Google Git
Sign in
llvm / llvm-project / llvm / refs/heads/master / . / lib / Transforms / Vectorize / VectorCombine.cpp
blob: 52cb1dbb33b863f58f5340dd0deb37cf435923b3 [file] [log] [blame]
//===------- VectorCombine.cpp - Optimize partial vector operations -------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This pass optimizes scalar/vector interactions using target cost models. The
// transforms implemented here may not fit in traditional loop-based or SLP
// vectorization passes.
//
//===----------------------------------------------------------------------===//
#include "llvm/Transforms/Vectorize/VectorCombine.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/ScopeExit.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/BasicAliasAnalysis.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/Analysis/GlobalsModRef.h"
#include "llvm/Analysis/Loads.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/Analysis/VectorUtils.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Transforms/Utils/Local.h"
#include "llvm/Transforms/Utils/LoopUtils.h"
#include <numeric>
#include <queue>
#include <set>
#define DEBUG_TYPE "vector-combine"
#include "llvm/Transforms/Utils/InstructionWorklist.h"
using namespace llvm;
using namespace llvm::PatternMatch;
STATISTIC(NumVecLoad, "Number of vector loads formed");
STATISTIC(NumVecCmp, "Number of vector compares formed");
STATISTIC(NumVecBO, "Number of vector binops formed");
STATISTIC(NumVecCmpBO, "Number of vector compare + binop formed");
STATISTIC(NumShufOfBitcast, "Number of shuffles moved after bitcast");
STATISTIC(NumScalarOps, "Number of scalar unary + binary ops formed");
STATISTIC(NumScalarCmp, "Number of scalar compares formed");
STATISTIC(NumScalarIntrinsic, "Number of scalar intrinsic calls formed");
static cl::opt<bool> DisableVectorCombine(
"disable-vector-combine", cl::init(false), cl::Hidden,
cl::desc("Disable all vector combine transforms"));
static cl::opt<bool> DisableBinopExtractShuffle(
"disable-binop-extract-shuffle", cl::init(false), cl::Hidden,
cl::desc("Disable binop extract to shuffle transforms"));
static cl::opt<unsigned> MaxInstrsToScan(
"vector-combine-max-scan-instrs", cl::init(30), cl::Hidden,
cl::desc("Max number of instructions to scan for vector combining."));
static const unsigned InvalidIndex = std::numeric_limits<unsigned>::max();
namespace {
class VectorCombine {
public:
VectorCombine(Function &F, const TargetTransformInfo &TTI,
const DominatorTree &DT, AAResults &AA, AssumptionCache &AC,
const DataLayout *DL, TTI::TargetCostKind CostKind,
bool TryEarlyFoldsOnly)
: F(F), Builder(F.getContext()), TTI(TTI), DT(DT), AA(AA), AC(AC), DL(DL),
CostKind(CostKind), TryEarlyFoldsOnly(TryEarlyFoldsOnly) {}
bool run();
private:
Function &F;
IRBuilder<> Builder;
const TargetTransformInfo &TTI;
const DominatorTree &DT;
AAResults &AA;
AssumptionCache &AC;
const DataLayout *DL;
TTI::TargetCostKind CostKind;
/// If true, only perform beneficial early IR transforms. Do not introduce new
/// vector operations.
bool TryEarlyFoldsOnly;
InstructionWorklist Worklist;
// TODO: Direct calls from the top-level "run" loop use a plain "Instruction"
// parameter. That should be updated to specific sub-classes because the
// run loop was changed to dispatch on opcode.
bool vectorizeLoadInsert(Instruction &I);
bool widenSubvectorLoad(Instruction &I);
ExtractElementInst *getShuffleExtract(ExtractElementInst *Ext0,
ExtractElementInst *Ext1,
unsigned PreferredExtractIndex) const;
bool isExtractExtractCheap(ExtractElementInst *Ext0, ExtractElementInst *Ext1,
const Instruction &I,
ExtractElementInst *&ConvertToShuffle,
unsigned PreferredExtractIndex);
void foldExtExtCmp(ExtractElementInst *Ext0, ExtractElementInst *Ext1,
Instruction &I);
void foldExtExtBinop(ExtractElementInst *Ext0, ExtractElementInst *Ext1,
Instruction &I);
bool foldExtractExtract(Instruction &I);
bool foldInsExtFNeg(Instruction &I);
bool foldInsExtBinop(Instruction &I);
bool foldInsExtVectorToShuffle(Instruction &I);
bool foldBitcastShuffle(Instruction &I);
bool scalarizeOpOrCmp(Instruction &I);
bool scalarizeVPIntrinsic(Instruction &I);
bool foldExtractedCmps(Instruction &I);
bool foldBinopOfReductions(Instruction &I);
bool foldSingleElementStore(Instruction &I);
bool scalarizeLoadExtract(Instruction &I);
bool foldConcatOfBoolMasks(Instruction &I);
bool foldPermuteOfBinops(Instruction &I);
bool foldShuffleOfBinops(Instruction &I);
bool foldShuffleOfSelects(Instruction &I);
bool foldShuffleOfCastops(Instruction &I);
bool foldShuffleOfShuffles(Instruction &I);
bool foldShuffleOfIntrinsics(Instruction &I);
bool foldShuffleToIdentity(Instruction &I);
bool foldShuffleFromReductions(Instruction &I);
bool foldCastFromReductions(Instruction &I);
bool foldSelectShuffle(Instruction &I, bool FromReduction = false);
bool foldInterleaveIntrinsics(Instruction &I);
bool shrinkType(Instruction &I);
void replaceValue(Value &Old, Value &New) {
LLVM_DEBUG(dbgs() << "VC: Replacing: " << Old << '\n');
LLVM_DEBUG(dbgs() << " With: " << New << '\n');
Old.replaceAllUsesWith(&New);
if (auto *NewI = dyn_cast<Instruction>(&New)) {
New.takeName(&Old);
Worklist.pushUsersToWorkList(*NewI);
Worklist.pushValue(NewI);
}
Worklist.pushValue(&Old);
}
void eraseInstruction(Instruction &I) {
LLVM_DEBUG(dbgs() << "VC: Erasing: " << I << '\n');
SmallVector<Value *> Ops(I.operands());
Worklist.remove(&I);
I.eraseFromParent();
// Push remaining users of the operands and then the operand itself - allows
// further folds that were hindered by OneUse limits.
