unsigned s_compMethodsCount; // to produce unique label names unsigned compGenTreeID; unsigned compBasicBlockID; #endif BasicBlock* compCurBB; // the current basic block in process GenTreeStmt* compCurStmt; // the current statement in process #ifdef DEBUG unsigned compCurStmtNum; // to give all statements an increasing StmtNum when printing dumps #endif // The following is used to create the 'method JIT info' block. size_t compInfoBlkSize; BYTE* compInfoBlkAddr; EHblkDsc* compHndBBtab; // array of EH data unsigned compHndBBtabCount; // element count of used elements in EH data array unsigned compHndBBtabAllocCount; // element count of allocated elements in EH data array #if defined(_TARGET_X86_) //------------------------------------------------------------------------- // Tracking of region covered by the monitor in synchronized methods void* syncStartEmitCookie; // the emitter cookie for first instruction after the call to MON_ENTER void* syncEndEmitCookie; // the emitter cookie for first instruction after the call to MON_EXIT #endif // !_TARGET_X86_ Phases previousCompletedPhase; // the most recently completed phase //------------------------------------------------------------------------- // The following keeps track of how many bytes of local frame space we've // grabbed so far in the current function, and how many argument bytes we // need to pop when we return. // unsigned compLclFrameSize; // secObject+lclBlk+locals+temps // Count of callee-saved regs we pushed in the prolog. // Does not include EBP for isFramePointerUsed() and double-aligned frames. // In case of Amd64 this doesn't include float regs saved on stack. unsigned compCalleeRegsPushed; #if defined(_TARGET_XARCH_) // Mask of callee saved float regs on stack. regMaskTP compCalleeFPRegsSavedMask; #endif #ifdef _TARGET_AMD64_ // Quirk for VS debug-launch scenario to work: // Bytes of padding between save-reg area and locals. #define VSQUIRK_STACK_PAD (2 * REGSIZE_BYTES) unsigned compVSQuirkStackPaddingNeeded; bool compQuirkForPPPflag; #endif unsigned compArgSize; // total size of arguments in bytes (including register args (lvIsRegArg)) unsigned compMapILargNum(unsigned ILargNum); // map accounting for hidden args unsigned compMapILvarNum(unsigned ILvarNum); // map accounting for hidden args unsigned compMap2ILvarNum(unsigned varNum) const; // map accounting for hidden args //------------------------------------------------------------------------- static void compStartup(); // One-time initialization static void compShutdown(); // One-time finalization void compInit(ArenaAllocator* pAlloc, InlineInfo* inlineInfo); void compDone(); static void compDisplayStaticSizes(FILE* fout); //------------ Some utility functions -------------- void* compGetHelperFtn(CorInfoHelpFunc ftnNum, /* IN */ void** ppIndirection); /* OUT */ // Several JIT/EE interface functions return a CorInfoType, and also return a // class handle as an out parameter if the type is a value class. Returns the // size of the type these describe. unsigned compGetTypeSize(CorInfoType cit, CORINFO_CLASS_HANDLE clsHnd); #ifdef DEBUG // Components used by the compiler may write unit test suites, and // have them run within this method. They will be run only once per process, and only // in debug. (Perhaps should be under the control of a COMPlus_ flag.) // These should fail by asserting. void compDoComponentUnitTestsOnce(); #endif // DEBUG int compCompile(CORINFO_METHOD_HANDLE methodHnd, CORINFO_MODULE_HANDLE classPtr, COMP_HANDLE compHnd, CORINFO_METHOD_INFO* methodInfo, void** methodCodePtr, ULONG* methodCodeSize, JitFlags* compileFlags); void compCompileFinish(); int compCompileHelper(CORINFO_MODULE_HANDLE classPtr, COMP_HANDLE compHnd, CORINFO_METHOD_INFO* methodInfo, void** methodCodePtr, ULONG* methodCodeSize, JitFlags* compileFlags, CorInfoInstantiationVerification instVerInfo); ArenaAllocator* compGetArenaAllocator(); #if MEASURE_MEM_ALLOC static bool s_dspMemStats; // Display per-phase memory statistics for every function #endif // MEASURE_MEM_ALLOC #if LOOP_HOIST_STATS unsigned m_loopsConsidered; bool m_curLoopHasHoistedExpression; unsigned m_loopsWithHoistedExpressions; unsigned m_totalHoistedExpressions; void AddLoopHoistStats(); void PrintPerMethodLoopHoistStats(); static CritSecObject s_loopHoistStatsLock; // This lock protects the data structures below. static unsigned s_loopsConsidered; static unsigned s_loopsWithHoistedExpressions; static unsigned s_totalHoistedExpressions; static void PrintAggregateLoopHoistStats(FILE* f); #endif // LOOP_HOIST_STATS bool compIsForImportOnly(); bool compIsForInlining() const; bool compDonotInline(); #ifdef DEBUG // Get the default fill char value we randomize this value when JitStress is enabled. static unsigned char compGetJitDefaultFill(Compiler* comp); const char* compLocalVarName(unsigned varNum, unsigned offs); VarName compVarName(regNumber reg, bool isFloatReg = false); const char* compRegVarName(regNumber reg, bool displayVar = false, bool isFloatReg = false); const char* compRegNameForSize(regNumber reg, size_t size); const char* compFPregVarName(unsigned fpReg, bool displayVar = false); void compDspSrcLinesByNativeIP(UNATIVE_OFFSET curIP); void compDspSrcLinesByLineNum(unsigned line, bool seek = false); #endif // DEBUG //------------------------------------------------------------------------- struct VarScopeListNode { VarScopeDsc* data; VarScopeListNode* next; static VarScopeListNode* Create(VarScopeDsc* value, CompAllocator alloc) { VarScopeListNode* node = new (alloc) VarScopeListNode; node->data = value; node->next = nullptr; return node; } }; struct VarScopeMapInfo { VarScopeListNode* head; VarScopeListNode* tail; static VarScopeMapInfo* Create(VarScopeListNode* node, CompAllocator alloc) { VarScopeMapInfo* info = new (alloc) VarScopeMapInfo; info->head = node; info->tail = node; return info; } }; // Max value of scope count for which we would use linear search; for larger values we would use hashtable lookup. static const unsigned MAX_LINEAR_FIND_LCL_SCOPELIST = 32; typedef JitHashTable, VarScopeMapInfo*> VarNumToScopeDscMap; // Map to keep variables' scope indexed by varNum containing it's scope dscs at the index. VarNumToScopeDscMap* compVarScopeMap; VarScopeDsc* compFindLocalVar(unsigned varNum, unsigned lifeBeg, unsigned lifeEnd); VarScopeDsc* compFindLocalVar(unsigned varNum, unsigned offs); VarScopeDsc* compFindLocalVarLinear(unsigned varNum, unsigned offs); void compInitVarScopeMap(); VarScopeDsc** compEnterScopeList; // List