blob: 3f9a52ab7638aff79130ba7cd0f1277f70c2dd2f [file] [log] [blame]
//===--- CodeGenFunction.cpp - Emit LLVM Code from ASTs for a Function ----===//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
// This coordinates the per-function state used while generating code.
#include "CodeGenFunction.h"
#include "CGBlocks.h"
#include "CGCleanup.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/ASTLambda.h"
#include "clang/AST/Decl.h"
#include "clang/AST/DeclCXX.h"
#include "clang/AST/StmtCXX.h"
#include "clang/AST/StmtObjC.h"
#include "clang/Basic/Builtins.h"
#include "clang/Basic/CodeGenOptions.h"
#include "clang/Basic/TargetInfo.h"
#include "clang/CodeGen/CGFunctionInfo.h"
#include "clang/Frontend/FrontendDiagnostic.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/MDBuilder.h"
#include "llvm/IR/Operator.h"
#include "llvm/Transforms/Utils/PromoteMemToReg.h"
using namespace clang;
using namespace CodeGen;
/// shouldEmitLifetimeMarkers - Decide whether we need emit the life-time
/// markers.
static bool shouldEmitLifetimeMarkers(const CodeGenOptions &CGOpts,
const LangOptions &LangOpts) {
if (CGOpts.DisableLifetimeMarkers)
return false;
// Sanitizers may use markers.
if (CGOpts.SanitizeAddressUseAfterScope ||
LangOpts.Sanitize.has(SanitizerKind::HWAddress) ||
return true;
// For now, only in optimized builds.
return CGOpts.OptimizationLevel != 0;
CodeGenFunction::CodeGenFunction(CodeGenModule &cgm, bool suppressNewContext)
: CodeGenTypeCache(cgm), CGM(cgm), Target(cgm.getTarget()),
Builder(cgm, cgm.getModule().getContext(), llvm::ConstantFolder(),
SanOpts(CGM.getLangOpts().Sanitize), DebugInfo(CGM.getModuleDebugInfo()),
PGO(cgm), ShouldEmitLifetimeMarkers(shouldEmitLifetimeMarkers(
CGM.getCodeGenOpts(), CGM.getLangOpts())) {
if (!suppressNewContext)
llvm::FastMathFlags FMF;
if (CGM.getLangOpts().FastMath)
if (CGM.getLangOpts().FiniteMathOnly) {
if (CGM.getCodeGenOpts().NoNaNsFPMath) {
if (CGM.getCodeGenOpts().NoSignedZeros) {
if (CGM.getCodeGenOpts().ReciprocalMath) {
if (CGM.getCodeGenOpts().Reassociate) {
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)
if (getLangOpts().OpenMP && CurFn)
CharUnits CodeGenFunction::getNaturalPointeeTypeAlignment(QualType T,
LValueBaseInfo *BaseInfo,
TBAAAccessInfo *TBAAInfo) {
return getNaturalTypeAlignment(T->getPointeeType(), BaseInfo, TBAAInfo,
/* forPointeeType= */ true);
CharUnits CodeGenFunction::getNaturalTypeAlignment(QualType T,
LValueBaseInfo *BaseInfo,
TBAAAccessInfo *TBAAInfo,
bool forPointeeType) {
if (TBAAInfo)
*TBAAInfo = CGM.getTBAAAccessInfo(T);
// 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<TypedefType>()) {
if (auto Align = TT->getDecl()->getMaxAlignment()) {
if (BaseInfo)
*BaseInfo = LValueBaseInfo(AlignmentSource::AttributedType);
return getContext().toCharUnitsFromBits(Align);
if (BaseInfo)
*BaseInfo = LValueBaseInfo(AlignmentSource::Type);
CharUnits Alignment;
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);
if (T.getQualifiers().hasUnaligned())
Alignment = CharUnits::One();
// 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 &&
Alignment = CharUnits::fromQuantity(MaxAlign);
return Alignment;
LValue CodeGenFunction::MakeNaturalAlignAddrLValue(llvm::Value *V, QualType T) {
LValueBaseInfo BaseInfo;
TBAAAccessInfo TBAAInfo;
CharUnits Alignment = getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo);
return LValue::MakeAddr(Address(V, Alignment), T, getContext(), BaseInfo,
/// 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.
CodeGenFunction::MakeNaturalAlignPointeeAddrLValue(llvm::Value *V, QualType T) {
LValueBaseInfo BaseInfo;
TBAAAccessInfo TBAAInfo;
CharUnits Align = getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo,
/* forPointeeType= */ true);
return MakeAddrLValue(Address(V, Align), T, BaseInfo, TBAAInfo);
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/"
llvm_unreachable("non-canonical or dependent type in IR-generation");
case Type::Auto:
case Type::DeducedTemplateSpecialization:
llvm_unreachable("undeduced 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:
case Type::Pipe:
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<AtomicType>(type)->getValueType();
llvm_unreachable("unknown type kind!");
llvm::DebugLoc 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()) {
delete ReturnBlock.getBlock();
ReturnBlock = JumpDest();
} else
return llvm::DebugLoc();
// 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 =
if (BI && BI->isUnconditional() &&
BI->getSuccessor(0) == ReturnBlock.getBlock()) {
// Record/return the DebugLoc of the simple 'return' expression to be used
// later by the actual 'ret' instruction.
llvm::DebugLoc Loc = BI->getDebugLoc();
delete ReturnBlock.getBlock();
ReturnBlock = JumpDest();
return Loc;
// 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.
return llvm::DebugLoc();
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);
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 HasCleanups = EHStack.stable_begin() != PrologueCleanupDepth;
bool HasOnlyLifetimeMarkers =
HasCleanups && EHStack.containsOnlyLifetimeMarkers(PrologueCleanupDepth);
bool EmitRetDbgLoc = !HasCleanups || HasOnlyLifetimeMarkers;
if (HasCleanups) {
// Make sure the line table doesn't jump back into the body for
// the ret after it's been at EndLoc.
if (CGDebugInfo *DI = getDebugInfo())
if (OnlySimpleReturnStmts)
DI->EmitLocation(Builder, EndLoc);
// Emit function epilog (to return).
llvm::DebugLoc Loc = EmitReturnBlock();
if (ShouldInstrumentFunction()) {
if (CGM.getCodeGenOpts().InstrumentFunctions)
CurFn->addFnAttr("instrument-function-exit", "__cyg_profile_func_exit");
if (CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining)
// Emit debug descriptor for function end.
