#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/IR/CallSite.h"
#include "llvm/IR/DataLayout.h"
+#include "llvm/IR/Dominators.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/Transforms/Utils/BuildLibCalls.h"
#include "llvm/Transforms/Utils/Local.h"
}
Instruction *InstCombiner::SimplifyMemTransfer(MemIntrinsic *MI) {
- unsigned DstAlign = getKnownAlignment(MI->getArgOperand(0), DL);
- unsigned SrcAlign = getKnownAlignment(MI->getArgOperand(1), DL);
+ unsigned DstAlign = getKnownAlignment(MI->getArgOperand(0), DL, AT, MI, DT);
+ unsigned SrcAlign = getKnownAlignment(MI->getArgOperand(1), DL, AT, MI, DT);
unsigned MinAlign = std::min(DstAlign, SrcAlign);
unsigned CopyAlign = MI->getAlignment();
}
Instruction *InstCombiner::SimplifyMemSet(MemSetInst *MI) {
- unsigned Alignment = getKnownAlignment(MI->getDest(), DL);
+ unsigned Alignment = getKnownAlignment(MI->getDest(), DL, AT, MI, DT);
if (MI->getAlignment() < Alignment) {
MI->setAlignment(ConstantInt::get(MI->getAlignmentType(),
Alignment, false));
uint32_t BitWidth = IT->getBitWidth();
APInt KnownZero(BitWidth, 0);
APInt KnownOne(BitWidth, 0);
- ComputeMaskedBits(II->getArgOperand(0), KnownZero, KnownOne);
+ computeKnownBits(II->getArgOperand(0), KnownZero, KnownOne, 0, II);
unsigned TrailingZeros = KnownOne.countTrailingZeros();
APInt Mask(APInt::getLowBitsSet(BitWidth, TrailingZeros));
if ((Mask & KnownZero) == Mask)
uint32_t BitWidth = IT->getBitWidth();
APInt KnownZero(BitWidth, 0);
APInt KnownOne(BitWidth, 0);
- ComputeMaskedBits(II->getArgOperand(0), KnownZero, KnownOne);
+ computeKnownBits(II->getArgOperand(0), KnownZero, KnownOne, 0, II);
unsigned LeadingZeros = KnownOne.countLeadingZeros();
APInt Mask(APInt::getHighBitsSet(BitWidth, LeadingZeros));
if ((Mask & KnownZero) == Mask)
uint32_t BitWidth = IT->getBitWidth();
APInt LHSKnownZero(BitWidth, 0);
APInt LHSKnownOne(BitWidth, 0);
- ComputeMaskedBits(LHS, LHSKnownZero, LHSKnownOne);
+ computeKnownBits(LHS, LHSKnownZero, LHSKnownOne, 0, II);
bool LHSKnownNegative = LHSKnownOne[BitWidth - 1];
bool LHSKnownPositive = LHSKnownZero[BitWidth - 1];
if (LHSKnownNegative || LHSKnownPositive) {
APInt RHSKnownZero(BitWidth, 0);
APInt RHSKnownOne(BitWidth, 0);
- ComputeMaskedBits(RHS, RHSKnownZero, RHSKnownOne);
+ computeKnownBits(RHS, RHSKnownZero, RHSKnownOne, 0, II);
bool RHSKnownNegative = RHSKnownOne[BitWidth - 1];
bool RHSKnownPositive = RHSKnownZero[BitWidth - 1];
if (LHSKnownNegative && RHSKnownNegative) {
return InsertValueInst::Create(Struct, II->getArgOperand(0), 0);
}
}
+
+ // We can strength reduce reduce this signed add into a regular add if we
+ // can prove that it will never overflow.