has the offsets where variables // enter scope, sorted by instr offset unsigned compNextEnterScope; VarScopeDsc** compExitScopeList; // List has the offsets where variables // go out of scope, sorted by instr offset unsigned compNextExitScope; void compInitScopeLists(); void compResetScopeLists(); VarScopeDsc* compGetNextEnterScope(unsigned offs, bool scan = false); VarScopeDsc* compGetNextExitScope(unsigned offs, bool scan = false); void compProcessScopesUntil(unsigned offset, VARSET_TP* inScope, void (Compiler::*enterScopeFn)(VARSET_TP* inScope, VarScopeDsc*), void (Compiler::*exitScopeFn)(VARSET_TP* inScope, VarScopeDsc*)); #ifdef DEBUG void compDispScopeLists(); #endif // DEBUG bool compIsProfilerHookNeeded(); //------------------------------------------------------------------------- /* Statistical Data Gathering */ void compJitStats(); // call this function and enable // various ifdef's below for statistical data #if CALL_ARG_STATS void compCallArgStats(); static void compDispCallArgStats(FILE* fout); #endif //------------------------------------------------------------------------- protected: #ifdef DEBUG bool skipMethod(); #endif ArenaAllocator* compArenaAllocator; public: void compFunctionTraceStart(); void compFunctionTraceEnd(void* methodCodePtr, ULONG methodCodeSize, bool isNYI); protected: size_t compMaxUncheckedOffsetForNullObject; void compInitOptions(JitFlags* compileFlags); void compSetProcessor(); void compInitDebuggingInfo(); void compSetOptimizationLevel(); #ifdef _TARGET_ARMARCH_ bool compRsvdRegCheck(FrameLayoutState curState); #endif void compCompile(void** methodCodePtr, ULONG* methodCodeSize, JitFlags* compileFlags); // Clear annotations produced during optimizations; to be used between iterations when repeating opts. void ResetOptAnnotations(); // Regenerate loop descriptors; to be used between iterations when repeating opts. void RecomputeLoopInfo(); #ifdef PROFILING_SUPPORTED // Data required for generating profiler Enter/Leave/TailCall hooks bool compProfilerHookNeeded; // Whether profiler Enter/Leave/TailCall hook needs to be generated for the method void* compProfilerMethHnd; // Profiler handle of the method being compiled. Passed as param to ELT callbacks bool compProfilerMethHndIndirected; // Whether compProfilerHandle is pointer to the handle or is an actual handle #endif #ifdef _TARGET_AMD64_ bool compQuirkForPPP(); // Check if this method should be Quirked for the PPP issue #endif public: // Assumes called as part of process shutdown; does any compiler-specific work associated with that. static void ProcessShutdownWork(ICorStaticInfo* statInfo); CompAllocator getAllocator(CompMemKind cmk = CMK_Generic) { return CompAllocator(compArenaAllocator, cmk); } CompAllocator getAllocatorGC() { return getAllocator(CMK_GC); } CompAllocator getAllocatorLoopHoist() { return getAllocator(CMK_LoopHoist); } #ifdef DEBUG CompAllocator getAllocatorDebugOnly() { return getAllocator(CMK_DebugOnly); } #endif // DEBUG /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX typeInfo XX XX XX XX Checks for type compatibility and merges types XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: // Set to TRUE if verification cannot be skipped for this method // If we detect unverifiable code, we will lazily check // canSkipMethodVerification() to see if verification is REALLY needed. BOOL tiVerificationNeeded; // It it initially TRUE, and it gets set to FALSE if we run into unverifiable code // Note that this is valid only if tiVerificationNeeded was ever TRUE. BOOL tiIsVerifiableCode; // Set to TRUE if runtime callout is needed for this method BOOL tiRuntimeCalloutNeeded; // Set to TRUE if security prolog/epilog callout is needed for this method // Note: This flag is different than compNeedSecurityCheck. // compNeedSecurityCheck means whether or not a security object needs // to be allocated on the stack, which is currently true for EnC as well. // tiSecurityCalloutNeeded means whether or not security callouts need // to be inserted in the jitted code. BOOL tiSecurityCalloutNeeded; // Returns TRUE if child is equal to or a subtype of parent for merge purposes // This support is necessary to suport attributes that are not described in // for example, signatures. For example, the permanent home byref (byref that // points to the gc heap), isn't a property of method signatures, therefore, // it is safe to have mismatches here (that tiCompatibleWith will not flag), // but when deciding if we need to reimport a block, we need to take these // in account BOOL tiMergeCompatibleWith(const typeInfo& pChild, const typeInfo& pParent, bool normalisedForStack) const; // Returns TRUE if child is equal to or a subtype of parent. // normalisedForStack indicates that both types are normalised for the stack BOOL tiCompatibleWith(const typeInfo& pChild, const typeInfo& pParent, bool normalisedForStack) const; // Merges pDest and pSrc. Returns FALSE if merge is undefined. // *pDest is modified to represent the merged type. Sets "*changed" to true // if this changes "*pDest". BOOL tiMergeToCommonParent(typeInfo* pDest, const typeInfo* pSrc, bool* changed) const; #ifdef DEBUG // VSW 471305 // IJW allows assigning REF to BYREF. The following allows us to temporarily // bypass the assert check in gcMarkRegSetGCref and gcMarkRegSetByref // We use a "short" as we need to push/pop this scope. // short compRegSetCheckLevel; #endif /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX IL verification stuff XX XX XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: // The following is used to track liveness of local variables, initialization // of valueclass constructors, and type safe use of IL instructions. // dynamic state info needed for verification EntryState verCurrentState; // this ptr of object type .ctors are considered intited only after // the base class ctor is called, or an alternate ctor is called. // An uninited this ptr can be used to access fields, but cannot // be used to call a member function. BOOL verTrackObjCtorInitState; void verInitBBEntryState(BasicBlock* block, EntryState* currentState); // Requires that "tis" is not TIS_Bottom -- it's a definite init/uninit state. void verSetThisInit(BasicBlock* block, ThisInitState tis); void verInitCurrentState(); void verResetCurrentState(BasicBlock* block, EntryState* currentState); // Merges the current verification state into the entry state of "block", return FALSE if that merge fails, // TRUE if it succeeds. Further sets "*changed" to true if this changes the entry state of "block". BOOL verMergeEntryStates(BasicBlock* block, bool* changed); void verConvertBBToThrowVerificationException(BasicBlock* block DEBUGARG(bool logMsg)); void verHandleVerificationFailure(BasicBlock* block DEBUGARG(bool logMsg)); typeInfo verMakeTypeInfo(CORINFO_CLASS_HANDLE clsHnd, bool bashStructToRef = false); // converts from jit type representation to typeInfo typeInfo verMakeTypeInfo(CorInfoType ciType, CORINFO_CLASS_HANDLE clsHnd); // converts from jit type representation to typeInfo BOOL verIsSDArray(typeInfo ti); typeInfo verGetArrayElemType(typeInfo ti); typeInfo verParseArgSigToTypeInfo(CORINFO_SIG_INFO* sig, CORINFO_ARG_LIST_HANDLE args); BOOL verNeedsVerification(); BOOL verIsByRefLike(const typeInfo& ti); BOOL verIsSafeToReturnByRef(const typeInfo& ti); // generic type variables range over types that satisfy IsBoxable BOOL verIsBoxable(const typeInfo& ti); void DECLSPEC_NORETURN verRaiseVerifyException(INDEBUG(const char* reason) DEBUGARG(const char* file) DEBUGARG(unsigned line)); void verRaiseVerifyExceptionIfNeeded(INDEBUG(const char* reason) DEBUGARG(const char* file) DEBUGARG(unsigned line)); bool verCheckTailCallConstraint(OPCODE opcode, CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_RESOLVED_TOKEN* pConstrainedResolvedToken, // Is this a "constrained." call // on a type parameter? bool speculative // If true, won't throw if verificatoin fails. Instead it will // return false to the caller. // If false, it will throw. ); bool verIsBoxedValueType(typeInfo ti); void verVerifyCall(OPCODE opcode, CORINFO_RESOLVED_TOKEN* pResolvedToken, CORINFO_RESOLVED_TOKEN* pConstrainedResolvedToken, bool tailCall, bool readonlyCall, // is this a "readonly." call? const BYTE* delegateCreateStart, const BYTE* codeAddr, CORINFO_CALL_INFO* callInfo DEBUGARG(const char* methodName)); BOOL verCheckDelegateCreation(const BYTE* delegateCreateStart, const BYTE* codeAddr, mdMemberRef& targetMemberRef); typeInfo verVerifySTIND(const typeInfo& ptr, const typeInfo& value, const typeInfo& instrType); typeInfo verVerifyLDIND(const typeInfo& ptr, const typeInfo& instrType); void verVerifyField(CORINFO_RESOLVED_TOKEN* pResolvedToken, const CORINFO_FIELD_INFO& fieldInfo, const typeInfo* tiThis, BOOL mutator, BOOL allowPlainStructAsThis = FALSE); void verVerifyCond(const typeInfo& tiOp1, const typeInfo& tiOp2, unsigned opcode); void verVerifyThisPtrInitialised(); BOOL verIsCallToInitThisPtr(CORINFO_CLASS_HANDLE context, CORINFO_CLASS_HANDLE target); #ifdef DEBUG // One line log function. Default level is 0. Increasing it gives you // more log information // levels are currently unused: #define JITDUMP(level,...) (); void JitLogEE(unsigned level, const char* fmt, ...); bool compDebugBreak; bool compJitHaltMethod(); #endif /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX GS Security checks for unsafe buffers XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ public: struct ShadowParamVarInfo { FixedBitVect* assignGroup; // the closure set of variables whose values depend on each other unsigned shadowCopy; // Lcl var num, valid only if not set to NO_SHADOW_COPY static bool mayNeedShadowCopy(LclVarDsc* varDsc) { #if defined(_TARGET_AMD64_) // GS cookie logic to create shadow slots, create trees to copy reg args to shadow // slots and update all trees to refer to shadow slots is done immediately after // fgMorph(). Lsra could potentially mark a param as DoNotEnregister after JIT determines // not to shadow a parameter. Also, LSRA could potentially spill a param which is passed // in register. Therefore, conservatively all params may need a shadow copy. Note that // GS cookie logic further checks whether the param is a ptr or an unsafe buffer before // creating a shadow slot even though this routine returns true. // // TODO-AMD64-CQ: Revisit this conservative approach as it could create more shadow slots than // required. There are two cases under which a reg arg could potentially be used from its // home location: // a) LSRA marks it as DoNotEnregister (see LinearScan::identifyCandidates()) // b) LSRA spills it // // Possible solution to address case (a) // - The conditions under which LSRA marks a varDsc as DoNotEnregister could be checked // in this routine. Note that live out of exception handler is something we may not be // able to do it here since GS cookie logic is invoked ahead of liveness computation. // Therefore, for methods with exception handling and need GS cookie check we might have // to take conservative approach. // // Possible solution to address case (b) // - Whenver a parameter passed in an argument register needs to be spilled by LSRA, we // create a new spill temp if the method needs GS cookie check. return varDsc->lvIsParam; #else // !defined(_TARGET_AMD64_) return varDsc->lvIsParam && !varDsc->lvIsRegArg; #endif } #ifdef DEBUG void Print() { printf("assignGroup [%p]; shadowCopy: [%d];\n", assignGroup, shadowCopy); } #endif }; GSCookie* gsGlobalSecurityCookieAddr; // Address of global cookie for unsafe buffer checks GSCookie gsGlobalSecurityCookieVal; // Value of global cookie if addr is NULL ShadowParamVarInfo* gsShadowVarInfo; // Table used by shadow param analysis code void gsGSChecksInitCookie(); // Grabs cookie variable void gsCopyShadowParams(); // Identify vulnerable params and create dhadow copies bool gsFindVulnerableParams(); // Shadow param analysis code void gsParamsToShadows(); // Insert copy code and replave param uses by shadow static fgWalkPreFn gsMarkPtrsAndAssignGroups; // Shadow param analysis tree-walk static fgWalkPreFn gsReplaceShadowParams; // Shadow param replacement tree-walk #define DEFAULT_MAX_INLINE_SIZE 100 // Methods with > DEFAULT_MAX_INLINE_SIZE IL bytes will never be inlined. // This can be overwritten by setting complus_JITInlineSize env variable. #define DEFAULT_MAX_INLINE_DEPTH 20 // Methods at more than this level deep will not be inlined #define DEFAULT_MAX_LOCALLOC_TO_LOCAL_SIZE 32 // fixed locallocs of this size or smaller will convert to local buffers private: #ifdef FEATURE_JIT_METHOD_PERF JitTimer* pCompJitTimer; // Timer data structure (by phases) for current compilation. static CompTimeSummaryInfo s_compJitTimerSummary; // Summary of the Timer information for the whole run. static LPCWSTR JitTimeLogCsv(); // Retrieve the file name for CSV from ConfigDWORD. static LPCWSTR compJitTimeLogFilename; // If a log file for JIT time is desired, filename to write it to. #endif inline void