diff options
Diffstat (limited to 'clang/lib/CodeGen')
22 files changed, 532 insertions, 246 deletions
diff --git a/clang/lib/CodeGen/BackendUtil.cpp b/clang/lib/CodeGen/BackendUtil.cpp index 2f6d4c414e73..1b7257857dd3 100644 --- a/clang/lib/CodeGen/BackendUtil.cpp +++ b/clang/lib/CodeGen/BackendUtil.cpp @@ -407,13 +407,13 @@ static bool initTargetOptions(const CompilerInstance &CI, // Set EABI version. Options.EABIVersion = TargetOpts.EABIVersion; - if (LangOpts.hasSjLjExceptions()) + if (CodeGenOpts.hasSjLjExceptions()) Options.ExceptionModel = llvm::ExceptionHandling::SjLj; - if (LangOpts.hasSEHExceptions()) + if (CodeGenOpts.hasSEHExceptions()) Options.ExceptionModel = llvm::ExceptionHandling::WinEH; - if (LangOpts.hasDWARFExceptions()) + if (CodeGenOpts.hasDWARFExceptions()) Options.ExceptionModel = llvm::ExceptionHandling::DwarfCFI; - if (LangOpts.hasWasmExceptions()) + if (CodeGenOpts.hasWasmExceptions()) Options.ExceptionModel = llvm::ExceptionHandling::Wasm; Options.NoInfsFPMath = LangOpts.NoHonorInfs; diff --git a/clang/lib/CodeGen/CGBlocks.cpp b/clang/lib/CodeGen/CGBlocks.cpp index f3ddf7bf9a46..0e80522536e1 100644 --- a/clang/lib/CodeGen/CGBlocks.cpp +++ b/clang/lib/CodeGen/CGBlocks.cpp @@ -853,9 +853,24 @@ llvm::Value *CodeGenFunction::EmitBlockLiteral(const CGBlockInfo &blockInfo) { offset += size; index++; }; + auto addSignedHeaderField = + [&](llvm::Value *Value, const PointerAuthSchema &Schema, + GlobalDecl Decl, QualType Type, CharUnits Size, const Twine &Name) { + auto StorageAddress = projectField(index, Name); + if (Schema) { + auto AuthInfo = EmitPointerAuthInfo( + Schema, StorageAddress.emitRawPointer(*this), Decl, Type); + Value = EmitPointerAuthSign(AuthInfo, Value); + } + Builder.CreateStore(Value, StorageAddress); + offset += Size; + index++; + }; if (!IsOpenCL) { - addHeaderField(isa, getPointerSize(), "block.isa"); + addSignedHeaderField( + isa, CGM.getCodeGenOpts().PointerAuth.ObjCIsaPointers, GlobalDecl(), + QualType(), getPointerSize(), "block.isa"); addHeaderField(llvm::ConstantInt::get(IntTy, flags.getBitMask()), getIntSize(), "block.flags"); addHeaderField(llvm::ConstantInt::get(IntTy, 0), getIntSize(), @@ -1285,7 +1300,9 @@ static llvm::Constant *buildGlobalBlock(CodeGenModule &CGM, if (IsWindows) fields.addNullPointer(CGM.Int8PtrPtrTy); else - fields.add(CGM.getNSConcreteGlobalBlock()); + fields.addSignedPointer(CGM.getNSConcreteGlobalBlock(), + CGM.getCodeGenOpts().PointerAuth.ObjCIsaPointers, + GlobalDecl(), QualType()); // __flags BlockFlags flags = BLOCK_IS_GLOBAL; @@ -1532,8 +1549,8 @@ llvm::Function *CodeGenFunction::GenerateBlockFunction( llvm::BasicBlock *resume = Builder.GetInsertBlock(); // Go back to the entry. - if (entry_ptr->getNextNonDebugInstruction()) - entry_ptr = entry_ptr->getNextNonDebugInstruction()->getIterator(); + if (entry_ptr->getNextNode()) + entry_ptr = entry_ptr->getNextNode()->getIterator(); else entry_ptr = entry->end(); Builder.SetInsertPoint(entry, entry_ptr); diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp index c8c3d6b20c49..19d8ba26d44d 100644 --- a/clang/lib/CodeGen/CGCall.cpp +++ b/clang/lib/CodeGen/CGCall.cpp @@ -2852,8 +2852,21 @@ void CodeGenModule::ConstructAttributeList(StringRef Name, if (AI.getInReg()) Attrs.addAttribute(llvm::Attribute::InReg); - if (AI.getIndirectByVal()) + // Depending on the ABI, this may be either a byval or a dead_on_return + // argument. + if (AI.getIndirectByVal()) { Attrs.addByValAttr(getTypes().ConvertTypeForMem(ParamType)); + } else { + // Add dead_on_return when the object's lifetime ends in the callee. + // This includes trivially-destructible objects, as well as objects + // whose destruction / clean-up is carried out within the callee (e.g., + // Obj-C ARC-managed structs, MSVC callee-destroyed objects). + if (!ParamType.isDestructedType() || !ParamType->isRecordType() || + ParamType->castAs<RecordType>() + ->getDecl() + ->isParamDestroyedInCallee()) + Attrs.addAttribute(llvm::Attribute::DeadOnReturn); + } auto *Decl = ParamType->getAsRecordDecl(); if (CodeGenOpts.PassByValueIsNoAlias && Decl && diff --git a/clang/lib/CodeGen/CGCoroutine.cpp b/clang/lib/CodeGen/CGCoroutine.cpp index 0fc488e98aaf..117ef3d16e21 100644 --- a/clang/lib/CodeGen/CGCoroutine.cpp +++ b/clang/lib/CodeGen/CGCoroutine.cpp @@ -707,11 +707,15 @@ struct GetReturnObjectManager { Builder.CreateStore(Builder.getFalse(), GroActiveFlag); GroEmission = CGF.EmitAutoVarAlloca(*GroVarDecl); - auto *GroAlloca = dyn_cast_or_null<llvm::AllocaInst>( - GroEmission.getOriginalAllocatedAddress().getPointer()); - assert(GroAlloca && "expected alloca to be emitted"); - GroAlloca->setMetadata(llvm::LLVMContext::MD_coro_outside_frame, - llvm::MDNode::get(CGF.CGM.getLLVMContext(), {})); + + if (!GroVarDecl->isNRVOVariable()) { + // NRVO variables don't have allocas and won't have the same issue. + auto *GroAlloca = dyn_cast_or_null<llvm::AllocaInst>( + GroEmission.getOriginalAllocatedAddress().getPointer()); + assert(GroAlloca && "expected alloca to be emitted"); + GroAlloca->setMetadata(llvm::LLVMContext::MD_coro_outside_frame, + llvm::MDNode::get(CGF.CGM.getLLVMContext(), {})); + } // Remember the top of EHStack before emitting the cleanup. auto old_top = CGF.EHStack.stable_begin(); diff --git a/clang/lib/CodeGen/CGDebugInfo.cpp b/clang/lib/CodeGen/CGDebugInfo.cpp index 8bf7d24d8551..446cf8d9e05c 100644 --- a/clang/lib/CodeGen/CGDebugInfo.cpp +++ b/clang/lib/CodeGen/CGDebugInfo.cpp @@ -170,6 +170,10 @@ void CGDebugInfo::addInstToSpecificSourceAtom(llvm::Instruction *KeyInstruction, if (!Group || !CGM.getCodeGenOpts().DebugKeyInstructions) return; + llvm::DISubprogram *SP = KeyInstruction->getFunction()->getSubprogram(); + if (!SP || !SP->getKeyInstructionsEnabled()) + return; + addInstSourceAtomMetadata(KeyInstruction, Group, /*Rank=*/1); llvm::Instruction *BackupI = @@ -514,6 +518,8 @@ CGDebugInfo::computeChecksum(FileID FID, SmallString<64> &Checksum) const { case clang::CodeGenOptions::DSH_SHA256: llvm::toHex(llvm::SHA256::hash(Data), /*LowerCase=*/true, Checksum); return llvm::DIFile::CSK_SHA256; + case clang::CodeGenOptions::DSH_NONE: + return std::nullopt; } llvm_unreachable("Unhandled DebugSrcHashKind enum"); } @@ -797,7 +803,8 @@ void CGDebugInfo::CreateCompileUnit() { // Create new compile unit. TheCU = DBuilder.createCompileUnit( LangTag, CUFile, CGOpts.EmitVersionIdentMetadata ? Producer : "", - LO.Optimize || CGOpts.PrepareForLTO || CGOpts.PrepareForThinLTO, + CGOpts.OptimizationLevel != 0 || CGOpts.PrepareForLTO || + CGOpts.PrepareForThinLTO, CGOpts.DwarfDebugFlags, RuntimeVers, CGOpts.SplitDwarfFile, EmissionKind, DwoId, CGOpts.SplitDwarfInlining, CGOpts.DebugInfoForProfiling, NameTableKind, CGOpts.DebugRangesBaseAddress, remapDIPath(Sysroot), SDK); @@ -2282,7 +2289,7 @@ llvm::DISubprogram *CGDebugInfo::CreateCXXMemberFunction( Flags |= llvm::DINode::FlagRValueReference; if (!Method->isExternallyVisible()) SPFlags |= llvm::DISubprogram::SPFlagLocalToUnit; - if (CGM.getLangOpts().Optimize) + if (CGM.getCodeGenOpts().OptimizationLevel != 0) SPFlags |= llvm::DISubprogram::SPFlagOptimized; // In this debug mode, emit type info for a class when its constructor type @@ -4353,7 +4360,7 @@ llvm::DISubprogram *CGDebugInfo::getFunctionFwdDeclOrStub(GlobalDecl GD, FD->getReturnType(), ArgTypes, FunctionProtoType::ExtProtoInfo(CC)); if (!FD->isExternallyVisible()) SPFlags |= llvm::DISubprogram::SPFlagLocalToUnit; - if (CGM.getLangOpts().Optimize) + if (CGM.getCodeGenOpts().OptimizationLevel != 0) SPFlags |= llvm::DISubprogram::SPFlagOptimized; if (Stub) { @@ -4629,6 +4636,7 @@ void CGDebugInfo::emitFunctionStart(GlobalDecl GD, SourceLocation Loc, llvm::DIFile *Unit = getOrCreateFile(Loc); llvm::DIScope *FDContext = Unit; llvm::DINodeArray TParamsArray; + bool KeyInstructions = CGM.getCodeGenOpts().DebugKeyInstructions; if (!HasDecl) { // Use llvm function