blob: 715e8e7aadc63f424928627d6f69238125355de3 [file] [log] [blame]
//===--- CGCall.cpp - Encapsulate calling convention details --------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// These classes wrap the information about a call or function
// definition used to handle ABI compliancy.
//
//===----------------------------------------------------------------------===//
#include "CGCall.h"
#include "ABIInfo.h"
#include "CGCXXABI.h"
#include "CGCleanup.h"
#include "CodeGenFunction.h"
#include "CodeGenModule.h"
#include "TargetInfo.h"
#include "clang/AST/Decl.h"
#include "clang/AST/DeclCXX.h"
#include "clang/AST/DeclObjC.h"
#include "clang/Basic/TargetBuiltins.h"
#include "clang/Basic/TargetInfo.h"
#include "clang/CodeGen/CGFunctionInfo.h"
#include "clang/Frontend/CodeGenOptions.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/IR/Attributes.h"
#include "llvm/IR/CallSite.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/InlineAsm.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/Transforms/Utils/Local.h"
using namespace clang;
using namespace CodeGen;
/***/
static unsigned ClangCallConvToLLVMCallConv(CallingConv CC) {
switch (CC) {
default: return llvm::CallingConv::C;
case CC_X86StdCall: return llvm::CallingConv::X86_StdCall;
case CC_X86FastCall: return llvm::CallingConv::X86_FastCall;
case CC_X86ThisCall: return llvm::CallingConv::X86_ThisCall;
case CC_X86_64Win64: return llvm::CallingConv::X86_64_Win64;
case CC_X86_64SysV: return llvm::CallingConv::X86_64_SysV;
case CC_AAPCS: return llvm::CallingConv::ARM_AAPCS;
case CC_AAPCS_VFP: return llvm::CallingConv::ARM_AAPCS_VFP;
case CC_IntelOclBicc: return llvm::CallingConv::Intel_OCL_BI;
// TODO: Add support for __pascal to LLVM.
case CC_X86Pascal: return llvm::CallingConv::C;
// TODO: Add support for __vectorcall to LLVM.
case CC_X86VectorCall: return llvm::CallingConv::X86_VectorCall;
case CC_SpirFunction: return llvm::CallingConv::SPIR_FUNC;
case CC_SpirKernel: return llvm::CallingConv::SPIR_KERNEL;
}
}
/// Derives the 'this' type for codegen purposes, i.e. ignoring method
/// qualification.
/// FIXME: address space qualification?
static CanQualType GetThisType(ASTContext &Context, const CXXRecordDecl *RD) {
QualType RecTy = Context.getTagDeclType(RD)->getCanonicalTypeInternal();
return Context.getPointerType(CanQualType::CreateUnsafe(RecTy));
}
/// Returns the canonical formal type of the given C++ method.
static CanQual<FunctionProtoType> GetFormalType(const CXXMethodDecl *MD) {
return MD->getType()->getCanonicalTypeUnqualified()
.getAs<FunctionProtoType>();
}
/// Returns the "extra-canonicalized" return type, which discards
/// qualifiers on the return type. Codegen doesn't care about them,
/// and it makes ABI code a little easier to be able to assume that
/// all parameter and return types are top-level unqualified.
static CanQualType GetReturnType(QualType RetTy) {
return RetTy->getCanonicalTypeUnqualified().getUnqualifiedType();
}
/// Arrange the argument and result information for a value of the given
/// unprototyped freestanding function type.
const CGFunctionInfo &
CodeGenTypes::arrangeFreeFunctionType(CanQual<FunctionNoProtoType> FTNP) {
// When translating an unprototyped function type, always use a
// variadic type.
return arrangeLLVMFunctionInfo(FTNP->getReturnType().getUnqualifiedType(),
/*instanceMethod=*/false,
/*chainCall=*/false, None,
FTNP->getExtInfo(), RequiredArgs(0));
}
/// Adds the formal paramaters in FPT to the given prefix. If any parameter in
/// FPT has pass_object_size attrs, then we'll add parameters for those, too.
static void appendParameterTypes(const CodeGenTypes &CGT,
SmallVectorImpl<CanQualType> &prefix,
const CanQual<FunctionProtoType> &FPT,
const FunctionDecl *FD) {
// Fast path: unknown target.
if (FD == nullptr) {
prefix.append(FPT->param_type_begin(), FPT->param_type_end());
return;
}
// In the vast majority cases, we'll have precisely FPT->getNumParams()
// parameters; the only thing that can change this is the presence of
// pass_object_size. So, we preallocate for the common case.
prefix.reserve(prefix.size() + FPT->getNumParams());
assert(FD->getNumParams() == FPT->getNumParams());
for (unsigned I = 0, E = FPT->getNumParams(); I != E; ++I) {
prefix.push_back(FPT->getParamType(I));
if (FD->getParamDecl(I)->hasAttr<PassObjectSizeAttr>())
prefix.push_back(CGT.getContext().getSizeType());
}
}
/// Arrange the LLVM function layout for a value of the given function
/// type, on top of any implicit parameters already stored.
static const CGFunctionInfo &
arrangeLLVMFunctionInfo(CodeGenTypes &CGT, bool instanceMethod,
SmallVectorImpl<CanQualType> &prefix,
CanQual<FunctionProtoType> FTP,
const FunctionDecl *FD) {
RequiredArgs required = RequiredArgs::forPrototypePlus(FTP, prefix.size());
// FIXME: Kill copy.
appendParameterTypes(CGT, prefix, FTP, FD);
CanQualType resultType = FTP->getReturnType().getUnqualifiedType();
return CGT.arrangeLLVMFunctionInfo(resultType, instanceMethod,
/*chainCall=*/false, prefix,
FTP->getExtInfo(), required);
}
/// Arrange the argument and result information for a value of the
/// given freestanding function type.
const CGFunctionInfo &
CodeGenTypes::arrangeFreeFunctionType(CanQual<FunctionProtoType> FTP,
const FunctionDecl *FD) {
SmallVector<CanQualType, 16> argTypes;
return ::arrangeLLVMFunctionInfo(*this, /*instanceMethod=*/false, argTypes,
FTP, FD);
}
static CallingConv getCallingConventionForDecl(const Decl *D, bool IsWindows) {
// Set the appropriate calling convention for the Function.
if (D->hasAttr<StdCallAttr>())
return CC_X86StdCall;
if (D->hasAttr<FastCallAttr>())
return CC_X86FastCall;
if (D->hasAttr<ThisCallAttr>())
return CC_X86ThisCall;
if (D->hasAttr<VectorCallAttr>())
return CC_X86VectorCall;
if (D->hasAttr<PascalAttr>())
return CC_X86Pascal;
if (PcsAttr *PCS = D->getAttr<PcsAttr>())
return (PCS->getPCS() == PcsAttr::AAPCS ? CC_AAPCS : CC_AAPCS_VFP);
if (D->hasAttr<IntelOclBiccAttr>())
return CC_IntelOclBicc;
if (D->hasAttr<MSABIAttr>())
return IsWindows ? CC_C : CC_X86_64Win64;
if (D->hasAttr<SysVABIAttr>())
return IsWindows ? CC_X86_64SysV : CC_C;
return CC_C;
}
/// Arrange the argument and result information for a call to an
/// unknown C++ non-static member function of the given abstract type.
/// (Zero value of RD means we don't have any meaningful "this" argument type,
/// so fall back to a generic pointer type).
/// The member function must be an ordinary function, i.e. not a
/// constructor or destructor.
const CGFunctionInfo &
CodeGenTypes::arrangeCXXMethodType(const CXXRecordDecl *RD,
const FunctionProtoType *FTP,
const CXXMethodDecl *MD) {
SmallVector<CanQualType, 16> argTypes;
// Add the 'this' pointer.
if (RD)
argTypes.push_back(GetThisType(Context, RD));
else
argTypes.push_back(Context.VoidPtrTy);
return ::arrangeLLVMFunctionInfo(
*this, true, argTypes,
FTP->getCanonicalTypeUnqualified().getAs<FunctionProtoType>(), MD);
}
/// Arrange the argument and result information for a declaration or
/// definition of the given C++ non-static member function. The
/// member function must be an ordinary function, i.e. not a
/// constructor or destructor.
const CGFunctionInfo &
CodeGenTypes::arrangeCXXMethodDeclaration(const CXXMethodDecl *MD) {
assert(!isa<CXXConstructorDecl>(MD) && "wrong method for constructors!");
assert(!isa<CXXDestructorDecl>(MD) && "wrong method for destructors!");
CanQual<FunctionProtoType> prototype = GetFormalType(MD);
if (MD->isInstance()) {
// The abstract case is perfectly fine.
const CXXRecordDecl *ThisType = TheCXXABI.getThisArgumentTypeForMethod(MD);
return arrangeCXXMethodType(ThisType, prototype.getTypePtr(), MD);
}
return arrangeFreeFunctionType(prototype, MD);
}
const CGFunctionInfo &
CodeGenTypes::arrangeCXXStructorDeclaration(const CXXMethodDecl *MD,
StructorType Type) {
SmallVector<CanQualType, 16> argTypes;
argTypes.push_back(GetThisType(Context, MD->getParent()));
GlobalDecl GD;
if (auto *CD = dyn_cast<CXXConstructorDecl>(MD)) {
GD = GlobalDecl(CD, toCXXCtorType(Type));
} else {
auto *DD = dyn_cast<CXXDestructorDecl>(MD);
GD = GlobalDecl(DD, toCXXDtorType(Type));
}
CanQual<FunctionProtoType> FTP = GetFormalType(MD);
// Add the formal parameters.
appendParameterTypes(*this, argTypes, FTP, MD);
TheCXXABI.buildStructorSignature(MD, Type, argTypes);
RequiredArgs required =
(MD->isVariadic() ? RequiredArgs(argTypes.size()) : RequiredArgs::All);
FunctionType::ExtInfo extInfo = FTP->getExtInfo();
CanQualType resultType = TheCXXABI.HasThisReturn(GD)
? argTypes.front()
: TheCXXABI.hasMostDerivedReturn(GD)
? CGM.getContext().VoidPtrTy
: Context.VoidTy;
return arrangeLLVMFunctionInfo(resultType, /*instanceMethod=*/true,
/*chainCall=*/false, argTypes, extInfo,
required);
}
/// Arrange a call to a C++ method, passing the given arguments.
const CGFunctionInfo &
CodeGenTypes::arrangeCXXConstructorCall(const CallArgList &args,
const CXXConstructorDecl *D,
CXXCtorType CtorKind,
unsigned ExtraArgs) {
// FIXME: Kill copy.
SmallVector<CanQualType, 16> ArgTypes;
for (const auto &Arg : args)
ArgTypes.push_back(Context.getCanonicalParamType(Arg.Ty));
CanQual<FunctionProtoType> FPT = GetFormalType(D);
RequiredArgs Required = RequiredArgs::forPrototypePlus(FPT, 1 + ExtraArgs);
GlobalDecl GD(D, CtorKind);
CanQualType ResultType = TheCXXABI.HasThisReturn(GD)
? ArgTypes.front()
: TheCXXABI.hasMostDerivedReturn(GD)
? CGM.getContext().VoidPtrTy
: Context.VoidTy;
FunctionType::ExtInfo Info = FPT->getExtInfo();
return arrangeLLVMFunctionInfo(ResultType, /*instanceMethod=*/true,
/*chainCall=*/false, ArgTypes, Info,
Required);
}
/// Arrange the argument and result information for the declaration or
/// definition of the given function.
const CGFunctionInfo &
CodeGenTypes::arrangeFunctionDeclaration(const FunctionDecl *FD) {
if (const CXXMethodDecl *MD = dyn_cast<CXXMethodDecl>(FD))
if (MD->isInstance())
return arrangeCXXMethodDeclaration(MD);
CanQualType FTy = FD->getType()->getCanonicalTypeUnqualified();
assert(isa<FunctionType>(FTy));
// When declaring a function without a prototype, always use a
// non-variadic type.
if (isa<FunctionNoProtoType>(FTy)) {
CanQual<FunctionNoProtoType> noProto = FTy.getAs<FunctionNoProtoType>();
return arrangeLLVMFunctionInfo(
noProto->getReturnType(), /*instanceMethod=*/false,
/*chainCall=*/false, None, noProto->getExtInfo(), RequiredArgs::All);
}
assert(isa<FunctionProtoType>(FTy));
return arrangeFreeFunctionType(FTy.getAs<FunctionProtoType>(), FD);
}
/// Arrange the argument and result information for the declaration or
/// definition of an Objective-C method.
const CGFunctionInfo &
CodeGenTypes::arrangeObjCMethodDeclaration(const ObjCMethodDecl *MD) {
// It happens that this is the same as a call with no optional
// arguments, except also using the formal 'self' type.
return arrangeObjCMessageSendSignature(MD, MD->getSelfDecl()->getType());
}
/// Arrange the argument and result information for the function type
/// through which to perform a send to the given Objective-C method,
/// using the given receiver type. The receiver type is not always
/// the 'self' type of the method or even an Objective-C pointer type.