if (CGDebugInfo *DI = getDebugInfo())
DI->EmitFunctionEnd(Builder, CurFn);
// 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);
EmitFunctionEpilog(*CurFnInfo, EmitRetDbgLoc, EndLoc);
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) {
// 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<llvm::Value *, 4> EscapeArgs;
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);
// Remove the AllocaInsertPt instruction, which is just a convenience for us.
llvm::Instruction *Ptr = AllocaInsertPt;
AllocaInsertPt = nullptr;
// 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<llvm::PHINode>(IndirectBranch->getAddress());
if (PN->getNumIncomingValues() == 0) {
EmitIfUsed(*this, EHResumeBlock);
EmitIfUsed(*this, TerminateLandingPad);
EmitIfUsed(*this, TerminateHandler);
EmitIfUsed(*this, UnreachableBlock);
for (const auto &FuncletAndParent : TerminateFunclets)
EmitIfUsed(*this, FuncletAndParent.second);
if (CGM.getCodeGenOpts().EmitDeclMetadata)
for (SmallVectorImpl<std::pair<llvm::Instruction *, llvm::Value *> >::iterator
I = DeferredReplacements.begin(),
E = DeferredReplacements.end();
I != E; ++I) {
// 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);
cast<llvm::AllocaInst>(NormalCleanupDest.getPointer()), DT);
NormalCleanupDest = Address::invalid();
// Scan function arguments for vector width.
for (llvm::Argument &A : CurFn->args())
if (auto *VT = dyn_cast<llvm::VectorType>(A.getType()))
LargestVectorWidth = std::max((uint64_t)LargestVectorWidth,
// Update vector width based on return type.
if (auto *VT = dyn_cast<llvm::VectorType>(CurFn->getReturnType()))
LargestVectorWidth = std::max((uint64_t)LargestVectorWidth,
// Add the required-vector-width attribute. This contains the max width from:
// 1. min-vector-width attribute used in the source program.
// 2. Any builtins used that have a vector width specified.
// 3. Values passed in and out of inline assembly.
// 4. Width of vector arguments and return types for this function.
// 5. Width of vector aguments and return types for functions called by this
// function.
CurFn->addFnAttr("min-legal-vector-width", llvm::utostr(LargestVectorWidth));
// If we generated an unreachable return block, delete it now.
if (ReturnBlock.isValid() && ReturnBlock.getBlock()->use_empty()) {
if (ReturnValue.isValid()) {
auto *RetAlloca = dyn_cast<llvm::AllocaInst>(ReturnValue.getPointer());
if (RetAlloca && RetAlloca->use_empty()) {
ReturnValue = Address::invalid();
/// ShouldInstrumentFunction - Return true if the current function should be
/// instrumented with __cyg_profile_func_* calls
bool CodeGenFunction::ShouldInstrumentFunction() {
if (!CGM.getCodeGenOpts().InstrumentFunctions &&
!CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining &&
return false;
if (!CurFuncDecl || CurFuncDecl->hasAttr<NoInstrumentFunctionAttr>())
return false;
return true;
/// ShouldXRayInstrument - Return true if the current function should be
/// instrumented with XRay nop sleds.
bool CodeGenFunction::ShouldXRayInstrumentFunction() const {
return CGM.getCodeGenOpts().XRayInstrumentFunctions;
/// 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 ==
bool CodeGenFunction::AlwaysEmitXRayTypedEvents() const {
return CGM.getCodeGenOpts().XRayInstrumentFunctions &&
(CGM.getCodeGenOpts().XRayAlwaysEmitTypedEvents ||
CGM.getCodeGenOpts().XRayInstrumentationBundle.Mask ==
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(),
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, "");
auto *GOTAsInt = Builder.CreateAdd(PCRelAsInt, FuncAsInt, "");
auto *GOTAddr = Builder.CreateIntToPtr(GOTAsInt, Int8PtrPtrTy, "global_addr");
// Load the original pointer through the global.
return Builder.CreateLoad(Address(GOTAddr, getPointerAlign()),
void CodeGenFunction::EmitOpenCLKernelMetadata(const FunctionDecl *FD,
llvm::Function *Fn)
if (!FD->hasAttr<OpenCLKernelAttr>())
llvm::LLVMContext &Context = getLLVMContext();
CGM.GenOpenCLArgMetadata(Fn, FD, this);
if (const VecTypeHintAttr *A = FD->getAttr<VecTypeHintAttr>()) {
QualType HintQTy = A->getTypeHint();
const ExtVectorType *HintEltQTy = HintQTy->getAs<ExtVectorType>();
bool IsSignedInteger =
HintQTy->isSignedIntegerType() ||
(HintEltQTy && HintEltQTy->getElementType()->isSignedIntegerType());
llvm::Metadata *AttrMDArgs[] = {
llvm::IntegerType::get(Context, 32),
llvm::APInt(32, (uint64_t)(IsSignedInteger ? 1 : 0))))};
Fn->setMetadata("vec_type_hint", llvm::MDNode::get(Context, AttrMDArgs));
if (const WorkGroupSizeHintAttr *A = FD->getAttr<WorkGroupSizeHintAttr>()) {
llvm::Metadata *AttrMDArgs[] = {
Fn->setMetadata("work_group_size_hint", llvm::MDNode::get(Context, AttrMDArgs));
if (const ReqdWorkGroupSizeAttr *A = FD->getAttr<ReqdWorkGroupSizeAttr>()) {
llvm::Metadata *AttrMDArgs[] = {
Fn->setMetadata("reqd_work_group_size", llvm::MDNode::get(Context, AttrMDArgs));
if (const OpenCLIntelReqdSubGroupSizeAttr *A =
FD->getAttr<OpenCLIntelReqdSubGroupSizeAttr>()) {
llvm::Metadata *AttrMDArgs[] = {
llvm::MDNode::get(Context, AttrMDArgs));
/// 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<FunctionDecl>(F))
Body = FD->getBody();
else if (auto *OMD = dyn_cast_or_null<ObjCMethodDecl>(F))
Body = OMD->getBody();
if (auto *CS = dyn_cast_or_null<CompoundStmt>(Body)) {
auto LastStmt = CS->body_rbegin();
if (LastStmt != CS->body_rend())
return isa<ReturnStmt>(*LastStmt);
return false;
void CodeGenFunction::markAsIgnoreThreadCheckingAtRuntime(llvm::Function *Fn) {
if (SanOpts.has(SanitizerKind::Thread)) {
static bool matchesStlAllocatorFn(const Decl *D, const ASTContext &Ctx) {
auto *MD = dyn_cast_or_null<CXXMethodDecl>(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<PointerType>();
if (!PT || !PT->isVoidPointerType() ||
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<CXXMethodDecl>(FD))
if (!MD->isStatic())
return nullptr;
return CGM.getTargetCodeGenInfo().getUBSanFunctionSignature(CGM);
void CodeGenFunction::StartFunction(GlobalDecl GD,
QualType RetTy,
llvm::Function *Fn,
const CGFunctionInfo &FnInfo,
const FunctionArgList &Args,
SourceLocation Loc,
SourceLocation StartLoc) {
assert(!CurFn &&
"Do not use a CodeGenFunction object for more than one function");
const Decl *D = GD.getDecl();
DidCallStackSave = false;
CurCodeDecl = D;
if (const auto *FD = dyn_cast_or_null<FunctionDecl>(D))
if (FD->usesSEHTry())
CurSEHParent = FD;
CurFuncDecl = (D ? D->getNonClosureContext() : nullptr);
FnRetTy = RetTy;
CurFn = Fn;
CurFnInfo = &FnInfo;
assert(CurFn->isDeclaration() && "Function already has body?");
// If this function has been blacklisted for any of the enabled sanitizers,
// disable the sanitizer for the function.