name. LinkageName = Fn->getName(); @@ -4645,6 +4653,9 @@ void CGDebugInfo::emitFunctionStart(GlobalDecl GD, SourceLocation Loc, } collectFunctionDeclProps(GD, Unit, Name, LinkageName, FDContext, TParamsArray, Flags); + // Disable KIs if this is a coroutine. + KeyInstructions = + KeyInstructions && !isa_and_present<CoroutineBodyStmt>(FD->getBody()); } else if (const auto *OMD = dyn_cast<ObjCMethodDecl>(D)) { Name = getObjCMethodName(OMD); Flags |= llvm::DINode::FlagPrototyped; @@ -4679,7 +4690,7 @@ void CGDebugInfo::emitFunctionStart(GlobalDecl GD, SourceLocation Loc, if (Fn->hasLocalLinkage()) SPFlags |= llvm::DISubprogram::SPFlagLocalToUnit; - if (CGM.getLangOpts().Optimize) + if (CGM.getCodeGenOpts().OptimizationLevel != 0) SPFlags |= llvm::DISubprogram::SPFlagOptimized; llvm::DINode::DIFlags FlagsForDef = Flags | getCallSiteRelatedAttrs(); @@ -4706,7 +4717,7 @@ void CGDebugInfo::emitFunctionStart(GlobalDecl GD, SourceLocation Loc, llvm::DISubprogram *SP = DBuilder.createFunction( FDContext, Name, LinkageName, Unit, LineNo, DIFnType, ScopeLine, FlagsForDef, SPFlagsForDef, TParamsArray.get(), Decl, nullptr, - Annotations, "", CGM.getCodeGenOpts().DebugKeyInstructions); + Annotations, "", KeyInstructions); Fn->setSubprogram(SP); // We might get here with a VarDecl in the case we're generating @@ -4762,7 +4773,7 @@ void CGDebugInfo::EmitFunctionDecl(GlobalDecl GD, SourceLocation Loc, unsigned LineNo = getLineNumber(Loc); unsigned ScopeLine = 0; llvm::DISubprogram::DISPFlags SPFlags = llvm::DISubprogram::SPFlagZero; - if (CGM.getLangOpts().Optimize) + if (CGM.getCodeGenOpts().OptimizationLevel != 0) SPFlags |= llvm::DISubprogram::SPFlagOptimized; llvm::DINodeArray Annotations = CollectBTFDeclTagAnnotations(D); @@ -5094,7 +5105,8 @@ llvm::DILocalVariable *CGDebugInfo::EmitDeclare(const VarDecl *VD, // Use VarDecl's Tag, Scope and Line number. auto FieldAlign = getDeclAlignIfRequired(Field, CGM.getContext()); auto *D = DBuilder.createAutoVariable( - Scope, FieldName, Unit, Line, FieldTy, CGM.getLangOpts().Optimize, + Scope, FieldName, Unit, Line, FieldTy, + CGM.getCodeGenOpts().OptimizationLevel != 0, Flags | llvm::DINode::FlagArtificial, FieldAlign); // Insert an llvm.dbg.declare into the current block. @@ -5120,9 +5132,9 @@ llvm::DILocalVariable *CGDebugInfo::EmitDeclare(const VarDecl *VD, llvm::DILocalVariable *D = nullptr; if (ArgNo) { llvm::DINodeArray Annotations = CollectBTFDeclTagAnnotations(VD); - D = DBuilder.createParameterVariable(Scope, Name, *ArgNo, Unit, Line, Ty, - CGM.getLangOpts().Optimize, Flags, - Annotations); + D = DBuilder.createParameterVariable( + Scope, Name, *ArgNo, Unit, Line, Ty, + CGM.getCodeGenOpts().OptimizationLevel != 0, Flags, Annotations); } else { // For normal local variable, we will try to find out whether 'VD' is the // copy parameter of coroutine. @@ -5163,8 +5175,9 @@ llvm::DILocalVariable *CGDebugInfo::EmitDeclare(const VarDecl *VD, D = RemapCoroArgToLocalVar(); // Or we will create a new DIVariable for this Decl if D dose not exists. if (!D) - D = DBuilder.createAutoVariable(Scope, Name, Unit, Line, Ty, - CGM.getLangOpts().Optimize, Flags, Align); + D = DBuilder.createAutoVariable( + Scope, Name, Unit, Line, Ty, + CGM.getCodeGenOpts().OptimizationLevel != 0, Flags, Align); } // Insert an llvm.dbg.declare into the current block. DBuilder.insertDeclare(Storage, D, DBuilder.createExpression(Expr), @@ -5218,7 +5231,7 @@ llvm::DILocalVariable *CGDebugInfo::EmitDeclare(const BindingDecl *BD, auto *Scope = cast<llvm::DIScope>(LexicalBlockStack.back()); // Create the descriptor for the variable. llvm::DILocalVariable *D = DBuilder.createAutoVariable( - Scope, Name, Unit, Line, Ty, CGM.getLangOpts().Optimize, + Scope, Name, Unit, Line, Ty, CGM.getCodeGenOpts().OptimizationLevel != 0, llvm::DINode::FlagZero, Align); if (const MemberExpr *ME = dyn_cast<MemberExpr>(BD->getBinding())) { @@ -5316,9 +5329,10 @@ void CGDebugInfo::EmitLabel(const LabelDecl *D, CGBuilderTy &Builder) { StringRef Name = D->getName(); // Create the descriptor for the label. - auto *L = DBuilder.createLabel( - Scope, Name, Unit, Line, Column, /*IsArtificial=*/false, - /*CoroSuspendIdx=*/std::nullopt, CGM.getLangOpts().Optimize); + auto *L = DBuilder.createLabel(Scope, Name, Unit, Line, Column, + /*IsArtificial=*/false, + /*CoroSuspendIdx=*/std::nullopt, + CGM.getCodeGenOpts().OptimizationLevel != 0); // Insert an llvm.dbg.label into the current block. DBuilder.insertLabel(L, @@ -5581,7 +5595,8 @@ void CGDebugInfo::EmitDeclareOfBlockLiteralArgVariable(const CGBlockInfo &block, // Create the descriptor for the parameter. auto *debugVar = DBuilder.createParameterVariable( - scope, Name, ArgNo, tunit, line, type, CGM.getLangOpts().Optimize, flags); + scope, Name, ArgNo, tunit, line, type, + CGM.getCodeGenOpts().OptimizationLevel != 0, flags); // Insert an llvm.dbg.declare into the current block. DBuilder.insertDeclare(Alloca, debugVar, DBuilder.createExpression(), @@ -6356,7 +6371,7 @@ llvm::DebugLoc CGDebugInfo::SourceLocToDebugLoc(SourceLocation Loc) { llvm::DINode::DIFlags CGDebugInfo::getCallSiteRelatedAttrs() const { // Call site-related attributes are only useful in optimized programs, and // when there's a possibility of debugging backtraces. - if (!CGM.getLangOpts().Optimize || + if (CGM.getCodeGenOpts().OptimizationLevel == 0 || DebugKind == llvm::codegenoptions::NoDebugInfo || DebugKind == llvm::codegenoptions::LocTrackingOnly) return llvm::DINode::FlagZero; @@ -6465,24 +6480,27 @@ SanitizerOrdinalToCheckLabel(SanitizerKind::SanitizerOrdinal Ordinal) { llvm::DILocation *CodeGenFunction::SanitizerAnnotateDebugInfo( ArrayRef<SanitizerKind::SanitizerOrdinal> Ordinals, SanitizerHandler Handler) { + llvm::DILocation *CheckDebugLoc = Builder.getCurrentDebugLocation(); + auto *DI = getDebugInfo(); + if (!DI) + return CheckDebugLoc; + std::string Label; if (Ordinals.size() == 1) Label = SanitizerOrdinalToCheckLabel(Ordinals[0]); else Label = SanitizerHandlerToCheckLabel(Handler); - llvm::DILocation *CheckDI = Builder.getCurrentDebugLocation(); - for (auto Ord : Ordinals) { // TODO: deprecate ClArrayBoundsPseudoFn if (((ClArrayBoundsPseudoFn && Ord == SanitizerKind::SO_ArrayBounds) || CGM.getCodeGenOpts().SanitizeAnnotateDebugInfo.has(Ord)) && - CheckDI) { - return getDebugInfo()->CreateSyntheticInlineAt(CheckDI, Label); + CheckDebugLoc) { + return DI->CreateSyntheticInlineAt(CheckDebugLoc, Label); } } - return CheckDI; + return CheckDebugLoc; } SanitizerDebugLocation::SanitizerDebugLocation( diff --git a/clang/lib/CodeGen/CGException.cpp b/clang/lib/CodeGen/CGException.cpp index ad138b9876e8..f86af4581c34 100644 --- a/clang/lib/CodeGen/CGException.cpp +++ b/clang/lib/CodeGen/CGException.cpp @@ -131,20 +131,21 @@ const EHPersonality EHPersonality::ZOS_CPlusPlus = {"__zos_cxx_personality_v2", nullptr}; static const EHPersonality &getCPersonality(const TargetInfo &Target, - const LangOptions &L) { + const CodeGenOptions &CGOpts) { const llvm::Triple &T = Target.getTriple(); if (T.isWindowsMSVCEnvironment()) return EHPersonality::MSVC_CxxFrameHandler3; - if (L.hasSjLjExceptions()) + if (CGOpts.hasSjLjExceptions()) return EHPersonality::GNU_C_SJLJ; - if (L.hasDWARFExceptions()) + if (CGOpts.hasDWARFExceptions()) return EHPersonality::GNU_C; - if (L.hasSEHExceptions()) + if (CGOpts.hasSEHExceptions()) return EHPersonality::GNU_C_SEH; return EHPersonality::GNU_C; } static const EHPersonality &getObjCPersonality(const TargetInfo &Target, + const CodeGenOptions &CGOpts, const LangOptions &L) { const llvm::Triple &T = Target.getTriple(); if (T.isWindowsMSVCEnvironment()) @@ -152,7 +153,7 @@ static const EHPersonality &getObjCPersonality(const TargetInfo &Target, switch (L.ObjCRuntime.getKind()) { case ObjCRuntime::FragileMacOSX: - return getCPersonality(Target, L); + return getCPersonality(Target, CGOpts); case ObjCRuntime::MacOSX: case ObjCRuntime::iOS: case ObjCRuntime::WatchOS: @@ -165,9 +166,9 @@ static const EHPersonality &getObjCPersonality(const TargetInfo &Target, [[fallthrough]]; case ObjCRuntime::GCC: case ObjCRuntime::ObjFW: - if (L.hasSjLjExceptions()) + if (CGOpts.hasSjLjExceptions()) return EHPersonality::GNU_ObjC_SJLJ; - if (L.hasSEHExceptions()) + if (CGOpts.hasSEHExceptions()) return EHPersonality::GNU_ObjC_SEH; return EHPersonality::GNU_ObjC; } @@ -175,19 +176,19 @@ static const EHPersonality &getObjCPersonality(const TargetInfo &Target, } static const EHPersonality &getCXXPersonality(const TargetInfo &Target, - const LangOptions &L) { + const CodeGenOptions &CGOpts) { const llvm::Triple &T = Target.getTriple(); if (T.isWindowsMSVCEnvironment()) return EHPersonality::MSVC_CxxFrameHandler3; if (T.isOSAIX()) return EHPersonality::XL_CPlusPlus; - if (L.hasSjLjExceptions()) + if (CGOpts.hasSjLjExceptions()) return EHPersonality::GNU_CPlusPlus_SJLJ; - if (L.hasDWARFExceptions()) + if (CGOpts.hasDWARFExceptions()) return EHPersonality::GNU_CPlusPlus; - if (L.hasSEHExceptions()) + if (CGOpts.hasSEHExceptions()) return EHPersonality::GNU_CPlusPlus_SEH; - if (L.hasWasmExceptions()) + if (CGOpts.hasWasmExceptions()) return EHPersonality::GNU_Wasm_CPlusPlus; if (T.isOSzOS()) return EHPersonality::ZOS_CPlusPlus; @@ -197,6 +198,7 @@ static const EHPersonality &getCXXPersonality(const TargetInfo &Target, /// Determines the personality function to use when both C++ /// and Objective-C exceptions are being caught. static const EHPersonality &getObjCXXPersonality(const TargetInfo &Target, + const CodeGenOptions &CGOpts, const LangOptions &L) { if (Target.getTriple().isWindowsMSVCEnvironment()) return EHPersonality::MSVC_CxxFrameHandler3; @@ -205,7 +207,7 @@ static const EHPersonality &getObjCXXPersonality(const TargetInfo &Target, // In the fragile ABI, just use C++ exception handling and hope // they're not doing crazy exception mixing. case ObjCRuntime::FragileMacOSX: - return getCXXPersonality(Target, L); + return getCXXPersonality(Target, CGOpts); // The ObjC personality defers to the C++ personality for non-ObjC // handlers. Unlike the C++ case, we use the same personality @@ -213,7 +215,7 @@ static const EHPersonality &getObjCXXPersonality(const TargetInfo &Target, case ObjCRuntime::MacOSX: case ObjCRuntime::iOS: case ObjCRuntime::WatchOS: - return getObjCPersonality(Target, L); + return getObjCPersonality(Target, CGOpts, L); case ObjCRuntime::GNUstep: return Target.getTriple().isOSCygMing() ? EHPersonality::GNU_CPlusPlus_SEH @@ -223,7 +225,7 @@ static const EHPersonality &getObjCXXPersonality(const TargetInfo &Target, // mixed EH. Use the ObjC personality just to avoid returning null. case ObjCRuntime::GCC: case ObjCRuntime::ObjFW: - return getObjCPersonality(Target, L); + return getObjCPersonality(Target, CGOpts, L); } llvm_unreachable("bad runtime kind"); } @@ -237,6 +239,7 @@ static const EHPersonality &getSEHPersonalityMSVC(const llvm::Triple &T) { const EHPersonality &EHPersonality::get(CodeGenModule &CGM, const FunctionDecl *FD) { const llvm::Triple &T = CGM.getTarget().getTriple(); + const CodeGenOptions &CGOpts = CGM.getCodeGenOpts(); const LangOptions &L = CGM.getLangOpts(); const TargetInfo &Target = CGM.getTarget(); @@ -245,10 +248,10 @@ const EHPersonality &EHPersonality::get(CodeGenModule &CGM, return getSEHPersonalityMSVC(T); if (L.ObjC) - return L.CPlusPlus ? getObjCXXPersonality(Target, L) - : getObjCPersonality(Target, L); - return L.CPlusPlus ? getCXXPersonality(Target, L) - : getCPersonality(Target, L); + return L.CPlusPlus ? getObjCXXPersonality(Target, CGOpts, L) + : getObjCPersonality(Target, CGOpts, L); + return L.CPlusPlus ? getCXXPersonality(Target, CGOpts) + : getCPersonality(Target, CGOpts); } const EHPersonality &EHPersonality::get(CodeGenFunction &CGF) { @@ -344,7 +347,7 @@ void CodeGenModule::SimplifyPersonality() { return; const EHPersonality &ObjCXX = EHPersonality::get(*this, /*FD=*/nullptr); - const EHPersonality &CXX = getCXXPersonality(getTarget(), LangOpts); + const EHPersonality &CXX = getCXXPersonality(getTarget(), CodeGenOpts); if (&ObjCXX == &CXX) return; @@ -500,7 +503,7 @@ void CodeGenFunction::EmitStartEHSpec(const Decl *D) { // In Wasm EH we currently treat 'throw()' in the same way as 'noexcept'. In // case of throw with types, we ignore it and print a warning for now. // TODO Correctly handle exception specification in Wasm EH - if (CGM.getLangOpts().hasWasmExceptions()) { + if (CGM.getCodeGenOpts().hasWasmExceptions()) { if (EST == EST_DynamicNone) EHStack.pushTerminate(); else @@ -515,8 +518,8 @@ void CodeGenFunction::EmitStartEHSpec(const Decl *D) { // throw with types. // TODO Correctly handle exception specification in Emscripten EH if (getTarget().getCXXABI() == TargetCXXABI::WebAssembly && - CGM.getLangOpts().getExceptionHandling() == - LangOptions::ExceptionHandlingKind::None && + CGM.getCodeGenOpts().getExceptionHandling() == + CodeGenOptions::ExceptionHandlingKind::None && EST == EST_Dynamic) CGM.getDiags().Report(D->getLocation(), diag::warn_wasm_dynamic_exception_spec_ignored) @@ -604,7 +607,7 @@ void CodeGenFunction::EmitEndEHSpec(const Decl *D) { // In wasm we currently treat 'throw()' in the same way as 'noexcept'. In // case of throw with types, we ignore it and print a warning for now. // TODO Correctly handle exception specification in wasm - if (CGM.getLangOpts().hasWasmExceptions()) { + if (CGM.getCodeGenOpts().hasWasmExceptions()) { if (EST == EST_DynamicNone) EHStack.popTerminate(); return; diff --git a/clang/lib/CodeGen/CGLoopInfo.cpp b/clang/lib/CodeGen/CGLoopInfo.cpp index 4a9092842858..b2b569a43038 100644 --- a/clang/lib/CodeGen/CGLoopInfo.cpp +++ b/clang/lib/CodeGen/CGLoopInfo.cpp @@ -221,18 +221,6 @@ LoopInfo::createLoopVectorizeMetadata(const LoopAttributes &Attrs, return createUnrollAndJamMetadata(Attrs, LoopProperties, HasUserTransforms); } - // Apply all loop properties to the vectorized loop. - SmallVector<Metadata *, 4> FollowupLoopProperties; - FollowupLoopProperties.append(LoopProperties.begin(), LoopProperties.end()); - - // Don't vectorize an already vectorized loop. - FollowupLoopProperties.push_back( - MDNode::get(Ctx, MDString::get(Ctx, "llvm.loop.isvectorized"))); - - bool FollowupHasTransforms = false; - SmallVector<Metadata *, 4> Followup = createUnrollAndJamMetadata( - Attrs, FollowupLoopProperties, FollowupHasTransforms); - SmallVector<Metadata *, 4> Args; Args.append(LoopProperties.begin(), LoopProperties.end()); @@ -286,22 +274,46 @@ LoopInfo::createLoopVectorizeMetadata(const LoopAttributes &Attrs, // 5) it is implied when vectorize.width is unset (0) and the user // explicitly requested fixed-width vectorization, i.e. // vectorize.scalable.enable is false. + bool VectorizeEnabled = false; if (Attrs.VectorizeEnable != LoopAttributes::Unspecified || (IsVectorPredicateEnabled && Attrs.VectorizeWidth != 1) || Attrs.VectorizeWidth > 1 || Attrs.VectorizeScalable == LoopAttributes::Enable || (Attrs.VectorizeScalable == LoopAttributes::Disable && Attrs.VectorizeWidth != 1)) { - bool AttrVal = Attrs.VectorizeEnable != LoopAttributes::Disable; + VectorizeEnabled = Attrs.VectorizeEnable != LoopAttributes::Disable; Args.push_back( MDNode::get(Ctx, {MDString::get(Ctx, "llvm.loop.vectorize.enable"), ConstantAsMetadata::get(ConstantInt::get( - llvm::Type::getInt1Ty(Ctx), AttrVal))})); + llvm::Type::getInt1Ty(Ctx), VectorizeEnabled))})); } - if (FollowupHasTransforms) - Args.push_back( - createFollowupMetadata("llvm.loop.vectorize.followup_all", Followup)); + // Apply all loop properties to the vectorized loop. + SmallVector<Metadata *, 4> FollowupLoopProperties; + + // If vectorization is not explicitly enabled, the follow-up metadata will be + // directly appended to the list currently being created. In that case, adding + // LoopProperties to FollowupLoopProperties would result in duplication. + if (VectorizeEnabled) + FollowupLoopProperties.append(LoopProperties.begin(), LoopProperties.end()); + + // Don't vectorize an already vectorized loop. + FollowupLoopProperties.push_back( + MDNode::get(Ctx, MDString::get(Ctx, "llvm.loop.isvectorized"))); + + bool FollowupHasTransforms = false; + SmallVector<Metadata *, 4> Followup = createUnrollAndJamMetadata( + Attrs, FollowupLoopProperties, FollowupHasTransforms); + + if (FollowupHasTransforms) { + // If vectorization is explicitly enabled, we create a follow-up metadata, + // otherwise directly add the contents of it to Args. + if (VectorizeEnabled) + Args.push_back( + createFollowupMetadata("llvm.loop.vectorize.followup_all", Followup)); + else + Args.append(Followup.begin(), Followup.end()); + } HasUserTransforms = true; return Args; diff --git a/clang/lib/CodeGen/CGObjC.cpp b/clang/lib/CodeGen/CGObjC.cpp index 6f87444d3f67..24b6ce7c1c70 100644 --- a/clang/lib/CodeGen/CGObjC.cpp +++ b/clang/lib/CodeGen/CGObjC.cpp @@ -1193,16 +1193,23 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl, ivarAddr = ivarAddr.withElementType(bitcastType); llvm::LoadInst *load = Builder.CreateLoad(ivarAddr, "load"); load->setAtomic(llvm::AtomicOrdering::Unordered); + llvm::Value *ivarVal = load; + if (PointerAuthQualifier PAQ = ivar->getType().getPointerAuth()) { + CGPointerAuthInfo SrcInfo = EmitPointerAuthInfo(PAQ, ivarAddr); + CGPointerAuthInfo TargetInfo = + CGM.getPointerAuthInfoForType(getterMethod->getReturnType()); + ivarVal = emitPointerAuthResign(ivarVal, ivar->getType(), SrcInfo, + TargetInfo, /*isKnownNonNull=*/false); + } // Store that value into the return address. Doing this with a // bitcast is likely to produce some pretty ugly IR, but it's not // the *most* terrible thing in the world. llvm::Type *retTy = ConvertType(getterMethod->getReturnType()); uint64_t retTySize = CGM.getDataLayout().getTypeSizeInBits(retTy); - llvm::Value *ivarVal = load; if (ivarSize > retTySize) { bitcastType = llvm::Type::getIntNTy(getLLVMContext(), retTySize); - ivarVal = Builder.CreateTrunc(load, bitcastType); + ivarVal = Builder.CreateTrunc(ivarVal, bitcastType); } Builder.CreateStore(ivarVal, ReturnValue.withElementType(bitcastType)); @@ -1214,6 +1221,16 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl, case PropertyImplStrategy::GetSetProperty: { llvm::FunctionCallee getPropertyFn = CGM.getObjCRuntime().GetPropertyGetFunction(); + + if (ivar->getType().getPointerAuth()) { + // This currently cannot be hit, but if we ever allow objc pointers + // to be signed, this will become possible. Reaching here would require + // a copy, weak, etc property backed by an authenticated pointer. + CGM.ErrorUnsupported(propImpl, + "Obj-C getter requiring pointer authentication"); + return; + } + if (!getPropertyFn) { CGM.ErrorUnsupported(propImpl, "Obj-C getter requiring atomic copy"); return; @@ -1269,7 +1286,9 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl, LValue LV = EmitLValueForIvar(TypeOfSelfObject(), LoadObjCSelf(), ivar, 0); QualType ivarType = ivar->getType(); - switch (getEvaluationKind(ivarType)) { + auto EvaluationKind = getEvaluationKind(ivarType); + assert(!ivarType.getPointerAuth() || EvaluationKind == TEK_Scalar); + switch (EvaluationKind) { case TEK_Complex: { ComplexPairTy pair = EmitLoadOfComplex(LV, SourceLocation()); EmitStoreOfComplex(pair, MakeAddrLValue(ReturnValue, ivarType), @@ -1287,6 +1306,11 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl, case TEK_Scalar: { llvm::Value *value; if (propType->isReferenceType()) { + if (ivarType.getPointerAuth()) { + CGM.ErrorUnsupported(propImpl, + "Obj-C getter for authenticated reference type"); + return; + } value = LV.getAddress().emitRawPointer(*this); } else { // We want to load and autoreleaseReturnValue ARC __weak ivars. @@ -1300,7 +1324,19 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl, // Otherwise we want to do a simple load, suppressing the // final autorelease. } else { - value = EmitLoadOfLValue(LV, SourceLocation()).getScalarVal(); + if (PointerAuthQualifier PAQ = ivar->getType().getPointerAuth()) { + Address ivarAddr = LV.getAddress(); + llvm::LoadInst *LoadInst = Builder.CreateLoad(ivarAddr, "load"); + llvm::Value *Load = LoadInst; + auto SrcInfo = EmitPointerAuthInfo(PAQ, ivarAddr); + auto TargetInfo = + CGM.getPointerAuthInfoForType(getterMethod->getReturnType()); + Load = emitPointerAuthResign(Load, ivarType, SrcInfo, TargetInfo, + /*isKnownNonNull=*/false); + value = Load; + } else + value = EmitLoadOfLValue(LV, SourceLocation()).getScalarVal(); + AutoreleaseResult = false; } @@ -1490,6 +1526,14 @@ CodeGenFunction::generateObjCSetterBody(const ObjCImplementationDecl *classImpl, llvm::Value *load = Builder.CreateLoad(argAddr); + if (PointerAuthQualifier PAQ = ivar->getType().getPointerAuth()) { + QualType PropertyType = propImpl->getPropertyDecl()->getType(); + CGPointerAuthInfo SrcInfo = CGM.getPointerAuthInfoForType(PropertyType); + CGPointerAuthInfo TargetInfo = EmitPointerAuthInfo(PAQ, ivarAddr); + load = emitPointerAuthResign(load, ivar->getType(), SrcInfo, TargetInfo, + /*isKnownNonNull=*/false); + } + // Perform an atomic store. There are no memory ordering requirements. llvm::StoreInst *store = Builder.CreateStore(load, ivarAddr); store->setAtomic(llvm::AtomicOrdering::Unordered); diff --git a/clang/lib/CodeGen/CGObjCGNU.cpp b/clang/lib/CodeGen/CGObjCGNU.cpp index d828702cbb87..8acf8d2ddec0 100644 --- a/clang/lib/CodeGen/CGObjCGNU.cpp +++ b/clang/lib/CodeGen/CGObjCGNU.cpp @@ -1942,8 +1942,9 @@ class CGObjCGNUstep2 : public CGObjCGNUstep { // struct objc_class *sibling_class classFields.addNullPointer(PtrTy); // struct objc_protocol_list *protocols; - auto RuntimeProtocols = GetRuntimeProtocolList(classDecl->protocol_begin(), - classDecl->protocol_end()); + auto RuntimeProtocols = + GetRuntimeProtocolList(classDecl->all_referenced_protocol_begin(), + classDecl->all_referenced_protocol_end()); SmallVector<llvm::Constant *, 16> Protocols; for (const auto *I : RuntimeProtocols) Protocols.push_back(GenerateProtocolRef(I)); diff --git a/clang/lib/CodeGen/CGObjCMac.cpp b/clang/lib/CodeGen/CGObjCMac.cpp index a52c92cdbc83..8e71a576552d 100644 --- a/clang/lib/CodeGen/CGObjCMac.cpp +++ b/clang/lib/CodeGen/CGObjCMac.cpp @@ -1935,7 +1935,9 @@ CGObjCCommonMac::GenerateConstantNSString(const StringLiteral *Literal) { auto Fields = Builder.beginStruct(NSConstantStringType); // Class pointer. - Fields.add(Class); + Fields.addSignedPointer(Class, + CGM.getCodeGenOpts().PointerAuth.ObjCIsaPointers, + GlobalDecl(), QualType()); // String pointer. llvm::Constant *C = @@ -4975,10 +4977,7 @@ enum ImageInfoFlags { eImageInfo_GCOnly = (1 << 2), eImageInfo_OptimizedByDyld = (1 << 3), // This flag is set by the dyld shared cache. - // A flag indicating that the module has no instances of a @synthesize of a - // superclass variable. This flag used to be consumed by the runtime to work - // around miscompile by gcc. - eImageInfo_CorrectedSynthesize = (1 << 4), // This flag is no longer set by clang. + eImageInfo_SignedClassRO = (1 << 4), // Reused (was _CorrectedSynthesize) eImageInfo_ImageIsSimulated = (1 << 5), eImageInfo_ClassProperties = (1 << 6) }; @@ -5036,6 +5035,17 @@ void CGObjCCommonMac::EmitImageInfo() { // Indicate whether we are generating class properties. Mod.addModuleFlag(llvm::Module::Error, "Objective-C Class Properties", eImageInfo_ClassProperties); + + // Indicate whether we want enforcement of pointer signing for class_ro_t + // pointers. + if (CGM.getLangOpts().PointerAuthObjcClassROPointers) + Mod.addModuleFlag(llvm::Module::Error, + "Objective-C Enforce ClassRO Pointer Signing", + eImageInfo_SignedClassRO); + else + Mod.addModuleFlag(llvm::Module::Error, + "Objective-C Enforce ClassRO Pointer Signing", + llvm::ConstantInt::get(Int8Ty, 0)); } // struct objc_module { @@ -6223,11 +6233,19 @@ llvm::GlobalVariable *CGObjCNonFragileABIMac::BuildClassRoTInitializer( methods.push_back(MD); } - values.add(emitMethodList(ID->getObjCRuntimeNameAsString(), - (flags & NonFragileABI_Class_Meta) - ? MethodListType::ClassMethods - : MethodListType::InstanceMethods, - methods)); + llvm::Constant *MethListPtr = emitMethodList( + ID->getObjCRuntimeNameAsString(), + (flags & NonFragileABI_Class_Meta) ? MethodListType::ClassMethods + : MethodListType::InstanceMethods, + methods); + + const PointerAuthSchema &MethListSchema = + CGM.getCodeGenOpts().PointerAuth.ObjCMethodListPointer; + if (!MethListPtr->isNullValue()) + values.addSignedPointer(MethListPtr, MethListSchema, GlobalDecl(), + QualType()); + else + values.add(MethListPtr); const ObjCInterfaceDecl *OID = ID->getClassInterface(); assert(OID && "CGObjCNonFragileABIMac::BuildClassRoTInitializer"); @@ -6275,15 +6293,20 @@ llvm::GlobalVariable *CGObjCNonFragileABIMac::BuildClassObject( bool HiddenVisibility) { ConstantInitBuilder builder(CGM); auto values = builder.beginStruct(ObjCTypes.ClassnfABITy); - values.add(IsAGV); - if (SuperClassGV) { - values.add(SuperClassGV); - } else { + const PointerAuthOptions &PointerAuthOpts = CGM.getCodeGenOpts().PointerAuth; + values.addSignedPointer(IsAGV, PointerAuthOpts.ObjCIsaPointers, GlobalDecl(), + QualType()); + if (SuperClassGV) + values.addSignedPointer(SuperClassGV, PointerAuthOpts.ObjCSuperPointers, + GlobalDecl(), QualType()); + else values.addNullPointer(ObjCTypes.ClassnfABIPtrTy); - } + values.add(ObjCEmptyCacheVar); values.add(ObjCEmptyVtableVar); - values.add(ClassRoGV); + + values.addSignedPointer(ClassRoGV, PointerAuthOpts.ObjCClassROPointers, + GlobalDecl(), QualType()); llvm::GlobalVariable *GV = cast<llvm::GlobalVariable>( GetClassGlobal(CI, isMetaclass, ForDefinition)); @@ -6543,15 +6566,27 @@ void CGObjCNonFragileABIMac::GenerateCategory(const ObjCCategoryImplDecl *OCD) { } } - auto instanceMethodList = emitMethodList( + llvm::Constant *InstanceMethodList = emitMethodList( listName, MethodListType::CategoryInstanceMethods, instanceMethods); - auto classMethodList = emitMethodList( + const PointerAuthSchema &MethListSchema = + CGM.getCodeGenOpts().PointerAuth.ObjCMethodListPointer; + if (!InstanceMethodList->isNullValue()) + values.addSignedPointer(InstanceMethodList, MethListSchema, GlobalDecl(), + QualType()); + else + values.add(InstanceMethodList); + + llvm::Constant *ClassMethodList = emitMethodList( listName, MethodListType::CategoryClassMethods, classMethods); - values.add(instanceMethodList); - values.add(classMethodList); + if (!ClassMethodList->isNullValue()) + values.addSignedPointer(ClassMethodList, MethListSchema, GlobalDecl(), + QualType()); + else + values.add(ClassMethodList); + // Keep track of whether we have actual metadata to emit. bool isEmptyCategory = - instanceMethodList->isNullValue() && classMethodList->isNullValue(); + InstanceMethodList->isNullValue() && ClassMethodList->isNullValue(); const ObjCCategoryDecl *Category = Interface->FindCategoryDeclaration(OCD->getIdentifier()); @@ -6629,7 +6664,13 @@ void CGObjCNonFragileABIMac::emitMethodConstant(ConstantArrayBuilder &builder, } else { llvm::Function *fn = GetMethodDefinition(MD); assert(fn && "no definition for method?"); - method.add(fn); + if (const PointerAuthSchema &Schema = + CGM.getCodeGenOpts().PointerAuth.ObjCMethodListFunctionPointers) { + llvm::Constant *Bitcast = + llvm::ConstantExpr::getBitCast(fn, ObjCTypes.Int8PtrProgramASTy); + method.addSignedPointer(Bitcast, Schema, GlobalDecl(), QualType()); + } else + method.add(fn); } method.finishAndAddTo(builder); @@ -7672,10 +7713,15 @@ CGObjCNonFragileABIMac::GetInterfaceEHType(const ObjCInterfaceDecl *ID, } llvm::Value *VTableIdx = llvm::ConstantInt::get(CGM.Int32Ty, 2); + llvm::Constant *VTablePtr = llvm::ConstantExpr::getInBoundsGetElementPtr( + VTableGV->getValueType(), VTableGV, VTableIdx); + ConstantInitBuilder builder(CGM); auto values = builder.beginStruct(ObjCTypes.EHTypeTy); - values.add(llvm::ConstantExpr::getInBoundsGetElementPtr( - VTableGV->getValueType(), VTableGV, VTableIdx)); + const PointerAuthSchema &TypeInfoSchema = + CGM.getCodeGenOpts().PointerAuth.CXXTypeInfoVTablePointer; + values.addSignedPointer(VTablePtr, TypeInfoSchema, GlobalDecl(), QualType()); + values.add(GetClassName(ClassName)); values.add(GetClassGlobal(ID, /*metaclass*/ false, NotForDefinition)); diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp index a5f2f0efa2c3..ce2dd4d76368 100644 --- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp +++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp @@ -1092,7 +1092,7 @@ emitCombinerOrInitializer(CodeGenModule &CGM, QualType Ty, auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage, Name, &CGM.getModule()); CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo); - if (CGM.getLangOpts().Optimize) { + if (CGM.getCodeGenOpts().OptimizationLevel != 0) { Fn->removeFnAttr(llvm::Attribute::NoInline); Fn->removeFnAttr(llvm::Attribute::OptimizeNone); Fn->addFnAttr(llvm::Attribute::AlwaysInline); @@ -3199,7 +3199,7 @@ emitTaskPrivateMappingFunction(CodeGenModule &CGM, SourceLocation Loc, &CGM.getModule()); CGM.SetInternalFunctionAttributes(GlobalDecl(), TaskPrivatesMap, TaskPrivatesMapFnInfo); - if (CGM.getLangOpts().Optimize) { + if (CGM.getCodeGenOpts().OptimizationLevel != 0) { TaskPrivatesMap->removeFnAttr(llvm::Attribute::NoInline); TaskPrivatesMap->removeFnAttr(llvm::Attribute::OptimizeNone); TaskPrivatesMap->addFnAttr(llvm::Attribute::AlwaysInline); diff --git a/clang/lib/CodeGen/CMakeLists.txt b/clang/lib/CodeGen/CMakeLists.txt index a05b31f971e1..0f2a352886e7 100644 --- a/clang/lib/CodeGen/CMakeLists.txt +++ b/clang/lib/CodeGen/CMakeLists.txt @@ -144,7 +144,6 @@ add_clang_library(clangCodeGen Targets/MSP430.cpp Targets/Mips.cpp Targets/NVPTX.cpp - Targets/PNaCl.cpp Targets/PPC.cpp Targets/RISCV.cpp Targets/SPIR.cpp diff --git a/clang/lib/CodeGen/CodeGenAction.cpp b/clang/lib/CodeGen/CodeGenAction.cpp index 5493cc92bd8b..eb5b6045510d 100644 --- a/clang/lib/CodeGen/CodeGenAction.cpp +++ b/clang/lib/CodeGen/CodeGenAction.cpp @@ -995,7 +995,7 @@ CodeGenAction::CreateASTConsumer(CompilerInstance &CI, StringRef InFile) { std::vector<std::unique_ptr<ASTConsumer>> Consumers(2); Consumers[0] = std::make_unique<ReducedBMIGenerator>( CI.getPreprocessor(), CI.getModuleCache(), - CI.getFrontendOpts().ModuleOutputPath); + CI.getFrontendOpts().ModuleOutputPath, CI.getCodeGenOpts()); Consumers[1] = std::move(Result); return std::make_unique<MultiplexConsumer>(std::move(Consumers)); } diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp index 776a646ceb32..0fda31c8e5fa 100644 --- a/clang/lib/CodeGen/CodeGenFunction.cpp +++ b/clang/lib/CodeGen/CodeGenFunction.cpp @@ -161,8 +161,7 @@ void CodeGenFunction::CGFPOptionsRAII::ConstructorHelper(FPOptions FPFeatures) { llvm::RoundingMode NewRoundingBehavior = FPFeatures.getRoundingMode(); CGF.Builder.setDefaultConstrainedRounding(NewRoundingBehavior); auto NewExceptionBehavior = - ToConstrainedExceptMD(static_cast<LangOptions::FPExceptionModeKind>( - FPFeatures.getExceptionMode())); + ToConstrainedExceptMD(FPFeatures.getExceptionMode()); CGF.Builder.setDefaultConstrainedExcept(NewExceptionBehavior); CGF.SetFastMathFlags(FPFeatures); diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp index c8866f15745c..236cc3d9e890 100644 --- a/clang/lib/CodeGen/CodeGenModule.cpp +++ b/clang/lib/CodeGen/CodeGenModule.cpp @@ -118,9 +118,7 @@ createTargetCodeGenInfo(CodeGenModule &CGM) { return createM68kTargetCodeGenInfo(CGM); case llvm::Triple::mips: case llvm::Triple::mipsel: - if (Triple.getOS() == llvm::Triple::NaCl) - return createPNaClTargetCodeGenInfo(CGM); - else if (Triple.getOS() == llvm::Triple::Win32) + if (Triple.getOS() == llvm::Triple::Win32) return createWindowsMIPSTargetCodeGenInfo(CGM, /*IsOS32=*/true); return