head 1.1; branch 1.1.1; access; symbols netbsd-11-0-RC4:1.1.1.15 netbsd-11-0-RC3:1.1.1.15 netbsd-11-0-RC2:1.1.1.15 netbsd-11-0-RC1:1.1.1.15 perseant-exfatfs-base-20250801:1.1.1.15 netbsd-11:1.1.1.15.0.10 netbsd-11-base:1.1.1.15 netbsd-10-1-RELEASE:1.1.1.15 perseant-exfatfs-base-20240630:1.1.1.15 perseant-exfatfs:1.1.1.15.0.8 perseant-exfatfs-base:1.1.1.15 netbsd-8-3-RELEASE:1.1.1.11 netbsd-9-4-RELEASE:1.1.1.14 netbsd-10-0-RELEASE:1.1.1.15 netbsd-10-0-RC6:1.1.1.15 netbsd-10-0-RC5:1.1.1.15 netbsd-10-0-RC4:1.1.1.15 netbsd-10-0-RC3:1.1.1.15 netbsd-10-0-RC2:1.1.1.15 netbsd-10-0-RC1:1.1.1.15 netbsd-10:1.1.1.15.0.6 netbsd-10-base:1.1.1.15 netbsd-9-3-RELEASE:1.1.1.14 cjep_sun2x:1.1.1.15.0.4 cjep_sun2x-base:1.1.1.15 cjep_staticlib_x-base1:1.1.1.15 netbsd-9-2-RELEASE:1.1.1.14 cjep_staticlib_x:1.1.1.15.0.2 cjep_staticlib_x-base:1.1.1.15 netbsd-9-1-RELEASE:1.1.1.14 phil-wifi-20200421:1.1.1.15 phil-wifi-20200411:1.1.1.15 phil-wifi-20200406:1.1.1.15 netbsd-8-2-RELEASE:1.1.1.11 netbsd-9-0-RELEASE:1.1.1.14 netbsd-9-0-RC2:1.1.1.14 netbsd-9-0-RC1:1.1.1.14 netbsd-9:1.1.1.14.0.2 netbsd-9-base:1.1.1.14 phil-wifi-20190609:1.1.1.14 netbsd-8-1-RELEASE:1.1.1.11 netbsd-8-1-RC1:1.1.1.11 pgoyette-compat-merge-20190127:1.1.1.13.2.1 pgoyette-compat-20190127:1.1.1.14 pgoyette-compat-20190118:1.1.1.14 pgoyette-compat-1226:1.1.1.14 pgoyette-compat-1126:1.1.1.14 pgoyette-compat-1020:1.1.1.14 pgoyette-compat-0930:1.1.1.14 pgoyette-compat-0906:1.1.1.14 netbsd-7-2-RELEASE:1.1.1.7.2.1 pgoyette-compat-0728:1.1.1.14 clang-337282:1.1.1.14 netbsd-8-0-RELEASE:1.1.1.11 phil-wifi:1.1.1.13.0.4 phil-wifi-base:1.1.1.13 pgoyette-compat-0625:1.1.1.13 netbsd-8-0-RC2:1.1.1.11 pgoyette-compat-0521:1.1.1.13 pgoyette-compat-0502:1.1.1.13 pgoyette-compat-0422:1.1.1.13 netbsd-8-0-RC1:1.1.1.11 pgoyette-compat-0415:1.1.1.13 pgoyette-compat-0407:1.1.1.13 pgoyette-compat-0330:1.1.1.13 pgoyette-compat-0322:1.1.1.13 pgoyette-compat-0315:1.1.1.13 netbsd-7-1-2-RELEASE:1.1.1.7.2.1 pgoyette-compat:1.1.1.13.0.2 pgoyette-compat-base:1.1.1.13 netbsd-7-1-1-RELEASE:1.1.1.7.2.1 clang-319952:1.1.1.13 matt-nb8-mediatek:1.1.1.11.0.8 matt-nb8-mediatek-base:1.1.1.11 clang-309604:1.1.1.12 perseant-stdc-iso10646:1.1.1.11.0.6 perseant-stdc-iso10646-base:1.1.1.11 netbsd-8:1.1.1.11.0.4 netbsd-8-base:1.1.1.11 prg-localcount2-base3:1.1.1.11 prg-localcount2-base2:1.1.1.11 prg-localcount2-base1:1.1.1.11 prg-localcount2:1.1.1.11.0.2 prg-localcount2-base:1.1.1.11 pgoyette-localcount-20170426:1.1.1.11 bouyer-socketcan-base1:1.1.1.11 pgoyette-localcount-20170320:1.1.1.11 netbsd-7-1:1.1.1.7.2.1.0.6 netbsd-7-1-RELEASE:1.1.1.7.2.1 netbsd-7-1-RC2:1.1.1.7.2.1 clang-294123:1.1.1.11 netbsd-7-nhusb-base-20170116:1.1.1.7.2.1 bouyer-socketcan:1.1.1.10.0.2 bouyer-socketcan-base:1.1.1.10 clang-291444:1.1.1.10 pgoyette-localcount-20170107:1.1.1.9 netbsd-7-1-RC1:1.1.1.7.2.1 pgoyette-localcount-20161104:1.1.1.9 netbsd-7-0-2-RELEASE:1.1.1.7.2.1 localcount-20160914:1.1.1.9 netbsd-7-nhusb:1.1.1.7.2.1.0.4 netbsd-7-nhusb-base:1.1.1.7.2.1 clang-280599:1.1.1.9 pgoyette-localcount-20160806:1.1.1.9 pgoyette-localcount-20160726:1.1.1.9 pgoyette-localcount:1.1.1.9.0.2 pgoyette-localcount-base:1.1.1.9 netbsd-7-0-1-RELEASE:1.1.1.7.2.1 clang-261930:1.1.1.9 netbsd-7-0:1.1.1.7.2.1.0.2 netbsd-7-0-RELEASE:1.1.1.7.2.1 netbsd-7-0-RC3:1.1.1.7.2.1 netbsd-7-0-RC2:1.1.1.7.2.1 netbsd-7-0-RC1:1.1.1.7.2.1 clang-237755:1.1.1.8 clang-232565:1.1.1.8 clang-227398:1.1.1.8 tls-maxphys-base:1.1.1.7 tls-maxphys:1.1.1.7.0.4 netbsd-7:1.1.1.7.0.2 netbsd-7-base:1.1.1.7 clang-215315:1.1.1.7 clang-209886:1.1.1.6 yamt-pagecache:1.1.1.5.0.4 yamt-pagecache-base9:1.1.1.5 tls-earlyentropy:1.1.1.5.0.2 tls-earlyentropy-base:1.1.1.6 riastradh-xf86-video-intel-2-7-1-pre-2-21-15:1.1.1.5 riastradh-drm2-base3:1.1.1.5 clang-202566:1.1.1.5 clang-201163:1.1.1.4 clang-199312:1.1.1.3 clang-198450:1.1.1.2 clang-196603:1.1.1.1 clang-195771:1.1.1.1 LLVM:1.1.1; locks; strict; comment @// @; 1.1 date 2013.11.28.14.14.53; author joerg; state Exp; branches 1.1.1.1; next ; commitid ow8OybrawrB1f3fx; 1.1.1.1 date 2013.11.28.14.14.53; author joerg; state Exp; branches; next 1.1.1.2; commitid ow8OybrawrB1f3fx; 1.1.1.2 date 2014.01.05.15.38.09; author joerg; state Exp; branches; next 1.1.1.3; commitid wh3aCSIWykURqWjx; 1.1.1.3 date 2014.01.15.21.26.26; author joerg; state Exp; branches; next 1.1.1.4; commitid NQXlzzA0SPkc5glx; 1.1.1.4 date 2014.02.14.20.07.10; author joerg; state Exp; branches; next 1.1.1.5; commitid annVkZ1sc17rF6px; 1.1.1.5 date 2014.03.04.19.53.53; author joerg; state Exp; branches 1.1.1.5.2.1 1.1.1.5.4.1; next 1.1.1.6; commitid 29z1hJonZISIXprx; 1.1.1.6 date 2014.05.30.18.14.41; author joerg; state Exp; branches; next 1.1.1.7; commitid 8q0kdlBlCn09GACx; 1.1.1.7 date 2014.08.10.17.08.36; author joerg; state Exp; branches 1.1.1.7.2.1 1.1.1.7.4.1; next 1.1.1.8; commitid N85tXAN6Ex9VZPLx; 1.1.1.8 date 2015.01.29.19.57.31; author joerg; state Exp; branches; next 1.1.1.9; commitid mlISSizlPKvepX7y; 1.1.1.9 date 2016.02.27.22.12.10; author joerg; state Exp; branches 1.1.1.9.2.1; next 1.1.1.10; commitid tIimz3oDlh1NpBWy; 1.1.1.10 date 2017.01.11.10.33.32; author joerg; state Exp; branches 1.1.1.10.2.1; next 1.1.1.11; commitid CNnUNfII1jgNmxBz; 1.1.1.11 date 2017.02.09.17.38.03; author joerg; state Exp; branches; next 1.1.1.12; commitid UxB8JMFWM7xyMiFz; 1.1.1.12 date 2017.08.01.19.35.33; author joerg; state Exp; branches; next 1.1.1.13; commitid pMuDy65V0VicSx1A; 1.1.1.13 date 2017.12.17.20.59.10; author joerg; state Exp; branches 1.1.1.13.2.1 1.1.1.13.4.1; next 1.1.1.14; commitid T7p9sjiUJyY1UhjA; 1.1.1.14 date 2018.07.17.18.31.00; author joerg; state Exp; branches; next 1.1.1.15; commitid wDzL46ALjrCZgwKA; 1.1.1.15 date 2019.11.13.22.19.21; author joerg; state dead; branches; next ; commitid QD8YATxuNG34YJKB; 1.1.1.5.2.1 date 2014.08.10.07.08.07; author tls; state Exp; branches; next ; commitid t01A1TLTYxkpGMLx; 1.1.1.5.4.1 date 2014.03.04.19.53.53; author yamt; state dead; branches; next 1.1.1.5.4.2; commitid WSrDtL5nYAUyiyBx; 1.1.1.5.4.2 date 2014.05.22.16.18.27; author yamt; state Exp; branches; next ; commitid WSrDtL5nYAUyiyBx; 1.1.1.7.2.1 date 2015.06.04.20.04.28; author snj; state Exp; branches; next ; commitid yRnjq9fueSo6n9oy; 1.1.1.7.4.1 date 2014.08.10.17.08.36; author tls; state dead; branches; next 1.1.1.7.4.2; commitid jTnpym9Qu0o4R1Nx; 1.1.1.7.4.2 date 2014.08.19.23.47.27; author tls; state Exp; branches; next ; commitid jTnpym9Qu0o4R1Nx; 1.1.1.9.2.1 date 2017.03.20.06.52.37; author pgoyette; state Exp; branches; next ; commitid jjw7cAwgyKq7RfKz; 1.1.1.10.2.1 date 2017.04.21.16.51.38; author bouyer; state Exp; branches; next ; commitid dUG7nkTKALCadqOz; 1.1.1.13.2.1 date 2018.07.28.04.33.18; author pgoyette; state Exp; branches; next ; commitid 1UP1xAIUxv1ZgRLA; 1.1.1.13.4.1 date 2019.06.10.21.45.22; author christos; state Exp; branches; next 1.1.1.13.4.2; commitid jtc8rnCzWiEEHGqB; 1.1.1.13.4.2 date 2020.04.13.07.46.32; author martin; state dead; branches; next ; commitid X01YhRUPVUDaec4C; desc @@ 1.1 log @Initial revision @ text @//===--- CodeGenFunction.cpp - Emit LLVM Code from ASTs for a Function ----===// // // The LLVM Compiler Infrastructure // // This file is distributed under the University of Illinois Open Source // License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// // // This coordinates the per-function state used while generating code. // //===----------------------------------------------------------------------===// #include "CodeGenFunction.h" #include "CGCUDARuntime.h" #include "CGCXXABI.h" #include "CGDebugInfo.h" #include "CodeGenModule.h" #include "TargetInfo.h" #include "clang/AST/ASTContext.h" #include "clang/AST/Decl.h" #include "clang/AST/DeclCXX.h" #include "clang/AST/StmtCXX.h" #include "clang/Basic/OpenCL.h" #include "clang/Basic/TargetInfo.h" #include "clang/CodeGen/CGFunctionInfo.h" #include "clang/Frontend/CodeGenOptions.h" #include "llvm/IR/DataLayout.h" #include "llvm/IR/Intrinsics.h" #include "llvm/IR/MDBuilder.h" #include "llvm/IR/Operator.h" using namespace clang; using namespace CodeGen; CodeGenFunction::CodeGenFunction(CodeGenModule &cgm, bool suppressNewContext) : CodeGenTypeCache(cgm), CGM(cgm), Target(cgm.getTarget()), Builder(cgm.getModule().getContext()), CapturedStmtInfo(0), SanitizePerformTypeCheck(CGM.getSanOpts().Null | CGM.getSanOpts().Alignment | CGM.getSanOpts().ObjectSize | CGM.getSanOpts().Vptr), SanOpts(&CGM.getSanOpts()), AutoreleaseResult(false), BlockInfo(0), BlockPointer(0), LambdaThisCaptureField(0), NormalCleanupDest(0), NextCleanupDestIndex(1), FirstBlockInfo(0), EHResumeBlock(0), ExceptionSlot(0), EHSelectorSlot(0), DebugInfo(CGM.getModuleDebugInfo()), DisableDebugInfo(false), DidCallStackSave(false), IndirectBranch(0), SwitchInsn(0), CaseRangeBlock(0), UnreachableBlock(0), NumReturnExprs(0), NumSimpleReturnExprs(0), CXXABIThisDecl(0), CXXABIThisValue(0), CXXThisValue(0), CXXDefaultInitExprThis(0), CXXStructorImplicitParamDecl(0), CXXStructorImplicitParamValue(0), OutermostConditional(0), CurLexicalScope(0), TerminateLandingPad(0), TerminateHandler(0), TrapBB(0) { if (!suppressNewContext) CGM.getCXXABI().getMangleContext().startNewFunction(); llvm::FastMathFlags FMF; if (CGM.getLangOpts().FastMath) FMF.setUnsafeAlgebra(); if (CGM.getLangOpts().FiniteMathOnly) { FMF.setNoNaNs(); FMF.setNoInfs(); } Builder.SetFastMathFlags(FMF); } CodeGenFunction::~CodeGenFunction() { assert(LifetimeExtendedCleanupStack.empty() && "failed to emit a cleanup"); // If there are any unclaimed block infos, go ahead and destroy them // now. This can happen if IR-gen gets clever and skips evaluating // something. if (FirstBlockInfo) destroyBlockInfos(FirstBlockInfo); } llvm::Type *CodeGenFunction::ConvertTypeForMem(QualType T) { return CGM.getTypes().ConvertTypeForMem(T); } llvm::Type *CodeGenFunction::ConvertType(QualType T) { return CGM.getTypes().ConvertType(T); } TypeEvaluationKind CodeGenFunction::getEvaluationKind(QualType type) { type = type.getCanonicalType(); while (true) { switch (type->getTypeClass()) { #define TYPE(name, parent) #define ABSTRACT_TYPE(name, parent) #define NON_CANONICAL_TYPE(name, parent) case Type::name: #define DEPENDENT_TYPE(name, parent) case Type::name: #define NON_CANONICAL_UNLESS_DEPENDENT_TYPE(name, parent) case Type::name: #include "clang/AST/TypeNodes.def" llvm_unreachable("non-canonical or dependent type in IR-generation"); case Type::Auto: llvm_unreachable("undeduced auto type in IR-generation"); // Various scalar types. case Type::Builtin: case Type::Pointer: case Type::BlockPointer: case Type::LValueReference: case Type::RValueReference: case Type::MemberPointer: case Type::Vector: case Type::ExtVector: case Type::FunctionProto: case Type::FunctionNoProto: case Type::Enum: case Type::ObjCObjectPointer: return TEK_Scalar; // Complexes. case Type::Complex: return TEK_Complex; // Arrays, records, and Objective-C objects. case Type::ConstantArray: case Type::IncompleteArray: case Type::VariableArray: case Type::Record: case Type::ObjCObject: case Type::ObjCInterface: return TEK_Aggregate; // We operate on atomic values according to their underlying type. case Type::Atomic: type = cast(type)->getValueType(); continue; } llvm_unreachable("unknown type kind!"); } } void CodeGenFunction::EmitReturnBlock() { // For cleanliness, we try to avoid emitting the return block for // simple cases. llvm::BasicBlock *CurBB = Builder.GetInsertBlock(); if (CurBB) { assert(!CurBB->getTerminator() && "Unexpected terminated block."); // We have a valid insert point, reuse it if it is empty or there are no // explicit jumps to the return block. if (CurBB->empty() || ReturnBlock.getBlock()->use_empty()) { ReturnBlock.getBlock()->replaceAllUsesWith(CurBB); delete ReturnBlock.getBlock(); } else EmitBlock(ReturnBlock.getBlock()); return; } // Otherwise, if the return block is the target of a single direct // branch then we can just put the code in that block instead. This // cleans up functions which started with a unified return block. if (ReturnBlock.getBlock()->hasOneUse()) { llvm::BranchInst *BI = dyn_cast(*ReturnBlock.getBlock()->use_begin()); if (BI && BI->isUnconditional() && BI->getSuccessor(0) == ReturnBlock.getBlock()) { // Reset insertion point, including debug location, and delete the // branch. This is really subtle and only works because the next change // in location will hit the caching in CGDebugInfo::EmitLocation and not // override this. Builder.SetCurrentDebugLocation(BI->getDebugLoc()); Builder.SetInsertPoint(BI->getParent()); BI->eraseFromParent(); delete ReturnBlock.getBlock(); return; } } // FIXME: We are at an unreachable point, there is no reason to emit the block // unless it has uses. However, we still need a place to put the debug // region.end for now. EmitBlock(ReturnBlock.getBlock()); } static void EmitIfUsed(CodeGenFunction &CGF, llvm::BasicBlock *BB) { if (!BB) return; if (!BB->use_empty()) return CGF.CurFn->getBasicBlockList().push_back(BB); delete BB; } void CodeGenFunction::FinishFunction(SourceLocation EndLoc) { assert(BreakContinueStack.empty() && "mismatched push/pop in break/continue stack!"); bool OnlySimpleReturnStmts = NumSimpleReturnExprs > 0 && NumSimpleReturnExprs == NumReturnExprs && ReturnBlock.getBlock()->use_empty(); // Usually the return expression is evaluated before the cleanup // code. If the function contains only a simple return statement, // such as a constant, the location before the cleanup code becomes // the last useful breakpoint in the function, because the simple // return expression will be evaluated after the cleanup code. To be // safe, set the debug location for cleanup code to the location of // the return statement. Otherwise the cleanup code should be at the // end of the function's lexical scope. // // If there are multiple branches to the return block, the branch // instructions will get the location of the return statements and // all will be fine. if (CGDebugInfo *DI = getDebugInfo()) { if (OnlySimpleReturnStmts) DI->EmitLocation(Builder, LastStopPoint); else DI->EmitLocation(Builder, EndLoc); } // Pop any cleanups that might have been associated with the // parameters. Do this in whatever block we're currently in; it's // important to do this before we enter the return block or return // edges will be *really* confused. bool EmitRetDbgLoc = true; if (EHStack.stable_begin() != PrologueCleanupDepth) { PopCleanupBlocks(PrologueCleanupDepth); // Make sure the line table doesn't jump back into the body for // the ret after it's been at EndLoc. EmitRetDbgLoc = false; if (CGDebugInfo *DI = getDebugInfo()) if (OnlySimpleReturnStmts) DI->EmitLocation(Builder, EndLoc); } // Emit function epilog (to return). EmitReturnBlock(); if (ShouldInstrumentFunction()) EmitFunctionInstrumentation("__cyg_profile_func_exit"); // Emit debug descriptor for function end. if (CGDebugInfo *DI = getDebugInfo()) { DI->EmitFunctionEnd(Builder); } EmitFunctionEpilog(*CurFnInfo, EmitRetDbgLoc, EndLoc); EmitEndEHSpec(CurCodeDecl); assert(EHStack.empty() && "did not remove all scopes from cleanup stack!"); // If someone did an indirect goto, emit the indirect goto block at the end of // the function. if (IndirectBranch) { EmitBlock(IndirectBranch->getParent()); Builder.ClearInsertionPoint(); } // Remove the AllocaInsertPt instruction, which is just a convenience for us. llvm::Instruction *Ptr = AllocaInsertPt; AllocaInsertPt = 0; Ptr->eraseFromParent(); // If someone took the address of a label but never did an indirect goto, we // made a zero entry PHI node, which is illegal, zap it now. if (IndirectBranch) { llvm::PHINode *PN = cast(IndirectBranch->getAddress()); if (PN->getNumIncomingValues() == 0) { PN->replaceAllUsesWith(llvm::UndefValue::get(PN->getType())); PN->eraseFromParent(); } } EmitIfUsed(*this, EHResumeBlock); EmitIfUsed(*this, TerminateLandingPad); EmitIfUsed(*this, TerminateHandler); EmitIfUsed(*this, UnreachableBlock); if (CGM.getCodeGenOpts().EmitDeclMetadata) EmitDeclMetadata(); } /// ShouldInstrumentFunction - Return true if the current function should be /// instrumented with __cyg_profile_func_* calls bool CodeGenFunction::ShouldInstrumentFunction() { if (!CGM.getCodeGenOpts().InstrumentFunctions) return false; if (!CurFuncDecl || CurFuncDecl->hasAttr()) return false; return true; } /// EmitFunctionInstrumentation - Emit LLVM code to call the specified /// instrumentation function with the current function and the call site, if /// function instrumentation is enabled. void CodeGenFunction::EmitFunctionInstrumentation(const char *Fn) { // void __cyg_profile_func_{enter,exit} (void *this_fn, void *call_site); llvm::PointerType *PointerTy = Int8PtrTy; llvm::Type *ProfileFuncArgs[] = { PointerTy, PointerTy }; llvm::FunctionType *FunctionTy = llvm::FunctionType::get(VoidTy, ProfileFuncArgs, false); llvm::Constant *F = CGM.CreateRuntimeFunction(FunctionTy, Fn); llvm::CallInst *CallSite = Builder.CreateCall( CGM.getIntrinsic(llvm::Intrinsic::returnaddress), llvm::ConstantInt::get(Int32Ty, 0), "callsite"); llvm::Value *args[] = { llvm::ConstantExpr::getBitCast(CurFn, PointerTy), CallSite }; EmitNounwindRuntimeCall(F, args); } void CodeGenFunction::EmitMCountInstrumentation() { llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy, false); llvm::Constant *MCountFn = CGM.CreateRuntimeFunction(FTy, getTarget().getMCountName()); EmitNounwindRuntimeCall(MCountFn); } // OpenCL v1.2 s5.6.4.6 allows the compiler to store kernel argument // information in the program executable. The argument information stored // includes the argument name, its type, the address and access qualifiers used. static void GenOpenCLArgMetadata(const FunctionDecl *FD, llvm::Function *Fn, CodeGenModule &CGM,llvm::LLVMContext &Context, SmallVector &kernelMDArgs, CGBuilderTy& Builder, ASTContext &ASTCtx) { // Create MDNodes that represent the kernel arg metadata. // Each MDNode is a list in the form of "key", N number of values which is // the same number of values as their are kernel arguments. // MDNode for the kernel argument address space qualifiers. SmallVector addressQuals; addressQuals.push_back(llvm::MDString::get(Context, "kernel_arg_addr_space")); // MDNode for the kernel argument access qualifiers (images only). SmallVector accessQuals; accessQuals.push_back(llvm::MDString::get(Context, "kernel_arg_access_qual")); // MDNode for the kernel argument type names. SmallVector argTypeNames; argTypeNames.push_back(llvm::MDString::get(Context, "kernel_arg_type")); // MDNode for the kernel argument type qualifiers. SmallVector argTypeQuals; argTypeQuals.push_back(llvm::MDString::get(Context, "kernel_arg_type_qual")); // MDNode for the kernel argument names. SmallVector argNames; argNames.push_back(llvm::MDString::get(Context, "kernel_arg_name")); for (unsigned i = 0, e = FD->getNumParams(); i != e; ++i) { const ParmVarDecl *parm = FD->getParamDecl(i); QualType ty = parm->getType(); std::string typeQuals; if (ty->isPointerType()) { QualType pointeeTy = ty->getPointeeType(); // Get address qualifier. addressQuals.push_back(Builder.getInt32(ASTCtx.getTargetAddressSpace( pointeeTy.getAddressSpace()))); // Get argument type name. std::string typeName = pointeeTy.getUnqualifiedType().getAsString() + "*"; // Turn "unsigned type" to "utype" std::string::size_type pos = typeName.find("unsigned"); if (pos != std::string::npos) typeName.erase(pos+1, 8); argTypeNames.push_back(llvm::MDString::get(Context, typeName)); // Get argument type qualifiers: if (ty.isRestrictQualified()) typeQuals = "restrict"; if (pointeeTy.isConstQualified() || (pointeeTy.getAddressSpace() == LangAS::opencl_constant)) typeQuals += typeQuals.empty() ? "const" : " const"; if (pointeeTy.isVolatileQualified()) typeQuals += typeQuals.empty() ? "volatile" : " volatile"; } else { addressQuals.push_back(Builder.getInt32(0)); // Get argument type name. std::string typeName = ty.getUnqualifiedType().getAsString(); // Turn "unsigned type" to "utype" std::string::size_type pos = typeName.find("unsigned"); if (pos != std::string::npos) typeName.erase(pos+1, 8); argTypeNames.push_back(llvm::MDString::get(Context, typeName)); // Get argument type qualifiers: if (ty.isConstQualified()) typeQuals = "const"; if (ty.isVolatileQualified()) typeQuals += typeQuals.empty() ? "volatile" : " volatile"; } argTypeQuals.push_back(llvm::MDString::get(Context, typeQuals)); // Get image access qualifier: if (ty->isImageType()) { if (parm->hasAttr() && parm->getAttr()->getAccess() == CLIA_write_only) accessQuals.push_back(llvm::MDString::get(Context, "write_only")); else accessQuals.push_back(llvm::MDString::get(Context, "read_only")); } else accessQuals.push_back(llvm::MDString::get(Context, "none")); // Get argument name. argNames.push_back(llvm::MDString::get(Context, parm->getName())); } kernelMDArgs.push_back(llvm::MDNode::get(Context, addressQuals)); kernelMDArgs.push_back(llvm::MDNode::get(Context, accessQuals)); kernelMDArgs.push_back(llvm::MDNode::get(Context, argTypeNames)); kernelMDArgs.push_back(llvm::MDNode::get(Context, argTypeQuals)); kernelMDArgs.push_back(llvm::MDNode::get(Context, argNames)); } void CodeGenFunction::EmitOpenCLKernelMetadata(const FunctionDecl *FD, llvm::Function *Fn) { if (!FD->hasAttr()) return; llvm::LLVMContext &Context = getLLVMContext(); SmallVector kernelMDArgs; kernelMDArgs.push_back(Fn); if (CGM.getCodeGenOpts().EmitOpenCLArgMetadata) GenOpenCLArgMetadata(FD, Fn, CGM, Context, kernelMDArgs, Builder, getContext()); if (FD->hasAttr()) { VecTypeHintAttr *attr = FD->getAttr(); QualType hintQTy = attr->getTypeHint(); const ExtVectorType *hintEltQTy = hintQTy->getAs(); bool isSignedInteger = hintQTy->isSignedIntegerType() || (hintEltQTy && hintEltQTy->getElementType()->isSignedIntegerType()); llvm::Value *attrMDArgs[] = { llvm::MDString::get(Context, "vec_type_hint"), llvm::UndefValue::get(CGM.getTypes().ConvertType(attr->getTypeHint())), llvm::ConstantInt::get( llvm::IntegerType::get(Context, 32), llvm::APInt(32, (uint64_t)(isSignedInteger ? 1 : 0))) }; kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs)); } if (FD->hasAttr()) { WorkGroupSizeHintAttr *attr = FD->getAttr(); llvm::Value *attrMDArgs[] = { llvm::MDString::get(Context, "work_group_size_hint"), Builder.getInt32(attr->getXDim()), Builder.getInt32(attr->getYDim()), Builder.getInt32(attr->getZDim()) }; kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs)); } if (FD->hasAttr()) { ReqdWorkGroupSizeAttr *attr = FD->getAttr(); llvm::Value *attrMDArgs[] = { llvm::MDString::get(Context, "reqd_work_group_size"), Builder.getInt32(attr->getXDim()), Builder.getInt32(attr->getYDim()), Builder.getInt32(attr->getZDim()) }; kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs)); } llvm::MDNode *kernelMDNode = llvm::MDNode::get(Context, kernelMDArgs); llvm::NamedMDNode *OpenCLKernelMetadata = CGM.getModule().getOrInsertNamedMetadata("opencl.kernels"); OpenCLKernelMetadata->addOperand(kernelMDNode); } void CodeGenFunction::StartFunction(GlobalDecl GD, QualType RetTy, llvm::Function *Fn, const CGFunctionInfo &FnInfo, const FunctionArgList &Args, SourceLocation StartLoc) { const Decl *D = GD.getDecl(); DidCallStackSave = false; CurCodeDecl = D; CurFuncDecl = (D ? D->getNonClosureContext() : 0); FnRetTy = RetTy; CurFn = Fn; CurFnInfo = &FnInfo; assert(CurFn->isDeclaration() && "Function already has body?"); if (CGM.getSanitizerBlacklist().isIn(*Fn)) { SanOpts = &SanitizerOptions::Disabled; SanitizePerformTypeCheck = false; } // Pass inline keyword to optimizer if it appears explicitly on any // declaration. if (!CGM.getCodeGenOpts().NoInline) if (const FunctionDecl *FD = dyn_cast_or_null(D)) for (FunctionDecl::redecl_iterator RI = FD->redecls_begin(), RE = FD->redecls_end(); RI != RE; ++RI) if (RI->isInlineSpecified()) { Fn->addFnAttr(llvm::Attribute::InlineHint); break; } if (getLangOpts().OpenCL) { // Add metadata for a kernel function. if (const FunctionDecl *FD = dyn_cast_or_null(D)) EmitOpenCLKernelMetadata(FD, Fn); } // If we are checking function types, emit a function type signature as // prefix data. if (getLangOpts().CPlusPlus && SanOpts->Function) { if (const FunctionDecl *FD = dyn_cast_or_null(D)) { if (llvm::Constant *PrefixSig = CGM.getTargetCodeGenInfo().getUBSanFunctionSignature(CGM)) { llvm::Constant *FTRTTIConst = CGM.GetAddrOfRTTIDescriptor(FD->getType(), /*ForEH=*/true); llvm::Constant *PrefixStructElems[] = { PrefixSig, FTRTTIConst }; llvm::Constant *PrefixStructConst = llvm::ConstantStruct::getAnon(PrefixStructElems, /*Packed=*/true); Fn->setPrefixData(PrefixStructConst); } } } llvm::BasicBlock *EntryBB = createBasicBlock("entry", CurFn); // Create a marker to make it easy to insert allocas into the entryblock // later. Don't create this with the builder, because we don't want it // folded. llvm::Value *Undef = llvm::UndefValue::get(Int32Ty); AllocaInsertPt = new llvm::BitCastInst(Undef, Int32Ty, "", EntryBB); if (Builder.isNamePreserving()) AllocaInsertPt->setName("allocapt"); ReturnBlock = getJumpDestInCurrentScope("return"); Builder.SetInsertPoint(EntryBB); // Emit subprogram debug descriptor. if (CGDebugInfo *DI = getDebugInfo()) { SmallVector ArgTypes; for (FunctionArgList::const_iterator i = Args.begin(), e = Args.end(); i != e; ++i) { ArgTypes.push_back((*i)->getType()); } QualType FnType = getContext().getFunctionType(RetTy, ArgTypes, FunctionProtoType::ExtProtoInfo()); DI->setLocation(StartLoc); DI->EmitFunctionStart(GD, FnType, CurFn, Builder); } if (ShouldInstrumentFunction()) EmitFunctionInstrumentation("__cyg_profile_func_enter"); if (CGM.getCodeGenOpts().InstrumentForProfiling) EmitMCountInstrumentation(); if (RetTy->isVoidType()) { // Void type; nothing to return. ReturnValue = 0; } else if (CurFnInfo->getReturnInfo().getKind() == ABIArgInfo::Indirect && !hasScalarEvaluationKind(CurFnInfo->getReturnType())) { // Indirect aggregate return; emit returned value directly into sret slot. // This reduces code size, and affects correctness in C++. ReturnValue = CurFn->arg_begin(); } else { ReturnValue = CreateIRTemp(RetTy, "retval"); // Tell the epilog emitter to autorelease the result. We do this // now so that various specialized functions can suppress it // during their IR-generation. if (getLangOpts().ObjCAutoRefCount && !CurFnInfo->isReturnsRetained() && RetTy->isObjCRetainableType()) AutoreleaseResult = true; } EmitStartEHSpec(CurCodeDecl); PrologueCleanupDepth = EHStack.stable_begin(); EmitFunctionProlog(*CurFnInfo, CurFn, Args); if (D && isa(D) && cast(D)->isInstance()) { CGM.getCXXABI().EmitInstanceFunctionProlog(*this); const CXXMethodDecl *MD = cast(D); if (MD->getParent()->isLambda() && MD->getOverloadedOperator() == OO_Call) { // We're in a lambda; figure out the captures. MD->getParent()->getCaptureFields(LambdaCaptureFields, LambdaThisCaptureField); if (LambdaThisCaptureField) { // If this lambda captures this, load it. LValue ThisLValue = EmitLValueForLambdaField(LambdaThisCaptureField); CXXThisValue = EmitLoadOfLValue(ThisLValue, SourceLocation()).getScalarVal(); } } else { // Not in a lambda; just use 'this' from the method. // FIXME: Should we generate a new load for each use of 'this'? The // fast register allocator would be happier... CXXThisValue = CXXABIThisValue; } } // If any of the arguments have a variably modified type, make sure to // emit the type size. for (FunctionArgList::const_iterator i = Args.begin(), e = Args.end(); i != e; ++i) { const VarDecl *VD = *i; // Dig out the type as written from ParmVarDecls; it's unclear whether // the standard (C99 6.9.1p10) requires this, but we're following the // precedent set by gcc. QualType Ty; if (const ParmVarDecl *PVD = dyn_cast(VD)) Ty = PVD->getOriginalType(); else Ty = VD->getType(); if (Ty->isVariablyModifiedType()) EmitVariablyModifiedType(Ty); } // Emit a location at the end of the prologue. if (CGDebugInfo *DI = getDebugInfo()) DI->EmitLocation(Builder, StartLoc); } void CodeGenFunction::EmitFunctionBody(FunctionArgList &Args, const Stmt *Body) { if (const CompoundStmt *S = dyn_cast(Body)) EmitCompoundStmtWithoutScope(*S); else EmitStmt(Body); } /// Tries to mark the given function nounwind based on the /// non-existence of any throwing calls within it. We believe this is /// lightweight enough to do at -O0. static void TryMarkNoThrow(llvm::Function *F) { // LLVM treats 'nounwind' on a function as part of the type, so we // can't do this on functions that can be overwritten. if (F->mayBeOverridden()) return; for (llvm::Function::iterator FI = F->begin(), FE = F->end(); FI != FE; ++FI) for (llvm::BasicBlock::iterator BI = FI->begin(), BE = FI->end(); BI != BE; ++BI) if (llvm::CallInst *Call = dyn_cast(&*BI)) { if (!Call->doesNotThrow()) return; } else if (isa(&*BI)) { return; } F->setDoesNotThrow(); } static void EmitSizedDeallocationFunction(CodeGenFunction &CGF, const FunctionDecl *UnsizedDealloc) { // This is a weak discardable definition of the sized deallocation function. CGF.CurFn->setLinkage(llvm::Function::LinkOnceAnyLinkage); // Call the unsized deallocation function and forward the first argument // unchanged. llvm::Constant *Unsized = CGF.CGM.GetAddrOfFunction(UnsizedDealloc); CGF.Builder.CreateCall(Unsized, &*CGF.CurFn->arg_begin()); } void CodeGenFunction::GenerateCode(GlobalDecl GD, llvm::Function *Fn, const CGFunctionInfo &FnInfo) { const FunctionDecl *FD = cast(GD.getDecl()); // Check if we should generate debug info for this function. if (FD->hasAttr()) DebugInfo = NULL; // disable debug info indefinitely for this function FunctionArgList Args; QualType ResTy = FD->getResultType(); CurGD = GD; const CXXMethodDecl *MD; if ((MD = dyn_cast(FD)) && MD->isInstance()) { if (CGM.getCXXABI().HasThisReturn(GD)) ResTy = MD->getThisType(getContext()); CGM.getCXXABI().BuildInstanceFunctionParams(*this, ResTy, Args); } for (unsigned i = 0, e = FD->getNumParams(); i != e; ++i) Args.push_back(FD->getParamDecl(i)); SourceRange BodyRange; if (Stmt *Body = FD->getBody()) BodyRange = Body->getSourceRange(); CurEHLocation = BodyRange.getEnd(); // Emit the standard function prologue. StartFunction(GD, ResTy, Fn, FnInfo, Args, BodyRange.getBegin()); // Generate the body of the function. if (isa(FD)) EmitDestructorBody(Args); else if (isa(FD)) EmitConstructorBody(Args); else if (getLangOpts().CUDA && !CGM.getCodeGenOpts().CUDAIsDevice && FD->hasAttr()) CGM.getCUDARuntime().EmitDeviceStubBody(*this, Args); else if (isa(FD) && cast(FD)->isLambdaToBlockPointerConversion()) { // The lambda conversion to block pointer is special; the semantics can't be // expressed in the AST, so IRGen needs to special-case it. EmitLambdaToBlockPointerBody(Args); } else if (isa(FD) && cast(FD)->isLambdaStaticInvoker()) { // The lambda static invoker function is special, because it forwards or // clones the body of the function call operator (but is actually static). EmitLambdaStaticInvokeFunction(cast(FD)); } else if (FD->isDefaulted() && isa(FD) && (cast(FD)->isCopyAssignmentOperator() || cast(FD)->isMoveAssignmentOperator())) { // Implicit copy-assignment gets the same special treatment as implicit // copy-constructors. emitImplicitAssignmentOperatorBody(Args); } else if (Stmt *Body = FD->getBody()) { EmitFunctionBody(Args, Body); } else if (FunctionDecl *UnsizedDealloc = FD->getCorrespondingUnsizedGlobalDeallocationFunction()) { // Global sized deallocation functions get an implicit weak definition if // they don't have an explicit definition. EmitSizedDeallocationFunction(*this, UnsizedDealloc); } else llvm_unreachable("no definition for emitted function"); // C++11 [stmt.return]p2: // Flowing off the end of a function [...] results in undefined behavior in // a value-returning function. // C11 6.9.1p12: // If the '}' that terminates a function is reached, and the value of the // function call is used by the caller, the behavior is undefined. if (getLangOpts().CPlusPlus && !FD->hasImplicitReturnZero() && !FD->getResultType()->isVoidType() && Builder.GetInsertBlock()) { if (SanOpts->Return) EmitCheck(Builder.getFalse(), "missing_return", EmitCheckSourceLocation(FD->getLocation()), ArrayRef(), CRK_Unrecoverable); else if (CGM.getCodeGenOpts().OptimizationLevel == 0) Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::trap)); Builder.CreateUnreachable(); Builder.ClearInsertionPoint(); } // Emit the standard function epilogue. FinishFunction(BodyRange.getEnd()); // If we haven't marked the function nothrow through other means, do // a quick pass now to see if we can. if (!CurFn->doesNotThrow()) TryMarkNoThrow(CurFn); } /// ContainsLabel - Return true if the statement contains a label in it. If /// this statement is not executed normally, it not containing a label means /// that we can just remove the code. bool CodeGenFunction::ContainsLabel(const Stmt *S, bool IgnoreCaseStmts) { // Null statement, not a label! if (S == 0) return false; // If this is a label, we have to emit the code, consider something like: // if (0) { ... foo: bar(); } goto foo; // // TODO: If anyone cared, we could track __label__'s, since we know that you // can't jump to one from outside their declared region. if (isa(S)) return true; // If this is a case/default statement, and we haven't seen a switch, we have // to emit the code. if (isa(S) && !IgnoreCaseStmts) return true; // If this is a switch statement, we want to ignore cases below it. if (isa(S)) IgnoreCaseStmts = true; // Scan subexpressions for verboten labels. for (Stmt::const_child_range I = S->children(); I; ++I) if (ContainsLabel(*I, IgnoreCaseStmts)) return true; return false; } /// containsBreak - Return true if the statement contains a break out of it. /// If the statement (recursively) contains a switch or loop with a break /// inside of it, this is fine. bool CodeGenFunction::containsBreak(const Stmt *S) { // Null statement, not a label! if (S == 0) return false; // If this is a switch or loop that defines its own break scope, then we can // include it and anything inside of it. if (isa(S) || isa(S) || isa(S) || isa(S)) return false; if (isa(S)) return true; // Scan subexpressions for verboten breaks. for (Stmt::const_child_range I = S->children(); I; ++I) if (containsBreak(*I)) return true; return false; } /// ConstantFoldsToSimpleInteger - If the specified expression does not fold /// to a constant, or if it does but contains a label, return false. If it /// constant folds return true and set the boolean result in Result. bool CodeGenFunction::ConstantFoldsToSimpleInteger(const Expr *Cond, bool &ResultBool) { llvm::APSInt ResultInt; if (!ConstantFoldsToSimpleInteger(Cond, ResultInt)) return