/// This is *not* the right method for actually performing such a
/// message send, due to the possibility of optional arguments.
const CGFunctionInfo &
CodeGenTypes::arrangeObjCMessageSendSignature(const ObjCMethodDecl *MD,
QualType receiverType) {
SmallVector<CanQualType, 16> argTys;
argTys.push_back(Context.getCanonicalParamType(receiverType));
argTys.push_back(Context.getCanonicalParamType(Context.getObjCSelType()));
// FIXME: Kill copy?
for (const auto *I : MD->params()) {
argTys.push_back(Context.getCanonicalParamType(I->getType()));
}
FunctionType::ExtInfo einfo;
bool IsWindows = getContext().getTargetInfo().getTriple().isOSWindows();
einfo = einfo.withCallingConv(getCallingConventionForDecl(MD, IsWindows));
if (getContext().getLangOpts().ObjCAutoRefCount &&
MD->hasAttr<NSReturnsRetainedAttr>())
einfo = einfo.withProducesResult(true);
RequiredArgs required =
(MD->isVariadic() ? RequiredArgs(argTys.size()) : RequiredArgs::All);
return arrangeLLVMFunctionInfo(
GetReturnType(MD->getReturnType()), /*instanceMethod=*/false,
/*chainCall=*/false, argTys, einfo, required);
}
const CGFunctionInfo &
CodeGenTypes::arrangeGlobalDeclaration(GlobalDecl GD) {
// FIXME: Do we need to handle ObjCMethodDecl?
const FunctionDecl *FD = cast<FunctionDecl>(GD.getDecl());
if (const CXXConstructorDecl *CD = dyn_cast<CXXConstructorDecl>(FD))
return arrangeCXXStructorDeclaration(CD, getFromCtorType(GD.getCtorType()));
if (const CXXDestructorDecl *DD = dyn_cast<CXXDestructorDecl>(FD))
return arrangeCXXStructorDeclaration(DD, getFromDtorType(GD.getDtorType()));
return arrangeFunctionDeclaration(FD);
}
/// Arrange a thunk that takes 'this' as the first parameter followed by
/// varargs. Return a void pointer, regardless of the actual return type.
/// The body of the thunk will end in a musttail call to a function of the
/// correct type, and the caller will bitcast the function to the correct
/// prototype.
const CGFunctionInfo &
CodeGenTypes::arrangeMSMemberPointerThunk(const CXXMethodDecl *MD) {
assert(MD->isVirtual() && "only virtual memptrs have thunks");
CanQual<FunctionProtoType> FTP = GetFormalType(MD);
CanQualType ArgTys[] = { GetThisType(Context, MD->getParent()) };
return arrangeLLVMFunctionInfo(Context.VoidTy, /*instanceMethod=*/false,
/*chainCall=*/false, ArgTys,
FTP->getExtInfo(), RequiredArgs(1));
}
const CGFunctionInfo &
CodeGenTypes::arrangeMSCtorClosure(const CXXConstructorDecl *CD,
CXXCtorType CT) {
assert(CT == Ctor_CopyingClosure || CT == Ctor_DefaultClosure);
CanQual<FunctionProtoType> FTP = GetFormalType(CD);
SmallVector<CanQualType, 2> ArgTys;
const CXXRecordDecl *RD = CD->getParent();
ArgTys.push_back(GetThisType(Context, RD));
if (CT == Ctor_CopyingClosure)
ArgTys.push_back(*FTP->param_type_begin());
if (RD->getNumVBases() > 0)
ArgTys.push_back(Context.IntTy);
CallingConv CC = Context.getDefaultCallingConvention(
/*IsVariadic=*/false, /*IsCXXMethod=*/true);
return arrangeLLVMFunctionInfo(Context.VoidTy, /*instanceMethod=*/true,
/*chainCall=*/false, ArgTys,
FunctionType::ExtInfo(CC), RequiredArgs::All);
}
/// Arrange a call as unto a free function, except possibly with an
/// additional number of formal parameters considered required.
static const CGFunctionInfo &
arrangeFreeFunctionLikeCall(CodeGenTypes &CGT,
CodeGenModule &CGM,
const CallArgList &args,
const FunctionType *fnType,
unsigned numExtraRequiredArgs,
bool chainCall) {
assert(args.size() >= numExtraRequiredArgs);
// In most cases, there are no optional arguments.
RequiredArgs required = RequiredArgs::All;
// If we have a variadic prototype, the required arguments are the
// extra prefix plus the arguments in the prototype.
if (const FunctionProtoType *proto = dyn_cast<FunctionProtoType>(fnType)) {
if (proto->isVariadic())
required = RequiredArgs(proto->getNumParams() + numExtraRequiredArgs);
// If we don't have a prototype at all, but we're supposed to
// explicitly use the variadic convention for unprototyped calls,
// treat all of the arguments as required but preserve the nominal
// possibility of variadics.
} else if (CGM.getTargetCodeGenInfo()
.isNoProtoCallVariadic(args,
cast<FunctionNoProtoType>(fnType))) {
required = RequiredArgs(args.size());
}
// FIXME: Kill copy.
SmallVector<CanQualType, 16> argTypes;
for (const auto &arg : args)
argTypes.push_back(CGT.getContext().getCanonicalParamType(arg.Ty));
return CGT.arrangeLLVMFunctionInfo(GetReturnType(fnType->getReturnType()),
/*instanceMethod=*/false, chainCall,
argTypes, fnType->getExtInfo(), required);
}
/// Figure out the rules for calling a function with the given formal
/// type using the given arguments. The arguments are necessary
/// because the function might be unprototyped, in which case it's
/// target-dependent in crazy ways.
const CGFunctionInfo &
CodeGenTypes::arrangeFreeFunctionCall(const CallArgList &args,
const FunctionType *fnType,
bool chainCall) {
return arrangeFreeFunctionLikeCall(*this, CGM, args, fnType,
chainCall ? 1 : 0, chainCall);
}
/// A block function call is essentially a free-function call with an
/// extra implicit argument.
const CGFunctionInfo &
CodeGenTypes::arrangeBlockFunctionCall(const CallArgList &args,
const FunctionType *fnType) {
return arrangeFreeFunctionLikeCall(*this, CGM, args, fnType, 1,
/*chainCall=*/false);
}
const CGFunctionInfo &
CodeGenTypes::arrangeFreeFunctionCall(QualType resultType,
const CallArgList &args,
FunctionType::ExtInfo info,
RequiredArgs required) {
// FIXME: Kill copy.
SmallVector<CanQualType, 16> argTypes;
for (const auto &Arg : args)
argTypes.push_back(Context.getCanonicalParamType(Arg.Ty));
return arrangeLLVMFunctionInfo(
GetReturnType(resultType), /*instanceMethod=*/false,
/*chainCall=*/false, argTypes, info, required);
}
/// Arrange a call to a C++ method, passing the given arguments.
const CGFunctionInfo &
CodeGenTypes::arrangeCXXMethodCall(const CallArgList &args,
const FunctionProtoType *FPT,
RequiredArgs required) {
// FIXME: Kill copy.
SmallVector<CanQualType, 16> argTypes;
for (const auto &Arg : args)
argTypes.push_back(Context.getCanonicalParamType(Arg.Ty));
FunctionType::ExtInfo info = FPT->getExtInfo();
return arrangeLLVMFunctionInfo(
GetReturnType(FPT->getReturnType()), /*instanceMethod=*/true,
/*chainCall=*/false, argTypes, info, required);
}
const CGFunctionInfo &CodeGenTypes::arrangeFreeFunctionDeclaration(
QualType resultType, const FunctionArgList &args,
const FunctionType::ExtInfo &info, bool isVariadic) {
// FIXME: Kill copy.
SmallVector<CanQualType, 16> argTypes;
for (auto Arg : args)
argTypes.push_back(Context.getCanonicalParamType(Arg->getType()));
RequiredArgs required =
(isVariadic ? RequiredArgs(args.size()) : RequiredArgs::All);
return arrangeLLVMFunctionInfo(
GetReturnType(resultType), /*instanceMethod=*/false,
/*chainCall=*/false, argTypes, info, required);
}
const CGFunctionInfo &CodeGenTypes::arrangeNullaryFunction() {
return arrangeLLVMFunctionInfo(
getContext().VoidTy, /*instanceMethod=*/false, /*chainCall=*/false,
None, FunctionType::ExtInfo(), RequiredArgs::All);
}
/// Arrange the argument and result information for an abstract value
/// of a given function type. This is the method which all of the
/// above functions ultimately defer to.
const CGFunctionInfo &
CodeGenTypes::arrangeLLVMFunctionInfo(CanQualType resultType,
bool instanceMethod,
bool chainCall,
ArrayRef<CanQualType> argTypes,
FunctionType::ExtInfo info,
RequiredArgs required) {
assert(std::all_of(argTypes.begin(), argTypes.end(),
std::mem_fun_ref(&CanQualType::isCanonicalAsParam)));
unsigned CC = ClangCallConvToLLVMCallConv(info.getCC());
// Lookup or create unique function info.
llvm::FoldingSetNodeID ID;
CGFunctionInfo::Profile(ID, instanceMethod, chainCall, info, required,
resultType, argTypes);
void *insertPos = nullptr;
CGFunctionInfo *FI = FunctionInfos.FindNodeOrInsertPos(ID, insertPos);
if (FI)
return *FI;
// Construct the function info. We co-allocate the ArgInfos.
FI = CGFunctionInfo::create(CC, instanceMethod, chainCall, info,
resultType, argTypes, required);
FunctionInfos.InsertNode(FI, insertPos);
bool inserted = FunctionsBeingProcessed.insert(FI).second;
(void)inserted;
assert(inserted && "Recursively being processed?");
// Compute ABI information.
getABIInfo().computeInfo(*FI);
// Loop over all of the computed argument and return value info. If any of
// them are direct or extend without a specified coerce type, specify the
// default now.
ABIArgInfo &retInfo = FI->getReturnInfo();
if (retInfo.canHaveCoerceToType() && retInfo.getCoerceToType() == nullptr)
retInfo.setCoerceToType(ConvertType(FI->getReturnType()));
for (auto &I : FI->arguments())
if (I.info.canHaveCoerceToType() && I.info.getCoerceToType() == nullptr)
I.info.setCoerceToType(ConvertType(I.type));
bool erased = FunctionsBeingProcessed.erase(FI); (void)erased;
assert(erased && "Not in set?");
return *FI;
}
CGFunctionInfo *CGFunctionInfo::create(unsigned llvmCC,
bool instanceMethod,
bool chainCall,
const FunctionType::ExtInfo &info,
CanQualType resultType,
ArrayRef<CanQualType> argTypes,
RequiredArgs required) {
void *buffer = operator new(sizeof(CGFunctionInfo) +
sizeof(ArgInfo) * (argTypes.size() + 1));
CGFunctionInfo *FI = new(buffer) CGFunctionInfo();
FI->CallingConvention = llvmCC;
FI->EffectiveCallingConvention = llvmCC;
FI->ASTCallingConvention = info.getCC();
FI->InstanceMethod = instanceMethod;
FI->ChainCall = chainCall;
FI->NoReturn = info.getNoReturn();
FI->ReturnsRetained = info.getProducesResult();
FI->Required = required;
FI->HasRegParm = info.getHasRegParm();
FI->RegParm = info.getRegParm();
FI->ArgStruct = nullptr;
FI->ArgStructAlign = 0;
FI->NumArgs = argTypes.size();
FI->getArgsBuffer()[0].type = resultType;
for (unsigned i = 0, e = argTypes.size(); i != e; ++i)
FI->getArgsBuffer()[i + 1].type = argTypes[i];
return FI;
}
/***/
namespace {
// ABIArgInfo::Expand implementation.