createMIPSTargetCodeGenInfo(CGM, /*IsOS32=*/true); @@ -6616,7 +6614,9 @@ CodeGenModule::GetAddrOfConstantCFString(const StringLiteral *Literal) { auto Fields = Builder.beginStruct(STy); // Class pointer. - Fields.add(cast<llvm::Constant>(CFConstantStringClassRef)); + Fields.addSignedPointer(cast<llvm::Constant>(CFConstantStringClassRef), + getCodeGenOpts().PointerAuth.ObjCIsaPointers, + GlobalDecl(), QualType()); // Flags. if (IsSwiftABI) { diff --git a/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp b/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp index f09b3b92c4ea..ee736a281621 100644 --- a/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp +++ b/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp @@ -411,15 +411,18 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, case AMDGPU::BI__builtin_amdgcn_rcp: case AMDGPU::BI__builtin_amdgcn_rcpf: case AMDGPU::BI__builtin_amdgcn_rcph: + case AMDGPU::BI__builtin_amdgcn_rcp_bf16: return emitBuiltinWithOneOverloadedType<1>(*this, E, Intrinsic::amdgcn_rcp); case AMDGPU::BI__builtin_amdgcn_sqrt: case AMDGPU::BI__builtin_amdgcn_sqrtf: case AMDGPU::BI__builtin_amdgcn_sqrth: + case AMDGPU::BI__builtin_amdgcn_sqrt_bf16: return emitBuiltinWithOneOverloadedType<1>(*this, E, Intrinsic::amdgcn_sqrt); case AMDGPU::BI__builtin_amdgcn_rsq: case AMDGPU::BI__builtin_amdgcn_rsqf: case AMDGPU::BI__builtin_amdgcn_rsqh: + case AMDGPU::BI__builtin_amdgcn_rsq_bf16: return emitBuiltinWithOneOverloadedType<1>(*this, E, Intrinsic::amdgcn_rsq); case AMDGPU::BI__builtin_amdgcn_rsq_clamp: case AMDGPU::BI__builtin_amdgcn_rsq_clampf: @@ -427,15 +430,19 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, Intrinsic::amdgcn_rsq_clamp); case AMDGPU::BI__builtin_amdgcn_sinf: case AMDGPU::BI__builtin_amdgcn_sinh: + case AMDGPU::BI__builtin_amdgcn_sin_bf16: return emitBuiltinWithOneOverloadedType<1>(*this, E, Intrinsic::amdgcn_sin); case AMDGPU::BI__builtin_amdgcn_cosf: case AMDGPU::BI__builtin_amdgcn_cosh: + case AMDGPU::BI__builtin_amdgcn_cos_bf16: return emitBuiltinWithOneOverloadedType<1>(*this, E, Intrinsic::amdgcn_cos); case AMDGPU::BI__builtin_amdgcn_dispatch_ptr: return EmitAMDGPUDispatchPtr(*this, E); case AMDGPU::BI__builtin_amdgcn_logf: + case AMDGPU::BI__builtin_amdgcn_log_bf16: return emitBuiltinWithOneOverloadedType<1>(*this, E, Intrinsic::amdgcn_log); case AMDGPU::BI__builtin_amdgcn_exp2f: + case AMDGPU::BI__builtin_amdgcn_exp2_bf16: return emitBuiltinWithOneOverloadedType<1>(*this, E, Intrinsic::amdgcn_exp2); case AMDGPU::BI__builtin_amdgcn_log_clampf: @@ -497,6 +504,11 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_ballot, { ResultType }); return Builder.CreateCall(F, { Src }); } + case AMDGPU::BI__builtin_amdgcn_tanhf: + case AMDGPU::BI__builtin_amdgcn_tanhh: + case AMDGPU::BI__builtin_amdgcn_tanh_bf16: + return emitBuiltinWithOneOverloadedType<1>(*this, E, + Intrinsic::amdgcn_tanh); case AMDGPU::BI__builtin_amdgcn_uicmp: case AMDGPU::BI__builtin_amdgcn_uicmpl: case AMDGPU::BI__builtin_amdgcn_sicmp: @@ -818,7 +830,46 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x32_bf8_fp8_w32: case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x32_bf8_fp8_w64: case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x32_bf8_bf8_w32: - case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x32_bf8_bf8_w64: { + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x32_bf8_bf8_w64: + // GFX1250 WMMA builtins + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x4_f32: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x32_bf16: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x32_f16: + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x32_f16: + case AMDGPU::BI__builtin_amdgcn_wmma_bf16_16x16x32_bf16: + case AMDGPU::BI__builtin_amdgcn_wmma_bf16f32_16x16x32_bf16: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x64_fp8_fp8: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x64_fp8_bf8: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x64_bf8_fp8: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x64_bf8_bf8: + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x64_fp8_fp8: + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x64_fp8_bf8: + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x64_bf8_fp8: + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x64_bf8_bf8: + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x128_fp8_fp8: + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x128_fp8_bf8: + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x128_bf8_fp8: + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x128_bf8_bf8: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x128_fp8_fp8: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x128_fp8_bf8: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x128_bf8_fp8: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x128_bf8_bf8: + case AMDGPU::BI__builtin_amdgcn_wmma_i32_16x16x64_iu8: + case AMDGPU::BI__builtin_amdgcn_wmma_f32_32x16x128_f4: + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x64_f16: + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x64_bf16: + case AMDGPU::BI__builtin_amdgcn_swmmac_f16_16x16x64_f16: + case AMDGPU::BI__builtin_amdgcn_swmmac_bf16_16x16x64_bf16: + case AMDGPU::BI__builtin_amdgcn_swmmac_bf16f32_16x16x64_bf16: + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x128_fp8_fp8: + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x128_fp8_bf8: + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x128_bf8_fp8: + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x128_bf8_bf8: + case AMDGPU::BI__builtin_amdgcn_swmmac_f16_16x16x128_fp8_fp8: + case AMDGPU::BI__builtin_amdgcn_swmmac_f16_16x16x128_fp8_bf8: + case AMDGPU::BI__builtin_amdgcn_swmmac_f16_16x16x128_bf8_fp8: + case AMDGPU::BI__builtin_amdgcn_swmmac_f16_16x16x128_bf8_bf8: + case AMDGPU::BI__builtin_amdgcn_swmmac_i32_16x16x128_iu8: { // These operations perform a matrix multiplication and accumulation of // the form: @@ -833,6 +884,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, // "false". bool AppendFalseForOpselArg = false; unsigned BuiltinWMMAOp; + // Need return type when D and C are of different types. + bool NeedReturnType = false; switch (BuiltinID) { case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x16_f16_w32: @@ -971,6 +1024,160 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, ArgsForMatchingMatrixTypes = {2, 0, 1, 3}; // CD, A, B, Index BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f32_16x16x32_bf8_bf8; break; + // GFX1250 WMMA builtins + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x4_f32: + ArgsForMatchingMatrixTypes = {5, 1}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x4_f32; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x32_bf16: + ArgsForMatchingMatrixTypes = {5, 1}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x32_bf16; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x32_f16: + ArgsForMatchingMatrixTypes = {5, 1}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x32_f16; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x32_f16: + ArgsForMatchingMatrixTypes = {5, 1}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f16_16x16x32_f16; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_bf16_16x16x32_bf16: + ArgsForMatchingMatrixTypes = {5, 1}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_bf16_16x16x32_bf16; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_bf16f32_16x16x32_bf16: + NeedReturnType = true; + ArgsForMatchingMatrixTypes = {1, 