+ if (II->getIntrinsicID() == Intrinsic::sadd_with_overflow) {
+ Value *LHS = II->getArgOperand(0), *RHS = II->getArgOperand(1);
+ if (WillNotOverflowSignedAdd(LHS, RHS, II)) {
+ Value *Add = Builder->CreateNSWAdd(LHS, RHS);
+ Add->takeName(&CI);
+ Constant *V[] = {UndefValue::get(Add->getType()), Builder->getFalse()};
+ StructType *ST = cast<StructType>(II->getType());
+ Constant *Struct = ConstantStruct::get(ST, V);
+ return InsertValueInst::Create(Struct, Add, 0);
+ }
+ }
+
break;
case Intrinsic::usub_with_overflow:
case Intrinsic::ssub_with_overflow:
APInt LHSKnownZero(BitWidth, 0);
APInt LHSKnownOne(BitWidth, 0);
- ComputeMaskedBits(LHS, LHSKnownZero, LHSKnownOne);
+ computeKnownBits(LHS, LHSKnownZero, LHSKnownOne, 0, II);
APInt RHSKnownZero(BitWidth, 0);
APInt RHSKnownOne(BitWidth, 0);
- ComputeMaskedBits(RHS, RHSKnownZero, RHSKnownOne);
+ computeKnownBits(RHS, RHSKnownZero, RHSKnownOne, 0, II);
// Get the largest possible values for each operand.
APInt LHSMax = ~LHSKnownZero;
case Intrinsic::ppc_altivec_lvx:
case Intrinsic::ppc_altivec_lvxl:
// Turn PPC lvx -> load if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, DL) >= 16) {
+ if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16,
+ DL, AT, II, DT) >= 16) {
Value *Ptr = Builder->CreateBitCast(II->getArgOperand(0),
PointerType::getUnqual(II->getType()));
return new LoadInst(Ptr);
case Intrinsic::ppc_altivec_stvx:
case Intrinsic::ppc_altivec_stvxl:
// Turn stvx -> store if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(1), 16, DL) >= 16) {
+ if (getOrEnforceKnownAlignment(II->getArgOperand(1), 16,
+ DL, AT, II, DT) >= 16) {
Type *OpPtrTy =
PointerType::getUnqual(II->getArgOperand(0)->getType());
Value *Ptr = Builder->CreateBitCast(II->getArgOperand(1), OpPtrTy);
case Intrinsic::x86_sse2_storeu_pd:
case Intrinsic::x86_sse2_storeu_dq:
// Turn X86 storeu -> store if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, DL) >= 16) {
+ if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16,
+ DL, AT, II, DT) >= 16) {
Type *OpPtrTy =
PointerType::getUnqual(II->getArgOperand(1)->getType());
Value *Ptr = Builder->CreateBitCast(II->getArgOperand(0), OpPtrTy);
CI,
Builder->CreateShuffleVector(
Vec, Undef, ConstantDataVector::get(
- II->getContext(), ArrayRef<uint32_t>(Mask))));
+ II->getContext(), makeArrayRef(Mask))));
} else if (auto Source =
dyn_cast<IntrinsicInst>(II->getArgOperand(0))) {
break;
}
+ case Intrinsic::x86_sse41_pblendvb:
+ case Intrinsic::x86_sse41_blendvps:
+ case Intrinsic::x86_sse41_blendvpd:
+ case Intrinsic::x86_avx_blendv_ps_256:
+ case Intrinsic::x86_avx_blendv_pd_256:
+ case Intrinsic::x86_avx2_pblendvb: {
+ // Convert blendv* to vector selects if the mask is constant.
+ // This optimization is convoluted because the intrinsic is defined as
+ // getting a vector of floats or doubles for the ps and pd versions.
+ // FIXME: That should be changed.
+ Value *Mask = II->getArgOperand(2);
+ if (auto C = dyn_cast<ConstantDataVector>(Mask)) {
+ auto Tyi1 = Builder->getInt1Ty();
+ auto SelectorType = cast<VectorType>(Mask->getType());
+ auto EltTy = SelectorType->getElementType();
+ unsigned Size = SelectorType->getNumElements();
+ unsigned BitWidth =
+ EltTy->isFloatTy()
+ ? 32
+ : (EltTy->isDoubleTy() ? 64 : EltTy->getIntegerBitWidth());
+ assert((BitWidth == 64 || BitWidth == 32 || BitWidth == 8) &&
+ "Wrong arguments for variable blend intrinsic");
+ SmallVector<Constant *, 32> Selectors;
+ for (unsigned I = 0; I < Size; ++I) {
+ // The intrinsics only read the top bit
+ uint64_t Selector;
+ if (BitWidth == 8)
+ Selector = C->getElementAsInteger(I);
+ else
+ Selector = C->getElementAsAPFloat(I).bitcastToAPInt().getZExtValue();
+ Selectors.push_back(ConstantInt::get(Tyi1, Selector >> (BitWidth - 1)));
+ }
+ auto NewSelector = ConstantVector::get(Selectors);
+ return SelectInst::Create(NewSelector, II->getArgOperand(1),
+ II->getArgOperand(0), "blendv");
+ } else {
+ break;
+ }
+ }
+
case Intrinsic::x86_avx_vpermilvar_ps:
case Intrinsic::x86_avx_vpermilvar_ps_256:
case Intrinsic::x86_avx_vpermilvar_pd:
case Intrinsic::x86_avx_vpermilvar_pd_256: {
// Convert vpermil* to shufflevector if the mask is constant.