// Specifies the way QualType passed as ABIArgInfo::Expand is expanded.
struct TypeExpansion {
enum TypeExpansionKind {
// Elements of constant arrays are expanded recursively.
TEK_ConstantArray,
// Record fields are expanded recursively (but if record is a union, only
// the field with the largest size is expanded).
TEK_Record,
// For complex types, real and imaginary parts are expanded recursively.
TEK_Complex,
// All other types are not expandable.
TEK_None
};
const TypeExpansionKind Kind;
TypeExpansion(TypeExpansionKind K) : Kind(K) {}
virtual ~TypeExpansion() {}
};
struct ConstantArrayExpansion : TypeExpansion {
QualType EltTy;
uint64_t NumElts;
ConstantArrayExpansion(QualType EltTy, uint64_t NumElts)
: TypeExpansion(TEK_ConstantArray), EltTy(EltTy), NumElts(NumElts) {}
static bool classof(const TypeExpansion *TE) {
return TE->Kind == TEK_ConstantArray;
}
};
struct RecordExpansion : TypeExpansion {
SmallVector<const CXXBaseSpecifier *, 1> Bases;
SmallVector<const FieldDecl *, 1> Fields;
RecordExpansion(SmallVector<const CXXBaseSpecifier *, 1> &&Bases,
SmallVector<const FieldDecl *, 1> &&Fields)
: TypeExpansion(TEK_Record), Bases(Bases), Fields(Fields) {}
static bool classof(const TypeExpansion *TE) {
return TE->Kind == TEK_Record;
}
};
struct ComplexExpansion : TypeExpansion {
QualType EltTy;
ComplexExpansion(QualType EltTy) : TypeExpansion(TEK_Complex), EltTy(EltTy) {}
static bool classof(const TypeExpansion *TE) {
return TE->Kind == TEK_Complex;
}
};
struct NoExpansion : TypeExpansion {
NoExpansion() : TypeExpansion(TEK_None) {}
static bool classof(const TypeExpansion *TE) {
return TE->Kind == TEK_None;
}
};
} // namespace
static std::unique_ptr<TypeExpansion>
getTypeExpansion(QualType Ty, const ASTContext &Context) {
if (const ConstantArrayType *AT = Context.getAsConstantArrayType(Ty)) {
return llvm::make_unique<ConstantArrayExpansion>(
AT->getElementType(), AT->getSize().getZExtValue());
}
if (const RecordType *RT = Ty->getAs<RecordType>()) {
SmallVector<const CXXBaseSpecifier *, 1> Bases;
SmallVector<const FieldDecl *, 1> Fields;
const RecordDecl *RD = RT->getDecl();
assert(!RD->hasFlexibleArrayMember() &&
"Cannot expand structure with flexible array.");
if (RD->isUnion()) {
// Unions can be here only in degenerative cases - all the fields are same
// after flattening. Thus we have to use the "largest" field.
const FieldDecl *LargestFD = nullptr;
CharUnits UnionSize = CharUnits::Zero();
for (const auto *FD : RD->fields()) {
// Skip zero length bitfields.
if (FD->isBitField() && FD->getBitWidthValue(Context) == 0)
continue;
assert(!FD->isBitField() &&
"Cannot expand structure with bit-field members.");
CharUnits FieldSize = Context.getTypeSizeInChars(FD->getType());
if (UnionSize < FieldSize) {
UnionSize = FieldSize;
LargestFD = FD;
}
}
if (LargestFD)
Fields.push_back(LargestFD);
} else {
if (const auto *CXXRD = dyn_cast<CXXRecordDecl>(RD)) {
assert(!CXXRD->isDynamicClass() &&
"cannot expand vtable pointers in dynamic classes");
for (const CXXBaseSpecifier &BS : CXXRD->bases())
Bases.push_back(&BS);
}
for (const auto *FD : RD->fields()) {
// Skip zero length bitfields.
if (FD->isBitField() && FD->getBitWidthValue(Context) == 0)
continue;
assert(!FD->isBitField() &&
"Cannot expand structure with bit-field members.");
Fields.push_back(FD);
}
}
return llvm::make_unique<RecordExpansion>(std::move(Bases),
std::move(Fields));
}
if (const ComplexType *CT = Ty->getAs<ComplexType>()) {
return llvm::make_unique<ComplexExpansion>(CT->getElementType());
}
return llvm::make_unique<NoExpansion>();
}
static int getExpansionSize(QualType Ty, const ASTContext &Context) {
auto Exp = getTypeExpansion(Ty, Context);
if (auto CAExp = dyn_cast<ConstantArrayExpansion>(Exp.get())) {
return CAExp->NumElts * getExpansionSize(CAExp->EltTy, Context);
}
if (auto RExp = dyn_cast<RecordExpansion>(Exp.get())) {
int Res = 0;
for (auto BS : RExp->Bases)
Res += getExpansionSize(BS->getType(), Context);
for (auto FD : RExp->Fields)
Res += getExpansionSize(FD->getType(), Context);
return Res;
}
if (isa<ComplexExpansion>(Exp.get()))
return 2;
assert(isa<NoExpansion>(Exp.get()));
return 1;
}
void
CodeGenTypes::getExpandedTypes(QualType Ty,
SmallVectorImpl<llvm::Type *>::iterator &TI) {
auto Exp = getTypeExpansion(Ty, Context);
if (auto CAExp = dyn_cast<ConstantArrayExpansion>(Exp.get())) {
for (int i = 0, n = CAExp->NumElts; i < n; i++) {
getExpandedTypes(CAExp->EltTy, TI);
}
} else if (auto RExp = dyn_cast<RecordExpansion>(Exp.get())) {
for (auto BS : RExp->Bases)
getExpandedTypes(BS->getType(), TI);
for (auto FD : RExp->Fields)
getExpandedTypes(FD->getType(), TI);
} else if (auto CExp = dyn_cast<ComplexExpansion>(Exp.get())) {
llvm::Type *EltTy = ConvertType(CExp->EltTy);
*TI++ = EltTy;
*TI++ = EltTy;
} else {
assert(isa<NoExpansion>(Exp.get()));
*TI++ = ConvertType(Ty);
}
}
static void forConstantArrayExpansion(CodeGenFunction &CGF,
ConstantArrayExpansion *CAE,
Address BaseAddr,
llvm::function_ref<void(Address)> Fn) {
CharUnits EltSize = CGF.getContext().getTypeSizeInChars(CAE->EltTy);
CharUnits EltAlign =
BaseAddr.getAlignment().alignmentOfArrayElement(EltSize);
for (int i = 0, n = CAE->NumElts; i < n; i++) {
llvm::Value *EltAddr =
CGF.Builder.CreateConstGEP2_32(nullptr, BaseAddr.getPointer(), 0, i);
Fn(Address(EltAddr, EltAlign));
}
}
void CodeGenFunction::ExpandTypeFromArgs(
QualType Ty, LValue LV, SmallVectorImpl<llvm::Argument *>::iterator &AI) {
assert(LV.isSimple() &&
"Unexpected non-simple lvalue during struct expansion.");
auto Exp = getTypeExpansion(Ty, getContext());
if (auto CAExp = dyn_cast<ConstantArrayExpansion>(Exp.get())) {
forConstantArrayExpansion(*this, CAExp, LV.getAddress(),
[&](Address EltAddr) {
LValue LV = MakeAddrLValue(EltAddr, CAExp->EltTy);
ExpandTypeFromArgs(CAExp->EltTy, LV, AI);
});
} else if (auto RExp = dyn_cast<RecordExpansion>(Exp.get())) {
Address This = LV.getAddress();
for (const CXXBaseSpecifier *BS : RExp->Bases) {
// Perform a single step derived-to-base conversion.
Address Base =
GetAddressOfBaseClass(This, Ty->getAsCXXRecordDecl(), &BS, &BS + 1,
/*NullCheckValue=*/false, SourceLocation());
LValue SubLV = MakeAddrLValue(Base, BS->getType());
// Recurse onto bases.
ExpandTypeFromArgs(BS->getType(), SubLV, AI);
}
for (auto FD : RExp->Fields) {
// FIXME: What are the right qualifiers here?
LValue SubLV = EmitLValueForField(LV, FD);
ExpandTypeFromArgs(FD->getType(), SubLV, AI);
}
} else if (isa<ComplexExpansion>(Exp.get())) {
auto realValue = *AI++;
auto imagValue = *AI++;
EmitStoreOfComplex(ComplexPairTy(realValue, imagValue), LV, /*init*/ true);
} else {
assert(isa<NoExpansion>(Exp.get()));
EmitStoreThroughLValue(RValue::get(*AI++), LV);
}
}
void CodeGenFunction::ExpandTypeToArgs(
QualType Ty, RValue RV, llvm::FunctionType *IRFuncTy,
SmallVectorImpl<llvm::Value *> &IRCallArgs, unsigned &IRCallArgPos) {
auto Exp = getTypeExpansion(Ty, getContext());
if (auto CAExp = dyn_cast<ConstantArrayExpansion>(Exp.get())) {
forConstantArrayExpansion(*this, CAExp, RV.getAggregateAddress(),
[&](Address EltAddr) {
RValue EltRV =
convertTempToRValue(EltAddr, CAExp->EltTy, SourceLocation());
ExpandTypeToArgs(CAExp->EltTy, EltRV, IRFuncTy, IRCallArgs, IRCallArgPos);
});
} else if (auto RExp = dyn_cast<RecordExpansion>(Exp.get())) {
Address This = RV.getAggregateAddress();
for (const CXXBaseSpecifier *BS : RExp->Bases) {
// Perform a single step derived-to-base conversion.
Address Base =
GetAddressOfBaseClass(This, Ty->getAsCXXRecordDecl(), &BS, &BS + 1,
/*NullCheckValue=*/false, SourceLocation());
RValue BaseRV = RValue::getAggregate(Base);
// Recurse onto bases.
ExpandTypeToArgs(BS->getType(), BaseRV, IRFuncTy, IRCallArgs,
IRCallArgPos);
}
LValue LV = MakeAddrLValue(This, Ty);
for (auto FD : RExp->Fields) {
RValue FldRV = EmitRValueForField(LV, FD, SourceLocation());
ExpandTypeToArgs(FD->getType(), FldRV, IRFuncTy, IRCallArgs,
IRCallArgPos);
}
} else if (isa<ComplexExpansion>(Exp.get())) {
ComplexPairTy CV = RV.getComplexVal();
IRCallArgs[IRCallArgPos++] = CV.first;
IRCallArgs[IRCallArgPos++] = CV.second;
} else {
assert(isa<NoExpansion>(Exp.get()));
assert(RV.isScalar() &&
"Unexpected non-scalar rvalue during struct expansion.");
// Insert a bitcast as needed.
llvm::Value *V = RV.getScalarVal();
if (IRCallArgPos < IRFuncTy->getNumParams() &&
V->getType() != IRFuncTy->getParamType(IRCallArgPos))
V = Builder.CreateBitCast(V, IRFuncTy->getParamType(IRCallArgPos));
IRCallArgs[IRCallArgPos++] = V;
}
}
/// Create a temporary allocation for the purposes of coercion.
static Address CreateTempAllocaForCoercion(CodeGenFunction &CGF, llvm::Type *Ty,
CharUnits MinAlign) {
// Don't use an alignment that's worse than what LLVM would prefer.
auto PrefAlign = CGF.CGM.getDataLayout().getPrefTypeAlignment(Ty);
CharUnits Align = std::max(MinAlign, CharUnits::fromQuantity(PrefAlign));
return CGF.CreateTempAlloca(Ty, Align);
}
/// EnterStructPointerForCoercedAccess - Given a struct pointer that we are
/// accessing some number of bytes out of it, try to gep into the struct to get
/// at its inner goodness. Dive as deep as possible without entering an element
/// with an in-memory size smaller than DstSize.
static Address
EnterStructPointerForCoercedAccess(Address SrcPtr,
llvm::StructType *SrcSTy,
uint64_t DstSize, CodeGenFunction &CGF) {
// We can't dive into a zero-element struct.
if (SrcSTy->getNumElements() == 0) return SrcPtr;
llvm::Type *FirstElt = SrcSTy->getElementType(0);
// If the first elt is at least as large as what we're looking for, or if the
// first element is the same size as the whole struct, we can enter it. The
// comparison must be made on the store size and not the alloca size. Using
// the alloca size may overstate the size of the load.
uint64_t FirstEltSize =
CGF.CGM.getDataLayout().getTypeStoreSize(FirstElt);
if (FirstEltSize < DstSize &&
FirstEltSize < CGF.CGM.getDataLayout().getTypeStoreSize(SrcSTy))
return SrcPtr;
// GEP into the first element.