5}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_bf16f32_16x16x32_bf16; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x64_fp8_fp8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x64_fp8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x64_fp8_bf8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x64_fp8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x64_bf8_fp8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x64_bf8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x64_bf8_bf8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x64_bf8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x64_fp8_fp8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f16_16x16x64_fp8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x64_fp8_bf8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f16_16x16x64_fp8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x64_bf8_fp8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f16_16x16x64_bf8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x64_bf8_bf8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f16_16x16x64_bf8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x128_fp8_fp8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f16_16x16x128_fp8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x128_fp8_bf8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f16_16x16x128_fp8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x128_bf8_fp8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f16_16x16x128_bf8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f16_16x16x128_bf8_bf8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f16_16x16x128_bf8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x128_fp8_fp8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x128_fp8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x128_fp8_bf8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x128_fp8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x128_bf8_fp8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x128_bf8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_16x16x128_bf8_bf8: + ArgsForMatchingMatrixTypes = {3, 0}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_16x16x128_bf8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_i32_16x16x64_iu8: + ArgsForMatchingMatrixTypes = {4, 1}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_i32_16x16x64_iu8; + break; + case AMDGPU::BI__builtin_amdgcn_wmma_f32_32x16x128_f4: + ArgsForMatchingMatrixTypes = {3, 0, 1}; + BuiltinWMMAOp = Intrinsic::amdgcn_wmma_f32_32x16x128_f4; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x64_f16: + ArgsForMatchingMatrixTypes = {4, 1, 3, 5}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f32_16x16x64_f16; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x64_bf16: + ArgsForMatchingMatrixTypes = {4, 1, 3, 5}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f32_16x16x64_bf16; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f16_16x16x64_f16: + ArgsForMatchingMatrixTypes = {4, 1, 3, 5}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f16_16x16x64_f16; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_bf16_16x16x64_bf16: + ArgsForMatchingMatrixTypes = {4, 1, 3, 5}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_bf16_16x16x64_bf16; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_bf16f32_16x16x64_bf16: + ArgsForMatchingMatrixTypes = {4, 1, 3, 5}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_bf16f32_16x16x64_bf16; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x128_fp8_fp8: + ArgsForMatchingMatrixTypes = {2, 0, 1, 3}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f32_16x16x128_fp8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x128_fp8_bf8: + ArgsForMatchingMatrixTypes = {2, 0, 1, 3}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f32_16x16x128_fp8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x128_bf8_fp8: + ArgsForMatchingMatrixTypes = {2, 0, 1, 3}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f32_16x16x128_bf8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f32_16x16x128_bf8_bf8: + ArgsForMatchingMatrixTypes = {2, 0, 1, 3}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f32_16x16x128_bf8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f16_16x16x128_fp8_fp8: + ArgsForMatchingMatrixTypes = {2, 0, 1, 3}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f16_16x16x128_fp8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f16_16x16x128_fp8_bf8: + ArgsForMatchingMatrixTypes = {2, 0, 1, 3}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f16_16x16x128_fp8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f16_16x16x128_bf8_fp8: + ArgsForMatchingMatrixTypes = {2, 0, 1, 3}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f16_16x16x128_bf8_fp8; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_f16_16x16x128_bf8_bf8: + ArgsForMatchingMatrixTypes = {2, 0, 1, 3}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_f16_16x16x128_bf8_bf8; + break; + case AMDGPU::BI__builtin_amdgcn_swmmac_i32_16x16x128_iu8: + ArgsForMatchingMatrixTypes = {4, 1, 3, 5}; + BuiltinWMMAOp = Intrinsic::amdgcn_swmmac_i32_16x16x128_iu8; + break; } SmallVector<Value *, 6> Args; @@ -980,6 +1187,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, Args.push_back(Builder.getFalse()); SmallVector<llvm::Type *, 6> ArgTypes; + if (NeedReturnType) + ArgTypes.push_back(ConvertType(E->getType())); for (auto ArgIdx : ArgsForMatchingMatrixTypes) ArgTypes.push_back(Args[ArgIdx]->getType()); diff --git a/clang/lib/CodeGen/TargetBuiltins/PPC.cpp b/clang/lib/CodeGen/TargetBuiltins/PPC.cpp index f9890285f0aa..270e9fc976f2 100644 --- a/clang/lib/CodeGen/TargetBuiltins/PPC.cpp +++ b/clang/lib/CodeGen/TargetBuiltins/PPC.cpp @@ -1151,6 +1151,11 @@ Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID, Value *Acc = Builder.CreateLoad(Addr); CallOps.push_back(Acc); } + if (BuiltinID == PPC::BI__builtin_mma_dmmr || + BuiltinID == PPC::BI__builtin_mma_dmxor) { + Address Addr = EmitPointerWithAlignment(E->getArg(1)); + Ops[1] = Builder.CreateLoad(Addr); + } for (unsigned i=1; i<Ops.size(); i++) CallOps.push_back(Ops[i]); llvm::Function *F = CGM.getIntrinsic(ID); diff --git a/clang/lib/CodeGen/TargetBuiltins/RISCV.cpp b/clang/lib/CodeGen/TargetBuiltins/RISCV.cpp index 89e3f6f203df..b08a0588c5ac 100644 --- a/clang/lib/CodeGen/TargetBuiltins/RISCV.cpp +++ b/clang/lib/CodeGen/TargetBuiltins/RISCV.cpp @@ -413,6 +413,12 @@ Value *CodeGenFunction::EmitRISCVBuiltinExpr(unsigned BuiltinID, ID = Intrinsic::riscv_cv_alu_subuRN; break; + // XAndesBFHCvt + case RISCV::BI__builtin_riscv_nds_fcvt_s_bf16: + return Builder.CreateFPExt(Ops[0], FloatTy); + case RISCV::BI__builtin_riscv_nds_fcvt_bf16_s: + return Builder.CreateFPTrunc(Ops[0], BFloatTy); + // Vector builtins are handled from here. #include "clang/Basic/riscv_vector_builtin_cg.inc" diff --git a/clang/lib/CodeGen/TargetBuiltins/SPIR.cpp b/clang/lib/CodeGen/TargetBuiltins/SPIR.cpp index 16243951c7be..243aad8bf708 100644 --- a/clang/lib/CodeGen/TargetBuiltins/SPIR.cpp +++ b/clang/lib/CodeGen/TargetBuiltins/SPIR.cpp @@ -58,6 +58,18 @@ Value *CodeGenFunction::EmitSPIRVBuiltinExpr(unsigned BuiltinID, /*ReturnType=*/I->getType(), Intrinsic::spv_reflect, ArrayRef<Value *>{I, N}, nullptr, "spv.reflect"); } + case SPIRV::BI__builtin_spirv_refract: { + Value *I = EmitScalarExpr(E->getArg(0)); + Value *N = EmitScalarExpr(E->getArg(1)); + Value *eta = EmitScalarExpr(E->getArg(2)); + assert(E->getArg(0)->getType()->hasFloatingRepresentation() && + E->getArg(1)->getType()->hasFloatingRepresentation() && + E->getArg(2)->getType()->isFloatingType() && + "refract operands must have a float representation"); + return Builder.CreateIntrinsic( + /*ReturnType=*/I->getType(), Intrinsic::spv_refract, + ArrayRef<Value *>{I, N, eta}, nullptr, "spv.refract"); + } case SPIRV::BI__builtin_spirv_smoothstep: { Value *Min = EmitScalarExpr(E->getArg(0)); Value *Max = EmitScalarExpr(E->getArg(1)); diff --git a/clang/lib/CodeGen/TargetInfo.h b/clang/lib/CodeGen/TargetInfo.h index b4057d369f98..d0edae129509 100644 --- a/clang/lib/CodeGen/TargetInfo.h +++ b/clang/lib/CodeGen/TargetInfo.h @@ -545,9 +545,6 @@ createMSP430TargetCodeGenInfo(CodeGenModule &CGM); std::unique_ptr<TargetCodeGenInfo> createNVPTXTargetCodeGenInfo(CodeGenModule &CGM); -std::unique_ptr<TargetCodeGenInfo> -createPNaClTargetCodeGenInfo(CodeGenModule &CGM); - enum class PPC64_SVR4_ABIKind { ELFv1 = 0, ELFv2, diff --git a/clang/lib/CodeGen/Targets/PNaCl.cpp b/clang/lib/CodeGen/Targets/PNaCl.cpp deleted file mode 100644 index 358010785850..000000000000 --- a/clang/lib/CodeGen/Targets/PNaCl.cpp +++ /dev/null @@ -1,114 +0,0 @@ -//===- PNaCl.cpp ----------------------------------------------------------===// -// -// 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 -// -//===----------------------------------------------------------------------===// - -#include "ABIInfoImpl.h" -#include "TargetInfo.h" - -using namespace clang; -using namespace clang::CodeGen; - -//===----------------------------------------------------------------------===// -// le32/PNaCl bitcode ABI Implementation -// -// This is a simplified version of the x86_32 ABI. Arguments and return values -// are always passed on the stack. -//===----------------------------------------------------------------------===// - -class PNaClABIInfo : public ABIInfo { - public: - PNaClABIInfo(CodeGen::CodeGenTypes &CGT) : ABIInfo(CGT) {} - - ABIArgInfo classifyReturnType(QualType RetTy) const; - ABIArgInfo classifyArgumentType(QualType RetTy) const; - - void computeInfo(CGFunctionInfo &FI) const override; - RValue EmitVAArg(CodeGenFunction &CGF, Address VAListAddr, QualType Ty, - AggValueSlot Slot) const override; -}; - -class PNaClTargetCodeGenInfo : public TargetCodeGenInfo { - public: - PNaClTargetCodeGenInfo(CodeGen::CodeGenTypes &CGT) - : TargetCodeGenInfo(std::make_unique<PNaClABIInfo>(CGT)) {} -}; - -void PNaClABIInfo::computeInfo(CGFunctionInfo &FI) const { - if (!getCXXABI().classifyReturnType(FI)) - FI.getReturnInfo() = classifyReturnType(FI.getReturnType()); - - for (auto &I : FI.arguments()) - I.info = classifyArgumentType(I.type); -} - -RValue PNaClABIInfo::EmitVAArg(CodeGenFunction &CGF, Address VAListAddr, - QualType Ty, AggValueSlot Slot) const { - // The PNaCL ABI is a bit odd, in that varargs don't use normal - // function classification. Structs get passed directly for varargs - // functions, through a rewriting transform in - // pnacl-llvm/lib/Transforms/NaCl/ExpandVarArgs.cpp, which allows - // this target to actually support a va_arg instructions with an - // aggregate type, unlike other targets. - return CGF.EmitLoadOfAnyValue( - CGF.MakeAddrLValue( - EmitVAArgInstr(CGF, VAListAddr, Ty, ABIArgInfo::getDirect()), Ty), - Slot); -} - -/// Classify argument of given type \p Ty. -ABIArgInfo PNaClABIInfo::classifyArgumentType(QualType Ty) const { - if (isAggregateTypeForABI(Ty)) { - if (CGCXXABI::RecordArgABI RAA = getRecordArgABI(Ty, getCXXABI())) - return getNaturalAlignIndirect(Ty, getDataLayout().getAllocaAddrSpace(), - RAA == CGCXXABI::RAA_DirectInMemory); - return getNaturalAlignIndirect(Ty, getDataLayout().getAllocaAddrSpace()); - } else if (const EnumType *EnumTy = Ty->getAs<EnumType>()) { - // Treat an enum type as its underlying type. - Ty = EnumTy->getDecl()->getIntegerType(); - } else if (Ty->isFloatingType()) { - // Floating-point types don't go inreg. - return ABIArgInfo::getDirect(); - } else if (const auto *EIT = Ty->getAs<BitIntType>()) { - // Treat bit-precise integers as integers if <= 64, otherwise pass - // indirectly. - if (EIT->getNumBits() > 64) - return getNaturalAlignIndirect(Ty, getDataLayout().getAllocaAddrSpace()); - return ABIArgInfo::getDirect(); - } - - return (isPromotableIntegerTypeForABI(Ty) ? ABIArgInfo::getExtend(Ty) - : ABIArgInfo::getDirect()); -} - -ABIArgInfo PNaClABIInfo::classifyReturnType(QualType RetTy) const { - if (RetTy->isVoidType()) - return ABIArgInfo::getIgnore(); - - // In the PNaCl ABI we always return records/structures on the stack. - if (isAggregateTypeForABI(RetTy)) - return getNaturalAlignIndirect(RetTy, getDataLayout().getAllocaAddrSpace()); - - // Treat bit-precise integers as integers if <= 64, otherwise pass indirectly. - if (const auto *EIT = RetTy->getAs<BitIntType>()) { - if (EIT->getNumBits() > 64) - return getNaturalAlignIndirect(RetTy, - getDataLayout().getAllocaAddrSpace()); - return ABIArgInfo::getDirect(); - } - - // Treat an enum type as its underlying type. - if (const EnumType *EnumTy = RetTy->getAs<EnumType>()) - RetTy = EnumTy->getDecl()->getIntegerType(); - - return (isPromotableIntegerTypeForABI(RetTy) ? ABIArgInfo::getExtend(RetTy) - : ABIArgInfo::getDirect()); -} - -std::unique_ptr<TargetCodeGenInfo> -CodeGen::createPNaClTargetCodeGenInfo(CodeGenModule &CGM) { - return std::make_unique<PNaClTargetCodeGenInfo>(CGM.getTypes()); -} diff --git a/clang/lib/CodeGen/Targets/X86.cpp b/clang/lib/CodeGen/Targets/X86.cpp index 0f59caac2323..abb91486e7ee 100644 --- a/clang/lib/CodeGen/Targets/X86.cpp +++ b/clang/lib/CodeGen/Targets/X86.cpp @@ -2470,13 +2470,12 @@ GetSSETypeAtOffset(llvm::Type *IRType, unsigned IROffset, return llvm::Type::getDoubleTy(getVMContext()); } - /// GetINTEGERTypeAtOffset - The ABI specifies that a value should be passed in -/// an 8-byte GPR. This means that we either have a scalar or we are talking -/// about the high or low part of an up-to-16-byte struct. This routine picks -/// the best LLVM IR type to represent this, which may be i64 or may be anything -/// else that the backend will pass in a GPR that works better (e.g. i8, %foo*, -/// etc). +/// one or more 8-byte GPRs. This means that we either have a scalar or we are +/// talking about the high and/or low part of an up-to-16-byte struct. This +/// routine picks the best LLVM IR type to represent this, which may be i64 or +/// may be anything else that the backend will pass in GPRs that works better +/// (e.g. i8, %foo*, etc). /// /// PrefType is an LLVM IR type that corresponds to (part of) the IR type for /// the source type. IROffset is an offset in bytes into the LLVM IR type that @@ -2534,6 +2533,13 @@ GetINTEGERTypeAtOffset(llvm::Type *IRType, unsigned IROffset, SourceOffset); } + // if we have a 128-bit integer, we can pass it safely using an i128 + // so we return that + if (IRType->isIntegerTy(128)) { + assert(IROffset == 0); + return IRType; + } + // Okay, we don't have any better idea of what to pass, so we pass this in an // integer register that isn't too big to fit the rest of the struct. unsigned TySizeInBytes = @@ -2572,8 +2578,7 @@ GetX86_64ByValArgumentPair(llvm::Type *Lo, llvm::Type *Hi, if (HiStart != 8) { // There are usually two sorts of types the ABI generation code can produce // for the low part of a pair that aren't 8 bytes in size: half, float or - // i8/i16/i32. This can also include pointers when they are 32-bit (X32 and - // NaCl). + // i8/i16/i32. This can also include pointers when they are 32-bit (X32). // Promote these to a larger type. if (Lo->isHalfTy() || Lo->isFloatTy()) Lo = llvm::Type::getDoubleTy(Lo->getContext()); @@ -2592,8 +2597,7 @@ GetX86_64ByValArgumentPair(llvm::Type *Lo, llvm::Type *Hi, return Result; } -ABIArgInfo X86_64ABIInfo:: -classifyReturnType(QualType RetTy) const { +ABIArgInfo X86_64ABIInfo::classifyReturnType(QualType RetTy) const { // AMD64-ABI 3.2.3p4: Rule 1. Classify the return type with the // classification algorithm. X86_64ABIInfo::Class Lo, Hi; @@ -2639,6 +2643,12 @@ classifyReturnType(QualType RetTy) const { isPromotableIntegerTypeForABI(RetTy)) return ABIArgInfo::getExtend(RetTy); } + + if (ResType->isIntegerTy(128)) { + // i128 are passed directly + assert(Hi == Integer); + return ABIArgInfo::getDirect(ResType); + } break; // AMD64-ABI 3.2.3p4: Rule 4. If the class is SSE, the next @@ -2784,6 +2794,11 @@ X86_64ABIInfo::classifyArgumentType(QualType Ty, unsigned freeIntRegs, return ABIArgInfo::getExtend(Ty, CGT.ConvertType(Ty)); } + if (ResType->isIntegerTy(128)) { + assert(Hi == Integer); + ++neededInt; + return ABIArgInfo::getDirect(ResType); + } break; // AMD64-ABI 3.2.3p3: Rule 3. If the class is SSE, the next |