do {
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"
} while (0);
if (D) {
// Apply the no_sanitize* attributes to SanOpts.
for (auto Attr : D->specific_attrs<NoSanitizeAttr>()) {
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);
// Apply sanitizer attributes to the function.
if (SanOpts.hasOneOf(SanitizerKind::Address | SanitizerKind::KernelAddress))
if (SanOpts.hasOneOf(SanitizerKind::HWAddress | SanitizerKind::KernelHWAddress))
if (SanOpts.has(SanitizerKind::MemTag))
if (SanOpts.has(SanitizerKind::Thread))
if (SanOpts.hasOneOf(SanitizerKind::Memory | SanitizerKind::KernelMemory))
if (SanOpts.has(SanitizerKind::SafeStack))
if (SanOpts.has(SanitizerKind::ShadowCallStack))
// Apply fuzzing attribute to the function.
if (SanOpts.hasOneOf(SanitizerKind::Fuzzer | SanitizerKind::FuzzerNoLink))
// Ignore TSan memory acesses from within ObjC/ObjC++ dealloc, initialize,
// .cxx_destruct, __destroy_helper_block_ and all of their calees at run time.
if (SanOpts.has(SanitizerKind::Thread)) {
if (const auto *OMD = dyn_cast_or_null<ObjCMethodDecl>(D)) {
IdentifierInfo *II = OMD->getSelector().getIdentifierInfoForSlot(0);
if (OMD->getMethodFamily() == OMF_dealloc ||
OMD->getMethodFamily() == OMF_initialize ||
(OMD->getSelector().isUnarySelector() && II->isStr(".cxx_destruct"))) {
// 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;
// Ignore null checks in coroutine functions since the coroutines passes
// are not aware of how to move the extra UBSan instructions across the split
// coroutine boundaries.
if (D && SanOpts.has(SanitizerKind::Null))
if (const auto *FD = dyn_cast<FunctionDecl>(D))
if (FD->getBody() &&
FD->getBody()->getStmtClass() == Stmt::CoroutineBodyStmtClass)
SanOpts.Mask &= ~SanitizerKind::Null;
// Apply xray attributes to the function (as a string, for now)
if (D) {
if (const auto *XRayAttr = D->getAttr<XRayInstrumentAttr>()) {
if (CGM.getCodeGenOpts().XRayInstrumentationBundle.has(
XRayInstrKind::Function)) {
if (XRayAttr->alwaysXRayInstrument() && ShouldXRayInstrumentFunction())
Fn->addFnAttr("function-instrument", "xray-always");
if (XRayAttr->neverXRayInstrument())
Fn->addFnAttr("function-instrument", "xray-never");
if (const auto *LogArgs = D->getAttr<XRayLogArgsAttr>())
if (ShouldXRayInstrumentFunction())
} else {
if (ShouldXRayInstrumentFunction() && !CGM.imbueXRayAttrs(Fn, Loc))
// Add no-jump-tables value.
// Add profile-sample-accurate value.
if (CGM.getCodeGenOpts().ProfileSampleAccurate)
if (D && D->hasAttr<CFICanonicalJumpTableAttr>())
if (getLangOpts().OpenCL) {
// Add metadata for a kernel function.
if (const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(D))
EmitOpenCLKernelMetadata(FD, Fn);
// If we are checking function types, emit a function type signature as
// prologue data.
if (getLangOpts().CPlusPlus && SanOpts.has(SanitizerKind::Function)) {
if (const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(D)) {
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 =
llvm::Constant *FTRTTIConst =
CGM.GetAddrOfRTTIDescriptor(ProtoTy, /*ForEH=*/true);
llvm::Constant *FTRTTIConstEncoded =
EncodeAddrForUseInPrologue(Fn, FTRTTIConst);
llvm::Constant *PrologueStructElems[] = {PrologueSig,
llvm::Constant *PrologueStructConst =
llvm::ConstantStruct::getAnon(PrologueStructElems, /*Packed=*/true);
// 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<ReturnsNonNullAttr>()))
RetValNullabilityPrecondition =
// If we're in C++ mode and the function name is "main", it is guaranteed
// to be norecurse by the standard ( "The function main shall not be
// used within a program").
if (getLangOpts().CPlusPlus)
if (const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(D))
if (FD->isMain())
// If a custom alignment is used, force realigning to this alignment on
// any main function which certainly will need it.
if (const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(D))
if ((FD->isMain() || FD->isMSVCRTEntryPoint()) &&
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, "allocapt", EntryBB);
ReturnBlock = getJumpDestInCurrentScope("return");
// 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));
// Emit subprogram debug descriptor.
if (CGDebugInfo *DI = getDebugInfo()) {
// 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<FunctionDecl>(D))
if (const auto *SrcFnTy = FD->getType()->getAs<FunctionType>())
CC = SrcFnTy->getCallConv();
SmallVector<QualType, 16> ArgTypes;
for (const VarDecl *VD : Args)
QualType FnType = getContext().getFunctionType(
RetTy, ArgTypes, FunctionProtoType::ExtProtoInfo(CC));
DI->EmitFunctionStart(GD, Loc, StartLoc, FnType, CurFn, CurFuncIsThunk,
if (ShouldInstrumentFunction()) {
if (CGM.getCodeGenOpts().InstrumentFunctions)
CurFn->addFnAttr("instrument-function-entry", "__cyg_profile_func_enter");
if (CGM.getCodeGenOpts().InstrumentFunctionsAfterInlining)
if (CGM.getCodeGenOpts().InstrumentFunctionEntryBare)
// 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.
if (CGM.getCodeGenOpts().InstrumentForProfiling) {
// Calls to fentry/mcount should not be generated if function has
// the no_instrument_function attribute.
if (!CurFuncDecl || !CurFuncDecl->hasAttr<NoInstrumentFunctionAttr>()) {
if (CGM.getCodeGenOpts().CallFEntry)
Fn->addFnAttr("fentry-call", "true");
else {
if (RetTy->isVoidType()) {
// Void type; nothing to return.