SrcPtr = CGF.Builder.CreateStructGEP(SrcPtr, 0, CharUnits(), "coerce.dive");
// If the first element is a struct, recurse.
llvm::Type *SrcTy = SrcPtr.getElementType();
if (llvm::StructType *SrcSTy = dyn_cast<llvm::StructType>(SrcTy))
return EnterStructPointerForCoercedAccess(SrcPtr, SrcSTy, DstSize, CGF);
return SrcPtr;
}
/// CoerceIntOrPtrToIntOrPtr - Convert a value Val to the specific Ty where both
/// are either integers or pointers. This does a truncation of the value if it
/// is too large or a zero extension if it is too small.
///
/// This behaves as if the value were coerced through memory, so on big-endian
/// targets the high bits are preserved in a truncation, while little-endian
/// targets preserve the low bits.
static llvm::Value *CoerceIntOrPtrToIntOrPtr(llvm::Value *Val,
llvm::Type *Ty,
CodeGenFunction &CGF) {
if (Val->getType() == Ty)
return Val;
if (isa<llvm::PointerType>(Val->getType())) {
// If this is Pointer->Pointer avoid conversion to and from int.
if (isa<llvm::PointerType>(Ty))
return CGF.Builder.CreateBitCast(Val, Ty, "coerce.val");
// Convert the pointer to an integer so we can play with its width.
Val = CGF.Builder.CreatePtrToInt(Val, CGF.IntPtrTy, "coerce.val.pi");
}
llvm::Type *DestIntTy = Ty;
if (isa<llvm::PointerType>(DestIntTy))
DestIntTy = CGF.IntPtrTy;
if (Val->getType() != DestIntTy) {
const llvm::DataLayout &DL = CGF.CGM.getDataLayout();
if (DL.isBigEndian()) {
// Preserve the high bits on big-endian targets.
// That is what memory coercion does.
uint64_t SrcSize = DL.getTypeSizeInBits(Val->getType());
uint64_t DstSize = DL.getTypeSizeInBits(DestIntTy);
if (SrcSize > DstSize) {
Val = CGF.Builder.CreateLShr(Val, SrcSize - DstSize, "coerce.highbits");
Val = CGF.Builder.CreateTrunc(Val, DestIntTy, "coerce.val.ii");
} else {
Val = CGF.Builder.CreateZExt(Val, DestIntTy, "coerce.val.ii");
Val = CGF.Builder.CreateShl(Val, DstSize - SrcSize, "coerce.highbits");
}
} else {
// Little-endian targets preserve the low bits. No shifts required.
Val = CGF.Builder.CreateIntCast(Val, DestIntTy, false, "coerce.val.ii");
}
}
if (isa<llvm::PointerType>(Ty))
Val = CGF.Builder.CreateIntToPtr(Val, Ty, "coerce.val.ip");
return Val;
}
/// CreateCoercedLoad - Create a load from \arg SrcPtr interpreted as
/// a pointer to an object of type \arg Ty, known to be aligned to
/// \arg SrcAlign bytes.
///
/// This safely handles the case when the src type is smaller than the
/// destination type; in this situation the values of bits which not
/// present in the src are undefined.
static llvm::Value *CreateCoercedLoad(Address Src, llvm::Type *Ty,
CodeGenFunction &CGF) {
llvm::Type *SrcTy = Src.getElementType();
// If SrcTy and Ty are the same, just do a load.
if (SrcTy == Ty)
return CGF.Builder.CreateLoad(Src);
uint64_t DstSize = CGF.CGM.getDataLayout().getTypeAllocSize(Ty);
if (llvm::StructType *SrcSTy = dyn_cast<llvm::StructType>(SrcTy)) {
Src = EnterStructPointerForCoercedAccess(Src, SrcSTy, DstSize, CGF);
SrcTy = Src.getType()->getElementType();
}
uint64_t SrcSize = CGF.CGM.getDataLayout().getTypeAllocSize(SrcTy);
// If the source and destination are integer or pointer types, just do an
// extension or truncation to the desired type.
if ((isa<llvm::IntegerType>(Ty) || isa<llvm::PointerType>(Ty)) &&
(isa<llvm::IntegerType>(SrcTy) || isa<llvm::PointerType>(SrcTy))) {
llvm::Value *Load = CGF.Builder.CreateLoad(Src);
return CoerceIntOrPtrToIntOrPtr(Load, Ty, CGF);
}
// If load is legal, just bitcast the src pointer.
if (SrcSize >= DstSize) {
// Generally SrcSize is never greater than DstSize, since this means we are
// losing bits. However, this can happen in cases where the structure has
// additional padding, for example due to a user specified alignment.
//
// FIXME: Assert that we aren't truncating non-padding bits when have access
// to that information.
Src = CGF.Builder.CreateBitCast(Src, llvm::PointerType::getUnqual(Ty));
return CGF.Builder.CreateLoad(Src);
}
// Otherwise do coercion through memory. This is stupid, but simple.
Address Tmp = CreateTempAllocaForCoercion(CGF, Ty, Src.getAlignment());
Address Casted = CGF.Builder.CreateBitCast(Tmp, CGF.Int8PtrTy);
Address SrcCasted = CGF.Builder.CreateBitCast(Src, CGF.Int8PtrTy);
CGF.Builder.CreateMemCpy(Casted, SrcCasted,
llvm::ConstantInt::get(CGF.IntPtrTy, SrcSize),
false);
return CGF.Builder.CreateLoad(Tmp);
}
// Function to store a first-class aggregate into memory. We prefer to
// store the elements rather than the aggregate to be more friendly to
// fast-isel.
// FIXME: Do we need to recurse here?
static void BuildAggStore(CodeGenFunction &CGF, llvm::Value *Val,
Address Dest, bool DestIsVolatile) {
// Prefer scalar stores to first-class aggregate stores.
if (llvm::StructType *STy =
dyn_cast<llvm::StructType>(Val->getType())) {
const llvm::StructLayout *Layout =
CGF.CGM.getDataLayout().getStructLayout(STy);
for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
auto EltOffset = CharUnits::fromQuantity(Layout->getElementOffset(i));
Address EltPtr = CGF.Builder.CreateStructGEP(Dest, i, EltOffset);
llvm::Value *Elt = CGF.Builder.CreateExtractValue(Val, i);
CGF.Builder.CreateStore(Elt, EltPtr, DestIsVolatile);
}
} else {
CGF.Builder.CreateStore(Val, Dest, DestIsVolatile);
}
}
/// CreateCoercedStore - Create a store to \arg DstPtr from \arg Src,
/// where the source and destination may have different types. The
/// destination is known to be aligned to \arg DstAlign bytes.
///
/// This safely handles the case when the src type is larger than the
/// destination type; the upper bits of the src will be lost.
static void CreateCoercedStore(llvm::Value *Src,
Address Dst,
bool DstIsVolatile,
CodeGenFunction &CGF) {
llvm::Type *SrcTy = Src->getType();
llvm::Type *DstTy = Dst.getType()->getElementType();
if (SrcTy == DstTy) {
CGF.Builder.CreateStore(Src, Dst, DstIsVolatile);
return;
}
uint64_t SrcSize = CGF.CGM.getDataLayout().getTypeAllocSize(SrcTy);
if (llvm::StructType *DstSTy = dyn_cast<llvm::StructType>(DstTy)) {
Dst = EnterStructPointerForCoercedAccess(Dst, DstSTy, SrcSize, CGF);
DstTy = Dst.getType()->getElementType();
}
// If the source and destination are integer or pointer types, just do an
// extension or truncation to the desired type.
if ((isa<llvm::IntegerType>(SrcTy) || isa<llvm::PointerType>(SrcTy)) &&
(isa<llvm::IntegerType>(DstTy) || isa<llvm::PointerType>(DstTy))) {
Src = CoerceIntOrPtrToIntOrPtr(Src, DstTy, CGF);
CGF.Builder.CreateStore(Src, Dst, DstIsVolatile);
return;
}
uint64_t DstSize = CGF.CGM.getDataLayout().getTypeAllocSize(DstTy);
// If store is legal, just bitcast the src pointer.
if (SrcSize <= DstSize) {
Dst = CGF.Builder.CreateBitCast(Dst, llvm::PointerType::getUnqual(SrcTy));
BuildAggStore(CGF, Src, Dst, DstIsVolatile);
} else {
// Otherwise do coercion through memory. This is stupid, but
// simple.
// Generally SrcSize is never greater than DstSize, since this means we are
// losing bits. However, this can happen in cases where the structure has
// additional padding, for example due to a user specified alignment.
//
// FIXME: Assert that we aren't truncating non-padding bits when have access
// to that information.
Address Tmp = CreateTempAllocaForCoercion(CGF, SrcTy, Dst.getAlignment());
CGF.Builder.CreateStore(Src, Tmp);
Address Casted = CGF.Builder.CreateBitCast(Tmp, CGF.Int8PtrTy);
Address DstCasted = CGF.Builder.CreateBitCast(Dst, CGF.Int8PtrTy);
CGF.Builder.CreateMemCpy(DstCasted, Casted,
llvm::ConstantInt::get(CGF.IntPtrTy, DstSize),
false);
}
}
static Address emitAddressAtOffset(CodeGenFunction &CGF, Address addr,
const ABIArgInfo &info) {
if (unsigned offset = info.getDirectOffset()) {
addr = CGF.Builder.CreateElementBitCast(addr, CGF.Int8Ty);
addr = CGF.Builder.CreateConstInBoundsByteGEP(addr,
CharUnits::fromQuantity(offset));
addr = CGF.Builder.CreateElementBitCast(addr, info.getCoerceToType());
}
return addr;
}
namespace {
/// Encapsulates information about the way function arguments from
/// CGFunctionInfo should be passed to actual LLVM IR function.
class ClangToLLVMArgMapping {
static const unsigned InvalidIndex = ~0U;
unsigned InallocaArgNo;
unsigned SRetArgNo;
unsigned TotalIRArgs;
/// Arguments of LLVM IR function corresponding to single Clang argument.
struct IRArgs {
unsigned PaddingArgIndex;
// Argument is expanded to IR arguments at positions
// [FirstArgIndex, FirstArgIndex + NumberOfArgs).
unsigned FirstArgIndex;
unsigned NumberOfArgs;
IRArgs()
: PaddingArgIndex(InvalidIndex), FirstArgIndex(InvalidIndex),
NumberOfArgs(0) {}
};
SmallVector<IRArgs, 8> ArgInfo;
public:
ClangToLLVMArgMapping(const ASTContext &Context, const CGFunctionInfo &FI,
bool OnlyRequiredArgs = false)
: InallocaArgNo(InvalidIndex), SRetArgNo(InvalidIndex), TotalIRArgs(0),
ArgInfo(OnlyRequiredArgs ? FI.getNumRequiredArgs() : FI.arg_size()) {
construct(Context, FI, OnlyRequiredArgs);
}
bool hasInallocaArg() const { return InallocaArgNo != InvalidIndex; }
unsigned getInallocaArgNo() const {
assert(hasInallocaArg());
return InallocaArgNo;
}
bool hasSRetArg() const { return SRetArgNo != InvalidIndex; }
unsigned getSRetArgNo() const {
assert(hasSRetArg());
return SRetArgNo;
}
unsigned totalIRArgs() const { return TotalIRArgs; }
bool hasPaddingArg(unsigned ArgNo) const {
assert(ArgNo < ArgInfo.size());
return ArgInfo[ArgNo].PaddingArgIndex != InvalidIndex;
}
unsigned getPaddingArgNo(unsigned ArgNo) const {
assert(hasPaddingArg(ArgNo));
return ArgInfo[ArgNo].PaddingArgIndex;
}
/// Returns index of first IR argument corresponding to ArgNo, and their
/// quantity.