false; ResultBool = ResultInt.getBoolValue(); return true; } /// ConstantFoldsToSimpleInteger - If the specified expression does not fold /// to a constant, or if it does but contains a label, return false. If it /// constant folds return true and set the folded value. bool CodeGenFunction:: ConstantFoldsToSimpleInteger(const Expr *Cond, llvm::APSInt &ResultInt) { // FIXME: Rename and handle conversion of other evaluatable things // to bool. llvm::APSInt Int; if (!Cond->EvaluateAsInt(Int, getContext())) return false; // Not foldable, not integer or not fully evaluatable. if (CodeGenFunction::ContainsLabel(Cond)) return false; // Contains a label. ResultInt = Int; return true; } /// EmitBranchOnBoolExpr - Emit a branch on a boolean condition (e.g. for an if /// statement) to the specified blocks. Based on the condition, this might try /// to simplify the codegen of the conditional based on the branch. /// void CodeGenFunction::EmitBranchOnBoolExpr(const Expr *Cond, llvm::BasicBlock *TrueBlock, llvm::BasicBlock *FalseBlock) { Cond = Cond->IgnoreParens(); if (const BinaryOperator *CondBOp = dyn_cast(Cond)) { // Handle X && Y in a condition. if (CondBOp->getOpcode() == BO_LAnd) { // If we have "1 && X", simplify the code. "0 && X" would have constant // folded if the case was simple enough. bool ConstantBool = false; if (ConstantFoldsToSimpleInteger(CondBOp->getLHS(), ConstantBool) && ConstantBool) { // br(1 && X) -> br(X). return EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock); } // If we have "X && 1", simplify the code to use an uncond branch. // "X && 0" would have been constant folded to 0. if (ConstantFoldsToSimpleInteger(CondBOp->getRHS(), ConstantBool) && ConstantBool) { // br(X && 1) -> br(X). return EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, FalseBlock); } // Emit the LHS as a conditional. If the LHS conditional is false, we // want to jump to the FalseBlock. llvm::BasicBlock *LHSTrue = createBasicBlock("land.lhs.true"); ConditionalEvaluation eval(*this); EmitBranchOnBoolExpr(CondBOp->getLHS(), LHSTrue, FalseBlock); EmitBlock(LHSTrue); // Any temporaries created here are conditional. eval.begin(*this); EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock); eval.end(*this); return; } if (CondBOp->getOpcode() == BO_LOr) { // If we have "0 || X", simplify the code. "1 || X" would have constant // folded if the case was simple enough. bool ConstantBool = false; if (ConstantFoldsToSimpleInteger(CondBOp->getLHS(), ConstantBool) && !ConstantBool) { // br(0 || X) -> br(X). return EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock); } // If we have "X || 0", simplify the code to use an uncond branch. // "X || 1" would have been constant folded to 1. if (ConstantFoldsToSimpleInteger(CondBOp->getRHS(), ConstantBool) && !ConstantBool) { // br(X || 0) -> br(X). return EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, FalseBlock); } // Emit the LHS as a conditional. If the LHS conditional is true, we // want to jump to the TrueBlock. llvm::BasicBlock *LHSFalse = createBasicBlock("lor.lhs.false"); ConditionalEvaluation eval(*this); EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, LHSFalse); EmitBlock(LHSFalse); // Any temporaries created here are conditional. eval.begin(*this); EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock); eval.end(*this); return; } } if (const UnaryOperator *CondUOp = dyn_cast(Cond)) { // br(!x, t, f) -> br(x, f, t) if (CondUOp->getOpcode() == UO_LNot) return EmitBranchOnBoolExpr(CondUOp->getSubExpr(), FalseBlock, TrueBlock); } if (const ConditionalOperator *CondOp = dyn_cast(Cond)) { // br(c ? x : y, t, f) -> br(c, br(x, t, f), br(y, t, f)) llvm::BasicBlock *LHSBlock = createBasicBlock("cond.true"); llvm::BasicBlock *RHSBlock = createBasicBlock("cond.false"); ConditionalEvaluation cond(*this); EmitBranchOnBoolExpr(CondOp->getCond(), LHSBlock, RHSBlock); cond.begin(*this); EmitBlock(LHSBlock); EmitBranchOnBoolExpr(CondOp->getLHS(), TrueBlock, FalseBlock); cond.end(*this); cond.begin(*this); EmitBlock(RHSBlock); EmitBranchOnBoolExpr(CondOp->getRHS(), TrueBlock, FalseBlock); cond.end(*this); return; } if (const CXXThrowExpr *Throw = dyn_cast(Cond)) { // Conditional operator handling can give us a throw expression as a // condition for a case like: // br(c ? throw x : y, t, f) -> br(c, br(throw x, t, f), br(y, t, f) // Fold this to: // br(c, throw x, br(y, t, f)) EmitCXXThrowExpr(Throw, /*KeepInsertionPoint*/false); return; } // Emit the code with the fully general case. llvm::Value *CondV = EvaluateExprAsBool(Cond); Builder.CreateCondBr(CondV, TrueBlock, FalseBlock); } /// ErrorUnsupported - Print out an error that codegen doesn't support the /// specified stmt yet. void CodeGenFunction::ErrorUnsupported(const Stmt *S, const char *Type) { CGM.ErrorUnsupported(S, Type); } /// emitNonZeroVLAInit - Emit the "zero" initialization of a /// variable-length array whose elements have a non-zero bit-pattern. /// /// \param baseType the inner-most element type of the array /// \param src - a char* pointing to the bit-pattern for a single /// base element of the array /// \param sizeInChars - the total size of the VLA, in chars static void emitNonZeroVLAInit(CodeGenFunction &CGF, QualType baseType, llvm::Value *dest, llvm::Value *src, llvm::Value *sizeInChars) { std::pair baseSizeAndAlign = CGF.getContext().getTypeInfoInChars(baseType); CGBuilderTy &Builder = CGF.Builder; llvm::Value *baseSizeInChars = llvm::ConstantInt::get(CGF.IntPtrTy, baseSizeAndAlign.first.getQuantity()); llvm::Type *i8p = Builder.getInt8PtrTy(); llvm::Value *begin = Builder.CreateBitCast(dest, i8p, "vla.begin"); llvm::Value *end = Builder.CreateInBoundsGEP(dest, sizeInChars, "vla.end"); llvm::BasicBlock *originBB = CGF.Builder.GetInsertBlock(); llvm::BasicBlock *loopBB = CGF.createBasicBlock("vla-init.loop"); llvm::BasicBlock *contBB = CGF.createBasicBlock("vla-init.cont"); // Make a loop over the VLA. C99 guarantees that the VLA element // count must be nonzero. CGF.EmitBlock(loopBB); llvm::PHINode *cur = Builder.CreatePHI(i8p, 2, "vla.cur"); cur->addIncoming(begin, originBB); // memcpy the individual element bit-pattern. Builder.CreateMemCpy(cur, src, baseSizeInChars, baseSizeAndAlign.second.getQuantity(), /*volatile*/ false); // Go to the next element. llvm::Value *next = Builder.CreateConstInBoundsGEP1_32(cur, 1, "vla.next"); // Leave if that's the end of the VLA. llvm::Value *done = Builder.CreateICmpEQ(next, end, "vla-init.isdone"); Builder.CreateCondBr(done, contBB, loopBB); cur->addIncoming(next, loopBB); CGF.EmitBlock(contBB); } void CodeGenFunction::EmitNullInitialization(llvm::Value *DestPtr, QualType Ty) { // Ignore empty classes in C++. if (getLangOpts().CPlusPlus) { if (const RecordType *RT = Ty->getAs()) { if (cast(RT->getDecl())->isEmpty()) return; } } // Cast the dest ptr to the appropriate i8 pointer type. unsigned DestAS = cast(DestPtr->getType())->getAddressSpace(); llvm::Type *BP = Builder.getInt8PtrTy(DestAS); if (DestPtr->getType() != BP) DestPtr = Builder.CreateBitCast(DestPtr, BP); // Get size and alignment info for this aggregate. std::pair TypeInfo = getContext().getTypeInfoInChars(Ty); CharUnits Size = TypeInfo.first; CharUnits Align = TypeInfo.second; llvm::Value *SizeVal; const VariableArrayType *vla; // Don't bother emitting a zero-byte memset. if (Size.isZero()) { // But note that getTypeInfo returns 0 for a VLA. if (const VariableArrayType *vlaType = dyn_cast_or_null( getContext().getAsArrayType(Ty))) { QualType eltType; llvm::Value *numElts; llvm::tie(numElts, eltType) = getVLASize(vlaType); SizeVal = numElts; CharUnits eltSize = getContext().getTypeSizeInChars(eltType); if (!eltSize.isOne()) SizeVal = Builder.CreateNUWMul(SizeVal, CGM.getSize(eltSize)); vla = vlaType; } else { return; } } else { SizeVal = CGM.getSize(Size); vla = 0; } // If the type contains a pointer to data member we can't memset it to zero. // Instead, create a null constant and copy it to the destination. // TODO: there are other patterns besides zero that we can usefully memset, // like -1, which happens to be the pattern used by member-pointers. if (!CGM.getTypes().isZeroInitializable(Ty)) { // For a VLA, emit a single element, then splat that over the VLA. if (vla) Ty = getContext().getBaseElementType(vla); llvm::Constant *NullConstant = CGM.EmitNullConstant(Ty); llvm::GlobalVariable *NullVariable = new llvm::GlobalVariable(CGM.getModule(), NullConstant->getType(), /*isConstant=*/true, llvm::GlobalVariable::PrivateLinkage, NullConstant, Twine()); llvm::Value *SrcPtr = Builder.CreateBitCast(NullVariable, Builder.getInt8PtrTy()); if (vla) return emitNonZeroVLAInit(*this, Ty, DestPtr, SrcPtr, SizeVal); // Get and call the appropriate llvm.memcpy overload. Builder.CreateMemCpy(DestPtr, SrcPtr, SizeVal, Align.getQuantity(), false); return; } // Otherwise, just memset the whole thing to zero. This is legal // because in LLVM, all default initializers (other than the ones we just // handled above) are guaranteed to have a bit pattern of all zeros. Builder.CreateMemSet(DestPtr, Builder.getInt8(0), SizeVal, Align.getQuantity(), false); } llvm::BlockAddress *CodeGenFunction::GetAddrOfLabel(const LabelDecl *L) { // Make sure that there is a block for the indirect goto. if (IndirectBranch == 0) GetIndirectGotoBlock(); llvm::BasicBlock *BB = getJumpDestForLabel(L).getBlock(); // Make sure the indirect branch includes all of the address-taken blocks. IndirectBranch->addDestination(BB); return llvm::BlockAddress::get(CurFn, BB); } llvm::BasicBlock *CodeGenFunction::GetIndirectGotoBlock() { // If we already made the indirect branch for indirect goto, return its block. if (IndirectBranch) return IndirectBranch->getParent(); CGBuilderTy TmpBuilder(createBasicBlock("indirectgoto")); // Create the PHI node that indirect gotos will add entries to. llvm::Value *DestVal = TmpBuilder.CreatePHI(Int8PtrTy, 0, "indirect.goto.dest"); // Create the indirect branch instruction. IndirectBranch = TmpBuilder.CreateIndirectBr(DestVal); return IndirectBranch->getParent(); } /// Computes the length of an array in elements, as well as the base /// element type and a properly-typed first element pointer. llvm::Value *CodeGenFunction::emitArrayLength(const ArrayType *origArrayType, QualType &baseType, llvm::Value *&addr) { const ArrayType *arrayType = origArrayType; // If it's a VLA, we have to load the stored size. Note that // this is the size of the VLA in bytes, not its size in elements. llvm::Value *numVLAElements = 0; if (isa(arrayType)) { numVLAElements = getVLASize(cast(arrayType)).first; // Walk into all VLAs. This doesn't require changes to addr, // which has type T* where T is the first non-VLA element type. do { QualType elementType = arrayType->getElementType(); arrayType = getContext().getAsArrayType(elementType); // If we only have VLA components, 'addr' requires no adjustment. if (!arrayType) { baseType = elementType; return numVLAElements; } } while (isa(arrayType)); // We get out here only if we find a constant array type // inside the VLA. } // We have some number of constant-length arrays, so addr should // have LLVM type [M x [N x [...]]]*. Build a GEP that walks // down to the first element of addr. SmallVector gepIndices; // GEP down to the array type. llvm::ConstantInt *zero = Builder.getInt32(0); gepIndices.push_back(zero); uint64_t countFromCLAs = 1; QualType eltType; llvm::ArrayType *llvmArrayType = dyn_cast( cast(addr->getType())->getElementType()); while (llvmArrayType) { assert(isa(arrayType)); assert(cast(arrayType)->getSize().getZExtValue() == llvmArrayType->getNumElements()); gepIndices.push_back(zero); countFromCLAs *= llvmArrayType->getNumElements(); eltType = arrayType->getElementType(); llvmArrayType = dyn_cast(llvmArrayType->getElementType()); arrayType = getContext().getAsArrayType(arrayType->getElementType()); assert((!llvmArrayType || arrayType) && "LLVM and Clang types are out-of-synch"); } if (arrayType) { // From this point onwards, the Clang array type has been emitted // as some other type (probably a packed struct). Compute the array // size, and just emit the 'begin' expression as a bitcast. while (arrayType) { countFromCLAs *= cast(arrayType)->getSize().getZExtValue(); eltType = arrayType->getElementType(); arrayType = getContext().getAsArrayType(eltType); } unsigned AddressSpace = addr->getType()->getPointerAddressSpace(); llvm::Type *BaseType = ConvertType(eltType)->getPointerTo(AddressSpace); addr = Builder.CreateBitCast(addr, BaseType, "array.begin"); } else { // Create the actual GEP. addr = Builder.CreateInBoundsGEP(addr, gepIndices, "array.begin"); } baseType = eltType; llvm::Value *numElements = llvm::ConstantInt::get(SizeTy, countFromCLAs); // If we had any VLA dimensions, factor them in. if (numVLAElements) numElements = Builder.CreateNUWMul(numVLAElements, numElements); return numElements; } std::pair CodeGenFunction::getVLASize(QualType type) { const VariableArrayType *vla = getContext().getAsVariableArrayType(type); assert(vla && "type was not a variable array type!"); return getVLASize(vla); } std::pair CodeGenFunction::getVLASize(const VariableArrayType *type) { // The number of elements so far; always size_t. llvm::Value *numElements = 0; QualType elementType; do { elementType = type->getElementType(); llvm::Value *vlaSize = VLASizeMap[type->getSizeExpr()]; assert(vlaSize && "no size for VLA!"); assert(vlaSize->getType() == SizeTy); if (!numElements) { numElements = vlaSize; } else { // It's undefined behavior if this wraps around, so mark it that way. // FIXME: Teach -fcatch-undefined-behavior to trap this. numElements = Builder.CreateNUWMul(numElements, vlaSize); } } while ((type = getContext().getAsVariableArrayType(elementType))); return std::pair(numElements, elementType); } void CodeGenFunction::EmitVariablyModifiedType(QualType type) { assert(type->isVariablyModifiedType() && "Must pass variably modified type to EmitVLASizes!"); EnsureInsertPoint(); // We're going to walk down into the type and look for VLA // expressions. do { assert(type->isVariablyModifiedType()); const Type *ty = type.getTypePtr(); switch (ty->getTypeClass()) { #define TYPE(Class, Base) #define ABSTRACT_TYPE(Class, Base) #define NON_CANONICAL_TYPE(Class, Base) #define DEPENDENT_TYPE(Class, Base) case Type::Class: #define NON_CANONICAL_UNLESS_DEPENDENT_TYPE(Class, Base) #include "clang/AST/TypeNodes.def" llvm_unreachable("unexpected dependent type!"); // These types are never variably-modified. case Type::Builtin: case Type::Complex: case Type::Vector: case Type::ExtVector: case Type::Record: case Type::Enum: case Type::Elaborated: case Type::TemplateSpecialization: case Type::ObjCObject: case Type::ObjCInterface: case Type::ObjCObjectPointer: llvm_unreachable("type class is never variably-modified!"); case Type::Decayed: type = cast(ty)->getPointeeType(); break; case Type::Pointer: type = cast(ty)->getPointeeType(); break; case Type::BlockPointer: type = cast(ty)->getPointeeType(); break; case Type::LValueReference: case Type::RValueReference: type = cast(ty)->getPointeeType(); break; case Type::MemberPointer: type = cast(ty)->getPointeeType(); break; case Type::ConstantArray: case Type::IncompleteArray: // Losing element qualification here is fine. type = cast(ty)->getElementType(); break; case Type::VariableArray: { // Losing element qualification here is fine. const VariableArrayType *vat = cast(ty); // Unknown size indication requires no size computation. // Otherwise, evaluate and record it. if (const Expr *size = vat->getSizeExpr()) { // It's possible that we might have emitted this already, // e.g. with a typedef and a pointer to it. llvm::Value *&entry = VLASizeMap[size]; if (!entry) { llvm::Value *Size = EmitScalarExpr(size); // C11 6.7.6.2p5: // If the size is an expression that is not an integer constant // expression [...] each time it is evaluated it shall have a value // greater than zero. if (SanOpts->VLABound && size->getType()->isSignedIntegerType()) { llvm::Value *Zero = llvm::Constant::getNullValue(Size->getType()); llvm::Constant *StaticArgs[] = { EmitCheckSourceLocation(size->getLocStart()), EmitCheckTypeDescriptor(size->getType()) }; EmitCheck(Builder.CreateICmpSGT(Size, Zero), "vla_bound_not_positive", StaticArgs, Size, CRK_Recoverable); } // Always zexting here would be wrong if it weren't // undefined behavior to have a negative bound. entry = Builder.CreateIntCast(Size, SizeTy, /*signed*/ false); } } type = vat->getElementType(); break; } case Type::FunctionProto: case Type::FunctionNoProto: type = cast(ty)->getResultType(); break; case Type::Paren: case Type::TypeOf: case Type::UnaryTransform: case Type::Attributed: case Type::SubstTemplateTypeParm: case Type::PackExpansion: // Keep walking after single level desugaring. type = type.getSingleStepDesugaredType(getContext()); break; case Type::Typedef: case Type::Decltype: case Type::Auto: // Stop walking: nothing to do. return; case Type::TypeOfExpr: // Stop walking: emit typeof expression. EmitIgnoredExpr(cast(ty)->getUnderlyingExpr()); return; case Type::Atomic: type = cast(ty)->getValueType(); break; } } while (type->isVariablyModifiedType()); } llvm::Value* CodeGenFunction::EmitVAListRef(const Expr* E) { if (getContext().getBuiltinVaListType()->isArrayType()) return EmitScalarExpr(E); return EmitLValue(E).getAddress(); } void CodeGenFunction::EmitDeclRefExprDbgValue(const DeclRefExpr *E, llvm::Constant *Init) { assert (Init && "Invalid DeclRefExpr initializer!"); if (CGDebugInfo *Dbg = getDebugInfo()) if (CGM.getCodeGenOpts().getDebugInfo() >= CodeGenOptions::LimitedDebugInfo) Dbg->EmitGlobalVariable(E->getDecl(), Init); } CodeGenFunction::PeepholeProtection CodeGenFunction::protectFromPeepholes(RValue rvalue) { // At the moment, the only aggressive peephole we do in IR gen // is trunc(zext) folding, but if we add more, we can easily // extend this protection. if (!rvalue.isScalar()) return PeepholeProtection(); llvm::Value *value = rvalue.getScalarVal(); if (!isa(value)) return PeepholeProtection(); // Just make an extra bitcast. assert(HaveInsertPoint()); llvm::Instruction *inst = new llvm::BitCastInst(value, value->getType(), "", Builder.GetInsertBlock()); PeepholeProtection protection; protection.Inst = inst; return protection; } void CodeGenFunction::unprotectFromPeepholes(PeepholeProtection protection) { if (!protection.Inst) return; // In theory, we could try to duplicate the peepholes now, but whatever. protection.Inst->eraseFromParent(); } llvm::Value *CodeGenFunction::EmitAnnotationCall(llvm::Value *AnnotationFn, llvm::Value *AnnotatedVal, StringRef AnnotationStr, SourceLocation Location) { llvm::Value *Args[4] = { AnnotatedVal, Builder.CreateBitCast(CGM.EmitAnnotationString(AnnotationStr), Int8PtrTy), Builder.CreateBitCast(CGM.EmitAnnotationUnit(Location), Int8PtrTy), CGM.EmitAnnotationLineNo(Location) }; return Builder.CreateCall(AnnotationFn, Args); } void CodeGenFunction::EmitVarAnnotations(const VarDecl *D, llvm::Value *V) { assert(D->hasAttr() && "no annotate attribute"); // FIXME We create a new bitcast for every annotation because that's what // llvm-gcc was doing. for (specific_attr_iterator ai = D->specific_attr_begin(), ae = D->specific_attr_end(); ai != ae; ++ai) EmitAnnotationCall(CGM.getIntrinsic(llvm::Intrinsic::var_annotation), Builder.CreateBitCast(V, CGM.Int8PtrTy, V->getName()), (*ai)->getAnnotation(), D->getLocation()); } llvm::Value *CodeGenFunction::EmitFieldAnnotations(const FieldDecl *D, llvm::Value *V) { assert(D->hasAttr() && "no annotate attribute"); llvm::Type *VTy = V->getType(); llvm::Value *F = CGM.getIntrinsic(llvm::Intrinsic::ptr_annotation, CGM.Int8PtrTy); for (specific_attr_iterator ai = D->specific_attr_begin(), ae = D->specific_attr_end(); ai != ae; ++ai) { // FIXME Always emit the cast inst so we can differentiate between // annotation on the first field of a struct and annotation on the struct // itself. if (VTy != CGM.Int8PtrTy) V = Builder.Insert(new llvm::BitCastInst(V, CGM.Int8PtrTy)); V = EmitAnnotationCall(F, V, (*ai)->getAnnotation(), D->getLocation()); V = Builder.CreateBitCast(V, VTy); } return V; } CodeGenFunction::CGCapturedStmtInfo::~CGCapturedStmtInfo() { } @ 1.1.1.1 log @Import Clang 3.4rc1 r195771. @ text @@ 1.1.1.2 log @Import clang 3.5svn r198450. @ text @d402 1 a402 1 d407 2 a408 2 const OpenCLImageAccessAttr *A = parm->getAttr(); if (A && A->getAccess() == CLIA_write_only) d441 3 a443 2 if (const VecTypeHintAttr *A = FD->getAttr()) { QualType hintQTy = A->getTypeHint(); d450 1 a450 1 llvm::UndefValue::get(CGM.getTypes().ConvertType(A->getTypeHint())), d458 2 a459 1 if (const WorkGroupSizeHintAttr *A = FD->getAttr()) { d462 3 a464 3 Builder.getInt32(A->getXDim()), Builder.getInt32(A->getYDim()), Builder.getInt32(A->getZDim()) d469 2 a470 1 if (const ReqdWorkGroupSizeAttr *A = FD->getAttr()) { d473 3 a475 3 Builder.getInt32(A->getXDim()), Builder.getInt32(A->getYDim()), Builder.getInt32(A->getZDim()) d697 2 a698 2 const CXXMethodDecl *MD = dyn_cast(FD); if (MD && MD->isInstance()) { d701 1 a701 1 CGM.getCXXABI().buildThisParam(*this, Args); a706 3 if (MD && (isa(MD) || isa(MD))) CGM.getCXXABI().addImplicitStructorParams(*this, ResTy, Args); a1310 4 case Type::Adjusted: type = cast(ty)->getAdjustedType(); break; @ 1.1.1.3 log @Import Clang 3.5svn r199312 @ text @a18 1 #include "CodeGenPGO.h" d24 1 d47 1 a47 2 PGO(cgm), SwitchInsn(0), SwitchWeights(0), CaseRangeBlock(0), UnreachableBlock(0), NumReturnExprs(0), d384 1 a384 6 uint32_t AddrSpc = 0; if (ty->isImageType()) AddrSpc = CGM.getContext().getTargetAddressSpace(LangAS::opencl_global); addressQuals.push_back(Builder.getInt32(AddrSpc)); d408 1 a408 1 if (A && A->isWriteOnly()) a411 1 // FIXME: what about read_write? a572 2 PGO.assignRegionCounters(GD); a644 2 RegionCounter Cnt = getPGORegionCounter(Body); Cnt.beginRegion(Builder); a773 3 PGO.emitWriteoutFunction(CurGD); PGO.destroyRegionCounters(); d872 1 a872 2 llvm::BasicBlock *FalseBlock, uint64_t TrueCount) { a875 2 RegionCounter Cnt = getPGORegionCounter(CondBOp); d884 1 a884 3 Cnt.beginRegion(Builder); return EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, TrueCount); d892 1 a892 2 return EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, FalseBlock, TrueCount); a897 3 // The counter tells us how often we evaluate RHS, and all of TrueCount // can be propagated to that branch. uint64_t RHSCount = Cnt.getCount(); d900 1 a900 1 EmitBranchOnBoolExpr(CondBOp->getLHS(), LHSTrue, FalseBlock, RHSCount); a903 1 Cnt.beginRegion(Builder); d905 1 a905 1 EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, TrueCount); a906 2 Cnt.adjustForControlFlow(); Cnt.applyAdjustmentsToRegion(); d918 1 a918 3 Cnt.beginRegion(Builder); return EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, TrueCount); d926 1 a926 2 return EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, FalseBlock, TrueCount); a931 5 // We have the count for entry to the RHS and for the whole expression // being true, so we can divy up True count between the short circuit and // the RHS. uint64_t LHSCount = TrueCount - Cnt.getCount(); uint64_t RHSCount = TrueCount - LHSCount; d934 1 a934 1 EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, LHSFalse, LHSCount); a937 1 Cnt.beginRegion(Builder); d939 1 a939 2 EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, RHSCount); a940 2 Cnt.adjustForControlFlow(); Cnt.applyAdjustmentsToRegion(); d948 2 a949 7 if (CondUOp->getOpcode() == UO_LNot) { // Negate the count. uint64_t FalseCount = PGO.getCurrentRegionCount() - TrueCount; // Negate the condition and swap the destination blocks. return EmitBranchOnBoolExpr(CondUOp->getSubExpr(), FalseBlock, TrueBlock, FalseCount); } a956 1 RegionCounter Cnt = getPGORegionCounter(CondOp); d958 1 a958 12 EmitBranchOnBoolExpr(CondOp->getCond(), LHSBlock, RHSBlock, Cnt.getCount()); // When computing PGO branch weights, we only know the overall count for // the true block. This code is essentially doing tail duplication of the // naive code-gen, introducing new edges for which counts are not // available. Divide the counts proportionally between the LHS and RHS of // the conditional operator. uint64_t LHSScaledTrueCount = 0; if (TrueCount) { double LHSRatio = Cnt.getCount() / (double) PGO.getCurrentRegionCount(); LHSScaledTrueCount = TrueCount * LHSRatio; } d962 1 a962 3 Cnt.beginRegion(Builder); EmitBranchOnBoolExpr(CondOp->getLHS(), TrueBlock, FalseBlock, LHSScaledTrueCount); d967 1 a967 3 Cnt.beginElseRegion(); EmitBranchOnBoolExpr(CondOp->getRHS(), TrueBlock, FalseBlock, TrueCount - LHSScaledTrueCount); a982 6 // Create branch weights based on the number of times we get here and the // number of times the condition should be true. uint64_t CurrentCount = PGO.getCurrentRegionCountWithMin(TrueCount); llvm::MDNode *Weights = PGO.createBranchWeights(TrueCount, CurrentCount - TrueCount); d985 1 a985 1 Builder.CreateCondBr(CondV, TrueBlock, FalseBlock, Weights); @ 1.1.1.4 log @Import Clang 3.5svn r201163. @ text @a278 8 for (SmallVectorImpl >::iterator I = DeferredReplacements.begin(), E = DeferredReplacements.end(); I != E; ++I) { I->first->replaceAllUsesWith(I->second); I->first->eraseFromParent(); } d512 3 a514 4 // declaration. Also, in the case of -fno-inline attach NoInline // attribute to all function that are not marked AlwaysInline or ForceInline. if (const FunctionDecl *FD = dyn_cast_or_null(D)) { if (!CGM.getCodeGenOpts().NoInline) { a520 4 } else if (!FD->hasAttr() && !FD->hasAttr()) Fn->addFnAttr(llvm::Attribute::NoInline); } a580 8 if (CGM.getPGOData() && D) { // Turn on InlineHint attribute for hot functions. if (CGM.getPGOData()->isHotFunction(CGM.getMangledName(GD))) Fn->addFnAttr(llvm::Attribute::InlineHint); // Turn on Cold attribute for cold functions. else if (CGM.getPGOData()->isColdFunction(CGM.getMangledName(GD))) Fn->addFnAttr(llvm::Attribute::Cold); } a589 8 } else if (CurFnInfo->getReturnInfo().getKind() == ABIArgInfo::InAlloca && !hasScalarEvaluationKind(CurFnInfo->getReturnType())) { // Load the sret pointer from the argument struct and return into that. unsigned Idx = CurFnInfo->getReturnInfo().getInAllocaFieldIndex(); llvm::Function::arg_iterator EI = CurFn->arg_end(); --EI; llvm::Value *Addr = Builder.CreateStructGEP(EI, Idx); ReturnValue = Builder.CreateLoad(Addr, "agg.result"); d702 1 a702 1 QualType ResTy = FD->getReturnType(); d767 1 a767 1 !FD->getReturnType()->isVoidType() && Builder.GetInsertBlock()) { d964 1 a964 1 uint64_t LHSCount = Cnt.getParentCount() - Cnt.getCount(); d1447 1 a1447 1 type = cast(ty)->getReturnType(); @ 1.1.1.5 log @Import Clang 3.5svn r202566. @ text @d397 1 a397 1 d521 1 a521 1 // attribute to all function that are not marked AlwaysInline. d530 2 a531 1 } else if (!FD->hasAttr()) a690 20 /// When instrumenting to collect profile data, the counts for some blocks /// such as switch cases need to not include the fall-through counts, so /// emit a branch around the instrumentation code. When not instrumenting, /// this just calls EmitBlock(). void CodeGenFunction::EmitBlockWithFallThrough(llvm::BasicBlock *BB, RegionCounter &Cnt) { llvm::BasicBlock *SkipCountBB = 0; if (HaveInsertPoint() && CGM.getCodeGenOpts().ProfileInstrGenerate) { // When instrumenting for profiling, the fallthrough to certain // statements needs to skip over the instrumentation code so that we // get an accurate count. SkipCountBB = createBasicBlock("skipcount"); EmitBranch(SkipCountBB); } EmitBlock(BB); Cnt.beginRegion(Builder, /*AddIncomingFallThrough=*/true); if (SkipCountBB) EmitBlock(SkipCountBB); } d920 1 a923 2 RegionCounter Cnt = getPGORegionCounter(CondBOp); d960 2 a966 2 RegionCounter Cnt = getPGORegionCounter(CondBOp); d1006 2 d1040 1 a1040 1 double LHSRatio = Cnt.getCount() / (double) Cnt.getParentCount(); d1053 1 d1073 1 a1073 1 uint64_t CurrentCount = std::max(PGO.getCurrentRegionCount(), TrueCount); @ 1.1.1.5.2.1 log @Rebase. @ text @a17 1 #include "CGOpenMPRuntime.h" d37 1 a37 2 Builder(cgm.getModule().getContext(), llvm::ConstantFolder(), CGBuilderInserterTy(this)), CapturedStmtInfo(nullptr), d42 12 a53 14 SanOpts(&CGM.getSanOpts()), AutoreleaseResult(false), BlockInfo(nullptr), BlockPointer(nullptr), LambdaThisCaptureField(nullptr), NormalCleanupDest(nullptr), NextCleanupDestIndex(1), FirstBlockInfo(nullptr), EHResumeBlock(nullptr), ExceptionSlot(nullptr), EHSelectorSlot(nullptr), DebugInfo(CGM.getModuleDebugInfo()), DisableDebugInfo(false), DidCallStackSave(false), IndirectBranch(nullptr), PGO(cgm), SwitchInsn(nullptr), SwitchWeights(nullptr), CaseRangeBlock(nullptr), UnreachableBlock(nullptr), NumReturnExprs(0), NumSimpleReturnExprs(0), CXXABIThisDecl(nullptr), CXXABIThisValue(nullptr), CXXThisValue(nullptr), CXXDefaultInitExprThis(nullptr), CXXStructorImplicitParamDecl(nullptr), CXXStructorImplicitParamValue(nullptr), OutermostConditional(nullptr), CurLexicalScope(nullptr), TerminateLandingPad(nullptr), TerminateHandler(nullptr), TrapBB(nullptr) { a74 4 if (getLangOpts().OpenMP) { CGM.getOpenMPRuntime().FunctionFinished(*this); } d161 1 a161 1 dyn_cast(*ReturnBlock.getBlock()->user_begin()); d259 1 a259 1 AllocaInsertPt = nullptr; a341 2 const PrintingPolicy &Policy = ASTCtx.getPrintingPolicy(); d375 1 a375 2 std::string typeName = pointeeTy.getUnqualifiedType().getAsString(Policy) + "*"; d401 1 a401 1 std::string typeName = ty.getUnqualifiedType().getAsString(Policy); a497 16 /// Determine whether the function F ends with a return stmt. static bool endsWithReturn(const Decl* F) { const Stmt *Body = nullptr; if (auto *FD = dyn_cast_or_null(F)) Body = FD->getBody(); else if (auto *OMD = dyn_cast_or_null(F)) Body = OMD->getBody(); if (auto *CS = dyn_cast_or_null(Body)) { auto LastStmt = CS->body_rbegin(); if (LastStmt != CS->body_rend()) return isa(*LastStmt); } return false; } a502 1 SourceLocation Loc, d508 1 a508 1 CurFuncDecl = (D ? D->getNonClosureContext() : nullptr); d524 2 a525 1 for (auto RI : FD->redecls()) d581 3 a583 1 DI->EmitFunctionStart(GD, Loc, StartLoc, FnType, CurFn, Builder); d592 10 d604 1 a604 5 ReturnValue = nullptr; // Count the implicit return. if (!endsWithReturn(D)) ++NumReturnExprs; d609 1 a609 4 auto AI = CurFn->arg_begin(); if (CurFnInfo->getReturnInfo().isSRetAfterThis()) ++AI; ReturnValue = AI; d696 1 a696 1 llvm::BasicBlock *SkipCountBB = nullptr; d747 1 a747 1 DebugInfo = nullptr; // disable debug info indefinitely for this function a769 16 // Use the location of the start of the function to determine where // the function definition is located. By default use the location // of the declaration as the location for the subprogram. A function // may lack a declaration in the source code if it is created by code // gen. (examples: _GLOBAL__I_a, __cxx_global_array_dtor, thunk). SourceLocation Loc; if (FD) { Loc = FD->getLocation(); // If this is a function specialization then use the pattern body // as the location for the function. if (const FunctionDecl *SpecDecl = FD->getTemplateInstantiationPattern()) if (SpecDecl->hasBody(SpecDecl)) Loc = SpecDecl->getLocation(); } d771 1 a771 1 StartFunction(GD, ResTy, Fn, FnInfo, Args, Loc, BodyRange.getBegin()); a773 1 PGO.assignRegionCounters(GD.getDecl(), CurFn); d834 1 a834 1 PGO.emitInstrumentationData(); d843 1 a843 1 if (!S) return false; d875 1 a875 1 if (!S) return false; d1189 1 a1189 1 std::tie(numElts, eltType) = getVLASize(vlaType); d1201 1 a1201 1 vla = nullptr; d1238 1 a1238 1 if (!IndirectBranch) d1272 1 a1272 1 llvm::Value *numVLAElements = nullptr; d1365 1 a1365 1 llvm::Value *numElements = nullptr; d1378 1 a1378 1 // FIXME: Teach -fsanitize=undefined to trap this. d1582 3 a1584 1 for (const auto *I : D->specific_attrs()) d1587 1 a1587 1 I->getAnnotation(), D->getLocation()); d1597 3 a1599 1 for (const auto *I : D->specific_attrs()) { d1605 1 a1605 1 V = EmitAnnotationCall(F, V, I->getAnnotation(), D->getLocation()); a1612 27 void CodeGenFunction::InsertHelper(llvm::Instruction *I, const llvm::Twine &Name, llvm::BasicBlock *BB, llvm::BasicBlock::iterator InsertPt) const { LoopStack.InsertHelper(I); } template void CGBuilderInserter::InsertHelper( llvm::Instruction *I, const llvm::Twine &Name, llvm::BasicBlock *BB, llvm::BasicBlock::iterator InsertPt) const { llvm::IRBuilderDefaultInserter::InsertHelper(I, Name, BB, InsertPt); if (CGF) CGF->InsertHelper(I, Name, BB, InsertPt); } #ifdef NDEBUG #define PreserveNames false #else #define PreserveNames true #endif template void CGBuilderInserter::InsertHelper( llvm::Instruction *I, const llvm::Twine &Name, llvm::BasicBlock *BB, llvm::BasicBlock::iterator InsertPt) const; #undef PreserveNames @ 1.1.1.6 log @Import Clang 3.5svn r209886. @ text @a17 1 #include "CGOpenMPRuntime.h" d37 1 a37 2 Builder(cgm.getModule().getContext(), llvm::ConstantFolder(), CGBuilderInserterTy(this)), CapturedStmtInfo(nullptr), d42 12 a53 14 SanOpts(&CGM.getSanOpts()), AutoreleaseResult(false), BlockInfo(nullptr), BlockPointer(nullptr), LambdaThisCaptureField(nullptr), NormalCleanupDest(nullptr), NextCleanupDestIndex(1), FirstBlockInfo(nullptr), EHResumeBlock(nullptr), ExceptionSlot(nullptr), EHSelectorSlot(nullptr), DebugInfo(CGM.getModuleDebugInfo()), DisableDebugInfo(false), DidCallStackSave(false), IndirectBranch(nullptr), PGO(cgm), SwitchInsn(nullptr), SwitchWeights(nullptr), CaseRangeBlock(nullptr), UnreachableBlock(nullptr), NumReturnExprs(0), NumSimpleReturnExprs(0), CXXABIThisDecl(nullptr), CXXABIThisValue(nullptr), CXXThisValue(nullptr), CXXDefaultInitExprThis(nullptr), CXXStructorImplicitParamDecl(nullptr), CXXStructorImplicitParamValue(nullptr), OutermostConditional(nullptr), CurLexicalScope(nullptr), TerminateLandingPad(nullptr), TerminateHandler(nullptr), TrapBB(nullptr) { a74 4 if (getLangOpts().OpenMP) { CGM.getOpenMPRuntime().FunctionFinished(*this); } d161 1 a161 1 dyn_cast(*ReturnBlock.getBlock()->user_begin()); d259 1 a259 1 AllocaInsertPt = nullptr; a341 2 const PrintingPolicy &Policy = ASTCtx.getPrintingPolicy(); d375 1 a375 2 std::string typeName = pointeeTy.getUnqualifiedType().getAsString(Policy) + "*"; d401 1 a401 1 std::string typeName = ty.getUnqualifiedType().getAsString(Policy); a497 16 /// Determine whether the function F ends with a return stmt. static bool endsWithReturn(const Decl* F) { const Stmt *Body = nullptr; if (auto *FD = dyn_cast_or_null(F)) Body = FD->getBody(); else if (auto *OMD = dyn_cast_or_null(F)) Body = OMD->getBody(); if (auto *CS = dyn_cast_or_null(Body)) { auto LastStmt = CS->body_rbegin(); if (LastStmt != CS->body_rend()) return isa(*LastStmt); } return false; } a502 1 SourceLocation Loc, d508 1 a508 1 CurFuncDecl = (D ? D->getNonClosureContext() : nullptr); d524 2 a525 1 for (auto RI : FD->redecls()) d581 3 a583 1 DI->EmitFunctionStart(GD, Loc, StartLoc, FnType, CurFn, Builder); d592 10 d604 1 a604 5 ReturnValue = nullptr; // Count the implicit return. if (!endsWithReturn(D)) ++NumReturnExprs; d609 1 a609 4 auto AI = CurFn->arg_begin(); if (CurFnInfo->getReturnInfo().isSRetAfterThis()) ++AI; ReturnValue = AI; d696 1 a696 1 llvm::BasicBlock *SkipCountBB = nullptr; d747 1 a747 1 DebugInfo = nullptr; // disable debug info indefinitely for this function a769 16 // Use the location of the start of the function to determine where // the function definition is located. By default use the location // of the declaration as the location for the subprogram. A function // may lack a declaration in the source code if it is created by code // gen. (examples: _GLOBAL__I_a, __cxx_global_array_dtor, thunk). SourceLocation Loc; if (FD) { Loc = FD->getLocation(); // If this is a function specialization then use the pattern body // as the location for the function. if (const FunctionDecl *SpecDecl = FD->getTemplateInstantiationPattern()) if (SpecDecl->hasBody(SpecDecl)) Loc = SpecDecl->getLocation(); } d771 1 a771 1 StartFunction(GD, ResTy, Fn, FnInfo, Args, Loc, BodyRange.getBegin()); a773 1 PGO.assignRegionCounters(GD.getDecl(), CurFn); d834 1 a834 1 PGO.emitInstrumentationData(); d843 1 a843 1 if (!S) return false; d875 1 a875 1 if (!S) return false; d1189 1 a1189 1 std::tie(numElts, eltType) = getVLASize(vlaType); d1201 1 a1201 1 vla = nullptr; d1238 1 a1238 1 if (!IndirectBranch) d1272 1 a1272 1 llvm::Value *numVLAElements = nullptr; d1365 1 a1365 1 llvm::Value *numElements = nullptr; d1378 1 a1378 1 // FIXME: Teach -fsanitize=undefined to trap this. d1582 3 a1584 1 for (const auto *I : D->specific_attrs()) d1587 1 a1587 1 I->getAnnotation(), D->getLocation()); d1597 3 a1599 1 for (const auto *I : D->specific_attrs()) { d1605 1 a1605 1 V = EmitAnnotationCall(F, V, I->getAnnotation(), D->getLocation()); a1612 27 void CodeGenFunction::InsertHelper(llvm::Instruction *I, const llvm::Twine &Name, llvm::BasicBlock *BB, llvm::BasicBlock::iterator InsertPt) const { LoopStack.InsertHelper(I); } template void CGBuilderInserter::InsertHelper( llvm::Instruction *I, const llvm::Twine &Name, llvm::BasicBlock *BB, llvm::BasicBlock::iterator InsertPt) const { llvm::IRBuilderDefaultInserter::InsertHelper(I, Name, BB, InsertPt); if (CGF) CGF->InsertHelper(I, Name, BB, InsertPt); } #ifdef NDEBUG #define PreserveNames false #else #define PreserveNames true #endif template void CGBuilderInserter::InsertHelper( llvm::Instruction *I, const llvm::Twine &Name, llvm::BasicBlock *BB, llvm::BasicBlock::iterator InsertPt) const; #undef PreserveNames @ 1.1.1.7 log @Import clang 3.6svn r215315. @ text @d39 15 a53 12 CGBuilderInserterTy(this)), CapturedStmtInfo(nullptr), SanOpts(&CGM.getLangOpts().Sanitize), IsSanitizerScope(false), CurFuncIsThunk(false), AutoreleaseResult(false), BlockInfo(nullptr), BlockPointer(nullptr), LambdaThisCaptureField(nullptr), NormalCleanupDest(nullptr), NextCleanupDestIndex(1), FirstBlockInfo(nullptr), EHResumeBlock(nullptr), ExceptionSlot(nullptr), EHSelectorSlot(nullptr), DebugInfo(CGM.getModuleDebugInfo()), DisableDebugInfo(false), DidCallStackSave(false), IndirectBranch(nullptr), PGO(cgm), SwitchInsn(nullptr), SwitchWeights(nullptr), CaseRangeBlock(nullptr), UnreachableBlock(nullptr), NumReturnExprs(0), NumSimpleReturnExprs(0), CXXABIThisDecl(nullptr), CXXABIThisValue(nullptr), CXXThisValue(nullptr), a363 5 // MDNode for the kernel argument base type names. SmallVector argBaseTypeNames; argBaseTypeNames.push_back( llvm::MDString::get(Context, "kernel_arg_base_type")); d390 1 a390 1 if (pointeeTy.isCanonical() && pos != std::string::npos) a394 12 std::string baseTypeName = pointeeTy.getUnqualifiedType().getCanonicalType().getAsString( Policy) + "*"; // Turn "unsigned type" to "utype" pos = baseTypeName.find("unsigned"); if (pos != std::string::npos) baseTypeName.erase(pos+1, 8); argBaseTypeNames.push_back(llvm::MDString::get(Context, baseTypeName)); d416 1 a416 1 if (ty.isCanonical() && pos != std::string::npos) a420 10 std::string baseTypeName = ty.getUnqualifiedType().getCanonicalType().getAsString(Policy); // Turn "unsigned type" to "utype" pos = baseTypeName.find("unsigned"); if (pos != std::string::npos) baseTypeName.erase(pos+1, 8); argBaseTypeNames.push_back(llvm::MDString::get(Context, baseTypeName)); a447 1 kernelMDArgs.push_back(llvm::MDNode::get(Context, argBaseTypeNames)); d542 1 a542 1 if (CGM.getSanitizerBlacklist().isIn(*Fn)) d544 2 d797 10 a806 7 SourceLocation Loc = FD->getLocation(); // If this is a function specialization then use the pattern body // as the location for the function. if (const FunctionDecl *SpecDecl = FD->getTemplateInstantiationPattern()) if (SpecDecl->hasBody(SpecDecl)) Loc = SpecDecl->getLocation(); a811 1 PGO.checkGlobalDecl(GD); d855 1 a855 2 if (SanOpts->Return) { SanitizerScope SanScope(this); d859 1 a859 1 } else if (CGM.getCodeGenOpts().OptimizationLevel == 0) a1510 1 SanitizerScope SanScope(this); a1648 10 CodeGenFunction::SanitizerScope::SanitizerScope(CodeGenFunction *CGF) : CGF(CGF) { assert(!CGF->IsSanitizerScope); CGF->IsSanitizerScope = true; } CodeGenFunction::SanitizerScope::~SanitizerScope() { CGF->IsSanitizerScope = false; } a1653 5 if (IsSanitizerScope) { I->setMetadata( CGM.getModule().getMDKindID("nosanitize"), llvm::MDNode::get(CGM.getLLVMContext(), ArrayRef())); } @ 1.1.1.7.2.1 log @Update LLVM to 3.6.1, requested by joerg in ticket 824. @ text @d40 2 a41 3 CurFn(nullptr), CapturedStmtInfo(nullptr), SanOpts(CGM.getLangOpts().Sanitize), IsSanitizerScope(false), CurFuncIsThunk(false), AutoreleaseResult(false), SawAsmBlock(false), a64 6 if (CGM.getCodeGenOpts().NoNaNsFPMath) { FMF.setNoNaNs(); } if (CGM.getCodeGenOpts().NoSignedZeros) { FMF.setNoSignedZeros(); } a81 11 LValue CodeGenFunction::MakeNaturalAlignAddrLValue(llvm::Value *V, QualType T) { CharUnits Alignment; if (CGM.getCXXABI().isTypeInfoCalculable(T)) { Alignment = getContext().getTypeAlignInChars(T); unsigned MaxAlign = getContext().getLangOpts().MaxTypeAlign; if (MaxAlign && Alignment.getQuantity() > MaxAlign && !getContext().isAlignmentRequired(T)) Alignment = CharUnits::fromQuantity(MaxAlign); } return LValue::MakeAddr(V, T, Alignment, getContext(), CGM.getTBAAInfo(T)); } d143 1 a143 1 llvm::DebugLoc CodeGenFunction::EmitReturnBlock() { d158 1 a158 1 return llvm::DebugLoc(); d169 5 a173 3 // Record/return the DebugLoc of the simple 'return' expression to be used // later by the actual 'ret' instruction. llvm::DebugLoc Loc = BI->getDebugLoc(); d177 1 a177 1 return Loc; a185 1 return llvm::DebugLoc(); d239 1 a239 1 llvm::DebugLoc Loc = EmitReturnBlock(); d245 1 a245 1 if (CGDebugInfo *DI = getDebugInfo()) d247 1 a248 3 // Reset the debug location to that of the simple 'return' expression, if any // rather than that of the end of the function's scope '}'. ApplyDebugLocation AL(*this, Loc); d340 3 a342 3 CodeGenModule &CGM, llvm::LLVMContext &Context, SmallVector &kernelMDArgs, CGBuilderTy &Builder, ASTContext &ASTCtx) { d350 1 a350 1 SmallVector addressQuals; d354 1 a354 1 SmallVector accessQuals; d358 1 a358 1 SmallVector argTypeNames; d362 1 a362 1 SmallVector argBaseTypeNames; d367 1 a367 1 SmallVector argTypeQuals; d371 1 a371 1 SmallVector argNames; d383 2 a384 2 addressQuals.push_back(llvm::ConstantAsMetadata::get(Builder.getInt32( ASTCtx.getTargetAddressSpace(pointeeTy.getAddressSpace())))); d423 1 a423 2 addressQuals.push_back( llvm::ConstantAsMetadata::get(Builder.getInt32(AddrSpc))); d474 1 a474 2 if (CGM.getCodeGenOpts().EmitOpenCLArgMetadata) kernelMDArgs.push_back(llvm::MDNode::get(Context, argNames)); d485 2 a486 2 SmallVector kernelMDArgs; kernelMDArgs.push_back(llvm::ConstantAsMetadata::get(Fn)); d488 3 a490 2 GenOpenCLArgMetadata(FD, Fn, CGM, Context, kernelMDArgs, Builder, getContext()); d498 7 a504 7 llvm::Metadata *attrMDArgs[] = { llvm::MDString::get(Context, "vec_type_hint"), llvm::ConstantAsMetadata::get(llvm::UndefValue::get( CGM.getTypes().ConvertType(A->getTypeHint()))), llvm::ConstantAsMetadata::get(llvm::ConstantInt::get( llvm::IntegerType::get(Context, 32), llvm::APInt(32, (uint64_t)(isSignedInteger ? 1 : 0))))}; d509 6 a514 5 llvm::Metadata *attrMDArgs[] = { llvm::MDString::get(Context, "work_group_size_hint"), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getXDim())), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getYDim())), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getZDim()))}; d519 6 a524 5 llvm::Metadata *attrMDArgs[] = { llvm::MDString::get(Context, "reqd_work_group_size"), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getXDim())), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getYDim())), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getZDim()))}; a556 3 assert(!CurFn && "Do not use a CodeGenFunction object for more than one function"); d567 2 a568 2 if (CGM.isInSanitizerBlacklist(Fn, Loc)) SanOpts.clear(); d591 2 a592 2 // prologue data. if (getLangOpts().CPlusPlus && SanOpts.has(SanitizerKind::Function)) { d594 1 a594 1 if (llvm::Constant *PrologueSig = d598 4 a601 4 llvm::Constant *PrologueStructElems[] = { PrologueSig, FTRTTIConst }; llvm::Constant *PrologueStructConst = llvm::ConstantStruct::getAnon(PrologueStructElems, /*Packed=*/true); Fn->setPrologueData(PrologueStructConst); a693 8 for (auto *FD : MD->getParent()->fields()) { if (FD->hasCapturedVLAType()) { auto *ExprArg = EmitLoadOfLValue(EmitLValueForLambdaField(FD), SourceLocation()).getScalarVal(); auto VAT = FD->getCapturedVLAType(); VLASizeMap[VAT->getSizeExpr()] = ExprArg; } } a801 2 else if (CGM.getCXXABI().hasMostDerivedReturn(GD)) ResTy = CGM.getContext().VoidPtrTy; d804 3 a806 2 Args.append(FD->param_begin(), FD->param_end()); d874 1 a874 1 if (getLangOpts().CPlusPlus && !FD->hasImplicitReturnZero() && !SawAsmBlock && d876 1 a876 1 if (SanOpts.has(SanitizerKind::Return)) { d878 3 a880 4 llvm::Value *IsFalse = Builder.getFalse(); EmitCheck(std::make_pair(IsFalse, SanitizerKind::Return), "missing_return", EmitCheckSourceLocation(FD->getLocation()), None); d894 3 d1531 1 a1531 1 if (SanOpts.has(SanitizerKind::VLABound) && d1539 3 a1541 3 EmitCheck(std::make_pair(Builder.CreateICmpSGT(Size, Zero), SanitizerKind::VLABound), "vla_bound_not_positive", StaticArgs, Size); d1687 5 a1691 2 if (IsSanitizerScope) CGM.getSanitizerMetadata()->disableSanitizerForInstruction(I); @ 1.1.1.8 log @Import Clang 3.6RC1 r227398. @ text @d40 2 a41 3 CurFn(nullptr), CapturedStmtInfo(nullptr), SanOpts(CGM.getLangOpts().Sanitize), IsSanitizerScope(false), CurFuncIsThunk(false), AutoreleaseResult(false), SawAsmBlock(false), a64 6 if (CGM.getCodeGenOpts().NoNaNsFPMath) { FMF.setNoNaNs(); } if (CGM.getCodeGenOpts().NoSignedZeros) { FMF.setNoSignedZeros(); } a81 11 LValue CodeGenFunction::MakeNaturalAlignAddrLValue(llvm::Value *V, QualType T) { CharUnits Alignment; if (CGM.getCXXABI().isTypeInfoCalculable(T)) { Alignment = getContext().getTypeAlignInChars(T); unsigned MaxAlign = getContext().getLangOpts().MaxTypeAlign; if (MaxAlign && Alignment.getQuantity() > MaxAlign && !getContext().isAlignmentRequired(T)) Alignment = CharUnits::fromQuantity(MaxAlign); } return LValue::MakeAddr(V, T, Alignment, getContext(), CGM.getTBAAInfo(T)); } d143 1 a143 1 llvm::DebugLoc CodeGenFunction::EmitReturnBlock() { d158 1 a158 1 return llvm::DebugLoc(); d169 5 a173 3 // Record/return the DebugLoc of the simple 'return' expression to be used // later by the actual 'ret' instruction. llvm::DebugLoc Loc = BI->getDebugLoc(); d177 1 a177 1 return Loc; a185 1 return llvm::DebugLoc(); d239 1 a239 1 llvm::DebugLoc Loc = EmitReturnBlock(); d245 1 a245 1 if (CGDebugInfo *DI = getDebugInfo()) d247 1 a248 3 // Reset the debug location to that of the simple 'return' expression, if any // rather than that of the end of the function's scope '}'. ApplyDebugLocation AL(*this, Loc); d340 3 a342 3 CodeGenModule &CGM, llvm::LLVMContext &Context, SmallVector &kernelMDArgs, CGBuilderTy &Builder, ASTContext &ASTCtx) { d350 1 a350 1 SmallVector addressQuals; d354 1 a354 1 SmallVector accessQuals; d358 1 a358 1 SmallVector argTypeNames; d362 1 a362 1 SmallVector argBaseTypeNames; d367 1 a367 1 SmallVector argTypeQuals; d371 1 a371 1 SmallVector argNames; d383 2 a384 2 addressQuals.push_back(llvm::ConstantAsMetadata::get(Builder.getInt32( ASTCtx.getTargetAddressSpace(pointeeTy.getAddressSpace())))); d423 1 a423 2 addressQuals.push_back( llvm::ConstantAsMetadata::get(Builder.getInt32(AddrSpc))); d474 1 a474 2 if (CGM.getCodeGenOpts().EmitOpenCLArgMetadata) kernelMDArgs.push_back(llvm::MDNode::get(Context, argNames)); d485 2 a486 2 SmallVector kernelMDArgs; kernelMDArgs.push_back(llvm::ConstantAsMetadata::get(Fn)); d488 3 a490 2 GenOpenCLArgMetadata(FD, Fn, CGM, Context, kernelMDArgs, Builder, getContext()); d498 7 a504 7 llvm::Metadata *attrMDArgs[] = { llvm::MDString::get(Context, "vec_type_hint"), llvm::ConstantAsMetadata::get(llvm::UndefValue::get( CGM.getTypes().ConvertType(A->getTypeHint()))), llvm::ConstantAsMetadata::get(llvm::ConstantInt::get( llvm::IntegerType::get(Context, 32), llvm::APInt(32, (uint64_t)(isSignedInteger ? 1 : 0))))}; d509 6 a514 5 llvm::Metadata *attrMDArgs[] = { llvm::MDString::get(Context, "work_group_size_hint"), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getXDim())), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getYDim())), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getZDim()))}; d519 6 a524 5 llvm::Metadata *attrMDArgs[] = { llvm::MDString::get(Context, "reqd_work_group_size"), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getXDim())), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getYDim())), llvm::ConstantAsMetadata::get(Builder.getInt32(A->getZDim()))}; a556 3 assert(!CurFn && "Do not use a CodeGenFunction object for more than one function"); d567 2 a568 2 if (CGM.isInSanitizerBlacklist(Fn, Loc)) SanOpts.clear(); d591 2 a592 2 // prologue data. if (getLangOpts().CPlusPlus && SanOpts.has(SanitizerKind::Function)) { d594 1 a594 1 if (llvm::Constant *PrologueSig = d598 4 a601 4 llvm::Constant *PrologueStructElems[] = { PrologueSig, FTRTTIConst }; llvm::Constant *PrologueStructConst = llvm::ConstantStruct::getAnon(PrologueStructElems, /*Packed=*/true); Fn->setPrologueData(PrologueStructConst); a693 8 for (auto *FD : MD->getParent()->fields()) { if (FD->hasCapturedVLAType()) { auto *ExprArg = EmitLoadOfLValue(EmitLValueForLambdaField(FD), SourceLocation()).getScalarVal(); auto VAT = FD->getCapturedVLAType(); VLASizeMap[VAT->getSizeExpr()] = ExprArg; } } a801 2 else if (CGM.getCXXABI().hasMostDerivedReturn(GD)) ResTy = CGM.getContext().VoidPtrTy; d804 3 a806 2 Args.append(FD->param_begin(), FD->param_end()); d874 1 a874 1 if (getLangOpts().CPlusPlus && !FD->hasImplicitReturnZero() && !SawAsmBlock && d876 1 a876 1 if (SanOpts.has(SanitizerKind::Return)) { d878 3 a880 4 llvm::Value *IsFalse = Builder.getFalse(); EmitCheck(std::make_pair(IsFalse, SanitizerKind::Return), "missing_return", EmitCheckSourceLocation(FD->getLocation()), None); d894 3 d1531 1 a1531 1 if (SanOpts.has(SanitizerKind::VLABound) && d1539 3 a1541 3 EmitCheck(std::make_pair(Builder.CreateICmpSGT(Size, Zero), SanitizerKind::VLABound), "vla_bound_not_positive", StaticArgs, Size); d1687 5 a1691 2 if (IsSanitizerScope) CGM.getSanitizerMetadata()->disableSanitizerForInstruction(I); @ 1.1.1.9 log @Import Clang 3.8.0rc3 r261930. @ text @a14 2 #include "CGBlocks.h" #include "CGCleanup.h" a25 1 #include "clang/Basic/Builtins.h" a28 1 #include "clang/Sema/SemaDiagnostic.h" d38 1 a38 1 Builder(cgm, cgm.getModule().getContext(), llvm::ConstantFolder(), d40 1 a40 2 CurFn(nullptr), ReturnValue(Address::invalid()), CapturedStmtInfo(nullptr), a42 1 IsOutlinedSEHHelper(false), d47 6 a52 7 DebugInfo(CGM.getModuleDebugInfo()), DisableDebugInfo(false), DidCallStackSave(false), IndirectBranch(nullptr), PGO(cgm), SwitchInsn(nullptr), SwitchWeights(nullptr), CaseRangeBlock(nullptr), UnreachableBlock(nullptr), NumReturnExprs(0), NumSimpleReturnExprs(0), CXXABIThisDecl(nullptr), CXXABIThisValue(nullptr), CXXThisValue(nullptr), CXXStructorImplicitParamDecl(nullptr), d72 1 a72 4 if (CGM.getCodeGenOpts().ReciprocalMath) { FMF.setAllowReciprocal(); } Builder.setFastMathFlags(FMF); d85 1 a85 1 CGM.getOpenMPRuntime().functionFinished(*this); d89 1 a89 21 CharUnits CodeGenFunction::getNaturalPointeeTypeAlignment(QualType T, AlignmentSource *Source) { return getNaturalTypeAlignment(T->getPointeeType(), Source, /*forPointee*/ true); } CharUnits CodeGenFunction::getNaturalTypeAlignment(QualType T, AlignmentSource *Source, bool forPointeeType) { // Honor alignment typedef attributes even on incomplete types. // We also honor them straight for C++ class types, even as pointees; // there's an expressivity gap here. if (auto TT = T->getAs()) { if (auto Align = TT->getDecl()->getMaxAlignment()) { if (Source) *Source = AlignmentSource::AttributedType; return getContext().toCharUnitsFromBits(Align); } } if (Source) *Source = AlignmentSource::Type; d91 6 a96 20 if (T->isIncompleteType()) { Alignment = CharUnits::One(); // Shouldn't be used, but pessimistic is best. } else { // For C++ class pointees, we don't know whether we're pointing at a // base or a complete object, so we generally need to use the // non-virtual alignment. const CXXRecordDecl *RD; if (forPointeeType && (RD = T->getAsCXXRecordDecl())) { Alignment = CGM.getClassPointerAlignment(RD); } else { Alignment = getContext().getTypeAlignInChars(T); } // Cap to the global maximum type alignment unless the alignment // was somehow explicit on the type. if (unsigned MaxAlign = getLangOpts().MaxTypeAlign) { if (Alignment.getQuantity() > MaxAlign && !getContext().isAlignmentRequired(T)) Alignment = CharUnits::fromQuantity(MaxAlign); } d98 1 a98 1 return Alignment; a100 17 LValue CodeGenFunction::MakeNaturalAlignAddrLValue(llvm::Value *V, QualType T) { AlignmentSource AlignSource; CharUnits Alignment = getNaturalTypeAlignment(T, &AlignSource); return LValue::MakeAddr(Address(V, Alignment), T, getContext(), AlignSource, CGM.getTBAAInfo(T)); } /// Given a value of type T* that may not be to a complete object, /// construct an l-value with the natural pointee alignment of T. LValue CodeGenFunction::MakeNaturalAlignPointeeAddrLValue(llvm::Value *V, QualType T) { AlignmentSource AlignSource; CharUnits Align = getNaturalTypeAlignment(T, &AlignSource, /*pointee*/ true); return MakeAddrLValue(Address(V, Align), T, AlignSource); } a136 1 case Type::Pipe: d242 4 a245 5 bool HasCleanups = EHStack.stable_begin() != PrologueCleanupDepth; bool HasOnlyLifetimeMarkers = HasCleanups && EHStack.containsOnlyLifetimeMarkers(PrologueCleanupDepth); bool EmitRetDbgLoc = !HasCleanups || HasOnlyLifetimeMarkers; if (HasCleanups) { d248 2 a252 2 PopCleanupBlocks(PrologueCleanupDepth); a280 14 // If some of our locals escaped, insert a call to llvm.localescape in the // entry block. if (!EscapedLocals.empty()) { // Invert the map from local to index into a simple vector. There should be // no holes. SmallVector EscapeArgs; EscapeArgs.resize(EscapedLocals.size()); for (auto &Pair : EscapedLocals) EscapeArgs[Pair.second] = Pair.first; llvm::Function *FrameEscapeFn = llvm::Intrinsic::getDeclaration( &CGM.getModule(), llvm::Intrinsic::localescape); CGBuilderTy(*this, AllocaInsertPt).CreateCall(FrameEscapeFn, EscapeArgs); } d438 1 a438 2 bool isPipe = ty->isPipeType(); if (ty->isImageType() || isPipe) d446 1 a446 5 std::string typeName; if (isPipe) typeName = cast(ty)->getElementType().getAsString(Policy); else typeName = ty.getUnqualifiedType().getAsString(Policy); d455 1 a455 6 std::string baseTypeName; if (isPipe) baseTypeName = cast(ty)->getElementType().getCanonicalType().getAsString(Policy); else baseTypeName = a469 2 if (isPipe) typeQuals = "pipe"; d474 2 a475 4 // Get image and pipe access qualifier: // FIXME: now image and pipe share the same access qualifier maybe we can // refine it to OpenCL access qualifier and also handle write_read if (ty->isImageType()|| ty->isPipeType()) { a590 16 if (D) { // Apply the no_sanitize* attributes to SanOpts. for (auto Attr : D->specific_attrs()) SanOpts.Mask &= ~Attr->getMask(); } // Apply sanitizer attributes to the function. if (SanOpts.hasOneOf(SanitizerKind::Address | SanitizerKind::KernelAddress)) Fn->addFnAttr(llvm::Attribute::SanitizeAddress); if (SanOpts.has(SanitizerKind::Thread)) Fn->addFnAttr(llvm::Attribute::SanitizeThread); if (SanOpts.has(SanitizerKind::Memory)) Fn->addFnAttr(llvm::Attribute::SanitizeMemory); if (SanOpts.has(SanitizerKind::SafeStack)) Fn->addFnAttr(llvm::Attribute::SafeStack); a626 8 // If we're in C++ mode and the function name is "main", it is guaranteed // to be norecurse by the standard (3.6.1.3 "The function main shall not be // used within a program"). if (getLangOpts().CPlusPlus) if (const FunctionDecl *FD = dyn_cast_or_null(D)) if (FD->isMain()) Fn->addFnAttr(llvm::Attribute::NoRecurse); d663 1 a663 1 ReturnValue = Address::invalid(); d675 1 a675 1 ReturnValue = Address(&*AI, CurFnInfo->getReturnInfo().getIndirectAlign()); d682 2 a683 3 llvm::Value *Addr = Builder.CreateStructGEP(nullptr, &*EI, Idx); Addr = Builder.CreateAlignedLoad(Addr, getPointerAlign(), "agg.result"); ReturnValue = Address(Addr, getNaturalTypeAlignment(RetTy)); d756 2 a757 1 incrementProfileCounter(Body); d769 1 a769 1 const Stmt *S) { d779 1 a779 3 uint64_t CurrentCount = getCurrentProfileCount(); incrementProfileCounter(S); setCurrentProfileCount(getCurrentProfileCount() + CurrentCount); d792 7 a798 3 for (llvm::BasicBlock &BB : *F) for (llvm::Instruction &I : BB) if (I.mayThrow()) d800 3 d804 9 a812 1 F->setDoesNotThrow(); d835 2 a836 13 for (auto *Param : FD->params()) { Args.push_back(Param); if (!Param->hasAttr()) continue; IdentifierInfo *NoID = nullptr; auto *Implicit = ImplicitParamDecl::Create( getContext(), Param->getDeclContext(), Param->getLocation(), NoID, getContext().getSizeType()); SizeArguments[Param] = Implicit; Args.push_back(Implicit); } d862 2 a863 1 PGO.assignRegionCounters(GD, CurFn); d869 1 a869 1 !getLangOpts().CUDAIsDevice && d871 1 a871 1 CGM.getCUDARuntime().emitDeviceStub(*this, Args); d890 5 d912 2 a913 3 } else if (CGM.getCodeGenOpts().OptimizationLevel == 0) { EmitTrapCall(llvm::Intrinsic::trap); } d952 2 a953 2 for (const Stmt *SubStmt : S->children()) if (ContainsLabel(SubStmt, IgnoreCaseStmts)) d976 2 a977 2 for (const Stmt *SubStmt : S->children()) if (containsBreak(SubStmt)) d1031 2 d1039 1 a1039 1 incrementProfileCounter(CondBOp); d1058 1 a1058 1 uint64_t RHSCount = getProfileCount(CondBOp->getRHS()); d1061 2 a1062 8 { ApplyDebugLocation DL(*this, Cond); EmitBranchOnBoolExpr(CondBOp->getLHS(), LHSTrue, FalseBlock, RHSCount); EmitBlock(LHSTrue); } incrementProfileCounter(CondBOp); setCurrentProfileCount(getProfileCount(CondBOp->getRHS())); d1065 1 d1074 2 d1082 1 a1082 1 incrementProfileCounter(CondBOp); d1102 1 a1102 2 uint64_t LHSCount = getCurrentProfileCount() - getProfileCount(CondBOp->getRHS()); d1106 2 a1107 8 { ApplyDebugLocation DL(*this, Cond); EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, LHSFalse, LHSCount); EmitBlock(LHSFalse); } incrementProfileCounter(CondBOp); setCurrentProfileCount(getProfileCount(CondBOp->getRHS())); d1110 1 d1124 1 a1124 1 uint64_t FalseCount = getCurrentProfileCount() - TrueCount; d1136 1 d1138 1 a1138 2 EmitBranchOnBoolExpr(CondOp->getCond(), LHSBlock, RHSBlock, getProfileCount(CondOp)); d1147 1 a1147 2 double LHSRatio = getProfileCount(CondOp) / (double)getCurrentProfileCount(); d1153 3 a1155 6 incrementProfileCounter(CondOp); { ApplyDebugLocation DL(*this, Cond); EmitBranchOnBoolExpr(CondOp->getLHS(), TrueBlock, FalseBlock, LHSScaledTrueCount); } a1176 16 // If the branch has a condition wrapped by __builtin_unpredictable, // create metadata that specifies that the branch is unpredictable. // Don't bother if not optimizing because that metadata would not be used. llvm::MDNode *Unpredictable = nullptr; if (CGM.getCodeGenOpts().OptimizationLevel != 0) { if (const CallExpr *Call = dyn_cast(Cond)) { const Decl *TargetDecl = Call->getCalleeDecl(); if (const FunctionDecl *FD = dyn_cast_or_null(TargetDecl)) { if (FD->getBuiltinID() == Builtin::BI__builtin_unpredictable) { llvm::MDBuilder MDHelper(getLLVMContext()); Unpredictable = MDHelper.createUnpredictable(); } } } } d1179 3 a1181 3 uint64_t CurrentCount = std::max(getCurrentProfileCount(), TrueCount); llvm::MDNode *Weights = createProfileWeights(TrueCount, CurrentCount - TrueCount); d1184 2 a1185 6 llvm::Value *CondV; { ApplyDebugLocation DL(*this, Cond); CondV = EvaluateExprAsBool(Cond); } Builder.CreateCondBr(CondV, TrueBlock, FalseBlock, Weights, Unpredictable); d1202 1 a1202 1 Address dest, Address src, d1204 3 a1208 1 CharUnits baseSize = CGF.getContext().getTypeSizeInChars(baseType); d1210 1 a1210 1 = llvm::ConstantInt::get(CGF.IntPtrTy, baseSize.getQuantity()); d1212 4 a1215 4 Address begin = Builder.CreateElementBitCast(dest, CGF.Int8Ty, "vla.begin"); llvm::Value *end = Builder.CreateInBoundsGEP(begin.getPointer(), sizeInChars, "vla.end"); d1225 2 a1226 5 llvm::PHINode *cur = Builder.CreatePHI(begin.getType(), 2, "vla.cur"); cur->addIncoming(begin.getPointer(), originBB); CharUnits curAlign = dest.getAlignment().alignmentOfArrayElement(baseSize); d1229 2 a1230 1 Builder.CreateMemCpy(Address(cur, curAlign), src, baseSizeInChars, d1234 1 a1234 2 llvm::Value *next = Builder.CreateInBoundsGEP(CGF.Int8Ty, cur, baseSizeInChars, "vla.next"); d1245 1 a1245 1 CodeGenFunction::EmitNullInitialization(Address DestPtr, QualType Ty) { d1255 5 a1259 2 if (DestPtr.getElementType() != Int8Ty) DestPtr = Builder.CreateElementBitCast(DestPtr, Int8Ty); d1262 4 a1265 1 CharUnits size = getContext().getTypeSizeInChars(Ty); d1271 1 a1271 1 if (size.isZero()) { d1289 1 a1289 1 SizeVal = CGM.getSize(size); d1308 2 a1309 4 CharUnits NullAlign = DestPtr.getAlignment(); NullVariable->setAlignment(NullAlign.getQuantity()); Address SrcPtr(Builder.CreateBitCast(NullVariable, Builder.getInt8PtrTy()), NullAlign); d1314 1 a1314 1 Builder.CreateMemCpy(DestPtr, SrcPtr, SizeVal, false); d1321 2 a1322 1 Builder.CreateMemSet(DestPtr, Builder.getInt8(0), SizeVal, false); d1341 1 a1341 1 CGBuilderTy TmpBuilder(*this, createBasicBlock("indirectgoto")); d1356 1 a1356 1 Address &addr) { d1395 2 a1396 1 dyn_cast(addr.getElementType()); d1424 3 a1426 2 llvm::Type *baseType = ConvertType(eltType); addr = Builder.CreateElementBitCast(addr, baseType, "array.begin"); d1429 1 a1429 3 addr = Address(Builder.CreateInBoundsGEP(addr.getPointer(), gepIndices, "array.begin"), addr.getAlignment()); a1609 4 case Type::Pipe: type = cast(ty)->getElementType(); break; d1614 1 a1614 1 Address CodeGenFunction::EmitVAListRef(const Expr* E) { d1616 1 a1616 5 return EmitPointerWithAlignment(E); return EmitLValue(E).getAddress(); } Address CodeGenFunction::EmitMSVAListRef(const Expr *E) { d1678 2 a1679 2 Address CodeGenFunction::EmitFieldAnnotations(const FieldDecl *D, Address Addr) { a1680 1 llvm::Value *V = Addr.getPointer(); d1695 1 a1695 1 return Address(V, Addr.getAlignment()); a1737 77 static bool hasRequiredFeatures(const SmallVectorImpl &ReqFeatures, CodeGenModule &CGM, const FunctionDecl *FD, std::string &FirstMissing) { // If there aren't any required features listed then go ahead and return. if (ReqFeatures.empty()) return false; // Now build up the set of caller features and verify that all the required // features are there. llvm::StringMap CallerFeatureMap; CGM.getFunctionFeatureMap(CallerFeatureMap, FD); // If we have at least one of the features in the feature list return // true, otherwise return false. return std::all_of( ReqFeatures.begin(), ReqFeatures.end(), [&](StringRef Feature) { SmallVector OrFeatures; Feature.split(OrFeatures, "|"); return std::any_of(OrFeatures.begin(), OrFeatures.end(), [&](StringRef Feature) { if (!CallerFeatureMap.lookup(Feature)) { FirstMissing = Feature.str(); return false; } return true; }); }); } // Emits an error if we don't have a valid set of target features for the // called function. void CodeGenFunction::checkTargetFeatures(const CallExpr *E, const FunctionDecl *TargetDecl) { // Early exit if this is an indirect call. if (!TargetDecl) return; // Get the current enclosing function if it exists. If it doesn't // we can't check the target features anyhow. const FunctionDecl *FD = dyn_cast_or_null(CurFuncDecl); if (!FD) return; // Grab the required features for the call. For a builtin this is listed in // the td file with the default cpu, for an always_inline function this is any // listed cpu and any listed features. unsigned BuiltinID = TargetDecl->getBuiltinID(); std::string MissingFeature; if (BuiltinID) { SmallVector ReqFeatures; const char *FeatureList = CGM.getContext().BuiltinInfo.getRequiredFeatures(BuiltinID); // Return if the builtin doesn't have any required features. if (!FeatureList || StringRef(FeatureList) == "") return; StringRef(FeatureList).split(ReqFeatures, ","); if (!hasRequiredFeatures(ReqFeatures, CGM, FD, MissingFeature)) CGM.getDiags().Report(E->getLocStart(), diag::err_builtin_needs_feature) << TargetDecl->getDeclName() << CGM.getContext().BuiltinInfo.getRequiredFeatures(BuiltinID); } else if (TargetDecl->hasAttr()) { // Get the required features for the callee. SmallVector ReqFeatures; llvm::StringMap CalleeFeatureMap; CGM.getFunctionFeatureMap(CalleeFeatureMap, TargetDecl); for (const auto &F : CalleeFeatureMap) { // Only positive features are "required". if (F.getValue()) ReqFeatures.push_back(F.getKey()); } if (!hasRequiredFeatures(ReqFeatures, CGM, FD, MissingFeature)) CGM.getDiags().Report(E->getLocStart(), diag::err_function_needs_feature) << FD->getDeclName() << TargetDecl->getDeclName() << MissingFeature; } } @ 1.1.1.9.2.1 log @Sync with HEAD @ text @a27 1 #include "clang/AST/StmtObjC.h" a39 20 /// shouldEmitLifetimeMarkers - Decide whether we need emit the life-time /// markers. static bool shouldEmitLifetimeMarkers(const CodeGenOptions &CGOpts, const LangOptions &LangOpts) { if (CGOpts.DisableLifetimeMarkers) return false; // Asan uses markers for use-after-scope checks. if (CGOpts.SanitizeAddressUseAfterScope) return true; // Disable lifetime markers in msan builds. // FIXME: Remove this when msan works with lifetime markers. if (LangOpts.Sanitize.has(SanitizerKind::Memory)) return false; // For now, only in optimized builds. return CGOpts.OptimizationLevel != 0; } d45 9 a53 7 CapturedStmtInfo(nullptr), SanOpts(CGM.getLangOpts().Sanitize), IsSanitizerScope(false), CurFuncIsThunk(false), AutoreleaseResult(false), SawAsmBlock(false), IsOutlinedSEHHelper(false), BlockInfo(nullptr), BlockPointer(nullptr), LambdaThisCaptureField(nullptr), NormalCleanupDest(nullptr), NextCleanupDestIndex(1), FirstBlockInfo(nullptr), EHResumeBlock(nullptr), ExceptionSlot(nullptr), EHSelectorSlot(nullptr), DebugInfo(CGM.getModuleDebugInfo()), d62 1 a62 3 TerminateHandler(nullptr), TrapBB(nullptr), ShouldEmitLifetimeMarkers( shouldEmitLifetimeMarkers(CGM.getCodeGenOpts(), CGM.getLangOpts())) { d94 1 a94 1 if (getLangOpts().OpenMP && CurFn) d96 1 a399 6 /// ShouldXRayInstrument - Return true if the current function should be /// instrumented with XRay nop sleds. bool CodeGenFunction::ShouldXRayInstrumentFunction() const { return CGM.getCodeGenOpts().XRayInstrumentFunctions; } a403 1 auto NL = ApplyDebugLocation::CreateArtificial(*this); d424 2 a425 19 static void removeImageAccessQualifier(std::string& TyName) { std::string ReadOnlyQual("__read_only"); std::string::size_type ReadOnlyPos = TyName.find(ReadOnlyQual); if (ReadOnlyPos != std::string::npos) // "+ 1" for the space after access qualifier. TyName.erase(ReadOnlyPos, ReadOnlyQual.size() + 1); else { std::string WriteOnlyQual("__write_only"); std::string::size_type WriteOnlyPos = TyName.find(WriteOnlyQual); if (WriteOnlyPos != std::string::npos) TyName.erase(WriteOnlyPos, WriteOnlyQual.size() + 1); else { std::string ReadWriteQual("__read_write"); std::string::size_type ReadWritePos = TyName.find(ReadWriteQual); if (ReadWritePos != std::string::npos) TyName.erase(ReadWritePos, ReadWriteQual.size() + 1); } } } d427 3 a429 15 // Returns the address space id that should be produced to the // kernel_arg_addr_space metadata. This is always fixed to the ids // as specified in the SPIR 2.0 specification in order to differentiate // for example in clGetKernelArgInfo() implementation between the address // spaces with targets without unique mapping to the OpenCL address spaces // (basically all single AS CPUs). static unsigned ArgInfoAddressSpace(unsigned LangAS) { switch (LangAS) { case LangAS::opencl_global: return 1; case LangAS::opencl_constant: return 2; case LangAS::opencl_local: return 3; case LangAS::opencl_generic: return 4; // Not in SPIR 2.0 specs. default: return 0; // Assume private. } d437 1 d447 1 d451 1 d455 1 d459 2 d464 1 d468 1 d480 1 a480 1 ArgInfoAddressSpace(pointeeTy.getAddressSpace())))); d517 2 a518 1 AddrSpc = ArgInfoAddressSpace(LangAS::opencl_global); d526 1 a526 2 typeName = ty.getCanonicalType()->getAs()->getElementType() .getAsString(Policy); d535 2 d539 2 a540 3 baseTypeName = ty.getCanonicalType()->getAs() ->getElementType().getCanonicalType() .getAsString(Policy); a544 11 // Remove access qualifiers on images // (as they are inseparable from type in clang implementation, // but OpenCL spec provides a special query to get access qualifier // via clGetKernelArgInfo with CL_KERNEL_ARG_ACCESS_QUALIFIER): if (ty->isImageType()) { removeImageAccessQualifier(typeName); removeImageAccessQualifier(baseTypeName); } argTypeNames.push_back(llvm::MDString::get(Context, typeName)); d564 2 d567 1 a567 1 const OpenCLAccessAttr *A = parm->getAttr(); a569 2 else if (A && A->isReadWrite()) accessQuals.push_back(llvm::MDString::get(Context, "read_write")); d572 1 d580 5 a584 10 Fn->setMetadata("kernel_arg_addr_space", llvm::MDNode::get(Context, addressQuals)); Fn->setMetadata("kernel_arg_access_qual", llvm::MDNode::get(Context, accessQuals)); Fn->setMetadata("kernel_arg_type", llvm::MDNode::get(Context, argTypeNames)); Fn->setMetadata("kernel_arg_base_type", llvm::MDNode::get(Context, argBaseTypeNames)); Fn->setMetadata("kernel_arg_type_qual", llvm::MDNode::get(Context, argTypeQuals)); d586 1 a586 2 