ReturnValue = Address::invalid();
// Count the implicit return.
if (!endsWithReturn(D))
} else if (CurFnInfo->getReturnInfo().getKind() == ABIArgInfo::Indirect) {
// Indirect 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())
ReturnValue = Address(&*AI, CurFnInfo->getReturnInfo().getIndirectAlign());
if (!CurFnInfo->getReturnInfo().getIndirectByVal()) {
ReturnValuePointer =
CreateDefaultAlignTempAlloca(Int8PtrTy, "result.ptr");
ReturnValue.getPointer(), Int8PtrTy),
} 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();
llvm::Value *Addr = Builder.CreateStructGEP(nullptr, &*EI, Idx);
ReturnValuePointer = Address(Addr, getPointerAlign());
Addr = Builder.CreateAlignedLoad(Addr, getPointerAlign(), "agg.result");
ReturnValue = Address(Addr, getNaturalTypeAlignment(RetTy));
} 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() &&
AutoreleaseResult = true;
PrologueCleanupDepth = EHStack.stable_begin();
// Emit OpenMP specific initialization of the device functions.
if (getLangOpts().OpenMP && CurCodeDecl)
CGM.getOpenMPRuntime().emitFunctionProlog(*this, CurCodeDecl);
EmitFunctionProlog(*CurFnInfo, CurFn, Args);
if (D && isa<CXXMethodDecl>(D) && cast<CXXMethodDecl>(D)->isInstance()) {
const CXXMethodDecl *MD = cast<CXXMethodDecl>(D);
if (MD->getParent()->isLambda() &&
MD->getOverloadedOperator() == OO_Call) {
// We're in a lambda; figure out the captures.
if (LambdaThisCaptureField) {
// 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();
for (auto *FD : MD->getParent()->fields()) {
if (FD->hasCapturedVLAType()) {
auto *ExprArg = EmitLoadOfLValue(EmitLValueForLambdaField(FD),
auto VAT = FD->getCapturedVLAType();
VLASizeMap[VAT->getSizeExpr()] = ExprArg;
} 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;
// Check the 'this' pointer once per function, if it's available.
if (CXXABIThisValue) {
SanitizerSet SkippedChecks;
SkippedChecks.set(SanitizerKind::ObjectSize, true);
QualType ThisTy = MD->getThisType();
// 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) &&
MD->getParent()->getLambdaCaptureDefault() == LCD_None)
SkippedChecks.set(SanitizerKind::Null, true);
EmitTypeCheck(isa<CXXConstructorDecl>(MD) ? TCK_ConstructorCall
: TCK_MemberCall,
Loc, CXXABIThisValue, ThisTy,
// 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<ParmVarDecl>(VD))
Ty = PVD->getOriginalType();
Ty = VD->getType();
if (Ty->isVariablyModifiedType())
// Emit a location at the end of the prologue.
if (CGDebugInfo *DI = getDebugInfo())
DI->EmitLocation(Builder, StartLoc);
// 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<MinVectorWidthAttr>())
LargestVectorWidth = VecWidth->getVectorWidth();
void CodeGenFunction::EmitFunctionBody(const Stmt *Body) {
if (const CompoundStmt *S = dyn_cast<CompoundStmt>(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,
const Stmt *S) {
llvm::BasicBlock *SkipCountBB = nullptr;
if (HaveInsertPoint() && CGM.getCodeGenOpts().hasProfileClangInstr()) {
// 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");
uint64_t CurrentCount = getCurrentProfileCount();
setCurrentProfileCount(getCurrentProfileCount() + CurrentCount);
if (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->isInterposable()) return;
for (llvm::BasicBlock &BB : *F)
for (llvm::Instruction &I : BB)
if (I.mayThrow())
QualType CodeGenFunction::BuildFunctionArgList(GlobalDecl GD,
FunctionArgList &Args) {
const FunctionDecl *FD = cast<FunctionDecl>(GD.getDecl());
QualType ResTy = FD->getReturnType();
const CXXMethodDecl *MD = dyn_cast<CXXMethodDecl>(FD);
if (MD && MD->isInstance()) {
if (CGM.getCXXABI().HasThisReturn(GD))
ResTy = MD->getThisType();
else if (CGM.getCXXABI().hasMostDerivedReturn(GD))
ResTy = CGM.getContext().VoidPtrTy;
CGM.getCXXABI().buildThisParam(*this, Args);
// 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<CXXConstructorDecl>(FD))
if (auto Inherited = CD->getInheritedConstructor())
PassedParams =
getTypes().inheritingCtorHasParams(Inherited, GD.getCtorType());
if (PassedParams) {
for (auto *Param : FD->parameters()) {
if (!Param->hasAttr<PassObjectSizeAttr>())
auto *Implicit = ImplicitParamDecl::Create(
getContext(), Param->getDeclContext(), Param->getLocation(),
/*Id=*/nullptr, getContext().getSizeType(), ImplicitParamDecl::Other);
SizeArguments[Param] = Implicit;
if (MD && (isa<CXXConstructorDecl>(MD) || isa<CXXDestructorDecl>(MD)))
CGM.getCXXABI().addImplicitStructorParams(*this, ResTy, Args);
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<RecordType>()) {
if (const auto *ClassDecl = dyn_cast<CXXRecordDecl>(RT->getDecl()))
return !ClassDecl->hasTrivialDestructor();
return !T.isTriviallyCopyableType(Context);
void CodeGenFunction::GenerateCode(GlobalDecl GD, llvm::Function *Fn,
const CGFunctionInfo &FnInfo) {
const FunctionDecl *FD = cast<FunctionDecl>(GD.getDecl());
CurGD = GD;
FunctionArgList Args;
QualType ResTy = BuildFunctionArgList(GD, Args);
// Check if we should generate debug info for this function.
if (FD->hasAttr<NoDebugAttr>())
DebugInfo = nullptr; // disable debug info indefinitely for this function
// The function might not have a body if we're generating thunks for a
// function declaration.