Value *V = II->getArgOperand(1);
+ unsigned Size = cast<VectorType>(V->getType())->getNumElements();
+ assert(Size == 8 || Size == 4 || Size == 2);
+ uint32_t Indexes[8];
if (auto C = dyn_cast<ConstantDataVector>(V)) {
- unsigned Size = C->getNumElements();
- assert(Size == 8 || Size == 4 || Size == 2);
- uint32_t Indexes[8];
-
// The intrinsics only read one or two bits, clear the rest.
for (unsigned I = 0; I < Size; ++I) {
uint32_t Index = C->getElementAsInteger(I) & 0x3;
Index >>= 1;
Indexes[I] = Index;
}
-
- // The _256 variants are a bit trickier since the mask bits always index
- // into the corresponding 128 half. In order to convert to a generic
- // shuffle, we have to make that explicit.
- if (II->getIntrinsicID() == Intrinsic::x86_avx_vpermilvar_ps_256 ||
- II->getIntrinsicID() == Intrinsic::x86_avx_vpermilvar_pd_256) {
- for (unsigned I = Size / 2; I < Size; ++I)
- Indexes[I] += Size / 2;
- }
- auto NewC =
- ConstantDataVector::get(C->getContext(), makeArrayRef(Indexes, Size));
- auto V1 = II->getArgOperand(0);
- auto V2 = UndefValue::get(V1->getType());
- auto Shuffle = Builder->CreateShuffleVector(V1, V2, NewC);
- return ReplaceInstUsesWith(CI, Shuffle);
+ } else if (isa<ConstantAggregateZero>(V)) {
+ for (unsigned I = 0; I < Size; ++I)
+ Indexes[I] = 0;
+ } else {
+ break;
}
- break;
+ // The _256 variants are a bit trickier since the mask bits always index
+ // into the corresponding 128 half. In order to convert to a generic
+ // shuffle, we have to make that explicit.
+ if (II->getIntrinsicID() == Intrinsic::x86_avx_vpermilvar_ps_256 ||
+ II->getIntrinsicID() == Intrinsic::x86_avx_vpermilvar_pd_256) {
+ for (unsigned I = Size / 2; I < Size; ++I)
+ Indexes[I] += Size / 2;
+ }
+ auto NewC =
+ ConstantDataVector::get(V->getContext(), makeArrayRef(Indexes, Size));
+ auto V1 = II->getArgOperand(0);
+ auto V2 = UndefValue::get(V1->getType());
+ auto Shuffle = Builder->CreateShuffleVector(V1, V2, NewC);
+ return ReplaceInstUsesWith(CI, Shuffle);
}
case Intrinsic::ppc_altivec_vperm:
// Turn vperm(V1,V2,mask) -> shuffle(V1,V2,mask) if mask is a constant.
+ // Note that ppc_altivec_vperm has a big-endian bias, so when creating
+ // a vectorshuffle for little endian, we must undo the transformation
+ // performed on vec_perm in altivec.h. That is, we must complement
+ // the permutation mask with respect to 31 and reverse the order of
+ // V1 and V2.
if (Constant *Mask = dyn_cast<Constant>(II->getArgOperand(2))) {
assert(Mask->getType()->getVectorNumElements() == 16 &&
"Bad type for intrinsic!");
unsigned Idx =
cast<ConstantInt>(Mask->getAggregateElement(i))->getZExtValue();
Idx &= 31; // Match the hardware behavior.
