diff options
Diffstat (limited to 'llvm/lib/CodeGen')
29 files changed, 774 insertions, 165 deletions
diff --git a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp index e2af0c5..8aa488f 100644 --- a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp +++ b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp @@ -41,6 +41,7 @@ #include "llvm/CodeGen/GCMetadataPrinter.h" #include "llvm/CodeGen/LazyMachineBlockFrequencyInfo.h" #include "llvm/CodeGen/MachineBasicBlock.h" +#include "llvm/CodeGen/MachineBlockHashInfo.h" #include "llvm/CodeGen/MachineBranchProbabilityInfo.h" #include "llvm/CodeGen/MachineConstantPool.h" #include "llvm/CodeGen/MachineDominators.h" @@ -184,6 +185,8 @@ static cl::opt<bool> PrintLatency( cl::desc("Print instruction latencies as verbose asm comments"), cl::Hidden, cl::init(false)); +extern cl::opt<bool> EmitBBHash; + STATISTIC(EmittedInsts, "Number of machine instrs printed"); char AsmPrinter::ID = 0; @@ -474,6 +477,8 @@ void AsmPrinter::getAnalysisUsage(AnalysisUsage &AU) const { AU.addRequired<GCModuleInfo>(); AU.addRequired<LazyMachineBlockFrequencyInfoPass>(); AU.addRequired<MachineBranchProbabilityInfoWrapperPass>(); + if (EmitBBHash) + AU.addRequired<MachineBlockHashInfo>(); } bool AsmPrinter::doInitialization(Module &M) { @@ -1434,13 +1439,11 @@ getBBAddrMapFeature(const MachineFunction &MF, int NumMBBSectionRanges, "BB entries info is required for BBFreq and BrProb " "features"); } - return {FuncEntryCountEnabled, - BBFreqEnabled, - BrProbEnabled, + return {FuncEntryCountEnabled, BBFreqEnabled, BrProbEnabled, MF.hasBBSections() && NumMBBSectionRanges > 1, - static_cast<bool>(BBAddrMapSkipEmitBBEntries), - HasCalls, - false}; + // Use static_cast to avoid breakage of tests on windows. + static_cast<bool>(BBAddrMapSkipEmitBBEntries), HasCalls, + static_cast<bool>(EmitBBHash)}; } void AsmPrinter::emitBBAddrMapSection(const MachineFunction &MF) { @@ -1499,6 +1502,9 @@ void AsmPrinter::emitBBAddrMapSection(const MachineFunction &MF) { PrevMBBEndSymbol = MBBSymbol; } + auto MBHI = + Features.BBHash ? &getAnalysis<MachineBlockHashInfo>() : nullptr; + if (!Features.OmitBBEntries) { OutStreamer->AddComment("BB id"); // Emit the BB ID for this basic block. @@ -1526,6 +1532,10 @@ void AsmPrinter::emitBBAddrMapSection(const MachineFunction &MF) { emitLabelDifferenceAsULEB128(MBB.getEndSymbol(), CurrentLabel); // Emit the Metadata. OutStreamer->emitULEB128IntValue(getBBAddrMapMetadata(MBB)); + // Emit the Hash. + if (MBHI) { + OutStreamer->emitInt64(MBHI->getMBBHash(MBB)); + } } PrevMBBEndSymbol = MBB.getEndSymbol(); } diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.h b/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.h index 9288d7e..9c0b68b 100644 --- a/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.h +++ b/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.h @@ -334,7 +334,7 @@ public: const DIE &TyDIE); protected: - ~DwarfUnit(); + ~DwarfUnit() override; /// Create new static data member DIE. DIE *getOrCreateStaticMemberDIE(const DIDerivedType *DT); diff --git a/llvm/lib/CodeGen/CMakeLists.txt b/llvm/lib/CodeGen/CMakeLists.txt index b6872605..4373c53 100644 --- a/llvm/lib/CodeGen/CMakeLists.txt +++ b/llvm/lib/CodeGen/CMakeLists.txt @@ -108,6 +108,7 @@ add_llvm_component_library(LLVMCodeGen LowerEmuTLS.cpp MachineBasicBlock.cpp MachineBlockFrequencyInfo.cpp + MachineBlockHashInfo.cpp MachineBlockPlacement.cpp MachineBranchProbabilityInfo.cpp MachineCFGPrinter.cpp diff --git a/llvm/lib/CodeGen/CodeGen.cpp b/llvm/lib/CodeGen/CodeGen.cpp index c438eae..9795a0b 100644 --- a/llvm/lib/CodeGen/CodeGen.cpp +++ b/llvm/lib/CodeGen/CodeGen.cpp @@ -98,6 +98,7 @@ void llvm::initializeCodeGen(PassRegistry &Registry) { initializeMachineUniformityAnalysisPassPass(Registry); initializeMIR2VecVocabLegacyAnalysisPass(Registry); initializeMIR2VecVocabPrinterLegacyPassPass(Registry); + initializeMIR2VecPrinterLegacyPassPass(Registry); initializeMachineUniformityInfoPrinterPassPass(Registry); initializeMachineVerifierLegacyPassPass(Registry); initializeObjCARCContractLegacyPassPass(Registry); diff --git a/llvm/lib/CodeGen/CommandFlags.cpp b/llvm/lib/CodeGen/CommandFlags.cpp index 0522698..c1365f4 100644 --- a/llvm/lib/CodeGen/CommandFlags.cpp +++ b/llvm/lib/CodeGen/CommandFlags.cpp @@ -64,7 +64,6 @@ CGOPT_EXP(uint64_t, LargeDataThreshold) CGOPT(ExceptionHandling, ExceptionModel) CGOPT_EXP(CodeGenFileType, FileType) CGOPT(FramePointerKind, FramePointerUsage) -CGOPT(bool, EnableUnsafeFPMath) CGOPT(bool, EnableNoInfsFPMath) CGOPT(bool, EnableNoNaNsFPMath) CGOPT(bool, EnableNoSignedZerosFPMath) @@ -219,12 +218,6 @@ codegen::RegisterCodeGenFlags::RegisterCodeGenFlags() { "Enable frame pointer elimination"))); CGBINDOPT(FramePointerUsage); - static cl::opt<bool> EnableUnsafeFPMath( - "enable-unsafe-fp-math", - cl::desc("Enable optimizations that may decrease FP precision"), - cl::init(false)); - CGBINDOPT(EnableUnsafeFPMath); - static cl::opt<bool> EnableNoInfsFPMath( "enable-no-infs-fp-math", cl::desc("Enable FP math optimizations that assume no +-Infs"), @@ -552,7 +545,6 @@ TargetOptions codegen::InitTargetOptionsFromCodeGenFlags(const Triple &TheTriple) { TargetOptions Options; Options.AllowFPOpFusion = getFuseFPOps(); - Options.UnsafeFPMath = getEnableUnsafeFPMath(); Options.NoInfsFPMath = getEnableNoInfsFPMath(); Options.NoNaNsFPMath = getEnableNoNaNsFPMath(); Options.NoSignedZerosFPMath = getEnableNoSignedZerosFPMath(); @@ -706,7 +698,6 @@ void codegen::setFunctionAttributes(StringRef CPU, StringRef Features, if (getStackRealign()) NewAttrs.addAttribute("stackrealign"); - HANDLE_BOOL_ATTR(EnableUnsafeFPMathView, "unsafe-fp-math"); HANDLE_BOOL_ATTR(EnableNoInfsFPMathView, "no-infs-fp-math"); HANDLE_BOOL_ATTR(EnableNoNaNsFPMathView, "no-nans-fp-math"); HANDLE_BOOL_ATTR(EnableNoSignedZerosFPMathView, "no-signed-zeros-fp-math"); diff --git a/llvm/lib/CodeGen/ExpandFp.cpp b/llvm/lib/CodeGen/ExpandFp.cpp index 2b5ced3..f44eb22 100644 --- a/llvm/lib/CodeGen/ExpandFp.cpp +++ b/llvm/lib/CodeGen/ExpandFp.cpp @@ -1108,8 +1108,8 @@ public: }; } // namespace -ExpandFpPass::ExpandFpPass(const TargetMachine *TM, CodeGenOptLevel OptLevel) - : TM(TM), OptLevel(OptLevel) {} +ExpandFpPass::ExpandFpPass(const TargetMachine &TM, CodeGenOptLevel OptLevel) + : TM(&TM), OptLevel(OptLevel) {} void ExpandFpPass::printPipeline( raw_ostream &OS, function_ref<StringRef(StringRef)> MapClassName2PassName) { diff --git a/llvm/lib/CodeGen/GlobalISel/Combiner.cpp b/llvm/lib/CodeGen/GlobalISel/Combiner.cpp index 2cba6f0..0665437 100644 --- a/llvm/lib/CodeGen/GlobalISel/Combiner.cpp +++ b/llvm/lib/CodeGen/GlobalISel/Combiner.cpp @@ -62,7 +62,7 @@ public: static std::unique_ptr<WorkListMaintainer> create(Level Lvl, WorkListTy &WorkList, MachineRegisterInfo &MRI); - virtual ~WorkListMaintainer() = default; + ~WorkListMaintainer() override = default; void reportFullyCreatedInstrs() { LLVM_DEBUG({ @@ -95,7 +95,7 @@ public: WorkListMaintainerImpl(WorkListTy &WorkList, MachineRegisterInfo &MRI) : WorkList(WorkList), MRI(MRI) {} - virtual ~WorkListMaintainerImpl() = default; + ~WorkListMaintainerImpl() override = default; void reset() override { DeferList.clear(); diff --git a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp index 884c3f1..b49040b 100644 --- a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp +++ b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp @@ -139,7 +139,7 @@ class DILocationVerifier : public GISelChangeObserver { public: DILocationVerifier() = default; - ~DILocationVerifier() = default; + ~DILocationVerifier() override = default; const Instruction *getCurrentInst() const { return CurrInst; } void setCurrentInst(const Instruction *Inst) { CurrInst = Inst; } @@ -1862,15 +1862,19 @@ bool IRTranslator::translateVectorDeinterleave2Intrinsic( void IRTranslator::getStackGuard(Register DstReg, MachineIRBuilder &MIRBuilder) { + Value *Global = TLI->getSDagStackGuard(*MF->getFunction().getParent()); + if (!Global) { + LLVMContext &Ctx = MIRBuilder.getContext(); + Ctx.diagnose(DiagnosticInfoGeneric("unable to lower stackguard")); + MIRBuilder.buildUndef(DstReg); + return; + } + const TargetRegisterInfo *TRI = MF->getSubtarget().getRegisterInfo(); MRI->setRegClass(DstReg, TRI->getPointerRegClass()); auto MIB = MIRBuilder.buildInstr(TargetOpcode::LOAD_STACK_GUARD, {DstReg}, {}); - Value *Global = TLI->getSDagStackGuard(*MF->getFunction().getParent()); - if (!Global) - return; - unsigned AddrSpace = Global->getType()->getPointerAddressSpace(); LLT PtrTy = LLT::pointer(AddrSpace, DL->getPointerSizeInBits(AddrSpace)); diff --git a/llvm/lib/CodeGen/InlineSpiller.cpp b/llvm/lib/CodeGen/InlineSpiller.cpp index d6e8505..c3e0964 100644 --- a/llvm/lib/CodeGen/InlineSpiller.cpp +++ b/llvm/lib/CodeGen/InlineSpiller.cpp @@ -721,6 +721,9 @@ bool InlineSpiller::reMaterializeFor(LiveInterval &VirtReg, MachineInstr &MI) { // Allocate a new register for the remat. Register NewVReg = Edit->createFrom(Original); + // Constrain it to the register class of MI. + MRI.constrainRegClass(NewVReg, MRI.getRegClass(VirtReg.reg())); + // Finally we can rematerialize OrigMI before MI. SlotIndex DefIdx = Edit->rematerializeAt(*MI.getParent(), MI, NewVReg, RM, TRI); diff --git a/llvm/lib/CodeGen/LiveDebugValues/LiveDebugValues.cpp b/llvm/lib/CodeGen/LiveDebugValues/LiveDebugValues.cpp index b655375..94e3a82 100644 --- a/llvm/lib/CodeGen/LiveDebugValues/LiveDebugValues.cpp +++ b/llvm/lib/CodeGen/LiveDebugValues/LiveDebugValues.cpp @@ -69,7 +69,7 @@ public: static char ID; LiveDebugValuesLegacy(); - ~LiveDebugValuesLegacy() = default; + ~LiveDebugValuesLegacy() override = default; /// Calculate the liveness information for the given machine function. bool runOnMachineFunction(MachineFunction &MF) override; diff --git a/llvm/lib/CodeGen/LiveDebugValues/VarLocBasedImpl.cpp b/llvm/lib/CodeGen/LiveDebugValues/VarLocBasedImpl.cpp index b9ea03f..1c4b2f9 100644 --- a/llvm/lib/CodeGen/LiveDebugValues/VarLocBasedImpl.cpp +++ b/llvm/lib/CodeGen/LiveDebugValues/VarLocBasedImpl.cpp @@ -1094,7 +1094,7 @@ public: /// Default construct and initialize the pass. VarLocBasedLDV(); - ~VarLocBasedLDV(); + ~VarLocBasedLDV() override; /// Print to ostream with a message. void printVarLocInMBB(const MachineFunction &MF, const VarLocInMBB &V, diff --git a/llvm/lib/CodeGen/MIR2Vec.cpp b/llvm/lib/CodeGen/MIR2Vec.cpp index 5c78d98..00b37e7 100644 --- a/llvm/lib/CodeGen/MIR2Vec.cpp +++ b/llvm/lib/CodeGen/MIR2Vec.cpp @@ -12,6 +12,7 @@ //===----------------------------------------------------------------------===// #include "llvm/CodeGen/MIR2Vec.h" +#include "llvm/ADT/DepthFirstIterator.h" #include "llvm/ADT/Statistic.h" #include "llvm/CodeGen/TargetInstrInfo.h" #include "llvm/IR/Module.h" @@ -29,42 +30,85 @@ using namespace mir2vec; STATISTIC(MIRVocabMissCounter, "Number of lookups to MIR entities not present in the vocabulary"); -cl::OptionCategory llvm::mir2vec::MIR2VecCategory("MIR2Vec Options"); +namespace llvm { +namespace mir2vec { +cl::OptionCategory MIR2VecCategory("MIR2Vec Options"); // FIXME: Use a default vocab when not specified static cl::opt<std::string> VocabFile("mir2vec-vocab-path", cl::Optional, cl::desc("Path to the vocabulary file for MIR2Vec"), cl::init(""), cl::cat(MIR2VecCategory)); +cl::opt<float> OpcWeight("mir2vec-opc-weight", cl::Optional, cl::init(1.0), + cl::desc("Weight for machine opcode embeddings"), + cl::cat(MIR2VecCategory)); +cl::opt<float> CommonOperandWeight( + "mir2vec-common-operand-weight", cl::Optional, cl::init(1.0), + cl::desc("Weight for common operand embeddings"), cl::cat(MIR2VecCategory)); cl::opt<float> - llvm::mir2vec::OpcWeight("mir2vec-opc-weight", cl::Optional, cl::init(1.0), - cl::desc("Weight for machine opcode embeddings"), - cl::cat(MIR2VecCategory)); + RegOperandWeight("mir2vec-reg-operand-weight", cl::Optional, cl::init(1.0), + cl::desc("Weight for register operand embeddings"), + cl::cat(MIR2VecCategory)); +cl::opt<MIR2VecKind> MIR2VecEmbeddingKind( + "mir2vec-kind", cl::Optional, + cl::values(clEnumValN(MIR2VecKind::Symbolic, "symbolic", + "Generate symbolic embeddings for MIR")), + cl::init(MIR2VecKind::Symbolic), cl::desc("MIR2Vec embedding kind"), + cl::cat(MIR2VecCategory)); + +} // namespace mir2vec +} // namespace llvm //===----------------------------------------------------------------------===// -// Vocabulary Implementation +// Vocabulary //===----------------------------------------------------------------------===// -MIRVocabulary::MIRVocabulary(VocabMap &&OpcodeEntries, - const TargetInstrInfo &TII) - : TII(TII) { +MIRVocabulary::MIRVocabulary(VocabMap &&OpcodeMap, VocabMap &&CommonOperandMap, + VocabMap &&PhysicalRegisterMap, + VocabMap &&VirtualRegisterMap, + const TargetInstrInfo &TII, + const TargetRegisterInfo &TRI, + const MachineRegisterInfo &MRI) + : TII(TII), TRI(TRI), MRI(MRI) { buildCanonicalOpcodeMapping(); - unsigned CanonicalOpcodeCount = UniqueBaseOpcodeNames.size(); assert(CanonicalOpcodeCount > 0 && "No canonical opcodes found for target - invalid vocabulary"); - Layout.OperandBase = CanonicalOpcodeCount; - generateStorage(OpcodeEntries); + + buildRegisterOperandMapping(); + + // Define layout of vocabulary sections + Layout.OpcodeBase = 0; + Layout.CommonOperandBase = CanonicalOpcodeCount; + // We expect same classes for physical and virtual registers + Layout.PhyRegBase = Layout.CommonOperandBase + std::size(CommonOperandNames); + Layout.VirtRegBase = Layout.PhyRegBase + RegisterOperandNames.size(); + + generateStorage(OpcodeMap, CommonOperandMap, PhysicalRegisterMap, + VirtualRegisterMap); Layout.TotalEntries = Storage.size(); } -Expected<MIRVocabulary> MIRVocabulary::create(VocabMap &&Entries, - const TargetInstrInfo &TII) { - if (Entries.empty()) +Expected<MIRVocabulary> +MIRVocabulary::create(VocabMap &&OpcodeMap, VocabMap &&CommonOperandMap, + VocabMap &&PhyRegMap, VocabMap &&VirtRegMap, + const TargetInstrInfo &TII, const TargetRegisterInfo &TRI, + const MachineRegisterInfo &MRI) { + if (OpcodeMap.empty() || CommonOperandMap.empty() || PhyRegMap.empty() || + VirtRegMap.empty()) return createStringError(errc::invalid_argument, "Empty vocabulary entries provided"); - return MIRVocabulary(std::move(Entries), TII); + MIRVocabulary Vocab(std::move(OpcodeMap), std::move(CommonOperandMap), + std::move(PhyRegMap), std::move(VirtRegMap), TII, TRI, + MRI); + + // Validate Storage after construction + if (!Vocab.Storage.isValid()) + return createStringError(errc::invalid_argument, + "Failed to create valid vocabulary storage"); + Vocab.ZeroEmbedding = Embedding(Vocab.Storage.getDimension(), 0.0); + return std::move(Vocab); } std::string MIRVocabulary::extractBaseOpcodeName(StringRef InstrName) { @@ -111,22 +155,74 @@ unsigned MIRVocabulary::getCanonicalOpcodeIndex(unsigned Opcode) const { return getCanonicalIndexForBaseName(BaseOpcode); } +unsigned +MIRVocabulary::getCanonicalIndexForOperandName(StringRef OperandName) const { + auto It = std::find(std::begin(CommonOperandNames), + std::end(CommonOperandNames), OperandName); + assert(It != std::end(CommonOperandNames) && + "Operand name not found in common operands"); + return Layout.CommonOperandBase + + std::distance(std::begin(CommonOperandNames), It); +} + +unsigned +MIRVocabulary::getCanonicalIndexForRegisterClass(StringRef RegName, + bool IsPhysical) const { + auto It = std::find(RegisterOperandNames.begin(), RegisterOperandNames.end(), + RegName); + assert(It != RegisterOperandNames.end() && + "Register name not found in register operands"); + unsigned LocalIndex = std::distance(RegisterOperandNames.begin(), It); + return (IsPhysical ? Layout.PhyRegBase : Layout.VirtRegBase) + LocalIndex; +} + std::string MIRVocabulary::getStringKey(unsigned Pos) const { assert(Pos < Layout.TotalEntries && "Position out of bounds in vocabulary"); - // For now, all entries are opcodes since we only have one section - if (Pos < Layout.OperandBase && Pos < UniqueBaseOpcodeNames.size()) { + // Handle opcodes section + if (Pos < Layout.CommonOperandBase) { // Convert canonical index back to base opcode name auto It = UniqueBaseOpcodeNames.begin(); std::advance(It, Pos); + assert(It != UniqueBaseOpcodeNames.end() && + "Canonical index out of bounds in opcode section"); return *It; } - llvm_unreachable("Invalid position in vocabulary"); - return ""; + auto getLocalIndex = [](unsigned Pos, size_t BaseOffset, size_t Bound, + const char *Msg) { + unsigned LocalIndex = Pos - BaseOffset; + assert(LocalIndex < Bound && Msg); + return LocalIndex; + }; + + // Handle common operands section + if (Pos < Layout.PhyRegBase) { + unsigned LocalIndex = getLocalIndex( + Pos, Layout.CommonOperandBase, std::size(CommonOperandNames), + "Local index out of bounds in common operands"); + return CommonOperandNames[LocalIndex].str(); + } + + // Handle physical registers section + if (Pos < Layout.VirtRegBase) { + unsigned LocalIndex = + getLocalIndex(Pos, Layout.PhyRegBase, RegisterOperandNames.size(), + "Local index out of bounds in physical registers"); + return "PhyReg_" + RegisterOperandNames[LocalIndex]; + } + + // Handle virtual registers section + unsigned LocalIndex = + getLocalIndex(Pos, Layout.VirtRegBase, RegisterOperandNames.size(), + "Local index out of bounds in virtual registers"); + return "VirtReg_" + RegisterOperandNames[LocalIndex]; } -void MIRVocabulary::generateStorage(const VocabMap &OpcodeMap) { +void MIRVocabulary::generateStorage(const VocabMap &OpcodeMap, + const VocabMap &CommonOperandsMap, + const VocabMap &PhyRegMap, + const VocabMap &VirtRegMap) { // Helper for handling missing entities in the vocabulary. // Currently, we use a zero vector. In the future, we will throw an error to @@ -140,14 +236,14 @@ void MIRVocabulary::generateStorage(const VocabMap &OpcodeMap) { // Initialize opcode embeddings section unsigned EmbeddingDim = OpcodeMap.begin()->second.size(); - std::vector<Embedding> OpcodeEmbeddings(Layout.OperandBase, + std::vector<Embedding> OpcodeEmbeddings(Layout.CommonOperandBase, Embedding(EmbeddingDim)); // Populate opcode embeddings using canonical mapping for (auto COpcodeName : UniqueBaseOpcodeNames) { if (auto It = OpcodeMap.find(COpcodeName); It != OpcodeMap.end()) { auto COpcodeIndex = getCanonicalIndexForBaseName(COpcodeName); - assert(COpcodeIndex < Layout.OperandBase && + assert(COpcodeIndex < Layout.CommonOperandBase && "Canonical index out of bounds"); OpcodeEmbeddings[COpcodeIndex] = It->second; } else { @@ -155,8 +251,39 @@ void MIRVocabulary::generateStorage(const VocabMap &OpcodeMap) { } } - // TODO: Add operand/argument embeddings as additional sections - // This will require extending the vocabulary format and layout + // Initialize common operand embeddings section + std::vector<Embedding> CommonOperandEmbeddings(std::size(CommonOperandNames), + Embedding(EmbeddingDim)); + unsigned OperandIndex = 0; + for (const auto &CommonOperandName : CommonOperandNames) { + if (auto It = CommonOperandsMap.find(CommonOperandName.str()); + It != CommonOperandsMap.end()) { + CommonOperandEmbeddings[OperandIndex] = It->second; + } else { + handleMissingEntity(CommonOperandName); + } + ++OperandIndex; + } + + // Helper lambda for creating register operand embeddings + auto createRegisterEmbeddings = [&](const VocabMap &RegMap) { + std::vector<Embedding> RegEmbeddings(TRI.getNumRegClasses(), + Embedding(EmbeddingDim)); + unsigned RegOperandIndex = 0; + for (const auto &RegOperandName : RegisterOperandNames) { + if (auto It = RegMap.find(RegOperandName); It != RegMap.end()) + RegEmbeddings[RegOperandIndex] = It->second; + else + handleMissingEntity(RegOperandName); + ++RegOperandIndex; + } + return RegEmbeddings; + }; + + // Initialize register operand embeddings sections + std::vector<Embedding> PhyRegEmbeddings = createRegisterEmbeddings(PhyRegMap); + std::vector<Embedding> VirtRegEmbeddings = + createRegisterEmbeddings(VirtRegMap); // Scale the vocabulary sections based on the provided weights auto scaleVocabSection = [](std::vector<Embedding> &Embeddings, @@ -165,9 +292,20 @@ void MIRVocabulary::generateStorage(const VocabMap &OpcodeMap) { Embedding *= Weight; }; scaleVocabSection(OpcodeEmbeddings, OpcWeight); - - std::vector<std::vector<Embedding>> Sections(1); - Sections[0] = std::move(OpcodeEmbeddings); + scaleVocabSection(CommonOperandEmbeddings, CommonOperandWeight); + scaleVocabSection(PhyRegEmbeddings, RegOperandWeight); + scaleVocabSection(VirtRegEmbeddings, RegOperandWeight); + + std::vector<std::vector<Embedding>> Sections( + static_cast<unsigned>(Section::MaxSections)); + Sections[static_cast<unsigned>(Section::Opcodes)] = + std::move(OpcodeEmbeddings); + Sections[static_cast<unsigned>(Section::CommonOperands)] = + std::move(CommonOperandEmbeddings); + Sections[static_cast<unsigned>(Section::PhyRegisters)] = + std::move(PhyRegEmbeddings); + Sections[static_cast<unsigned>(Section::VirtRegisters)] = + std::move(VirtRegEmbeddings); Storage = ir2vec::VocabStorage(std::move(Sections)); } @@ -188,24 +326,130 @@ void MIRVocabulary::buildCanonicalOpcodeMapping() { << " unique base opcodes\n"); } +void MIRVocabulary::buildRegisterOperandMapping() { + // Check if already built + if (!RegisterOperandNames.empty()) + return; + + for (unsigned RC = 0; RC < TRI.getNumRegClasses(); ++RC) { + const TargetRegisterClass *RegClass = TRI.getRegClass(RC); + if (!RegClass) + continue; + + // Get the register class name + StringRef ClassName = TRI.getRegClassName(RegClass); + RegisterOperandNames.push_back(ClassName.str()); + } +} + +unsigned MIRVocabulary::getCommonOperandIndex( + MachineOperand::MachineOperandType OperandType) const { + assert(OperandType != MachineOperand::MO_Register && + "Expected non-register operand type"); + assert(OperandType > MachineOperand::MO_Register && + OperandType < MachineOperand::MO_Last && "Operand type out of bounds"); + return static_cast<unsigned>(OperandType) - 1; +} + +unsigned MIRVocabulary::getRegisterOperandIndex(Register Reg) const { + assert(!RegisterOperandNames.empty() && "Register operand mapping not built"); + assert(Reg.isValid() && "Invalid register; not expected here"); + assert((Reg.isPhysical() || Reg.isVirtual()) && + "Expected a physical or virtual register"); + + const TargetRegisterClass *RegClass = nullptr; + + // For physical registers, use TRI to get minimal register class as a + // physical register can belong to multiple classes. For virtual + // registers, use MRI to uniquely identify the assigned register class. + if (Reg.isPhysical()) + RegClass = TRI.getMinimalPhysRegClass(Reg); + else + RegClass = MRI.getRegClass(Reg); + + if (RegClass) + return RegClass->getID(); + // Fallback for registers without a class (shouldn't happen) + llvm_unreachable("Register operand without a valid register class"); + return 0; +} + +Expected<MIRVocabulary> MIRVocabulary::createDummyVocabForTest( + const TargetInstrInfo &TII, const TargetRegisterInfo &TRI, + const MachineRegisterInfo &MRI, unsigned Dim) { + assert(Dim > 0 && "Dimension must be greater than zero"); + + float DummyVal = 0.1f; + + VocabMap DummyOpcMap, DummyOperandMap, DummyPhyRegMap, DummyVirtRegMap; + + // Process