std::pair<unsigned, unsigned> getIRArgs(unsigned ArgNo) const {
assert(ArgNo < ArgInfo.size());
return std::make_pair(ArgInfo[ArgNo].FirstArgIndex,
ArgInfo[ArgNo].NumberOfArgs);
}
private:
void construct(const ASTContext &Context, const CGFunctionInfo &FI,
bool OnlyRequiredArgs);
};
void ClangToLLVMArgMapping::construct(const ASTContext &Context,
const CGFunctionInfo &FI,
bool OnlyRequiredArgs) {
unsigned IRArgNo = 0;
bool SwapThisWithSRet = false;
const ABIArgInfo &RetAI = FI.getReturnInfo();
if (RetAI.getKind() == ABIArgInfo::Indirect) {
SwapThisWithSRet = RetAI.isSRetAfterThis();
SRetArgNo = SwapThisWithSRet ? 1 : IRArgNo++;
}
unsigned ArgNo = 0;
unsigned NumArgs = OnlyRequiredArgs ? FI.getNumRequiredArgs() : FI.arg_size();
for (CGFunctionInfo::const_arg_iterator I = FI.arg_begin(); ArgNo < NumArgs;
++I, ++ArgNo) {
assert(I != FI.arg_end());
QualType ArgType = I->type;
const ABIArgInfo &AI = I->info;
// Collect data about IR arguments corresponding to Clang argument ArgNo.
auto &IRArgs = ArgInfo[ArgNo];
if (AI.getPaddingType())
IRArgs.PaddingArgIndex = IRArgNo++;
switch (AI.getKind()) {
case ABIArgInfo::Extend:
case ABIArgInfo::Direct: {
// FIXME: handle sseregparm someday...
llvm::StructType *STy = dyn_cast<llvm::StructType>(AI.getCoerceToType());
if (AI.isDirect() && AI.getCanBeFlattened() && STy) {
IRArgs.NumberOfArgs = STy->getNumElements();
} else {
IRArgs.NumberOfArgs = 1;
}
break;
}
case ABIArgInfo::Indirect:
IRArgs.NumberOfArgs = 1;
break;
case ABIArgInfo::Ignore:
case ABIArgInfo::InAlloca:
// ignore and inalloca doesn't have matching LLVM parameters.
IRArgs.NumberOfArgs = 0;
break;
case ABIArgInfo::Expand: {
IRArgs.NumberOfArgs = getExpansionSize(ArgType, Context);
break;
}
}
if (IRArgs.NumberOfArgs > 0) {
IRArgs.FirstArgIndex = IRArgNo;
IRArgNo += IRArgs.NumberOfArgs;
}
// Skip over the sret parameter when it comes second. We already handled it
// above.
if (IRArgNo == 1 && SwapThisWithSRet)
IRArgNo++;
}
assert(ArgNo == ArgInfo.size());
if (FI.usesInAlloca())
InallocaArgNo = IRArgNo++;
TotalIRArgs = IRArgNo;
}
} // namespace
/***/
bool CodeGenModule::ReturnTypeUsesSRet(const CGFunctionInfo &FI) {
return FI.getReturnInfo().isIndirect();
}
bool CodeGenModule::ReturnSlotInterferesWithArgs(const CGFunctionInfo &FI) {
return ReturnTypeUsesSRet(FI) &&
getTargetCodeGenInfo().doesReturnSlotInterfereWithArgs();
}
bool CodeGenModule::ReturnTypeUsesFPRet(QualType ResultType) {
if (const BuiltinType *BT = ResultType->getAs<BuiltinType>()) {
switch (BT->getKind()) {
default:
return false;
case BuiltinType::Float:
return getTarget().useObjCFPRetForRealType(TargetInfo::Float);
case BuiltinType::Double:
return getTarget().useObjCFPRetForRealType(TargetInfo::Double);
case BuiltinType::LongDouble:
return getTarget().useObjCFPRetForRealType(TargetInfo::LongDouble);
}
}
return false;
}
bool CodeGenModule::ReturnTypeUsesFP2Ret(QualType ResultType) {
if (const ComplexType *CT = ResultType->getAs<ComplexType>()) {
if (const BuiltinType *BT = CT->getElementType()->getAs<BuiltinType>()) {
if (BT->getKind() == BuiltinType::LongDouble)
return getTarget().useObjCFP2RetForComplexLongDouble();
}
}
return false;
}
llvm::FunctionType *CodeGenTypes::GetFunctionType(GlobalDecl GD) {
const CGFunctionInfo &FI = arrangeGlobalDeclaration(GD);
return GetFunctionType(FI);
}
llvm::FunctionType *
CodeGenTypes::GetFunctionType(const CGFunctionInfo &FI) {
bool Inserted = FunctionsBeingProcessed.insert(&FI).second;
(void)Inserted;
assert(Inserted && "Recursively being processed?");
llvm::Type *resultType = nullptr;
const ABIArgInfo &retAI = FI.getReturnInfo();
switch (retAI.getKind()) {
case ABIArgInfo::Expand:
llvm_unreachable("Invalid ABI kind for return argument");
case ABIArgInfo::Extend:
case ABIArgInfo::Direct:
resultType = retAI.getCoerceToType();
break;
case ABIArgInfo::InAlloca:
if (retAI.getInAllocaSRet()) {
// sret things on win32 aren't void, they return the sret pointer.
QualType ret = FI.getReturnType();
llvm::Type *ty = ConvertType(ret);
unsigned addressSpace = Context.getTargetAddressSpace(ret);
resultType = llvm::PointerType::get(ty, addressSpace);
} else {
resultType = llvm::Type::getVoidTy(getLLVMContext());
}
break;
case ABIArgInfo::Indirect:
case ABIArgInfo::Ignore:
resultType = llvm::Type::getVoidTy(getLLVMContext());
break;
}
ClangToLLVMArgMapping IRFunctionArgs(getContext(), FI, true);
SmallVector<llvm::Type*, 8> ArgTypes(IRFunctionArgs.totalIRArgs());
// Add type for sret argument.
if (IRFunctionArgs.hasSRetArg()) {
QualType Ret = FI.getReturnType();
llvm::Type *Ty = ConvertType(Ret);
unsigned AddressSpace = Context.getTargetAddressSpace(Ret);
ArgTypes[IRFunctionArgs.getSRetArgNo()] =
llvm::PointerType::get(Ty, AddressSpace);
}
// Add type for inalloca argument.
if (IRFunctionArgs.hasInallocaArg()) {
auto ArgStruct = FI.getArgStruct();
assert(ArgStruct);
ArgTypes[IRFunctionArgs.getInallocaArgNo()] = ArgStruct->getPointerTo();
}
// Add in all of the required arguments.
unsigned ArgNo = 0;
CGFunctionInfo::const_arg_iterator it = FI.arg_begin(),
ie = it + FI.getNumRequiredArgs();
for (; it != ie; ++it, ++ArgNo) {
const ABIArgInfo &ArgInfo = it->info;
// Insert a padding type to ensure proper alignment.
if (IRFunctionArgs.hasPaddingArg(ArgNo))
ArgTypes[IRFunctionArgs.getPaddingArgNo(ArgNo)] =
ArgInfo.getPaddingType();
unsigned FirstIRArg, NumIRArgs;
std::tie(FirstIRArg, NumIRArgs) = IRFunctionArgs.getIRArgs(ArgNo);
switch (ArgInfo.getKind()) {
case ABIArgInfo::Ignore:
case ABIArgInfo::InAlloca:
assert(NumIRArgs == 0);
break;
case ABIArgInfo::Indirect: {
assert(NumIRArgs == 1);
// indirect arguments are always on the stack, which is addr space #0.
llvm::Type *LTy = ConvertTypeForMem(it->type);
ArgTypes[FirstIRArg] = LTy->getPointerTo();
break;
}
case ABIArgInfo::Extend:
case ABIArgInfo::Direct: {
// Fast-isel and the optimizer generally like scalar values better than
// FCAs, so we flatten them if this is safe to do for this argument.
llvm::Type *argType = ArgInfo.getCoerceToType();
llvm::StructType *st = dyn_cast<llvm::StructType>(argType);
if (st && ArgInfo.isDirect() && ArgInfo.getCanBeFlattened()) {
assert(NumIRArgs == st->getNumElements());
for (unsigned i = 0, e = st->getNumElements(); i != e; ++i)
ArgTypes[FirstIRArg + i] = st->getElementType(i);
} else {
assert(NumIRArgs == 1);
ArgTypes[FirstIRArg] = argType;
}
break;
}
case ABIArgInfo::Expand:
auto ArgTypesIter = ArgTypes.begin() + FirstIRArg;
getExpandedTypes(it->type, ArgTypesIter);
assert(ArgTypesIter == ArgTypes.begin() + FirstIRArg + NumIRArgs);
break;
}
}
bool Erased = FunctionsBeingProcessed.erase(&FI); (void)Erased;
assert(Erased && "Not in set?");
return llvm::FunctionType::get(resultType, ArgTypes, FI.isVariadic());
}
llvm::Type *CodeGenTypes::GetFunctionTypeForVTable(GlobalDecl GD) {
const CXXMethodDecl *MD = cast<CXXMethodDecl>(GD.getDecl());
const FunctionProtoType *FPT = MD->getType()->getAs<FunctionProtoType>();
if (!isFuncTypeConvertible(FPT))
return llvm::StructType::get(getLLVMContext());
const CGFunctionInfo *Info;
if (isa<CXXDestructorDecl>(MD))
Info =
&arrangeCXXStructorDeclaration(MD, getFromDtorType(GD.getDtorType()));
else
Info = &arrangeCXXMethodDeclaration(MD);
return GetFunctionType(*Info);
}
static void AddAttributesFromFunctionProtoType(ASTContext &Ctx,
llvm::AttrBuilder &FuncAttrs,
const FunctionProtoType *FPT) {
if (!FPT)
return;
if (!isUnresolvedExceptionSpec(FPT->getExceptionSpecType()) &&
FPT->isNothrow(Ctx))
FuncAttrs.addAttribute(llvm::Attribute::NoUnwind);
}
void CodeGenModule::ConstructAttributeList(
StringRef Name, const CGFunctionInfo &FI, CGCalleeInfo CalleeInfo,
AttributeListType &PAL, unsigned &CallingConv, bool AttrOnCallSite) {
llvm::AttrBuilder FuncAttrs;
llvm::AttrBuilder RetAttrs;
bool HasOptnone = false;
CallingConv = FI.getEffectiveCallingConvention();
if (FI.isNoReturn())
FuncAttrs.addAttribute(llvm::Attribute::NoReturn);
// If we have information about the function prototype, we can learn
// attributes form there.
AddAttributesFromFunctionProtoType(getContext(), FuncAttrs,
CalleeInfo.getCalleeFunctionProtoType());
const Decl *TargetDecl = CalleeInfo.getCalleeDecl();
// FIXME: handle sseregparm someday...
if (TargetDecl) {
if (TargetDecl->hasAttr<ReturnsTwiceAttr>())
FuncAttrs.addAttribute(llvm::Attribute::ReturnsTwice);
if (TargetDecl->hasAttr<NoThrowAttr>())
FuncAttrs.addAttribute(llvm::Attribute::NoUnwind);
if (TargetDecl->hasAttr<NoReturnAttr>())
FuncAttrs.addAttribute(llvm::Attribute::NoReturn);
if (TargetDecl->hasAttr<NoDuplicateAttr>())
FuncAttrs.addAttribute(llvm::Attribute::NoDuplicate);
if (const FunctionDecl *Fn = dyn_cast<FunctionDecl>(TargetDecl)) {
AddAttributesFromFunctionProtoType(
getContext(), FuncAttrs, Fn->getType()->getAs<FunctionProtoType>());
// Don't use [[noreturn]] or _Noreturn for a call to a virtual function.