SourceRange BodyRange;
if (Stmt *Body = FD->getBody())
BodyRange = Body->getSourceRange();
BodyRange = FD->getLocation();
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();
Stmt *Body = FD->getBody();
// Initialize helper which will detect jumps which can cause invalid lifetime
// markers.
if (Body && ShouldEmitLifetimeMarkers)
// Emit the standard function prologue.
StartFunction(GD, ResTy, Fn, FnInfo, Args, Loc, BodyRange.getBegin());
// Generate the body of the function.
PGO.assignRegionCounters(GD, CurFn);
if (isa<CXXDestructorDecl>(FD))
else if (isa<CXXConstructorDecl>(FD))
else if (getLangOpts().CUDA &&
!getLangOpts().CUDAIsDevice &&
CGM.getCUDARuntime().emitDeviceStub(*this, Args);
else if (isa<CXXMethodDecl>(FD) &&
cast<CXXMethodDecl>(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).
} else if (FD->isDefaulted() && isa<CXXMethodDecl>(FD) &&
(cast<CXXMethodDecl>(FD)->isCopyAssignmentOperator() ||
cast<CXXMethodDecl>(FD)->isMoveAssignmentOperator())) {
// Implicit copy-assignment gets the same special treatment as implicit
// copy-constructors.
} else if (Body) {
} 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() && !SawAsmBlock &&
!FD->getReturnType()->isVoidType() && Builder.GetInsertBlock()) {
bool ShouldEmitUnreachable =
CGM.getCodeGenOpts().StrictReturn ||
shouldUseUndefinedBehaviorReturnOptimization(FD, getContext());
if (SanOpts.has(SanitizerKind::Return)) {
SanitizerScope SanScope(this);
llvm::Value *IsFalse = Builder.getFalse();
EmitCheck(std::make_pair(IsFalse, SanitizerKind::Return),
EmitCheckSourceLocation(FD->getLocation()), None);
} else if (ShouldEmitUnreachable) {
if (CGM.getCodeGenOpts().OptimizationLevel == 0)
if (SanOpts.has(SanitizerKind::Return) || ShouldEmitUnreachable) {
// Emit the standard function epilogue.
// If we haven't marked the function nothrow through other means, do
// a quick pass now to see if we can.
if (!CurFn->doesNotThrow())
/// 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<LabelStmt>(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<SwitchCase>(S) && !IgnoreCaseStmts)
return true;
// If this is a switch statement, we want to ignore cases below it.
if (isa<SwitchStmt>(S))
IgnoreCaseStmts = true;
// Scan subexpressions for verboten labels.
for (const Stmt *SubStmt : S->children())
if (ContainsLabel(SubStmt, 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<SwitchStmt>(S) || isa<WhileStmt>(S) || isa<DoStmt>(S) ||
return false;
if (isa<BreakStmt>(S))
return true;
// Scan subexpressions for verboten breaks.
for (const Stmt *SubStmt : S->children())
if (containsBreak(SubStmt))
return true;
return false;
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<IfStmt>(S) || isa<SwitchStmt>(S) || isa<WhileStmt>(S) ||
isa<DoStmt>(S) || isa<ForStmt>(S) || isa<CompoundStmt>(S) ||
isa<CXXForRangeStmt>(S) || isa<CXXTryStmt>(S) ||
isa<ObjCForCollectionStmt>(S) || isa<ObjCAtTryStmt>(S))
return false;
if (isa<DeclStmt>(S))
return true;
for (const Stmt *SubStmt : S->children())
if (mightAddDeclToScope(SubStmt))
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,
bool AllowLabels) {
llvm::APSInt ResultInt;
if (!ConstantFoldsToSimpleInteger(Cond, ResultInt, AllowLabels))
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,
bool AllowLabels) {
// FIXME: Rename and handle conversion of other evaluatable things
// to bool.
Expr::EvalResult Result;
if (!Cond->EvaluateAsInt(Result, getContext()))
return false; // Not foldable, not integer or not fully evaluatable.
llvm::APSInt Int = Result.Val.getInt();
if (!AllowLabels && 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<BinaryOperator>(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");
// The counter tells us how often we evaluate RHS, and all of TrueCount
// can be propagated to that branch.
uint64_t RHSCount = getProfileCount(CondBOp->getRHS());
ConditionalEvaluation eval(*this);
ApplyDebugLocation DL(*this, Cond);
EmitBranchOnBoolExpr(CondBOp->getLHS(), LHSTrue, FalseBlock, RHSCount);
// Any temporaries created here are conditional.
EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, TrueCount);
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");
// 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 =
getCurrentProfileCount() - getProfileCount(CondBOp->getRHS());
uint64_t RHSCount = TrueCount - LHSCount;
ConditionalEvaluation eval(*this);
ApplyDebugLocation DL(*this, Cond);
EmitBranchOnBoolExpr(CondBOp->getLHS(), TrueBlock, LHSFalse, LHSCount);
// Any temporaries created here are conditional.
EmitBranchOnBoolExpr(CondBOp->getRHS(), TrueBlock, FalseBlock, RHSCount);
if (const UnaryOperator *CondUOp = dyn_cast<UnaryOperator>(Cond)) {
// br(!x, t, f) -> br(x, f, t)
if (CondUOp->getOpcode() == UO_LNot) {
// Negate the count.
uint64_t FalseCount = getCurrentProfileCount() - TrueCount;
// Negate the condition and swap the destination blocks.