opcodes directly without creating temporary vocabulary + for (unsigned Opcode = 0; Opcode < TII.getNumOpcodes(); ++Opcode) { + std::string BaseOpcode = extractBaseOpcodeName(TII.getName(Opcode)); + if (DummyOpcMap.count(BaseOpcode) == 0) { // Only add if not already present + DummyOpcMap[BaseOpcode] = Embedding(Dim, DummyVal); + DummyVal += 0.1f; + } + } + + // Add common operands + for (const auto &CommonOperandName : CommonOperandNames) { + DummyOperandMap[CommonOperandName.str()] = Embedding(Dim, DummyVal); + DummyVal += 0.1f; + } + + // Process register classes directly + for (unsigned RC = 0; RC < TRI.getNumRegClasses(); ++RC) { + const TargetRegisterClass *RegClass = TRI.getRegClass(RC); + if (!RegClass) + continue; + + std::string ClassName = TRI.getRegClassName(RegClass); + DummyPhyRegMap[ClassName] = Embedding(Dim, DummyVal); + DummyVirtRegMap[ClassName] = Embedding(Dim, DummyVal); + DummyVal += 0.1f; + } + + // Create vocabulary directly without temporary instance + return MIRVocabulary::create( + std::move(DummyOpcMap), std::move(DummyOperandMap), + std::move(DummyPhyRegMap), std::move(DummyVirtRegMap), TII, TRI, MRI); +} + //===----------------------------------------------------------------------===// -// MIR2VecVocabLegacyAnalysis Implementation +// MIR2VecVocabProvider and MIR2VecVocabLegacyAnalysis //===----------------------------------------------------------------------===// -char MIR2VecVocabLegacyAnalysis::ID = 0; -INITIALIZE_PASS_BEGIN(MIR2VecVocabLegacyAnalysis, "mir2vec-vocab-analysis", - "MIR2Vec Vocabulary Analysis", false, true) -INITIALIZE_PASS_DEPENDENCY(MachineModuleInfoWrapperPass) -INITIALIZE_PASS_END(MIR2VecVocabLegacyAnalysis, "mir2vec-vocab-analysis", - "MIR2Vec Vocabulary Analysis", false, true) +Expected<mir2vec::MIRVocabulary> +MIR2VecVocabProvider::getVocabulary(const Module &M) { + VocabMap OpcVocab, CommonOperandVocab, PhyRegVocabMap, VirtRegVocabMap; -StringRef MIR2VecVocabLegacyAnalysis::getPassName() const { - return "MIR2Vec Vocabulary Analysis"; + if (Error Err = readVocabulary(OpcVocab, CommonOperandVocab, PhyRegVocabMap, + VirtRegVocabMap)) + return std::move(Err); + + for (const auto &F : M) { + if (F.isDeclaration()) + continue; + + if (auto *MF = MMI.getMachineFunction(F)) { + auto &Subtarget = MF->getSubtarget(); + if (const auto *TII = Subtarget.getInstrInfo()) + if (const auto *TRI = Subtarget.getRegisterInfo()) + return mir2vec::MIRVocabulary::create( + std::move(OpcVocab), std::move(CommonOperandVocab), + std::move(PhyRegVocabMap), std::move(VirtRegVocabMap), *TII, *TRI, + MF->getRegInfo()); + } + } + return createStringError(errc::invalid_argument, + "No machine functions found in module"); } -Error MIR2VecVocabLegacyAnalysis::readVocabulary() { - // TODO: Extend vocabulary format to support multiple sections - // (opcodes, operands, etc.) similar to IR2Vec structure +Error MIR2VecVocabProvider::readVocabulary(VocabMap &OpcodeVocab, + VocabMap &CommonOperandVocab, + VocabMap &PhyRegVocabMap, + VocabMap &VirtRegVocabMap) { if (VocabFile.empty()) return createStringError( errc::invalid_argument, @@ -222,43 +466,122 @@ Error MIR2VecVocabLegacyAnalysis::readVocabulary() { if (!ParsedVocabValue) return ParsedVocabValue.takeError(); - unsigned Dim = 0; + unsigned OpcodeDim = 0, CommonOperandDim = 0, PhyRegOperandDim = 0, + VirtRegOperandDim = 0; + if (auto Err = ir2vec::VocabStorage::parseVocabSection( + "Opcodes", *ParsedVocabValue, OpcodeVocab, OpcodeDim)) + return Err; + + if (auto Err = ir2vec::VocabStorage::parseVocabSection( + "CommonOperands", *ParsedVocabValue, CommonOperandVocab, + CommonOperandDim)) + return Err; + + if (auto Err = ir2vec::VocabStorage::parseVocabSection( + "PhysicalRegisters", *ParsedVocabValue, PhyRegVocabMap, + PhyRegOperandDim)) + return Err; + if (auto Err = ir2vec::VocabStorage::parseVocabSection( - "entities", *ParsedVocabValue, StrVocabMap, Dim)) + "VirtualRegisters", *ParsedVocabValue, VirtRegVocabMap, + VirtRegOperandDim)) return Err; + // All sections must have the same embedding dimension + if (!(OpcodeDim == CommonOperandDim && CommonOperandDim == PhyRegOperandDim && + PhyRegOperandDim == VirtRegOperandDim)) { + return createStringError( + errc::illegal_byte_sequence, + "MIR2Vec vocabulary sections have different dimensions"); + } + return Error::success(); } -Expected<mir2vec::MIRVocabulary> -MIR2VecVocabLegacyAnalysis::getMIR2VecVocabulary(const Module &M) { - if (StrVocabMap.empty()) { - if (Error Err = readVocabulary()) { - return std::move(Err); - } +char MIR2VecVocabLegacyAnalysis::ID = 0; +INITIALIZE_PASS_BEGIN(MIR2VecVocabLegacyAnalysis, "mir2vec-vocab-analysis", + "MIR2Vec Vocabulary Analysis", false, true) +INITIALIZE_PASS_DEPENDENCY(MachineModuleInfoWrapperPass) +INITIALIZE_PASS_END(MIR2VecVocabLegacyAnalysis, "mir2vec-vocab-analysis", + "MIR2Vec Vocabulary Analysis", false, true) + +StringRef MIR2VecVocabLegacyAnalysis::getPassName() const { + return "MIR2Vec Vocabulary Analysis"; +} + +//===----------------------------------------------------------------------===// +// MIREmbedder and its subclasses +//===----------------------------------------------------------------------===// + +std::unique_ptr<MIREmbedder> MIREmbedder::create(MIR2VecKind Mode, + const MachineFunction &MF, + const MIRVocabulary &Vocab) { + switch (Mode) { + case MIR2VecKind::Symbolic: + return std::make_unique<SymbolicMIREmbedder>(MF, Vocab); } + return nullptr; +} - // Get machine module info to access machine functions and target info - MachineModuleInfo &MMI = getAnalysis<MachineModuleInfoWrapperPass>().getMMI(); +Embedding MIREmbedder::computeEmbeddings(const MachineBasicBlock &MBB) const { + Embedding MBBVector(Dimension, 0); - // Find first available machine function to get target instruction info - for (const auto &F : M) { - if (F.isDeclaration()) - continue; + // Get instruction info for opcode name resolution + const auto &Subtarget = MF.getSubtarget(); + const auto *TII = Subtarget.getInstrInfo(); + if (!TII) { + MF.getFunction().getContext().emitError( + "MIR2Vec: No TargetInstrInfo available; cannot compute embeddings"); + return MBBVector; + } - if (auto *MF = MMI.getMachineFunction(F)) { - const TargetInstrInfo *TII = MF->getSubtarget().getInstrInfo(); - return mir2vec::MIRVocabulary::create(std::move(StrVocabMap), *TII); - } + // Process each machine instruction in the basic block + for (const auto &MI : MBB) { + // Skip debug instructions and other metadata + if (MI.isDebugInstr()) + continue; + MBBVector += computeEmbeddings(MI); } - // No machine functions available - return error - return createStringError(errc::invalid_argument, - "No machine functions found in module"); + return MBBVector; +} + +Embedding MIREmbedder::computeEmbeddings() const { + Embedding MFuncVector(Dimension, 0); + + // Consider all reachable machine basic blocks in the function + for (const auto *MBB : depth_first(&MF)) + MFuncVector += computeEmbeddings(*MBB); + return MFuncVector; +} + +SymbolicMIREmbedder::SymbolicMIREmbedder(const MachineFunction &MF, + const MIRVocabulary &Vocab) + : MIREmbedder(MF, Vocab) {} + +std::unique_ptr<SymbolicMIREmbedder> +SymbolicMIREmbedder::create(const MachineFunction &MF, + const MIRVocabulary &Vocab) { + return std::make_unique<SymbolicMIREmbedder>(MF, Vocab); +} + +Embedding SymbolicMIREmbedder::computeEmbeddings(const MachineInstr &MI) const { + // Skip debug instructions and other metadata + if (MI.isDebugInstr()) + return Embedding(Dimension, 0); + + // Opcode embedding + Embedding InstructionEmbedding = Vocab[MI.getOpcode()]; + + // Add operand contributions + for (const MachineOperand &MO : MI.operands()) + InstructionEmbedding += Vocab[MO]; + + return