Fn->setMetadata("kernel_arg_name", llvm::MDNode::get(Context, argNames)); d597 5 a601 1 GenOpenCLArgMetadata(FD, Fn, CGM, Context, Builder, getContext()); d610 1 d616 1 a616 1 Fn->setMetadata("vec_type_hint", llvm::MDNode::get(Context, attrMDArgs)); d621 1 d625 1 a625 1 Fn->setMetadata("work_group_size_hint", llvm::MDNode::get(Context, attrMDArgs)); d630 1 d634 1 a634 1 Fn->setMetadata("reqd_work_group_size", llvm::MDNode::get(Context, attrMDArgs)); d636 5 a672 3 if (const auto *FD = dyn_cast_or_null(D)) if (FD->usesSEHTry()) CurSEHParent = FD; d698 12 a709 26 // Ignore TSan memory acesses from within ObjC/ObjC++ dealloc, initialize, // .cxx_destruct and all of their calees at run time. if (SanOpts.has(SanitizerKind::Thread)) { if (const auto *OMD = dyn_cast_or_null(D)) { IdentifierInfo *II = OMD->getSelector().getIdentifierInfoForSlot(0); if (OMD->getMethodFamily() == OMF_dealloc || OMD->getMethodFamily() == OMF_initialize || (OMD->getSelector().isUnarySelector() && II->isStr(".cxx_destruct"))) { Fn->addFnAttr("sanitize_thread_no_checking_at_run_time"); Fn->removeFnAttr(llvm::Attribute::SanitizeThread); } } } // Apply xray attributes to the function (as a string, for now) if (D && ShouldXRayInstrumentFunction()) { if (const auto *XRayAttr = D->getAttr()) { if (XRayAttr->alwaysXRayInstrument()) Fn->addFnAttr("function-instrument", "xray-always"); if (XRayAttr->neverXRayInstrument()) Fn->addFnAttr("function-instrument", "xray-never"); } else { Fn->addFnAttr( "xray-instruction-threshold", llvm::itostr(CGM.getCodeGenOpts().XRayInstructionThreshold)); } a711 8 if (const FunctionDecl *FD = dyn_cast_or_null(D)) if (CGM.getLangOpts().OpenMP && FD->hasAttr()) CGM.getOpenMPRuntime().emitDeclareSimdFunction(FD, Fn); // Add no-jump-tables value. Fn->addFnAttr("no-jump-tables", llvm::toStringRef(CGM.getCodeGenOpts().NoUseJumpTables)); d741 1 a741 1 d748 3 a750 1 AllocaInsertPt = new llvm::BitCastInst(Undef, Int32Ty, "allocapt", EntryBB); a757 7 // Reconstruct the type from the argument list so that implicit parameters, // such as 'this' and 'vtt', show up in the debug info. Preserve the calling // convention. CallingConv CC = CallingConv::CC_C; if (auto *FD = dyn_cast_or_null(D)) if (const auto *SrcFnTy = FD->getType()->getAs()) CC = SrcFnTy->getCallConv(); d759 8 a766 4 for (const VarDecl *VD : Args) ArgTypes.push_back(VD->getType()); QualType FnType = getContext().getFunctionType( RetTy, ArgTypes, FunctionProtoType::ExtProtoInfo(CC)); a772 4 // Since emitting the mcount call here impacts optimizations such as function // inlining, we just add an attribute to insert a mcount call in backend. // The attribute "counting-function" is set to mcount function name which is // architecture dependent. d774 1 a774 1 Fn->addFnAttr("counting-function", getTarget().getMCountName()); d826 4 a829 16 // If the lambda captures the object referred to by '*this' - either by // value or by reference, make sure CXXThisValue points to the correct // object. // Get the lvalue for the field (which is a copy of the enclosing object // or contains the address of the enclosing object). LValue ThisFieldLValue = EmitLValueForLambdaField(LambdaThisCaptureField); if (!LambdaThisCaptureField->getType()->isPointerType()) { // If the enclosing object was captured by value, just use its address. CXXThisValue = ThisFieldLValue.getAddress().getPointer(); } else { // Load the lvalue pointed to by the field, since '*this' was captured // by reference. CXXThisValue = EmitLoadOfLValue(ThisFieldLValue, SourceLocation()).getScalarVal(); } d886 1 a886 1 if (HaveInsertPoint() && CGM.getCodeGenOpts().hasProfileClangInstr()) { d907 1 a907 1 if (F->isInterposable()) return; d917 2 a918 2 QualType CodeGenFunction::BuildFunctionArgList(GlobalDecl GD, FunctionArgList &Args) { d920 6 d928 1 d938 11 a948 22 // The base version of an inheriting constructor whose constructed base is a // virtual base is not passed any arguments (because it doesn't actually call // the inherited constructor). bool PassedParams = true; if (const CXXConstructorDecl *CD = dyn_cast(FD)) if (auto Inherited = CD->getInheritedConstructor()) PassedParams = getTypes().inheritingCtorHasParams(Inherited, GD.getCtorType()); if (PassedParams) { for (auto *Param : FD->parameters()) { Args.push_back(Param); if (!Param->hasAttr()) continue; IdentifierInfo *NoID = nullptr; auto *Implicit = ImplicitParamDecl::Create( getContext(), Param->getDeclContext(), Param->getLocation(), NoID, getContext().getSizeType()); SizeArguments[Param] = Implicit; Args.push_back(Implicit); } a953 28 return ResTy; } static bool shouldUseUndefinedBehaviorReturnOptimization(const FunctionDecl *FD, const ASTContext &Context) { QualType T = FD->getReturnType(); // Avoid the optimization for functions that return a record type with a // trivial destructor or another trivially copyable type. if (const RecordType *RT = T.getCanonicalType()->getAs()) { if (const auto *ClassDecl = dyn_cast(RT->getDecl())) return !ClassDecl->hasTrivialDestructor(); } return !T.isTriviallyCopyableType(Context); } void CodeGenFunction::GenerateCode(GlobalDecl GD, llvm::Function *Fn, const CGFunctionInfo &FnInfo) { const FunctionDecl *FD = cast(GD.getDecl()); CurGD = GD; FunctionArgList Args; QualType ResTy = BuildFunctionArgList(GD, Args); // Check if we should generate debug info for this function. if (FD->hasAttr()) DebugInfo = nullptr; // disable debug info indefinitely for this function a970 7 Stmt *Body = FD->getBody(); // Initialize helper which will detect jumps which can cause invalid lifetime // markers. if (Body && ShouldEmitLifetimeMarkers) Bypasses.Init(Body); d1000 1 a1000 1 } else if (Body) { a1012 3 bool ShouldEmitUnreachable = CGM.getCodeGenOpts().StrictReturn || shouldUseUndefinedBehaviorReturnOptimization(FD, getContext()); d1017 4 a1020 9 SanitizerHandler::MissingReturn, EmitCheckSourceLocation(FD->getLocation()), None); } else if (ShouldEmitUnreachable) { if (CGM.getCodeGenOpts().OptimizationLevel == 0) EmitTrapCall(llvm::Intrinsic::trap); } if (SanOpts.has(SanitizerKind::Return) || ShouldEmitUnreachable) { Builder.CreateUnreachable(); Builder.ClearInsertionPoint(); d1022 2 a1090 22 bool CodeGenFunction::mightAddDeclToScope(const Stmt *S) { if (!S) return false; // Some statement kinds add a scope and thus never add a decl to the current // scope. Note, this list is longer than the list of statements that might // have an unscoped decl nested within them, but this way is conservatively // correct even if more statement kinds are added. if (isa(S) || isa(S) || isa(S) || isa(S) || isa(S) || isa(S) || isa(S) || isa(S) || isa(S) || isa(S)) return false; if (isa(S)) return true; for (const Stmt *SubStmt : S->children()) if (mightAddDeclToScope(SubStmt)) return true; return false; } d1096 1 a1096 2 bool &ResultBool, bool AllowLabels) { d1098 1 a1098 1 if (!ConstantFoldsToSimpleInteger(Cond, ResultInt, AllowLabels)) d1108 2 a1109 3 bool CodeGenFunction::ConstantFoldsToSimpleInteger(const Expr *Cond, llvm::APSInt &ResultInt, bool AllowLabels) { d1116 1 a1116 1 if (!AllowLabels && CodeGenFunction::ContainsLabel(Cond)) d1300 9 a1308 6 auto *Call = dyn_cast(Cond); if (Call && CGM.getCodeGenOpts().OptimizationLevel != 0) { auto *FD = dyn_cast_or_null(Call->getCalleeDecl()); if (FD && FD->getBuiltinID() == Builtin::BI__builtin_unpredictable) { llvm::MDBuilder MDHelper(getLLVMContext()); Unpredictable = MDHelper.createUnpredictable(); a1640 1 case Type::ObjCTypeParam: d1704 1 a1704 1 SanitizerHandler::VLABoundNotPositive, StaticArgs, Size); d1764 2 a1765 2 const APValue &Init) { assert(!Init.isUninit() && "Invalid DeclRefExpr initializer!"); d1767 1 a1767 1 if (CGM.getCodeGenOpts().getDebugInfo() >= codegenoptions::LimitedDebugInfo) d1863 2 a1864 1 void CGBuilderInserter::InsertHelper( d1867 2 a1868 1 llvm::IRBuilderDefaultInserter::InsertHelper(I, Name, BB, InsertPt); d1873 10 a1958 16 void CodeGenFunction::EmitSanitizerStatReport(llvm::SanitizerStatKind SSK) { if (!CGM.getCodeGenOpts().SanitizeStats) return; llvm::IRBuilder<> IRB(Builder.GetInsertBlock(), Builder.GetInsertPoint()); IRB.SetCurrentDebugLocation(Builder.getCurrentDebugLocation()); CGM.getSanStats().create(IRB, SSK); } llvm::DebugLoc CodeGenFunction::SourceLocToDebugLoc(SourceLocation Location) { if (CGDebugInfo *DI = getDebugInfo()) return DI->SourceLocToDebugLoc(Location); return llvm::DebugLoc(); } @ 1.1.1.10 log @Import Clang pre-4.0.0 r291444. @ text @a27 1 #include "clang/AST/StmtObjC.h" a39 20 /// shouldEmitLifetimeMarkers - Decide whether we need emit the life-time /// markers. static bool shouldEmitLifetimeMarkers(const CodeGenOptions &CGOpts, const LangOptions &LangOpts) { if (CGOpts.DisableLifetimeMarkers) return false; // Asan uses markers for use-after-scope checks. if (CGOpts.SanitizeAddressUseAfterScope) return true; // Disable lifetime markers in msan builds. // FIXME: Remove this when msan works with lifetime markers. if (LangOpts.Sanitize.has(SanitizerKind::Memory)) return false; // For now, only in optimized builds. return CGOpts.OptimizationLevel != 0; } d45 9 a53 7 CapturedStmtInfo(nullptr), SanOpts(CGM.getLangOpts().Sanitize), IsSanitizerScope(false), CurFuncIsThunk(false), AutoreleaseResult(false), SawAsmBlock(false), IsOutlinedSEHHelper(false), BlockInfo(nullptr), BlockPointer(nullptr), LambdaThisCaptureField(nullptr), NormalCleanupDest(nullptr), NextCleanupDestIndex(1), FirstBlockInfo(nullptr), EHResumeBlock(nullptr), ExceptionSlot(nullptr), EHSelectorSlot(nullptr), DebugInfo(CGM.getModuleDebugInfo()), d62 1 a62 3 TerminateHandler(nullptr), TrapBB(nullptr), ShouldEmitLifetimeMarkers( shouldEmitLifetimeMarkers(CGM.getCodeGenOpts(), CGM.getLangOpts())) { a399 6 /// ShouldXRayInstrument - Return true if the current function should be /// instrumented with XRay nop sleds. bool CodeGenFunction::ShouldXRayInstrumentFunction() const { return CGM.getCodeGenOpts().XRayInstrumentFunctions; } a403 1 auto NL = ApplyDebugLocation::CreateArtificial(*this); d424 2 a425 19 static void removeImageAccessQualifier(std::string& TyName) { std::string ReadOnlyQual("__read_only"); std::string::size_type ReadOnlyPos = TyName.find(ReadOnlyQual); if (ReadOnlyPos != std::string::npos) // "+ 1" for the space after access qualifier. TyName.erase(ReadOnlyPos, ReadOnlyQual.size() + 1); else { std::string WriteOnlyQual("__write_only"); std::string::size_type WriteOnlyPos = TyName.find(WriteOnlyQual); if (WriteOnlyPos != std::string::npos) TyName.erase(WriteOnlyPos, WriteOnlyQual.size() + 1); else { std::string ReadWriteQual("__read_write"); std::string::size_type ReadWritePos = TyName.find(ReadWriteQual); if (ReadWritePos != std::string::npos) TyName.erase(ReadWritePos, ReadWriteQual.size() + 1); } } } d427 3 a429 15 // Returns the address space id that should be produced to the // kernel_arg_addr_space metadata. This is always fixed to the ids // as specified in the SPIR 2.0 specification in order to differentiate // for example in clGetKernelArgInfo() implementation between the address // spaces with targets without unique mapping to the OpenCL address spaces // (basically all single AS CPUs). static unsigned ArgInfoAddressSpace(unsigned LangAS) { switch (LangAS) { case LangAS::opencl_global: return 1; case LangAS::opencl_constant: return 2; case LangAS::opencl_local: return 3; case LangAS::opencl_generic: return 4; // Not in SPIR 2.0 specs. default: return 0; // Assume private. } d437 1 d447 1 d451 1 d455 1 d459 2 d464 1 d468 1 d480 1 a480 1 ArgInfoAddressSpace(pointeeTy.getAddressSpace())))); d517 2 a518 1 AddrSpc = ArgInfoAddressSpace(LangAS::opencl_global); d526 1 a526 2 typeName = ty.getCanonicalType()->getAs()->getElementType() .getAsString(Policy); d535 2 d539 2 a540 3 baseTypeName = ty.getCanonicalType()->getAs() ->getElementType().getCanonicalType() .getAsString(Policy); a544 11 // Remove access qualifiers on images // (as they are inseparable from type in clang implementation, // but OpenCL spec provides a special query to get access qualifier // via clGetKernelArgInfo with CL_KERNEL_ARG_ACCESS_QUALIFIER): if (ty->isImageType()) { removeImageAccessQualifier(typeName); removeImageAccessQualifier(baseTypeName); } argTypeNames.push_back(llvm::MDString::get(Context, typeName)); d564 2 d567 1 a567 1 const OpenCLAccessAttr *A = parm->getAttr(); a569 2 else if (A && A->isReadWrite()) accessQuals.push_back(llvm::MDString::get(Context, "read_write")); d572 1 d580 5 a584 10 Fn->setMetadata("kernel_arg_addr_space", llvm::MDNode::get(Context, addressQuals)); Fn->setMetadata("kernel_arg_access_qual", llvm::MDNode::get(Context, accessQuals)); Fn->setMetadata("kernel_arg_type", llvm::MDNode::get(Context, argTypeNames)); Fn->setMetadata("kernel_arg_base_type", llvm::MDNode::get(Context, argBaseTypeNames)); Fn->setMetadata("kernel_arg_type_qual", llvm::MDNode::get(Context, argTypeQuals)); d586 1 a586 2 Fn->setMetadata("kernel_arg_name", llvm::MDNode::get(Context, argNames)); d597 5 a601 1 GenOpenCLArgMetadata(FD, Fn, CGM, Context, Builder, getContext()); d610 1 d616 1 a616 1 Fn->setMetadata("vec_type_hint", llvm::MDNode::get(Context, attrMDArgs)); d621 1 d625 1 a625 1 Fn->setMetadata("work_group_size_hint", llvm::MDNode::get(Context, attrMDArgs)); d630 1 d634 1 a634 1 Fn->setMetadata("reqd_work_group_size", llvm::MDNode::get(Context, attrMDArgs)); d636 5 a672 3 if (const auto *FD = dyn_cast_or_null(D)) if (FD->usesSEHTry()) CurSEHParent = FD; d698 12 a709 26 // Ignore TSan memory acesses from within ObjC/ObjC++ dealloc, initialize, // .cxx_destruct and all of their calees at run time. if (SanOpts.has(SanitizerKind::Thread)) { if (const auto *OMD = dyn_cast_or_null(D)) { IdentifierInfo *II = OMD->getSelector().getIdentifierInfoForSlot(0); if (OMD->getMethodFamily() == OMF_dealloc || OMD->getMethodFamily() == OMF_initialize || (OMD->getSelector().isUnarySelector() && II->isStr(".cxx_destruct"))) { Fn->addFnAttr("sanitize_thread_no_checking_at_run_time"); Fn->removeFnAttr(llvm::Attribute::SanitizeThread); } } } // Apply xray attributes to the function (as a string, for now) if (D && ShouldXRayInstrumentFunction()) { if (const auto *XRayAttr = D->getAttr()) { if (XRayAttr->alwaysXRayInstrument()) Fn->addFnAttr("function-instrument", "xray-always"); if (XRayAttr->neverXRayInstrument()) Fn->addFnAttr("function-instrument", "xray-never"); } else { Fn->addFnAttr( "xray-instruction-threshold", llvm::itostr(CGM.getCodeGenOpts().XRayInstructionThreshold)); } a711 8 if (const FunctionDecl *FD = dyn_cast_or_null(D)) if (CGM.getLangOpts().OpenMP && FD->hasAttr()) CGM.getOpenMPRuntime().emitDeclareSimdFunction(FD, Fn); // Add no-jump-tables value. Fn->addFnAttr("no-jump-tables", llvm::toStringRef(CGM.getCodeGenOpts().NoUseJumpTables)); d741 1 a741 1 d748 3 a750 1 AllocaInsertPt = new llvm::BitCastInst(Undef, Int32Ty, "allocapt", EntryBB); a757 7 // Reconstruct the type from the argument list so that implicit parameters, // such as 'this' and 'vtt', show up in the debug info. Preserve the calling // convention. CallingConv CC = CallingConv::CC_C; if (auto *FD = dyn_cast_or_null(D)) if (const auto *SrcFnTy = FD->getType()->getAs()) CC = SrcFnTy->getCallConv(); d759 8 a766 4 for (const VarDecl *VD : Args) ArgTypes.push_back(VD->getType()); QualType FnType = getContext().getFunctionType( RetTy, ArgTypes, FunctionProtoType::ExtProtoInfo(CC)); a772 4 // Since emitting the mcount call here impacts optimizations such as function // inlining, we just add an attribute to insert a mcount call in backend. // The attribute "counting-function" is set to mcount function name which is // architecture dependent. d774 1 a774 1 Fn->addFnAttr("counting-function", getTarget().getMCountName()); d826 4 a829 16 // If the lambda captures the object referred to by '*this' - either by // value or by reference, make sure CXXThisValue points to the correct // object. // Get the lvalue for the field (which is a copy of the enclosing object // or contains the address of the enclosing object). LValue ThisFieldLValue = EmitLValueForLambdaField(LambdaThisCaptureField); if (!LambdaThisCaptureField->getType()->isPointerType()) { // If the enclosing object was captured by value, just use its address. CXXThisValue = ThisFieldLValue.getAddress().getPointer(); } else { // Load the lvalue pointed to by the field, since '*this' was captured // by reference. CXXThisValue = EmitLoadOfLValue(ThisFieldLValue, SourceLocation()).getScalarVal(); } d886 1 a886 1 if (HaveInsertPoint() && CGM.getCodeGenOpts().hasProfileClangInstr()) { d907 1 a907 1 if (F->isInterposable()) return; d917 2 a918 2 QualType CodeGenFunction::BuildFunctionArgList(GlobalDecl GD, FunctionArgList &Args) { d920 6 d928 1 d938 11 a948 22 // The base version of an inheriting constructor whose constructed base is a // virtual base is not passed any arguments (because it doesn't actually call // the inherited constructor). bool PassedParams = true; if (const CXXConstructorDecl *CD = dyn_cast(FD)) if (auto Inherited = CD->getInheritedConstructor()) PassedParams = getTypes().inheritingCtorHasParams(Inherited, GD.getCtorType()); if (PassedParams) { for (auto *Param : FD->parameters()) { Args.push_back(Param); if (!Param->hasAttr()) continue; IdentifierInfo *NoID = nullptr; auto *Implicit = ImplicitParamDecl::Create( getContext(), Param->getDeclContext(), Param->getLocation(), NoID, getContext().getSizeType()); SizeArguments[Param] = Implicit; Args.push_back(Implicit); } a953 28 return ResTy; } static bool shouldUseUndefinedBehaviorReturnOptimization(const FunctionDecl *FD, const ASTContext &Context) { QualType T = FD->getReturnType(); // Avoid the optimization for functions that return a record type with a // trivial destructor or another trivially copyable type. if (const RecordType *RT = T.getCanonicalType()->getAs()) { if (const auto *ClassDecl = dyn_cast(RT->getDecl())) return !ClassDecl->hasTrivialDestructor(); } return !T.isTriviallyCopyableType(Context); } void CodeGenFunction::GenerateCode(GlobalDecl GD, llvm::Function *Fn, const CGFunctionInfo &FnInfo) { const FunctionDecl *FD = cast(GD.getDecl()); CurGD = GD; FunctionArgList Args; QualType ResTy = BuildFunctionArgList(GD, Args); // Check if we should generate debug info for this function. if (FD->hasAttr()) DebugInfo = nullptr; // disable debug info indefinitely for this function a970 7 Stmt *Body = FD->getBody(); // Initialize helper which will detect jumps which can cause invalid lifetime // markers. if (Body && ShouldEmitLifetimeMarkers) Bypasses.Init(Body); d1000 1 a1000 1 } else if (Body) { a1012 3 bool ShouldEmitUnreachable = CGM.getCodeGenOpts().StrictReturn || shouldUseUndefinedBehaviorReturnOptimization(FD, getContext()); d1017 4 a1020 9 SanitizerHandler::MissingReturn, EmitCheckSourceLocation(FD->getLocation()), None); } else if (ShouldEmitUnreachable) { if (CGM.getCodeGenOpts().OptimizationLevel == 0) EmitTrapCall(llvm::Intrinsic::trap); } if (SanOpts.has(SanitizerKind::Return) || ShouldEmitUnreachable) { Builder.CreateUnreachable(); Builder.ClearInsertionPoint(); d1022 2 a1090 22 bool CodeGenFunction::mightAddDeclToScope(const Stmt *S) { if (!S) return false; // Some statement kinds add a scope and thus never add a decl to the current // scope. Note, this list is longer than the list of statements that might // have an unscoped decl nested within them, but this way is conservatively // correct even if more statement kinds are added. if (isa(S) || isa(S) || isa(S) || isa(S) || isa(S) || isa(S) || isa(S) || isa(S) || isa(S) || isa(S)) return false; if (isa(S)) return true; for (const Stmt *SubStmt : S->children()) if (mightAddDeclToScope(SubStmt)) return true; return false; } d1096 1 a1096 2 bool &ResultBool, bool AllowLabels) { d1098 1 a1098 1 if (!ConstantFoldsToSimpleInteger(Cond, ResultInt, AllowLabels)) d1108 2 a1109 3 bool CodeGenFunction::ConstantFoldsToSimpleInteger(const Expr *Cond, llvm::APSInt &ResultInt, bool AllowLabels) { d1116 1 a1116 1 if (!AllowLabels && CodeGenFunction::ContainsLabel(Cond)) d1300 9 a1308 6 auto *Call = dyn_cast(Cond); if (Call && CGM.getCodeGenOpts().OptimizationLevel != 0) { auto *FD = dyn_cast_or_null(Call->getCalleeDecl()); if (FD && FD->getBuiltinID() == Builtin::BI__builtin_unpredictable) { llvm::MDBuilder MDHelper(getLLVMContext()); Unpredictable = MDHelper.createUnpredictable(); a1640 1 case Type::ObjCTypeParam: d1704 1 a1704 1 SanitizerHandler::VLABoundNotPositive, StaticArgs, Size); d1764 2 a1765 2 const APValue &Init) { assert(!Init.isUninit() && "Invalid DeclRefExpr initializer!"); d1767 1 a1767 1 if (CGM.getCodeGenOpts().getDebugInfo() >= codegenoptions::LimitedDebugInfo) d1863 2 a1864 1 void CGBuilderInserter::InsertHelper( d1867 2 a1868 1 llvm::IRBuilderDefaultInserter::InsertHelper(I, Name, BB, InsertPt); d1873 10 a1958 16 void CodeGenFunction::EmitSanitizerStatReport(llvm::SanitizerStatKind SSK) { if (!CGM.getCodeGenOpts().SanitizeStats) return; llvm::IRBuilder<> IRB(Builder.GetInsertBlock(), Builder.GetInsertPoint()); IRB.SetCurrentDebugLocation(Builder.getCurrentDebugLocation()); CGM.getSanStats().create(IRB, SSK); } llvm::DebugLoc CodeGenFunction::SourceLocToDebugLoc(SourceLocation Location) { if (CGDebugInfo *DI = getDebugInfo()) return DI->SourceLocToDebugLoc(Location); return llvm::DebugLoc(); } @ 1.1.1.10.2.1 log @Sync with HEAD @ text @d115 1 a115 1 if (getLangOpts().OpenMP && CurFn) d117 1 @ 1.1.1.11 log @Import Clang 4.0RC1 r294123. @ text @d115 1 a115 1 if (getLangOpts().OpenMP && CurFn) d117 1 @ 1.1.1.12 log @Import clang r309604 from branches/release_50 @ text @d48 4 a56 4 // Asan uses markers for use-after-scope checks. if (CGOpts.SanitizeAddressUseAfterScope) return true; d120 2 a121 2 LValueBaseInfo *BaseInfo) { return getNaturalTypeAlignment(T->getPointeeType(), BaseInfo, d126 1 a126 1 LValueBaseInfo *BaseInfo, d133 1 a133 2 if (BaseInfo) *BaseInfo = LValueBaseInfo(AlignmentSource::AttributedType, false); d138 1 a138 2 if (BaseInfo) *BaseInfo = LValueBaseInfo(AlignmentSource::Type, false); a151 2 if (T.getQualifiers().hasUnaligned()) Alignment = CharUnits::One(); d166 3 a168 3 LValueBaseInfo BaseInfo; CharUnits Alignment = getNaturalTypeAlignment(T, &BaseInfo); return LValue::MakeAddr(Address(V, Alignment), T, getContext(), BaseInfo, d176 3 a178 3 LValueBaseInfo BaseInfo; CharUnits Align = getNaturalTypeAlignment(T, &BaseInfo, /*pointee*/ true); return MakeAddrLValue(Address(V, Align), T, BaseInfo); d203 1 a203 2 case Type::DeducedTemplateSpecialization: llvm_unreachable("undeduced type in IR-generation"); d346 1 a346 1 DI->EmitFunctionEnd(Builder, CurFn); d610 5 d663 6 a668 6 QualType HintQTy = A->getTypeHint(); const ExtVectorType *HintEltQTy = HintQTy->getAs(); bool IsSignedInteger = HintQTy->isSignedIntegerType() || (HintEltQTy && HintEltQTy->getElementType()->isSignedIntegerType()); llvm::Metadata *AttrMDArgs[] = { d673 2 a674 2 llvm::APInt(32, (uint64_t)(IsSignedInteger ? 1 : 0))))}; Fn->setMetadata("vec_type_hint", llvm::MDNode::get(Context, AttrMDArgs)); d678 1 a678 1 llvm::Metadata *AttrMDArgs[] = { d682 1 a682 1 Fn->setMetadata("work_group_size_hint", llvm::MDNode::get(Context, AttrMDArgs)); d686 1 a686 1 llvm::Metadata *AttrMDArgs[] = { d690 1 a690 9 Fn->setMetadata("reqd_work_group_size", llvm::MDNode::get(Context, AttrMDArgs)); } if (const OpenCLIntelReqdSubGroupSizeAttr *A = FD->getAttr()) { llvm::Metadata *AttrMDArgs[] = { llvm::ConstantAsMetadata::get(Builder.getInt32(A->getSubGroupSize()))}; Fn->setMetadata("intel_reqd_sub_group_size", llvm::MDNode::get(Context, AttrMDArgs)); a709 5 static void markAsIgnoreThreadCheckingAtRuntime(llvm::Function *Fn) { Fn->addFnAttr("sanitize_thread_no_checking_at_run_time"); Fn->removeFnAttr(llvm::Attribute::SanitizeThread); } d753 1 a753 1 // .cxx_destruct, __destroy_helper_block_ and all of their calees at run time. d760 2 a761 1 markAsIgnoreThreadCheckingAtRuntime(Fn); a762 4 } else if (const auto *FD = dyn_cast_or_null(D)) { IdentifierInfo *II = FD->getIdentifier(); if (II && II->isStr("__destroy_helper_block_")) markAsIgnoreThreadCheckingAtRuntime(Fn); a772 4 if (const auto *LogArgs = D->getAttr()) { Fn->addFnAttr("xray-log-args", llvm::utostr(LogArgs->getArgumentCount())); } d774 3 a776 4 if (!CGM.imbueXRayAttrs(Fn, Loc)) Fn->addFnAttr( "xray-instruction-threshold", llvm::itostr(CGM.getCodeGenOpts().XRayInstructionThreshold)); a809 12 // If we're checking nullability, we need to know whether we can check the // return value. Initialize the flag to 'true' and refine it in EmitParmDecl. if (SanOpts.has(SanitizerKind::NullabilityReturn)) { auto Nullability = FnRetTy->getNullability(getContext()); if (Nullability && *Nullability == NullabilityKind::NonNull) { if (!(SanOpts.has(SanitizerKind::ReturnsNonnullAttribute) && CurCodeDecl && CurCodeDecl->getAttr())) RetValNullabilityPrecondition = llvm::ConstantInt::getTrue(getLLVMContext()); } } a829 7 // If we're checking the return value, allocate space for a pointer to a // precise source location of the checked return statement. if (requiresReturnValueCheck()) { ReturnLocation = CreateDefaultAlignTempAlloca(Int8PtrTy, "return.sloc.ptr"); InitTempAlloca(ReturnLocation, llvm::ConstantPointerNull::get(Int8PtrTy)); } d854 2 a855 8 if (CGM.getCodeGenOpts().InstrumentForProfiling) { if (CGM.getCodeGenOpts().CallFEntry) Fn->addFnAttr("fentry-call", "true"); else { if (!CurFuncDecl || !CurFuncDecl->hasAttr()) Fn->addFnAttr("counting-function", getTarget().getMCountName()); } } a937 10 // Check the 'this' pointer once per function, if it's available. if (CXXThisValue) { SanitizerSet SkippedChecks; SkippedChecks.set(SanitizerKind::ObjectSize, true); QualType ThisTy = MD->getThisType(getContext()); EmitTypeCheck(TCK_Load, Loc, CXXThisValue, ThisTy, getContext().getTypeAlignInChars(ThisTy->getPointeeType()), SkippedChecks); } d1039 1 d1041 2 a1042 2 getContext(), Param->getDeclContext(), Param->getLocation(), /*Id=*/nullptr, getContext().getSizeType(), ImplicitParamDecl::Other); a1078 2 // The function might not have a body if we're generating thunks for a // function declaration. d1080 1 a1080 4 if (Stmt *Body = FD->getBody()) BodyRange = Body->getSourceRange(); else BodyRange = FD->getLocation(); a1893 1 case Type::DeducedTemplateSpecialization: @ 1.1.1.13 log @Import clang r319952 from branches/release_50 @ text @a24 1 #include "clang/AST/ASTLambda.h" d986 1 a986 1 if (CXXABIThisValue) { d990 1 a990 12 // If this is the call operator of a lambda with no capture-default, it // may have a static invoker function, which may call this operator with // a null 'this' pointer. if (isLambdaCallOperator(MD) && cast(MD->getParent())->getLambdaCaptureDefault() == LCD_None) SkippedChecks.set(SanitizerKind::Null, true); EmitTypeCheck(isa(MD) ? TCK_ConstructorCall : TCK_MemberCall, Loc, CXXABIThisValue, ThisTy, @ 1.1.1.13.4.1 log @Sync with HEAD @ text @a35 1 #include "llvm/IR/Dominators.h" a38 1 #include "llvm/Transforms/Utils/PromoteMemToReg.h" d66 19 a84 3 SanOpts(CGM.getLangOpts().Sanitize), DebugInfo(CGM.getModuleDebugInfo()), PGO(cgm), ShouldEmitLifetimeMarkers(shouldEmitLifetimeMarkers( CGM.getCodeGenOpts(), CGM.getLangOpts())) { d90 1 a90 1 FMF.setFast(); a103 3 if (CGM.getCodeGenOpts().Reassociate) { FMF.setAllowReassoc(); } d121 3 a123 4 LValueBaseInfo *BaseInfo, TBAAAccessInfo *TBAAInfo) { return getNaturalTypeAlignment(T->getPointeeType(), BaseInfo, TBAAInfo, /* forPointeeType= */ true); a127 1 TBAAAccessInfo *TBAAInfo, a128 3 if (TBAAInfo) *TBAAInfo = CGM.getTBAAAccessInfo(T); d135 1 a135 1 *BaseInfo = LValueBaseInfo(AlignmentSource::AttributedType); d141 1 a141 1 *BaseInfo = LValueBaseInfo(AlignmentSource::Type); d172 1 a172 2 TBAAAccessInfo TBAAInfo; CharUnits Alignment = getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo); d174 1 a174 1 TBAAInfo); d182 2 a183 4 TBAAAccessInfo TBAAInfo; CharUnits Align = getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo, /* forPointeeType= */ true); return MakeAddrLValue(Address(V, Align), T, BaseInfo, TBAAInfo); d347 2 a348 7 if (ShouldInstrumentFunction()) { if (CGM.getCodeGenOpts().InstrumentFunctions) CurFn->addFnAttr("instrument-function-exit", "__cyg_profile_func_exit"); if (CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining) CurFn->addFnAttr("instrument-function-exit-inlined", "__cyg_profile_func_exit"); } a403 3 for (const auto &FuncletAndParent : TerminateFunclets) EmitIfUsed(*this, FuncletAndParent.second); a413 18 // Eliminate CleanupDestSlot alloca by replacing it with SSA values and // PHIs if the current function is a coroutine. We don't do it for all // functions as it may result in slight increase in numbers of instructions // if compiled with no optimizations. We do it for coroutine as the lifetime // of CleanupDestSlot alloca make correct coroutine frame building very // difficult. if (NormalCleanupDest.isValid() && isCoroutine()) { llvm::DominatorTree DT(*CurFn); llvm::PromoteMemToReg( cast(NormalCleanupDest.getPointer()), DT); NormalCleanupDest = Address::invalid(); } // Add the required-vector-width attribute. if (LargestVectorWidth != 0) CurFn->addFnAttr("min-legal-vector-width", llvm::utostr(LargestVectorWidth)); d419 1 a419 3 if (!CGM.getCodeGenOpts().InstrumentFunctions && !CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining && !CGM.getCodeGenOpts().InstrumentFunctionEntryBare) d432 23 a454 51 /// AlwaysEmitXRayCustomEvents - Return true if we should emit IR for calls to /// the __xray_customevent(...) builtin calls, when doing XRay instrumentation. bool CodeGenFunction::AlwaysEmitXRayCustomEvents() const { return CGM.getCodeGenOpts().XRayInstrumentFunctions && (CGM.getCodeGenOpts().XRayAlwaysEmitCustomEvents || CGM.getCodeGenOpts().XRayInstrumentationBundle.Mask == XRayInstrKind::Custom); } bool CodeGenFunction::AlwaysEmitXRayTypedEvents() const { return CGM.getCodeGenOpts().XRayInstrumentFunctions && (CGM.getCodeGenOpts().XRayAlwaysEmitTypedEvents || CGM.getCodeGenOpts().XRayInstrumentationBundle.Mask == XRayInstrKind::Typed); } llvm::Constant * CodeGenFunction::EncodeAddrForUseInPrologue(llvm::Function *F, llvm::Constant *Addr) { // Addresses stored in prologue data can't require run-time fixups and must // be PC-relative. Run-time fixups are undesirable because they necessitate // writable text segments, which are unsafe. And absolute addresses are // undesirable because they break PIE mode. // Add a layer of indirection through a private global. Taking its address // won't result in a run-time fixup, even if Addr has linkonce_odr linkage. auto *GV = new llvm::GlobalVariable(CGM.getModule(), Addr->getType(), /*isConstant=*/true, llvm::GlobalValue::PrivateLinkage, Addr); // Create a PC-relative address. auto *GOTAsInt = llvm::ConstantExpr::getPtrToInt(GV, IntPtrTy); auto *FuncAsInt = llvm::ConstantExpr::getPtrToInt(F, IntPtrTy); auto *PCRelAsInt = llvm::ConstantExpr::getSub(GOTAsInt, FuncAsInt); return (IntPtrTy == Int32Ty) ? PCRelAsInt : llvm::ConstantExpr::getTrunc(PCRelAsInt, Int32Ty); } llvm::Value * CodeGenFunction::DecodeAddrUsedInPrologue(llvm::Value *F, llvm::Value *EncodedAddr) { // Reconstruct the address of the global. auto *PCRelAsInt = Builder.CreateSExt(EncodedAddr, IntPtrTy); auto *FuncAsInt = Builder.CreatePtrToInt(F, IntPtrTy, "func_addr.int"); auto *GOTAsInt = Builder.CreateAdd(PCRelAsInt, FuncAsInt, "global_addr.int"); auto *GOTAddr = Builder.CreateIntToPtr(GOTAsInt, Int8PtrPtrTy, "global_addr"); // Load the original pointer through the global. return Builder.CreateLoad(Address(GOTAddr, getPointerAlign()), "decoded_addr"); d483 2 a484 2 static unsigned ArgInfoAddressSpace(LangAS AS) { switch (AS) { d624 1 a624 4 const Decl *PDecl = parm; if (auto *TD = dyn_cast(ty)) PDecl = TD->getDecl(); const OpenCLAccessAttr *A = PDecl->getAttr(); a723 29 static bool matchesStlAllocatorFn(const Decl *D, const ASTContext &Ctx) { auto *MD = dyn_cast_or_null(D); if (!MD || !MD->getDeclName().getAsIdentifierInfo() || !MD->getDeclName().getAsIdentifierInfo()->isStr("allocate") || (MD->getNumParams() != 1 && MD->getNumParams() != 2)) return false; if (MD->parameters()[0]->getType().getCanonicalType() != Ctx.getSizeType()) return false; if (MD->getNumParams() == 2) { auto *PT = MD->parameters()[1]->getType()->getAs(); if (!PT || !PT->isVoidPointerType() || !PT->getPointeeType().isConstQualified()) return false; } return true; } /// Return the UBSan prologue signature for \p FD if one is available. static llvm::Constant *getPrologueSignature(CodeGenModule &CGM, const FunctionDecl *FD) { if (const auto *MD = dyn_cast(FD)) if (!MD->isStatic()) return nullptr; return CGM.getTargetCodeGenInfo().getUBSanFunctionSignature(CGM); } d747 2 a748 13 // If this function has been blacklisted for any of the enabled sanitizers, // disable the sanitizer for the function. do { #define SANITIZER(NAME, ID) \ if (SanOpts.empty()) \ break; \ if (SanOpts.has(SanitizerKind::ID)) \ if (CGM.isInSanitizerBlacklist(SanitizerKind::ID, Fn, Loc)) \ SanOpts.set(SanitizerKind::ID, false); #include "clang/Basic/Sanitizers.def" #undef SANITIZER } while (0); d752 2 a753 12 for (auto Attr : D->specific_attrs()) { SanitizerMask mask = Attr->getMask(); SanOpts.Mask &= ~mask; if (mask & SanitizerKind::Address) SanOpts.set(SanitizerKind::KernelAddress, false); if (mask & SanitizerKind::KernelAddress) SanOpts.set(SanitizerKind::Address, false); if (mask & SanitizerKind::HWAddress) SanOpts.set(SanitizerKind::KernelHWAddress, false); if (mask & SanitizerKind::KernelHWAddress) SanOpts.set(SanitizerKind::HWAddress, false); } a758 2 if (SanOpts.hasOneOf(SanitizerKind::HWAddress | SanitizerKind::KernelHWAddress)) Fn->addFnAttr(llvm::Attribute::SanitizeHWAddress); a764 6 if (SanOpts.has(SanitizerKind::ShadowCallStack)) Fn->addFnAttr(llvm::Attribute::ShadowCallStack); // Apply fuzzing attribute to the function. if (SanOpts.hasOneOf(SanitizerKind::Fuzzer | SanitizerKind::FuzzerNoLink)) Fn->addFnAttr(llvm::Attribute::OptForFuzzing); a782 8 // Ignore unrelated casts in STL allocate() since the allocator must cast // from void* to T* before object initialization completes. Don't match on the // namespace because not all allocators are in std:: if (D && SanOpts.has(SanitizerKind::CFIUnrelatedCast)) { if (matchesStlAllocatorFn(D, getContext())) SanOpts.Mask &= ~SanitizerKind::CFIUnrelatedCast; } d784 1 a784 4 bool InstrumentXray = ShouldXRayInstrumentFunction() && CGM.getCodeGenOpts().XRayInstrumentationBundle.has( XRayInstrKind::Function); if (D && InstrumentXray) { d802 4 a809 4 // Add profile-sample-accurate value. if (CGM.getCodeGenOpts().ProfileSampleAccurate) Fn->addFnAttr("profile-sample-accurate"); d820 