return EmitBranchOnBoolExpr(CondUOp->getSubExpr(), FalseBlock, TrueBlock,
if (const ConditionalOperator *CondOp = dyn_cast<ConditionalOperator>(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,
// 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 =
getProfileCount(CondOp) / (double)getCurrentProfileCount();
LHSScaledTrueCount = TrueCount * LHSRatio;
ApplyDebugLocation DL(*this, Cond);
EmitBranchOnBoolExpr(CondOp->getLHS(), TrueBlock, FalseBlock,
EmitBranchOnBoolExpr(CondOp->getRHS(), TrueBlock, FalseBlock,
TrueCount - LHSScaledTrueCount);
if (const CXXThrowExpr *Throw = dyn_cast<CXXThrowExpr>(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);
// 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;
auto *Call = dyn_cast<CallExpr>(Cond->IgnoreImpCasts());
if (Call && CGM.getCodeGenOpts().OptimizationLevel != 0) {
auto *FD = dyn_cast_or_null<FunctionDecl>(Call->getCalleeDecl());
if (FD && FD->getBuiltinID() == Builtin::BI__builtin_unpredictable) {
llvm::MDBuilder MDHelper(getLLVMContext());
Unpredictable = MDHelper.createUnpredictable();
// 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(getCurrentProfileCount(), TrueCount);
llvm::MDNode *Weights =
createProfileWeights(TrueCount, CurrentCount - TrueCount);
// Emit the code with the fully general case.
llvm::Value *CondV;
ApplyDebugLocation DL(*this, Cond);
CondV = EvaluateExprAsBool(Cond);
Builder.CreateCondBr(CondV, TrueBlock, FalseBlock, Weights, Unpredictable);
/// 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,
Address dest, Address src,
llvm::Value *sizeInChars) {
CGBuilderTy &Builder = CGF.Builder;
CharUnits baseSize = CGF.getContext().getTypeSizeInChars(baseType);
llvm::Value *baseSizeInChars
= llvm::ConstantInt::get(CGF.IntPtrTy, baseSize.getQuantity());
Address begin =
Builder.CreateElementBitCast(dest, CGF.Int8Ty, "vla.begin");
llvm::Value *end =
Builder.CreateInBoundsGEP(begin.getPointer(), 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.
llvm::PHINode *cur = Builder.CreatePHI(begin.getType(), 2, "vla.cur");
cur->addIncoming(begin.getPointer(), originBB);
CharUnits curAlign =
// memcpy the individual element bit-pattern.
Builder.CreateMemCpy(Address(cur, curAlign), src, baseSizeInChars,
/*volatile*/ false);
// Go to the next element.
llvm::Value *next =
Builder.CreateInBoundsGEP(CGF.Int8Ty, cur, baseSizeInChars, "");
// 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);
CodeGenFunction::EmitNullInitialization(Address DestPtr, QualType Ty) {
// Ignore empty classes in C++.
if (getLangOpts().CPlusPlus) {
if (const RecordType *RT = Ty->getAs<RecordType>()) {
if (cast<CXXRecordDecl>(RT->getDecl())->isEmpty())
// Cast the dest ptr to the appropriate i8 pointer type.
if (DestPtr.getElementType() != Int8Ty)
DestPtr = Builder.CreateElementBitCast(DestPtr, Int8Ty);
// Get size and alignment info for this aggregate.
CharUnits size = getContext().getTypeSizeInChars(Ty);
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 =
getContext().getAsArrayType(Ty))) {
auto VlaSize = getVLASize(vlaType);
SizeVal = VlaSize.NumElts;
CharUnits eltSize = getContext().getTypeSizeInChars(VlaSize.Type);
if (!eltSize.isOne())
SizeVal = Builder.CreateNUWMul(SizeVal, CGM.getSize(eltSize));
vla = vlaType;
} else {
} 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(),
NullConstant, Twine());
CharUnits NullAlign = DestPtr.getAlignment();
Address 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, false);
// 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, false);
llvm::BlockAddress *CodeGenFunction::GetAddrOfLabel(const LabelDecl *L) {
// Make sure that there is a block for the indirect goto.
if (!IndirectBranch)
llvm::BasicBlock *BB = getJumpDestForLabel(L).getBlock();
// Make sure the indirect branch includes all of the address-taken blocks.
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(*this, createBasicBlock("indirectgoto"));
// Create the PHI node that indirect gotos will add entries to.
llvm::Value *DestVal = TmpBuilder.CreatePHI(Int8PtrTy, 0,
// 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,
Address &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<VariableArrayType>(arrayType)) {
numVLAElements = getVLASize(cast<VariableArrayType>(arrayType)).NumElts;
// 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<VariableArrayType>(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<llvm::Value*, 8> gepIndices;
// GEP down to the array type.
llvm::ConstantInt *zero = Builder.getInt32(0);
uint64_t countFromCLAs = 1;
QualType eltType;
llvm::ArrayType *llvmArrayType =
while (llvmArrayType) {
== llvmArrayType->getNumElements());
countFromCLAs *= llvmArrayType->getNumElements();
eltType = arrayType->getElementType();
llvmArrayType =
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 *=
eltType = arrayType->getElementType();
arrayType = getContext().getAsArrayType(eltType);
llvm::Type *baseType = ConvertType(eltType);
addr = Builder.CreateElementBitCast(addr, baseType, "array.begin");
} else {
// Create the actual GEP.
addr = Address(Builder.CreateInBoundsGEP(addr.getPointer(),
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;
CodeGenFunction::VlaSizePair CodeGenFunction::getVLASize(QualType type) {
const VariableArrayType *vla = getContext().getAsVariableArrayType(type);
assert(vla && "type was not a variable array type!");
return getVLASize(vla);
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 { numElements, elementType };
CodeGenFunction::getVLAElements1D(QualType type) {
const VariableArrayType *vla = getContext().getAsVariableArrayType(type);
assert(vla && "type was not a variable array type!");
return getVLAElements1D(vla);
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() };
void CodeGenFunction::EmitVariablyModifiedType(QualType type) {
assert(type->isVariablyModifiedType() &&
"Must pass variably modified type to EmitVLASizes!");
// We're going to walk down into the type and look for VLA
// expressions.
do {
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:
#include "clang/AST/"
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::ObjCTypeParam:
case Type::ObjCObject:
case Type::ObjCInterface:
case Type::ObjCObjectPointer:
llvm_unreachable("type class is never variably-modified!");
case Type::Adjusted:
type = cast<AdjustedType>(ty)->getAdjustedType();
case Type::Decayed:
type = cast<DecayedType>(ty)->getPointeeType();
case Type::Pointer:
type = cast<PointerType>(ty)->getPointeeType();
case Type::BlockPointer:
type = cast<BlockPointerType>(ty)->getPointeeType();
case Type::LValueReference:
case Type::RValueReference:
type = cast<ReferenceType>(ty)->getPointeeType();
case Type::MemberPointer:
type = cast<MemberPointerType>(ty)->getPointeeType();
case Type::ConstantArray:
case Type::IncompleteArray:
// Losing element qualification here is fine.