// These attributes are not inherited by overloads.
const CXXMethodDecl *MD = dyn_cast<CXXMethodDecl>(Fn);
if (Fn->isNoReturn() && !(AttrOnCallSite && MD && MD->isVirtual()))
FuncAttrs.addAttribute(llvm::Attribute::NoReturn);
}
// 'const', 'pure' and 'noalias' attributed functions are also nounwind.
if (TargetDecl->hasAttr<ConstAttr>()) {
FuncAttrs.addAttribute(llvm::Attribute::ReadNone);
FuncAttrs.addAttribute(llvm::Attribute::NoUnwind);
} else if (TargetDecl->hasAttr<PureAttr>()) {
FuncAttrs.addAttribute(llvm::Attribute::ReadOnly);
FuncAttrs.addAttribute(llvm::Attribute::NoUnwind);
} else if (TargetDecl->hasAttr<NoAliasAttr>()) {
FuncAttrs.addAttribute(llvm::Attribute::ArgMemOnly);
FuncAttrs.addAttribute(llvm::Attribute::NoUnwind);
}
if (TargetDecl->hasAttr<RestrictAttr>())
RetAttrs.addAttribute(llvm::Attribute::NoAlias);
if (TargetDecl->hasAttr<ReturnsNonNullAttr>())
RetAttrs.addAttribute(llvm::Attribute::NonNull);
HasOptnone = TargetDecl->hasAttr<OptimizeNoneAttr>();
}
// OptimizeNoneAttr takes precedence over -Os or -Oz. No warning needed.
if (!HasOptnone) {
if (CodeGenOpts.OptimizeSize)
FuncAttrs.addAttribute(llvm::Attribute::OptimizeForSize);
if (CodeGenOpts.OptimizeSize == 2)
FuncAttrs.addAttribute(llvm::Attribute::MinSize);
}
if (CodeGenOpts.DisableRedZone)
FuncAttrs.addAttribute(llvm::Attribute::NoRedZone);
if (CodeGenOpts.NoImplicitFloat)
FuncAttrs.addAttribute(llvm::Attribute::NoImplicitFloat);
if (CodeGenOpts.EnableSegmentedStacks &&
!(TargetDecl && TargetDecl->hasAttr<NoSplitStackAttr>()))
FuncAttrs.addAttribute("split-stack");
if (AttrOnCallSite) {
// Attributes that should go on the call site only.
if (!CodeGenOpts.SimplifyLibCalls ||
CodeGenOpts.isNoBuiltinFunc(Name.data()))
FuncAttrs.addAttribute(llvm::Attribute::NoBuiltin);
if (!CodeGenOpts.TrapFuncName.empty())
FuncAttrs.addAttribute("trap-func-name", CodeGenOpts.TrapFuncName);
} else {
// Attributes that should go on the function, but not the call site.
if (!CodeGenOpts.DisableFPElim) {
FuncAttrs.addAttribute("no-frame-pointer-elim", "false");
} else if (CodeGenOpts.OmitLeafFramePointer) {
FuncAttrs.addAttribute("no-frame-pointer-elim", "false");
FuncAttrs.addAttribute("no-frame-pointer-elim-non-leaf");
} else {
FuncAttrs.addAttribute("no-frame-pointer-elim", "true");
FuncAttrs.addAttribute("no-frame-pointer-elim-non-leaf");
}
bool DisableTailCalls =
CodeGenOpts.DisableTailCalls ||
(TargetDecl && TargetDecl->hasAttr<DisableTailCallsAttr>());
FuncAttrs.addAttribute("disable-tail-calls",
llvm::toStringRef(DisableTailCalls));
FuncAttrs.addAttribute("less-precise-fpmad",
llvm::toStringRef(CodeGenOpts.LessPreciseFPMAD));
FuncAttrs.addAttribute("no-infs-fp-math",
llvm::toStringRef(CodeGenOpts.NoInfsFPMath));
FuncAttrs.addAttribute("no-nans-fp-math",
llvm::toStringRef(CodeGenOpts.NoNaNsFPMath));
FuncAttrs.addAttribute("unsafe-fp-math",
llvm::toStringRef(CodeGenOpts.UnsafeFPMath));
FuncAttrs.addAttribute("use-soft-float",
llvm::toStringRef(CodeGenOpts.SoftFloat));
FuncAttrs.addAttribute("stack-protector-buffer-size",
llvm::utostr(CodeGenOpts.SSPBufferSize));
if (CodeGenOpts.StackRealignment)
FuncAttrs.addAttribute("stackrealign");
// Add target-cpu and target-features attributes to functions. If
// we have a decl for the function and it has a target attribute then
// parse that and add it to the feature set.
StringRef TargetCPU = getTarget().getTargetOpts().CPU;
const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(TargetDecl);
if (FD && FD->hasAttr<TargetAttr>()) {
llvm::StringMap<bool> FeatureMap;
getFunctionFeatureMap(FeatureMap, FD);
// Produce the canonical string for this set of features.
std::vector<std::string> Features;
for (llvm::StringMap<bool>::const_iterator it = FeatureMap.begin(),
ie = FeatureMap.end();
it != ie; ++it)
Features.push_back((it->second ? "+" : "-") + it->first().str());
// Now add the target-cpu and target-features to the function.
// While we populated the feature map above, we still need to
// get and parse the target attribute so we can get the cpu for
// the function.
const auto *TD = FD->getAttr<TargetAttr>();
TargetAttr::ParsedTargetAttr ParsedAttr = TD->parse();
if (ParsedAttr.second != "")
TargetCPU = ParsedAttr.second;
if (TargetCPU != "")
FuncAttrs.addAttribute("target-cpu", TargetCPU);
if (!Features.empty()) {
std::sort(Features.begin(), Features.end());
FuncAttrs.addAttribute(
"target-features",
llvm::join(Features.begin(), Features.end(), ","));
}
} else {
// Otherwise just add the existing target cpu and target features to the
// function.
std::vector<std::string> &Features = getTarget().getTargetOpts().Features;
if (TargetCPU != "")
FuncAttrs.addAttribute("target-cpu", TargetCPU);
if (!Features.empty()) {
std::sort(Features.begin(), Features.end());
FuncAttrs.addAttribute(
"target-features",
llvm::join(Features.begin(), Features.end(), ","));
}
}
}
ClangToLLVMArgMapping IRFunctionArgs(getContext(), FI);
QualType RetTy = FI.getReturnType();
const ABIArgInfo &RetAI = FI.getReturnInfo();
switch (RetAI.getKind()) {
case ABIArgInfo::Extend:
if (RetTy->hasSignedIntegerRepresentation())
RetAttrs.addAttribute(llvm::Attribute::SExt);
else if (RetTy->hasUnsignedIntegerRepresentation())
RetAttrs.addAttribute(llvm::Attribute::ZExt);
// FALL THROUGH
case ABIArgInfo::Direct:
if (RetAI.getInReg())
RetAttrs.addAttribute(llvm::Attribute::InReg);
break;
case ABIArgInfo::Ignore:
break;
case ABIArgInfo::InAlloca:
case ABIArgInfo::Indirect: {
// inalloca and sret disable readnone and readonly
FuncAttrs.removeAttribute(llvm::Attribute::ReadOnly)
.removeAttribute(llvm::Attribute::ReadNone);
break;
}
case ABIArgInfo::Expand:
llvm_unreachable("Invalid ABI kind for return argument");
}
if (const auto *RefTy = RetTy->getAs<ReferenceType>()) {
QualType PTy = RefTy->getPointeeType();
if (!PTy->isIncompleteType() && PTy->isConstantSizeType())
RetAttrs.addDereferenceableAttr(getContext().getTypeSizeInChars(PTy)
.getQuantity());
else if (getContext().getTargetAddressSpace(PTy) == 0)
RetAttrs.addAttribute(llvm::Attribute::NonNull);
}
// Attach return attributes.
if (RetAttrs.hasAttributes()) {
PAL.push_back(llvm::AttributeSet::get(
getLLVMContext(), llvm::AttributeSet::ReturnIndex, RetAttrs));
}
// Attach attributes to sret.
if (IRFunctionArgs.hasSRetArg()) {
llvm::AttrBuilder SRETAttrs;
SRETAttrs.addAttribute(llvm::Attribute::StructRet);
if (RetAI.getInReg())
SRETAttrs.addAttribute(llvm::Attribute::InReg);
PAL.push_back(llvm::AttributeSet::get(
getLLVMContext(), IRFunctionArgs.getSRetArgNo() + 1, SRETAttrs));
}
// Attach attributes to inalloca argument.
if (IRFunctionArgs.hasInallocaArg()) {
llvm::AttrBuilder Attrs;
Attrs.addAttribute(llvm::Attribute::InAlloca);
PAL.push_back(llvm::AttributeSet::get(
getLLVMContext(), IRFunctionArgs.getInallocaArgNo() + 1, Attrs));
}
unsigned ArgNo = 0;
for (CGFunctionInfo::const_arg_iterator I = FI.arg_begin(),
E = FI.arg_end();
I != E; ++I, ++ArgNo) {
QualType ParamType = I->type;
const ABIArgInfo &AI = I->info;
llvm::AttrBuilder Attrs;
// Add attribute for padding argument, if necessary.
if (IRFunctionArgs.hasPaddingArg(ArgNo)) {
if (AI.getPaddingInReg())
PAL.push_back(llvm::AttributeSet::get(
getLLVMContext(), IRFunctionArgs.getPaddingArgNo(ArgNo) + 1,
llvm::Attribute::InReg));
}
// 'restrict' -> 'noalias' is done in EmitFunctionProlog when we
// have the corresponding parameter variable. It doesn't make
// sense to do it here because parameters are so messed up.
switch (AI.getKind()) {
case ABIArgInfo::Extend:
if (ParamType->isSignedIntegerOrEnumerationType())
Attrs.addAttribute(llvm::Attribute::SExt);
else if (ParamType->isUnsignedIntegerOrEnumerationType()) {
if (getTypes().getABIInfo().shouldSignExtUnsignedType(ParamType))
Attrs.addAttribute(llvm::Attribute::SExt);
else
Attrs.addAttribute(llvm::Attribute::ZExt);
}
// FALL THROUGH
case ABIArgInfo::Direct:
if (ArgNo == 0 && FI.isChainCall())
Attrs.addAttribute(llvm::Attribute::Nest);
else if (AI.getInReg())
Attrs.addAttribute(llvm::Attribute::InReg);
break;
case ABIArgInfo::Indirect: {
if (AI.getInReg())
Attrs.addAttribute(llvm::Attribute::InReg);
if (AI.getIndirectByVal())
Attrs.addAttribute(llvm::Attribute::ByVal);
CharUnits Align = AI.getIndirectAlign();
// In a byval argument, it is important that the required
// alignment of the type is honored, as LLVM might be creating a
// *new* stack object, and needs to know what alignment to give
// it. (Sometimes it can deduce a sensible alignment on its own,
// but not if clang decides it must emit a packed struct, or the
// user specifies increased alignment requirements.)
//
// This is different from indirect *not* byval, where the object
// exists already, and the align attribute is purely
// informative.
assert(!Align.isZero());
// For now, only add this when we have a byval argument.
// TODO: be less lazy about updating test cases.
if (AI.getIndirectByVal())
Attrs.addAlignmentAttr(Align.getQuantity());
// byval disables readnone and readonly.
FuncAttrs.removeAttribute(llvm::Attribute::ReadOnly)
.removeAttribute(llvm::Attribute::ReadNone);
break;
}
case ABIArgInfo::Ignore:
case ABIArgInfo::Expand:
continue;
case ABIArgInfo::InAlloca:
// inalloca disables readnone and readonly.