EndPhase(Phases phase); // Indicate the end of the given phase. #if MEASURE_CLRAPI_CALLS // Thin wrappers that call into JitTimer (if present). inline void CLRApiCallEnter(unsigned apix); inline void CLRApiCallLeave(unsigned apix); public: inline void CLR_API_Enter(API_ICorJitInfo_Names ename); inline void CLR_API_Leave(API_ICorJitInfo_Names ename); private: #endif #if defined(DEBUG) || defined(INLINE_DATA) || defined(FEATURE_CLRSQM) // These variables are associated with maintaining SQM data about compile time. unsigned __int64 m_compCyclesAtEndOfInlining; // The thread-virtualized cycle count at the end of the inlining phase // in the current compilation. unsigned __int64 m_compCycles; // Net cycle count for current compilation DWORD m_compTickCountAtEndOfInlining; // The result of GetTickCount() (# ms since some epoch marker) at the end of // the inlining phase in the current compilation. #endif // defined(DEBUG) || defined(INLINE_DATA) || defined(FEATURE_CLRSQM) // Records the SQM-relevant (cycles and tick count). Should be called after inlining is complete. // (We do this after inlining because this marks the last point at which the JIT is likely to cause // type-loading and class initialization). void RecordStateAtEndOfInlining(); // Assumes being called at the end of compilation. Update the SQM state. void RecordStateAtEndOfCompilation(); #ifdef FEATURE_CLRSQM // Does anything SQM related necessary at process shutdown time. static void ProcessShutdownSQMWork(ICorStaticInfo* statInfo); #endif // FEATURE_CLRSQM public: #if FUNC_INFO_LOGGING static LPCWSTR compJitFuncInfoFilename; // If a log file for per-function information is required, this is the // filename to write it to. static FILE* compJitFuncInfoFile; // And this is the actual FILE* to write to. #endif // FUNC_INFO_LOGGING Compiler* prevCompiler; // Previous compiler on stack for TLS Compiler* linked list for reentrant compilers. // Is the compilation in a full trust context? bool compIsFullTrust(); #if MEASURE_NOWAY void RecordNowayAssert(const char* filename, unsigned line, const char* condStr); #endif // MEASURE_NOWAY #ifndef FEATURE_TRACELOGGING // Should we actually fire the noway assert body and the exception handler? bool compShouldThrowOnNoway(); #else // FEATURE_TRACELOGGING // Should we actually fire the noway assert body and the exception handler? bool compShouldThrowOnNoway(const char* filename, unsigned line); // Telemetry instance to use per method compilation. JitTelemetry compJitTelemetry; // Get common parameters that have to be logged with most telemetry data. void compGetTelemetryDefaults(const char** assemblyName, const char** scopeName, const char** methodName, unsigned* methodHash); #endif // !FEATURE_TRACELOGGING #ifdef DEBUG private: NodeToTestDataMap* m_nodeTestData; static const unsigned FIRST_LOOP_HOIST_CSE_CLASS = 1000; unsigned m_loopHoistCSEClass; // LoopHoist test annotations turn into CSE requirements; we // label them with CSE Class #'s starting at FIRST_LOOP_HOIST_CSE_CLASS. // Current kept in this. public: NodeToTestDataMap* GetNodeTestData() { Compiler* compRoot = impInlineRoot(); if (compRoot->m_nodeTestData == nullptr) { compRoot->m_nodeTestData = new (getAllocatorDebugOnly()) NodeToTestDataMap(getAllocatorDebugOnly()); } return compRoot->m_nodeTestData; } typedef JitHashTable, int> NodeToIntMap; // Returns the set (i.e., the domain of the result map) of nodes that are keys in m_nodeTestData, and // currently occur in the AST graph. NodeToIntMap* FindReachableNodesInNodeTestData(); // Node "from" is being eliminated, and being replaced by node "to". If "from" had any associated // test data, associate that data with "to". void TransferTestDataToNode(GenTree* from, GenTree* to); // Requires that "to" is a clone of "from". If any nodes in the "from" tree // have annotations, attach similar annotations to the corresponding nodes in "to". void CopyTestDataToCloneTree(GenTree* from, GenTree* to); // These are the methods that test that the various conditions implied by the // test attributes are satisfied. void JitTestCheckSSA(); // SSA builder tests. void JitTestCheckVN(); // Value numbering tests. #endif // DEBUG // The "FieldSeqStore", for canonicalizing field sequences. See the definition of FieldSeqStore for // operations. FieldSeqStore* m_fieldSeqStore; FieldSeqStore* GetFieldSeqStore() { Compiler* compRoot = impInlineRoot(); if (compRoot->m_fieldSeqStore == nullptr) { // Create a CompAllocator that labels sub-structure with CMK_FieldSeqStore, and use that for allocation. CompAllocator ialloc(getAllocator(CMK_FieldSeqStore)); compRoot->m_fieldSeqStore = new (ialloc) FieldSeqStore(ialloc); } return compRoot->m_fieldSeqStore; } typedef JitHashTable, FieldSeqNode*> NodeToFieldSeqMap; // Some nodes of "TYP_BYREF" or "TYP_I_IMPL" actually represent the address of a field within a struct, but since // the offset of the field is zero, there's no "GT_ADD" node. We normally attach a field sequence to the constant // that is added, but what do we do when that constant is zero, and is thus not present? We use this mechanism to // attach the field sequence directly to the address node. NodeToFieldSeqMap* m_zeroOffsetFieldMap; NodeToFieldSeqMap* GetZeroOffsetFieldMap() { // Don't need to worry about inlining here if (m_zeroOffsetFieldMap == nullptr) { // Create a CompAllocator that labels sub-structure with CMK_ZeroOffsetFieldMap, and use that for // allocation. CompAllocator ialloc(getAllocator(CMK_ZeroOffsetFieldMap)); m_zeroOffsetFieldMap = new (ialloc) NodeToFieldSeqMap(ialloc); } return m_zeroOffsetFieldMap; } // Requires that "op1" is a node of type "TYP_BYREF" or "TYP_I_IMPL". We are dereferencing this with the fields in // "fieldSeq", whose offsets are required all to be zero. Ensures that any field sequence annotation currently on // "op1" or its components is augmented by appending "fieldSeq". In practice, if "op1" is a GT_LCL_FLD, it has // a field sequence as a member; otherwise, it may be the addition of an a byref and a constant, where the const // has a field sequence -- in this case "fieldSeq" is appended to that of the constant; otherwise, we // record the the field sequence using the ZeroOffsetFieldMap described above. // // One exception above is that "op1" is a node of type "TYP_REF" where "op1" is a GT_LCL_VAR. // This happens when System.Object vtable pointer is a regular field at offset 0 in System.Private.CoreLib in // CoreRT. Such case is handled same as the default case. void