type = cast<ArrayType>(ty)->getElementType();
case Type::VariableArray: {
// Losing element qualification here is fine.
const VariableArrayType *vat = cast<VariableArrayType>(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
// 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.has(SanitizerKind::VLABound) &&
size->getType()->isSignedIntegerType()) {
SanitizerScope SanScope(this);
llvm::Value *Zero = llvm::Constant::getNullValue(Size->getType());
llvm::Constant *StaticArgs[] = {
EmitCheck(std::make_pair(Builder.CreateICmpSGT(Size, Zero),
SanitizerHandler::VLABoundNotPositive, StaticArgs, Size);
// 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();
case Type::FunctionProto:
case Type::FunctionNoProto:
type = cast<FunctionType>(ty)->getReturnType();
case Type::Paren:
case Type::TypeOf:
case Type::UnaryTransform:
case Type::Attributed:
case Type::SubstTemplateTypeParm:
case Type::PackExpansion:
case Type::MacroQualified:
// Keep walking after single level desugaring.
type = type.getSingleStepDesugaredType(getContext());
case Type::Typedef:
case Type::Decltype:
case Type::Auto:
case Type::DeducedTemplateSpecialization:
// Stop walking: nothing to do.
case Type::TypeOfExpr:
// Stop walking: emit typeof expression.
case Type::Atomic:
type = cast<AtomicType>(ty)->getValueType();
case Type::Pipe:
type = cast<PipeType>(ty)->getElementType();
} while (type->isVariablyModifiedType());
Address CodeGenFunction::EmitVAListRef(const Expr* E) {
if (getContext().getBuiltinVaListType()->isArrayType())
return EmitPointerWithAlignment(E);
return EmitLValue(E).getAddress();
Address CodeGenFunction::EmitMSVAListRef(const Expr *E) {
return EmitLValue(E).getAddress();
void CodeGenFunction::EmitDeclRefExprDbgValue(const DeclRefExpr *E,
const APValue &Init) {
assert(Init.hasValue() && "Invalid DeclRefExpr initializer!");
if (CGDebugInfo *Dbg = getDebugInfo())
if (CGM.getCodeGenOpts().getDebugInfo() >= codegenoptions::LimitedDebugInfo)
Dbg->EmitGlobalVariable(E->getDecl(), Init);
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<llvm::ZExtInst>(value)) return PeepholeProtection();
// Just make an extra bitcast.
llvm::Instruction *inst = new llvm::BitCastInst(value, value->getType(), "",
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.
void CodeGenFunction::EmitAlignmentAssumption(llvm::Value *PtrValue,
QualType Ty, SourceLocation Loc,
SourceLocation AssumptionLoc,
llvm::Value *Alignment,
llvm::Value *OffsetValue) {
llvm::Value *TheCheck;
llvm::Instruction *Assumption = Builder.CreateAlignmentAssumption(
CGM.getDataLayout(), PtrValue, Alignment, OffsetValue, &TheCheck);
if (SanOpts.has(SanitizerKind::Alignment)) {
EmitAlignmentAssumptionCheck(PtrValue, Ty, Loc, AssumptionLoc, Alignment,
OffsetValue, TheCheck, Assumption);
void CodeGenFunction::EmitAlignmentAssumption(llvm::Value *PtrValue,
const Expr *E,
SourceLocation AssumptionLoc,
llvm::Value *Alignment,
llvm::Value *OffsetValue) {
if (auto *CE = dyn_cast<CastExpr>(E))
E = CE->getSubExprAsWritten();
QualType Ty = E->getType();
SourceLocation Loc = E->getExprLoc();
EmitAlignmentAssumption(PtrValue, Ty, Loc, AssumptionLoc, Alignment,
llvm::Value *CodeGenFunction::EmitAnnotationCall(llvm::Function *AnnotationFn,
llvm::Value *AnnotatedVal,
StringRef AnnotationStr,
SourceLocation Location) {
llvm::Value *Args[4] = {
Builder.CreateBitCast(CGM.EmitAnnotationString(AnnotationStr), Int8PtrTy),
Builder.CreateBitCast(CGM.EmitAnnotationUnit(Location), Int8PtrTy),
return Builder.CreateCall(AnnotationFn, Args);
void CodeGenFunction::EmitVarAnnotations(const VarDecl *D, llvm::Value *V) {
assert(D->hasAttr<AnnotateAttr>() && "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<AnnotateAttr>())
Builder.CreateBitCast(V, CGM.Int8PtrTy, V->getName()),
I->getAnnotation(), D->getLocation());
Address CodeGenFunction::EmitFieldAnnotations(const FieldDecl *D,
Address Addr) {
assert(D->hasAttr<AnnotateAttr>() && "no annotate attribute");
llvm::Value *V = Addr.getPointer();
llvm::Type *VTy = V->getType();
llvm::Function *F = CGM.getIntrinsic(llvm::Intrinsic::ptr_annotation,
for (const auto *I : D->specific_attrs<AnnotateAttr>()) {
// 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.CreateBitCast(V, CGM.Int8PtrTy);
V = EmitAnnotationCall(F, V, I->getAnnotation(), D->getLocation());
V = Builder.CreateBitCast(V, VTy);
return Address(V, Addr.getAlignment());
CodeGenFunction::CGCapturedStmtInfo::~CGCapturedStmtInfo() { }
CodeGenFunction::SanitizerScope::SanitizerScope(CodeGenFunction *CGF)
: CGF(CGF) {
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 {
if (IsSanitizerScope)
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);
static bool hasRequiredFeatures(const SmallVectorImpl<StringRef> &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<bool> CallerFeatureMap;
CGM.getFunctionFeatureMap(CallerFeatureMap, GlobalDecl().getWithDecl(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<StringRef, 1> OrFeatures;
Feature.split(OrFeatures, '|');
return llvm::any_of(OrFeatures, [&](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) {
return checkTargetFeatures(E->getBeginLoc(), TargetDecl);
// Emits an error if we don't have a valid set of target features for the
// called function.
void CodeGenFunction::checkTargetFeatures(SourceLocation Loc,
const FunctionDecl *TargetDecl) {
// Early exit if this is an indirect call.
if (!TargetDecl)
// 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<FunctionDecl>(CurCodeDecl);
if (!FD)
// 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<StringRef, 1> ReqFeatures;
const char *FeatureList =
// Return if the builtin doesn't have any required features.
