blob: d8e363387b0767471162565102b84ba087f364eb [file] [log] [blame]
//===---- CGBuiltin.cpp - Emit LLVM Code for builtins ---------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This contains code to emit Builtin calls as LLVM code.
//
//===----------------------------------------------------------------------===//
#include "CodeGenFunction.h"
#include "CGObjCRuntime.h"
#include "CodeGenModule.h"
#include "TargetInfo.h"
#include "clang/AST/ASTContext.h"
#include "clang/AST/Decl.h"
#include "clang/Basic/TargetBuiltins.h"
#include "clang/Basic/TargetInfo.h"
#include "clang/CodeGen/CGFunctionInfo.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Intrinsics.h"
using namespace clang;
using namespace CodeGen;
using namespace llvm;
/// getBuiltinLibFunction - Given a builtin id for a function like
/// "__builtin_fabsf", return a Function* for "fabsf".
llvm::Value *CodeGenModule::getBuiltinLibFunction(const FunctionDecl *FD,
unsigned BuiltinID) {
assert(Context.BuiltinInfo.isLibFunction(BuiltinID));
// Get the name, skip over the __builtin_ prefix (if necessary).
StringRef Name;
GlobalDecl D(FD);
// If the builtin has been declared explicitly with an assembler label,
// use the mangled name. This differs from the plain label on platforms
// that prefix labels.
if (FD->hasAttr<AsmLabelAttr>())
Name = getMangledName(D);
else
Name = Context.BuiltinInfo.GetName(BuiltinID) + 10;
llvm::FunctionType *Ty =
cast<llvm::FunctionType>(getTypes().ConvertType(FD->getType()));
return GetOrCreateLLVMFunction(Name, Ty, D, /*ForVTable=*/false);
}
/// Emit the conversions required to turn the given value into an
/// integer of the given size.
static Value *EmitToInt(CodeGenFunction &CGF, llvm::Value *V,
QualType T, llvm::IntegerType *IntType) {
V = CGF.EmitToMemory(V, T);
if (V->getType()->isPointerTy())
return CGF.Builder.CreatePtrToInt(V, IntType);
assert(V->getType() == IntType);
return V;
}
static Value *EmitFromInt(CodeGenFunction &CGF, llvm::Value *V,
QualType T, llvm::Type *ResultType) {
V = CGF.EmitFromMemory(V, T);
if (ResultType->isPointerTy())
return CGF.Builder.CreateIntToPtr(V, ResultType);
assert(V->getType() == ResultType);
return V;
}
/// Utility to insert an atomic instruction based on Instrinsic::ID
/// and the expression node.
static RValue EmitBinaryAtomic(CodeGenFunction &CGF,
llvm::AtomicRMWInst::BinOp Kind,
const CallExpr *E) {
QualType T = E->getType();
assert(E->getArg(0)->getType()->isPointerType());
assert(CGF.getContext().hasSameUnqualifiedType(T,
E->getArg(0)->getType()->getPointeeType()));
assert(CGF.getContext().hasSameUnqualifiedType(T, E->getArg(1)->getType()));
llvm::Value *DestPtr = CGF.EmitScalarExpr(E->getArg(0));
unsigned AddrSpace = DestPtr->getType()->getPointerAddressSpace();
llvm::IntegerType *IntType =
llvm::IntegerType::get(CGF.getLLVMContext(),
CGF.getContext().getTypeSize(T));
llvm::Type *IntPtrType = IntType->getPointerTo(AddrSpace);
llvm::Value *Args[2];
Args[0] = CGF.Builder.CreateBitCast(DestPtr, IntPtrType);
Args[1] = CGF.EmitScalarExpr(E->getArg(1));
llvm::Type *ValueType = Args[1]->getType();
Args[1] = EmitToInt(CGF, Args[1], T, IntType);
llvm::Value *Result =
CGF.Builder.CreateAtomicRMW(Kind, Args[0], Args[1],
llvm::SequentiallyConsistent);
Result = EmitFromInt(CGF, Result, T, ValueType);
return RValue::get(Result);
}
/// Utility to insert an atomic instruction based Instrinsic::ID and
/// the expression node, where the return value is the result of the
/// operation.
static RValue EmitBinaryAtomicPost(CodeGenFunction &CGF,
llvm::AtomicRMWInst::BinOp Kind,
const CallExpr *E,
Instruction::BinaryOps Op) {
QualType T = E->getType();
assert(E->getArg(0)->getType()->isPointerType());
assert(CGF.getContext().hasSameUnqualifiedType(T,
E->getArg(0)->getType()->getPointeeType()));
assert(CGF.getContext().hasSameUnqualifiedType(T, E->getArg(1)->getType()));
llvm::Value *DestPtr = CGF.EmitScalarExpr(E->getArg(0));
unsigned AddrSpace = DestPtr->getType()->getPointerAddressSpace();
llvm::IntegerType *IntType =
llvm::IntegerType::get(CGF.getLLVMContext(),
CGF.getContext().getTypeSize(T));
llvm::Type *IntPtrType = IntType->getPointerTo(AddrSpace);
llvm::Value *Args[2];
Args[1] = CGF.EmitScalarExpr(E->getArg(1));
llvm::Type *ValueType = Args[1]->getType();
Args[1] = EmitToInt(CGF, Args[1], T, IntType);
Args[0] = CGF.Builder.CreateBitCast(DestPtr, IntPtrType);
llvm::Value *Result =
CGF.Builder.CreateAtomicRMW(Kind, Args[0], Args[1],
llvm::SequentiallyConsistent);
Result = CGF.Builder.CreateBinOp(Op, Result, Args[1]);
Result = EmitFromInt(CGF, Result, T, ValueType);
return RValue::get(Result);
}
/// EmitFAbs - Emit a call to fabs/fabsf/fabsl, depending on the type of ValTy,
/// which must be a scalar floating point type.
static Value *EmitFAbs(CodeGenFunction &CGF, Value *V, QualType ValTy) {
const BuiltinType *ValTyP = ValTy->getAs<BuiltinType>();
assert(ValTyP && "isn't scalar fp type!");
StringRef FnName;
switch (ValTyP->getKind()) {
default: llvm_unreachable("Isn't a scalar fp type!");
case BuiltinType::Float: FnName = "fabsf"; break;
case BuiltinType::Double: FnName = "fabs"; break;
case BuiltinType::LongDouble: FnName = "fabsl"; break;
}
// The prototype is something that takes and returns whatever V's type is.
llvm::FunctionType *FT = llvm::FunctionType::get(V->getType(), V->getType(),
false);
llvm::Value *Fn = CGF.CGM.CreateRuntimeFunction(FT, FnName);
return CGF.EmitNounwindRuntimeCall(Fn, V, "abs");
}
static RValue emitLibraryCall(CodeGenFunction &CGF, const FunctionDecl *Fn,
const CallExpr *E, llvm::Value *calleeValue) {
return CGF.EmitCall(E->getCallee()->getType(), calleeValue, E->getLocStart(),
ReturnValueSlot(), E->arg_begin(), E->arg_end(), Fn);
}
/// \brief Emit a call to llvm.{sadd,uadd,ssub,usub,smul,umul}.with.overflow.*
/// depending on IntrinsicID.
///
/// \arg CGF The current codegen function.
/// \arg IntrinsicID The ID for the Intrinsic we wish to generate.
/// \arg X The first argument to the llvm.*.with.overflow.*.
/// \arg Y The second argument to the llvm.*.with.overflow.*.
/// \arg Carry The carry returned by the llvm.*.with.overflow.*.
/// \returns The result (i.e. sum/product) returned by the intrinsic.
static llvm::Value *EmitOverflowIntrinsic(CodeGenFunction &CGF,
const llvm::Intrinsic::ID IntrinsicID,
llvm::Value *X, llvm::Value *Y,
llvm::Value *&Carry) {
// Make sure we have integers of the same width.
assert(X->getType() == Y->getType() &&
"Arguments must be the same type. (Did you forget to make sure both "
"arguments have the same integer width?)");
llvm::Value *Callee = CGF.CGM.getIntrinsic(IntrinsicID, X->getType());
llvm::Value *Tmp = CGF.Builder.CreateCall2(Callee, X, Y);
Carry = CGF.Builder.CreateExtractValue(Tmp, 1);
return CGF.Builder.CreateExtractValue(Tmp, 0);
}
RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
unsigned BuiltinID, const CallExpr *E) {
// See if we can constant fold this builtin. If so, don't emit it at all.
Expr::EvalResult Result;
if (E->EvaluateAsRValue(Result, CGM.getContext()) &&
!Result.hasSideEffects()) {
if (Result.Val.isInt())
return RValue::get(llvm::ConstantInt::get(getLLVMContext(),
Result.Val.getInt()));
if (Result.Val.isFloat())
return RValue::get(llvm::ConstantFP::get(getLLVMContext(),
Result.Val.getFloat()));
}
switch (BuiltinID) {
default: break; // Handle intrinsics and libm functions below.
case Builtin::BI__builtin___CFStringMakeConstantString:
case Builtin::BI__builtin___NSStringMakeConstantString:
return RValue::get(CGM.EmitConstantExpr(E, E->getType(), 0));
case Builtin::BI__builtin_stdarg_start:
case Builtin::BI__builtin_va_start:
case Builtin::BI__va_start:
case Builtin::BI__builtin_va_end: {
Value *ArgValue = (BuiltinID == Builtin::BI__va_start)
? EmitScalarExpr(E->getArg(0))
: EmitVAListRef(E->getArg(0));
llvm::Type *DestType = Int8PtrTy;
if (ArgValue->getType() != DestType)
ArgValue = Builder.CreateBitCast(ArgValue, DestType,
ArgValue->getName().data());
Intrinsic::ID inst = (BuiltinID == Builtin::BI__builtin_va_end) ?
Intrinsic::vaend : Intrinsic::vastart;
return RValue::get(Builder.CreateCall(CGM.getIntrinsic(inst), ArgValue));
}
case Builtin::BI__builtin_va_copy: {
Value *DstPtr = EmitVAListRef(E->getArg(0));
Value *SrcPtr = EmitVAListRef(E->getArg(1));
llvm::Type *Type = Int8PtrTy;
DstPtr = Builder.CreateBitCast(DstPtr, Type);
SrcPtr = Builder.CreateBitCast(SrcPtr, Type);
return RValue::get(Builder.CreateCall2(CGM.getIntrinsic(Intrinsic::vacopy),
DstPtr, SrcPtr));
}
case Builtin::BI__builtin_abs:
case Builtin::BI__builtin_labs:
case Builtin::BI__builtin_llabs: {
Value *ArgValue = EmitScalarExpr(E->getArg(0));
Value *NegOp = Builder.CreateNeg(ArgValue, "neg");
Value *CmpResult =
Builder.CreateICmpSGE(ArgValue,
llvm::Constant::getNullValue(ArgValue->getType()),
"abscond");
Value *Result =
Builder.CreateSelect(CmpResult, ArgValue, NegOp, "abs");
return RValue::get(Result);
}
case Builtin::BI__builtin_conj:
case Builtin::BI__builtin_conjf:
case Builtin::BI__builtin_conjl: {
ComplexPairTy ComplexVal = EmitComplexExpr(E->getArg(0));
Value *Real = ComplexVal.first;
Value *Imag = ComplexVal.second;
Value *Zero =
Imag->getType()->isFPOrFPVectorTy()
? llvm::ConstantFP::getZeroValueForNegation(Imag->getType())
: llvm::Constant::getNullValue(Imag->getType());
Imag = Builder.CreateFSub(Zero, Imag, "sub");
return RValue::getComplex(std::make_pair(Real, Imag));
}
case Builtin::BI__builtin_creal:
case Builtin::BI__builtin_crealf:
case Builtin::BI__builtin_creall:
case Builtin::BIcreal:
case Builtin::BIcrealf:
case Builtin::BIcreall: {
ComplexPairTy ComplexVal = EmitComplexExpr(E->getArg(0));
return RValue::get(ComplexVal.first);
}
case Builtin::BI__builtin_cimag:
case Builtin::BI__builtin_cimagf:
case Builtin::BI__builtin_cimagl:
case Builtin::BIcimag:
case Builtin::BIcimagf:
case Builtin::BIcimagl: {
ComplexPairTy ComplexVal = EmitComplexExpr(E->getArg(0));
return RValue::get(ComplexVal.second);
}
case Builtin::BI__builtin_ctzs:
case Builtin::BI__builtin_ctz:
case Builtin::BI__builtin_ctzl:
case Builtin::BI__builtin_ctzll: {
Value *ArgValue = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgType = ArgValue->getType();
Value *F = CGM.getIntrinsic(Intrinsic::cttz, ArgType);
llvm::Type *ResultType = ConvertType(E->getType());
Value *ZeroUndef = Builder.getInt1(getTarget().isCLZForZeroUndef());
Value *Result = Builder.CreateCall2(F, ArgValue, ZeroUndef);
if (Result->getType() != ResultType)
Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/true,
"cast");
return RValue::get(Result);
}
case Builtin::BI__builtin_clzs:
case Builtin::BI__builtin_clz:
case Builtin::BI__builtin_clzl:
case Builtin::BI__builtin_clzll: {
Value *ArgValue = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgType = ArgValue->getType();
Value *F = CGM.getIntrinsic(Intrinsic::ctlz, ArgType);
llvm::Type *ResultType = ConvertType(E->getType());
Value *ZeroUndef = Builder.getInt1(getTarget().isCLZForZeroUndef());
Value *Result = Builder.CreateCall2(F, ArgValue, ZeroUndef);
if (Result->getType() != ResultType)
Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/true,
"cast");
return RValue::get(Result);
}
case Builtin::BI__builtin_ffs:
case Builtin::BI__builtin_ffsl:
case Builtin::BI__builtin_ffsll: {
// ffs(x) -> x ? cttz(x) + 1 : 0
Value *ArgValue = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgType = ArgValue->getType();
Value *F = CGM.getIntrinsic(Intrinsic::cttz, ArgType);
llvm::Type *ResultType = ConvertType(E->getType());
Value *Tmp = Builder.CreateAdd(Builder.CreateCall2(F, ArgValue,
Builder.getTrue()),
llvm::ConstantInt::get(ArgType, 1));
Value *Zero = llvm::Constant::getNullValue(ArgType);
Value *IsZero = Builder.CreateICmpEQ(ArgValue, Zero, "iszero");
Value *Result = Builder.CreateSelect(IsZero, Zero, Tmp, "ffs");
if (Result->getType() != ResultType)
Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/true,
"cast");
return RValue::get(Result);
}
case Builtin::BI__builtin_parity:
case Builtin::BI__builtin_parityl:
case Builtin::BI__builtin_parityll: {
// parity(x) -> ctpop(x) & 1
Value *ArgValue = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgType = ArgValue->getType();
Value *F = CGM.getIntrinsic(Intrinsic::ctpop, ArgType);
llvm::Type *ResultType = ConvertType(E->getType());
Value *Tmp = Builder.CreateCall(F, ArgValue);
Value *Result = Builder.CreateAnd(Tmp, llvm::ConstantInt::get(ArgType, 1));
if (Result->getType() != ResultType)
Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/true,
"cast");
return RValue::get(Result);
}
case Builtin::BI__builtin_popcount:
case Builtin::BI__builtin_popcountl:
case Builtin::BI__builtin_popcountll: {
Value *ArgValue = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgType = ArgValue->getType();
Value *F = CGM.getIntrinsic(Intrinsic::ctpop, ArgType);
llvm::Type *ResultType = ConvertType(E->getType());
Value *Result = Builder.CreateCall(F, ArgValue);
if (Result->getType() != ResultType)
Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/true,
"cast");
return RValue::get(Result);
}
case Builtin::BI__builtin_expect: {
Value *ArgValue = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgType = ArgValue->getType();
Value *FnExpect = CGM.getIntrinsic(Intrinsic::expect, ArgType);
Value *ExpectedValue = EmitScalarExpr(E->getArg(1));
Value *Result = Builder.CreateCall2(FnExpect, ArgValue, ExpectedValue,
"expval");
return RValue::get(Result);
}
case Builtin::BI__builtin_bswap16:
case Builtin::BI__builtin_bswap32:
case Builtin::BI__builtin_bswap64: {
Value *ArgValue = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgType = ArgValue->getType();
Value *F = CGM.getIntrinsic(Intrinsic::bswap, ArgType);
return RValue::get(Builder.CreateCall(F, ArgValue));
}
case Builtin::BI__builtin_object_size: {
// We rely on constant folding to deal with expressions with side effects.