InstructionEmbedding; } //===----------------------------------------------------------------------===// -// Printer Passes Implementation +// Printer Passes //===----------------------------------------------------------------------===// char MIR2VecVocabPrinterLegacyPass::ID = 0; @@ -297,3 +620,56 @@ MachineFunctionPass * llvm::createMIR2VecVocabPrinterLegacyPass(raw_ostream &OS) { return new MIR2VecVocabPrinterLegacyPass(OS); } + +char MIR2VecPrinterLegacyPass::ID = 0; +INITIALIZE_PASS_BEGIN(MIR2VecPrinterLegacyPass, "print-mir2vec", + "MIR2Vec Embedder Printer Pass", false, true) +INITIALIZE_PASS_DEPENDENCY(MIR2VecVocabLegacyAnalysis) +INITIALIZE_PASS_DEPENDENCY(MachineModuleInfoWrapperPass) +INITIALIZE_PASS_END(MIR2VecPrinterLegacyPass, "print-mir2vec", + "MIR2Vec Embedder Printer Pass", false, true) + +bool MIR2VecPrinterLegacyPass::runOnMachineFunction(MachineFunction &MF) { + auto &Analysis = getAnalysis<MIR2VecVocabLegacyAnalysis>(); + auto VocabOrErr = + Analysis.getMIR2VecVocabulary(*MF.getFunction().getParent()); + assert(VocabOrErr && "Failed to get MIR2Vec vocabulary"); + auto &MIRVocab = *VocabOrErr; + + auto Emb = mir2vec::MIREmbedder::create(MIR2VecEmbeddingKind, MF, MIRVocab); + if (!Emb) { + OS << "Error creating MIR2Vec embeddings for function " << MF.getName() + << "\n"; + return false; + } + + OS << "MIR2Vec embeddings for machine function " << MF.getName() << ":\n"; + OS << "Machine Function vector: "; + Emb->getMFunctionVector().print(OS); + + OS << "Machine basic block vectors:\n"; + for (const MachineBasicBlock &MBB : MF) { + OS << "Machine basic block: " << MBB.getFullName() << ":\n"; + Emb->getMBBVector(MBB).print(OS); + } + + OS << "Machine instruction vectors:\n"; + for (const MachineBasicBlock &MBB : MF) { + for (const MachineInstr &MI : MBB) { + // Skip debug instructions as they are not + // embedded + if (MI.isDebugInstr()) + continue; + + OS << "Machine instruction: "; + MI.print(OS); + Emb->getMInstVector(MI).print(OS); + } + } + + return false; +} + +MachineFunctionPass *llvm::createMIR2VecPrinterLegacyPass(raw_ostream &OS) { + return new MIR2VecPrinterLegacyPass(OS); +} diff --git a/llvm/lib/CodeGen/MachineBasicBlock.cpp b/llvm/lib/CodeGen/MachineBasicBlock.cpp index 1cb57a4..ba0b025 100644 --- a/llvm/lib/CodeGen/MachineBasicBlock.cpp +++ b/llvm/lib/CodeGen/MachineBasicBlock.cpp @@ -1137,7 +1137,7 @@ public: MF.setDelegate(this); } - ~SlotIndexUpdateDelegate() { + ~SlotIndexUpdateDelegate() override { MF.resetDelegate(this); for (auto MI : Insertions) Indexes->insertMachineInstrInMaps(*MI); diff --git a/llvm/lib/CodeGen/MachineBlockHashInfo.cpp b/llvm/lib/CodeGen/MachineBlockHashInfo.cpp new file mode 100644 index 0000000..c4d9c0f --- /dev/null +++ b/llvm/lib/CodeGen/MachineBlockHashInfo.cpp @@ -0,0 +1,115 @@ +//===- llvm/CodeGen/MachineBlockHashInfo.cpp---------------------*- C++ -*-===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// Compute the hashes of basic blocks. +// +//===----------------------------------------------------------------------===// + +#include "llvm/CodeGen/MachineBlockHashInfo.h" +#include "llvm/CodeGen/Passes.h" +#include "llvm/InitializePasses.h" +#include "llvm/Target/TargetMachine.h" + +using namespace llvm; + +uint64_t hashBlock(const MachineBasicBlock &MBB, bool HashOperands) { + uint64_t Hash = 0; + for (const MachineInstr &MI : MBB) { + if (MI.isMetaInstruction() || MI.isTerminator()) + continue; + Hash = hashing::detail::hash_16_bytes(Hash, MI.getOpcode()); + if (HashOperands) { + for (unsigned i = 0; i < MI.getNumOperands(); i++) { + Hash = + hashing::detail::hash_16_bytes(Hash, hash_value(MI.getOperand(i))); + } + } + } + return Hash; +} + +/// Fold a 64-bit integer to a 16-bit one. +uint16_t fold_64_to_16(const uint64_t Value) { + uint16_t Res = static_cast<uint16_t>(Value); + Res ^= static_cast<uint16_t>(Value >> 16); + Res ^= static_cast<uint16_t>(Value >> 32); + Res ^= static_cast<uint16_t>(Value >> 48); + return Res; +} + +INITIALIZE_PASS(MachineBlockHashInfo, "machine-block-hash", + "Machine Block Hash Analysis", true, true) + +char MachineBlockHashInfo::ID = 0; + +MachineBlockHashInfo::MachineBlockHashInfo() : MachineFunctionPass(ID) { + initializeMachineBlockHashInfoPass(*PassRegistry::getPassRegistry()); +} + +void MachineBlockHashInfo::getAnalysisUsage(AnalysisUsage &AU) const { + AU.setPreservesAll(); + MachineFunctionPass::getAnalysisUsage(AU); +} + +struct CollectHashInfo { + uint64_t Offset; + uint64_t OpcodeHash; + uint64_t InstrHash; + uint64_t NeighborHash; +}; + +bool MachineBlockHashInfo::runOnMachineFunction(MachineFunction &F) { + DenseMap<const MachineBasicBlock *, CollectHashInfo> HashInfos; + uint16_t Offset = 0; + // Initialize hash components + for (const MachineBasicBlock &MBB : F) { + // offset of the machine basic block + HashInfos[&MBB].Offset = Offset; + Offset += MBB.size(); + // Hashing opcodes + HashInfos[&MBB].OpcodeHash = hashBlock(MBB, /*HashOperands=*/false); + // Hash complete instructions + HashInfos[&MBB].InstrHash = hashBlock(MBB, /*HashOperands=*/true); + } + + // Initialize neighbor hash + for (const MachineBasicBlock &MBB : F) { + uint64_t Hash = HashInfos[&MBB].OpcodeHash; + // Append hashes of successors + for (const MachineBasicBlock *SuccMBB : MBB.successors()) { + uint64_t SuccHash = HashInfos[SuccMBB].OpcodeHash; + Hash = hashing::detail::hash_16_bytes(Hash, SuccHash); + } + // Append hashes of predecessors + for (const MachineBasicBlock *PredMBB : MBB.predecessors()) { + uint64_t PredHash = HashInfos[PredMBB].OpcodeHash; + Hash = hashing::detail::hash_16_bytes(Hash, PredHash); + } + HashInfos[&MBB].NeighborHash = Hash; + } + + // Assign hashes + for (const MachineBasicBlock &MBB : F) { + const auto &HashInfo = HashInfos[&MBB]; + BlendedBlockHash BlendedHash(fold_64_to_16(HashInfo.Offset), + fold_64_to_16(HashInfo.OpcodeHash), + fold_64_to_16(HashInfo.InstrHash), + fold_64_to_16(HashInfo.NeighborHash)); + MBBHashInfo[&MBB] = BlendedHash.combine(); + } + + return false; +} + +uint64_t MachineBlockHashInfo::getMBBHash(const MachineBasicBlock &MBB) { + return MBBHashInfo[&MBB]; +} + +MachineFunctionPass *llvm::createMachineBlockHashInfoPass() { + return new MachineBlockHashInfo(); +} diff --git a/llvm/lib/CodeGen/MachineLICM.cpp b/llvm/lib/CodeGen/MachineLICM.cpp index 7acddff..729e73c 100644 --- a/llvm/lib/CodeGen/MachineLICM.cpp +++ b/llvm/lib/CodeGen/MachineLICM.cpp @@ -932,12 +932,11 @@ void MachineLICMImpl::InitRegPressure(MachineBasicBlock *BB) { void MachineLICMImpl::UpdateRegPressure(const MachineInstr *MI, bool ConsiderUnseenAsDef) { auto Cost = calcRegisterCost(MI, /*ConsiderSeen=*/true, ConsiderUnseenAsDef); - for (const auto &RPIdAndCost : Cost) { - unsigned Class = RPIdAndCost.first; - if (static_cast<int>(RegPressure[Class]) < -RPIdAndCost.second) + for (const auto &[Class, Weight] : Cost) { + if (static_cast<int>(RegPressure[Class]) < -Weight) RegPressure[Class] = 0; else - RegPressure[Class] += RPIdAndCost.second; + RegPressure[Class] += Weight; } } @@ -1215,11 +1214,10 @@ bool MachineLICMImpl::IsCheapInstruction(MachineInstr &MI) const { /// given cost matrix can cause high register pressure. bool MachineLICMImpl::CanCauseHighRegPressure( const SmallDenseMap<unsigned, int> &Cost, bool CheapInstr) { - for (const auto &RPIdAndCost : Cost) { - if (RPIdAndCost.second <= 0) + for (const auto &[Class, Weight] : Cost) { + if (Weight <= 0) continue; - unsigned Class = RPIdAndCost.first; int Limit = RegLimit[Class]; // Don't hoist cheap instructions if they would increase register pressure, @@ -1228,7 +1226,7 @@ bool MachineLICMImpl::CanCauseHighRegPressure( return true; for (const auto &RP : BackTrace) - if (static_cast<int>(RP[Class]) + RPIdAndCost.second >= Limit) + if (static_cast<int>(RP[Class]) + Weight >= Limit) return true; } @@ -1246,8 +1244,8 @@ void MachineLICMImpl::UpdateBackTraceRegPressure(const MachineInstr *MI) { // Update register pressure of blocks from loop header to current block. for (auto &RP : BackTrace) - for (const auto &RPIdAndCost : Cost) - RP[RPIdAndCost.first] += RPIdAndCost.second; + for (const auto &[Class, Weight] : Cost) + RP[Class] += Weight; } /// Return true if it is potentially profitable to hoist the given loop diff --git a/llvm/lib/CodeGen/PeepholeOptimizer.cpp b/llvm/lib/CodeGen/PeepholeOptimizer.cpp index e1d39d6..493d8df 100644 --- a/llvm/lib/CodeGen/PeepholeOptimizer.cpp +++ b/llvm/lib/CodeGen/PeepholeOptimizer.cpp @@ -196,7 +196,7 @@ public: CopyRewriter(MachineInstr &MI) : Rewriter(MI) { assert(MI.isCopy() && "Expected copy instruction"); } - virtual ~CopyRewriter() = default; + ~CopyRewriter() override = default; bool getNextRewritableSource(RegSubRegPair &Src, RegSubRegPair &Dst) override { diff --git a/llvm/lib/CodeGen/PreISelIntrinsicLowering.cpp b/llvm/lib/CodeGen/PreISelIntrinsicLowering.cpp index f54e2f2..620d3d3 100644 --- a/llvm/lib/CodeGen/PreISelIntrinsicLowering.cpp +++ b/llvm/lib/CodeGen/PreISelIntrinsicLowering.cpp @@ -593,7 +593,7 @@ bool PreISelIntrinsicLowering::lowerIntrinsics(Module &M) const { case Intrinsic::log: Changed |= forEachCall(F, [&](CallInst *CI) { Type *Ty = CI->getArgOperand(0)->getType(); - if (!isa<ScalableVectorType>(Ty)) + if (!TM || !isa<ScalableVectorType>(Ty)) return false; const TargetLowering *TL = TM->getSubtargetImpl(F)->getTargetLowering(); unsigned Op = TL->IntrinsicIDToISD(F.getIntrinsicID()); diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp index 310d35d..d2ea652 100644 --- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp @@ -2476,16 +2476,17 @@ static bool canFoldInAddressingMode(SDNode *N, SDNode *Use, SelectionDAG &DAG, /// masked vector operation if the target supports it. static SDValue foldSelectWithIdentityConstant(SDNode *N, SelectionDAG &DAG, bool ShouldCommuteOperands) { - // Match a select as operand 1. The identity constant that we are looking for - // is only valid as operand 1 of a non-commutative binop. SDValue N0 = N->getOperand(0); SDValue N1 = N->getOperand(1); + + // Match a select as operand 1. The identity constant that we are looking for + // is only valid as operand 1 of a non-commutative binop. if (ShouldCommuteOperands) std::swap(N0, N1); - unsigned SelOpcode = N1.getOpcode(); - if ((SelOpcode != ISD::VSELECT && SelOpcode != ISD::SELECT) || - !N1.hasOneUse()) + SDValue Cond, TVal, FVal; + if (!sd_match(N1, m_OneUse(m_SelectLike(m_Value(Cond), m_Value(TVal), + m_Value(FVal))))) return SDValue(); // We can't hoist all instructions because of immediate UB (not speculatable). @@ -2493,11 +2494,9 @@ static SDValue foldSelectWithIdentityConstant(SDNode *N, SelectionDAG &DAG, if (!DAG.isSafeToSpeculativelyExecuteNode(N)) return SDValue(); + unsigned SelOpcode = N1.getOpcode(); unsigned Opcode = N->getOpcode(); EVT VT = N->getValueType(0); - SDValue Cond = N1.getOperand(0); - SDValue TVal = N1.getOperand(1); - SDValue FVal = N1.getOperand(2); const TargetLowering &TLI = DAG.getTargetLoweringInfo(); // This transform increases uses of N0, so freeze it to be safe. @@ -13856,12 +13855,11 @@ static SDValue tryToFoldExtendSelectLoad(SDNode *N, const TargetLowering &TLI, Opcode == ISD::ANY_EXTEND) && "Expected EXTEND dag node in input!"); - if (!(N0->getOpcode() == ISD::SELECT || N0->getOpcode() == ISD::VSELECT) || - !N0.hasOneUse()) + SDValue Cond, Op1, Op2; + if (!sd_match(N0, m_OneUse(m_SelectLike(m_Value(Cond), m_Value(Op1), + m_Value(Op2))))) return SDValue(); - SDValue Op1 = N0->getOperand(1); - SDValue Op2 = N0->getOperand(2); if (!isCompatibleLoad(Op1, Opcode) || !isCompatibleLoad(Op2, Opcode)) return SDValue(); @@ -13883,7 +13881,7 @@ static SDValue tryToFoldExtendSelectLoad(SDNode *N, const TargetLowering &TLI, SDValue Ext1 = DAG.getNode(Opcode, DL, VT, Op1); SDValue Ext2 = DAG.getNode(Opcode, DL, VT, Op2); - return DAG.getSelect(DL, VT, N0->getOperand(0), Ext1, Ext2); + return DAG.getSelect(DL, VT, Cond, Ext1, Ext2); } /// Try to fold a sext/zext/aext dag node into a ConstantSDNode or @@ -17462,8 +17460,8 @@ SDValue DAGCombiner::visitFSUBForFMACombine(SDNode *N) { // fold (fsub (fpext (fneg (fmul, x, y))), z) // -> (fneg (fma (fpext x), (fpext y), z)) // Note: This could be removed with appropriate canonicalization of the - // input expression into (fneg (fadd (fpext (fmul, x, y)), z). However, the - // orthogonal flags -fp-contract=fast and -enable-unsafe-fp-math prevent + // input expression into (fneg (fadd (fpext (fmul, x, y)), z)). However, the + // command line flag -fp-contract=fast and fast-math flag contract prevent // from implementing the canonicalization in visitFSUB. if (matcher.match(N0, ISD::FP_EXTEND)) { SDValue N00 = N0.getOperand(0); @@ -17487,7 +17485,7 @@ SDValue DAGCombiner::visitFSUBForFMACombine(SDNode *N) { // -> (fneg (fma (fpext x)), (fpext y), z) // Note: This could be removed with appropriate canonicalization of the // input expression into (fneg (fadd (fpext (fmul, x, y)), z). However, the - // orthogonal flags -fp-contract=fast and -enable-unsafe-fp-math prevent + // command line flag -fp-contract=fast and fast-math flag contract prevent // from implementing the canonicalization in visitFSUB. if (matcher.match(N0, ISD::FNEG)) { SDValue N00 = N0.getOperand(0); @@ -29620,13 +29618,14 @@ static SDValue takeInexpensiveLog2(SelectionDAG &DAG, const SDLoc &DL, EVT VT, } // c ? X : Y -> c ? Log2(X) : Log2(Y) - if ((Op.getOpcode() == ISD::SELECT || Op.getOpcode() == ISD::VSELECT) && - Op.hasOneUse()) { - if (SDValue LogX = takeInexpensiveLog2(DAG, DL, VT, Op.getOperand(1), - Depth + 1, AssumeNonZero)) - if (SDValue LogY = takeInexpensiveLog2(DAG, DL, VT, Op.getOperand(2), - Depth + 1, AssumeNonZero)) - return DAG.getSelect(DL, VT, Op.getOperand(0), LogX, LogY); + SDValue Cond, TVal, FVal; + if (sd_match(Op, m_OneUse(m_SelectLike(m_Value(Cond), m_Value(TVal), + m_Value(FVal))))) { + if (SDValue LogX = + takeInexpensiveLog2(DAG, DL, VT, TVal, Depth + 1, AssumeNonZero)) + if (SDValue LogY = + takeInexpensiveLog2(DAG, DL, VT, FVal, Depth + 1, AssumeNonZero)) + return DAG.getSelect(DL, VT, Cond, LogX, LogY); } // log2(umin(X, Y)) -> umin(log2(X), log2(Y)) diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp index 437d0f4..bf1abfe 100644 --- a/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp @@ -3765,6 +3765,8 @@ bool DAGTypeLegalizer::SoftPromoteHalfOperand(SDNode *N, unsigned OpNo) { case ISD::FP_TO_UINT: case ISD::LRINT: case ISD::LLRINT: + case ISD::LROUND: + case ISD::LLROUND: Res = SoftPromoteHalfOp_Op0WithStrict(N); break; case ISD::FP_TO_SINT_SAT: diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp index 88a4a8b..b1776ea 100644 --- a/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp @@ -429,7 +429,20 @@ SDValue DAGTypeLegalizer::PromoteIntRes_Atomic0(AtomicSDNode *N) { } SDValue DAGTypeLegalizer::PromoteIntRes_Atomic1(AtomicSDNode *N) { - SDValue Op2 = GetPromotedInteger(N->getOperand(2)); + SDValue Op2 = N->getOperand(2); + switch (TLI.getExtendForAtomicRMWArg(N->getOpcode())) { + case ISD::SIGN_EXTEND: + Op2 = SExtPromotedInteger(Op2); + break; + case ISD::ZERO_EXTEND: + Op2 = ZExtPromotedInteger(Op2); + break; + case ISD::ANY_EXTEND: + Op2 = GetPromotedInteger(Op2); + break; + default: + llvm_unreachable("Invalid atomic op extension"); + } SDValue Res = DAG.getAtomic(N->getOpcode(), SDLoc(N), N->getMemoryVT(), N->getChain(), N->getBasePtr(), diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h