+ if (DL && DL->isLittleEndian())
+ Idx = 31 - Idx;
if (!ExtractedElts[Idx]) {
+ Value *Op0ToUse = (DL && DL->isLittleEndian()) ? Op1 : Op0;
+ Value *Op1ToUse = (DL && DL->isLittleEndian()) ? Op0 : Op1;
ExtractedElts[Idx] =
- Builder->CreateExtractElement(Idx < 16 ? Op0 : Op1,
+ Builder->CreateExtractElement(Idx < 16 ? Op0ToUse : Op1ToUse,
Builder->getInt32(Idx&15));
}
case Intrinsic::arm_neon_vst2lane:
case Intrinsic::arm_neon_vst3lane:
case Intrinsic::arm_neon_vst4lane: {
- unsigned MemAlign = getKnownAlignment(II->getArgOperand(0), DL);
+ unsigned MemAlign = getKnownAlignment(II->getArgOperand(0), DL, AT, II, DT);
unsigned AlignArg = II->getNumArgOperands() - 1;
ConstantInt *IntrAlign = dyn_cast<ConstantInt>(II->getArgOperand(AlignArg));
if (IntrAlign && IntrAlign->getZExtValue() < MemAlign) {
case Intrinsic::arm_neon_vmulls:
case Intrinsic::arm_neon_vmullu:
- case Intrinsic::arm64_neon_smull:
- case Intrinsic::arm64_neon_umull: {
+ case Intrinsic::aarch64_neon_smull:
+ case Intrinsic::aarch64_neon_umull: {
Value *Arg0 = II->getArgOperand(0);
Value *Arg1 = II->getArgOperand(1);
// Check for constant LHS & RHS - in this case we just simplify.
bool Zext = (II->getIntrinsicID() == Intrinsic::arm_neon_vmullu ||
- II->getIntrinsicID() == Intrinsic::arm64_neon_umull);
+ II->getIntrinsicID() == Intrinsic::aarch64_neon_umull);
VectorType *NewVT = cast<VectorType>(II->getType());
if (Constant *CV0 = dyn_cast<Constant>(Arg0)) {
if (Constant *CV1 = dyn_cast<Constant>(Arg1)) {
break;
}
+ case Intrinsic::AMDGPU_rcp: {
+ if (const ConstantFP *C = dyn_cast<ConstantFP>(II->getArgOperand(0))) {
+ const APFloat &ArgVal = C->getValueAPF();
+ APFloat Val(ArgVal.getSemantics(), 1.0);
+ APFloat::opStatus Status = Val.divide(ArgVal,
+ APFloat::rmNearestTiesToEven);
+ // Only do this if it was exact and therefore not dependent on the
+ // rounding mode.
+ if (Status == APFloat::opOK)
+ return ReplaceInstUsesWith(CI, ConstantFP::get(II->getContext(), Val));
+ }
+
+ break;
+ }
case Intrinsic::stackrestore: {
// If the save is right next to the restore, remove the restore. This can
// happen when variable allocas are DCE'd.
return EraseInstFromFunction(CI);
break;
}
+ case Intrinsic::assume: {
+ // Canonicalize assume(a && b) -> assume(a); assume(b);
+ // Note: New assumption intrinsics created here are registered by
+ // the InstCombineIRInserter object.
+ Value *IIOperand = II->getArgOperand(0), *A, *B,
+ *AssumeIntrinsic = II->getCalledValue();
+ if (match(IIOperand, m_And(m_Value(A), m_Value(B)))) {
+ Builder->CreateCall(AssumeIntrinsic, A, II->getName());
+ Builder->CreateCall(AssumeIntrinsic, B, II->getName());
+ return EraseInstFromFunction(*II);
+ }
+ // assume(!(a || b)) -> assume(!a); assume(!b);
+ if (match(IIOperand, m_Not(m_Or(m_Value(A), m_Value(B))))) {
+ Builder->CreateCall(AssumeIntrinsic, Builder->CreateNot(A),
+ II->getName());
+ Builder->CreateCall(AssumeIntrinsic, Builder->CreateNot(B),
+ II->getName());
+ return EraseInstFromFunction(*II);
+ }
+
+ // If there is a dominating assume with the same condition as this one,
+ // then this one is redundant, and should be removed.
+ APInt KnownZero(1, 0), KnownOne(1, 0);
+ computeKnownBits(IIOperand, KnownZero, KnownOne, 0, II);
+ if (KnownOne.isAllOnesValue())
+ return EraseInstFromFunction(*II);
+
+ break;
+ }
}
return visitCallSite(II);
projects / oota-llvm.git / blobdiff