assert(!E->getArg(0)->HasSideEffects(getContext()) &&
"should have been constant folded");
// We pass this builtin onto the optimizer so that it can
// figure out the object size in more complex cases.
llvm::Type *ResType = ConvertType(E->getType());
// LLVM only supports 0 and 2, make sure that we pass along that
// as a boolean.
Value *Ty = EmitScalarExpr(E->getArg(1));
ConstantInt *CI = dyn_cast<ConstantInt>(Ty);
assert(CI);
uint64_t val = CI->getZExtValue();
CI = ConstantInt::get(Builder.getInt1Ty(), (val & 0x2) >> 1);
// FIXME: Get right address space.
llvm::Type *Tys[] = { ResType, Builder.getInt8PtrTy(0) };
Value *F = CGM.getIntrinsic(Intrinsic::objectsize, Tys);
return RValue::get(Builder.CreateCall2(F, EmitScalarExpr(E->getArg(0)),CI));
}
case Builtin::BI__builtin_prefetch: {
Value *Locality, *RW, *Address = EmitScalarExpr(E->getArg(0));
// FIXME: Technically these constants should of type 'int', yes?
RW = (E->getNumArgs() > 1) ? EmitScalarExpr(E->getArg(1)) :
llvm::ConstantInt::get(Int32Ty, 0);
Locality = (E->getNumArgs() > 2) ? EmitScalarExpr(E->getArg(2)) :
llvm::ConstantInt::get(Int32Ty, 3);
Value *Data = llvm::ConstantInt::get(Int32Ty, 1);
Value *F = CGM.getIntrinsic(Intrinsic::prefetch);
return RValue::get(Builder.CreateCall4(F, Address, RW, Locality, Data));
}
case Builtin::BI__builtin_readcyclecounter: {
Value *F = CGM.getIntrinsic(Intrinsic::readcyclecounter);
return RValue::get(Builder.CreateCall(F));
}
case Builtin::BI__builtin___clear_cache: {
Value *Begin = EmitScalarExpr(E->getArg(0));
Value *End = EmitScalarExpr(E->getArg(1));
Value *F = CGM.getIntrinsic(Intrinsic::clear_cache);
return RValue::get(Builder.CreateCall2(F, Begin, End));
}
case Builtin::BI__builtin_trap: {
Value *F = CGM.getIntrinsic(Intrinsic::trap);
return RValue::get(Builder.CreateCall(F));
}
case Builtin::BI__debugbreak: {
Value *F = CGM.getIntrinsic(Intrinsic::debugtrap);
return RValue::get(Builder.CreateCall(F));
}
case Builtin::BI__builtin_unreachable: {
if (SanOpts->Unreachable)
EmitCheck(Builder.getFalse(), "builtin_unreachable",
EmitCheckSourceLocation(E->getExprLoc()),
ArrayRef<llvm::Value *>(), CRK_Unrecoverable);
else
Builder.CreateUnreachable();
// We do need to preserve an insertion point.
EmitBlock(createBasicBlock("unreachable.cont"));
return RValue::get(0);
}
case Builtin::BI__builtin_powi:
case Builtin::BI__builtin_powif:
case Builtin::BI__builtin_powil: {
Value *Base = EmitScalarExpr(E->getArg(0));
Value *Exponent = EmitScalarExpr(E->getArg(1));
llvm::Type *ArgType = Base->getType();
Value *F = CGM.getIntrinsic(Intrinsic::powi, ArgType);
return RValue::get(Builder.CreateCall2(F, Base, Exponent));
}
case Builtin::BI__builtin_isgreater:
case Builtin::BI__builtin_isgreaterequal:
case Builtin::BI__builtin_isless:
case Builtin::BI__builtin_islessequal:
case Builtin::BI__builtin_islessgreater:
case Builtin::BI__builtin_isunordered: {
// Ordered comparisons: we know the arguments to these are matching scalar
// floating point values.
Value *LHS = EmitScalarExpr(E->getArg(0));
Value *RHS = EmitScalarExpr(E->getArg(1));
switch (BuiltinID) {
default: llvm_unreachable("Unknown ordered comparison");
case Builtin::BI__builtin_isgreater:
LHS = Builder.CreateFCmpOGT(LHS, RHS, "cmp");
break;
case Builtin::BI__builtin_isgreaterequal:
LHS = Builder.CreateFCmpOGE(LHS, RHS, "cmp");
break;
case Builtin::BI__builtin_isless:
LHS = Builder.CreateFCmpOLT(LHS, RHS, "cmp");
break;
case Builtin::BI__builtin_islessequal:
LHS = Builder.CreateFCmpOLE(LHS, RHS, "cmp");
break;
case Builtin::BI__builtin_islessgreater:
LHS = Builder.CreateFCmpONE(LHS, RHS, "cmp");
break;
case Builtin::BI__builtin_isunordered:
LHS = Builder.CreateFCmpUNO(LHS, RHS, "cmp");
break;
}
// ZExt bool to int type.
return RValue::get(Builder.CreateZExt(LHS, ConvertType(E->getType())));
}
case Builtin::BI__builtin_isnan: {
Value *V = EmitScalarExpr(E->getArg(0));
V = Builder.CreateFCmpUNO(V, V, "cmp");
return RValue::get(Builder.CreateZExt(V, ConvertType(E->getType())));
}
case Builtin::BI__builtin_isinf: {
// isinf(x) --> fabs(x) == infinity
Value *V = EmitScalarExpr(E->getArg(0));
V = EmitFAbs(*this, V, E->getArg(0)->getType());
V = Builder.CreateFCmpOEQ(V, ConstantFP::getInfinity(V->getType()),"isinf");
return RValue::get(Builder.CreateZExt(V, ConvertType(E->getType())));
}
// TODO: BI__builtin_isinf_sign
// isinf_sign(x) -> isinf(x) ? (signbit(x) ? -1 : 1) : 0
case Builtin::BI__builtin_isnormal: {
// isnormal(x) --> x == x && fabsf(x) < infinity && fabsf(x) >= float_min
Value *V = EmitScalarExpr(E->getArg(0));
Value *Eq = Builder.CreateFCmpOEQ(V, V, "iseq");
Value *Abs = EmitFAbs(*this, V, E->getArg(0)->getType());
Value *IsLessThanInf =
Builder.CreateFCmpULT(Abs, ConstantFP::getInfinity(V->getType()),"isinf");
APFloat Smallest = APFloat::getSmallestNormalized(
getContext().getFloatTypeSemantics(E->getArg(0)->getType()));
Value *IsNormal =
Builder.CreateFCmpUGE(Abs, ConstantFP::get(V->getContext(), Smallest),
"isnormal");
V = Builder.CreateAnd(Eq, IsLessThanInf, "and");
V = Builder.CreateAnd(V, IsNormal, "and");
return RValue::get(Builder.CreateZExt(V, ConvertType(E->getType())));
}
case Builtin::BI__builtin_isfinite: {
// isfinite(x) --> x == x && fabs(x) != infinity;
Value *V = EmitScalarExpr(E->getArg(0));
Value *Eq = Builder.CreateFCmpOEQ(V, V, "iseq");
Value *Abs = EmitFAbs(*this, V, E->getArg(0)->getType());
Value *IsNotInf =
Builder.CreateFCmpUNE(Abs, ConstantFP::getInfinity(V->getType()),"isinf");
V = Builder.CreateAnd(Eq, IsNotInf, "and");
return RValue::get(Builder.CreateZExt(V, ConvertType(E->getType())));
}
case Builtin::BI__builtin_fpclassify: {
Value *V = EmitScalarExpr(E->getArg(5));
llvm::Type *Ty = ConvertType(E->getArg(5)->getType());
// Create Result
BasicBlock *Begin = Builder.GetInsertBlock();
BasicBlock *End = createBasicBlock("fpclassify_end", this->CurFn);
Builder.SetInsertPoint(End);
PHINode *Result =
Builder.CreatePHI(ConvertType(E->getArg(0)->getType()), 4,
"fpclassify_result");
// if (V==0) return FP_ZERO
Builder.SetInsertPoint(Begin);
Value *IsZero = Builder.CreateFCmpOEQ(V, Constant::getNullValue(Ty),
"iszero");
Value *ZeroLiteral = EmitScalarExpr(E->getArg(4));
BasicBlock *NotZero = createBasicBlock("fpclassify_not_zero", this->CurFn);
Builder.CreateCondBr(IsZero, End, NotZero);
Result->addIncoming(ZeroLiteral, Begin);
// if (V != V) return FP_NAN
Builder.SetInsertPoint(NotZero);
Value *IsNan = Builder.CreateFCmpUNO(V, V, "cmp");
Value *NanLiteral = EmitScalarExpr(E->getArg(0));
BasicBlock *NotNan = createBasicBlock("fpclassify_not_nan", this->CurFn);
Builder.CreateCondBr(IsNan, End, NotNan);
Result->addIncoming(NanLiteral, NotZero);
// if (fabs(V) == infinity) return FP_INFINITY
Builder.SetInsertPoint(NotNan);
Value *VAbs = EmitFAbs(*this, V, E->getArg(5)->getType());
Value *IsInf =
Builder.CreateFCmpOEQ(VAbs, ConstantFP::getInfinity(V->getType()),
"isinf");
Value *InfLiteral = EmitScalarExpr(E->getArg(1));
BasicBlock *NotInf = createBasicBlock("fpclassify_not_inf", this->CurFn);
Builder.CreateCondBr(IsInf, End, NotInf);
Result->addIncoming(InfLiteral, NotNan);
// if (fabs(V) >= MIN_NORMAL) return FP_NORMAL else FP_SUBNORMAL
Builder.SetInsertPoint(NotInf);
APFloat Smallest = APFloat::getSmallestNormalized(
getContext().getFloatTypeSemantics(E->getArg(5)->getType()));
Value *IsNormal =
Builder.CreateFCmpUGE(VAbs, ConstantFP::get(V->getContext(), Smallest),
"isnormal");
Value *NormalResult =
Builder.CreateSelect(IsNormal, EmitScalarExpr(E->getArg(2)),
EmitScalarExpr(E->getArg(3)));
Builder.CreateBr(End);
Result->addIncoming(NormalResult, NotInf);
// return Result
Builder.SetInsertPoint(End);
return RValue::get(Result);
}
case Builtin::BIalloca:
case Builtin::BI_alloca:
case Builtin::BI__builtin_alloca: {
Value *Size = EmitScalarExpr(E->getArg(0));
return RValue::get(Builder.CreateAlloca(Builder.getInt8Ty(), Size));
}
case Builtin::BIbzero:
case Builtin::BI__builtin_bzero: {
std::pair<llvm::Value*, unsigned> Dest =
EmitPointerWithAlignment(E->getArg(0));
Value *SizeVal = EmitScalarExpr(E->getArg(1));
Builder.CreateMemSet(Dest.first, Builder.getInt8(0), SizeVal,
Dest.second, false);
return RValue::get(Dest.first);
}
case Builtin::BImemcpy:
case Builtin::BI__builtin_memcpy: {
std::pair<llvm::Value*, unsigned> Dest =
EmitPointerWithAlignment(E->getArg(0));
std::pair<llvm::Value*, unsigned> Src =
EmitPointerWithAlignment(E->getArg(1));
Value *SizeVal = EmitScalarExpr(E->getArg(2));
unsigned Align = std::min(Dest.second, Src.second);
Builder.CreateMemCpy(Dest.first, Src.first, SizeVal, Align, false);
return RValue::get(Dest.first);
}
case Builtin::BI__builtin___memcpy_chk: {
// fold __builtin_memcpy_chk(x, y, cst1, cst2) to memcpy iff cst1<=cst2.