b/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h index 603dc34..9656a30 100644 --- a/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h +++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h @@ -890,6 +890,7 @@ private: SDValue ScalarizeVecRes_UnaryOpWithExtraInput(SDNode *N); SDValue ScalarizeVecRes_INSERT_VECTOR_ELT(SDNode *N); SDValue ScalarizeVecRes_LOAD(LoadSDNode *N); + SDValue ScalarizeVecRes_ATOMIC_LOAD(AtomicSDNode *N); SDValue ScalarizeVecRes_SCALAR_TO_VECTOR(SDNode *N); SDValue ScalarizeVecRes_VSELECT(SDNode *N); SDValue ScalarizeVecRes_SELECT(SDNode *N); diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp index 3b5f83f..bb4a8d9 100644 --- a/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp @@ -69,6 +69,9 @@ void DAGTypeLegalizer::ScalarizeVectorResult(SDNode *N, unsigned ResNo) { R = ScalarizeVecRes_UnaryOpWithExtraInput(N); break; case ISD::INSERT_VECTOR_ELT: R = ScalarizeVecRes_INSERT_VECTOR_ELT(N); break; + case ISD::ATOMIC_LOAD: + R = ScalarizeVecRes_ATOMIC_LOAD(cast<AtomicSDNode>(N)); + break; case ISD::LOAD: R = ScalarizeVecRes_LOAD(cast<LoadSDNode>(N));break; case ISD::SCALAR_TO_VECTOR: R = ScalarizeVecRes_SCALAR_TO_VECTOR(N); break; case ISD::SIGN_EXTEND_INREG: R = ScalarizeVecRes_InregOp(N); break; @@ -475,6 +478,18 @@ SDValue DAGTypeLegalizer::ScalarizeVecRes_INSERT_VECTOR_ELT(SDNode *N) { return Op; } +SDValue DAGTypeLegalizer::ScalarizeVecRes_ATOMIC_LOAD(AtomicSDNode *N) { + SDValue Result = DAG.getAtomicLoad( + N->getExtensionType(), SDLoc(N), N->getMemoryVT().getVectorElementType(), + N->getValueType(0).getVectorElementType(), N->getChain(), N->getBasePtr(), + N->getMemOperand()); + + // Legalize the chain result - switch anything that used the old chain to + // use the new one. + ReplaceValueWith(SDValue(N, 1), Result.getValue(1)); + return Result; +} + SDValue DAGTypeLegalizer::ScalarizeVecRes_LOAD(LoadSDNode *N) { assert(N->isUnindexed() && "Indexed vector load?"); diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp index 90edaf3..379242e 100644 --- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp @@ -8620,7 +8620,10 @@ SDValue SelectionDAG::getMemBasePlusOffset(SDValue Ptr, SDValue Offset, if (TLI->shouldPreservePtrArith(this->getMachineFunction().getFunction(), BasePtrVT)) return getNode(ISD::PTRADD, DL, BasePtrVT, Ptr, Offset, Flags); - return getNode(ISD::ADD, DL, BasePtrVT, Ptr, Offset, Flags); + // InBounds only applies to PTRADD, don't set it if we generate ADD. + SDNodeFlags AddFlags = Flags; + AddFlags.setInBounds(false); + return getNode(ISD::ADD, DL, BasePtrVT, Ptr, Offset, AddFlags); } /// Returns true if memcpy source is constant data. diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp index 20a0efd..dee0909 100644 --- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp @@ -1162,6 +1162,43 @@ SDValue SelectionDAGBuilder::getMemoryRoot() { return updateRoot(PendingLoads); } +SDValue SelectionDAGBuilder::getFPOperationRoot(fp::ExceptionBehavior EB) { + // If the new exception behavior differs from that of the pending + // ones, chain up them and update the root. + switch (EB) { + case fp::ExceptionBehavior::ebMayTrap: + case fp::ExceptionBehavior::ebIgnore: + // Floating-point exceptions produced by such operations are not intended + // to be observed, so the sequence of these operations does not need to be + // preserved. + // + // They however must not be mixed with the instructions that have strict + // exception behavior. Placing an operation with 'ebIgnore' behavior between + // 'ebStrict' operations could distort the observed exception behavior. + if (!PendingConstrainedFPStrict.empty()) { + assert(PendingConstrainedFP.empty()); + updateRoot(PendingConstrainedFPStrict); + } + break; + case fp::ExceptionBehavior::ebStrict: + // Floating-point exception produced by these operations may be observed, so + // they must be correctly chained. If trapping on FP exceptions is + // disabled, the exceptions can be observed only by functions that read + // exception flags, like 'llvm.get_fpenv' or 'fetestexcept'. It means that + // the order of operations is not significant between barriers. + // + // If trapping is enabled, each operation becomes an implicit observation + // point, so the operations must be sequenced according their original + // source order. + if (!PendingConstrainedFP.empty()) { + assert(PendingConstrainedFPStrict.empty()); + updateRoot(PendingConstrainedFP); + } + // TODO: Add support for trapping-enabled scenarios. + } + return DAG.getRoot(); +} + SDValue SelectionDAGBuilder::getRoot() { // Chain up all pending constrained intrinsics together with all // pending loads, by simply appending them to PendingLoads and @@ -1977,8 +2014,13 @@ SDValue SelectionDAGBuilder::getValueImpl(const Value *V) { if (const Instruction *Inst = dyn_cast<Instruction>(V)) { Register InReg = FuncInfo.InitializeRegForValue(Inst); + std::optional<CallingConv::ID> CallConv; + auto *CI = dyn_cast<CallInst>(Inst); + if (CI && !CI->isInlineAsm()) + CallConv = CI->getCallingConv(); + RegsForValue RFV(*DAG.getContext(), TLI, DAG.getDataLayout(), InReg, - Inst->getType(), std::nullopt); + Inst->getType(), CallConv); SDValue Chain = DAG.getEntryNode(); return RFV.getCopyFromRegs(DAG, FuncInfo, getCurSDLoc(), Chain, nullptr, V); } @@ -3126,12 +3168,16 @@ void SelectionDAGBuilder::visitSPDescriptorParent(StackProtectorDescriptor &SPD, if (TLI.useLoadStackGuardNode(M)) { Guard = getLoadStackGuard(DAG, dl, Chain); } else { - const Value *IRGuard = TLI.getSDagStackGuard(M); - SDValue GuardPtr = getValue(IRGuard); - - Guard = DAG.getLoad(PtrMemTy, dl, Chain, GuardPtr, - MachinePointerInfo(IRGuard, 0), Align, - MachineMemOperand::MOVolatile); + if (const Value *IRGuard = TLI.getSDagStackGuard(M)) { + SDValue GuardPtr = getValue(IRGuard); + Guard = DAG.getLoad(PtrMemTy, dl, Chain, GuardPtr, + MachinePointerInfo(IRGuard, 0), Align, + MachineMemOperand::MOVolatile); + } else { + LLVMContext &Ctx = *DAG.getContext(); + Ctx.diagnose(DiagnosticInfoGeneric("unable to lower stackguard")); + Guard = DAG.getPOISON(PtrMemTy); + } } // Perform the comparison via a getsetcc. @@ -4381,6 +4427,7 @@ void SelectionDAGBuilder::visitGetElementPtr(const User &I) { if (NW.hasNoUnsignedWrap() || (int64_t(Offset) >= 0 && NW.hasNoUnsignedSignedWrap())) Flags |= SDNodeFlags::NoUnsignedWrap; + Flags.setInBounds(NW.isInBounds()); N = DAG.getMemBasePlusOffset( N, DAG.getConstant(Offset, dl, N.getValueType()), dl, Flags); @@ -4424,6 +4471,7 @@ void SelectionDAGBuilder::visitGetElementPtr(const User &I) { if (NW.hasNoUnsignedWrap() || (Offs.isNonNegative() && NW.hasNoUnsignedSignedWrap())) Flags.setNoUnsignedWrap(true); + Flags.setInBounds(NW.isInBounds()); OffsVal = DAG.getSExtOrTrunc(OffsVal, dl, N.getValueType()); @@ -4493,6 +4541,7 @@ void SelectionDAGBuilder::visitGetElementPtr(const User &I) { // pointer index type (add nuw). SDNodeFlags AddFlags; AddFlags.setNoUnsignedWrap(NW.hasNoUnsignedWrap()); + AddFlags.setInBounds(NW.isInBounds()); N = DAG.getMemBasePlusOffset(N, IdxN, dl, AddFlags); } @@ -7319,6 +7368,13 @@ void SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I, Res = DAG.getPtrExtOrTrunc(Res, sdl, PtrTy); } else { const Value *Global = TLI.getSDagStackGuard(M); + if (!Global) { + LLVMContext &Ctx = *DAG.getContext(); + Ctx.diagnose(DiagnosticInfoGeneric("unable to lower stackguard")); + setValue(&I, DAG.getPOISON(PtrTy)); + return; + } + Align Align = DAG.getDataLayout().getPrefTypeAlign(Global->getType()); Res = DAG.getLoad(PtrTy, sdl, Chain, getValue(Global), MachinePointerInfo(Global, 0), Align, @@ -8279,6 +8335,30 @@ void SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I, } } +void SelectionDAGBuilder::pushFPOpOutChain(SDValue