if (!FeatureList || StringRef(FeatureList) == "")
StringRef(FeatureList).split(ReqFeatures, ',');
if (!hasRequiredFeatures(ReqFeatures, CGM, FD, MissingFeature))
CGM.getDiags().Report(Loc, diag::err_builtin_needs_feature)
<< TargetDecl->getDeclName()
<< CGM.getContext().BuiltinInfo.getRequiredFeatures(BuiltinID);
} else if (TargetDecl->hasAttr<TargetAttr>() ||
TargetDecl->hasAttr<CPUSpecificAttr>()) {
// Get the required features for the callee.
const TargetAttr *TD = TargetDecl->getAttr<TargetAttr>();
TargetAttr::ParsedTargetAttr ParsedAttr = CGM.filterFunctionTargetAttrs(TD);
SmallVector<StringRef, 1> ReqFeatures;
llvm::StringMap<bool> CalleeFeatureMap;
CGM.getFunctionFeatureMap(CalleeFeatureMap, TargetDecl);
for (const auto &F : ParsedAttr.Features) {
if (F[0] == '+' && CalleeFeatureMap.lookup(F.substr(1)))
for (const auto &F : CalleeFeatureMap) {
// Only positive features are "required".
if (F.getValue())
if (!hasRequiredFeatures(ReqFeatures, CGM, FD, MissingFeature))
CGM.getDiags().Report(Loc, diag::err_function_needs_feature)
<< FD->getDeclName() << TargetDecl->getDeclName() << MissingFeature;
void CodeGenFunction::EmitSanitizerStatReport(llvm::SanitizerStatKind SSK) {
if (!CGM.getCodeGenOpts().SanitizeStats)
llvm::IRBuilder<> IRB(Builder.GetInsertBlock(), Builder.GetInsertPoint());
CGM.getSanStats().create(IRB, SSK);
llvm::Value *
CodeGenFunction::FormResolverCondition(const MultiVersionResolverOption &RO) {
llvm::Value *Condition = nullptr;
if (!RO.Conditions.Architecture.empty())
Condition = EmitX86CpuIs(RO.Conditions.Architecture);
if (!RO.Conditions.Features.empty()) {
llvm::Value *FeatureCond = EmitX86CpuSupports(RO.Conditions.Features);
Condition =
Condition ? Builder.CreateAnd(Condition, FeatureCond) : FeatureCond;
return Condition;
static void CreateMultiVersionResolverReturn(CodeGenModule &CGM,
llvm::Function *Resolver,
CGBuilderTy &Builder,
llvm::Function *FuncToReturn,
bool SupportsIFunc) {
if (SupportsIFunc) {
llvm::SmallVector<llvm::Value *, 10> Args;
[&](llvm::Argument &Arg) { Args.push_back(&Arg); });
llvm::CallInst *Result = Builder.CreateCall(FuncToReturn, Args);
if (Resolver->getReturnType()->isVoidTy())
void CodeGenFunction::EmitMultiVersionResolver(
llvm::Function *Resolver, ArrayRef<MultiVersionResolverOption> Options) {
assert((getContext().getTargetInfo().getTriple().getArch() ==
llvm::Triple::x86 ||
getContext().getTargetInfo().getTriple().getArch() ==
llvm::Triple::x86_64) &&
"Only implemented for x86 targets");
bool SupportsIFunc = getContext().getTargetInfo().supportsIFunc();
// Main function's basic block.
llvm::BasicBlock *CurBlock = createBasicBlock("resolver_entry", Resolver);
for (const MultiVersionResolverOption &RO : Options) {
llvm::Value *Condition = FormResolverCondition(RO);
// The 'default' or 'generic' case.
if (!Condition) {
assert(&RO == Options.end() - 1 &&
"Default or Generic case must be last");
CreateMultiVersionResolverReturn(CGM, Resolver, Builder, RO.Function,
llvm::BasicBlock *RetBlock = createBasicBlock("resolver_return", Resolver);
CGBuilderTy RetBuilder(*this, RetBlock);
CreateMultiVersionResolverReturn(CGM, Resolver, RetBuilder, RO.Function,
CurBlock = createBasicBlock("resolver_else", Resolver);
Builder.CreateCondBr(Condition, RetBlock, CurBlock);
// If no generic/default, emit an unreachable.
llvm::CallInst *TrapCall = EmitTrapCall(llvm::Intrinsic::trap);
// Loc - where the diagnostic will point, where in the source code this
// alignment has failed.
// SecondaryLoc - if present (will be present if sufficiently different from
// Loc), the diagnostic will additionally point a "Note:" to this location.
// It should be the location where the __attribute__((assume_aligned))
// was written e.g.
void CodeGenFunction::EmitAlignmentAssumptionCheck(
llvm::Value *Ptr, QualType Ty, SourceLocation Loc,
SourceLocation SecondaryLoc, llvm::Value *Alignment,
llvm::Value *OffsetValue, llvm::Value *TheCheck,
llvm::Instruction *Assumption) {
assert(Assumption && isa<llvm::CallInst>(Assumption) &&
cast<llvm::CallInst>(Assumption)->getCalledValue() ==
llvm::Intrinsic::assume) &&
"Assumption should be a call to llvm.assume().");
assert(&(Builder.GetInsertBlock()->back()) == Assumption &&
"Assumption should be the last instruction of the basic block, "
"since the basic block is still being generated.");
if (!SanOpts.has(SanitizerKind::Alignment))
// Don't check pointers to volatile data. The behavior here is implementation-
// defined.
if (Ty->getPointeeType().isVolatileQualified())
// We need to temorairly remove the assumption so we can insert the
// sanitizer check before it, else the check will be dropped by optimizations.
SanitizerScope SanScope(this);
if (!OffsetValue)
OffsetValue = Builder.getInt1(0); // no offset.
llvm::Constant *StaticData[] = {EmitCheckSourceLocation(Loc),
llvm::Value *DynamicData[] = {EmitCheckValue(Ptr),
EmitCheck({std::make_pair(TheCheck, SanitizerKind::Alignment)},
SanitizerHandler::AlignmentAssumption, StaticData, DynamicData);
// We are now in the (new, empty) "cont" basic block.
// Reintroduce the assumption.
// FIXME: Assumption still has it's original basic block as it's Parent.
llvm::DebugLoc CodeGenFunction::SourceLocToDebugLoc(SourceLocation Location) {
if (CGDebugInfo *DI = getDebugInfo())
return DI->SourceLocToDebugLoc(Location);
return llvm::DebugLoc();