FuncAttrs.removeAttribute(llvm::Attribute::ReadOnly)
.removeAttribute(llvm::Attribute::ReadNone);
continue;
}
if (const auto *RefTy = ParamType->getAs<ReferenceType>()) {
QualType PTy = RefTy->getPointeeType();
if (!PTy->isIncompleteType() && PTy->isConstantSizeType())
Attrs.addDereferenceableAttr(getContext().getTypeSizeInChars(PTy)
.getQuantity());
else if (getContext().getTargetAddressSpace(PTy) == 0)
Attrs.addAttribute(llvm::Attribute::NonNull);
}
if (Attrs.hasAttributes()) {
unsigned FirstIRArg, NumIRArgs;
std::tie(FirstIRArg, NumIRArgs) = IRFunctionArgs.getIRArgs(ArgNo);
for (unsigned i = 0; i < NumIRArgs; i++)
PAL.push_back(llvm::AttributeSet::get(getLLVMContext(),
FirstIRArg + i + 1, Attrs));
}
}
assert(ArgNo == FI.arg_size());
if (FuncAttrs.hasAttributes())
PAL.push_back(llvm::
AttributeSet::get(getLLVMContext(),
llvm::AttributeSet::FunctionIndex,
FuncAttrs));
}
/// An argument came in as a promoted argument; demote it back to its
/// declared type.
static llvm::Value *emitArgumentDemotion(CodeGenFunction &CGF,
const VarDecl *var,
llvm::Value *value) {
llvm::Type *varType = CGF.ConvertType(var->getType());
// This can happen with promotions that actually don't change the
// underlying type, like the enum promotions.
if (value->getType() == varType) return value;
assert((varType->isIntegerTy() || varType->isFloatingPointTy())
&& "unexpected promotion type");
if (isa<llvm::IntegerType>(varType))
return CGF.Builder.CreateTrunc(value, varType, "arg.unpromote");
return CGF.Builder.CreateFPCast(value, varType, "arg.unpromote");
}
/// Returns the attribute (either parameter attribute, or function
/// attribute), which declares argument ArgNo to be non-null.
static const NonNullAttr *getNonNullAttr(const Decl *FD, const ParmVarDecl *PVD,
QualType ArgType, unsigned ArgNo) {
// FIXME: __attribute__((nonnull)) can also be applied to:
// - references to pointers, where the pointee is known to be
// nonnull (apparently a Clang extension)
// - transparent unions containing pointers
// In the former case, LLVM IR cannot represent the constraint. In
// the latter case, we have no guarantee that the transparent union
// is in fact passed as a pointer.
if (!ArgType->isAnyPointerType() && !ArgType->isBlockPointerType())
return nullptr;
// First, check attribute on parameter itself.
if (PVD) {
if (auto ParmNNAttr = PVD->getAttr<NonNullAttr>())
return ParmNNAttr;
}
// Check function attributes.
if (!FD)
return nullptr;
for (const auto *NNAttr : FD->specific_attrs<NonNullAttr>()) {
if (NNAttr->isNonNull(ArgNo))
return NNAttr;
}
return nullptr;
}
void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
llvm::Function *Fn,
const FunctionArgList &Args) {
if (CurCodeDecl && CurCodeDecl->hasAttr<NakedAttr>())
// Naked functions don't have prologues.
return;
// If this is an implicit-return-zero function, go ahead and
// initialize the return value. TODO: it might be nice to have
// a more general mechanism for this that didn't require synthesized
// return statements.
if (const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(CurCodeDecl)) {
if (FD->hasImplicitReturnZero()) {
QualType RetTy = FD->getReturnType().getUnqualifiedType();
llvm::Type* LLVMTy = CGM.getTypes().ConvertType(RetTy);
llvm::Constant* Zero = llvm::Constant::getNullValue(LLVMTy);
Builder.CreateStore(Zero, ReturnValue);
}
}
// FIXME: We no longer need the types from FunctionArgList; lift up and
// simplify.
ClangToLLVMArgMapping IRFunctionArgs(CGM.getContext(), FI);
// Flattened function arguments.
SmallVector<llvm::Argument *, 16> FnArgs;
FnArgs.reserve(IRFunctionArgs.totalIRArgs());
for (auto &Arg : Fn->args()) {
FnArgs.push_back(&Arg);
}
assert(FnArgs.size() == IRFunctionArgs.totalIRArgs());
// If we're using inalloca, all the memory arguments are GEPs off of the last
// parameter, which is a pointer to the complete memory area.
Address ArgStruct = Address::invalid();
const llvm::StructLayout *ArgStructLayout = nullptr;
if (IRFunctionArgs.hasInallocaArg()) {
ArgStructLayout = CGM.getDataLayout().getStructLayout(FI.getArgStruct());
ArgStruct = Address(FnArgs[IRFunctionArgs.getInallocaArgNo()],
FI.getArgStructAlignment());
assert(ArgStruct.getType() == FI.getArgStruct()->getPointerTo());
}
// Name the struct return parameter.
if (IRFunctionArgs.hasSRetArg()) {
auto AI = FnArgs[IRFunctionArgs.getSRetArgNo()];
AI->setName("agg.result");
AI->addAttr(llvm::AttributeSet::get(getLLVMContext(), AI->getArgNo() + 1,
llvm::Attribute::NoAlias));
}
// Track if we received the parameter as a pointer (indirect, byval, or
// inalloca). If already have a pointer, EmitParmDecl doesn't need to copy it
// into a local alloca for us.
SmallVector<ParamValue, 16> ArgVals;
ArgVals.reserve(Args.size());
// Create a pointer value for every parameter declaration. This usually
// entails copying one or more LLVM IR arguments into an alloca. Don't push
// any cleanups or do anything that might unwind. We do that separately, so
// we can push the cleanups in the correct order for the ABI.
assert(FI.arg_size() == Args.size() &&
"Mismatch between function signature & arguments.");
unsigned ArgNo = 0;
CGFunctionInfo::const_arg_iterator info_it = FI.arg_begin();
for (FunctionArgList::const_iterator i = Args.begin(), e = Args.end();
i != e; ++i, ++info_it, ++ArgNo) {
const VarDecl *Arg = *i;
QualType Ty = info_it->type;
const ABIArgInfo &ArgI = info_it->info;
bool isPromoted =
isa<ParmVarDecl>(Arg) && cast<ParmVarDecl>(Arg)->isKNRPromoted();
unsigned FirstIRArg, NumIRArgs;
std::tie(FirstIRArg, NumIRArgs) = IRFunctionArgs.getIRArgs(ArgNo);
switch (ArgI.getKind()) {
case ABIArgInfo::InAlloca: {
assert(NumIRArgs == 0);
auto FieldIndex = ArgI.getInAllocaFieldIndex();
CharUnits FieldOffset =
CharUnits::fromQuantity(ArgStructLayout->getElementOffset(FieldIndex));
Address V = Builder.CreateStructGEP(ArgStruct, FieldIndex, FieldOffset,
Arg->getName());
ArgVals.push_back(ParamValue::forIndirect(V));
break;
}
case ABIArgInfo::Indirect: {
assert(NumIRArgs == 1);
Address ParamAddr = Address(FnArgs[FirstIRArg], ArgI.getIndirectAlign());
if (!hasScalarEvaluationKind(Ty)) {
// Aggregates and complex variables are accessed by reference. All we
// need to do is realign the value, if requested.
Address V = ParamAddr;
if (ArgI.getIndirectRealign()) {
Address AlignedTemp = CreateMemTemp(Ty, "coerce");
// Copy from the incoming argument pointer to the temporary with the
// appropriate alignment.
//
// FIXME: We should have a common utility for generating an aggregate
// copy.
CharUnits Size = getContext().getTypeSizeInChars(Ty);
auto SizeVal = llvm::ConstantInt::get(IntPtrTy, Size.getQuantity());
Address Dst = Builder.CreateBitCast(AlignedTemp, Int8PtrTy);
Address Src = Builder.CreateBitCast(ParamAddr, Int8PtrTy);
Builder.CreateMemCpy(Dst, Src, SizeVal, false);
V = AlignedTemp;
}
ArgVals.push_back(ParamValue::forIndirect(V));
} else {
// Load scalar value from indirect argument.
llvm::Value *V =
EmitLoadOfScalar(ParamAddr, false, Ty, Arg->getLocStart());
if (isPromoted)
V = emitArgumentDemotion(*this, Arg, V);
ArgVals.push_back(ParamValue::forDirect(V));
}
break;
}
case ABIArgInfo::Extend:
case ABIArgInfo::Direct: {
// If we have the trivial case, handle it with no muss and fuss.
if (!isa<llvm::StructType>(ArgI.getCoerceToType()) &&
ArgI.getCoerceToType() == ConvertType(Ty) &&
ArgI.getDirectOffset() == 0) {
assert(NumIRArgs == 1);
auto AI = FnArgs[FirstIRArg];
llvm::Value *V = AI;
if (const ParmVarDecl *PVD = dyn_cast<ParmVarDecl>(Arg)) {
if (getNonNullAttr(CurCodeDecl, PVD, PVD->getType(),
PVD->getFunctionScopeIndex()))
AI->addAttr(llvm::AttributeSet::get(getLLVMContext(),
AI->getArgNo() + 1,
llvm::Attribute::NonNull));
QualType OTy = PVD->getOriginalType();
if (const auto *ArrTy =
getContext().getAsConstantArrayType(OTy)) {
// A C99 array parameter declaration with the static keyword also
// indicates dereferenceability, and if the size is constant we can
// use the dereferenceable attribute (which requires the size in
// bytes).
if (ArrTy->getSizeModifier() == ArrayType::Static) {
QualType ETy = ArrTy->getElementType();
uint64_t ArrSize = ArrTy->getSize().getZExtValue();
if (!ETy->isIncompleteType() && ETy->isConstantSizeType() &&
ArrSize) {
llvm::AttrBuilder Attrs;
Attrs.addDereferenceableAttr(
getContext().getTypeSizeInChars(ETy).getQuantity()*ArrSize);
AI->addAttr(llvm::AttributeSet::get(getLLVMContext(),
AI->getArgNo() + 1, Attrs));
} else if (getContext().getTargetAddressSpace(ETy) == 0) {
AI->addAttr(llvm::AttributeSet::get(getLLVMContext(),
AI->getArgNo() + 1,
llvm::Attribute::NonNull));
}
}
} else if (const auto *ArrTy =
getContext().getAsVariableArrayType(OTy)) {
// For C99 VLAs with the static keyword, we don't know the size so
// we can't use the dereferenceable attribute, but in addrspace(0)
// we know that it must be nonnull.
if (ArrTy->getSizeModifier() == VariableArrayType::Static &&
!getContext().getTargetAddressSpace(ArrTy->getElementType()))
AI->addAttr(llvm::AttributeSet::get(getLLVMContext(),
AI->getArgNo() + 1,
llvm::Attribute::NonNull));
}
const auto *AVAttr = PVD->getAttr<AlignValueAttr>();
if (!AVAttr)
if (const auto *TOTy = dyn_cast<TypedefType>(OTy))
AVAttr = TOTy->getDecl()->getAttr<AlignValueAttr>();
if (AVAttr) {
llvm::Value *AlignmentValue =
EmitScalarExpr(AVAttr->getAlignment());
llvm::ConstantInt *AlignmentCI =
cast<llvm::ConstantInt>(AlignmentValue);
unsigned Alignment =
std::min((unsigned) AlignmentCI->getZExtValue(),
+llvm::Value::MaximumAlignment);
llvm::AttrBuilder Attrs;
Attrs.addAlignmentAttr(Alignment);
AI->addAttr(llvm::AttributeSet::get(getLLVMContext(),
AI->getArgNo() + 1, Attrs));
}
}
if (Arg->getType().isRestrictQualified())
AI->addAttr(llvm::AttributeSet::get(getLLVMContext(),
AI->getArgNo() + 1,
llvm::Attribute::NoAlias));
// Ensure the argument is the correct type.
if (V->getType() != ArgI.getCoerceToType())
V = Builder.CreateBitCast(V, ArgI.getCoerceToType());
if (isPromoted)
V = emitArgumentDemotion(*this, Arg, V);
if (const CXXMethodDecl *MD =
dyn_cast_or_null<CXXMethodDecl>(CurCodeDecl)) {
if (MD->isVirtual() && Arg == CXXABIThisDecl)
V = CGM.getCXXABI().
adjustThisParameterInVirtualFunctionPrologue(*this, CurGD, V);
}
// Because of merging of function types from multiple decls it is
// possible for the type of an argument to not match the corresponding
// type in the function type. Since we are codegening the callee
// in here, add a cast to the argument type.
llvm::Type *LTy = ConvertType(Arg->getType());
if (V->getType() != LTy)
V = Builder.CreateBitCast(V, LTy);
ArgVals.push_back(ParamValue::forDirect(V));
break;
}
Address Alloca = CreateMemTemp(Ty, getContext().getDeclAlign(Arg),
Arg->getName());
// Pointer to store into.