llvm::APSInt Size, DstSize;
if (!E->getArg(2)->EvaluateAsInt(Size, CGM.getContext()) ||
!E->getArg(3)->EvaluateAsInt(DstSize, CGM.getContext()))
break;
if (Size.ugt(DstSize))
break;
std::pair<llvm::Value*, unsigned> Dest =
EmitPointerWithAlignment(E->getArg(0));
std::pair<llvm::Value*, unsigned> Src =
EmitPointerWithAlignment(E->getArg(1));
Value *SizeVal = llvm::ConstantInt::get(Builder.getContext(), Size);
unsigned Align = std::min(Dest.second, Src.second);
Builder.CreateMemCpy(Dest.first, Src.first, SizeVal, Align, false);
return RValue::get(Dest.first);
}
case Builtin::BI__builtin_objc_memmove_collectable: {
Value *Address = EmitScalarExpr(E->getArg(0));
Value *SrcAddr = EmitScalarExpr(E->getArg(1));
Value *SizeVal = EmitScalarExpr(E->getArg(2));
CGM.getObjCRuntime().EmitGCMemmoveCollectable(*this,
Address, SrcAddr, SizeVal);
return RValue::get(Address);
}
case Builtin::BI__builtin___memmove_chk: {
// fold __builtin_memmove_chk(x, y, cst1, cst2) to memmove iff cst1<=cst2.
llvm::APSInt Size, DstSize;
if (!E->getArg(2)->EvaluateAsInt(Size, CGM.getContext()) ||
!E->getArg(3)->EvaluateAsInt(DstSize, CGM.getContext()))
break;
if (Size.ugt(DstSize))
break;
std::pair<llvm::Value*, unsigned> Dest =
EmitPointerWithAlignment(E->getArg(0));
std::pair<llvm::Value*, unsigned> Src =
EmitPointerWithAlignment(E->getArg(1));
Value *SizeVal = llvm::ConstantInt::get(Builder.getContext(), Size);
unsigned Align = std::min(Dest.second, Src.second);
Builder.CreateMemMove(Dest.first, Src.first, SizeVal, Align, false);
return RValue::get(Dest.first);
}
case Builtin::BImemmove:
case Builtin::BI__builtin_memmove: {
std::pair<llvm::Value*, unsigned> Dest =
EmitPointerWithAlignment(E->getArg(0));
std::pair<llvm::Value*, unsigned> Src =
EmitPointerWithAlignment(E->getArg(1));
Value *SizeVal = EmitScalarExpr(E->getArg(2));
unsigned Align = std::min(Dest.second, Src.second);
Builder.CreateMemMove(Dest.first, Src.first, SizeVal, Align, false);
return RValue::get(Dest.first);
}
case Builtin::BImemset:
case Builtin::BI__builtin_memset: {
std::pair<llvm::Value*, unsigned> Dest =
EmitPointerWithAlignment(E->getArg(0));
Value *ByteVal = Builder.CreateTrunc(EmitScalarExpr(E->getArg(1)),
Builder.getInt8Ty());
Value *SizeVal = EmitScalarExpr(E->getArg(2));
Builder.CreateMemSet(Dest.first, ByteVal, SizeVal, Dest.second, false);
return RValue::get(Dest.first);
}
case Builtin::BI__builtin___memset_chk: {
// fold __builtin_memset_chk(x, y, cst1, cst2) to memset iff cst1<=cst2.
llvm::APSInt Size, DstSize;
if (!E->getArg(2)->EvaluateAsInt(Size, CGM.getContext()) ||
!E->getArg(3)->EvaluateAsInt(DstSize, CGM.getContext()))
break;
if (Size.ugt(DstSize))
break;
std::pair<llvm::Value*, unsigned> Dest =
EmitPointerWithAlignment(E->getArg(0));
Value *ByteVal = Builder.CreateTrunc(EmitScalarExpr(E->getArg(1)),
Builder.getInt8Ty());
Value *SizeVal = llvm::ConstantInt::get(Builder.getContext(), Size);
Builder.CreateMemSet(Dest.first, ByteVal, SizeVal, Dest.second, false);
return RValue::get(Dest.first);
}
case Builtin::BI__builtin_dwarf_cfa: {
// The offset in bytes from the first argument to the CFA.
//
// Why on earth is this in the frontend? Is there any reason at
// all that the backend can't reasonably determine this while
// lowering llvm.eh.dwarf.cfa()?
//
// TODO: If there's a satisfactory reason, add a target hook for
// this instead of hard-coding 0, which is correct for most targets.
int32_t Offset = 0;
Value *F = CGM.getIntrinsic(Intrinsic::eh_dwarf_cfa);
return RValue::get(Builder.CreateCall(F,
llvm::ConstantInt::get(Int32Ty, Offset)));
}
case Builtin::BI__builtin_return_address: {
Value *Depth = EmitScalarExpr(E->getArg(0));
Depth = Builder.CreateIntCast(Depth, Int32Ty, false);
Value *F = CGM.getIntrinsic(Intrinsic::returnaddress);
return RValue::get(Builder.CreateCall(F, Depth));
}
case Builtin::BI__builtin_frame_address: {
Value *Depth = EmitScalarExpr(E->getArg(0));
Depth = Builder.CreateIntCast(Depth, Int32Ty, false);
Value *F = CGM.getIntrinsic(Intrinsic::frameaddress);
return RValue::get(Builder.CreateCall(F, Depth));
}
case Builtin::BI__builtin_extract_return_addr: {
Value *Address = EmitScalarExpr(E->getArg(0));
Value *Result = getTargetHooks().decodeReturnAddress(*this, Address);
return RValue::get(Result);
}
case Builtin::BI__builtin_frob_return_addr: {
Value *Address = EmitScalarExpr(E->getArg(0));
Value *Result = getTargetHooks().encodeReturnAddress(*this, Address);
return RValue::get(Result);
}
case Builtin::BI__builtin_dwarf_sp_column: {
llvm::IntegerType *Ty
= cast<llvm::IntegerType>(ConvertType(E->getType()));
int Column = getTargetHooks().getDwarfEHStackPointer(CGM);
if (Column == -1) {
CGM.ErrorUnsupported(E, "__builtin_dwarf_sp_column");
return RValue::get(llvm::UndefValue::get(Ty));
}
return RValue::get(llvm::ConstantInt::get(Ty, Column, true));
}
case Builtin::BI__builtin_init_dwarf_reg_size_table: {
Value *Address = EmitScalarExpr(E->getArg(0));
if (getTargetHooks().initDwarfEHRegSizeTable(*this, Address))
CGM.ErrorUnsupported(E, "__builtin_init_dwarf_reg_size_table");
return RValue::get(llvm::UndefValue::get(ConvertType(E->getType())));
}
case Builtin::BI__builtin_eh_return: {
Value *Int = EmitScalarExpr(E->getArg(0));
Value *Ptr = EmitScalarExpr(E->getArg(1));
llvm::IntegerType *IntTy = cast<llvm::IntegerType>(Int->getType());
assert((IntTy->getBitWidth() == 32 || IntTy->getBitWidth() == 64) &&
"LLVM's __builtin_eh_return only supports 32- and 64-bit variants");
Value *F = CGM.getIntrinsic(IntTy->getBitWidth() == 32
? Intrinsic::eh_return_i32
: Intrinsic::eh_return_i64);
Builder.CreateCall2(F, Int, Ptr);
Builder.CreateUnreachable();
// We do need to preserve an insertion point.
EmitBlock(createBasicBlock("builtin_eh_return.cont"));
return RValue::get(0);
}
case Builtin::BI__builtin_unwind_init: {
Value *F = CGM.getIntrinsic(Intrinsic::eh_unwind_init);
return RValue::get(Builder.CreateCall(F));
}
case Builtin::BI__builtin_extend_pointer: {
// Extends a pointer to the size of an _Unwind_Word, which is
// uint64_t on all platforms. Generally this gets poked into a
// register and eventually used as an address, so if the
// addressing registers are wider than pointers and the platform
// doesn't implicitly ignore high-order bits when doing
// addressing, we need to make sure we zext / sext based on
// the platform's expectations.
//
// See: http://gcc.gnu.org/ml/gcc-bugs/2002-02/msg00237.html
// Cast the pointer to intptr_t.
Value *Ptr = EmitScalarExpr(E->getArg(0));
Value *Result = Builder.CreatePtrToInt(Ptr, IntPtrTy, "extend.cast");
// If that's 64 bits, we're done.
if (IntPtrTy->getBitWidth() == 64)
return RValue::get(Result);
// Otherwise, ask the codegen data what to do.
if (getTargetHooks().extendPointerWithSExt())
return RValue::get(Builder.CreateSExt(Result, Int64Ty, "extend.sext"));
else
return RValue::get(Builder.CreateZExt(Result, Int64Ty, "extend.zext"));
}
case Builtin::BI__builtin_setjmp: {
// Buffer is a void**.
Value *Buf = EmitScalarExpr(E->getArg(0));
// Store the frame pointer to the setjmp buffer.
Value *FrameAddr =
Builder.CreateCall(CGM.getIntrinsic(Intrinsic::frameaddress),
ConstantInt::get(Int32Ty, 0));
Builder.CreateStore(FrameAddr, Buf);
// Store the stack pointer to the setjmp buffer.
Value *StackAddr =
Builder.CreateCall(CGM.getIntrinsic(Intrinsic::stacksave));
Value *StackSaveSlot =
Builder.CreateGEP(Buf, ConstantInt::get(Int32Ty, 2));
Builder.CreateStore(StackAddr, StackSaveSlot);
// Call LLVM's EH setjmp, which is lightweight.
Value *F = CGM.getIntrinsic(Intrinsic::eh_sjlj_setjmp);
Buf = Builder.CreateBitCast(Buf, Int8PtrTy);
return RValue::get(Builder.CreateCall(F, Buf));
}
case Builtin::BI__builtin_longjmp: {
Value *Buf = EmitScalarExpr(E->getArg(0));
Buf = Builder.CreateBitCast(Buf, Int8PtrTy);
// Call LLVM's EH longjmp, which is lightweight.
Builder.CreateCall(CGM.getIntrinsic(Intrinsic::eh_sjlj_longjmp), Buf);
// longjmp doesn't return; mark this as unreachable.
Builder.CreateUnreachable();
// We do need to preserve an insertion point.
EmitBlock(createBasicBlock("longjmp.cont"));
return RValue::get(0);
}
case Builtin::BI__sync_fetch_and_add:
case Builtin::BI__sync_fetch_and_sub:
case Builtin::BI__sync_fetch_and_or:
case Builtin::BI__sync_fetch_and_and:
case Builtin::BI__sync_fetch_and_xor:
case Builtin::BI__sync_add_and_fetch:
case Builtin::BI__sync_sub_and_fetch:
case Builtin::BI__sync_and_and_fetch:
case Builtin::BI__sync_or_and_fetch:
case Builtin::BI__sync_xor_and_fetch:
case Builtin::BI__sync_val_compare_and_swap:
case Builtin::BI__sync_bool_compare_and_swap:
case Builtin::BI__sync_lock_test_and_set:
case Builtin::BI__sync_lock_release:
case Builtin::BI__sync_swap:
llvm_unreachable("Shouldn't make it through sema");
case Builtin::BI__sync_fetch_and_add_1:
case Builtin::BI__sync_fetch_and_add_2:
case Builtin::BI__sync_fetch_and_add_4:
case Builtin::BI__sync_fetch_and_add_8:
case Builtin::BI__sync_fetch_and_add_16:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::Add, E);
case Builtin::BI__sync_fetch_and_sub_1:
case Builtin::BI__sync_fetch_and_sub_2:
case Builtin::BI__sync_fetch_and_sub_4:
case Builtin::BI__sync_fetch_and_sub_8:
case Builtin::BI__sync_fetch_and_sub_16:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::Sub, E);
case Builtin::BI__sync_fetch_and_or_1:
case Builtin::BI__sync_fetch_and_or_2:
case Builtin::BI__sync_fetch_and_or_4:
case Builtin::BI__sync_fetch_and_or_8:
case Builtin::BI__sync_fetch_and_or_16:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::Or, E);
case Builtin::BI__sync_fetch_and_and_1:
case Builtin::BI__sync_fetch_and_and_2:
case Builtin::BI__sync_fetch_and_and_4:
case Builtin::BI__sync_fetch_and_and_8:
case Builtin::BI__sync_fetch_and_and_16:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::And, E);
case Builtin::BI__sync_fetch_and_xor_1:
case Builtin::BI__sync_fetch_and_xor_2:
case Builtin::BI__sync_fetch_and_xor_4:
case Builtin::BI__sync_fetch_and_xor_8:
case Builtin::BI__sync_fetch_and_xor_16:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::Xor, E);
// Clang extensions: not overloaded yet.