2 a821 6 if (llvm::Constant *PrologueSig = getPrologueSignature(CGM, FD)) { // Remove any (C++17) exception specifications, to allow calling e.g. a // noexcept function through a non-noexcept pointer. auto ProtoTy = getContext().getFunctionTypeWithExceptionSpec(FD->getType(), EST_None); d823 2 a824 5 CGM.GetAddrOfRTTIDescriptor(ProtoTy, /*ForEH=*/true); llvm::Constant *FTRTTIConstEncoded = EncodeAddrForUseInPrologue(Fn, FTRTTIConst); llvm::Constant *PrologueStructElems[] = {PrologueSig, FTRTTIConstEncoded}; d885 1 a885 2 DI->EmitFunctionStart(GD, Loc, StartLoc, FnType, CurFn, CurFuncIsThunk, Builder); d888 2 a889 10 if (ShouldInstrumentFunction()) { if (CGM.getCodeGenOpts().InstrumentFunctions) CurFn->addFnAttr("instrument-function-entry", "__cyg_profile_func_enter"); if (CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining) CurFn->addFnAttr("instrument-function-entry-inlined", "__cyg_profile_func_enter"); if (CGM.getCodeGenOpts().InstrumentFunctionEntryBare) CurFn->addFnAttr("instrument-function-entry-inlined", "__cyg_profile_func_enter_bare"); } d896 5 a900 9 // Calls to fentry/mcount should not be generated if function has // the no_instrument_function attribute. if (!CurFuncDecl || !CurFuncDecl->hasAttr()) { if (CGM.getCodeGenOpts().CallFEntry) Fn->addFnAttr("fentry-call", "true"); else { Fn->addFnAttr("instrument-function-entry-inlined", getTarget().getMCountName()); } a942 5 // Emit OpenMP specific initialization of the device functions. if (getLangOpts().OpenMP && CurCodeDecl) CGM.getOpenMPRuntime().emitFunctionProlog(*this, CurCodeDecl); d996 2 a997 1 MD->getParent()->getLambdaCaptureDefault() == LCD_None) a1028 6 // TODO: Do we need to handle this in two places like we do with // target-features/target-cpu? if (CurFuncDecl) if (const auto *VecWidth = CurFuncDecl->getAttr()) LargestVectorWidth = VecWidth->getVectorWidth(); d1188 7 a1194 2 else if (isa(FD) && cast(FD)->isLambdaStaticInvoker()) { d1197 1 a1197 1 EmitLambdaStaticInvokeBody(cast(FD)); d1641 6 a1646 3 auto VlaSize = getVLASize(vlaType); SizeVal = VlaSize.NumElts; CharUnits eltSize = getContext().getTypeSizeInChars(VlaSize.Type); d1729 1 a1729 1 numVLAElements = getVLASize(cast(arrayType)).NumElts; d1810 2 a1811 1 CodeGenFunction::VlaSizePair CodeGenFunction::getVLASize(QualType type) { d1817 1 a1817 1 CodeGenFunction::VlaSizePair d1838 1 a1838 16 return { numElements, elementType }; } CodeGenFunction::VlaSizePair CodeGenFunction::getVLAElements1D(QualType type) { const VariableArrayType *vla = getContext().getAsVariableArrayType(type); assert(vla && "type was not a variable array type!"); return getVLAElements1D(vla); } CodeGenFunction::VlaSizePair CodeGenFunction::getVLAElements1D(const VariableArrayType *Vla) { llvm::Value *VlaSize = VLASizeMap[Vla->getSizeExpr()]; assert(VlaSize && "no size for VLA!"); assert(VlaSize->getType() == SizeTy); return { VlaSize, Vla->getElementType() }; d2121 1 a2121 1 Feature.split(OrFeatures, '|'); d2159 1 a2159 1 StringRef(FeatureList).split(ReqFeatures, ','); a2166 4 const TargetAttr *TD = TargetDecl->getAttr(); TargetAttr::ParsedTargetAttr ParsedAttr = CGM.filterFunctionTargetAttrs(TD); a2169 6 for (const auto &F : ParsedAttr.Features) { if (F[0] == '+' && CalleeFeatureMap.lookup(F.substr(1))) ReqFeatures.push_back(StringRef(F).substr(1)); } a2189 54 llvm::Value * CodeGenFunction::FormResolverCondition(const MultiVersionResolverOption &RO) { llvm::Value *TrueCondition = nullptr; if (!RO.ParsedAttribute.Architecture.empty()) TrueCondition = EmitX86CpuIs(RO.ParsedAttribute.Architecture); if (!RO.ParsedAttribute.Features.empty()) { SmallVector FeatureList; llvm::for_each(RO.ParsedAttribute.Features, [&FeatureList](const std::string &Feature) { FeatureList.push_back(StringRef{Feature}.substr(1)); }); llvm::Value *FeatureCmp = EmitX86CpuSupports(FeatureList); TrueCondition = TrueCondition ? Builder.CreateAnd(TrueCondition, FeatureCmp) : FeatureCmp; } return TrueCondition; } void CodeGenFunction::EmitMultiVersionResolver( llvm::Function *Resolver, ArrayRef Options) { assert((getContext().getTargetInfo().getTriple().getArch() == llvm::Triple::x86 || getContext().getTargetInfo().getTriple().getArch() == llvm::Triple::x86_64) && "Only implemented for x86 targets"); // Main function's basic block. llvm::BasicBlock *CurBlock = createBasicBlock("entry", Resolver); Builder.SetInsertPoint(CurBlock); EmitX86CpuInit(); llvm::Function *DefaultFunc = nullptr; for (const MultiVersionResolverOption &RO : Options) { Builder.SetInsertPoint(CurBlock); llvm::Value *TrueCondition = FormResolverCondition(RO); if (!TrueCondition) { DefaultFunc = RO.Function; } else { llvm::BasicBlock *RetBlock = createBasicBlock("ro_ret", Resolver); llvm::IRBuilder<> RetBuilder(RetBlock); RetBuilder.CreateRet(RO.Function); CurBlock = createBasicBlock("ro_else", Resolver); Builder.CreateCondBr(TrueCondition, RetBlock, CurBlock); } } assert(DefaultFunc && "No default version?"); // Emit return from the 'else-ist' block. Builder.SetInsertPoint(CurBlock); Builder.CreateRet(DefaultFunc); } @ 1.1.1.13.4.2 log @Mostly merge changes from HEAD upto 20200411 @ text @@ 1.1.1.13.2.1 log @Sync with HEAD @ text @a35 1 #include "llvm/IR/Dominators.h" a38 1 #include "llvm/Transforms/Utils/PromoteMemToReg.h" d66 19 a84 3 SanOpts(CGM.getLangOpts().Sanitize), DebugInfo(CGM.getModuleDebugInfo()), PGO(cgm), ShouldEmitLifetimeMarkers(shouldEmitLifetimeMarkers( CGM.getCodeGenOpts(), CGM.getLangOpts())) { d90 1 a90 1 FMF.setFast(); a103 3 if (CGM.getCodeGenOpts().Reassociate) { FMF.setAllowReassoc(); } d121 3 a123 4 LValueBaseInfo *BaseInfo, TBAAAccessInfo *TBAAInfo) { return getNaturalTypeAlignment(T->getPointeeType(), BaseInfo, TBAAInfo, /* forPointeeType= */ true); a127 1 TBAAAccessInfo *TBAAInfo, a128 3 if (TBAAInfo) *TBAAInfo = CGM.getTBAAAccessInfo(T); d135 1 a135 1 *BaseInfo = LValueBaseInfo(AlignmentSource::AttributedType); d141 1 a141 1 *BaseInfo = LValueBaseInfo(AlignmentSource::Type); d172 1 a172 2 TBAAAccessInfo TBAAInfo; CharUnits Alignment = getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo); d174 1 a174 1 TBAAInfo); d182 2 a183 4 TBAAAccessInfo TBAAInfo; CharUnits Align = getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo, /* forPointeeType= */ true); return MakeAddrLValue(Address(V, Align), T, BaseInfo, TBAAInfo); d347 2 a348 7 if (ShouldInstrumentFunction()) { if (CGM.getCodeGenOpts().InstrumentFunctions) CurFn->addFnAttr("instrument-function-exit", "__cyg_profile_func_exit"); if (CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining) CurFn->addFnAttr("instrument-function-exit-inlined", "__cyg_profile_func_exit"); } a403 3 for (const auto &FuncletAndParent : TerminateFunclets) EmitIfUsed(*this, FuncletAndParent.second); a413 18 // Eliminate CleanupDestSlot alloca by replacing it with SSA values and // PHIs if the current function is a coroutine. We don't do it for all // functions as it may result in slight increase in numbers of instructions // if compiled with no optimizations. We do it for coroutine as the lifetime // of CleanupDestSlot alloca make correct coroutine frame building very // difficult. if (NormalCleanupDest.isValid() && isCoroutine()) { llvm::DominatorTree DT(*CurFn); llvm::PromoteMemToReg( cast(NormalCleanupDest.getPointer()), DT); NormalCleanupDest = Address::invalid(); } // Add the required-vector-width attribute. if (LargestVectorWidth != 0) CurFn->addFnAttr("min-legal-vector-width", llvm::utostr(LargestVectorWidth)); d419 1 a419 3 if (!CGM.getCodeGenOpts().InstrumentFunctions && !CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining && !CGM.getCodeGenOpts().InstrumentFunctionEntryBare) d432 23 a454 51 /// AlwaysEmitXRayCustomEvents - Return true if we should emit IR for calls to /// the __xray_customevent(...) builtin calls, when doing XRay instrumentation. bool CodeGenFunction::AlwaysEmitXRayCustomEvents() const { return CGM.getCodeGenOpts().XRayInstrumentFunctions && (CGM.getCodeGenOpts().XRayAlwaysEmitCustomEvents || CGM.getCodeGenOpts().XRayInstrumentationBundle.Mask == XRayInstrKind::Custom); } bool CodeGenFunction::AlwaysEmitXRayTypedEvents() const { return CGM.getCodeGenOpts().XRayInstrumentFunctions && (CGM.getCodeGenOpts().XRayAlwaysEmitTypedEvents || CGM.getCodeGenOpts().XRayInstrumentationBundle.Mask == XRayInstrKind::Typed); } llvm::Constant * CodeGenFunction::EncodeAddrForUseInPrologue(llvm::Function *F, llvm::Constant *Addr) { // Addresses stored in prologue data can't require run-time fixups and must // be PC-relative. Run-time fixups are undesirable because they necessitate // writable text segments, which are unsafe. And absolute addresses are // undesirable because they break PIE mode. // Add a layer of indirection through a private global. Taking its address // won't result in a run-time fixup, even if Addr has linkonce_odr linkage. auto *GV = new llvm::GlobalVariable(CGM.getModule(), Addr->getType(), /*isConstant=*/true, llvm::GlobalValue::PrivateLinkage, Addr); // Create a PC-relative address. auto *GOTAsInt = llvm::ConstantExpr::getPtrToInt(GV, IntPtrTy); auto *FuncAsInt = llvm::ConstantExpr::getPtrToInt(F, IntPtrTy); auto *PCRelAsInt = llvm::ConstantExpr::getSub(GOTAsInt, FuncAsInt); return (IntPtrTy == Int32Ty) ? PCRelAsInt : llvm::ConstantExpr::getTrunc(PCRelAsInt, Int32Ty); } llvm::Value * CodeGenFunction::DecodeAddrUsedInPrologue(llvm::Value *F, llvm::Value *EncodedAddr) { // Reconstruct the address of the global. auto *PCRelAsInt = Builder.CreateSExt(EncodedAddr, IntPtrTy); auto *FuncAsInt = Builder.CreatePtrToInt(F, IntPtrTy, "func_addr.int"); auto *GOTAsInt = Builder.CreateAdd(PCRelAsInt, FuncAsInt, "global_addr.int"); auto *GOTAddr = Builder.CreateIntToPtr(GOTAsInt, Int8PtrPtrTy, "global_addr"); // Load the original pointer through the global. return Builder.CreateLoad(Address(GOTAddr, getPointerAlign()), "decoded_addr"); d483 2 a484 2 static unsigned ArgInfoAddressSpace(LangAS AS) { switch (AS) { d624 1 a624 4 const Decl *PDecl = parm; if (auto *TD = dyn_cast(ty)) PDecl = TD->getDecl(); const OpenCLAccessAttr *A = PDecl->getAttr(); a723 29 static bool matchesStlAllocatorFn(const Decl *D, const ASTContext &Ctx) { auto *MD = dyn_cast_or_null(D); if (!MD || !MD->getDeclName().getAsIdentifierInfo() || !MD->getDeclName().getAsIdentifierInfo()->isStr("allocate") || (MD->getNumParams() != 1 && MD->getNumParams() != 2)) return false; if (MD->parameters()[0]->getType().getCanonicalType() != Ctx.getSizeType()) return false; if (MD->getNumParams() == 2) { auto *PT = MD->parameters()[1]->getType()->getAs(); if (!PT || !PT->isVoidPointerType() || !PT->getPointeeType().isConstQualified()) return false; } return true; } /// Return the UBSan prologue signature for \p FD if one is available. static llvm::Constant *getPrologueSignature(CodeGenModule &CGM, const FunctionDecl *FD) { if (const auto *MD = dyn_cast(FD)) if (!MD->isStatic()) return nullptr; return CGM.getTargetCodeGenInfo().getUBSanFunctionSignature(CGM); } d747 2 a748 13 // If this function has been blacklisted for any of the enabled sanitizers, // disable the sanitizer for the function. do { #define SANITIZER(NAME, ID) \ if (SanOpts.empty()) \ break; \ if (SanOpts.has(SanitizerKind::ID)) \ if (CGM.isInSanitizerBlacklist(SanitizerKind::ID, Fn, Loc)) \ SanOpts.set(SanitizerKind::ID, false); #include "clang/Basic/Sanitizers.def" #undef SANITIZER } while (0); d752 2 a753 12 for (auto Attr : D->specific_attrs()) { SanitizerMask mask = Attr->getMask(); SanOpts.Mask &= ~mask; if (mask & SanitizerKind::Address) SanOpts.set(SanitizerKind::KernelAddress, false); if (mask & SanitizerKind::KernelAddress) SanOpts.set(SanitizerKind::Address, false); if (mask & SanitizerKind::HWAddress) SanOpts.set(SanitizerKind::KernelHWAddress, false); if (mask & SanitizerKind::KernelHWAddress) SanOpts.set(SanitizerKind::HWAddress, false); } a758 2 if (SanOpts.hasOneOf(SanitizerKind::HWAddress | SanitizerKind::KernelHWAddress)) Fn->addFnAttr(llvm::Attribute::SanitizeHWAddress); a764 6 if (SanOpts.has(SanitizerKind::ShadowCallStack)) Fn->addFnAttr(llvm::Attribute::ShadowCallStack); // Apply fuzzing attribute to the function. if (SanOpts.hasOneOf(SanitizerKind::Fuzzer | SanitizerKind::FuzzerNoLink)) Fn->addFnAttr(llvm::Attribute::OptForFuzzing); a782 8 // Ignore unrelated casts in STL allocate() since the allocator must cast // from void* to T* before object initialization completes. Don't match on the // namespace because not all allocators are in std:: if (D && SanOpts.has(SanitizerKind::CFIUnrelatedCast)) { if (matchesStlAllocatorFn(D, getContext())) SanOpts.Mask &= ~SanitizerKind::CFIUnrelatedCast; } d784 1 a784 4 bool InstrumentXray = ShouldXRayInstrumentFunction() && CGM.getCodeGenOpts().XRayInstrumentationBundle.has( XRayInstrKind::Function); if (D && InstrumentXray) { d802 4 a809 4 // Add profile-sample-accurate value. if (CGM.getCodeGenOpts().ProfileSampleAccurate) Fn->addFnAttr("profile-sample-accurate"); d820 2 a821 6 if (llvm::Constant *PrologueSig = getPrologueSignature(CGM, FD)) { // Remove any (C++17) exception specifications, to allow calling e.g. a // noexcept function through a non-noexcept pointer. auto ProtoTy = getContext().getFunctionTypeWithExceptionSpec(FD->getType(), EST_None); d823 2 a824 5 CGM.GetAddrOfRTTIDescriptor(ProtoTy, /*ForEH=*/true); llvm::Constant *FTRTTIConstEncoded = EncodeAddrForUseInPrologue(Fn, FTRTTIConst); llvm::Constant *PrologueStructElems[] = {PrologueSig, FTRTTIConstEncoded}; d885 1 a885 2 DI->EmitFunctionStart(GD, Loc, StartLoc, FnType, CurFn, CurFuncIsThunk, Builder); d888 2 a889 10 if (ShouldInstrumentFunction()) { if (CGM.getCodeGenOpts().InstrumentFunctions) CurFn->addFnAttr("instrument-function-entry", "__cyg_profile_func_enter"); if (CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining) CurFn->addFnAttr("instrument-function-entry-inlined", "__cyg_profile_func_enter"); if (CGM.getCodeGenOpts().InstrumentFunctionEntryBare) CurFn->addFnAttr("instrument-function-entry-inlined", "__cyg_profile_func_enter_bare"); } d896 5 a900 9 // Calls to fentry/mcount should not be generated if function has // the no_instrument_function attribute. if (!CurFuncDecl || !CurFuncDecl->hasAttr()) { if (CGM.getCodeGenOpts().CallFEntry) Fn->addFnAttr("fentry-call", "true"); else { Fn->addFnAttr("instrument-function-entry-inlined", getTarget().getMCountName()); } a942 5 // Emit OpenMP specific initialization of the device functions. if (getLangOpts().OpenMP && CurCodeDecl) CGM.getOpenMPRuntime().emitFunctionProlog(*this, CurCodeDecl); d996 2 a997 1 MD->getParent()->getLambdaCaptureDefault() == LCD_None) a1028 6 // TODO: Do we need to handle this in two places like we do with // target-features/target-cpu? if (CurFuncDecl) if (const auto *VecWidth = CurFuncDecl->getAttr()) LargestVectorWidth = VecWidth->getVectorWidth(); d1188 7 a1194 2 else if (isa(FD) && cast(FD)->isLambdaStaticInvoker()) { d1197 1 a1197 1 EmitLambdaStaticInvokeBody(cast(FD)); d1641 6 a1646 3 auto VlaSize = getVLASize(vlaType); SizeVal = VlaSize.NumElts; CharUnits eltSize = getContext().getTypeSizeInChars(VlaSize.Type); d1729 1 a1729 1 numVLAElements = getVLASize(cast(arrayType)).NumElts; d1810 2 a1811 1 CodeGenFunction::VlaSizePair CodeGenFunction::getVLASize(QualType type) { d1817 1 a1817 1 CodeGenFunction::VlaSizePair d1838 1 a1838 16 return { numElements, elementType }; } CodeGenFunction::VlaSizePair CodeGenFunction::getVLAElements1D(QualType type) { const VariableArrayType *vla = getContext().getAsVariableArrayType(type); assert(vla && "type was not a variable array type!"); return getVLAElements1D(vla); } CodeGenFunction::VlaSizePair CodeGenFunction::getVLAElements1D(const VariableArrayType *Vla) { llvm::Value *VlaSize = VLASizeMap[Vla->getSizeExpr()]; assert(VlaSize && "no size for VLA!"); assert(VlaSize->getType() == SizeTy); return { VlaSize, Vla->getElementType() }; d2121 1 a2121 1 Feature.split(OrFeatures, '|'); d2159 1 a2159 1 StringRef(FeatureList).split(ReqFeatures, ','); a2166 4 const TargetAttr *TD = TargetDecl->getAttr(); TargetAttr::ParsedTargetAttr ParsedAttr = CGM.filterFunctionTargetAttrs(TD); a2169 6 for (const auto &F : ParsedAttr.Features) { if (F[0] == '+' && CalleeFeatureMap.lookup(F.substr(1))) ReqFeatures.push_back(StringRef(F).substr(1)); } a2189 54 llvm::Value * CodeGenFunction::FormResolverCondition(const MultiVersionResolverOption &RO) { llvm::Value *TrueCondition = nullptr; if (!RO.ParsedAttribute.Architecture.empty()) TrueCondition = EmitX86CpuIs(RO.ParsedAttribute.Architecture); if (!RO.ParsedAttribute.Features.empty()) { SmallVector FeatureList; llvm::for_each(RO.ParsedAttribute.Features, [&FeatureList](const std::string &Feature) { FeatureList.push_back(StringRef{Feature}.substr(1)); }); llvm::Value *FeatureCmp = EmitX86CpuSupports(FeatureList); TrueCondition = TrueCondition ? Builder.CreateAnd(TrueCondition, FeatureCmp) : FeatureCmp; } return TrueCondition; } void CodeGenFunction::EmitMultiVersionResolver( llvm::Function *Resolver, ArrayRef Options) { assert((getContext().getTargetInfo().getTriple().getArch() == llvm::Triple::x86 || getContext().getTargetInfo().getTriple().getArch() == llvm::Triple::x86_64) && "Only implemented for x86 targets"); // Main function's basic block. llvm::BasicBlock *CurBlock = createBasicBlock("entry", Resolver); Builder.SetInsertPoint(CurBlock); EmitX86CpuInit(); llvm::Function *DefaultFunc = nullptr; for (const MultiVersionResolverOption &RO : Options) { Builder.SetInsertPoint(CurBlock); llvm::Value *TrueCondition = FormResolverCondition(RO); if (!TrueCondition) { DefaultFunc = RO.Function; } else { llvm::BasicBlock *RetBlock = createBasicBlock("ro_ret", Resolver); llvm::IRBuilder<> RetBuilder(RetBlock); RetBuilder.CreateRet(RO.Function); CurBlock = createBasicBlock("ro_else", Resolver); Builder.CreateCondBr(TrueCondition, RetBlock, CurBlock); } } assert(DefaultFunc && "No default version?"); // Emit return from the 'else-ist' block. Builder.SetInsertPoint(CurBlock); Builder.CreateRet(DefaultFunc); } @ 1.1.1.14 log @Import clang r337282 from trunk @ text @a35 1 #include "llvm/IR/Dominators.h" a38 1 #include "llvm/Transforms/Utils/PromoteMemToReg.h" d66 19 a84 3 SanOpts(CGM.getLangOpts().Sanitize), DebugInfo(CGM.getModuleDebugInfo()), PGO(cgm), ShouldEmitLifetimeMarkers(shouldEmitLifetimeMarkers( CGM.getCodeGenOpts(), CGM.getLangOpts())) { d90 1 a90 1 FMF.setFast(); a103 3 if (CGM.getCodeGenOpts().Reassociate) { FMF.setAllowReassoc(); } d121 3 a123 4 LValueBaseInfo *BaseInfo, TBAAAccessInfo *TBAAInfo) { return getNaturalTypeAlignment(T->getPointeeType(), BaseInfo, TBAAInfo, /* forPointeeType= */ true); a127 1 TBAAAccessInfo *TBAAInfo, a128 3 if (TBAAInfo) *TBAAInfo = CGM.getTBAAAccessInfo(T); d135 1 a135 1 *BaseInfo = LValueBaseInfo(AlignmentSource::AttributedType); d141 1 a141 1 *BaseInfo = LValueBaseInfo(AlignmentSource::Type); d172 1 a172 2 TBAAAccessInfo TBAAInfo; CharUnits Alignment = getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo); d174 1 a174 1 TBAAInfo); d182 2 a183 4 TBAAAccessInfo TBAAInfo; CharUnits Align = getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo, /* forPointeeType= */ true); return MakeAddrLValue(Address(V, Align), T, BaseInfo, TBAAInfo); d347 2 a348 7 if (ShouldInstrumentFunction()) { if (CGM.getCodeGenOpts().InstrumentFunctions) CurFn->addFnAttr("instrument-function-exit", "__cyg_profile_func_exit"); if (CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining) CurFn->addFnAttr("instrument-function-exit-inlined", "__cyg_profile_func_exit"); } a403 3 for (const auto &FuncletAndParent : TerminateFunclets) EmitIfUsed(*this, FuncletAndParent.second); a413 18 // Eliminate CleanupDestSlot alloca by replacing it with SSA values and // PHIs if the current function is a coroutine. We don't do it for all // functions as it may result in slight increase in numbers of instructions // if compiled with no optimizations. We do it for coroutine as the lifetime // of CleanupDestSlot alloca make correct coroutine frame building very // difficult. if (NormalCleanupDest.isValid() && isCoroutine()) { llvm::DominatorTree DT(*CurFn); llvm::PromoteMemToReg( cast(NormalCleanupDest.getPointer()), DT); NormalCleanupDest = Address::invalid(); } // Add the required-vector-width attribute. if (LargestVectorWidth != 0) CurFn->addFnAttr("min-legal-vector-width", llvm::utostr(LargestVectorWidth)); d419 1 a419 3 if (!CGM.getCodeGenOpts().InstrumentFunctions && !CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining && !CGM.getCodeGenOpts().InstrumentFunctionEntryBare) d432 23 a454 51 /// AlwaysEmitXRayCustomEvents - Return true if we should emit IR for calls to /// the __xray_customevent(...) builtin calls, when doing XRay instrumentation. bool CodeGenFunction::AlwaysEmitXRayCustomEvents() const { return CGM.getCodeGenOpts().XRayInstrumentFunctions && (CGM.getCodeGenOpts().XRayAlwaysEmitCustomEvents || CGM.getCodeGenOpts().XRayInstrumentationBundle.Mask == XRayInstrKind::Custom); } bool CodeGenFunction::AlwaysEmitXRayTypedEvents() const { return CGM.getCodeGenOpts().XRayInstrumentFunctions && (CGM.getCodeGenOpts().XRayAlwaysEmitTypedEvents || CGM.getCodeGenOpts().XRayInstrumentationBundle.Mask == XRayInstrKind::Typed); } llvm::Constant * CodeGenFunction::EncodeAddrForUseInPrologue(llvm::Function *F, llvm::Constant *Addr) { // Addresses stored in prologue data can't require run-time fixups and must // be PC-relative. Run-time fixups are undesirable because they necessitate // writable text segments, which are unsafe. And absolute addresses are // undesirable because they break PIE mode. // Add a layer of indirection through a private global. Taking its address // won't result in a run-time fixup, even if Addr has linkonce_odr linkage. auto *GV = new llvm::GlobalVariable(CGM.getModule(), Addr->getType(), /*isConstant=*/true, llvm::GlobalValue::PrivateLinkage, Addr); // Create a PC-relative address. auto *GOTAsInt = llvm::ConstantExpr::getPtrToInt(GV, IntPtrTy); auto *FuncAsInt = llvm::ConstantExpr::getPtrToInt(F, IntPtrTy); auto *PCRelAsInt = llvm::ConstantExpr::getSub(GOTAsInt, FuncAsInt); return (IntPtrTy == Int32Ty) ? PCRelAsInt : llvm::ConstantExpr::getTrunc(PCRelAsInt, Int32Ty); } llvm::Value * CodeGenFunction::DecodeAddrUsedInPrologue(llvm::Value *F, llvm::Value *EncodedAddr) { // Reconstruct the address of the global. auto *PCRelAsInt = Builder.CreateSExt(EncodedAddr, IntPtrTy); auto *FuncAsInt = Builder.CreatePtrToInt(F, IntPtrTy, "func_addr.int"); auto *GOTAsInt = Builder.CreateAdd(PCRelAsInt, FuncAsInt, "global_addr.int"); auto *GOTAddr = Builder.CreateIntToPtr(GOTAsInt, Int8PtrPtrTy, "global_addr"); // Load the original pointer through the global. return Builder.CreateLoad(Address(GOTAddr, getPointerAlign()), "decoded_addr"); d483 2 a484 2 static unsigned ArgInfoAddressSpace(LangAS AS) { switch (AS) { d624 1 a624 4 const Decl *PDecl = parm; if (auto *TD = dyn_cast(ty)) PDecl = TD->getDecl(); const OpenCLAccessAttr *A = PDecl->getAttr(); a723 29 static bool matchesStlAllocatorFn(const Decl *D, const ASTContext &Ctx) { auto *MD = dyn_cast_or_null(D); if (!MD || !MD->getDeclName().getAsIdentifierInfo() || !MD->getDeclName().getAsIdentifierInfo()->isStr("allocate") || (MD->getNumParams() != 1 && MD->getNumParams() != 2)) return false; if (MD->parameters()[0]->getType().getCanonicalType() != Ctx.getSizeType()) return false; if (MD->getNumParams() == 2) { auto *PT = MD->parameters()[1]->getType()->getAs(); if (!PT || !PT->isVoidPointerType() || !PT->getPointeeType().isConstQualified()) return false; } return true; } /// Return the UBSan prologue signature for \p FD if one is available. static llvm::Constant *getPrologueSignature(CodeGenModule &CGM, const FunctionDecl *FD) { if (const auto *MD = dyn_cast(FD)) if (!MD->isStatic()) return nullptr; return CGM.getTargetCodeGenInfo().getUBSanFunctionSignature(CGM); } d747 2 a748 13 // If this function has been blacklisted for any of the enabled sanitizers, // disable the sanitizer for the function. do { #define SANITIZER(NAME, ID) \ if (SanOpts.empty()) \ break; \ if (SanOpts.has(SanitizerKind::ID)) \ if (CGM.isInSanitizerBlacklist(SanitizerKind::ID, Fn, Loc)) \ SanOpts.set(SanitizerKind::ID, false); #include "clang/Basic/Sanitizers.def" #undef SANITIZER } while (0); d752 2 a753 12 for (auto Attr : D->specific_attrs()) { SanitizerMask mask = Attr->getMask(); SanOpts.Mask &= ~mask; if (mask & SanitizerKind::Address) SanOpts.set(SanitizerKind::KernelAddress, false); if (mask & SanitizerKind::KernelAddress) SanOpts.set(SanitizerKind::Address, false); if (mask & SanitizerKind::HWAddress) SanOpts.set(SanitizerKind::KernelHWAddress, false); if (mask & SanitizerKind::KernelHWAddress) SanOpts.set(SanitizerKind::HWAddress, false); } a758 2 if (SanOpts.hasOneOf(SanitizerKind::HWAddress | SanitizerKind::KernelHWAddress)) Fn->addFnAttr(llvm::Attribute::SanitizeHWAddress); a764 6 if (SanOpts.has(SanitizerKind::ShadowCallStack)) Fn->addFnAttr(llvm::Attribute::ShadowCallStack); // Apply fuzzing attribute to the function. if (SanOpts.hasOneOf(SanitizerKind::Fuzzer | SanitizerKind::FuzzerNoLink)) Fn->addFnAttr(llvm::Attribute::OptForFuzzing); a782 8 // Ignore unrelated casts in STL allocate() since the allocator must cast // from void* to T* before object initialization completes. Don't match on the // namespace because not all allocators are in std:: if (D && SanOpts.has(SanitizerKind::CFIUnrelatedCast)) { if (matchesStlAllocatorFn(D, getContext())) SanOpts.Mask &= ~SanitizerKind::CFIUnrelatedCast; } d784 1 a784 4 bool InstrumentXray = ShouldXRayInstrumentFunction() && CGM.getCodeGenOpts().XRayInstrumentationBundle.has( XRayInstrKind::Function); if (D && InstrumentXray) { d802 4 a809 4 // Add profile-sample-accurate value. if (CGM.getCodeGenOpts().ProfileSampleAccurate) Fn->addFnAttr("profile-sample-accurate"); d820 2 a821 6 if (llvm::Constant *PrologueSig = getPrologueSignature(CGM, FD)) { // Remove any (C++17) exception specifications, to allow calling e.g. a // noexcept function through a non-noexcept pointer. auto ProtoTy = getContext().getFunctionTypeWithExceptionSpec(FD->getType(), EST_None); d823 2 a824 5 CGM.GetAddrOfRTTIDescriptor(ProtoTy, /*ForEH=*/true); llvm::Constant *FTRTTIConstEncoded = EncodeAddrForUseInPrologue(Fn, FTRTTIConst); llvm::Constant *PrologueStructElems[] = {PrologueSig, FTRTTIConstEncoded}; d885 1 a885 2 DI->EmitFunctionStart(GD, Loc, StartLoc, FnType, CurFn, CurFuncIsThunk, Builder); d888 2 a889 10 if (ShouldInstrumentFunction()) { if (CGM.getCodeGenOpts().InstrumentFunctions) CurFn->addFnAttr("instrument-function-entry", "__cyg_profile_func_enter"); if (CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining) CurFn->addFnAttr("instrument-function-entry-inlined", "__cyg_profile_func_enter"); if (CGM.getCodeGenOpts().InstrumentFunctionEntryBare) CurFn->addFnAttr("instrument-function-entry-inlined", "__cyg_profile_func_enter_bare"); } d896 5 a900 9 // Calls to fentry/mcount should not be generated if function has // the no_instrument_function attribute. if (!CurFuncDecl || !CurFuncDecl->hasAttr()) { if (CGM.getCodeGenOpts().CallFEntry) Fn->addFnAttr("fentry-call", "true"); else { Fn->addFnAttr("instrument-function-entry-inlined", getTarget().getMCountName()); } a942 5 // Emit OpenMP specific initialization of the device functions. if (getLangOpts().OpenMP && CurCodeDecl) CGM.getOpenMPRuntime().emitFunctionProlog(*this, CurCodeDecl); d996 2 a997 1 MD->getParent()->getLambdaCaptureDefault() == LCD_None) a1028 6 // TODO: Do we need to handle this in two places like we do with // target-features/target-cpu? if (CurFuncDecl) if (const auto *VecWidth = CurFuncDecl->getAttr()) LargestVectorWidth = VecWidth->getVectorWidth(); d1188 7 a1194 2 else if (isa(FD) && cast(FD)->isLambdaStaticInvoker()) { d1197 1 a1197 1 EmitLambdaStaticInvokeBody(cast(FD)); d1641 6 a1646 3 auto VlaSize = getVLASize(vlaType); SizeVal = VlaSize.NumElts; CharUnits eltSize = getContext().getTypeSizeInChars(VlaSize.Type); d1729 1 a1729 1 numVLAElements = getVLASize(cast(arrayType)).NumElts; d1810 2 a1811 1 CodeGenFunction::VlaSizePair CodeGenFunction::getVLASize(QualType type) { d1817 1 a1817 1 CodeGenFunction::VlaSizePair d1838 1 a1838 16 return { numElements, elementType }; } CodeGenFunction::VlaSizePair CodeGenFunction::getVLAElements1D(QualType type) { const VariableArrayType *vla = getContext().getAsVariableArrayType(type); assert(vla && "type was not a variable array type!"); return getVLAElements1D(vla); } CodeGenFunction::VlaSizePair CodeGenFunction::getVLAElements1D(const VariableArrayType *Vla) { llvm::Value *VlaSize = VLASizeMap[Vla->getSizeExpr()]; assert(VlaSize && "no size for VLA!"); assert(VlaSize->getType() == SizeTy); return { VlaSize, Vla->getElementType() }; d2121 1 a2121 1 Feature.split(OrFeatures, '|'); d2159 1 a2159 1 StringRef(FeatureList).split(ReqFeatures, ','); a2166 4 const TargetAttr *TD = TargetDecl->getAttr(); TargetAttr::ParsedTargetAttr ParsedAttr = CGM.filterFunctionTargetAttrs(TD); a2169 6 for (const auto &F : ParsedAttr.Features) { if (F[0] == '+' && CalleeFeatureMap.lookup(F.substr(1))) ReqFeatures.push_back(StringRef(F).substr(1)); } a2189 54 llvm::Value * CodeGenFunction::FormResolverCondition(const MultiVersionResolverOption &RO) { llvm::Value *TrueCondition = nullptr; if (!RO.ParsedAttribute.Architecture.empty()) TrueCondition = EmitX86CpuIs(RO.ParsedAttribute.Architecture); if (!RO.ParsedAttribute.Features.empty()) { SmallVector FeatureList; llvm::for_each(RO.ParsedAttribute.Features, [&FeatureList](const std::string &Feature) { FeatureList.push_back(StringRef{Feature}.substr(1)); }); llvm::Value *FeatureCmp = EmitX86CpuSupports(FeatureList); TrueCondition = TrueCondition ? Builder.CreateAnd(TrueCondition, FeatureCmp) : FeatureCmp; } return TrueCondition; } void CodeGenFunction::EmitMultiVersionResolver( llvm::Function *Resolver, ArrayRef Options) { assert((getContext().getTargetInfo().getTriple().getArch() == llvm::Triple::x86 || getContext().getTargetInfo().getTriple().getArch() == llvm::Triple::x86_64) && "Only implemented for x86 targets"); // Main function's basic block. llvm::BasicBlock *CurBlock = createBasicBlock("entry", Resolver); Builder.SetInsertPoint(CurBlock); EmitX86CpuInit(); llvm::Function *DefaultFunc = nullptr; for (const MultiVersionResolverOption &RO : Options) { Builder.SetInsertPoint(CurBlock); llvm::Value *TrueCondition = FormResolverCondition(RO); if (!TrueCondition) { DefaultFunc = RO.Function; } else { llvm::BasicBlock *RetBlock = createBasicBlock("ro_ret", Resolver); llvm::IRBuilder<> RetBuilder(RetBlock); RetBuilder.CreateRet(RO.Function); CurBlock = createBasicBlock("ro_else", Resolver); Builder.CreateCondBr(TrueCondition, RetBlock, CurBlock); } } assert(DefaultFunc && "No default version?"); // Emit return from the 'else-ist' block. Builder.SetInsertPoint(CurBlock); Builder.CreateRet(DefaultFunc); } @ 1.1.1.15 log @Mark old LLVM instance as dead. @ text @@ 1.1.1.7.4.1 log @file CodeGenFunction.cpp was added on branch tls-maxphys on 2014-08-19 23:47:27 +0000 @ text @d1 1712 @ 1.1.1.7.4.2 log @Rebase to HEAD as of a few days ago. @ text @a0 1712 //===--- CodeGenFunction.cpp - Emit LLVM Code from ASTs for a Function ----===// // // The LLVM Compiler Infrastructure // // This file is distributed under the University of Illinois Open Source // License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// // // This coordinates the per-function state used while generating code. // //===----------------------------------------------------------------------===// #include "CodeGenFunction.h" #include "CGCUDARuntime.h" #include "CGCXXABI.h" #include "CGDebugInfo.h" #include "CGOpenMPRuntime.h" #include "CodeGenModule.h" #include "CodeGenPGO.h" #include "TargetInfo.h" #include "clang/AST/ASTContext.h" #include "clang/AST/Decl.h" #include "clang/AST/DeclCXX.h" #include "clang/AST/StmtCXX.h" #include "clang/Basic/TargetInfo.h" #include "clang/CodeGen/CGFunctionInfo.h" #include "clang/Frontend/CodeGenOptions.h" #include "llvm/IR/DataLayout.h" #include "llvm/IR/Intrinsics.h" #include "llvm/IR/MDBuilder.h" #include "llvm/IR/Operator.h" using namespace clang; using namespace CodeGen; CodeGenFunction::CodeGenFunction(CodeGenModule &cgm, bool suppressNewContext) : CodeGenTypeCache(cgm), CGM(cgm), Target(cgm.getTarget()), Builder(cgm.getModule().getContext(), llvm::ConstantFolder(), CGBuilderInserterTy(this)), CapturedStmtInfo(nullptr), SanOpts(&CGM.getLangOpts().Sanitize), IsSanitizerScope(false), CurFuncIsThunk(false), AutoreleaseResult(false), BlockInfo(nullptr), BlockPointer(nullptr), LambdaThisCaptureField(nullptr), NormalCleanupDest(nullptr), NextCleanupDestIndex(1), FirstBlockInfo(nullptr), EHResumeBlock(nullptr), ExceptionSlot(nullptr), EHSelectorSlot(nullptr), DebugInfo(CGM.getModuleDebugInfo()), DisableDebugInfo(false), DidCallStackSave(false), IndirectBranch(nullptr), PGO(cgm), SwitchInsn(nullptr), SwitchWeights(nullptr), CaseRangeBlock(nullptr), UnreachableBlock(nullptr), NumReturnExprs(0), NumSimpleReturnExprs(0), CXXABIThisDecl(nullptr), CXXABIThisValue(nullptr), CXXThisValue(nullptr), CXXDefaultInitExprThis(nullptr), CXXStructorImplicitParamDecl(nullptr), CXXStructorImplicitParamValue(nullptr), OutermostConditional(nullptr), CurLexicalScope(nullptr), TerminateLandingPad(nullptr), TerminateHandler(nullptr), TrapBB(nullptr) { if (!suppressNewContext) CGM.getCXXABI().getMangleContext().startNewFunction(); llvm::FastMathFlags FMF; if (CGM.getLangOpts().FastMath) FMF.setUnsafeAlgebra(); if (CGM.getLangOpts().FiniteMathOnly) { FMF.setNoNaNs(); FMF.setNoInfs(); } Builder.SetFastMathFlags(FMF); } CodeGenFunction::~CodeGenFunction() { assert(LifetimeExtendedCleanupStack.empty() && "failed to emit a cleanup"); // If there are any unclaimed block infos, go ahead and destroy them // now. This can happen if IR-gen gets clever and skips evaluating // something. if (FirstBlockInfo) destroyBlockInfos(FirstBlockInfo); if (getLangOpts().OpenMP) { CGM.getOpenMPRuntime().FunctionFinished(*this); } } llvm::Type *CodeGenFunction::ConvertTypeForMem(QualType T) { return CGM.getTypes().ConvertTypeForMem(T); } llvm::Type *CodeGenFunction::ConvertType(QualType T) { return CGM.getTypes().ConvertType(T); } TypeEvaluationKind CodeGenFunction::getEvaluationKind(QualType type) { type = type.getCanonicalType(); while (true) { switch (type->getTypeClass()) { #define TYPE(name, parent) #define ABSTRACT_TYPE(name, parent) #define NON_CANONICAL_TYPE(name, parent) case Type::name: #define DEPENDENT_TYPE(name, parent) case Type::name: #define