Address Ptr = emitAddressAtOffset(*this, Alloca, ArgI);
// Fast-isel and the optimizer generally like scalar values better than
// FCAs, so we flatten them if this is safe to do for this argument.
llvm::StructType *STy = dyn_cast<llvm::StructType>(ArgI.getCoerceToType());
if (ArgI.isDirect() && ArgI.getCanBeFlattened() && STy &&
STy->getNumElements() > 1) {
auto SrcLayout = CGM.getDataLayout().getStructLayout(STy);
uint64_t SrcSize = CGM.getDataLayout().getTypeAllocSize(STy);
llvm::Type *DstTy = Ptr.getElementType();
uint64_t DstSize = CGM.getDataLayout().getTypeAllocSize(DstTy);
Address AddrToStoreInto = Address::invalid();
if (SrcSize <= DstSize) {
AddrToStoreInto =
Builder.CreateBitCast(Ptr, llvm::PointerType::getUnqual(STy));
} else {
AddrToStoreInto =
CreateTempAlloca(STy, Alloca.getAlignment(), "coerce");
}
assert(STy->getNumElements() == NumIRArgs);
for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
auto AI = FnArgs[FirstIRArg + i];
AI->setName(Arg->getName() + ".coerce" + Twine(i));
auto Offset = CharUnits::fromQuantity(SrcLayout->getElementOffset(i));
Address EltPtr =
Builder.CreateStructGEP(AddrToStoreInto, i, Offset);
Builder.CreateStore(AI, EltPtr);
}
if (SrcSize > DstSize) {
Builder.CreateMemCpy(Ptr, AddrToStoreInto, DstSize);
}
} else {
// Simple case, just do a coerced store of the argument into the alloca.
assert(NumIRArgs == 1);
auto AI = FnArgs[FirstIRArg];
AI->setName(Arg->getName() + ".coerce");
CreateCoercedStore(AI, Ptr, /*DestIsVolatile=*/false, *this);
}
// Match to what EmitParmDecl is expecting for this type.
if (CodeGenFunction::hasScalarEvaluationKind(Ty)) {
llvm::Value *V =
EmitLoadOfScalar(Alloca, false, Ty, Arg->getLocStart());
if (isPromoted)
V = emitArgumentDemotion(*this, Arg, V);
ArgVals.push_back(ParamValue::forDirect(V));
} else {
ArgVals.push_back(ParamValue::forIndirect(Alloca));
}
break;
}
case ABIArgInfo::Expand: {
// If this structure was expanded into multiple arguments then
// we need to create a temporary and reconstruct it from the
// arguments.
Address Alloca = CreateMemTemp(Ty, getContext().getDeclAlign(Arg));
LValue LV = MakeAddrLValue(Alloca, Ty);
ArgVals.push_back(ParamValue::forIndirect(Alloca));
auto FnArgIter = FnArgs.begin() + FirstIRArg;
ExpandTypeFromArgs(Ty, LV, FnArgIter);
assert(FnArgIter == FnArgs.begin() + FirstIRArg + NumIRArgs);
for (unsigned i = 0, e = NumIRArgs; i != e; ++i) {
auto AI = FnArgs[FirstIRArg + i];
AI->setName(Arg->getName() + "." + Twine(i));
}
break;
}
case ABIArgInfo::Ignore:
assert(NumIRArgs == 0);
// Initialize the local variable appropriately.
if (!hasScalarEvaluationKind(Ty)) {
ArgVals.push_back(ParamValue::forIndirect(CreateMemTemp(Ty)));
} else {
llvm::Value *U = llvm::UndefValue::get(ConvertType(Arg->getType()));
ArgVals.push_back(ParamValue::forDirect(U));
}
break;
}
}
if (getTarget().getCXXABI().areArgsDestroyedLeftToRightInCallee()) {
for (int I = Args.size() - 1; I >= 0; --I)
EmitParmDecl(*Args[I], ArgVals[I], I + 1);
} else {
for (unsigned I = 0, E = Args.size(); I != E; ++I)
EmitParmDecl(*Args[I], ArgVals[I], I + 1);
}
}
static void eraseUnusedBitCasts(llvm::Instruction *insn) {
while (insn->use_empty()) {
llvm::BitCastInst *bitcast = dyn_cast<llvm::BitCastInst>(insn);
if (!bitcast) return;
// This is "safe" because we would have used a ConstantExpr otherwise.
insn = cast<llvm::Instruction>(bitcast->getOperand(0));
bitcast->eraseFromParent();
}
}
/// Try to emit a fused autorelease of a return result.
static llvm::Value *tryEmitFusedAutoreleaseOfResult(CodeGenFunction &CGF,
llvm::Value *result) {
// We must be immediately followed the cast.
llvm::BasicBlock *BB = CGF.Builder.GetInsertBlock();
if (BB->empty()) return nullptr;
if (&BB->back() != result) return nullptr;
llvm::Type *resultType = result->getType();
// result is in a BasicBlock and is therefore an Instruction.
llvm::Instruction *generator = cast<llvm::Instruction>(result);
SmallVector<llvm::Instruction*,4> insnsToKill;
// Look for:
// %generator = bitcast %type1* %generator2 to %type2*
while (llvm::BitCastInst *bitcast = dyn_cast<llvm::BitCastInst>(generator)) {
// We would have emitted this as a constant if the operand weren't
// an Instruction.
generator = cast<llvm::Instruction>(bitcast->getOperand(0));
// Require the generator to be immediately followed by the cast.
if (generator->getNextNode() != bitcast)
return nullptr;
insnsToKill.push_back(bitcast);
}
// Look for:
// %generator = call i8* @objc_retain(i8* %originalResult)
// or
// %generator = call i8* @objc_retainAutoreleasedReturnValue(i8* %originalResult)
llvm::CallInst *call = dyn_cast<llvm::CallInst>(generator);
if (!call) return nullptr;
bool doRetainAutorelease;
if (call->getCalledValue() == CGF.CGM.getObjCEntrypoints().objc_retain) {
doRetainAutorelease = true;
} else if (call->getCalledValue() == CGF.CGM.getObjCEntrypoints()
.objc_retainAutoreleasedReturnValue) {
doRetainAutorelease = false;
// If we emitted an assembly marker for this call (and the
// ARCEntrypoints field should have been set if so), go looking
// for that call. If we can't find it, we can't do this
// optimization. But it should always be the immediately previous
// instruction, unless we needed bitcasts around the call.
if (CGF.CGM.getObjCEntrypoints().retainAutoreleasedReturnValueMarker) {
llvm::Instruction *prev = call->getPrevNode();
assert(prev);
if (isa<llvm::BitCastInst>(prev)) {
prev = prev->getPrevNode();
assert(prev);
}
assert(isa<llvm::CallInst>(prev));
assert(cast<llvm::CallInst>(prev)->getCalledValue() ==
CGF.CGM.getObjCEntrypoints().retainAutoreleasedReturnValueMarker);
insnsToKill.push_back(prev);
}
} else {
return nullptr;
}
result = call->getArgOperand(0);
insnsToKill.push_back(call);
// Keep killing bitcasts, for sanity. Note that we no longer care
// about precise ordering as long as there's exactly one use.
while (llvm::BitCastInst *bitcast = dyn_cast<llvm::BitCastInst>(result)) {
if (!bitcast->hasOneUse()) break;
insnsToKill.push_back(bitcast);
result = bitcast->getOperand(0);
}
// Delete all the unnecessary instructions, from latest to earliest.
for (SmallVectorImpl<llvm::Instruction*>::iterator
i = insnsToKill.begin(), e = insnsToKill.end(); i != e; ++i)
(*i)->eraseFromParent();
// Do the fused retain/autorelease if we were asked to.
if (doRetainAutorelease)
result = CGF.EmitARCRetainAutoreleaseReturnValue(result);
// Cast back to the result type.
return CGF.Builder.CreateBitCast(result, resultType);
}
/// If this is a +1 of the value of an immutable 'self', remove it.
static llvm::Value *tryRemoveRetainOfSelf(CodeGenFunction &CGF,
llvm::Value *result) {
// This is only applicable to a method with an immutable 'self'.
const ObjCMethodDecl *method =
dyn_cast_or_null<ObjCMethodDecl>(CGF.CurCodeDecl);
if (!method) return nullptr;
const VarDecl *self = method->getSelfDecl();
if (!self->getType().isConstQualified()) return nullptr;
// Look for a retain call.
llvm::CallInst *retainCall =
dyn_cast<llvm::CallInst>(result->stripPointerCasts());
if (!retainCall ||
retainCall->getCalledValue() != CGF.CGM.getObjCEntrypoints().objc_retain)
return nullptr;
// Look for an ordinary load of 'self'.
llvm::Value *retainedValue = retainCall->getArgOperand(0);
llvm::LoadInst *load =
dyn_cast<llvm::LoadInst>(retainedValue->stripPointerCasts());
if (!load || load->isAtomic() || load->isVolatile() ||
load->getPointerOperand() != CGF.GetAddrOfLocalVar(self).getPointer())
return nullptr;
// Okay! Burn it all down. This relies for correctness on the
// assumption that the retain is emitted as part of the return and
// that thereafter everything is used "linearly".
llvm::Type *resultType = result->getType();
eraseUnusedBitCasts(cast<llvm::Instruction>(result));
assert(retainCall->use_empty());
retainCall->eraseFromParent();
eraseUnusedBitCasts(cast<llvm::Instruction>(retainedValue));
return CGF.Builder.CreateBitCast(load, resultType);
}
/// Emit an ARC autorelease of the result of a function.
///
/// \return the value to actually return from the function
static llvm::Value *emitAutoreleaseOfResult(CodeGenFunction &CGF,
llvm::Value *result) {
// If we're returning 'self', kill the initial retain. This is a
// heuristic attempt to "encourage correctness" in the really unfortunate
// case where we have a return of self during a dealloc and we desperately
// need to avoid the possible autorelease.
if (llvm::Value *self = tryRemoveRetainOfSelf(CGF, result))
return self;
// At -O0, try to emit a fused retain/autorelease.
if (CGF.shouldUseFusedARCCalls())
if (llvm::Value *fused = tryEmitFusedAutoreleaseOfResult(CGF, result))
return fused;
return CGF.EmitARCAutoreleaseReturnValue(result);
}
/// Heuristically search for a dominating store to the return-value slot.
static llvm::StoreInst *findDominatingStoreToReturnValue(CodeGenFunction &CGF) {
// Check if a User is a store which pointerOperand is the ReturnValue.
// We are looking for stores to the ReturnValue, not for stores of the
// ReturnValue to some other location.
auto GetStoreIfValid = [&CGF](llvm::User *U) -> llvm::StoreInst * {
auto *SI = dyn_cast<llvm::StoreInst>(U);
if (!SI || SI->getPointerOperand() != CGF.ReturnValue.getPointer())
return nullptr;
// These aren't actually possible for non-coerced returns, and we
// only care about non-coerced returns on this code path.
assert(!SI->isAtomic() && !SI->isVolatile());
return SI;
};
// If there are multiple uses of the return-value slot, just check
// for something immediately preceding the IP. Sometimes this can
// happen with how we generate implicit-returns; it can also happen
// with noreturn cleanups.
if (!CGF.ReturnValue.getPointer()->hasOneUse()) {
llvm::BasicBlock *IP = CGF.Builder.GetInsertBlock();
if (IP->empty()) return nullptr;
llvm::Instruction *I = &IP->back();
// Skip lifetime markers
for (llvm::BasicBlock::reverse_iterator II = IP->rbegin(),
IE = IP->rend();
II != IE; ++II) {
if (llvm::IntrinsicInst *Intrinsic =
dyn_cast<llvm::IntrinsicInst>(&*II)) {
if (Intrinsic->getIntrinsicID() == llvm::Intrinsic::lifetime_end) {
const llvm::Value *CastAddr = Intrinsic->getArgOperand(1);
++II;
if (II == IE)
break;
if (isa<llvm::BitCastInst>(&*II) && (CastAddr == &*II))
continue;
}
}
I = &*II;
break;
}
return GetStoreIfValid(I);
}
llvm::StoreInst *store =
GetStoreIfValid(CGF.ReturnValue.getPointer()->user_back());
if (!store) return nullptr;
// Now do a first-and-dirty dominance check: just walk up the
// single-predecessors chain from the current insertion point.
llvm::BasicBlock *StoreBB = store->getParent();
llvm::BasicBlock *IP = CGF.Builder.GetInsertBlock();
while (IP != StoreBB) {
if (!(IP = IP->getSinglePredecessor()))
return nullptr;
}
// Okay, the store's basic block dominates the insertion point; we
// can do our thing.
return store;
}
void CodeGenFunction::EmitFunctionEpilog(const CGFunctionInfo &FI,
bool EmitRetDbgLoc,
SourceLocation EndLoc) {
if (CurCodeDecl && CurCodeDecl->hasAttr<NakedAttr>()) {
// Naked functions don't have epilogues.
Builder.CreateUnreachable();
return;
}
// Functions with no result always return void.
if (!ReturnValue.isValid()) {
Builder.CreateRetVoid();
return;
}
llvm::DebugLoc RetDbgLoc;
llvm::Value *RV = nullptr;
QualType RetTy = FI.getReturnType();