case Builtin::BI__sync_fetch_and_min:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::Min, E);
case Builtin::BI__sync_fetch_and_max:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::Max, E);
case Builtin::BI__sync_fetch_and_umin:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::UMin, E);
case Builtin::BI__sync_fetch_and_umax:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::UMax, E);
case Builtin::BI__sync_add_and_fetch_1:
case Builtin::BI__sync_add_and_fetch_2:
case Builtin::BI__sync_add_and_fetch_4:
case Builtin::BI__sync_add_and_fetch_8:
case Builtin::BI__sync_add_and_fetch_16:
return EmitBinaryAtomicPost(*this, llvm::AtomicRMWInst::Add, E,
llvm::Instruction::Add);
case Builtin::BI__sync_sub_and_fetch_1:
case Builtin::BI__sync_sub_and_fetch_2:
case Builtin::BI__sync_sub_and_fetch_4:
case Builtin::BI__sync_sub_and_fetch_8:
case Builtin::BI__sync_sub_and_fetch_16:
return EmitBinaryAtomicPost(*this, llvm::AtomicRMWInst::Sub, E,
llvm::Instruction::Sub);
case Builtin::BI__sync_and_and_fetch_1:
case Builtin::BI__sync_and_and_fetch_2:
case Builtin::BI__sync_and_and_fetch_4:
case Builtin::BI__sync_and_and_fetch_8:
case Builtin::BI__sync_and_and_fetch_16:
return EmitBinaryAtomicPost(*this, llvm::AtomicRMWInst::And, E,
llvm::Instruction::And);
case Builtin::BI__sync_or_and_fetch_1:
case Builtin::BI__sync_or_and_fetch_2:
case Builtin::BI__sync_or_and_fetch_4:
case Builtin::BI__sync_or_and_fetch_8:
case Builtin::BI__sync_or_and_fetch_16:
return EmitBinaryAtomicPost(*this, llvm::AtomicRMWInst::Or, E,
llvm::Instruction::Or);
case Builtin::BI__sync_xor_and_fetch_1:
case Builtin::BI__sync_xor_and_fetch_2:
case Builtin::BI__sync_xor_and_fetch_4:
case Builtin::BI__sync_xor_and_fetch_8:
case Builtin::BI__sync_xor_and_fetch_16:
return EmitBinaryAtomicPost(*this, llvm::AtomicRMWInst::Xor, E,
llvm::Instruction::Xor);
case Builtin::BI__sync_val_compare_and_swap_1:
case Builtin::BI__sync_val_compare_and_swap_2:
case Builtin::BI__sync_val_compare_and_swap_4:
case Builtin::BI__sync_val_compare_and_swap_8:
case Builtin::BI__sync_val_compare_and_swap_16: {
QualType T = E->getType();
llvm::Value *DestPtr = EmitScalarExpr(E->getArg(0));
unsigned AddrSpace = DestPtr->getType()->getPointerAddressSpace();
llvm::IntegerType *IntType =
llvm::IntegerType::get(getLLVMContext(),
getContext().getTypeSize(T));
llvm::Type *IntPtrType = IntType->getPointerTo(AddrSpace);
Value *Args[3];
Args[0] = Builder.CreateBitCast(DestPtr, IntPtrType);
Args[1] = EmitScalarExpr(E->getArg(1));
llvm::Type *ValueType = Args[1]->getType();
Args[1] = EmitToInt(*this, Args[1], T, IntType);
Args[2] = EmitToInt(*this, EmitScalarExpr(E->getArg(2)), T, IntType);
Value *Result = Builder.CreateAtomicCmpXchg(Args[0], Args[1], Args[2],
llvm::SequentiallyConsistent,
llvm::SequentiallyConsistent);
Result = EmitFromInt(*this, Result, T, ValueType);
return RValue::get(Result);
}
case Builtin::BI__sync_bool_compare_and_swap_1:
case Builtin::BI__sync_bool_compare_and_swap_2:
case Builtin::BI__sync_bool_compare_and_swap_4:
case Builtin::BI__sync_bool_compare_and_swap_8:
case Builtin::BI__sync_bool_compare_and_swap_16: {
QualType T = E->getArg(1)->getType();
llvm::Value *DestPtr = EmitScalarExpr(E->getArg(0));
unsigned AddrSpace = DestPtr->getType()->getPointerAddressSpace();
llvm::IntegerType *IntType =
llvm::IntegerType::get(getLLVMContext(),
getContext().getTypeSize(T));
llvm::Type *IntPtrType = IntType->getPointerTo(AddrSpace);
Value *Args[3];
Args[0] = Builder.CreateBitCast(DestPtr, IntPtrType);
Args[1] = EmitToInt(*this, EmitScalarExpr(E->getArg(1)), T, IntType);
Args[2] = EmitToInt(*this, EmitScalarExpr(E->getArg(2)), T, IntType);
Value *OldVal = Args[1];
Value *PrevVal = Builder.CreateAtomicCmpXchg(Args[0], Args[1], Args[2],
llvm::SequentiallyConsistent,
llvm::SequentiallyConsistent);
Value *Result = Builder.CreateICmpEQ(PrevVal, OldVal);
// zext bool to int.
Result = Builder.CreateZExt(Result, ConvertType(E->getType()));
return RValue::get(Result);
}
case Builtin::BI__sync_swap_1:
case Builtin::BI__sync_swap_2:
case Builtin::BI__sync_swap_4:
case Builtin::BI__sync_swap_8:
case Builtin::BI__sync_swap_16:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::Xchg, E);
case Builtin::BI__sync_lock_test_and_set_1:
case Builtin::BI__sync_lock_test_and_set_2:
case Builtin::BI__sync_lock_test_and_set_4:
case Builtin::BI__sync_lock_test_and_set_8:
case Builtin::BI__sync_lock_test_and_set_16:
return EmitBinaryAtomic(*this, llvm::AtomicRMWInst::Xchg, E);
case Builtin::BI__sync_lock_release_1:
case Builtin::BI__sync_lock_release_2:
case Builtin::BI__sync_lock_release_4:
case Builtin::BI__sync_lock_release_8:
case Builtin::BI__sync_lock_release_16: {
Value *Ptr = EmitScalarExpr(E->getArg(0));
QualType ElTy = E->getArg(0)->getType()->getPointeeType();
CharUnits StoreSize = getContext().getTypeSizeInChars(ElTy);
llvm::Type *ITy = llvm::IntegerType::get(getLLVMContext(),
StoreSize.getQuantity() * 8);
Ptr = Builder.CreateBitCast(Ptr, ITy->getPointerTo());
llvm::StoreInst *Store =
Builder.CreateStore(llvm::Constant::getNullValue(ITy), Ptr);
Store->setAlignment(StoreSize.getQuantity());
Store->setAtomic(llvm::Release);
return RValue::get(0);
}
case Builtin::BI__sync_synchronize: {
// We assume this is supposed to correspond to a C++0x-style
// sequentially-consistent fence (i.e. this is only usable for
// synchonization, not device I/O or anything like that). This intrinsic
// is really badly designed in the sense that in theory, there isn't
// any way to safely use it... but in practice, it mostly works
// to use it with non-atomic loads and stores to get acquire/release
// semantics.
Builder.CreateFence(llvm::SequentiallyConsistent);
return RValue::get(0);
}
case Builtin::BI__c11_atomic_is_lock_free:
case Builtin::BI__atomic_is_lock_free: {
// Call "bool __atomic_is_lock_free(size_t size, void *ptr)". For the
// __c11 builtin, ptr is 0 (indicating a properly-aligned object), since
// _Atomic(T) is always properly-aligned.
const char *LibCallName = "__atomic_is_lock_free";
CallArgList Args;
Args.add(RValue::get(EmitScalarExpr(E->getArg(0))),
getContext().getSizeType());
if (BuiltinID == Builtin::BI__atomic_is_lock_free)
Args.add(RValue::get(EmitScalarExpr(E->getArg(1))),
getContext().VoidPtrTy);
else
Args.add(RValue::get(llvm::Constant::getNullValue(VoidPtrTy)),
getContext().VoidPtrTy);
const CGFunctionInfo &FuncInfo =
CGM.getTypes().arrangeFreeFunctionCall(E->getType(), Args,
FunctionType::ExtInfo(),
RequiredArgs::All);
llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FuncInfo);
llvm::Constant *Func = CGM.CreateRuntimeFunction(FTy, LibCallName);
return EmitCall(FuncInfo, Func, ReturnValueSlot(), Args);
}
case Builtin::BI__atomic_test_and_set: {
// Look at the argument type to determine whether this is a volatile
// operation. The parameter type is always volatile.
QualType PtrTy = E->getArg(0)->IgnoreImpCasts()->getType();
bool Volatile =
PtrTy->castAs<PointerType>()->getPointeeType().isVolatileQualified();
Value *Ptr = EmitScalarExpr(E->getArg(0));
unsigned AddrSpace = Ptr->getType()->getPointerAddressSpace();
Ptr = Builder.CreateBitCast(Ptr, Int8Ty->getPointerTo(AddrSpace));
Value *NewVal = Builder.getInt8(1);
Value *Order = EmitScalarExpr(E->getArg(1));
if (isa<llvm::ConstantInt>(Order)) {
int ord = cast<llvm::ConstantInt>(Order)->getZExtValue();
AtomicRMWInst *Result = 0;
switch (ord) {
case 0: // memory_order_relaxed
default: // invalid order
Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
Ptr, NewVal,
llvm::Monotonic);
break;
case 1: // memory_order_consume
case 2: // memory_order_acquire
Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
Ptr, NewVal,
llvm::Acquire);
break;
case 3: // memory_order_release
Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
Ptr, NewVal,
llvm::Release);
break;
case 4: // memory_order_acq_rel
Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
Ptr, NewVal,
llvm::AcquireRelease);
break;
case 5: // memory_order_seq_cst
Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
Ptr, NewVal,
llvm::SequentiallyConsistent);
break;
}
Result->setVolatile(Volatile);
return RValue::get(Builder.CreateIsNotNull(Result, "tobool"));
}
llvm::BasicBlock *ContBB = createBasicBlock("atomic.continue", CurFn);
llvm::BasicBlock *BBs[5] = {
createBasicBlock("monotonic", CurFn),
createBasicBlock("acquire", CurFn),
createBasicBlock("release", CurFn),
createBasicBlock("acqrel", CurFn),
createBasicBlock("seqcst", CurFn)
};
llvm::AtomicOrdering Orders[5] = {
llvm::Monotonic, llvm::Acquire, llvm::Release,
llvm::AcquireRelease, llvm::SequentiallyConsistent
};
Order = Builder.CreateIntCast(Order, Builder.getInt32Ty(), false);
llvm::SwitchInst *SI = Builder.CreateSwitch(Order, BBs[0]);
Builder.SetInsertPoint(ContBB);
PHINode *Result = Builder.CreatePHI(Int8Ty, 5, "was_set");
for (unsigned i = 0; i < 5; ++i) {
Builder.SetInsertPoint(BBs[i]);
AtomicRMWInst *RMW = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
Ptr, NewVal, Orders[i]);
RMW->setVolatile(Volatile);
Result->addIncoming(RMW, BBs[i]);
Builder.CreateBr(ContBB);
}
SI->addCase(Builder.getInt32(0), BBs[0]);
SI->addCase(Builder.getInt32(1), BBs[1]);
SI->addCase(Builder.getInt32(2), BBs[1]);
SI->addCase(Builder.getInt32(3), BBs[2]);
SI->addCase(Builder.getInt32(4), BBs[3]);
SI->addCase(Builder.getInt32(5), BBs[4]);
Builder.SetInsertPoint(ContBB);
return RValue::get(Builder.CreateIsNotNull(Result, "tobool"));
}
case Builtin::BI__atomic_clear: {
QualType PtrTy = E->getArg(0)->IgnoreImpCasts()->getType();
bool Volatile =
PtrTy->castAs<PointerType>()->getPointeeType().isVolatileQualified();
Value *Ptr = EmitScalarExpr(E->getArg(0));
unsigned AddrSpace = Ptr->getType()->getPointerAddressSpace();
Ptr = Builder.CreateBitCast(Ptr, Int8Ty->getPointerTo(AddrSpace));
Value *NewVal = Builder.getInt8(0);
Value *Order = EmitScalarExpr(E->getArg(1));
if (isa<llvm::ConstantInt>(Order)) {
int ord = cast<llvm::ConstantInt>(Order)->getZExtValue();
StoreInst *Store = Builder.CreateStore(NewVal, Ptr, Volatile);
Store->setAlignment(1);
switch (ord) {
case 0: // memory_order_relaxed
default: // invalid order
Store->setOrdering(llvm::Monotonic);
break;
case 3: // memory_order_release
Store->setOrdering(llvm::Release);
break;
case 5: // memory_order_seq_cst
Store->setOrdering(llvm::SequentiallyConsistent);
break;
}
return RValue::get(0);
}
llvm::BasicBlock *ContBB = createBasicBlock("atomic.continue", CurFn);
llvm::BasicBlock *BBs[3] = {
createBasicBlock("monotonic", CurFn),
createBasicBlock("release", CurFn),
createBasicBlock("seqcst", CurFn)
};
llvm::AtomicOrdering Orders[3] = {
llvm::Monotonic, llvm::Release, llvm::SequentiallyConsistent
};
Order = Builder.CreateIntCast(Order, Builder.getInt32Ty(), false);
llvm::SwitchInst *SI = Builder.CreateSwitch(Order, BBs[0]);
for (unsigned i = 0; i < 3; ++i) {
Builder.SetInsertPoint(BBs[i]);
StoreInst *Store = Builder.CreateStore(NewVal, Ptr, Volatile);
Store->setAlignment(1);
Store->setOrdering(Orders[i]);
Builder.CreateBr(ContBB);
}
SI->addCase(Builder.getInt32(0), BBs[0]);
SI->addCase(Builder.getInt32(3), BBs[1]);
SI->addCase(Builder.getInt32(5), BBs[2]);
Builder.SetInsertPoint(ContBB);
return RValue::get(0);
}
case Builtin::BI__atomic_thread_fence:
case Builtin::BI__atomic_signal_fence:
case Builtin::BI__c11_atomic_thread_fence:
case Builtin::BI__c11_atomic_signal_fence: {
llvm::SynchronizationScope Scope;
if (BuiltinID == Builtin::BI__atomic_signal_fence ||
BuiltinID == Builtin::BI__c11_atomic_signal_fence)
Scope = llvm::SingleThread;
else
Scope = llvm::CrossThread;
Value *Order = EmitScalarExpr(E->getArg(0));
if (isa<llvm::ConstantInt>(Order)) {
int ord = cast<llvm::ConstantInt>(Order)->getZExtValue();
switch (ord) {
case 0: // memory_order_relaxed
default: // invalid order
break;
case 1: // memory_order_consume
case 2: // memory_order_acquire
Builder.CreateFence(llvm::Acquire, Scope);
break;
case 3: // memory_order_release
Builder.CreateFence(llvm::Release, Scope);
break;
case 4: // memory_order_acq_rel
Builder.CreateFence(llvm::AcquireRelease, Scope);
break;
case 5: // memory_order_seq_cst
Builder.CreateFence(llvm::SequentiallyConsistent, Scope);
break;
}
return RValue::get(0);
}
llvm::BasicBlock *AcquireBB, *ReleaseBB, *AcqRelBB, *SeqCstBB;
AcquireBB = createBasicBlock("acquire", CurFn);
ReleaseBB = createBasicBlock("release", CurFn);
AcqRelBB = createBasicBlock("acqrel", CurFn);
SeqCstBB = createBasicBlock("seqcst", CurFn);
llvm::BasicBlock *ContBB = createBasicBlock("atomic.continue", CurFn);
Order = Builder.CreateIntCast(Order, Builder.getInt32Ty(), false);
llvm::SwitchInst *SI = Builder.CreateSwitch(Order, ContBB);
Builder.SetInsertPoint(AcquireBB);
Builder.CreateFence(llvm::Acquire, Scope);
Builder.CreateBr(ContBB);
SI->addCase(Builder.getInt32(1), AcquireBB);
SI->addCase(Builder.getInt32(2), AcquireBB);
Builder.SetInsertPoint(ReleaseBB);
Builder.CreateFence(llvm::Release, Scope);
Builder.CreateBr(ContBB);
SI->addCase(Builder.getInt32(3), ReleaseBB);
Builder.SetInsertPoint(AcqRelBB);
Builder.CreateFence(llvm::AcquireRelease, Scope);
Builder.CreateBr(ContBB);
SI->addCase(Builder.getInt32(4), AcqRelBB);
Builder.SetInsertPoint(SeqCstBB);
Builder.CreateFence(llvm::SequentiallyConsistent, Scope);
Builder.CreateBr(ContBB);
SI->addCase(Builder.getInt32(5), SeqCstBB);
Builder.SetInsertPoint(ContBB);
return RValue::get(0);
}
// Library functions with special handling.