fgAddFieldSeqForZeroOffset(GenTree* op1, FieldSeqNode* fieldSeq); typedef JitHashTable, ArrayInfo> NodeToArrayInfoMap; NodeToArrayInfoMap* m_arrayInfoMap; NodeToArrayInfoMap* GetArrayInfoMap() { Compiler* compRoot = impInlineRoot(); if (compRoot->m_arrayInfoMap == nullptr) { // Create a CompAllocator that labels sub-structure with CMK_ArrayInfoMap, and use that for allocation. CompAllocator ialloc(getAllocator(CMK_ArrayInfoMap)); compRoot->m_arrayInfoMap = new (ialloc) NodeToArrayInfoMap(ialloc); } return compRoot->m_arrayInfoMap; } //----------------------------------------------------------------------------------------------------------------- // Compiler::TryGetArrayInfo: // Given an indirection node, checks to see whether or not that indirection represents an array access, and // if so returns information about the array. // // Arguments: // indir - The `GT_IND` node. // arrayInfo (out) - Information about the accessed array if this function returns true. Undefined otherwise. // // Returns: // True if the `GT_IND` node represents an array access; false otherwise. bool TryGetArrayInfo(GenTreeIndir* indir, ArrayInfo* arrayInfo) { if ((indir->gtFlags & GTF_IND_ARR_INDEX) == 0) { return false; } if (indir->gtOp1->OperIs(GT_INDEX_ADDR)) { GenTreeIndexAddr* const indexAddr = indir->gtOp1->AsIndexAddr(); *arrayInfo = ArrayInfo(indexAddr->gtElemType, indexAddr->gtElemSize, indexAddr->gtElemOffset, indexAddr->gtStructElemClass); return true; } bool found = GetArrayInfoMap()->Lookup(indir, arrayInfo); assert(found); return true; } NodeToUnsignedMap* m_memorySsaMap[MemoryKindCount]; // In some cases, we want to assign intermediate SSA #'s to memory states, and know what nodes create those memory // states. (We do this for try blocks, where, if the try block doesn't do a call that loses track of the memory // state, all the possible memory states are possible initial states of the corresponding catch block(s).) NodeToUnsignedMap* GetMemorySsaMap(MemoryKind memoryKind) { if (memoryKind == GcHeap && byrefStatesMatchGcHeapStates) { // Use the same map for GCHeap and ByrefExposed when their states match. memoryKind = ByrefExposed; } assert(memoryKind < MemoryKindCount); Compiler* compRoot = impInlineRoot(); if (compRoot->m_memorySsaMap[memoryKind] == nullptr) { // Create a CompAllocator that labels sub-structure with CMK_ArrayInfoMap, and use that for allocation. CompAllocator ialloc(getAllocator(CMK_ArrayInfoMap)); compRoot->m_memorySsaMap[memoryKind] = new (ialloc) NodeToUnsignedMap(ialloc); } return compRoot->m_memorySsaMap[memoryKind]; } // The Refany type is the only struct type whose structure is implicitly assumed by IL. We need its fields. CORINFO_CLASS_HANDLE m_refAnyClass; CORINFO_FIELD_HANDLE GetRefanyDataField() { if (m_refAnyClass == nullptr) { m_refAnyClass = info.compCompHnd->getBuiltinClass(CLASSID_TYPED_BYREF); } return info.compCompHnd->getFieldInClass(m_refAnyClass, 0); } CORINFO_FIELD_HANDLE GetRefanyTypeField() { if (m_refAnyClass == nullptr) { m_refAnyClass = info.compCompHnd->getBuiltinClass(CLASSID_TYPED_BYREF); } return info.compCompHnd->getFieldInClass(m_refAnyClass, 1); } #if VARSET_COUNTOPS static BitSetSupport::BitSetOpCounter m_varsetOpCounter; #endif #if ALLVARSET_COUNTOPS static BitSetSupport::BitSetOpCounter m_allvarsetOpCounter; #endif static HelperCallProperties s_helperCallProperties; #ifdef UNIX_AMD64_ABI static var_types GetTypeFromClassificationAndSizes(SystemVClassificationType classType, int size); static var_types GetEightByteType(const SYSTEMV_AMD64_CORINFO_STRUCT_REG_PASSING_DESCRIPTOR& structDesc, unsigned slotNum); static void GetStructTypeOffset(const SYSTEMV_AMD64_CORINFO_STRUCT_REG_PASSING_DESCRIPTOR& structDesc, var_types* type0, var_types* type1, unsigned __int8* offset0, unsigned __int8* offset1); void GetStructTypeOffset(CORINFO_CLASS_HANDLE typeHnd, var_types* type0, var_types* type1, unsigned __int8* offset0, unsigned __int8* offset1); #endif // defined(UNIX_AMD64_ABI) void fgMorphMultiregStructArgs(GenTreeCall* call); GenTree* fgMorphMultiregStructArg(GenTree* arg, fgArgTabEntry* fgEntryPtr); bool killGCRefs(GenTree* tree); }; // end of class Compiler //--------------------------------------------------------------------------------------------------------------------- // GenTreeVisitor: a flexible tree walker implemented using the curiosly-recurring-template pattern. // // This class implements a configurable walker for IR trees. There are five configuration options (defaults values are // shown in parentheses): // // - ComputeStack (false): when true, the walker will push each node onto the `m_ancestors` stack. "Ancestors" is a bit // of a misnomer, as the first entry will always be the current node. // // - DoPreOrder (false): when true, the walker will invoke `TVisitor::PreOrderVisit` with the current node as an // argument before visiting the node's operands. // // - DoPostOrder (false): when true, the walker will invoke `TVisitor::PostOrderVisit` with the current node as an // argument after visiting the node's operands. // // - DoLclVarsOnly (false): when true, the walker will only invoke `TVisitor::PreOrderVisit` for lclVar nodes. // `DoPreOrder` must be true if this option is true. // // - UseExecutionOrder (false): when true, then walker will visit a node's operands in execution order (e.g. if a // binary operator has the `GTF_REVERSE_OPS` flag set, the second operand will be // visited before the first). // // At least one of `DoPreOrder` and `DoPostOrder` must be specified. // // A simple pre-order visitor might look something like the following: // // class CountingVisitor final : public GenTreeVisitor // { // public: // enum // { // DoPreOrder = true // }; // // unsigned m_count; // // CountingVisitor(Compiler* compiler) // : GenTreeVisitor(compiler), m_count(0) // { // } // // Compiler::fgWalkResult PreOrderVisit(GenTree* node) // { // m_count++; // } // }; // // This visitor would then be used like so: // // CountingVisitor countingVisitor(compiler); // countingVisitor.WalkTree(root); // template class GenTreeVisitor { protected: typedef Compiler::fgWalkResult fgWalkResult; enum { ComputeStack = false, DoPreOrder = false, DoPostOrder = false, DoLclVarsOnly = false, UseExecutionOrder = false, }; Compiler* m_compiler; ArrayStack m_ancestors; GenTreeVisitor(Compiler* compiler) : m_compiler(compiler), m_ancestors(compiler->getAllocator(CMK_ArrayStack)) { assert(compiler != nullptr); static_assert_no_msg(TVisitor::DoPreOrder || TVisitor::DoPostOrder); static_assert_no_msg(!TVisitor::DoLclVarsOnly || TVisitor::DoPreOrder); } fgWalkResult PreOrderVisit(GenTree** use, GenTree* user) { return fgWalkResult::WALK_CONTINUE; } fgWalkResult PostOrderVisit(GenTree** use, GenTree* user) { return fgWalkResult::WALK_CONTINUE; } public: fgWalkResult WalkTree(GenTree** use, GenTree* user) { assert(use != nullptr); GenTree* node = *use; if (TVisitor::ComputeStack) { m_ancestors.Push(node); } fgWalkResult result = fgWalkResult::WALK_CONTINUE; if (TVisitor::DoPreOrder && !TVisitor::DoLclVarsOnly) { result = reinterpret_cast(this)->PreOrderVisit(use, user); if (result == fgWalkResult::WALK_ABORT) { return result; } node = *use; if ((node == nullptr) || (result == fgWalkResult::WALK_SKIP_SUBTREES)) { goto DONE; } } switch (node->OperGet()) { // Leaf lclVars case GT_LCL_VAR: case GT_LCL_FLD: case GT_LCL_VAR_ADDR: case GT_LCL_FLD_ADDR: if (TVisitor::DoLclVarsOnly) { result = reinterpret_cast(this)->PreOrderVisit(use, user); if (result == fgWalkResult::WALK_ABORT) { return result; } } __fallthrough; // Leaf nodes case GT_CATCH_ARG: case GT_LABEL: case GT_FTN_ADDR: case GT_RET_EXPR: case GT_CNS_INT: case GT_CNS_LNG: case GT_CNS_DBL: case GT_CNS_STR: case GT_MEMORYBARRIER: case GT_JMP: case GT_JCC: case GT_SETCC: case GT_NO_OP: case GT_START_NONGC: case GT_START_PREEMPTGC: case GT_PROF_HOOK: #if !FEATURE_EH_FUNCLETS case GT_END_LFIN: #endif // !FEATURE_EH_FUNCLETS case GT_PHI_ARG: case GT_JMPTABLE: case GT_CLS_VAR: case GT_CLS_VAR_ADDR: case GT_ARGPLACE: case GT_PHYSREG: case GT_EMITNOP: case GT_PINVOKE_PROLOG: case GT_PINVOKE_EPILOG: case GT_IL_OFFSET: break; // Lclvar unary operators case GT_STORE_LCL_VAR: case GT_STORE_LCL_FLD: if (TVisitor::DoLclVarsOnly) { result = reinterpret_cast(this)->PreOrderVisit(use, user); if (result == fgWalkResult::WALK_ABORT) { return result; } } __fallthrough; // Standard unary operators case GT_NOT: case GT_NEG: case GT_BSWAP: case GT_BSWAP16: case GT_COPY: case GT_RELOAD: case GT_ARR_LENGTH: case GT_CAST: case GT_BITCAST: case GT_CKFINITE: case GT_LCLHEAP: case GT_ADDR: case GT_IND: case GT_OBJ: case GT_BLK: case GT_BOX: case GT_ALLOCOBJ: case GT_INIT_VAL: case GT_JTRUE: case GT_SWITCH: case GT_NULLCHECK: case GT_PUTARG_REG: case GT_PUTARG_STK: case GT_RETURNTRAP: case GT_NOP: case GT_RETURN: case GT_RETFILT: case GT_PHI: case GT_RUNTIMELOOKUP: { GenTreeUnOp* const unOp = node->AsUnOp(); if (unOp->gtOp1 != nullptr) { result = WalkTree(&unOp->gtOp1, unOp); if (result == fgWalkResult::WALK_ABORT) { return result; } } break; } // Special nodes case GT_CMPXCHG: { GenTreeCmpXchg* const cmpXchg = node->AsCmpXchg(); result = WalkTree(&cmpXchg->gtOpLocation, cmpXchg); if (result == fgWalkResult::WALK_ABORT) { return result; } result = WalkTree(&cmpXchg->gtOpValue, cmpXchg); if (result == fgWalkResult::WALK_ABORT) { return result; } result = WalkTree(&cmpXchg->gtOpComparand, cmpXchg); if (result == fgWalkResult::WALK_ABORT) { return result; } break; } case GT_ARR_BOUNDS_CHECK: #ifdef FEATURE_SIMD case GT_SIMD_CHK: #endif // FEATURE_SIMD #ifdef FEATURE_HW_INTRINSICS case GT_HW_INTRINSIC_CHK: #endif // FEATURE_HW_INTRINSICS { GenTreeBoundsChk* const boundsChk = node->AsBoundsChk(); result = WalkTree(&boundsChk->gtIndex, boundsChk); if (result == fgWalkResult::WALK_ABORT) { return result; } result = WalkTree(&boundsChk->gtArrLen, boundsChk); if (result == fgWalkResult::WALK_ABORT) { return result; } break; } case GT_FIELD: { GenTreeField* const field = node->AsField(); if (field->gtFldObj != nullptr) { result = WalkTree(&field->gtFldObj, field); if (result == fgWalkResult::WALK_ABORT) { return result; } } break; } case GT_ARR_ELEM: { GenTreeArrElem* const arrElem = node->AsArrElem(); result = WalkTree(&arrElem->gtArrObj, arrElem); if (result == fgWalkResult::WALK_ABORT) { return result; } const unsigned rank = arrElem->gtArrRank; for (unsigned dim = 0; dim < rank; dim++) { result = WalkTree(&arrElem->gtArrInds[dim], arrElem); if (result == fgWalkResult::WALK_ABORT) { return result; } } break; } case GT_ARR_OFFSET: { GenTreeArrOffs* const arrOffs = node->AsArrOffs(); result = WalkTree(&arrOffs->gtOffset, arrOffs); if (result == fgWalkResult::WALK_ABORT) { return result; } result = WalkTree(&arrOffs->gtIndex, arrOffs); if (result == fgWalkResult::WALK_ABORT) { return result; } result = WalkTree(&arrOffs->gtArrObj, arrOffs); if (result == fgWalkResult::WALK_ABORT) { return result; } break; } case GT_DYN_BLK: { GenTreeDynBlk* const dynBlock = node->AsDynBlk(); GenTree** op1Use = &dynBlock->gtOp1; GenTree** op2Use = &dynBlock->gtDynamicSize; if (TVisitor::UseExecutionOrder && dynBlock->gtEvalSizeFirst) { std::swap(op1Use, op2Use); } result = WalkTree(op1Use, dynBlock); if (result == fgWalkResult::WALK_ABORT) { return result; } result = WalkTree(op2Use, dynBlock); if (result == fgWalkResult::WALK_ABORT) { return result; } break; } case GT_STORE_DYN_BLK: { GenTreeDynBlk* const dynBlock = node->AsDynBlk(); GenTree** op1Use = &dynBlock->gtOp1; GenTree** op2Use = &dynBlock->gtOp2; GenTree** op3Use = &dynBlock->gtDynamicSize; if (TVisitor::UseExecutionOrder) { if (dynBlock->IsReverseOp()) { std::swap(op1Use, op2Use); } if (dynBlock->gtEvalSizeFirst) { std::swap(op3Use, op2Use); std::swap(op2Use, op1Use); } } result = WalkTree(op1Use, dynBlock); if (result == fgWalkResult::WALK_ABORT) { return result; } result = WalkTree(op2Use, dynBlock); if (result == fgWalkResult::WALK_ABORT) { return result; } result = WalkTree(op3Use, dynBlock); if (result == fgWalkResult::WALK_ABORT) { return result; } break; } case GT_CALL: { GenTreeCall* const call = node->AsCall(); if (call->gtCallObjp != nullptr) { result = WalkTree(&call->gtCallObjp, call); if (result == fgWalkResult::WALK_ABORT) { return result; } } for (GenTreeArgList* args = call->gtCallArgs; args != nullptr; args = args->Rest()) { result = WalkTree(args->pCurrent(), call); if (result == fgWalkResult::WALK_ABORT) { return result; } } for (GenTreeArgList* args = call->gtCallLateArgs; args != nullptr; args = args->Rest()) { result = WalkTree(args->pCurrent(), call); if (result == fgWalkResult::WALK_ABORT) { return result; } } if (call->gtCallType == CT_INDIRECT) { if (call->gtCallCookie != nullptr) { result = WalkTree(&call->gtCallCookie, call); if (result == fgWalkResult::WALK_ABORT) { return result; } } result = WalkTree(&call->gtCallAddr, call); if (result == fgWalkResult::WALK_ABORT) { return result; } } if (call->gtControlExpr != nullptr) { result = WalkTree(&call->gtControlExpr, call); if (result == fgWalkResult::WALK_ABORT) { return result; } } break; } // Binary nodes default: { assert(node->OperIsBinary()); GenTreeOp* const op = node->AsOp(); GenTree** op1Use = &op->gtOp1; GenTree** op2Use = &op->gtOp2; if (TVisitor::UseExecutionOrder && node->IsReverseOp()) { std::swap(op1Use, op2Use); } if (*op1Use != nullptr) { result = WalkTree(op1Use, op); if (result == fgWalkResult::WALK_ABORT) { return result; } } if (*op2Use != nullptr) { result = WalkTree(op2Use, op); if (result == fgWalkResult::WALK_ABORT) { return result; } } break; } } DONE: // Finally, visit the current node if (TVisitor::DoPostOrder) { result = reinterpret_cast(this)->PostOrderVisit(use, user); } if (TVisitor::ComputeStack) { m_ancestors.Pop(); } return result; } }; template class GenericTreeWalker final : public GenTreeVisitor> { public: enum { ComputeStack = computeStack, DoPreOrder = doPreOrder, DoPostOrder = doPostOrder, DoLclVarsOnly = doLclVarsOnly, UseExecutionOrder = useExecutionOrder, }; private: Compiler::fgWalkData* m_walkData; public: GenericTreeWalker(Compiler::fgWalkData* walkData) : GenTreeVisitor>( walkData->compiler) , m_walkData(walkData) { assert(walkData != nullptr); if (computeStack) { walkData->parentStack = &this->m_ancestors; } } Compiler::fgWalkResult PreOrderVisit(GenTree** use, GenTree* user) { m_walkData->parent = user; return m_walkData->wtprVisitorFn(use, m_walkData); } Compiler::fgWalkResult PostOrderVisit(GenTree** use, GenTree* user) { m_walkData->parent = user; return m_walkData->wtpoVisitorFn(use, m_walkData); } }; /* XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX Miscellaneous Compiler stuff XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ // Values used to mark the types a stack slot is used for const unsigned TYPE_REF_INT = 0x01; // slot used as a 32-bit int const unsigned TYPE_REF_LNG = 0x02; // slot used as a 64-bit long const unsigned TYPE_REF_FLT = 0x04; // slot used as a 32-bit float const unsigned TYPE_REF_DBL = 0x08; // slot used as a 64-bit float const unsigned TYPE_REF_PTR = 0x10; // slot used as a 32-bit pointer const unsigned TYPE_REF_BYR = 0x20; // slot used as a byref pointer const unsigned TYPE_REF_STC = 0x40; // slot used as a struct const unsigned TYPE_REF_TYPEMASK = 0x7F; // bits that represent the type // const unsigned TYPE_REF_ADDR_TAKEN = 0x80; // slots address was taken /***************************************************************************** * * Variables to keep track of total code amounts. */ #if DISPLAY_SIZES extern size_t grossVMsize; extern size_t grossNCsize; extern size_t totalNCsize; extern unsigned genMethodICnt; extern unsigned genMethodNCnt; extern size_t gcHeaderISize; extern size_t gcPtrMapISize; extern size_t gcHeaderNSize; extern size_t gcPtrMapNSize; #endif // DISPLAY_SIZES /***************************************************************************** * * Variables to keep track of basic block counts (more data on 1 BB methods) */ #if COUNT_BASIC_BLOCKS extern Histogram bbCntTable; extern Histogram bbOneBBSizeTable; #endif /***************************************************************************** * * Used by optFindNaturalLoops to gather statistical information such as * - total number of natural loops * - number of loops with 1, 2, ... exit conditions * - number of loops that have an iterator (for like) * - number of loops that have a constant iterator */ #if COUNT_LOOPS extern unsigned totalLoopMethods; // counts the total number of methods that have natural loops extern unsigned maxLoopsPerMethod; // counts the maximum number of loops a method has extern unsigned totalLoopOverflows; // # of methods that identified more loops than we can represent extern unsigned totalLoopCount; // counts the total number of natural loops extern unsigned totalUnnatLoopCount; // counts the total number of (not-necessarily natural) loops extern unsigned totalUnnatLoopOverflows; // # of methods that identified more unnatural loops than we can represent extern unsigned iterLoopCount; // counts the # of loops with an iterator (for like) extern unsigned simpleTestLoopCount; // counts the # of loops with an iterator and a simple loop condition (iter < // const) extern unsigned constIterLoopCount; // counts the # of loops with a constant iterator (for like) extern bool hasMethodLoops; // flag to keep track if we already counted a method as having loops extern unsigned loopsThisMethod; // counts the number of loops in the current method extern bool loopOverflowThisMethod; // True if we exceeded the max # of loops in the method. extern Histogram loopCountTable; // Histogram of loop counts extern Histogram loopExitCountTable; // Histogram of loop exit counts #endif // COUNT_LOOPS /***************************************************************************** * variables to keep track of how many iterations we go in a dataflow pass */ #if DATAFLOW_ITER extern unsigned CSEiterCount; // counts the # of iteration for the CSE dataflow extern unsigned CFiterCount; // counts the # of iteration for the Const Folding dataflow #endif // DATAFLOW_ITER #if MEASURE_BLOCK_SIZE extern size_t genFlowNodeSize; extern size_t genFlowNodeCnt; #endif // MEASURE_BLOCK_SIZE #if MEASURE_NODE_SIZE struct NodeSizeStats { void Init() { genTreeNodeCnt = 0; genTreeNodeSize = 0; genTreeNodeActualSize = 0; } // Count of tree nodes allocated. unsigned __int64 genTreeNodeCnt; // The size we allocate. unsigned __int64 genTreeNodeSize; // The actual size of the node. Note that the actual size will likely be smaller // than the allocated size, but we sometimes use SetOper()/ChangeOper() to change // a smaller node to a larger one. TODO-Cleanup: add stats on // SetOper()/ChangeOper() usage to quantify this. unsigned __int64 genTreeNodeActualSize; }; extern NodeSizeStats genNodeSizeStats; // Total node size stats extern NodeSizeStats genNodeSizeStatsPerFunc; // Per-function node size stats extern Histogram genTreeNcntHist; extern Histogram genTreeNsizHist; #endif // MEASURE_NODE_SIZE /***************************************************************************** * Count fatal errors (including noway_asserts). */ #if MEASURE_FATAL extern unsigned fatal_badCode; extern unsigned fatal_noWay; extern unsigned fatal_NOMEM; extern unsigned fatal_noWayAssertBody; #ifdef DEBUG extern unsigned fatal_noWayAssertBodyArgs; #endif // DEBUG extern unsigned fatal_NYI; #endif // MEASURE_FATAL /***************************************************************************** * Codegen */ #ifdef _TARGET_XARCH_ const instruction