NON_CANONICAL_UNLESS_DEPENDENT_TYPE(name, parent) case Type::name: #include "clang/AST/TypeNodes.def" llvm_unreachable("non-canonical or dependent type in IR-generation"); case Type::Auto: llvm_unreachable("undeduced auto type in IR-generation"); // Various scalar types. case Type::Builtin: case Type::Pointer: case Type::BlockPointer: case Type::LValueReference: case Type::RValueReference: case Type::MemberPointer: case Type::Vector: case Type::ExtVector: case Type::FunctionProto: case Type::FunctionNoProto: case Type::Enum: case Type::ObjCObjectPointer: return TEK_Scalar; // Complexes. case Type::Complex: return TEK_Complex; // Arrays, records, and Objective-C objects. case Type::ConstantArray: case Type::IncompleteArray: case Type::VariableArray: case Type::Record: case Type::ObjCObject: case Type::ObjCInterface: return TEK_Aggregate; // We operate on atomic values according to their underlying type. case Type::Atomic: type = cast(type)->getValueType(); continue; } llvm_unreachable("unknown type kind!"); } } void CodeGenFunction::EmitReturnBlock() { // For cleanliness, we try to avoid emitting the return block for // simple cases. llvm::BasicBlock *CurBB = Builder.GetInsertBlock(); if (CurBB) { assert(!CurBB->getTerminator() && "Unexpected terminated block."); // We have a valid insert point, reuse it if it is empty or there are no // explicit jumps to the return block. if (CurBB->empty() || ReturnBlock.getBlock()->use_empty()) { ReturnBlock.getBlock()->replaceAllUsesWith(CurBB); delete ReturnBlock.getBlock(); } else EmitBlock(ReturnBlock.getBlock()); return; } // Otherwise, if the return block is the target of a single direct // branch then we can just put the code in that block instead. This // cleans up functions which started with a unified return block. if (ReturnBlock.getBlock()->hasOneUse()) { llvm::BranchInst *BI = dyn_cast(*ReturnBlock.getBlock()->user_begin()); if (BI && BI->isUnconditional() && BI->getSuccessor(0) == ReturnBlock.getBlock()) { // Reset insertion point, including debug location, and delete the // branch. This is really subtle and only works because the next change // in location will hit the caching in CGDebugInfo::EmitLocation and not // override this. Builder.SetCurrentDebugLocation(BI->getDebugLoc()); Builder.SetInsertPoint(BI->getParent()); BI->eraseFromParent(); delete ReturnBlock.getBlock(); return; } } // FIXME: We are at an unreachable point, there is no reason to emit the block // unless it has uses. However, we still need a place to put the debug // region.end for now. EmitBlock(ReturnBlock.getBlock()); } static void EmitIfUsed(CodeGenFunction &CGF, llvm::BasicBlock *BB) { if (!BB) return; if (!BB->use_empty()) return CGF.CurFn->getBasicBlockList().push_back(BB); delete BB; } void CodeGenFunction::FinishFunction(SourceLocation EndLoc) { assert(BreakContinueStack.empty() && "mismatched push/pop in break/continue stack!"); bool OnlySimpleReturnStmts = NumSimpleReturnExprs > 0 && NumSimpleReturnExprs == NumReturnExprs && ReturnBlock.getBlock()->use_empty(); // Usually the return expression is evaluated before the cleanup // code. If the function contains only a simple return statement, // such as a constant, the location before the cleanup code becomes // the last useful breakpoint in the function, because the simple // return expression will be evaluated after the cleanup code. To be // safe, set the debug location for cleanup code to the location of // the return statement. Otherwise the cleanup code should be at the // end of the function's lexical scope. // // If there are multiple branches to the return block, the branch // instructions will get the location of the return statements and // all will be fine. if (CGDebugInfo *DI = getDebugInfo()) { if (OnlySimpleReturnStmts) DI->EmitLocation(Builder, LastStopPoint); else DI->EmitLocation(Builder, EndLoc); } // Pop any cleanups that might have been associated with the // parameters. Do this in whatever block we're currently in; it's // important to do this before we enter the return block or return // edges will be *really* confused. bool EmitRetDbgLoc = true; if (EHStack.stable_begin() != PrologueCleanupDepth) { PopCleanupBlocks(PrologueCleanupDepth); // Make sure the line table doesn't jump back into the body for // the ret after it's been at EndLoc. EmitRetDbgLoc = false; if (CGDebugInfo *DI = getDebugInfo()) if (OnlySimpleReturnStmts) DI->EmitLocation(Builder, EndLoc); } // Emit function epilog (to return). EmitReturnBlock(); if (ShouldInstrumentFunction()) EmitFunctionInstrumentation("__cyg_profile_func_exit"); // Emit debug descriptor for function end. if (CGDebugInfo *DI = getDebugInfo()) { DI->EmitFunctionEnd(Builder); } EmitFunctionEpilog(*CurFnInfo, EmitRetDbgLoc, EndLoc); EmitEndEHSpec(CurCodeDecl); assert(EHStack.empty() && "did not remove all scopes from cleanup stack!"); // If someone did an indirect goto, emit the indirect goto block at the end of // the function. if (IndirectBranch) { EmitBlock(IndirectBranch->getParent()); Builder.ClearInsertionPoint(); } // Remove the AllocaInsertPt instruction, which is just a convenience for us. llvm::Instruction *Ptr = AllocaInsertPt; AllocaInsertPt = nullptr; Ptr->eraseFromParent(); // If someone took the address of a label but never did an indirect goto, we // made a zero entry PHI node, which is illegal, zap it now. if (IndirectBranch) { llvm::PHINode *PN = cast(IndirectBranch->getAddress()); if (PN->getNumIncomingValues() == 0) { PN->replaceAllUsesWith(llvm::UndefValue::get(PN->getType())); PN->eraseFromParent(); } } EmitIfUsed(*this, EHResumeBlock); EmitIfUsed(*this, TerminateLandingPad); EmitIfUsed(*this, TerminateHandler); EmitIfUsed(*this, UnreachableBlock); if (CGM.getCodeGenOpts().EmitDeclMetadata) EmitDeclMetadata(); for (SmallVectorImpl >::iterator I = DeferredReplacements.begin(), E = DeferredReplacements.end(); I != E; ++I) { I->first->replaceAllUsesWith(I->second); I->first->eraseFromParent(); } } /// ShouldInstrumentFunction - Return true if the current function should be /// instrumented with __cyg_profile_func_* calls bool CodeGenFunction::ShouldInstrumentFunction() { if (!CGM.getCodeGenOpts().InstrumentFunctions) return false; if (!CurFuncDecl || CurFuncDecl->hasAttr()) return false; return true; } /// EmitFunctionInstrumentation - Emit LLVM code to call the specified /// instrumentation function with the current function and the call site, if /// function instrumentation is enabled. void CodeGenFunction::EmitFunctionInstrumentation(const char *Fn) { // void __cyg_profile_func_{enter,exit} (void *this_fn, void *call_site); llvm::PointerType *PointerTy = Int8PtrTy; llvm::Type *ProfileFuncArgs[] = { PointerTy, PointerTy }; llvm::FunctionType *FunctionTy = llvm::FunctionType::get(VoidTy, ProfileFuncArgs, false); llvm::Constant *F = CGM.CreateRuntimeFunction(FunctionTy, Fn); llvm::CallInst *CallSite = Builder.CreateCall( CGM.getIntrinsic(llvm::Intrinsic::returnaddress), llvm::ConstantInt::get(Int32Ty, 0), "callsite"); llvm::Value *args[] = { llvm::ConstantExpr::getBitCast(CurFn, PointerTy), CallSite }; EmitNounwindRuntimeCall(F, args); } void CodeGenFunction::EmitMCountInstrumentation() { llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy, false); llvm::Constant *MCountFn = CGM.CreateRuntimeFunction(FTy, getTarget().getMCountName()); EmitNounwindRuntimeCall(MCountFn); } // OpenCL v1.2 s5.6.4.6 allows the compiler to store kernel argument // information in the program executable. The argument information stored // includes the argument name, its type, the address and access qualifiers used. static void GenOpenCLArgMetadata(const FunctionDecl *FD, llvm::Function *Fn, CodeGenModule &CGM,llvm::LLVMContext &Context, SmallVector &kernelMDArgs, CGBuilderTy& Builder, ASTContext &ASTCtx) { // Create MDNodes that represent the kernel arg metadata. // Each MDNode is a list in the form of "key", N number of values which is // the same number of values as their are kernel arguments. const PrintingPolicy &Policy = ASTCtx.getPrintingPolicy(); // MDNode for the kernel argument address space qualifiers. SmallVector addressQuals; addressQuals.push_back(llvm::MDString::get(Context, "kernel_arg_addr_space")); // MDNode for the kernel argument access qualifiers (images only). SmallVector accessQuals; accessQuals.push_back(llvm::MDString::get(Context, "kernel_arg_access_qual")); // MDNode for the kernel argument type names. SmallVector argTypeNames; argTypeNames.push_back(llvm::MDString::get(Context, "kernel_arg_type")); // MDNode for the kernel argument base type names. SmallVector argBaseTypeNames; argBaseTypeNames.push_back( llvm::MDString::get(Context, "kernel_arg_base_type")); // MDNode for the kernel argument type qualifiers. SmallVector argTypeQuals; argTypeQuals.push_back(llvm::MDString::get(Context, "kernel_arg_type_qual")); // MDNode for the kernel argument names. SmallVector argNames; argNames.push_back(llvm::MDString::get(Context, "kernel_arg_name")); for (unsigned i = 0, e = FD->getNumParams(); i != e; ++i) { const ParmVarDecl *parm = FD->getParamDecl(i); QualType ty = parm->getType(); std::string typeQuals; if (ty->isPointerType()) { QualType pointeeTy = ty->getPointeeType(); // Get address qualifier. addressQuals.push_back(Builder.getInt32(ASTCtx.getTargetAddressSpace( pointeeTy.getAddressSpace()))); // Get argument type name. std::string typeName = pointeeTy.getUnqualifiedType().getAsString(Policy) + "*"; // Turn "unsigned type" to "utype" std::string::size_type pos = typeName.find("unsigned"); if (pointeeTy.isCanonical() && pos != std::string::npos) typeName.erase(pos+1, 8); argTypeNames.push_back(llvm::MDString::get(Context, typeName)); std::string baseTypeName = pointeeTy.getUnqualifiedType().getCanonicalType().getAsString( Policy) + "*"; // Turn "unsigned type" to "utype" pos = baseTypeName.find("unsigned"); if (pos != std::string::npos) baseTypeName.erase(pos+1, 8); argBaseTypeNames.push_back(llvm::MDString::get(Context, baseTypeName)); // Get argument type qualifiers: if (ty.isRestrictQualified()) typeQuals = "restrict"; if (pointeeTy.isConstQualified() || (pointeeTy.getAddressSpace() == LangAS::opencl_constant)) typeQuals += typeQuals.empty() ? "const" : " const"; if (pointeeTy.isVolatileQualified()) typeQuals += typeQuals.empty() ? "volatile" : " volatile"; } else { uint32_t AddrSpc = 0; if (ty->isImageType()) AddrSpc = CGM.getContext().getTargetAddressSpace(LangAS::opencl_global); addressQuals.push_back(Builder.getInt32(AddrSpc)); // Get argument type name. std::string typeName = ty.getUnqualifiedType().getAsString(Policy); // Turn "unsigned type" to "utype" std::string::size_type pos = typeName.find("unsigned"); if (ty.isCanonical() && pos != std::string::npos) typeName.erase(pos+1, 8); argTypeNames.push_back(llvm::MDString::get(Context, typeName)); std::string baseTypeName = ty.getUnqualifiedType().getCanonicalType().getAsString(Policy); // Turn "unsigned type" to "utype" pos = baseTypeName.find("unsigned"); if (pos != std::string::npos) baseTypeName.erase(pos+1, 8); argBaseTypeNames.push_back(llvm::MDString::get(Context, baseTypeName)); // Get argument type qualifiers: if (ty.isConstQualified()) typeQuals = "const"; if (ty.isVolatileQualified()) typeQuals += typeQuals.empty() ? "volatile" : " volatile"; } argTypeQuals.push_back(llvm::MDString::get(Context, typeQuals)); // Get image access qualifier: if (ty->isImageType()) { const OpenCLImageAccessAttr *A = parm->getAttr(); if (A && A->isWriteOnly()) accessQuals.push_back(llvm::MDString::get(Context, "write_only")); else accessQuals.push_back(llvm::MDString::get(Context, "read_only")); // FIXME: what about read_write? } else accessQuals.push_back(llvm::MDString::get(Context, "none")); // Get argument name. argNames.push_back(llvm::MDString::get(Context, parm->getName())); } kernelMDArgs.push_back(llvm::MDNode::get(Context, addressQuals)); kernelMDArgs.push_back(llvm::MDNode::get(Context, accessQuals)); kernelMDArgs.push_back(llvm::MDNode::get(Context, argTypeNames)); kernelMDArgs.push_back(llvm::MDNode::get(Context, argBaseTypeNames)); kernelMDArgs.push_back(llvm::MDNode::get(Context, argTypeQuals)); kernelMDArgs.push_back(llvm::MDNode::get(Context, argNames)); } void CodeGenFunction::EmitOpenCLKernelMetadata(const FunctionDecl *FD, llvm::Function *Fn) { if (!FD->hasAttr()) return; llvm::LLVMContext &Context = getLLVMContext(); SmallVector kernelMDArgs; kernelMDArgs.push_back(Fn); if (CGM.getCodeGenOpts().EmitOpenCLArgMetadata) GenOpenCLArgMetadata(FD, Fn, CGM, Context, kernelMDArgs, Builder, getContext()); if (const VecTypeHintAttr *A = FD->getAttr()) { QualType hintQTy = A->getTypeHint(); const ExtVectorType *hintEltQTy = hintQTy->getAs(); bool isSignedInteger = hintQTy->isSignedIntegerType() || (hintEltQTy && hintEltQTy->getElementType()->isSignedIntegerType()); llvm::Value *attrMDArgs[] = { llvm::MDString::get(Context, "vec_type_hint"), llvm::UndefValue::get(CGM.getTypes().ConvertType(A->getTypeHint())), llvm::ConstantInt::get( llvm::IntegerType::get(Context, 32), llvm::APInt(32, (uint64_t)(isSignedInteger ? 1 : 0))) }; kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs)); } if (const WorkGroupSizeHintAttr *A = FD->getAttr()) { llvm::Value *attrMDArgs[] = { llvm::MDString::get(Context, "work_group_size_hint"), Builder.getInt32(A->getXDim()), Builder.getInt32(A->getYDim()), Builder.getInt32(A->getZDim()) }; kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs)); } if (const ReqdWorkGroupSizeAttr *A = FD->getAttr()) { llvm::Value *attrMDArgs[] = { llvm::MDString::get(Context, "reqd_work_group_size"), Builder.getInt32(A->getXDim()), Builder.getInt32(A->getYDim()), Builder.getInt32(A->getZDim()) }; kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs)); } llvm::MDNode *kernelMDNode = llvm::MDNode::get(Context, kernelMDArgs); llvm::NamedMDNode *OpenCLKernelMetadata = CGM.getModule().getOrInsertNamedMetadata("opencl.kernels"); OpenCLKernelMetadata->addOperand(kernelMDNode); } /// Determine whether the function F ends with a return stmt. static bool endsWithReturn(const Decl* F) { const Stmt *Body = nullptr; if (auto *FD = dyn_cast_or_null(F)) Body = FD->getBody(); else if (auto *OMD = dyn_cast_or_null(F)) Body = OMD->getBody(); if (auto *CS = dyn_cast_or_null(Body)) { auto LastStmt = CS->body_rbegin(); if (LastStmt != CS->body_rend()) return isa(*LastStmt); } return false; } void CodeGenFunction::StartFunction(GlobalDecl GD, QualType RetTy, llvm::Function *Fn, const CGFunctionInfo &FnInfo, const FunctionArgList &Args, SourceLocation Loc, SourceLocation StartLoc) { const Decl *D = GD.getDecl(); DidCallStackSave = false; CurCodeDecl = D; CurFuncDecl = (D ? D->getNonClosureContext() : nullptr); FnRetTy = RetTy; CurFn = Fn; CurFnInfo = &FnInfo; assert(CurFn->isDeclaration() && "Function already has body?"); if (CGM.getSanitizerBlacklist().isIn(*Fn)) SanOpts = &SanitizerOptions::Disabled; // Pass inline keyword to optimizer if it appears explicitly on any // declaration. Also, in the case of -fno-inline attach NoInline // attribute to all function that are not marked AlwaysInline. if (const FunctionDecl *FD = dyn_cast_or_null(D)) { if (!CGM.getCodeGenOpts().NoInline) { for (auto RI : FD->redecls()) if (RI->isInlineSpecified()) { Fn->addFnAttr(llvm::Attribute::InlineHint); break; } } else if (!FD->hasAttr()) Fn->addFnAttr(llvm::Attribute::NoInline); } if (getLangOpts().OpenCL) { // Add metadata for a kernel function. if (const FunctionDecl *FD = dyn_cast_or_null(D)) EmitOpenCLKernelMetadata(FD, Fn); } // If we are checking function types, emit a function type signature as // prefix data. if (getLangOpts().CPlusPlus && SanOpts->Function) { if (const FunctionDecl *FD = dyn_cast_or_null(D)) { if (llvm::Constant *PrefixSig = CGM.getTargetCodeGenInfo().getUBSanFunctionSignature(CGM)) { llvm::Constant *FTRTTIConst = CGM.GetAddrOfRTTIDescriptor(FD->getType(), /*ForEH=*/true); llvm::Constant *PrefixStructElems[] = { PrefixSig, FTRTTIConst }; llvm::Constant *PrefixStructConst = llvm::ConstantStruct::getAnon(PrefixStructElems, /*Packed=*/true); Fn->setPrefixData(PrefixStructConst); } } } llvm::BasicBlock *EntryBB = createBasicBlock("entry", CurFn); // Create a marker to make it easy to insert allocas into the entryblock // later. Don't create this with the builder, because we don't want it // folded. llvm::Value *Undef = llvm::UndefValue::get(Int32Ty); AllocaInsertPt = new llvm::BitCastInst(Undef, Int32Ty, "", EntryBB); if (Builder.isNamePreserving()) AllocaInsertPt->setName("allocapt"); ReturnBlock = getJumpDestInCurrentScope("return"); Builder.SetInsertPoint(EntryBB); // Emit subprogram debug descriptor. if (CGDebugInfo *DI = getDebugInfo()) { SmallVector ArgTypes; for (FunctionArgList::const_iterator i = Args.begin(), e = Args.end(); i != e; ++i) { ArgTypes.push_back((*i)->getType()); } QualType FnType = getContext().getFunctionType(RetTy, ArgTypes, FunctionProtoType::ExtProtoInfo()); DI->EmitFunctionStart(GD, Loc, StartLoc, FnType, CurFn, Builder); } if (ShouldInstrumentFunction()) EmitFunctionInstrumentation("__cyg_profile_func_enter"); if (CGM.getCodeGenOpts().InstrumentForProfiling) EmitMCountInstrumentation(); if (RetTy->isVoidType()) { // Void type; nothing to return. ReturnValue = nullptr; // Count the implicit return. if (!endsWithReturn(D)) ++NumReturnExprs; } else if (CurFnInfo->getReturnInfo().getKind() == ABIArgInfo::Indirect && !hasScalarEvaluationKind(CurFnInfo->getReturnType())) { // Indirect aggregate return; emit returned value directly into sret slot. // This reduces code size, and affects correctness in C++. auto AI = CurFn->arg_begin(); if (CurFnInfo->getReturnInfo().isSRetAfterThis()) ++AI; ReturnValue = AI; } else if (CurFnInfo->getReturnInfo().getKind() == ABIArgInfo::InAlloca && !hasScalarEvaluationKind(CurFnInfo->getReturnType())) { // Load the sret pointer from the argument struct and return into that. unsigned Idx = CurFnInfo->getReturnInfo().getInAllocaFieldIndex(); llvm::Function::arg_iterator EI = CurFn->arg_end(); --EI; llvm::Value *Addr = Builder.CreateStructGEP(EI, Idx); ReturnValue = Builder.CreateLoad(Addr, "agg.result"); } else { ReturnValue = CreateIRTemp(RetTy, "retval"); // Tell the epilog emitter to autorelease the result. We do this // now so that various specialized functions can suppress it // during their IR-generation. if (getLangOpts().ObjCAutoRefCount && !CurFnInfo->isReturnsRetained() && RetTy->isObjCRetainableType()) AutoreleaseResult = true; } EmitStartEHSpec(CurCodeDecl); PrologueCleanupDepth = EHStack.stable_begin(); EmitFunctionProlog(*CurFnInfo, CurFn, Args); if (D && isa(D) && cast(D)->isInstance()) { CGM.getCXXABI().EmitInstanceFunctionProlog(*this); const CXXMethodDecl *MD = cast(D); if (MD->getParent()->isLambda() && MD->getOverloadedOperator() == OO_Call) { // We're in a lambda; figure out the captures. MD->getParent()->getCaptureFields(LambdaCaptureFields, LambdaThisCaptureField); if (LambdaThisCaptureField) { // If this lambda captures this, load it. LValue ThisLValue = EmitLValueForLambdaField(LambdaThisCaptureField); CXXThisValue = EmitLoadOfLValue(ThisLValue, SourceLocation()).getScalarVal(); } } else { // Not in a lambda; just use 'this' from the method. // FIXME: Should we generate a new load for each use of 'this'? The // fast register allocator would be happier... CXXThisValue = CXXABIThisValue; } } // If any of the arguments have a variably modified type, make sure to // emit the type size. for (FunctionArgList::const_iterator i = Args.begin(), e = Args.end(); i != e; ++i) { const VarDecl *VD = *i; // Dig out the type as written from ParmVarDecls; it's unclear whether // the standard (C99 6.9.1p10) requires this, but we're following the // precedent set by gcc. QualType Ty; if (const ParmVarDecl *PVD = dyn_cast(VD)) Ty = PVD->getOriginalType(); else Ty = VD->getType(); if (Ty->isVariablyModifiedType()) EmitVariablyModifiedType(Ty); } // Emit a location at the end of the prologue. if (CGDebugInfo *DI = getDebugInfo()) DI->EmitLocation(Builder, StartLoc); } void CodeGenFunction::EmitFunctionBody(FunctionArgList &Args, const Stmt *Body) { RegionCounter Cnt = getPGORegionCounter(Body); Cnt.beginRegion(Builder); if (const CompoundStmt *S = dyn_cast(Body)) EmitCompoundStmtWithoutScope(*S); else EmitStmt(Body); } /// When instrumenting to collect profile data, the counts for some blocks /// such as switch cases need to not include the fall-through counts, so /// emit a branch around the instrumentation code. When not instrumenting, /// this just calls EmitBlock(). void CodeGenFunction::EmitBlockWithFallThrough(llvm::BasicBlock *BB, RegionCounter &Cnt) { llvm::BasicBlock *SkipCountBB = nullptr; if (HaveInsertPoint() && CGM.getCodeGenOpts().ProfileInstrGenerate) { // When instrumenting for profiling, the fallthrough to certain // statements needs to skip over the instrumentation code so that we // get an accurate count. SkipCountBB = createBasicBlock("skipcount"); EmitBranch(SkipCountBB); } EmitBlock(BB); Cnt.beginRegion(Builder, /*AddIncomingFallThrough=*/true); if (SkipCountBB) EmitBlock(SkipCountBB); } /// Tries to mark the given function nounwind based on the /// non-existence of any throwing calls within it. We believe this is /// lightweight enough to do at -O0. static void TryMarkNoThrow(llvm::Function *F) { // LLVM treats 'nounwind' on a function as part of the type, so we // can't do this on functions that can be overwritten. if (F->mayBeOverridden()) return; for (llvm::Function::iterator FI = F->begin(), FE = F->end(); FI != FE; ++FI) for (llvm::BasicBlock::iterator BI = FI->begin(), BE = FI->end(); BI != BE; ++BI) if (llvm::CallInst *Call = dyn_cast(&*BI)) { if (!Call->doesNotThrow()) return; } else if (isa(&*BI)) { return; } F->setDoesNotThrow(); } static void EmitSizedDeallocationFunction(CodeGenFunction &CGF, const FunctionDecl *UnsizedDealloc) { // This is a weak discardable definition of the sized deallocation function. CGF.CurFn->setLinkage(llvm::Function::LinkOnceAnyLinkage); // Call the unsized deallocation function and forward the first argument // unchanged. llvm::Constant *Unsized = CGF.CGM.GetAddrOfFunction(UnsizedDealloc); CGF.Builder.CreateCall(Unsized, &*CGF.CurFn->arg_begin()); } void CodeGenFunction::GenerateCode(GlobalDecl GD, llvm::Function *Fn, const CGFunctionInfo &FnInfo) { const FunctionDecl *FD = cast(GD.getDecl()); // Check if we should generate debug info for this function. if (FD->hasAttr()) DebugInfo = nullptr; // disable debug info indefinitely for this function FunctionArgList Args; QualType ResTy = FD->getReturnType(); CurGD = GD; const CXXMethodDecl *MD = dyn_cast(FD); if (MD && MD->isInstance()) { if (CGM.getCXXABI().HasThisReturn(GD)) ResTy = MD->getThisType(getContext()); CGM.getCXXABI().buildThisParam(*this, Args); } for (unsigned i = 0, e = FD->getNumParams(); i != e; ++i) Args.push_back(FD->getParamDecl(i)); if (MD && (isa(MD) || isa(MD))) CGM.getCXXABI().addImplicitStructorParams(*this, ResTy, Args); SourceRange BodyRange; if (Stmt *Body = FD->getBody()) BodyRange = Body->getSourceRange(); CurEHLocation = BodyRange.getEnd(); // Use the location of the start of the function to determine where // the function definition is located. By default use the location // of the declaration as the location for the subprogram. A function // may lack a declaration in the source code if it is created by code // gen. (examples: _GLOBAL__I_a, __cxx_global_array_dtor, thunk). SourceLocation Loc = FD->getLocation(); // If this is a function specialization then use the pattern body // as the location for the function. if (const FunctionDecl *SpecDecl = FD->getTemplateInstantiationPattern()) if (SpecDecl->hasBody(SpecDecl)) Loc = SpecDecl->getLocation(); // Emit the standard function prologue. StartFunction(GD, ResTy, Fn, FnInfo, Args, Loc, BodyRange.getBegin()); // Generate the body of the function. PGO.checkGlobalDecl(GD); PGO.assignRegionCounters(GD.getDecl(), CurFn); if (isa(FD)) EmitDestructorBody(Args); else if (isa(FD)) EmitConstructorBody(Args); else if (getLangOpts().CUDA && !CGM.getCodeGenOpts().CUDAIsDevice && FD->hasAttr()) CGM.getCUDARuntime().EmitDeviceStubBody(*this, Args); else if (isa(FD) && cast(FD)->isLambdaToBlockPointerConversion()) { // The lambda conversion to block pointer is special; the semantics can't be // expressed in the AST, so IRGen needs to special-case it. EmitLambdaToBlockPointerBody(Args); } else if (isa(FD) && cast(FD)->isLambdaStaticInvoker()) { // The lambda static invoker function is special, because it forwards or // clones the body of the function call operator (but is actually static). EmitLambdaStaticInvokeFunction(cast(FD)); } else if (FD->isDefaulted() && isa(FD) && (cast(FD)->isCopyAssignmentOperator() || cast(FD)->isMoveAssignmentOperator())) { // Implicit copy-assignment gets the same special treatment as implicit // copy-constructors. emitImplicitAssignmentOperatorBody(Args); } else if (Stmt *Body = FD->getBody()) { EmitFunctionBody(Args, Body); } else if (FunctionDecl *UnsizedDealloc = FD->getCorrespondingUnsizedGlobalDeallocationFunction()) { // Global sized deallocation functions get an implicit weak definition if // they don't have an explicit definition. EmitSizedDeallocationFunction(*this, UnsizedDealloc); } else llvm_unreachable("no definition for emitted function"); // C++11 [stmt.return]p2: // Flowing off the end of a function [...] results in undefined behavior in // a value-returning function. // C11 6.9.1p12: // If the '}' that terminates a function is reached, and the value of the // function call is used by the caller, the behavior is undefined. if (getLangOpts().CPlusPlus && !FD->hasImplicitReturnZero() && !FD->getReturnType()->isVoidType() && Builder.GetInsertBlock()) { if (SanOpts->Return) { SanitizerScope SanScope(this); EmitCheck(Builder.getFalse(), "missing_return", EmitCheckSourceLocation(FD->getLocation()), ArrayRef(), CRK_Unrecoverable); } else if (CGM.getCodeGenOpts().OptimizationLevel == 0) Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::trap)); Builder.CreateUnreachable(); Builder.ClearInsertionPoint(); } // Emit the standard function epilogue. FinishFunction(BodyRange.getEnd()); // If we haven't marked the function nothrow through other means, do // a quick pass now to see if we can. if (!CurFn->doesNotThrow()) TryMarkNoThrow(CurFn); PGO.emitInstrumentationData(); PGO.destroyRegionCounters(); } /// ContainsLabel - Return true if the statement contains a label in it. If /// this statement is not executed normally, it not containing a label means /// that we can just remove the code. bool CodeGenFunction::ContainsLabel(const Stmt *S, bool IgnoreCaseStmts) { // Null statement, not a label! if (!S) return false; // If this is a label, we have to emit the code, consider something like: // if (0) { ... foo: bar(); } goto foo; // // TODO: If anyone cared, we could track __label__'s, since we know that you // can't jump to one from outside their declared region. if (isa(S)) return true; // If this is a case/default statement, and we haven't seen a switch, we have // to emit the code. if (isa(S) && !IgnoreCaseStmts) return true; // If this is a switch statement, we want to ignore cases below it. if (isa(S)) IgnoreCaseStmts = true; // Scan subexpressions for verboten labels. for (Stmt::const_child_range I = S->children(); I; ++I) if (ContainsLabel(*I, IgnoreCaseStmts)) return true; return false; } /// containsBreak - Return true if the statement contains a break out of it. /// If the statement (recursively) contains a switch or loop with a break /// inside of it, this is fine. bool CodeGenFunction::containsBreak(const Stmt *S) { // Null statement, not a label! if (!S) return false; // If this is a switch or loop that defines its own break scope, then we can // include it and anything inside of it. if (isa(S) || isa(S) || isa(S) || isa(S)) return false; if (isa(S)) return true; // Scan subexpressions for verboten breaks. for (Stmt::const_child_range I = S->children(); I; ++I) if (containsBreak(*I)) return true; return false; } /// ConstantFoldsToSimpleInteger - If the specified expression does not fold /// to a constant, or if it does but contains a label, return false. If it /// constant folds return true and set the boolean result in Result. bool CodeGenFunction::ConstantFoldsToSimpleInteger(const Expr *Cond, bool &ResultBool) { llvm::APSInt ResultInt; if (!ConstantFoldsToSimpleInteger(Cond, ResultInt)) return false; ResultBool = ResultInt.getBoolValue(); return true; } /// ConstantFoldsToSimpleInteger - If the specified expression does not fold /// to a constant, or if it does but contains a label, return false. If it /// constant folds return true and set the folded value. bool CodeGenFunction:: ConstantFoldsToSimpleInteger(const Expr *Cond, llvm::APSInt &ResultInt) { // FIXME: Rename and handle conversion of other evaluatable things // to bool. llvm::APSInt Int; if (!Cond->EvaluateAsInt(Int, getContext())) return false; // Not foldable, not integer or not fully evaluatable. if (CodeGenFunction::ContainsLabel(Cond)) return false; // Contains a label. ResultInt = Int; return true; } /// EmitBranchOnBoolExpr - Emit a branch on a boolean condition (e.g. for an if /// statement) to the specified blocks. Based on the condition, this might try /// to simplify the codegen of the conditional based on the branch. /// void CodeGenFunction::EmitBranchOnBoolExpr(const Expr *Cond, llvm::BasicBlock *TrueBlock, llvm::BasicBlock *FalseBlock, uint64_t TrueCount) { Cond = Cond->IgnoreParens(); if (const BinaryOperator *CondBOp = dyn_cast(Cond)) { // Handle X && Y in a condition. if (CondBOp->getOpcode() == BO_LAnd) { RegionCounter Cnt = getPGORegionCounter(CondBOp); // If we have "1 && X", simplify the code. "0 && X" would have constant // folded if the case was simple enough. bool ConstantBool = false; if (ConstantFoldsToSimpleInteger(CondBOp->getLHS(), ConstantBool) && ConstantBool) { // br(1 && X) -> br(X). Cnt.beginRegion(Builder); return EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, TrueCount); } // If we have "X && 1", simplify the code to use an uncond branch. // "X && 0" would have been constant folded to 0. if (ConstantFoldsToSimpleInteger(CondBOp->getRHS(), ConstantBool) && ConstantBool) { // br(X && 1) -> br(X). return EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, FalseBlock, TrueCount); } // Emit the LHS as a conditional. If the LHS conditional is false, we // want to jump to the FalseBlock. llvm::BasicBlock *LHSTrue = createBasicBlock("land.lhs.true"); // The counter tells us how often we evaluate RHS, and all of TrueCount // can be propagated to that branch. uint64_t RHSCount = Cnt.getCount(); ConditionalEvaluation eval(*this); EmitBranchOnBoolExpr(CondBOp->getLHS(), LHSTrue, FalseBlock, RHSCount); EmitBlock(LHSTrue); // Any temporaries created here are conditional. Cnt.beginRegion(Builder); eval.begin(*this); EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, TrueCount); eval.end(*this); return; } if (CondBOp->getOpcode() == BO_LOr) { RegionCounter Cnt = getPGORegionCounter(CondBOp); // If we have "0 || X", simplify the code. "1 || X" would have constant // folded if the case was simple enough. bool ConstantBool = false; if (ConstantFoldsToSimpleInteger(CondBOp->getLHS(), ConstantBool) && !ConstantBool) { // br(0 || X) -> br(X). Cnt.beginRegion(Builder); return EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, TrueCount); } // If we have "X || 0", simplify the code to use an uncond branch. // "X || 1" would have been constant folded to 1. if (ConstantFoldsToSimpleInteger(CondBOp->getRHS(), ConstantBool) && !ConstantBool) { // br(X || 0) -> br(X). return EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, FalseBlock, TrueCount); } // Emit the LHS as a conditional. If the LHS conditional is true, we // want to jump to the TrueBlock. llvm::BasicBlock *LHSFalse = createBasicBlock("lor.lhs.false"); // We have the count for entry to the RHS and for the whole expression // being true, so we can divy up True count between the short circuit and // the RHS. uint64_t LHSCount = Cnt.getParentCount() - Cnt.getCount(); uint64_t RHSCount = TrueCount - LHSCount; ConditionalEvaluation eval(*this); EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, LHSFalse, LHSCount); EmitBlock(LHSFalse); // Any temporaries created here are conditional. Cnt.beginRegion(Builder); eval.begin(*this); EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, RHSCount); eval.end(*this); return; } } if (const UnaryOperator *CondUOp = dyn_cast(Cond)) { // br(!x, t, f) -> br(x, f, t) if (CondUOp->getOpcode() == UO_LNot) { // Negate the count. uint64_t FalseCount = PGO.getCurrentRegionCount() - TrueCount; // Negate the condition and swap the destination blocks. return EmitBranchOnBoolExpr(CondUOp->getSubExpr(), FalseBlock, TrueBlock, FalseCount); } } if (const ConditionalOperator *CondOp = dyn_cast(Cond)) { // br(c ? x : y, t, f) -> br(c, br(x, t, f), br(y, t, f)) llvm::BasicBlock *LHSBlock = createBasicBlock("cond.true"); llvm::BasicBlock *RHSBlock = createBasicBlock("cond.false"); RegionCounter Cnt = getPGORegionCounter(CondOp); ConditionalEvaluation cond(*this); EmitBranchOnBoolExpr(CondOp->getCond(), LHSBlock, RHSBlock, Cnt.getCount()); // When computing PGO branch weights, we only know the overall count for // the true block. This code is essentially doing tail duplication of the // naive code-gen, introducing new edges for which counts are not // available. Divide the counts proportionally between the LHS and RHS of // the conditional operator. uint64_t LHSScaledTrueCount = 0; if (TrueCount) { double LHSRatio = Cnt.getCount() / (double) Cnt.getParentCount(); LHSScaledTrueCount = TrueCount * LHSRatio; } cond.begin(*this); EmitBlock(LHSBlock); Cnt.beginRegion(Builder); EmitBranchOnBoolExpr(CondOp->getLHS(), TrueBlock, FalseBlock, LHSScaledTrueCount); cond.end(*this); cond.begin(*this); EmitBlock(RHSBlock); EmitBranchOnBoolExpr(CondOp->getRHS(), TrueBlock, FalseBlock, TrueCount - LHSScaledTrueCount); cond.end(*this); return; } if (const CXXThrowExpr *Throw = dyn_cast(Cond)) { // Conditional operator handling can give us a throw expression as a // condition for a case like: // br(c ? throw x : y, t, f) -> br(c, br(throw x, t, f), br(y, t, f) // Fold this to: // br(c, throw x, br(y, t, f)) EmitCXXThrowExpr(Throw, /*KeepInsertionPoint*/false); return; } // Create branch weights based on the number of times we get here and the // number of times the condition should be true. uint64_t CurrentCount = std::max(PGO.getCurrentRegionCount(), TrueCount); llvm::MDNode *Weights = PGO.createBranchWeights(TrueCount, CurrentCount - TrueCount); // Emit the code with the fully general case. llvm::Value *CondV = EvaluateExprAsBool(Cond); Builder.CreateCondBr(CondV, TrueBlock, FalseBlock, Weights); } /// ErrorUnsupported - Print out an error that codegen doesn't support the /// specified stmt yet. void CodeGenFunction::ErrorUnsupported(const Stmt *S, const char *Type) { CGM.ErrorUnsupported(S, Type); } /// emitNonZeroVLAInit - Emit the "zero" initialization of a /// variable-length array whose elements have a non-zero bit-pattern. /// /// \param baseType the inner-most element type of the array /// \param src - a char* pointing to the bit-pattern for a single /// base element of the array /// \param sizeInChars - the total size of the VLA, in chars static void emitNonZeroVLAInit(CodeGenFunction &CGF, QualType baseType, llvm::Value *dest, llvm::Value *src, llvm::Value *sizeInChars) { std::pair baseSizeAndAlign = CGF.getContext().getTypeInfoInChars(baseType); CGBuilderTy &Builder = CGF.Builder; llvm::Value *baseSizeInChars = llvm::ConstantInt::get(CGF.IntPtrTy, baseSizeAndAlign.first.getQuantity()); llvm::Type *i8p = Builder.getInt8PtrTy(); llvm::Value *begin = Builder.CreateBitCast(dest, i8p, "vla.begin"); llvm::Value *end = Builder.CreateInBoundsGEP(dest, sizeInChars, "vla.end"); llvm::BasicBlock *originBB = CGF.Builder.GetInsertBlock(); llvm::BasicBlock *loopBB = CGF.createBasicBlock("vla-init.loop"); llvm::BasicBlock *contBB = CGF.createBasicBlock("vla-init.cont"); // Make a loop over the VLA. C99 guarantees that the VLA element // count must be nonzero. CGF.EmitBlock(loopBB); llvm::PHINode *cur = Builder.CreatePHI(i8p, 2, "vla.cur"); cur->addIncoming(begin, originBB); // memcpy the individual element bit-pattern. Builder.CreateMemCpy(cur, src, baseSizeInChars, baseSizeAndAlign.second.getQuantity(), /*volatile*/ false); // Go to the next element. llvm::Value *next = Builder.CreateConstInBoundsGEP1_32(cur, 1, "vla.next"); // Leave if that's the end of the VLA. llvm::Value *done = Builder.CreateICmpEQ(next, end, "vla-init.isdone"); Builder.CreateCondBr(done, contBB, loopBB); cur->addIncoming(next, loopBB); CGF.EmitBlock(contBB); } void CodeGenFunction::EmitNullInitialization(llvm::Value *DestPtr, QualType Ty) { // Ignore empty classes in C++. if (getLangOpts().CPlusPlus) { if (const RecordType *RT = Ty->getAs()) { if (cast(RT->getDecl())->isEmpty()) return; } } // Cast the dest ptr to the appropriate i8 pointer type. unsigned DestAS = cast(DestPtr->getType())->getAddressSpace(); llvm::Type *BP = Builder.getInt8PtrTy(DestAS); if (DestPtr->getType() != BP) DestPtr = Builder.CreateBitCast(DestPtr, BP); // Get size and alignment info for this aggregate. std::pair TypeInfo = getContext().getTypeInfoInChars(Ty); CharUnits Size = TypeInfo.first; CharUnits Align = TypeInfo.second; llvm::Value *SizeVal; const VariableArrayType *vla; // Don't bother emitting a zero-byte memset. if (Size.isZero()) { // But note that getTypeInfo returns 0 for a VLA. if (const VariableArrayType *vlaType = dyn_cast_or_null( getContext().getAsArrayType(Ty))) { QualType eltType; llvm::Value *numElts; std::tie(numElts, eltType) = getVLASize(vlaType); SizeVal = numElts; CharUnits eltSize = getContext().getTypeSizeInChars(eltType); if (!eltSize.isOne()) SizeVal = Builder.CreateNUWMul(SizeVal, CGM.getSize(eltSize)); vla = vlaType; } else { return; } } else { SizeVal = CGM.getSize(Size); vla = nullptr; } // If the type contains a pointer to data member we can't memset it to zero. // Instead, create a null constant and copy it to the destination. // TODO: there are other patterns besides zero that we can usefully memset, // like -1, which happens to be the pattern used by member-pointers. if (!CGM.getTypes().isZeroInitializable(Ty)) { // For a VLA, emit a single element, then splat that over the VLA. if (vla) Ty = getContext().getBaseElementType(vla); llvm::Constant *NullConstant = CGM.EmitNullConstant(Ty); llvm::GlobalVariable *NullVariable = new llvm::GlobalVariable(CGM.getModule(), NullConstant->getType(), /*isConstant=*/true, llvm::GlobalVariable::PrivateLinkage, NullConstant, Twine()); llvm::Value *SrcPtr = Builder.CreateBitCast(NullVariable, Builder.getInt8PtrTy()); if (vla) return emitNonZeroVLAInit(*this, Ty, DestPtr, SrcPtr, SizeVal); // Get and call the appropriate llvm.memcpy overload. Builder.CreateMemCpy(DestPtr, SrcPtr, SizeVal, Align.getQuantity(), false); return; } // Otherwise, just memset the whole thing to zero. This is legal // because in LLVM, all default initializers (other than the ones we just // handled above) are guaranteed to have a bit pattern of all zeros. Builder.CreateMemSet(DestPtr, Builder.getInt8(0), SizeVal, Align.getQuantity(), false); } llvm::BlockAddress *CodeGenFunction::GetAddrOfLabel(const LabelDecl *L) { // Make sure that there is a block for the indirect goto. if (!IndirectBranch) GetIndirectGotoBlock(); llvm::BasicBlock *BB = getJumpDestForLabel(L).getBlock(); // Make sure the indirect branch includes all of the address-taken blocks. IndirectBranch->addDestination(BB); return llvm::BlockAddress::get(CurFn, BB); } llvm::BasicBlock *CodeGenFunction::GetIndirectGotoBlock() { // If we already made the indirect branch for indirect goto, return its block. if (IndirectBranch) return IndirectBranch->getParent(); CGBuilderTy TmpBuilder(createBasicBlock("indirectgoto")); // Create the PHI node that indirect gotos will add entries to. llvm::Value *DestVal = TmpBuilder.CreatePHI(Int8PtrTy, 0, "indirect.goto.dest"); // Create the indirect branch instruction. IndirectBranch = TmpBuilder.CreateIndirectBr(DestVal); return IndirectBranch->getParent(); } /// Computes the length of an array in elements, as well as the base /// element type and a properly-typed first element pointer. llvm::Value *CodeGenFunction::emitArrayLength(const ArrayType *origArrayType, QualType &baseType, llvm::Value *&addr) { const ArrayType *arrayType = origArrayType; // If it's a VLA, we have to load the stored size. Note that // this is the size of the VLA in bytes, not its size in elements. llvm::Value *numVLAElements = nullptr; if (isa(arrayType)) { numVLAElements = getVLASize(cast(arrayType)).first; // Walk into all VLAs. This doesn't require changes to addr, // which has type T* where T is the first non-VLA element type. do { QualType elementType = arrayType->getElementType(); arrayType = getContext().getAsArrayType(elementType); // If we only have VLA components, 'addr' requires no adjustment. if (!arrayType) { baseType = elementType; return numVLAElements; } } while (isa(arrayType)); // We get out here only if we find a constant array type // inside the VLA. } // We have some number of constant-length arrays, so addr should // have LLVM type [M x [N x [...]]]*. Build a GEP that walks // down to the first element of addr. SmallVector gepIndices; // GEP down to the array type. llvm::ConstantInt *zero = Builder.getInt32(0); gepIndices.push_back(zero); uint64_t countFromCLAs = 1; QualType eltType; llvm::ArrayType *llvmArrayType = dyn_cast( cast(addr->getType())->getElementType()); while (llvmArrayType) { assert(isa(arrayType)); assert(cast(arrayType)->getSize().getZExtValue() == llvmArrayType->getNumElements()); gepIndices.push_back(zero); countFromCLAs *= llvmArrayType->getNumElements(); eltType = arrayType->getElementType(); llvmArrayType = dyn_cast(llvmArrayType->getElementType()); arrayType = getContext().getAsArrayType(arrayType->getElementType()); assert((!llvmArrayType || arrayType) && "LLVM and Clang types are out-of-synch"); } if (arrayType) { // From this point onwards, the Clang array type has been emitted // as some other type (probably a packed struct). Compute the array // size, and just emit the 'begin' expression as a bitcast. while (arrayType) { countFromCLAs *= cast(arrayType)->getSize().getZExtValue(); eltType = arrayType->getElementType(); arrayType = getContext().getAsArrayType(eltType); } unsigned AddressSpace = addr->getType()->getPointerAddressSpace(); llvm::Type *BaseType = ConvertType(eltType)->getPointerTo(AddressSpace); addr = Builder.CreateBitCast(addr, BaseType, "array.begin"); } else { // Create the actual GEP. addr = Builder.CreateInBoundsGEP(addr, gepIndices, "array.begin"); } baseType = eltType; llvm::Value *numElements = llvm::ConstantInt::get(SizeTy, countFromCLAs); // If we had any VLA dimensions, factor them in. if (numVLAElements) numElements = Builder.CreateNUWMul(numVLAElements, numElements); return numElements; } std::pair CodeGenFunction::getVLASize(QualType type) { const VariableArrayType *vla = getContext().getAsVariableArrayType(type); assert(vla && "type was not a variable array type!"); return getVLASize(vla); } std::pair CodeGenFunction::getVLASize(const VariableArrayType *type) { // The number of elements so far; always size_t. llvm::Value *numElements = nullptr; QualType elementType; do { elementType = type->getElementType(); llvm::Value *vlaSize = VLASizeMap[type->getSizeExpr()]; assert(vlaSize && "no size for VLA!"); assert(vlaSize->getType() == SizeTy); if (!numElements) { numElements = vlaSize; } else { // It's undefined behavior if this wraps around, so mark it that way. // FIXME: Teach -fsanitize=undefined to trap this. numElements = Builder.CreateNUWMul(numElements, vlaSize); } } while ((type = getContext().getAsVariableArrayType(elementType))); return std::pair(numElements, elementType); } void CodeGenFunction::EmitVariablyModifiedType(QualType type) { assert(type->isVariablyModifiedType() && "Must pass variably modified type to EmitVLASizes!"); EnsureInsertPoint(); // We're going to walk down into the type and look for VLA // expressions. do { assert(type->isVariablyModifiedType()); const Type *ty = type.getTypePtr(); switch (ty->getTypeClass()) { #define TYPE(Class, Base) #define ABSTRACT_TYPE(Class, Base) #define NON_CANONICAL_TYPE(Class, Base) #define DEPENDENT_TYPE(Class, Base) case Type::Class: #define NON_CANONICAL_UNLESS_DEPENDENT_TYPE(Class, Base) #include "clang/AST/TypeNodes.def" llvm_unreachable("unexpected dependent type!"); // These types are never variably-modified. case Type::Builtin: case Type::Complex: case Type::Vector: case Type::ExtVector: case Type::Record: case Type::Enum: case Type::Elaborated: case Type::TemplateSpecialization: case Type::ObjCObject: case Type::ObjCInterface: case Type::ObjCObjectPointer: llvm_unreachable("type class is never variably-modified!"); case Type::Adjusted: type = cast(ty)->getAdjustedType(); break; case Type::Decayed: type = cast(ty)->getPointeeType(); break; case Type::Pointer: type = cast(ty)->getPointeeType(); break; case Type::BlockPointer: type = cast(ty)->getPointeeType(); break; case Type::LValueReference: case Type::RValueReference: type = cast(ty)->getPointeeType(); break; case Type::MemberPointer: type = cast(ty)->getPointeeType(); break; case Type::ConstantArray: case Type::IncompleteArray: // Losing element qualification here is fine. type = cast(ty)->getElementType(); break; case Type::VariableArray: { // Losing element qualification here is fine. const VariableArrayType *vat = cast(ty); // Unknown size indication requires no size computation. // Otherwise, evaluate and record it. if (const Expr *size = vat->getSizeExpr()) { // It's possible that we might have emitted this already, // e.g. with a typedef and a pointer to it. llvm::Value *&entry = VLASizeMap[size]; if (!entry) { llvm::Value *Size = EmitScalarExpr(size); // C11 6.7.6.2p5: // If the size is an expression that is not an integer constant // expression [...] each time it is evaluated it shall have a value // greater than zero. if (SanOpts->VLABound && size->getType()->isSignedIntegerType()) { SanitizerScope SanScope(this); llvm::Value *Zero = llvm::Constant::getNullValue(Size->getType()); llvm::Constant *StaticArgs[] = { EmitCheckSourceLocation(size->getLocStart()), EmitCheckTypeDescriptor(size->getType()) }; EmitCheck(Builder.CreateICmpSGT(Size, Zero), "vla_bound_not_positive", StaticArgs, Size, CRK_Recoverable); } // Always zexting here would be wrong if it weren't // undefined behavior to have a negative bound. entry = Builder.CreateIntCast(Size, SizeTy, /*signed*/ false); } } type = vat->getElementType(); break; } case Type::FunctionProto: case Type::FunctionNoProto: type = cast(ty)->getReturnType(); break; case Type::Paren: case Type::TypeOf: case Type::UnaryTransform: case Type::Attributed: case Type::SubstTemplateTypeParm: case Type::PackExpansion: // Keep walking after single level desugaring. type = type.getSingleStepDesugaredType(getContext()); break; case Type::Typedef: case Type::Decltype: case Type::Auto: // Stop walking: nothing to do. return; case Type::TypeOfExpr: // Stop walking: emit typeof expression. EmitIgnoredExpr(cast(ty)->getUnderlyingExpr()); return; case Type::Atomic: type = cast(ty)->getValueType(); break; } } while (type->isVariablyModifiedType()); } llvm::Value* CodeGenFunction::EmitVAListRef(const Expr* E) { if (getContext().getBuiltinVaListType()->isArrayType()) return EmitScalarExpr(E); return EmitLValue(E).getAddress(); } void CodeGenFunction::EmitDeclRefExprDbgValue(const DeclRefExpr *E, llvm::Constant *Init) { assert (Init && "Invalid DeclRefExpr initializer!"); if (CGDebugInfo *Dbg = getDebugInfo()) if (CGM.getCodeGenOpts().getDebugInfo() >= CodeGenOptions::LimitedDebugInfo) Dbg->EmitGlobalVariable(E->getDecl(), Init); } CodeGenFunction::PeepholeProtection CodeGenFunction::protectFromPeepholes(RValue rvalue) { // At the moment, the only aggressive peephole we do in IR gen // is trunc(zext) folding, but if we add more, we can easily // extend this protection. if (!rvalue.isScalar()) return PeepholeProtection(); llvm::Value *value = rvalue.getScalarVal(); if (!isa(value)) return PeepholeProtection(); // Just make an extra bitcast. assert(HaveInsertPoint()); llvm::Instruction *inst = new llvm::BitCastInst(value, value->getType(), "", Builder.GetInsertBlock()); PeepholeProtection protection; protection.Inst = inst; return protection; } void CodeGenFunction::unprotectFromPeepholes(PeepholeProtection protection) { if (!protection.Inst) return; // In theory, we could try to duplicate the peepholes now, but whatever. protection.Inst->eraseFromParent(); } llvm::Value *CodeGenFunction::EmitAnnotationCall(llvm::Value *AnnotationFn, llvm::Value *AnnotatedVal, StringRef AnnotationStr, SourceLocation Location) { llvm::Value *Args[4] = { AnnotatedVal, Builder.CreateBitCast(CGM.EmitAnnotationString(AnnotationStr), Int8PtrTy), Builder.CreateBitCast(CGM.EmitAnnotationUnit(Location), Int8PtrTy), CGM.EmitAnnotationLineNo(Location) }; return Builder.CreateCall(AnnotationFn, Args); } void CodeGenFunction::EmitVarAnnotations(const VarDecl *D, llvm::Value *V) { assert(D->hasAttr() && "no annotate attribute"); // FIXME We create a new bitcast for every annotation because that's what // llvm-gcc was doing. for (const auto *I : D->specific_attrs()) EmitAnnotationCall(CGM.getIntrinsic(llvm::Intrinsic::var_annotation), Builder.CreateBitCast(V, CGM.Int8PtrTy, V->getName()), I->getAnnotation(), D->getLocation()); } llvm::Value *CodeGenFunction::EmitFieldAnnotations(const FieldDecl *D, llvm::Value *V) { assert(D->hasAttr() && "no annotate attribute"); llvm::Type *VTy = V->getType(); llvm::Value *F = CGM.getIntrinsic(llvm::Intrinsic::ptr_annotation, CGM.Int8PtrTy); for (const auto *I : D->specific_attrs()) { // FIXME Always emit the cast inst so we can differentiate between // annotation on the first field of a struct and annotation on the struct // itself. if (VTy != CGM.Int8PtrTy) V = Builder.Insert(new llvm::BitCastInst(V, CGM.Int8PtrTy)); V = EmitAnnotationCall(F, V, I->getAnnotation(), D->getLocation()); V = Builder.CreateBitCast(V, VTy); } return V; } CodeGenFunction::CGCapturedStmtInfo::~CGCapturedStmtInfo() { } CodeGenFunction::SanitizerScope::SanitizerScope(CodeGenFunction *CGF) : CGF(CGF) { assert(!CGF->IsSanitizerScope); CGF->IsSanitizerScope = true; } CodeGenFunction::SanitizerScope::~SanitizerScope() { CGF->IsSanitizerScope = false; } void CodeGenFunction::InsertHelper(llvm::Instruction *I, const llvm::Twine &Name, llvm::BasicBlock *BB, llvm::BasicBlock::iterator InsertPt) const { LoopStack.InsertHelper(I); if (IsSanitizerScope) { I->setMetadata( CGM.getModule().getMDKindID("nosanitize"), llvm::MDNode::get(CGM.getLLVMContext(), ArrayRef())); } } template void CGBuilderInserter::InsertHelper( llvm::Instruction *I, const llvm::Twine &Name, llvm::BasicBlock *BB, llvm::BasicBlock::iterator InsertPt) const { llvm::IRBuilderDefaultInserter::InsertHelper(I, Name, BB, InsertPt); if (CGF) CGF->InsertHelper(I, Name, BB, InsertPt); } #ifdef NDEBUG #define PreserveNames false #else #define PreserveNames true #endif template void CGBuilderInserter::InsertHelper( llvm::Instruction *I, const llvm::Twine &Name, llvm::BasicBlock *BB, llvm::BasicBlock::iterator InsertPt) const; #undef PreserveNames @ 1.1.1.5.4.1 log @file CodeGenFunction.cpp was added on branch yamt-pagecache on 2014-05-22 16:18:27 +0000 @ text @d1 1612 @ 1.1.1.5.4.2 log @sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments") @ text @a0 1612 //===--- CodeGenFunction.cpp - Emit LLVM Code from ASTs for a Function ----===// // // The LLVM Compiler Infrastructure // // This file is distributed under the University of Illinois Open Source // License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// // // This coordinates the per-function state used while generating code. // //===----------------------------------------------------------------------===// #include "CodeGenFunction.h" #include "CGCUDARuntime.h" #include "CGCXXABI.h" #include "CGDebugInfo.h" #include "CodeGenModule.h" #include "CodeGenPGO.h" #include "TargetInfo.h" #include "clang/AST/ASTContext.h" #include "clang/AST/Decl.h" #include "clang/AST/DeclCXX.h" #include "clang/AST/StmtCXX.h" #include "clang/Basic/TargetInfo.h" #include "clang/CodeGen/CGFunctionInfo.h" #include "clang/Frontend/CodeGenOptions.h" #include "llvm/IR/DataLayout.h" #include "llvm/IR/Intrinsics.h" #include "llvm/IR/MDBuilder.h" #include "llvm/IR/Operator.h" using namespace clang; using namespace CodeGen; CodeGenFunction::CodeGenFunction(CodeGenModule &cgm, bool suppressNewContext) : CodeGenTypeCache(cgm), CGM(cgm), Target(cgm.getTarget()), Builder(cgm.getModule().getContext()), CapturedStmtInfo(0), SanitizePerformTypeCheck(CGM.getSanOpts().Null | CGM.getSanOpts().Alignment | CGM.getSanOpts().ObjectSize | CGM.getSanOpts().Vptr), SanOpts(&CGM.getSanOpts()), AutoreleaseResult(false), BlockInfo(0), BlockPointer(0), LambdaThisCaptureField(0), NormalCleanupDest(0), NextCleanupDestIndex(1), FirstBlockInfo(0), EHResumeBlock(0), ExceptionSlot(0), EHSelectorSlot(0), DebugInfo(CGM.getModuleDebugInfo()), DisableDebugInfo(false), DidCallStackSave(false), IndirectBranch(0), PGO(cgm), SwitchInsn(0), SwitchWeights(0), CaseRangeBlock(0), UnreachableBlock(0), NumReturnExprs(0), NumSimpleReturnExprs(0), CXXABIThisDecl(0), CXXABIThisValue(0), CXXThisValue(0), CXXDefaultInitExprThis(0), CXXStructorImplicitParamDecl(0), CXXStructorImplicitParamValue(0), OutermostConditional(0), CurLexicalScope(0), TerminateLandingPad(0), TerminateHandler(0), TrapBB(0) { if (!suppressNewContext) CGM.getCXXABI().getMangleContext().startNewFunction(); llvm::FastMathFlags FMF; if (CGM.getLangOpts().FastMath) FMF.setUnsafeAlgebra(); if (CGM.getLangOpts().FiniteMathOnly) { FMF.setNoNaNs(); FMF.setNoInfs(); } Builder.SetFastMathFlags(FMF); } CodeGenFunction::~CodeGenFunction() { assert(LifetimeExtendedCleanupStack.empty() && "failed to emit a cleanup"); // If there are any unclaimed block infos, go ahead and destroy them // now. This can happen if IR-gen gets clever and skips evaluating // something. if (FirstBlockInfo) destroyBlockInfos(FirstBlockInfo); } llvm::Type *CodeGenFunction::ConvertTypeForMem(QualType T) { return CGM.getTypes().ConvertTypeForMem(T); } llvm::Type *CodeGenFunction::ConvertType(QualType T) { return CGM.getTypes().ConvertType(T); } TypeEvaluationKind CodeGenFunction::getEvaluationKind(QualType type) { type = type.getCanonicalType(); while (true) { switch (type->getTypeClass()) { #define TYPE(name, parent) #define ABSTRACT_TYPE(name, parent) #define NON_CANONICAL_TYPE(name, parent) case Type::name: #define DEPENDENT_TYPE(name, parent) case Type::name: #define NON_CANONICAL_UNLESS_DEPENDENT_TYPE(name, parent) case Type::name: #include "clang/AST/TypeNodes.def" llvm_unreachable("non-canonical or dependent type in IR-generation"); case Type::Auto: llvm_unreachable("undeduced auto type in IR-generation"); // Various scalar types. case Type::Builtin: case Type::Pointer: case Type::BlockPointer: case Type::LValueReference: case Type::RValueReference: case Type::MemberPointer: case Type::Vector: case Type::ExtVector: case Type::FunctionProto: case Type::FunctionNoProto: case Type::Enum: case Type::ObjCObjectPointer: return TEK_Scalar; // Complexes. case Type::Complex: return TEK_Complex; // Arrays, records, and Objective-C objects. case Type::ConstantArray: case Type::IncompleteArray: case Type::VariableArray: case Type::Record: case Type::ObjCObject: case Type::ObjCInterface: return TEK_Aggregate; // We operate on atomic values according to their underlying type. case Type::Atomic: type = cast(type)->getValueType(); continue; } llvm_unreachable("unknown type kind!"); } } void CodeGenFunction::EmitReturnBlock() { // For cleanliness, we try to avoid emitting the return block for // simple cases. llvm::BasicBlock *CurBB = Builder.GetInsertBlock(); if (CurBB) { assert(!CurBB->getTerminator() && "Unexpected terminated block."); // We have a valid insert point, reuse it if it is empty or there are no // explicit jumps to the return block. if (CurBB->empty() || ReturnBlock.getBlock()->use_empty()) { ReturnBlock.getBlock()->replaceAllUsesWith(CurBB); delete ReturnBlock.getBlock(); } else EmitBlock(ReturnBlock.getBlock()); return; } // Otherwise, if the return block is the target of a single direct // branch then we can just put the code in that block instead. This // cleans up functions which started with a unified return block. if (ReturnBlock.getBlock()->hasOneUse()) { llvm::BranchInst *BI = dyn_cast(*ReturnBlock.getBlock()->use_begin()); if (BI && BI->isUnconditional() && BI->getSuccessor(0) == ReturnBlock.getBlock()) { // Reset insertion point, including debug location, and delete the // branch. This is really subtle and only works because the next change // in location will hit the caching in CGDebugInfo::EmitLocation and not // override this. Builder.SetCurrentDebugLocation(BI->getDebugLoc()); Builder.SetInsertPoint(BI->getParent()); BI->eraseFromParent(); delete ReturnBlock.getBlock(); return; } } // FIXME: We are at an unreachable point, there is no reason to emit the block // unless it has uses. However, we still need a place to put the debug // region.end for now. EmitBlock(ReturnBlock.getBlock()); } static void EmitIfUsed(CodeGenFunction &CGF, llvm::BasicBlock *BB) { if (!BB) return; if (!BB->use_empty()) return CGF.CurFn->getBasicBlockList().push_back(BB); delete BB; } void CodeGenFunction::FinishFunction(SourceLocation EndLoc) { assert(BreakContinueStack.empty() && "mismatched push/pop in break/continue stack!"); bool OnlySimpleReturnStmts = NumSimpleReturnExprs > 0 && NumSimpleReturnExprs == NumReturnExprs && ReturnBlock.getBlock()->use_empty(); // Usually the return expression is evaluated before the cleanup // code. If the function contains only a simple return statement, // such as a constant, the location before the cleanup code becomes // the last useful breakpoint in the function, because the simple // return expression will be evaluated after the cleanup code. To be // safe, set the debug location for cleanup code to the location of // the return statement. Otherwise the cleanup code should be at the // end of the function's lexical scope. // // If there are multiple branches to the return block, the branch // instructions will get the location of the return statements and // all will be fine. if (CGDebugInfo *DI = getDebugInfo()) { if (OnlySimpleReturnStmts) DI->EmitLocation(Builder, LastStopPoint); else DI->EmitLocation(Builder, EndLoc); } // Pop any cleanups that might have been associated with the // parameters. Do this in whatever block we're currently in; it's // important to do this before we enter the return block or return // edges will be *really* confused. bool EmitRetDbgLoc = true; if (EHStack.stable_begin() != PrologueCleanupDepth) { PopCleanupBlocks(PrologueCleanupDepth); // Make sure the line table doesn't jump back into the body for // the ret after it's been at EndLoc. EmitRetDbgLoc = false; if (CGDebugInfo *DI = getDebugInfo()) if (OnlySimpleReturnStmts) DI->EmitLocation(Builder, EndLoc); } // Emit function epilog (to return). EmitReturnBlock(); if (ShouldInstrumentFunction()) EmitFunctionInstrumentation("__cyg_profile_func_exit"); // Emit debug descriptor for function end. if (CGDebugInfo *DI = getDebugInfo()) { DI->EmitFunctionEnd(Builder); } EmitFunctionEpilog(*CurFnInfo, EmitRetDbgLoc, EndLoc); EmitEndEHSpec(CurCodeDecl); assert(EHStack.empty() && "did not remove all scopes from cleanup stack!"); // If someone did an indirect goto, emit the indirect goto block at the end of // the function. if (IndirectBranch) { EmitBlock(IndirectBranch->getParent()); Builder.ClearInsertionPoint(); } // Remove the AllocaInsertPt instruction, which is just a convenience for us. llvm::Instruction *Ptr = AllocaInsertPt; AllocaInsertPt = 0; Ptr->eraseFromParent(); // If someone took the address of a label but never did an indirect goto, we // made a zero entry PHI node, which is illegal, zap it now. if (IndirectBranch) { llvm::PHINode *PN = cast(IndirectBranch->getAddress()); if (PN->getNumIncomingValues() == 0) { PN->replaceAllUsesWith(llvm::UndefValue::get(PN->getType())); PN->eraseFromParent(); } } EmitIfUsed(*this, EHResumeBlock); EmitIfUsed(*this, TerminateLandingPad); EmitIfUsed(*this, TerminateHandler); EmitIfUsed(*this, UnreachableBlock); if (CGM.getCodeGenOpts().EmitDeclMetadata) EmitDeclMetadata(); for (SmallVectorImpl >::iterator I = DeferredReplacements.begin(), E = DeferredReplacements.end(); I != E; ++I) { I->first->replaceAllUsesWith(I->second); I->first->eraseFromParent(); } } /// ShouldInstrumentFunction - Return true if the current function should be /// instrumented with __cyg_profile_func_* calls bool CodeGenFunction::ShouldInstrumentFunction() { if (!CGM.getCodeGenOpts().InstrumentFunctions) return false; if (!CurFuncDecl || CurFuncDecl->hasAttr()) return false; return true; } /// EmitFunctionInstrumentation - Emit LLVM code to call the specified /// instrumentation function with the current function and the call site, if /// function instrumentation is enabled. void CodeGenFunction::EmitFunctionInstrumentation(const char *Fn) { // void __cyg_profile_func_{enter,exit} (void *this_fn, void *call_site); llvm::PointerType *PointerTy = Int8PtrTy; llvm::Type *ProfileFuncArgs[] = { PointerTy, PointerTy }; llvm::FunctionType *FunctionTy = llvm::FunctionType::get(VoidTy, ProfileFuncArgs, false); llvm::Constant *F = CGM.CreateRuntimeFunction(FunctionTy, Fn); llvm::CallInst *CallSite = Builder.CreateCall( CGM.getIntrinsic(llvm::Intrinsic::returnaddress), llvm::ConstantInt::get(Int32Ty, 0), "callsite"); llvm::Value *args[] = { llvm::ConstantExpr::getBitCast(CurFn, PointerTy), CallSite }; EmitNounwindRuntimeCall(F, args); } void CodeGenFunction::EmitMCountInstrumentation() { llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy, false); llvm::Constant *MCountFn = CGM.CreateRuntimeFunction(FTy, getTarget().getMCountName()); EmitNounwindRuntimeCall(MCountFn); } // OpenCL v1.2 s5.6.4.6 allows the compiler to store kernel argument // information in the program executable. The argument information stored // includes the argument name, its type, the address and access qualifiers used. static void GenOpenCLArgMetadata(const FunctionDecl *FD, llvm::Function *Fn, CodeGenModule &CGM,llvm::LLVMContext &Context, SmallVector &kernelMDArgs, CGBuilderTy& Builder, ASTContext &ASTCtx) { // Create MDNodes that represent the kernel arg metadata. // Each MDNode is a list in the form of "key", N number of values which is // the same number of values as their are kernel arguments. // MDNode for the kernel argument address space qualifiers. SmallVector addressQuals; addressQuals.push_back(llvm::MDString::get(Context, "kernel_arg_addr_space")); // MDNode for the kernel argument access qualifiers (images only). SmallVector accessQuals; accessQuals.push_back(llvm::MDString::get(Context, "kernel_arg_access_qual")); // MDNode for the kernel argument type names. SmallVector argTypeNames; argTypeNames.push_back(llvm::MDString::get(Context, "kernel_arg_type")); // MDNode for the kernel argument type qualifiers. SmallVector argTypeQuals; argTypeQuals.push_back(llvm::MDString::get(Context, "kernel_arg_type_qual")); // MDNode for the kernel argument names. SmallVector argNames; argNames.push_back(llvm::MDString::get(Context, "kernel_arg_name")); for (unsigned i = 0, e = FD->getNumParams(); i != e; ++i) { const ParmVarDecl *parm = FD->getParamDecl(i); QualType ty = parm->getType(); std::string typeQuals; if (ty->isPointerType()) { QualType pointeeTy = ty->getPointeeType(); // Get address qualifier. addressQuals.push_back(Builder.getInt32(ASTCtx.getTargetAddressSpace( pointeeTy.getAddressSpace()))); // Get argument type name. std::string typeName = pointeeTy.getUnqualifiedType().getAsString() + "*"; // Turn "unsigned type" to "utype" std::string::size_type pos = typeName.find("unsigned"); if (pos != std::string::npos) typeName.erase(pos+1, 8); argTypeNames.push_back(llvm::MDString::get(Context, typeName)); // Get argument type qualifiers: if (ty.isRestrictQualified()) typeQuals = "restrict"; if (pointeeTy.isConstQualified() || (pointeeTy.getAddressSpace() == LangAS::opencl_constant)) typeQuals += typeQuals.empty() ? "const" : " const"; if (pointeeTy.isVolatileQualified()) typeQuals += typeQuals.empty() ? "volatile" : " volatile"; } else { uint32_t AddrSpc = 0; if (ty->isImageType()) AddrSpc = CGM.getContext().getTargetAddressSpace(LangAS::opencl_global); addressQuals.push_back(Builder.getInt32(AddrSpc)); // Get argument type name. std::string typeName = ty.getUnqualifiedType().getAsString(); // Turn "unsigned type" to "utype" std::string::size_type pos = typeName.find("unsigned"); if (pos != std::string::npos) typeName.erase(pos+1, 8); argTypeNames.push_back(llvm::MDString::get(Context, typeName)); // Get argument type qualifiers: if (ty.isConstQualified()) typeQuals = "const"; if (ty.isVolatileQualified()) typeQuals += typeQuals.empty() ? "volatile" : " volatile"; } argTypeQuals.push_back(llvm::MDString::get(Context, typeQuals)); // Get image access qualifier: if (ty->isImageType()) { const OpenCLImageAccessAttr *A = parm->getAttr(); if (A && A->isWriteOnly()) accessQuals.push_back(llvm::MDString::get(Context, "write_only")); else accessQuals.push_back(llvm::MDString::get(Context, "read_only")); // FIXME: what about read_write? } else accessQuals.push_back(llvm::MDString::get(Context, "none")); // Get argument name. argNames.push_back(llvm::MDString::get(Context, parm->getName())); } kernelMDArgs.push_back(llvm::MDNode::get(Context, addressQuals)); kernelMDArgs.push_back(llvm::MDNode::get(Context, accessQuals)); kernelMDArgs.push_back(llvm::MDNode::get(Context, argTypeNames)); kernelMDArgs.push_back(llvm::MDNode::get(Context, argTypeQuals)); kernelMDArgs.push_back(llvm::MDNode::get(Context, argNames)); } void CodeGenFunction::EmitOpenCLKernelMetadata(const FunctionDecl *FD, llvm::Function *Fn) { if (!FD->hasAttr()) return; llvm::LLVMContext &Context = getLLVMContext(); SmallVector kernelMDArgs; kernelMDArgs.push_back(Fn); if (CGM.getCodeGenOpts().EmitOpenCLArgMetadata) GenOpenCLArgMetadata(FD, Fn, CGM, Context, kernelMDArgs, Builder, getContext()); if (const VecTypeHintAttr *A = FD->getAttr()) { QualType hintQTy = A->getTypeHint(); const ExtVectorType *hintEltQTy = hintQTy->getAs(); bool isSignedInteger = hintQTy->isSignedIntegerType() || (hintEltQTy && hintEltQTy->getElementType()->isSignedIntegerType()); llvm::Value *attrMDArgs[] = { llvm::MDString::get(Context, "vec_type_hint"), llvm::UndefValue::get(CGM.getTypes().ConvertType(A->getTypeHint())), llvm::ConstantInt::get( llvm::IntegerType::get(Context, 32), llvm::APInt(32, (uint64_t)(isSignedInteger ? 