case Builtin::BIsqrt:
case Builtin::BIsqrtf:
case Builtin::BIsqrtl: {
// Transform a call to sqrt* into a @llvm.sqrt.* intrinsic call, but only
// in finite- or unsafe-math mode (the intrinsic has different semantics
// for handling negative numbers compared to the library function, so
// -fmath-errno=0 is not enough).
if (!FD->hasAttr<ConstAttr>())
break;
if (!(CGM.getCodeGenOpts().UnsafeFPMath ||
CGM.getCodeGenOpts().NoNaNsFPMath))
break;
Value *Arg0 = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgType = Arg0->getType();
Value *F = CGM.getIntrinsic(Intrinsic::sqrt, ArgType);
return RValue::get(Builder.CreateCall(F, Arg0));
}
case Builtin::BIpow:
case Builtin::BIpowf:
case Builtin::BIpowl: {
// Transform a call to pow* into a @llvm.pow.* intrinsic call.
if (!FD->hasAttr<ConstAttr>())
break;
Value *Base = EmitScalarExpr(E->getArg(0));
Value *Exponent = EmitScalarExpr(E->getArg(1));
llvm::Type *ArgType = Base->getType();
Value *F = CGM.getIntrinsic(Intrinsic::pow, ArgType);
return RValue::get(Builder.CreateCall2(F, Base, Exponent));
}
case Builtin::BIfma:
case Builtin::BIfmaf:
case Builtin::BIfmal:
case Builtin::BI__builtin_fma:
case Builtin::BI__builtin_fmaf:
case Builtin::BI__builtin_fmal: {
// Rewrite fma to intrinsic.
Value *FirstArg = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgType = FirstArg->getType();
Value *F = CGM.getIntrinsic(Intrinsic::fma, ArgType);
return RValue::get(Builder.CreateCall3(F, FirstArg,
EmitScalarExpr(E->getArg(1)),
EmitScalarExpr(E->getArg(2))));
}
case Builtin::BI__builtin_signbit:
case Builtin::BI__builtin_signbitf:
case Builtin::BI__builtin_signbitl: {
LLVMContext &C = CGM.getLLVMContext();
Value *Arg = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgTy = Arg->getType();
if (ArgTy->isPPC_FP128Ty())
break; // FIXME: I'm not sure what the right implementation is here.
int ArgWidth = ArgTy->getPrimitiveSizeInBits();
llvm::Type *ArgIntTy = llvm::IntegerType::get(C, ArgWidth);
Value *BCArg = Builder.CreateBitCast(Arg, ArgIntTy);
Value *ZeroCmp = llvm::Constant::getNullValue(ArgIntTy);
Value *Result = Builder.CreateICmpSLT(BCArg, ZeroCmp);
return RValue::get(Builder.CreateZExt(Result, ConvertType(E->getType())));
}
case Builtin::BI__builtin_annotation: {
llvm::Value *AnnVal = EmitScalarExpr(E->getArg(0));
llvm::Value *F = CGM.getIntrinsic(llvm::Intrinsic::annotation,
AnnVal->getType());
// Get the annotation string, go through casts. Sema requires this to be a
// non-wide string literal, potentially casted, so the cast<> is safe.
const Expr *AnnotationStrExpr = E->getArg(1)->IgnoreParenCasts();
StringRef Str = cast<StringLiteral>(AnnotationStrExpr)->getString();
return RValue::get(EmitAnnotationCall(F, AnnVal, Str, E->getExprLoc()));
}
case Builtin::BI__builtin_addcb:
case Builtin::BI__builtin_addcs:
case Builtin::BI__builtin_addc:
case Builtin::BI__builtin_addcl:
case Builtin::BI__builtin_addcll:
case Builtin::BI__builtin_subcb:
case Builtin::BI__builtin_subcs:
case Builtin::BI__builtin_subc:
case Builtin::BI__builtin_subcl:
case Builtin::BI__builtin_subcll: {
// We translate all of these builtins from expressions of the form:
// int x = ..., y = ..., carryin = ..., carryout, result;
// result = __builtin_addc(x, y, carryin, &carryout);
//
// to LLVM IR of the form:
//
// %tmp1 = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %x, i32 %y)
// %tmpsum1 = extractvalue {i32, i1} %tmp1, 0
// %carry1 = extractvalue {i32, i1} %tmp1, 1
// %tmp2 = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %tmpsum1,
// i32 %carryin)
// %result = extractvalue {i32, i1} %tmp2, 0
// %carry2 = extractvalue {i32, i1} %tmp2, 1
// %tmp3 = or i1 %carry1, %carry2
// %tmp4 = zext i1 %tmp3 to i32
// store i32 %tmp4, i32* %carryout
// Scalarize our inputs.
llvm::Value *X = EmitScalarExpr(E->getArg(0));
llvm::Value *Y = EmitScalarExpr(E->getArg(1));
llvm::Value *Carryin = EmitScalarExpr(E->getArg(2));
std::pair<llvm::Value*, unsigned> CarryOutPtr =
EmitPointerWithAlignment(E->getArg(3));
// Decide if we are lowering to a uadd.with.overflow or usub.with.overflow.
llvm::Intrinsic::ID IntrinsicId;
switch (BuiltinID) {
default: llvm_unreachable("Unknown multiprecision builtin id.");
case Builtin::BI__builtin_addcb:
case Builtin::BI__builtin_addcs:
case Builtin::BI__builtin_addc:
case Builtin::BI__builtin_addcl:
case Builtin::BI__builtin_addcll:
IntrinsicId = llvm::Intrinsic::uadd_with_overflow;
break;
case Builtin::BI__builtin_subcb:
case Builtin::BI__builtin_subcs:
case Builtin::BI__builtin_subc:
case Builtin::BI__builtin_subcl:
case Builtin::BI__builtin_subcll:
IntrinsicId = llvm::Intrinsic::usub_with_overflow;
break;
}
// Construct our resulting LLVM IR expression.
llvm::Value *Carry1;
llvm::Value *Sum1 = EmitOverflowIntrinsic(*this, IntrinsicId,
X, Y, Carry1);
llvm::Value *Carry2;
llvm::Value *Sum2 = EmitOverflowIntrinsic(*this, IntrinsicId,
Sum1, Carryin, Carry2);
llvm::Value *CarryOut = Builder.CreateZExt(Builder.CreateOr(Carry1, Carry2),
X->getType());
llvm::StoreInst *CarryOutStore = Builder.CreateStore(CarryOut,
CarryOutPtr.first);
CarryOutStore->setAlignment(CarryOutPtr.second);
return RValue::get(Sum2);
}
case Builtin::BI__builtin_uadd_overflow:
case Builtin::BI__builtin_uaddl_overflow:
case Builtin::BI__builtin_uaddll_overflow:
case Builtin::BI__builtin_usub_overflow:
case Builtin::BI__builtin_usubl_overflow:
case Builtin::BI__builtin_usubll_overflow:
case Builtin::BI__builtin_umul_overflow:
case Builtin::BI__builtin_umull_overflow:
case Builtin::BI__builtin_umulll_overflow:
case Builtin::BI__builtin_sadd_overflow:
case Builtin::BI__builtin_saddl_overflow:
case Builtin::BI__builtin_saddll_overflow:
case Builtin::BI__builtin_ssub_overflow:
case Builtin::BI__builtin_ssubl_overflow:
case Builtin::BI__builtin_ssubll_overflow:
case Builtin::BI__builtin_smul_overflow:
case Builtin::BI__builtin_smull_overflow:
case Builtin::BI__builtin_smulll_overflow: {
// We translate all of these builtins directly to the relevant llvm IR node.
// Scalarize our inputs.
llvm::Value *X = EmitScalarExpr(E->getArg(0));
llvm::Value *Y = EmitScalarExpr(E->getArg(1));
std::pair<llvm::Value *, unsigned> SumOutPtr =
EmitPointerWithAlignment(E->getArg(2));
// Decide which of the overflow intrinsics we are lowering to:
llvm::Intrinsic::ID IntrinsicId;
switch (BuiltinID) {
default: llvm_unreachable("Unknown security overflow builtin id.");
case Builtin::BI__builtin_uadd_overflow:
case Builtin::BI__builtin_uaddl_overflow:
case Builtin::BI__builtin_uaddll_overflow:
IntrinsicId = llvm::Intrinsic::uadd_with_overflow;
break;
case Builtin::BI__builtin_usub_overflow:
case Builtin::BI__builtin_usubl_overflow:
case Builtin::BI__builtin_usubll_overflow:
IntrinsicId = llvm::Intrinsic::usub_with_overflow;
break;
case Builtin::BI__builtin_umul_overflow:
case Builtin::BI__builtin_umull_overflow:
case Builtin::BI__builtin_umulll_overflow:
IntrinsicId = llvm::Intrinsic::umul_with_overflow;
break;
case Builtin::BI__builtin_sadd_overflow:
case Builtin::BI__builtin_saddl_overflow:
case Builtin::BI__builtin_saddll_overflow:
IntrinsicId = llvm::Intrinsic::sadd_with_overflow;
break;
case Builtin::BI__builtin_ssub_overflow:
case Builtin::BI__builtin_ssubl_overflow:
case Builtin::BI__builtin_ssubll_overflow:
IntrinsicId = llvm::Intrinsic::ssub_with_overflow;
break;
case Builtin::BI__builtin_smul_overflow:
case Builtin::BI__builtin_smull_overflow:
case Builtin::BI__builtin_smulll_overflow:
IntrinsicId = llvm::Intrinsic::smul_with_overflow;
break;
}
llvm::Value *Carry;
llvm::Value *Sum = EmitOverflowIntrinsic(*this, IntrinsicId, X, Y, Carry);
llvm::StoreInst *SumOutStore = Builder.CreateStore(Sum, SumOutPtr.first);
SumOutStore->setAlignment(SumOutPtr.second);
return RValue::get(Carry);
}
case Builtin::BI__builtin_addressof:
return RValue::get(EmitLValue(E->getArg(0)).getAddress());
case Builtin::BI__noop:
return RValue::get(0);
case Builtin::BI_InterlockedCompareExchange: {
AtomicCmpXchgInst *CXI = Builder.CreateAtomicCmpXchg(
EmitScalarExpr(E->getArg(0)),
EmitScalarExpr(E->getArg(2)),
EmitScalarExpr(E->getArg(1)),
SequentiallyConsistent,
SequentiallyConsistent);
CXI->setVolatile(true);
return RValue::get(CXI);
}
case Builtin::BI_InterlockedIncrement: {
AtomicRMWInst *RMWI = Builder.CreateAtomicRMW(
AtomicRMWInst::Add,
EmitScalarExpr(E->getArg(0)),
ConstantInt::get(Int32Ty, 1),
llvm::SequentiallyConsistent);
RMWI->setVolatile(true);
return RValue::get(Builder.CreateAdd(RMWI, ConstantInt::get(Int32Ty, 1)));
}
case Builtin::BI_InterlockedDecrement: {
AtomicRMWInst *RMWI = Builder.CreateAtomicRMW(
AtomicRMWInst::Sub,
EmitScalarExpr(E->getArg(0)),
ConstantInt::get(Int32Ty, 1),
llvm::SequentiallyConsistent);
RMWI->setVolatile(true);
return RValue::get(Builder.CreateSub(RMWI, ConstantInt::get(Int32Ty, 1)));
}
case Builtin::BI_InterlockedExchangeAdd: {
AtomicRMWInst *RMWI = Builder.CreateAtomicRMW(
AtomicRMWInst::Add,
EmitScalarExpr(E->getArg(0)),
EmitScalarExpr(E->getArg(1)),
llvm::SequentiallyConsistent);
RMWI->setVolatile(true);
return RValue::get(RMWI);
}
}
// If this is an alias for a lib function (e.g. __builtin_sin), emit
// the call using the normal call path, but using the unmangled
// version of the function name.