INS_SHIFT_LEFT_LOGICAL = INS_shl; const instruction INS_SHIFT_RIGHT_LOGICAL = INS_shr; const instruction INS_SHIFT_RIGHT_ARITHM = INS_sar; const instruction INS_AND = INS_and; const instruction INS_OR = INS_or; const instruction INS_XOR = INS_xor; const instruction INS_NEG = INS_neg; const instruction INS_TEST = INS_test; const instruction INS_MUL = INS_imul; const instruction INS_SIGNED_DIVIDE = INS_idiv; const instruction INS_UNSIGNED_DIVIDE = INS_div; const instruction INS_BREAKPOINT = INS_int3; const instruction INS_ADDC = INS_adc; const instruction INS_SUBC = INS_sbb; const instruction INS_NOT = INS_not; #endif // _TARGET_XARCH_ #ifdef _TARGET_ARM_ const instruction INS_SHIFT_LEFT_LOGICAL = INS_lsl; const instruction INS_SHIFT_RIGHT_LOGICAL = INS_lsr; const instruction INS_SHIFT_RIGHT_ARITHM = INS_asr; const instruction INS_AND = INS_and; const instruction INS_OR = INS_orr; const instruction INS_XOR = INS_eor; const instruction INS_NEG = INS_rsb; const instruction INS_TEST = INS_tst; const instruction INS_MUL = INS_mul; const instruction INS_MULADD = INS_mla; const instruction INS_SIGNED_DIVIDE = INS_sdiv; const instruction INS_UNSIGNED_DIVIDE = INS_udiv; const instruction INS_BREAKPOINT = INS_bkpt; const instruction INS_ADDC = INS_adc; const instruction INS_SUBC = INS_sbc; const instruction INS_NOT = INS_mvn; const instruction INS_ABS = INS_vabs; const instruction INS_SQRT = INS_vsqrt; #endif // _TARGET_ARM_ #ifdef _TARGET_ARM64_ const instruction INS_MULADD = INS_madd; const instruction INS_BREAKPOINT = INS_bkpt; const instruction INS_ABS = INS_fabs; const instruction INS_SQRT = INS_fsqrt; #endif // _TARGET_ARM64_ /*****************************************************************************/ extern const BYTE genTypeSizes[]; extern const BYTE genTypeAlignments[]; extern const BYTE genTypeStSzs[]; extern const BYTE genActualTypes[]; /*****************************************************************************/ // VERY_LARGE_FRAME_SIZE_REG_MASK is the set of registers we need to use for // the probing loop generated for very large stack frames (see `getVeryLargeFrameSize`). // We only use this to ensure that if we need to reserve a callee-saved register, // it will be reserved. For ARM32, only R12 and LR are non-callee-saved, non-argument // registers, so we save at least one more callee-saved register. For ARM64, however, // we already know we have at least three non-callee-saved, non-argument integer registers, // so we don't need to save any more. #ifdef _TARGET_ARM_ #define VERY_LARGE_FRAME_SIZE_REG_MASK (RBM_R4) #endif /*****************************************************************************/ extern BasicBlock dummyBB; /*****************************************************************************/ /*****************************************************************************/ // foreach_block: An iterator over all blocks in the function. // __compiler: the Compiler* object // __block : a BasicBlock*, already declared, that gets updated each iteration. #define foreach_block(__compiler, __block) \ for ((__block) = (__compiler)->fgFirstBB; (__block); (__block) = (__block)->bbNext) /*****************************************************************************/ /*****************************************************************************/ #ifdef DEBUG void dumpConvertedVarSet(Compiler* comp, VARSET_VALARG_TP vars); /*XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XX XX XX Debugging helpers XX XX XX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX */ /*****************************************************************************/ /* The following functions are intended to be called from the debugger, to dump * various data structures. The can be used in the debugger Watch or Quick Watch * windows. They are designed to be short to type and take as few arguments as * possible. The 'c' versions take a Compiler*, whereas the 'd' versions use the TlsCompiler. * See the function definition comment for more details. */ void cBlock(Compiler* comp, BasicBlock* block); void cBlocks(Compiler* comp); void cBlocksV(Compiler* comp); void cTree(Compiler* comp, GenTree* tree); void cTrees(Compiler* comp); void cEH(Compiler* comp); void cVar(Compiler* comp, unsigned lclNum); void cVarDsc(Compiler* comp, LclVarDsc* varDsc); void cVars(Compiler* comp); void cVarsFinal(Compiler* comp); void cBlockPreds(Compiler* comp, BasicBlock* block); void cReach(Compiler* comp); void cDoms(Compiler* comp); void cLiveness(Compiler* comp); void cCVarSet(Compiler* comp, VARSET_VALARG_TP vars); void cFuncIR(Compiler* comp); void cBlockIR(Compiler* comp, BasicBlock* block); void cLoopIR(Compiler* comp, Compiler::LoopDsc* loop); void cTreeIR(Compiler* comp, GenTree* tree); int cTreeTypeIR(Compiler* comp, GenTree* tree); int cTreeKindsIR(Compiler* comp, GenTree* tree); int cTreeFlagsIR(Compiler* comp, GenTree* tree); int cOperandIR(Compiler* comp, GenTree* operand); int cLeafIR(Compiler* comp, GenTree* tree); int cIndirIR(Compiler* comp, GenTree* tree); int cListIR(Compiler* comp, GenTree* list); int cSsaNumIR(Compiler* comp, GenTree* tree); int cValNumIR(Compiler* comp, GenTree* tree); int cDependsIR(Compiler* comp, GenTree* comma, bool* first); void dBlock(BasicBlock* block); void dBlocks(); void dBlocksV(); void dTree(GenTree* tree); void dTrees(); void dEH(); void dVar(unsigned lclNum); void dVarDsc(LclVarDsc* varDsc); void dVars(); void dVarsFinal(); void dBlockPreds(BasicBlock* block); void dReach(); void dDoms(); void dLiveness(); void dCVarSet(VARSET_VALARG_TP vars); void dRegMask(regMaskTP mask); void dFuncIR(); void dBlockIR(BasicBlock* block); void dTreeIR(GenTree* tree); void dLoopIR(Compiler::LoopDsc* loop); void dLoopNumIR(unsigned loopNum); int dTabStopIR(int curr, int tabstop); int dTreeTypeIR(GenTree* tree); int dTreeKindsIR(GenTree* tree); int dTreeFlagsIR(GenTree* tree); int dOperandIR(GenTree* operand); int dLeafIR(GenTree* tree); int dIndirIR(GenTree* tree); int dListIR(GenTree* list); int dSsaNumIR(GenTree* tree); int dValNumIR(GenTree* tree); int dDependsIR(GenTree* comma); void dFormatIR(); GenTree* dFindTree(GenTree* tree, unsigned id); GenTree* dFindTree(unsigned id); GenTreeStmt* dFindStmt(unsigned id); BasicBlock* dFindBlock(unsigned bbNum); #endif // DEBUG #include "compiler.hpp" // All the shared inline functions /*****************************************************************************/ #endif //_COMPILER_H_ /*****************************************************************************/