1 : 0))) }; kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs)); } if (const WorkGroupSizeHintAttr *A = FD->getAttr()) { llvm::Value *attrMDArgs[] = { llvm::MDString::get(Context, "work_group_size_hint"), Builder.getInt32(A->getXDim()), Builder.getInt32(A->getYDim()), Builder.getInt32(A->getZDim()) }; kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs)); } if (const ReqdWorkGroupSizeAttr *A = FD->getAttr()) { llvm::Value *attrMDArgs[] = { llvm::MDString::get(Context, "reqd_work_group_size"), Builder.getInt32(A->getXDim()), Builder.getInt32(A->getYDim()), Builder.getInt32(A->getZDim()) }; kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs)); } llvm::MDNode *kernelMDNode = llvm::MDNode::get(Context, kernelMDArgs); llvm::NamedMDNode *OpenCLKernelMetadata = CGM.getModule().getOrInsertNamedMetadata("opencl.kernels"); OpenCLKernelMetadata->addOperand(kernelMDNode); } void CodeGenFunction::StartFunction(GlobalDecl GD, QualType RetTy, llvm::Function *Fn, const CGFunctionInfo &FnInfo, const FunctionArgList &Args, SourceLocation StartLoc) { const Decl *D = GD.getDecl(); DidCallStackSave = false; CurCodeDecl = D; CurFuncDecl = (D ? D->getNonClosureContext() : 0); FnRetTy = RetTy; CurFn = Fn; CurFnInfo = &FnInfo; assert(CurFn->isDeclaration() && "Function already has body?"); if (CGM.getSanitizerBlacklist().isIn(*Fn)) { SanOpts = &SanitizerOptions::Disabled; SanitizePerformTypeCheck = false; } // Pass inline keyword to optimizer if it appears explicitly on any // declaration. Also, in the case of -fno-inline attach NoInline // attribute to all function that are not marked AlwaysInline. if (const FunctionDecl *FD = dyn_cast_or_null(D)) { if (!CGM.getCodeGenOpts().NoInline) { for (FunctionDecl::redecl_iterator RI = FD->redecls_begin(), RE = FD->redecls_end(); RI != RE; ++RI) if (RI->isInlineSpecified()) { Fn->addFnAttr(llvm::Attribute::InlineHint); break; } } else if (!FD->hasAttr()) Fn->addFnAttr(llvm::Attribute::NoInline); } if (getLangOpts().OpenCL) { // Add metadata for a kernel function. if (const FunctionDecl *FD = dyn_cast_or_null(D)) EmitOpenCLKernelMetadata(FD, Fn); } // If we are checking function types, emit a function type signature as // prefix data. if (getLangOpts().CPlusPlus && SanOpts->Function) { if (const FunctionDecl *FD = dyn_cast_or_null(D)) { if (llvm::Constant *PrefixSig = CGM.getTargetCodeGenInfo().getUBSanFunctionSignature(CGM)) { llvm::Constant *FTRTTIConst = CGM.GetAddrOfRTTIDescriptor(FD->getType(), /*ForEH=*/true); llvm::Constant *PrefixStructElems[] = { PrefixSig, FTRTTIConst }; llvm::Constant *PrefixStructConst = llvm::ConstantStruct::getAnon(PrefixStructElems, /*Packed=*/true); Fn->setPrefixData(PrefixStructConst); } } } llvm::BasicBlock *EntryBB = createBasicBlock("entry", CurFn); // Create a marker to make it easy to insert allocas into the entryblock // later. Don't create this with the builder, because we don't want it // folded. llvm::Value *Undef = llvm::UndefValue::get(Int32Ty); AllocaInsertPt = new llvm::BitCastInst(Undef, Int32Ty, "", EntryBB); if (Builder.isNamePreserving()) AllocaInsertPt->setName("allocapt"); ReturnBlock = getJumpDestInCurrentScope("return"); Builder.SetInsertPoint(EntryBB); // Emit subprogram debug descriptor. if (CGDebugInfo *DI = getDebugInfo()) { SmallVector ArgTypes; for (FunctionArgList::const_iterator i = Args.begin(), e = Args.end(); i != e; ++i) { ArgTypes.push_back((*i)->getType()); } QualType FnType = getContext().getFunctionType(RetTy, ArgTypes, FunctionProtoType::ExtProtoInfo()); DI->setLocation(StartLoc); DI->EmitFunctionStart(GD, FnType, CurFn, Builder); } if (ShouldInstrumentFunction()) EmitFunctionInstrumentation("__cyg_profile_func_enter"); if (CGM.getCodeGenOpts().InstrumentForProfiling) EmitMCountInstrumentation(); PGO.assignRegionCounters(GD); if (CGM.getPGOData() && D) { // Turn on InlineHint attribute for hot functions. if (CGM.getPGOData()->isHotFunction(CGM.getMangledName(GD))) Fn->addFnAttr(llvm::Attribute::InlineHint); // Turn on Cold attribute for cold functions. else if (CGM.getPGOData()->isColdFunction(CGM.getMangledName(GD))) Fn->addFnAttr(llvm::Attribute::Cold); } if (RetTy->isVoidType()) { // Void type; nothing to return. ReturnValue = 0; } else if (CurFnInfo->getReturnInfo().getKind() == ABIArgInfo::Indirect && !hasScalarEvaluationKind(CurFnInfo->getReturnType())) { // Indirect aggregate return; emit returned value directly into sret slot. // This reduces code size, and affects correctness in C++. ReturnValue = CurFn->arg_begin(); } else if (CurFnInfo->getReturnInfo().getKind() == ABIArgInfo::InAlloca && !hasScalarEvaluationKind(CurFnInfo->getReturnType())) { // Load the sret pointer from the argument struct and return into that. unsigned Idx = CurFnInfo->getReturnInfo().getInAllocaFieldIndex(); llvm::Function::arg_iterator EI = CurFn->arg_end(); --EI; llvm::Value *Addr = Builder.CreateStructGEP(EI, Idx); ReturnValue = Builder.CreateLoad(Addr, "agg.result"); } else { ReturnValue = CreateIRTemp(RetTy, "retval"); // Tell the epilog emitter to autorelease the result. We do this // now so that various specialized functions can suppress it // during their IR-generation. if (getLangOpts().ObjCAutoRefCount && !CurFnInfo->isReturnsRetained() && RetTy->isObjCRetainableType()) AutoreleaseResult = true; } EmitStartEHSpec(CurCodeDecl); PrologueCleanupDepth = EHStack.stable_begin(); EmitFunctionProlog(*CurFnInfo, CurFn, Args); if (D && isa(D) && cast(D)->isInstance()) { CGM.getCXXABI().EmitInstanceFunctionProlog(*this); const CXXMethodDecl *MD = cast(D); if (MD->getParent()->isLambda() && MD->getOverloadedOperator() == OO_Call) { // We're in a lambda; figure out the captures. MD->getParent()->getCaptureFields(LambdaCaptureFields, LambdaThisCaptureField); if (LambdaThisCaptureField) { // If this lambda captures this, load it. LValue ThisLValue = EmitLValueForLambdaField(LambdaThisCaptureField); CXXThisValue = EmitLoadOfLValue(ThisLValue, SourceLocation()).getScalarVal(); } } else { // Not in a lambda; just use 'this' from the method. // FIXME: Should we generate a new load for each use of 'this'? The // fast register allocator would be happier... CXXThisValue = CXXABIThisValue; } } // If any of the arguments have a variably modified type, make sure to // emit the type size. for (FunctionArgList::const_iterator i = Args.begin(), e = Args.end(); i != e; ++i) { const VarDecl *VD = *i; // Dig out the type as written from ParmVarDecls; it's unclear whether // the standard (C99 6.9.1p10) requires this, but we're following the // precedent set by gcc. QualType Ty; if (const ParmVarDecl *PVD = dyn_cast(VD)) Ty = PVD->getOriginalType(); else Ty = VD->getType(); if (Ty->isVariablyModifiedType()) EmitVariablyModifiedType(Ty); } // Emit a location at the end of the prologue. if (CGDebugInfo *DI = getDebugInfo()) DI->EmitLocation(Builder, StartLoc); } void CodeGenFunction::EmitFunctionBody(FunctionArgList &Args, const Stmt *Body) { RegionCounter Cnt = getPGORegionCounter(Body); Cnt.beginRegion(Builder); if (const CompoundStmt *S = dyn_cast(Body)) EmitCompoundStmtWithoutScope(*S); else EmitStmt(Body); } /// When instrumenting to collect profile data, the counts for some blocks /// such as switch cases need to not include the fall-through counts, so /// emit a branch around the instrumentation code. When not instrumenting, /// this just calls EmitBlock(). void CodeGenFunction::EmitBlockWithFallThrough(llvm::BasicBlock *BB, RegionCounter &Cnt) { llvm::BasicBlock *SkipCountBB = 0; if (HaveInsertPoint() && CGM.getCodeGenOpts().ProfileInstrGenerate) { // When instrumenting for profiling, the fallthrough to certain // statements needs to skip over the instrumentation code so that we // get an accurate count. SkipCountBB = createBasicBlock("skipcount"); EmitBranch(SkipCountBB); } EmitBlock(BB); Cnt.beginRegion(Builder, /*AddIncomingFallThrough=*/true); if (SkipCountBB) EmitBlock(SkipCountBB); } /// Tries to mark the given function nounwind based on the /// non-existence of any throwing calls within it. We believe this is /// lightweight enough to do at -O0. static void TryMarkNoThrow(llvm::Function *F) { // LLVM treats 'nounwind' on a function as part of the type, so we // can't do this on functions that can be overwritten. if (F->mayBeOverridden()) return; for (llvm::Function::iterator FI = F->begin(), FE = F->end(); FI != FE; ++FI) for (llvm::BasicBlock::iterator BI = FI->begin(), BE = FI->end(); BI != BE; ++BI) if (llvm::CallInst *Call = dyn_cast(&*BI)) { if (!Call->doesNotThrow()) return; } else if (isa(&*BI)) { return; } F->setDoesNotThrow(); } static void EmitSizedDeallocationFunction(CodeGenFunction &CGF, const FunctionDecl *UnsizedDealloc) { // This is a weak discardable definition of the sized deallocation function. CGF.CurFn->setLinkage(llvm::Function::LinkOnceAnyLinkage); // Call the unsized deallocation function and forward the first argument // unchanged. llvm::Constant *Unsized = CGF.CGM.GetAddrOfFunction(UnsizedDealloc); CGF.Builder.CreateCall(Unsized, &*CGF.CurFn->arg_begin()); } void CodeGenFunction::GenerateCode(GlobalDecl GD, llvm::Function *Fn, const CGFunctionInfo &FnInfo) { const FunctionDecl *FD = cast(GD.getDecl()); // Check if we should generate debug info for this function. if (FD->hasAttr()) DebugInfo = NULL; // disable debug info indefinitely for this function FunctionArgList Args; QualType ResTy = FD->getReturnType(); CurGD = GD; const CXXMethodDecl *MD = dyn_cast(FD); if (MD && MD->isInstance()) { if (CGM.getCXXABI().HasThisReturn(GD)) ResTy = MD->getThisType(getContext()); CGM.getCXXABI().buildThisParam(*this, Args); } for (unsigned i = 0, e = FD->getNumParams(); i != e; ++i) Args.push_back(FD->getParamDecl(i)); if (MD && (isa(MD) || isa(MD))) CGM.getCXXABI().addImplicitStructorParams(*this, ResTy, Args); SourceRange BodyRange; if (Stmt *Body = FD->getBody()) BodyRange = Body->getSourceRange(); CurEHLocation = BodyRange.getEnd(); // Emit the standard function prologue. StartFunction(GD, ResTy, Fn, FnInfo, Args, BodyRange.getBegin()); // Generate the body of the function. if (isa(FD)) EmitDestructorBody(Args); else if (isa(FD)) EmitConstructorBody(Args); else if (getLangOpts().CUDA && !CGM.getCodeGenOpts().CUDAIsDevice && FD->hasAttr()) CGM.getCUDARuntime().EmitDeviceStubBody(*this, Args); else if (isa(FD) && cast(FD)->isLambdaToBlockPointerConversion()) { // The lambda conversion to block pointer is special; the semantics can't be // expressed in the AST, so IRGen needs to special-case it. EmitLambdaToBlockPointerBody(Args); } else if (isa(FD) && cast(FD)->isLambdaStaticInvoker()) { // The lambda static invoker function is special, because it forwards or // clones the body of the function call operator (but is actually static). EmitLambdaStaticInvokeFunction(cast(FD)); } else if (FD->isDefaulted() && isa(FD) && (cast(FD)->isCopyAssignmentOperator() || cast(FD)->isMoveAssignmentOperator())) { // Implicit copy-assignment gets the same special treatment as implicit // copy-constructors. emitImplicitAssignmentOperatorBody(Args); } else if (Stmt *Body = FD->getBody()) { EmitFunctionBody(Args, Body); } else if (FunctionDecl *UnsizedDealloc = FD->getCorrespondingUnsizedGlobalDeallocationFunction()) { // Global sized deallocation functions get an implicit weak definition if // they don't have an explicit definition. EmitSizedDeallocationFunction(*this, UnsizedDealloc); } else llvm_unreachable("no definition for emitted function"); // C++11 [stmt.return]p2: // Flowing off the end of a function [...] results in undefined behavior in // a value-returning function. // C11 6.9.1p12: // If the '}' that terminates a function is reached, and the value of the // function call is used by the caller, the behavior is undefined. if (getLangOpts().CPlusPlus && !FD->hasImplicitReturnZero() && !FD->getReturnType()->isVoidType() && Builder.GetInsertBlock()) { if (SanOpts->Return) EmitCheck(Builder.getFalse(), "missing_return", EmitCheckSourceLocation(FD->getLocation()), ArrayRef(), CRK_Unrecoverable); else if (CGM.getCodeGenOpts().OptimizationLevel == 0) Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::trap)); Builder.CreateUnreachable(); Builder.ClearInsertionPoint(); } // Emit the standard function epilogue. FinishFunction(BodyRange.getEnd()); // If we haven't marked the function nothrow through other means, do // a quick pass now to see if we can. if (!CurFn->doesNotThrow()) TryMarkNoThrow(CurFn); PGO.emitWriteoutFunction(CurGD); PGO.destroyRegionCounters(); } /// ContainsLabel - Return true if the statement contains a label in it. If /// this statement is not executed normally, it not containing a label means /// that we can just remove the code. bool CodeGenFunction::ContainsLabel(const Stmt *S, bool IgnoreCaseStmts) { // Null statement, not a label! if (S == 0) return false; // If this is a label, we have to emit the code, consider something like: // if (0) { ... foo: bar(); } goto foo; // // TODO: If anyone cared, we could track __label__'s, since we know that you // can't jump to one from outside their declared region. if (isa(S)) return true; // If this is a case/default statement, and we haven't seen a switch, we have // to emit the code. if (isa(S) && !IgnoreCaseStmts) return true; // If this is a switch statement, we want to ignore cases below it. if (isa(S)) IgnoreCaseStmts = true; // Scan subexpressions for verboten labels. for (Stmt::const_child_range I = S->children(); I; ++I) if (ContainsLabel(*I, IgnoreCaseStmts)) return true; return false; } /// containsBreak - Return true if the statement contains a break out of it. /// If the statement (recursively) contains a switch or loop with a break /// inside of it, this is fine. bool CodeGenFunction::containsBreak(const Stmt *S) { // Null statement, not a label! if (S == 0) return false; // If this is a switch or loop that defines its own break scope, then we can // include it and anything inside of it. if (isa(S) || isa(S) || isa(S) || isa(S)) return false; if (isa(S)) return true; // Scan subexpressions for verboten breaks. for (Stmt::const_child_range I = S->children(); I; ++I) if (containsBreak(*I)) return true; return false; } /// ConstantFoldsToSimpleInteger - If the specified expression does not fold /// to a constant, or if it does but contains a label, return false. If it /// constant folds return true and set the boolean result in Result. bool CodeGenFunction::ConstantFoldsToSimpleInteger(const Expr *Cond, bool &ResultBool) { llvm::APSInt ResultInt; if (!ConstantFoldsToSimpleInteger(Cond, ResultInt)) return false; ResultBool = ResultInt.getBoolValue(); return true; } /// ConstantFoldsToSimpleInteger - If the specified expression does not fold /// to a constant, or if it does but contains a label, return false. If it /// constant folds return true and set the folded value. bool CodeGenFunction:: ConstantFoldsToSimpleInteger(const Expr *Cond, llvm::APSInt &ResultInt) { // FIXME: Rename and handle conversion of other evaluatable things // to bool. llvm::APSInt Int; if (!Cond->EvaluateAsInt(Int, getContext())) return false; // Not foldable, not integer or not fully evaluatable. if (CodeGenFunction::ContainsLabel(Cond)) return false; // Contains a label. ResultInt = Int; return true; } /// EmitBranchOnBoolExpr - Emit a branch on a boolean condition (e.g. for an if /// statement) to the specified blocks. Based on the condition, this might try /// to simplify the codegen of the conditional based on the branch. /// void CodeGenFunction::EmitBranchOnBoolExpr(const Expr *Cond, llvm::BasicBlock *TrueBlock, llvm::BasicBlock *FalseBlock, uint64_t TrueCount) { Cond = Cond->IgnoreParens(); if (const BinaryOperator *CondBOp = dyn_cast(Cond)) { // Handle X && Y in a condition. if (CondBOp->getOpcode() == BO_LAnd) { RegionCounter Cnt = getPGORegionCounter(CondBOp); // If we have "1 && X", simplify the code. "0 && X" would have constant // folded if the case was simple enough. bool ConstantBool = false; if (ConstantFoldsToSimpleInteger(CondBOp->getLHS(), ConstantBool) && ConstantBool) { // br(1 && X) -> br(X). Cnt.beginRegion(Builder); return EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, TrueCount); } // If we have "X && 1", simplify the code to use an uncond branch. // "X && 0" would have been constant folded to 0. if (ConstantFoldsToSimpleInteger(CondBOp->getRHS(), ConstantBool) && ConstantBool) { // br(X && 1) -> br(X). return EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, FalseBlock, TrueCount); } // Emit the LHS as a conditional. If the LHS conditional is false, we // want to jump to the FalseBlock. llvm::BasicBlock *LHSTrue = createBasicBlock("land.lhs.true"); // The counter tells us how often we evaluate RHS, and all of TrueCount // can be propagated to that branch. uint64_t RHSCount = Cnt.getCount(); ConditionalEvaluation eval(*this); EmitBranchOnBoolExpr(CondBOp->getLHS(), LHSTrue, FalseBlock, RHSCount); EmitBlock(LHSTrue); // Any temporaries created here are conditional. Cnt.beginRegion(Builder); eval.begin(*this); EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, TrueCount); eval.end(*this); return; } if (CondBOp->getOpcode() == BO_LOr) { RegionCounter Cnt = getPGORegionCounter(CondBOp); // If we have "0 || X", simplify the code. "1 || X" would have constant // folded if the case was simple enough. bool ConstantBool = false; if (ConstantFoldsToSimpleInteger(CondBOp->getLHS(), ConstantBool) && !ConstantBool) { // br(0 || X) -> br(X). Cnt.beginRegion(Builder); return EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, TrueCount); } // If we have "X || 0", simplify the code to use an uncond branch. // "X || 1" would have been constant folded to 1. if (ConstantFoldsToSimpleInteger(CondBOp->getRHS(), ConstantBool) && !ConstantBool) { // br(X || 0) -> br(X). return EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, FalseBlock, TrueCount); } // Emit the LHS as a conditional. If the LHS conditional is true, we // want to jump to the TrueBlock. llvm::BasicBlock *LHSFalse = createBasicBlock("lor.lhs.false"); // We have the count for entry to the RHS and for the whole expression // being true, so we can divy up True count between the short circuit and // the RHS. uint64_t LHSCount = Cnt.getParentCount() - Cnt.getCount(); uint64_t RHSCount = TrueCount - LHSCount; ConditionalEvaluation eval(*this); EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, LHSFalse, LHSCount); EmitBlock(LHSFalse); // Any temporaries created here are conditional. Cnt.beginRegion(Builder); eval.begin(*this); EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, RHSCount); eval.end(*this); return; } } if (const UnaryOperator *CondUOp = dyn_cast(Cond)) { // br(!x, t, f) -> br(x, f, t) if (CondUOp->getOpcode() == UO_LNot) { // Negate the count. uint64_t FalseCount = PGO.getCurrentRegionCount() - TrueCount; // Negate the condition and swap the destination blocks. return EmitBranchOnBoolExpr(CondUOp->getSubExpr(), FalseBlock, TrueBlock, FalseCount); } } if (const ConditionalOperator *CondOp = dyn_cast(Cond)) { // br(c ? x : y, t, f) -> br(c, br(x, t, f), br(y, t, f)) llvm::BasicBlock *LHSBlock = createBasicBlock("cond.true"); llvm::BasicBlock *RHSBlock = createBasicBlock("cond.false"); RegionCounter Cnt = getPGORegionCounter(CondOp); ConditionalEvaluation cond(*this); EmitBranchOnBoolExpr(CondOp->getCond(), LHSBlock, RHSBlock, Cnt.getCount()); // When computing PGO branch weights, we only know the overall count for // the true block. This code is essentially doing tail duplication of the // naive code-gen, introducing new edges for which counts are not // available. Divide the counts proportionally between the LHS and RHS of // the conditional operator. uint64_t LHSScaledTrueCount = 0; if (TrueCount) { double LHSRatio = Cnt.getCount() / (double) Cnt.getParentCount(); LHSScaledTrueCount = TrueCount * LHSRatio; } cond.begin(*this); EmitBlock(LHSBlock); Cnt.beginRegion(Builder); EmitBranchOnBoolExpr(CondOp->getLHS(), TrueBlock, FalseBlock, LHSScaledTrueCount); cond.end(*this); cond.begin(*this); EmitBlock(RHSBlock); EmitBranchOnBoolExpr(CondOp->getRHS(), TrueBlock, FalseBlock, TrueCount - LHSScaledTrueCount); cond.end(*this); return; } if (const CXXThrowExpr *Throw = dyn_cast(Cond)) { // Conditional operator handling can give us a throw expression as a // condition for a case like: // br(c ? throw x : y, t, f) -> br(c, br(throw x, t, f), br(y, t, f) // Fold this to: // br(c, throw x, br(y, t, f)) EmitCXXThrowExpr(Throw, /*KeepInsertionPoint*/false); return; } // Create branch weights based on the number of times we get here and the // number of times the condition should be true. uint64_t CurrentCount = std::max(PGO.getCurrentRegionCount(), TrueCount); llvm::MDNode *Weights = PGO.createBranchWeights(TrueCount, CurrentCount - TrueCount); // Emit the code with the fully general case. llvm::Value *CondV = EvaluateExprAsBool(Cond); Builder.CreateCondBr(CondV, TrueBlock, FalseBlock, Weights); } /// ErrorUnsupported - Print out an error that codegen doesn't support the /// specified stmt yet. void CodeGenFunction::ErrorUnsupported(const Stmt *S, const char *Type) { CGM.ErrorUnsupported(S, Type); } /// emitNonZeroVLAInit - Emit the "zero" initialization of a /// variable-length array whose elements have a non-zero bit-pattern. /// /// \param baseType the inner-most element type of the array /// \param src - a char* pointing to the bit-pattern for a single /// base element of the array /// \param sizeInChars - the total size of the VLA, in chars static void emitNonZeroVLAInit(CodeGenFunction &CGF, QualType baseType, llvm::Value *dest, llvm::Value *src, llvm::Value *sizeInChars) { std::pair baseSizeAndAlign = CGF.getContext().getTypeInfoInChars(baseType); CGBuilderTy &Builder = CGF.Builder; llvm::Value *baseSizeInChars = llvm::ConstantInt::get(CGF.IntPtrTy, baseSizeAndAlign.first.getQuantity()); llvm::Type *i8p = Builder.getInt8PtrTy(); llvm::Value *begin = Builder.CreateBitCast(dest, i8p, "vla.begin"); llvm::Value *end = Builder.CreateInBoundsGEP(dest, sizeInChars, "vla.end"); llvm::BasicBlock *originBB = CGF.Builder.GetInsertBlock(); llvm::BasicBlock *loopBB = CGF.createBasicBlock("vla-init.loop"); llvm::BasicBlock *contBB = CGF.createBasicBlock("vla-init.cont"); // Make a loop over the VLA. C99 guarantees that the VLA element // count must be nonzero. CGF.EmitBlock(loopBB); llvm::PHINode *cur = Builder.CreatePHI(i8p, 2, "vla.cur"); cur->addIncoming(begin, originBB); // memcpy the individual element bit-pattern. Builder.CreateMemCpy(cur, src, baseSizeInChars, baseSizeAndAlign.second.getQuantity(), /*volatile*/ false); // Go to the next element. llvm::Value *next = Builder.CreateConstInBoundsGEP1_32(cur, 1, "vla.next"); // Leave if that's the end of the VLA. llvm::Value *done = Builder.CreateICmpEQ(next, end, "vla-init.isdone"); Builder.CreateCondBr(done, contBB, loopBB); cur->addIncoming(next, loopBB); CGF.EmitBlock(contBB); } void CodeGenFunction::EmitNullInitialization(llvm::Value *DestPtr, QualType Ty) { // Ignore empty classes in C++. if (getLangOpts().CPlusPlus) { if (const RecordType *RT = Ty->getAs()) { if (cast(RT->getDecl())->isEmpty()) return; } } // Cast the dest ptr to the appropriate i8 pointer type. unsigned DestAS = cast(DestPtr->getType())->getAddressSpace(); llvm::Type *BP = Builder.getInt8PtrTy(DestAS); if (DestPtr->getType() != BP) DestPtr = Builder.CreateBitCast(DestPtr, BP); // Get size and alignment info for this aggregate. std::pair TypeInfo = getContext().getTypeInfoInChars(Ty); CharUnits Size = TypeInfo.first; CharUnits Align = TypeInfo.second; llvm::Value *SizeVal; const VariableArrayType *vla; // Don't bother emitting a zero-byte memset. if (Size.isZero()) { // But note that getTypeInfo returns 0 for a VLA. if (const VariableArrayType *vlaType = dyn_cast_or_null( getContext().getAsArrayType(Ty))) { QualType eltType; llvm::Value *numElts; llvm::tie(numElts, eltType) = getVLASize(vlaType); SizeVal = numElts; CharUnits eltSize = getContext().getTypeSizeInChars(eltType); if (!eltSize.isOne()) SizeVal = Builder.CreateNUWMul(SizeVal, CGM.getSize(eltSize)); vla = vlaType; } else { return; } } else { SizeVal = CGM.getSize(Size); vla = 0; } // If the type contains a pointer to data member we can't memset it to zero. // Instead, create a null constant and copy it to the destination. // TODO: there are other patterns besides zero that we can usefully memset, // like -1, which happens to be the pattern used by member-pointers. if (!CGM.getTypes().isZeroInitializable(Ty)) { // For a VLA, emit a single element, then splat that over the VLA. if (vla) Ty = getContext().getBaseElementType(vla); llvm::Constant *NullConstant = CGM.EmitNullConstant(Ty); llvm::GlobalVariable *NullVariable = new llvm::GlobalVariable(CGM.getModule(), NullConstant->getType(), /*isConstant=*/true, llvm::GlobalVariable::PrivateLinkage, NullConstant, Twine()); llvm::Value *SrcPtr = Builder.CreateBitCast(NullVariable, Builder.getInt8PtrTy()); if (vla) return emitNonZeroVLAInit(*this, Ty, DestPtr, SrcPtr, SizeVal); // Get and call the appropriate llvm.memcpy overload. Builder.CreateMemCpy(DestPtr, SrcPtr, SizeVal, Align.getQuantity(), false); return; } // Otherwise, just memset the whole thing to zero. This is legal // because in LLVM, all default initializers (other than the ones we just // handled above) are guaranteed to have a bit pattern of all zeros. Builder.CreateMemSet(DestPtr, Builder.getInt8(0), SizeVal, Align.getQuantity(), false); } llvm::BlockAddress *CodeGenFunction::GetAddrOfLabel(const LabelDecl *L) { // Make sure that there is a block for the indirect goto. if (IndirectBranch == 0) GetIndirectGotoBlock(); llvm::BasicBlock *BB = getJumpDestForLabel(L).getBlock(); // Make sure the indirect branch includes all of the address-taken blocks. IndirectBranch->addDestination(BB); return llvm::BlockAddress::get(CurFn, BB); } llvm::BasicBlock *CodeGenFunction::GetIndirectGotoBlock() { // If we already made the indirect branch for indirect goto, return its block. if (IndirectBranch) return IndirectBranch->getParent(); CGBuilderTy TmpBuilder(createBasicBlock("indirectgoto")); // Create the PHI node that indirect gotos will add entries to. llvm::Value *DestVal = TmpBuilder.CreatePHI(Int8PtrTy, 0, "indirect.goto.dest"); // Create the indirect branch instruction. IndirectBranch = TmpBuilder.CreateIndirectBr(DestVal); return IndirectBranch->getParent(); } /// Computes the length of an array in elements, as well as the base /// element type and a properly-typed first element pointer. llvm::Value *CodeGenFunction::emitArrayLength(const ArrayType *origArrayType, QualType &baseType, llvm::Value *&addr) { const ArrayType *arrayType = origArrayType; // If it's a VLA, we have to load the stored size. Note that // this is the size of the VLA in bytes, not its size in elements. llvm::Value *numVLAElements = 0; if (isa(arrayType)) { numVLAElements = getVLASize(cast(arrayType)).first; // Walk into all VLAs. This doesn't require changes to addr, // which has type T* where T is the first non-VLA element type. do { QualType elementType = arrayType->getElementType(); arrayType = getContext().getAsArrayType(elementType); // If we only have VLA components, 'addr' requires no adjustment. if (!arrayType) { baseType = elementType; return numVLAElements; } } while (isa(arrayType)); // We get out here only if we find a constant array type // inside the VLA. } // We have some number of constant-length arrays, so addr should // have LLVM type [M x [N x [...]]]*. Build a GEP that walks // down to the first element of addr. SmallVector gepIndices; // GEP down to the array type. llvm::ConstantInt *zero = Builder.getInt32(0); gepIndices.push_back(zero); uint64_t countFromCLAs = 1; QualType eltType; llvm::ArrayType *llvmArrayType = dyn_cast( cast(addr->getType())->getElementType()); while (llvmArrayType) { assert(isa(arrayType)); assert(cast(arrayType)->getSize().getZExtValue() == llvmArrayType->getNumElements()); gepIndices.push_back(zero); countFromCLAs *= llvmArrayType->getNumElements(); eltType = arrayType->getElementType(); llvmArrayType = dyn_cast(llvmArrayType->getElementType()); arrayType = getContext().getAsArrayType(arrayType->getElementType()); assert((!llvmArrayType || arrayType) && "LLVM and Clang types are out-of-synch"); } if (arrayType) { // From this point onwards, the Clang array type has been emitted // as some other type (probably a packed struct). Compute the array // size, and just emit the 'begin' expression as a bitcast. while (arrayType) { countFromCLAs *= cast(arrayType)->getSize().getZExtValue(); eltType = arrayType->getElementType(); arrayType = getContext().getAsArrayType(eltType); } unsigned AddressSpace = addr->getType()->getPointerAddressSpace(); llvm::Type *BaseType = ConvertType(eltType)->getPointerTo(AddressSpace); addr = Builder.CreateBitCast(addr, BaseType, "array.begin"); } else { // Create the actual GEP. addr = Builder.CreateInBoundsGEP(addr, gepIndices, "array.begin"); } baseType = eltType; llvm::Value *numElements = llvm::ConstantInt::get(SizeTy, countFromCLAs); // If we had any VLA dimensions, factor them in. if (numVLAElements) numElements = Builder.CreateNUWMul(numVLAElements, numElements); return numElements; } std::pair CodeGenFunction::getVLASize(QualType type) { const VariableArrayType *vla = getContext().getAsVariableArrayType(type); assert(vla && "type was not a variable array type!"); return getVLASize(vla); } std::pair CodeGenFunction::getVLASize(const VariableArrayType *type) { // The number of elements so far; always size_t. llvm::Value *numElements = 0; QualType elementType; do { elementType = type->getElementType(); llvm::Value *vlaSize = VLASizeMap[type->getSizeExpr()]; assert(vlaSize && "no size for VLA!"); assert(vlaSize->getType() == SizeTy); if (!numElements) { numElements = vlaSize; } else { // It's undefined behavior if this wraps around, so mark it that way. // FIXME: Teach -fcatch-undefined-behavior to trap this. numElements = Builder.CreateNUWMul(numElements, vlaSize); } } while ((type = getContext().getAsVariableArrayType(elementType))); return std::pair(numElements, elementType); } void CodeGenFunction::EmitVariablyModifiedType(QualType type) { assert(type->isVariablyModifiedType() && "Must pass variably modified type to EmitVLASizes!"); EnsureInsertPoint(); // We're going to walk down into the type and look for VLA // expressions. do { assert(type->isVariablyModifiedType()); const Type *ty = type.getTypePtr(); switch (ty->getTypeClass()) { #define TYPE(Class, Base) #define ABSTRACT_TYPE(Class, Base) #define NON_CANONICAL_TYPE(Class, Base) #define DEPENDENT_TYPE(Class, Base) case Type::Class: #define NON_CANONICAL_UNLESS_DEPENDENT_TYPE(Class, Base) #include "clang/AST/TypeNodes.def" llvm_unreachable("unexpected dependent type!"); // These types are never variably-modified. case Type::Builtin: case Type::Complex: case Type::Vector: case Type::ExtVector: case Type::Record: case Type::Enum: case Type::Elaborated: case Type::TemplateSpecialization: case Type::ObjCObject: case Type::ObjCInterface: case Type::ObjCObjectPointer: llvm_unreachable("type class is never variably-modified!"); case Type::Adjusted: type = cast(ty)->getAdjustedType(); break; case Type::Decayed: type = cast(ty)->getPointeeType(); break; case Type::Pointer: type = cast(ty)->getPointeeType(); break; case Type::BlockPointer: type = cast(ty)->getPointeeType(); break; case Type::LValueReference: case Type::RValueReference: type = cast(ty)->getPointeeType(); break; case Type::MemberPointer: type = cast(ty)->getPointeeType(); break; case Type::ConstantArray: case Type::IncompleteArray: // Losing element qualification here is fine. type = cast(ty)->getElementType(); break; case Type::VariableArray: { // Losing element qualification here is fine. const VariableArrayType *vat = cast(ty); // Unknown size indication requires no size computation. // Otherwise, evaluate and record it. if (const Expr *size = vat->getSizeExpr()) { // It's possible that we might have emitted this already, // e.g. with a typedef and a pointer to it. llvm::Value *&entry = VLASizeMap[size]; if (!entry) { llvm::Value *Size = EmitScalarExpr(size); // C11 6.7.6.2p5: // If the size is an expression that is not an integer constant // expression [...] each time it is evaluated it shall have a value // greater than zero. if (SanOpts->VLABound && size->getType()->isSignedIntegerType()) { llvm::Value *Zero = llvm::Constant::getNullValue(Size->getType()); llvm::Constant *StaticArgs[] = { EmitCheckSourceLocation(size->getLocStart()), EmitCheckTypeDescriptor(size->getType()) }; EmitCheck(Builder.CreateICmpSGT(Size, Zero), "vla_bound_not_positive", StaticArgs, Size, CRK_Recoverable); } // Always zexting here would be wrong if it weren't // undefined behavior to have a negative bound. entry = Builder.CreateIntCast(Size, SizeTy, /*signed*/ false); } } type = vat->getElementType(); break; } case Type::FunctionProto: case Type::FunctionNoProto: type = cast(ty)->getReturnType(); break; case Type::Paren: case Type::TypeOf: case Type::UnaryTransform: case Type::Attributed: case Type::SubstTemplateTypeParm: case Type::PackExpansion: // Keep walking after single level desugaring. type = type.getSingleStepDesugaredType(getContext()); break; case Type::Typedef: case Type::Decltype: case Type::Auto: // Stop walking: nothing to do. return; case Type::TypeOfExpr: // Stop walking: emit typeof expression. EmitIgnoredExpr(cast(ty)->getUnderlyingExpr()); return; case Type::Atomic: type = cast(ty)->getValueType(); break; } } while (type->isVariablyModifiedType()); } llvm::Value* CodeGenFunction::EmitVAListRef(const Expr* E) { if (getContext().getBuiltinVaListType()->isArrayType()) return EmitScalarExpr(E); return EmitLValue(E).getAddress(); } void CodeGenFunction::EmitDeclRefExprDbgValue(const DeclRefExpr *E, llvm::Constant *Init) { assert (Init && "Invalid DeclRefExpr initializer!"); if (CGDebugInfo *Dbg = getDebugInfo()) if (CGM.getCodeGenOpts().getDebugInfo() >= CodeGenOptions::LimitedDebugInfo) Dbg->EmitGlobalVariable(E->getDecl(), Init); } CodeGenFunction::PeepholeProtection CodeGenFunction::protectFromPeepholes(RValue rvalue) { // At the moment, the only aggressive peephole we do in IR gen // is trunc(zext) folding, but if we add more, we can easily // extend this protection. if (!rvalue.isScalar()) return PeepholeProtection(); llvm::Value *value = rvalue.getScalarVal(); if (!isa(value)) return PeepholeProtection(); // Just make an extra bitcast. assert(HaveInsertPoint()); llvm::Instruction *inst = new llvm::BitCastInst(value, value->getType(), "", Builder.GetInsertBlock()); PeepholeProtection protection; protection.Inst = inst; return protection; } void CodeGenFunction::unprotectFromPeepholes(PeepholeProtection protection) { if (!protection.Inst) return; // In theory, we could try to duplicate the peepholes now, but whatever. protection.Inst->eraseFromParent(); } llvm::Value *CodeGenFunction::EmitAnnotationCall(llvm::Value *AnnotationFn, llvm::Value *AnnotatedVal, StringRef AnnotationStr, SourceLocation Location) { llvm::Value *Args[4] = { AnnotatedVal, Builder.CreateBitCast(CGM.EmitAnnotationString(AnnotationStr), Int8PtrTy), Builder.CreateBitCast(CGM.EmitAnnotationUnit(Location), Int8PtrTy), CGM.EmitAnnotationLineNo(Location) }; return Builder.CreateCall(AnnotationFn, Args); } void CodeGenFunction::EmitVarAnnotations(const VarDecl *D, llvm::Value *V) { assert(D->hasAttr() && "no annotate attribute"); // FIXME We create a new bitcast for every annotation because that's what // llvm-gcc was doing. for (specific_attr_iterator ai = D->specific_attr_begin(), ae = D->specific_attr_end(); ai != ae; ++ai) EmitAnnotationCall(CGM.getIntrinsic(llvm::Intrinsic::var_annotation), Builder.CreateBitCast(V, CGM.Int8PtrTy, V->getName()), (*ai)->getAnnotation(), D->getLocation()); } llvm::Value *CodeGenFunction::EmitFieldAnnotations(const FieldDecl *D, llvm::Value *V) { assert(D->hasAttr() && "no annotate attribute"); llvm::Type *VTy = V->getType(); llvm::Value *F = CGM.getIntrinsic(llvm::Intrinsic::ptr_annotation, CGM.Int8PtrTy); for (specific_attr_iterator ai = D->specific_attr_begin(), ae = D->specific_attr_end(); ai != ae; ++ai) { // FIXME Always emit the cast inst so we can differentiate between // annotation on the first field of a struct and annotation on the struct // itself. if (VTy != CGM.Int8PtrTy) V = Builder.Insert(new llvm::BitCastInst(V, CGM.Int8PtrTy)); V = EmitAnnotationCall(F, V, (*ai)->getAnnotation(), D->getLocation()); V = Builder.CreateBitCast(V, VTy); } return V; } CodeGenFunction::CGCapturedStmtInfo::~CGCapturedStmtInfo() { } @