if (getContext().BuiltinInfo.isLibFunction(BuiltinID))
return emitLibraryCall(*this, FD, E,
CGM.getBuiltinLibFunction(FD, BuiltinID));
// If this is a predefined lib function (e.g. malloc), emit the call
// using exactly the normal call path.
if (getContext().BuiltinInfo.isPredefinedLibFunction(BuiltinID))
return emitLibraryCall(*this, FD, E, EmitScalarExpr(E->getCallee()));
// See if we have a target specific intrinsic.
const char *Name = getContext().BuiltinInfo.GetName(BuiltinID);
Intrinsic::ID IntrinsicID = Intrinsic::not_intrinsic;
if (const char *Prefix =
llvm::Triple::getArchTypePrefix(getTarget().getTriple().getArch()))
IntrinsicID = Intrinsic::getIntrinsicForGCCBuiltin(Prefix, Name);
if (IntrinsicID != Intrinsic::not_intrinsic) {
SmallVector<Value*, 16> Args;
// Find out if any arguments are required to be integer constant
// expressions.
unsigned ICEArguments = 0;
ASTContext::GetBuiltinTypeError Error;
getContext().GetBuiltinType(BuiltinID, Error, &ICEArguments);
assert(Error == ASTContext::GE_None && "Should not codegen an error");
Function *F = CGM.getIntrinsic(IntrinsicID);
llvm::FunctionType *FTy = F->getFunctionType();
for (unsigned i = 0, e = E->getNumArgs(); i != e; ++i) {
Value *ArgValue;
// If this is a normal argument, just emit it as a scalar.
if ((ICEArguments & (1 << i)) == 0) {
ArgValue = EmitScalarExpr(E->getArg(i));
} else {
// If this is required to be a constant, constant fold it so that we
// know that the generated intrinsic gets a ConstantInt.
llvm::APSInt Result;
bool IsConst = E->getArg(i)->isIntegerConstantExpr(Result,getContext());
assert(IsConst && "Constant arg isn't actually constant?");
(void)IsConst;
ArgValue = llvm::ConstantInt::get(getLLVMContext(), Result);
}
// If the intrinsic arg type is different from the builtin arg type
// we need to do a bit cast.
llvm::Type *PTy = FTy->getParamType(i);
if (PTy != ArgValue->getType()) {
assert(PTy->canLosslesslyBitCastTo(FTy->getParamType(i)) &&
"Must be able to losslessly bit cast to param");
ArgValue = Builder.CreateBitCast(ArgValue, PTy);
}
Args.push_back(ArgValue);
}
Value *V = Builder.CreateCall(F, Args);
QualType BuiltinRetType = E->getType();
llvm::Type *RetTy = VoidTy;
if (!BuiltinRetType->isVoidType())
RetTy = ConvertType(BuiltinRetType);
if (RetTy != V->getType()) {
assert(V->getType()->canLosslesslyBitCastTo(RetTy) &&
"Must be able to losslessly bit cast result type");
V = Builder.CreateBitCast(V, RetTy);
}
return RValue::get(V);
}
// See if we have a target specific builtin that needs to be lowered.
if (Value *V = EmitTargetBuiltinExpr(BuiltinID, E))
return RValue::get(V);
ErrorUnsupported(E, "builtin function");
// Unknown builtin, for now just dump it out and return undef.
return GetUndefRValue(E->getType());
}
Value *CodeGenFunction::EmitTargetBuiltinExpr(unsigned BuiltinID,
const CallExpr *E) {
switch (getTarget().getTriple().getArch()) {
case llvm::Triple::aarch64:
case llvm::Triple::aarch64_be:
return EmitAArch64BuiltinExpr(BuiltinID, E);
case llvm::Triple::arm:
case llvm::Triple::armeb:
case llvm::Triple::thumb:
case llvm::Triple::thumbeb:
return EmitARMBuiltinExpr(BuiltinID, E);
case llvm::Triple::arm64:
return EmitARM64BuiltinExpr(BuiltinID, E);
case llvm::Triple::x86:
case llvm::Triple::x86_64:
return EmitX86BuiltinExpr(BuiltinID, E);
case llvm::Triple::ppc:
case llvm::Triple::ppc64:
case llvm::Triple::ppc64le:
return EmitPPCBuiltinExpr(BuiltinID, E);
default:
return 0;
}
}
static llvm::VectorType *GetNeonType(CodeGenFunction *CGF,
NeonTypeFlags TypeFlags,
bool V1Ty=false) {
int IsQuad = TypeFlags.isQuad();
switch (TypeFlags.getEltType()) {
case NeonTypeFlags::Int8:
case NeonTypeFlags::Poly8:
return llvm::VectorType::get(CGF->Int8Ty, V1Ty ? 1 : (8 << IsQuad));
case NeonTypeFlags::Int16:
case NeonTypeFlags::Poly16:
case NeonTypeFlags::Float16:
return llvm::VectorType::get(CGF->Int16Ty, V1Ty ? 1 : (4 << IsQuad));
case NeonTypeFlags::Int32:
return llvm::VectorType::get(CGF->Int32Ty, V1Ty ? 1 : (2 << IsQuad));
case NeonTypeFlags::Int64:
case NeonTypeFlags::Poly64:
return llvm::VectorType::get(CGF->Int64Ty, V1Ty ? 1 : (1 << IsQuad));
case NeonTypeFlags::Poly128:
// FIXME: i128 and f128 doesn't get fully support in Clang and llvm.
// There is a lot of i128 and f128 API missing.
// so we use v16i8 to represent poly128 and get pattern matched.
return llvm::VectorType::get(CGF->Int8Ty, 16);
case NeonTypeFlags::Float32:
return llvm::VectorType::get(CGF->FloatTy, V1Ty ? 1 : (2 << IsQuad));
case NeonTypeFlags::Float64:
return llvm::VectorType::get(CGF->DoubleTy, V1Ty ? 1 : (1 << IsQuad));
}
llvm_unreachable("Unknown vector element type!");
}
Value *CodeGenFunction::EmitNeonSplat(Value *V, Constant *C) {
unsigned nElts = cast<llvm::VectorType>(V->getType())->getNumElements();
Value* SV = llvm::ConstantVector::getSplat(nElts, C);
return Builder.CreateShuffleVector(V, V, SV, "lane");
}
Value *CodeGenFunction::EmitNeonCall(Function *F, SmallVectorImpl<Value*> &Ops,
const char *name,
unsigned shift, bool rightshift) {
unsigned j = 0;
for (Function::const_arg_iterator ai = F->arg_begin(), ae = F->arg_end();
ai != ae; ++ai, ++j)
if (shift > 0 && shift == j)
Ops[j] = EmitNeonShiftVector(Ops[j], ai->getType(), rightshift);
else
Ops[j] = Builder.CreateBitCast(Ops[j], ai->getType(), name);
return Builder.CreateCall(F, Ops, name);
}
Value *CodeGenFunction::EmitNeonShiftVector(Value *V, llvm::Type *Ty,
bool neg) {
int SV = cast<ConstantInt>(V)->getSExtValue();
llvm::VectorType *VTy = cast<llvm::VectorType>(Ty);
llvm::Constant *C = ConstantInt::get(VTy->getElementType(), neg ? -SV : SV);
return llvm::ConstantVector::getSplat(VTy->getNumElements(), C);
}
// \brief Right-shift a vector by a constant.
Value *CodeGenFunction::EmitNeonRShiftImm(Value *Vec, Value *Shift,
llvm::Type *Ty, bool usgn,
const char *name) {
llvm::VectorType *VTy = cast<llvm::VectorType>(Ty);
int ShiftAmt = cast<ConstantInt>(Shift)->getSExtValue();
int EltSize = VTy->getScalarSizeInBits();
Vec = Builder.CreateBitCast(Vec, Ty);
// lshr/ashr are undefined when the shift amount is equal to the vector
// element size.
if (ShiftAmt == EltSize) {
if (usgn) {
// Right-shifting an unsigned value by its size yields 0.
llvm::Constant *Zero = ConstantInt::get(VTy->getElementType(), 0);
return llvm::ConstantVector::getSplat(VTy->getNumElements(), Zero);
} else {
// Right-shifting a signed value by its size is equivalent
// to a shift of size-1.
--ShiftAmt;
Shift = ConstantInt::get(VTy->getElementType(), ShiftAmt);
}
}
Shift = EmitNeonShiftVector(Shift, Ty, false);
if (usgn)
return Builder.CreateLShr(Vec, Shift, name);
else
return Builder.CreateAShr(Vec, Shift, name);
}
Value *CodeGenFunction::EmitConcatVectors(Value *Lo, Value *Hi,
llvm::Type *ArgTy) {
unsigned NumElts = ArgTy->getVectorNumElements();
SmallVector<Constant *, 16> Indices;
for (unsigned i = 0; i < 2 * NumElts; ++i)
Indices.push_back(ConstantInt::get(Int32Ty, i));
Constant *Mask = ConstantVector::get(Indices);
Value *LoCast = Builder.CreateBitCast(Lo, ArgTy);
Value *HiCast = Builder.CreateBitCast(Hi, ArgTy);
return Builder.CreateShuffleVector(LoCast, HiCast, Mask, "concat");
}
Value *CodeGenFunction::EmitExtractHigh(Value *Vec, llvm::Type *ResTy) {
unsigned NumElts = ResTy->getVectorNumElements();
SmallVector<Constant *, 8> Indices;
llvm::Type *InTy = llvm::VectorType::get(ResTy->getVectorElementType(),
NumElts * 2);
Value *VecCast = Builder.CreateBitCast(Vec, InTy);
// extract_high is a shuffle on the second half of the input indices: E.g. 4,
// 5, 6, 7 if we're extracting <4 x i16> from <8 x i16>.
for (unsigned i = 0; i < NumElts; ++i)
Indices.push_back(ConstantInt::get(Int32Ty, NumElts + i));
Constant *Mask = ConstantVector::get(Indices);
return Builder.CreateShuffleVector(VecCast, VecCast, Mask, "concat");
}
/// GetPointeeAlignment - Given an expression with a pointer type, find the
/// alignment of the type referenced by the pointer. Skip over implicit
/// casts.
std::pair<llvm::Value*, unsigned>
CodeGenFunction::EmitPointerWithAlignment(const Expr *Addr) {
assert(Addr->getType()->isPointerType());
Addr = Addr->IgnoreParens();
if (const ImplicitCastExpr *ICE = dyn_cast<ImplicitCastExpr>(Addr)) {
if ((ICE->getCastKind() == CK_BitCast || ICE->getCastKind() == CK_NoOp) &&
ICE->getSubExpr()->getType()->isPointerType()) {
std::pair<llvm::Value*, unsigned> Ptr =
EmitPointerWithAlignment(ICE->getSubExpr());
Ptr.first = Builder.CreateBitCast(Ptr.first,
ConvertType(Addr->getType()));
return Ptr;
} else if (ICE->getCastKind() == CK_ArrayToPointerDecay) {
LValue LV = EmitLValue(ICE->getSubExpr());
unsigned Align = LV.getAlignment().getQuantity();
if (!Align) {
// FIXME: Once LValues are fixed to always set alignment,
// zap this code.
QualType PtTy = ICE->getSubExpr()->getType();
if (!PtTy->isIncompleteType())
Align = getContext().getTypeAlignInChars(PtTy).getQuantity();
else
Align = 1;
}
return std::make_pair(LV.getAddress(), Align);
}
}
if (const UnaryOperator *UO = dyn_cast<UnaryOperator>(Addr)) {
if (UO->getOpcode() == UO_AddrOf) {
LValue LV = EmitLValue(UO->getSubExpr());
unsigned Align = LV.getAlignment().getQuantity();
if (!Align) {
// FIXME: Once LValues are fixed to always set alignment,
// zap this code.