Result, + fp::ExceptionBehavior EB) { + assert(Result.getNode()->getNumValues() == 2); + SDValue OutChain = Result.getValue(1); + assert(OutChain.getValueType() == MVT::Other); + + // Instead of updating the root immediately, push the produced chain to the + // appropriate list, deferring the update until the root is requested. In this + // case, the nodes from the lists are chained using TokenFactor, indicating + // that the operations are independent. + // + // In particular, the root is updated before any call that might access the + // floating-point environment, except for constrained intrinsics. + switch (EB) { + case fp::ExceptionBehavior::ebMayTrap: + case fp::ExceptionBehavior::ebIgnore: + PendingConstrainedFP.push_back(OutChain); + break; + case fp::ExceptionBehavior::ebStrict: + PendingConstrainedFPStrict.push_back(OutChain); + break; + } +} + void SelectionDAGBuilder::visitConstrainedFPIntrinsic( const ConstrainedFPIntrinsic &FPI) { SDLoc sdl = getCurSDLoc(); @@ -8286,42 +8366,16 @@ void SelectionDAGBuilder::visitConstrainedFPIntrinsic( // We do not need to serialize constrained FP intrinsics against // each other or against (nonvolatile) loads, so they can be // chained like loads. - SDValue Chain = DAG.getRoot(); + fp::ExceptionBehavior EB = *FPI.getExceptionBehavior(); + SDValue Chain = getFPOperationRoot(EB); SmallVector<SDValue, 4> Opers; Opers.push_back(Chain); for (unsigned I = 0, E = FPI.getNonMetadataArgCount(); I != E; ++I) Opers.push_back(getValue(FPI.getArgOperand(I))); - auto pushOutChain = [this](SDValue Result, fp::ExceptionBehavior EB) { - assert(Result.getNode()->getNumValues() == 2); - - // Push node to the appropriate list so that future instructions can be - // chained up correctly. - SDValue OutChain = Result.getValue(1); - switch (EB) { - case fp::ExceptionBehavior::ebIgnore: - // The only reason why ebIgnore nodes still need to be chained is that - // they might depend on the current rounding mode, and therefore must - // not be moved across instruction that may change that mode. - [[fallthrough]]; - case fp::ExceptionBehavior::ebMayTrap: - // These must not be moved across calls or instructions that may change - // floating-point exception masks. - PendingConstrainedFP.push_back(OutChain); - break; - case fp::ExceptionBehavior::ebStrict: - // These must not be moved across calls or instructions that may change - // floating-point exception masks or read floating-point exception flags. - // In addition, they cannot be optimized out even if unused. - PendingConstrainedFPStrict.push_back(OutChain); - break; - } - }; - const TargetLowering &TLI = DAG.getTargetLoweringInfo(); EVT VT = TLI.getValueType(DAG.getDataLayout(), FPI.getType()); SDVTList VTs = DAG.getVTList(VT, MVT::Other); - fp::ExceptionBehavior EB = *FPI.getExceptionBehavior(); SDNodeFlags Flags; if (EB == fp::ExceptionBehavior::ebIgnore) @@ -8345,7 +8399,7 @@ void SelectionDAGBuilder::visitConstrainedFPIntrinsic( !TLI.isFMAFasterThanFMulAndFAdd(DAG.getMachineFunction(), VT)) { Opers.pop_back(); SDValue Mul = DAG.getNode(ISD::STRICT_FMUL, sdl, VTs, Opers, Flags); - pushOutChain(Mul, EB); + pushFPOpOutChain(Mul, EB); Opcode = ISD::STRICT_FADD; Opers.clear(); Opers.push_back(Mul.getValue(1)); @@ -8376,7 +8430,7 @@ void SelectionDAGBuilder::visitConstrainedFPIntrinsic( } SDValue Result = DAG.getNode(Opcode, sdl, VTs, Opers, Flags); - pushOutChain(Result, EB); + pushFPOpOutChain(Result, EB); SDValue FPResult = Result.getValue(0); setValue(&FPI, FPResult); diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h index c7577fa..47e19f7 100644 --- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h +++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h @@ -195,6 +195,11 @@ private: /// Update root to include all chains from the Pending list. SDValue updateRoot(SmallVectorImpl<SDValue> &Pending); + /// Given a node representing a floating-point operation and its specified + /// exception behavior, this either updates the root or stores the node in + /// a list to be added to chains latter. + void pushFPOpOutChain(SDValue Result, fp::ExceptionBehavior EB); + /// A unique monotonically increasing number used to order the SDNodes we /// create. unsigned SDNodeOrder; @@ -300,6 +305,13 @@ public: /// memory node that may need to be ordered after any prior load instructions. SDValue getMemoryRoot(); + /// Return the current virtual root of the Selection DAG, flushing + /// PendingConstrainedFP or PendingConstrainedFPStrict items if the new + /// exception behavior (specified by \p EB) differs from that of the pending + /// instructions. This must be done before emitting constrained FP operation + /// call. + SDValue getFPOperationRoot(fp::ExceptionBehavior EB); + /// Similar to getMemoryRoot, but also flushes PendingConstrainedFP(Strict) /// items. This must be done before emitting any call other any other node /// that may need to be ordered after FP instructions due to other side diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp index 39cbfad..77377d3 100644 --- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp @@ -689,6 +689,9 @@ void SDNode::print_details(raw_ostream &OS, const SelectionDAG *G) const { if (getFlags().hasSameSign()) OS << " samesign"; + if (getFlags().hasInBounds()) + OS << " inbounds"; + if (getFlags().hasNonNeg()) OS << " nneg"; diff --git a/llvm/lib/CodeGen/ShrinkWrap.cpp b/llvm/lib/CodeGen/ShrinkWrap.cpp index 826e412..8358105 100644 --- a/llvm/lib/CodeGen/ShrinkWrap.cpp +++ b/llvm/lib/CodeGen/ShrinkWrap.cpp @@ -319,7 +319,7 @@ bool ShrinkWrapImpl::useOrDefCSROrFI(const MachineInstr &MI, RegScavenger *RS, return isa<GlobalValue>(UO); } if (const PseudoSourceValue *PSV = Op->getPseudoValue()) - return PSV->isJumpTable(); + return PSV->isJumpTable() || PSV->isConstantPool(); return false; }; // Load/store operations may access the stack indirectly when we previously diff --git a/llvm/lib/CodeGen/TargetOptionsImpl.cpp b/llvm/lib/CodeGen/TargetOptionsImpl.cpp index 5eb86e7..049efe8 100644 --- a/llvm/lib/CodeGen/TargetOptionsImpl.cpp +++ b/llvm/lib/CodeGen/TargetOptionsImpl.cpp @@ -51,7 +51,7 @@ bool TargetOptions::FramePointerIsReserved(const MachineFunction &MF) const { /// HonorSignDependentRoundingFPMath - Return true if the codegen must assume /// that the rounding mode of the FPU can change from its default. bool TargetOptions::HonorSignDependentRoundingFPMath() const { - return !UnsafeFPMath && HonorSignDependentRoundingFPMathOption; + return HonorSignDependentRoundingFPMathOption; } /// NOTE: There are targets that still do not support the debug entry values diff --git a/llvm/lib/CodeGen/TargetPassConfig.cpp b/llvm/lib/CodeGen/TargetPassConfig.cpp index b6169e6..10b7238 100644 --- a/llvm/lib/CodeGen/TargetPassConfig.cpp +++ b/llvm/lib/CodeGen/TargetPassConfig.cpp @@ -272,6 +272,12 @@ static cl::opt<bool> cl::desc("Split static data sections into hot and cold " "sections using profile information")); +cl::opt<bool> EmitBBHash( + "emit-bb-hash", + cl::desc( + "Emit the hash of basic block in the SHT_LLVM_BB_ADDR_MAP section."), + cl::init(false), cl::Optional); + /// Allow standard passes to be disabled by command line options. This supports /// simple binary flags that either suppress the pass or do nothing. /// i.e. -disable-mypass=false has no effect. @@ -1281,6 +1287,8 @@ void TargetPassConfig::addMachinePasses() { // address map (or both). if (TM->getBBSectionsType() != llvm::BasicBlockSection::None || TM->Options.BBAddrMap) { + if (EmitBBHash) + addPass(llvm::createMachineBlockHashInfoPass()); if (TM->getBBSectionsType() == llvm::BasicBlockSection::List) { addPass(llvm::createBasicBlockSectionsProfileReaderWrapperPass( TM->getBBSectionsFuncListBuf())); |