QualType PtTy = UO->getSubExpr()->getType();
if (!PtTy->isIncompleteType())
Align = getContext().getTypeAlignInChars(PtTy).getQuantity();
else
Align = 1;
}
return std::make_pair(LV.getAddress(), Align);
}
}
unsigned Align = 1;
QualType PtTy = Addr->getType()->getPointeeType();
if (!PtTy->isIncompleteType())
Align = getContext().getTypeAlignInChars(PtTy).getQuantity();
return std::make_pair(EmitScalarExpr(Addr), Align);
}
enum {
AddRetType = (1 << 0),
Add1ArgType = (1 << 1),
Add2ArgTypes = (1 << 2),
VectorizeRetType = (1 << 3),
VectorizeArgTypes = (1 << 4),
InventFloatType = (1 << 5),
UnsignedAlts = (1 << 6),
Use64BitVectors = (1 << 7),
Use128BitVectors = (1 << 8),
Vectorize1ArgType = Add1ArgType | VectorizeArgTypes,
VectorRet = AddRetType | VectorizeRetType,
VectorRetGetArgs01 =
AddRetType | Add2ArgTypes | VectorizeRetType | VectorizeArgTypes,
FpCmpzModifiers =
AddRetType | VectorizeRetType | Add1ArgType | InventFloatType
};
struct NeonIntrinsicInfo {
unsigned BuiltinID;
unsigned LLVMIntrinsic;
unsigned AltLLVMIntrinsic;
const char *NameHint;
unsigned TypeModifier;
bool operator<(unsigned RHSBuiltinID) const {
return BuiltinID < RHSBuiltinID;
}
};
#define NEONMAP0(NameBase) \
{ NEON::BI__builtin_neon_ ## NameBase, 0, 0, #NameBase, 0 }
#define NEONMAP1(NameBase, LLVMIntrinsic, TypeModifier) \
{ NEON:: BI__builtin_neon_ ## NameBase, \
Intrinsic::LLVMIntrinsic, 0, #NameBase, TypeModifier }
#define NEONMAP2(NameBase, LLVMIntrinsic, AltLLVMIntrinsic, TypeModifier) \
{ NEON:: BI__builtin_neon_ ## NameBase, \
Intrinsic::LLVMIntrinsic, Intrinsic::AltLLVMIntrinsic, \
#NameBase, TypeModifier }
static const NeonIntrinsicInfo AArch64SISDIntrinsicInfo[] = {
NEONMAP1(vabdd_f64, aarch64_neon_vabd, AddRetType),
NEONMAP1(vabds_f32, aarch64_neon_vabd, AddRetType),
NEONMAP1(vabsd_s64, aarch64_neon_vabs, 0),
NEONMAP1(vaddd_s64, aarch64_neon_vaddds, 0),
NEONMAP1(vaddd_u64, aarch64_neon_vadddu, 0),
NEONMAP1(vaddlv_s16, aarch64_neon_saddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlv_s32, aarch64_neon_saddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlv_s8, aarch64_neon_saddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlv_u16, aarch64_neon_uaddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlv_u32, aarch64_neon_uaddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlv_u8, aarch64_neon_uaddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlvq_s16, aarch64_neon_saddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlvq_s32, aarch64_neon_saddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlvq_s8, aarch64_neon_saddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlvq_u16, aarch64_neon_uaddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlvq_u32, aarch64_neon_uaddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddlvq_u8, aarch64_neon_uaddlv, VectorRet | Add1ArgType),
NEONMAP1(vaddv_f32, aarch64_neon_vpfadd, AddRetType | Add1ArgType),
NEONMAP1(vaddv_s16, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddv_s32, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddv_s8, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddv_u16, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddv_u32, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddv_u8, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddvq_f32, aarch64_neon_vpfadd, AddRetType | Add1ArgType),
NEONMAP1(vaddvq_f64, aarch64_neon_vpfadd, AddRetType | Add1ArgType),
NEONMAP1(vaddvq_s16, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddvq_s32, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddvq_s64, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddvq_s8, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddvq_u16, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddvq_u32, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddvq_u64, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vaddvq_u8, aarch64_neon_vaddv, VectorRet | Add1ArgType),
NEONMAP1(vcaged_f64, aarch64_neon_fcage, VectorRet | Add2ArgTypes),
NEONMAP1(vcages_f32, aarch64_neon_fcage, VectorRet | Add2ArgTypes),
NEONMAP1(vcagtd_f64, aarch64_neon_fcagt, VectorRet | Add2ArgTypes),
NEONMAP1(vcagts_f32, aarch64_neon_fcagt, VectorRet | Add2ArgTypes),
NEONMAP1(vcaled_f64, aarch64_neon_fcage, VectorRet | Add2ArgTypes),
NEONMAP1(vcales_f32, aarch64_neon_fcage, VectorRet | Add2ArgTypes),
NEONMAP1(vcaltd_f64, aarch64_neon_fcagt, VectorRet | Add2ArgTypes),
NEONMAP1(vcalts_f32, aarch64_neon_fcagt, VectorRet | Add2ArgTypes),
NEONMAP1(vceqd_f64, aarch64_neon_fceq, VectorRet | Add2ArgTypes),
NEONMAP1(vceqd_s64, aarch64_neon_vceq, VectorRetGetArgs01),
NEONMAP1(vceqd_u64, aarch64_neon_vceq, VectorRetGetArgs01),
NEONMAP1(vceqs_f32, aarch64_neon_fceq, VectorRet | Add2ArgTypes),
NEONMAP1(vceqzd_f64, aarch64_neon_fceq, FpCmpzModifiers),
NEONMAP1(vceqzd_s64, aarch64_neon_vceq, VectorRetGetArgs01),
NEONMAP1(vceqzd_u64, aarch64_neon_vceq, VectorRetGetArgs01),
NEONMAP1(vceqzs_f32, aarch64_neon_fceq, FpCmpzModifiers),
NEONMAP1(vcged_f64, aarch64_neon_fcge, VectorRet | Add2ArgTypes),
NEONMAP1(vcged_s64, aarch64_neon_vcge, VectorRetGetArgs01),
NEONMAP1(vcged_u64, aarch64_neon_vchs, VectorRetGetArgs01),
NEONMAP1(vcges_f32, aarch64_neon_fcge, VectorRet | Add2ArgTypes),
NEONMAP1(vcgezd_f64, aarch64_neon_fcge, FpCmpzModifiers),
NEONMAP1(vcgezd_s64, aarch64_neon_vcge, VectorRetGetArgs01),
NEONMAP1(vcgezs_f32, aarch64_neon_fcge, FpCmpzModifiers),
NEONMAP1(vcgtd_f64, aarch64_neon_fcgt, VectorRet | Add2ArgTypes),
NEONMAP1(vcgtd_s64, aarch64_neon_vcgt, VectorRetGetArgs01),
NEONMAP1(vcgtd_u64, aarch64_neon_vchi, VectorRetGetArgs01),
NEONMAP1(vcgts_f32, aarch64_neon_fcgt, VectorRet | Add2ArgTypes),
NEONMAP1(vcgtzd_f64, aarch64_neon_fcgt, FpCmpzModifiers),
NEONMAP1(vcgtzd_s64, aarch64_neon_vcgt, VectorRetGetArgs01),
NEONMAP1(vcgtzs_f32, aarch64_neon_fcgt, FpCmpzModifiers),
NEONMAP1(vcled_f64, aarch64_neon_fcge, VectorRet | Add2ArgTypes),
NEONMAP1(vcled_s64, aarch64_neon_vcge, VectorRetGetArgs01),
NEONMAP1(vcled_u64, aarch64_neon_vchs, VectorRetGetArgs01),
NEONMAP1(vcles_f32, aarch64_neon_fcge, VectorRet | Add2ArgTypes),
NEONMAP1(vclezd_f64, aarch64_neon_fclez, FpCmpzModifiers),
NEONMAP1(vclezd_s64, aarch64_neon_vclez, VectorRetGetArgs01),
NEONMAP1(vclezs_f32, aarch64_neon_fclez, FpCmpzModifiers),
NEONMAP1(vcltd_f64, aarch64_neon_fcgt, VectorRet | Add2ArgTypes),
NEONMAP1(vcltd_s64, aarch64_neon_vcgt, VectorRetGetArgs01),
NEONMAP1(vcltd_u64, aarch64_neon_vchi, VectorRetGetArgs01),
NEONMAP1(vclts_f32, aarch64_neon_fcgt, VectorRet | Add2ArgTypes),
NEONMAP1(vcltzd_f64, aarch64_neon_fcltz, FpCmpzModifiers),
NEONMAP1(vcltzd_s64, aarch64_neon_vcltz, VectorRetGetArgs01),
NEONMAP1(vcltzs_f32, aarch64_neon_fcltz, FpCmpzModifiers),
NEONMAP1(vcvtad_s64_f64, aarch64_neon_fcvtas, VectorRet | Add1ArgType),
NEONMAP1(vcvtad_u64_f64, aarch64_neon_fcvtau, VectorRet | Add1ArgType),
NEONMAP1(vcvtas_s32_f32, aarch64_neon_fcvtas, VectorRet | Add1ArgType),
NEONMAP1(vcvtas_u32_f32, aarch64_neon_fcvtau, VectorRet | Add1ArgType),
NEONMAP1(vcvtd_f64_s64, aarch64_neon_vcvtint2fps, AddRetType | Vectorize1ArgType),
NEONMAP1(vcvtd_f64_u64, aarch64_neon_vcvtint2fpu, AddRetType | Vectorize1ArgType),
NEONMAP1(vcvtd_n_f64_s64, aarch64_neon_vcvtfxs2fp_n, AddRetType | Vectorize1ArgType),
NEONMAP1(vcvtd_n_f64_u64, aarch64_neon_vcvtfxu2fp_n, AddRetType | Vectorize1ArgType),
NEONMAP1(vcvtd_n_s64_f64, aarch64_neon_vcvtfp2fxs_n, VectorRet | Add1ArgType),
NEONMAP1(vcvtd_n_u64_f64, aarch64_neon_vcvtfp2fxu_n, VectorRet | Add1ArgType),
NEONMAP1(vcvtd_s64_f64, aarch64_neon_fcvtzs, VectorRet | Add1ArgType),
NEONMAP1(vcvtd_u64_f64, aarch64_neon_fcvtzu, VectorRet | Add1ArgType),
NEONMAP1(vcvtmd_s64_f64, aarch64_neon_fcvtms, VectorRet | Add1ArgType),
NEONMAP1(vcvtmd_u64_f64, aarch64_neon_fcvtmu, VectorRet | Add1ArgType),
NEONMAP1(vcvtms_s32_f32, aarch64_neon_fcvtms, VectorRet | Add1ArgType),
NEONMAP1(vcvtms_u32_f32, aarch64_neon_fcvtmu, VectorRet | Add1ArgType),
NEONMAP1(vcvtnd_s64_f64, aarch64_neon_fcvtns, VectorRet | Add1ArgType),
NEONMAP1(vcvtnd_u64_f64, aarch64_neon_fcvtnu, VectorRet | Add1ArgType),
NEONMAP1(vcvtns_s32_f32, aarch64_neon_fcvtns, VectorRet | Add1ArgType),
NEONMAP1(vcvtns_u32_f32, aarch64_neon_fcvtnu, VectorRet | Add1ArgType),
NEONMAP1(vcvtpd_s64_f64, aarch64_neon_fcvtps, VectorRet | Add1ArgType),
NEONMAP1(vcvtpd_u64_f64, aarch64_neon_fcvtpu, VectorRet | Add1ArgType),
NEONMAP1(vcvtps_s32_f32, aarch64_neon_fcvtps, VectorRet | Add1ArgType),
NEONMAP1(vcvtps_u32_f32, aarch64_neon_fcvtpu, VectorRet | Add1ArgType),
NEONMAP1(vcvts_f32_s32, aarch64_neon_vcvtint2fps, AddRetType | Vectorize1ArgType),
NEONMAP1(vcvts_f32_u32, aarch64_neon_vcvtint2fpu, AddRetType | Vectorize1ArgType),
NEONMAP1(vcvts_n_f32_s32, aarch64_neon_vcvtfxs2fp_n, AddRetType | Vectorize1ArgType),
NEONMAP1(vcvts_n_f32_u32, aarch64_neon_vcvtfxu2fp_n, AddRetType | Vectorize1ArgType),
NEONMAP1(vcvts_n_s32_f32, aarch64_neon_vcvtfp2fxs_n, VectorRet | Add1ArgType),
NEONMAP1(vcvts_n_u32_f32, aarch64_neon_vcvtfp2fxu_n, VectorRet | Add1ArgType),
NEONMAP1(vcvts_s32_f32, aarch64_neon_fcvtzs, VectorRet | Add1ArgType),
NEONMAP1(vcvts_u32_f32, aarch64_neon_fcvtzu, VectorRet | Add1ArgType),
NEONMAP1(vcvtxd_f32_f64, aarch64_neon_fcvtxn, 0),
NEONMAP0(vdupb_lane_i8),
NEONMAP0(vdupb_laneq_i8),
NEONMAP0(vdupd_lane_f64),
NEONMAP0(vdupd_lane_i64),
NEONMAP0(vdupd_laneq_f64),
NEONMAP0(vdupd_laneq_i64),
NEONMAP0(vduph_lane_i16),
NEONMAP0(vduph_laneq_i16),
NEONMAP0(vdups_lane_f32),
NEONMAP0(vdups_lane_i32),
NEONMAP0(vdups_laneq_f32),
NEONMAP0(vdups_laneq_i32),
NEONMAP0(vfmad_lane_f64),
NEONMAP0(vfmad_laneq_f64),
NEONMAP0(vfmas_lane_f32),
NEONMAP0(vfmas_laneq_f32),
NEONMAP0(vget_lane_f32),
NEONMAP0(vget_lane_f64),
NEONMAP0(vget_lane_i16),
NEONMAP0(vget_lane_i32),
NEONMAP0(vget_lane_i64),
NEONMAP0(vget_lane_i8),
NEONMAP0(vgetq_lane_f32),
NEONMAP0(vgetq_lane_f64),
NEONMAP0(vgetq_lane_i16),
NEONMAP0(vgetq_lane_i32),
NEONMAP0(vgetq_lane_i64),
NEONMAP0(vgetq_lane_i8),
NEONMAP1(vmaxnmv_f32, aarch64_neon_vpfmaxnm, AddRetType | Add1ArgType),
NEONMAP1(vmaxnmvq_f32, aarch64_neon_vmaxnmv, 0),
NEONMAP1(vmaxnmvq_f64, aarch64_neon_vpfmaxnm, AddRetType | Add1ArgType),
NEONMAP1(vmaxv_f32, aarch64_neon_vpmax, AddRetType | Add1ArgType),
NEONMAP1(vmaxv_s16, aarch64_neon_smaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxv_s32, aarch64_neon_smaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxv_s8, aarch64_neon_smaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxv_u16, aarch64_neon_umaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxv_u32, aarch64_neon_umaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxv_u8, aarch64_neon_umaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxvq_f32, aarch64_neon_vmaxv, 0),
NEONMAP1(vmaxvq_f64, aarch64_neon_vpmax, AddRetType | Add1ArgType),
NEONMAP1(vmaxvq_s16, aarch64_neon_smaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxvq_s32, aarch64_neon_smaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxvq_s8, aarch64_neon_smaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxvq_u16, aarch64_neon_umaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxvq_u32, aarch64_neon_umaxv, VectorRet | Add1ArgType),
NEONMAP1(vmaxvq_u8, aarch64_neon_umaxv, VectorRet | Add1ArgType),
NEONMAP1(vminnmv_f32, aarch64_neon_vpfminnm, AddRetType | Add1ArgType),
NEONMAP1(vminnmvq_f32, aarch64_neon_vminnmv, 0),
NEONMAP1(vminnmvq_f64, aarch64_neon_vpfminnm, AddRetType | Add1ArgType),
NEONMAP1(vminv_f32, aarch64_neon_vpmin, AddRetType | Add1ArgType),
NEONMAP1(vminv_s16, aarch64_neon_sminv, VectorRet | Add1ArgType),
NEONMAP1(vminv_s32, aarch64_neon_sminv, VectorRet | Add1ArgType),
NEONMAP1(vminv_s8, aarch64_neon_sminv, VectorRet | Add1ArgType),
NEONMAP1(vminv_u16, aarch64_neon_uminv, VectorRet | Add1ArgType),
NEONMAP1(vminv_u32, aarch64_neon_uminv, VectorRet | Add1ArgType),
NEONMAP1(vminv_u8, aarch64_neon_uminv, VectorRet | Add1ArgType),
NEONMAP1(vminvq_f32, aarch64_neon_vminv, 0),
NEONMAP1(vminvq_f64, aarch64_neon_vpmin, AddRetType | Add1ArgType),
NEONMAP1(vminvq_s16, aarch64_neon_sminv, VectorRet | Add1ArgType),
NEONMAP1(vminvq_s32, aarch64_neon_sminv, VectorRet | Add1ArgType),
NEONMAP1(vminvq_s8, aarch64_neon_sminv, VectorRet | Add1ArgType),
NEONMAP1(vminvq_u16, aarch64_neon_uminv, VectorRet | Add1ArgType),
NEONMAP1(vminvq_u32, aarch64_neon_uminv, VectorRet | Add1ArgType),
NEONMAP1(vminvq_u8, aarch64_neon_uminv, VectorRet | Add1ArgType),
NEONMAP0(vmul_n_f64),
NEONMAP1(vmull_p64, aarch64_neon_vmull_p64, 0),
NEONMAP0(vmulxd_f64),
NEONMAP0(vmulxs_f32),
NEONMAP1(vnegd_s64, aarch64_neon_vneg, 0),
NEONMAP1(vpaddd_f64, aarch64_neon_vpfadd, AddRetType | Add1ArgType),
NEONMAP1(vpaddd_s64, aarch64_neon_vpadd, 0),
NEONMAP1(vpaddd_u64, aarch64_neon_vpadd, 0),
NEONMAP1(vpadds_f32, aarch64_neon_vpfadd, AddRetType | Add1ArgType),
NEONMAP1(vpmaxnmqd_f64, aarch64_neon_vpfmaxnm, AddRetType | Add1ArgType),
NEONMAP1(vpmaxnms_f32, aarch64_neon_vpfmaxnm, AddRetType | Add1ArgType),
NEONMAP1(vpmaxqd_f64, aarch64_neon_vpmax, AddRetType | Add1ArgType),
NEONMAP1(vpmaxs_f32, aarch64_neon_vpmax, AddRetType | Add1ArgType),
NEONMAP1(vpminnmqd_f64, aarch64_neon_vpfminnm, AddRetType | Add1ArgType),
NEONMAP1(vpminnms_f32, aarch64_neon_vpfminnm, AddRetType | Add1ArgType),
NEONMAP1(vpminqd_f64, aarch64_neon_vpmin, AddRetType | Add1ArgType),
NEONMAP1(vpmins_f32, aarch64_neon_vpmin, AddRetType | Add1ArgType),
NEONMAP1(vqabsb_s8, arm_neon_vqabs, VectorRet),
NEONMAP1(vqabsd_s64, arm_neon_vqabs, VectorRet),
NEONMAP1(vqabsh_s16, arm_neon_vqabs, VectorRet),
NEONMAP1(vqabss_s32, arm_neon_vqabs, VectorRet),
NEONMAP1(vqaddb_s8, arm_neon_vqadds, VectorRet),
NEONMAP1(vqaddb_u8, arm_neon_vqaddu, VectorRet),
NEONMAP1(vqaddd_s64, arm_neon_vqadds, VectorRet),
NEONMAP1(vqaddd_u64, arm_neon_vqaddu, VectorRet),
NEONMAP1(vqaddh_s16, arm_neon_vqadds, VectorRet),
NEONMAP1(vqaddh_u16, arm_neon_vqaddu, VectorRet),
NEONMAP1(vqadds_s32, arm_neon_vqadds, VectorRet),
NEONMAP1(vqadds_u32, arm_neon_vqaddu, VectorRet),
NEONMAP0(vqdmlalh_lane_s16),
NEONMAP0(vqdmlalh_laneq_s16),
NEONMAP1(vqdmlalh_s16, aarch64_neon_vqdmlal, VectorRet),
NEONMAP0(vqdmlals_lane_s32),
NEONMAP0(vqdmlals_laneq_s32),
NEONMAP1(vqdmlals_s32, aarch64_neon_vqdmlal, VectorRet),
NEONMAP0(vqdmlslh_lane_s16),
NEONMAP0(vqdmlslh_laneq_s16),
NEONMAP1(vqdmlslh_s16, aarch64_neon_vqdmlsl, VectorRet),
NEONMAP0(vqdmlsls_lane_s32),
NEONMAP0(vqdmlsls_laneq_s32),
NEONMAP1(vqdmlsls_s32, aarch64_neon_vqdmlsl, VectorRet),
NEONMAP1(vqdmulhh_s16, arm_neon_vqdmulh, VectorRet),
NEONMAP1(vqdmulhs_s32, arm_neon_vqdmulh, VectorRet),
NEONMAP1(vqdmullh_s16, arm_neon_vqdmull, VectorRet),
NEONMAP1(vqdmulls_s32, arm_neon_vqdmull, VectorRet),
NEONMAP1(vqmovnd_s64, arm_neon_vqmovns, VectorRet),
NEONMAP1(vqmovnd_u64, arm_neon_vqmovnu, VectorRet),
NEONMAP1(vqmovnh_s16, arm_neon_vqmovns, VectorRet),
NEONMAP1(vqmovnh_u16, arm_neon_vqmovnu, VectorRet),
NEONMAP1(vqmovns_s32, arm_neon_vqmovns, VectorRet),
NEONMAP1(vqmovns_u32, arm_neon_vqmovnu, VectorRet),
NEONMAP1(vqmovund_s64, arm_neon_vqmovnsu, VectorRet),
NEONMAP1(vqmovunh_s16, arm_neon_vqmovnsu, VectorRet),
NEONMAP1(vqmovuns_s32, arm_neon_vqmovnsu, VectorRet),
NEONMAP1(vqnegb_s8, arm_neon_vqneg, VectorRet),
NEONMAP1(vqnegd_s64, arm_neon_vqneg, VectorRet),
NEONMAP1(vqnegh_s16, arm_neon_vqneg, VectorRet),
NEONMAP1(vqnegs_s32, arm_neon_vqneg, VectorRet),
NEONMAP1(vqrdmulhh_s16, arm_neon_vqrdmulh, VectorRet),
NEONMAP1(vqrdmulhs_s32, arm_neon_vqrdmulh, VectorRet),
NEONMAP1(vqrshlb_s8, aarch64_neon_vqrshls, VectorRet),
NEONMAP1(vqrshlb_u8, aarch64_neon_vqrshlu, VectorRet),
NEONMAP1(vqrshld_s64, aarch64_neon_vqrshls, VectorRet),
NEONMAP1(vqrshld_u64, aarch64_neon_vqrshlu, VectorRet),
NEONMAP1(vqrshlh_s16, aarch64_neon_vqrshls, VectorRet),
NEONMAP1(vqrshlh_u16, aarch64_neon_vqrshlu, VectorRet),
NEONMAP1(vqrshls_s32, aarch64_neon_vqrshls, VectorRet),
NEONMAP1(vqrshls_u32, aarch64_neon_vqrshlu, VectorRet),
NEONMAP1(vqrshrnd_n_s64, aarch64_neon_vsqrshrn, VectorRet),
NEONMAP1(vqrshrnd_n_u64, aarch64_neon_vuqrshrn, VectorRet),
NEONMAP1(vqrshrnh_n_s16, aarch64_neon_vsqrshrn, VectorRet),
NEONMAP1(vqrshrnh_n_u16, aarch64_neon_vuqrshrn, VectorRet),
NEONMAP1(vqrshrns_n_s32, aarch64_neon_vsqrshrn, VectorRet),
NEONMAP1(vqrshrns_n_u32, aarch64_neon_vuqrshrn, VectorRet),
NEONMAP1(vqrshrund_n_s64, aarch64_neon_vsqrshrun, VectorRet),
NEONMAP1(vqrshrunh_n_s16, aarch64_neon_vsqrshrun, VectorRet),
NEONMAP1(vqrshruns_n_s32, aarch64_neon_vsqrshrun, VectorRet),
NEONMAP1(vqshlb_n_s8, aarch64_neon_vqshls_n, VectorRet),
NEONMAP1(vqshlb_n_u8, aarch64_neon_vqshlu_n, VectorRet),
NEONMAP1(vqshlb_s8, aarch64_neon_vqshls, VectorRet),
NEONMAP1(vqshlb_u8, aarch64_neon_vqshlu, VectorRet),
NEONMAP1(vqshld_n_s64, aarch64_neon_vqshls_n, VectorRet),
NEONMAP1(vqshld_n_u64, aarch64_neon_vqshlu_n, VectorRet),
NEONMAP1(vqshld_s64, aarch64_neon_vqshls, VectorRet),
NEONMAP1(vqshld_u64, aarch64_neon_vqshlu, VectorRet),
NEONMAP1(vqshlh_n_s16, aarch64_neon_vqshls_n, VectorRet),
NEONMAP1(vqshlh_n_u16, aarch64_neon_vqshlu_n, VectorRet),
NEONMAP1(vqshlh_s16, aarch64_neon_vqshls, VectorRet),
NEONMAP1(vqshlh_u16, aarch64_neon_vqshlu, VectorRet),
NEONMAP1(vqshls_n_s32, aarch64_neon_vqshls_n, VectorRet),
NEONMAP1(vqshls_n_u32, aarch64_neon_vqshlu_n, VectorRet),
NEONMAP1(vqshls_s32, aarch64_neon_vqshls, VectorRet),
NEONMAP1(vqshls_u32, aarch64_neon_vqshlu, VectorRet),
NEONMAP1(vqshlub_n_s8, aarch64_neon_vsqshlu, VectorRet),
NEONMAP1(vqshlud_n_s64, aarch64_neon_vsqshlu, VectorRet),
NEONMAP1(vqshluh_n_s16, aarch64_neon_vsqshlu, VectorRet),
NEONMAP1(vqshlus_n_s32, aarch64_neon_vsqshlu, VectorRet),
NEONMAP1(vqshrnd_n_s64, aarch64_neon_vsqshrn, VectorRet),
NEONMAP1(vqshrnd_n_u64, aarch64_neon_vuqshrn, VectorRet),
NEONMAP1(vqshrnh_n_s16, aarch64_neon_vsqshrn, VectorRet),