for (Value *Op : Ops)
if (auto *OpI = dyn_cast<Instruction>(Op)) {
Worklist.pushUsersToWorkList(*OpI);
Worklist.pushValue(OpI);
}
}
};
} // namespace
/// Return the source operand of a potentially bitcasted value. If there is no
/// bitcast, return the input value itself.
static Value *peekThroughBitcasts(Value *V) {
while (auto *BitCast = dyn_cast<BitCastInst>(V))
V = BitCast->getOperand(0);
return V;
}
static bool canWidenLoad(LoadInst *Load, const TargetTransformInfo &TTI) {
// Do not widen load if atomic/volatile or under asan/hwasan/memtag/tsan.
// The widened load may load data from dirty regions or create data races
// non-existent in the source.
if (!Load || !Load->isSimple() || !Load->hasOneUse() ||
Load->getFunction()->hasFnAttribute(Attribute::SanitizeMemTag) ||
mustSuppressSpeculation(*Load))
return false;
// We are potentially transforming byte-sized (8-bit) memory accesses, so make
// sure we have all of our type-based constraints in place for this target.
Type *ScalarTy = Load->getType()->getScalarType();
uint64_t ScalarSize = ScalarTy->getPrimitiveSizeInBits();
unsigned MinVectorSize = TTI.getMinVectorRegisterBitWidth();
if (!ScalarSize || !MinVectorSize || MinVectorSize % ScalarSize != 0 ||
ScalarSize % 8 != 0)
return false;
return true;
}
bool VectorCombine::vectorizeLoadInsert(Instruction &I) {
// Match insert into fixed vector of scalar value.
// TODO: Handle non-zero insert index.
Value *Scalar;
if (!match(&I,
m_InsertElt(m_Poison(), m_OneUse(m_Value(Scalar)), m_ZeroInt())))
return false;
// Optionally match an extract from another vector.
Value *X;
bool HasExtract = match(Scalar, m_ExtractElt(m_Value(X), m_ZeroInt()));
if (!HasExtract)
X = Scalar;
auto *Load = dyn_cast<LoadInst>(X);
if (!canWidenLoad(Load, TTI))
return false;
Type *ScalarTy = Scalar->getType();
uint64_t ScalarSize = ScalarTy->getPrimitiveSizeInBits();
unsigned MinVectorSize = TTI.getMinVectorRegisterBitWidth();
// Check safety of replacing the scalar load with a larger vector load.
// We use minimal alignment (maximum flexibility) because we only care about
// the dereferenceable region. When calculating cost and creating a new op,
// we may use a larger value based on alignment attributes.
Value *SrcPtr = Load->getPointerOperand()->stripPointerCasts();
assert(isa<PointerType>(SrcPtr->getType()) && "Expected a pointer type");
unsigned MinVecNumElts = MinVectorSize / ScalarSize;
auto *MinVecTy = VectorType::get(ScalarTy, MinVecNumElts, false);
unsigned OffsetEltIndex = 0;
Align Alignment = Load->getAlign();
if (!isSafeToLoadUnconditionally(SrcPtr, MinVecTy, Align(1), *DL, Load, &AC,
&DT)) {
// It is not safe to load directly from the pointer, but we can still peek
// through gep offsets and check if it safe to load from a base address with
// updated alignment. If it is, we can shuffle the element(s) into place
// after loading.
unsigned OffsetBitWidth = DL->getIndexTypeSizeInBits(SrcPtr->getType());
APInt Offset(OffsetBitWidth, 0);
SrcPtr = SrcPtr->stripAndAccumulateInBoundsConstantOffsets(*DL, Offset);
// We want to shuffle the result down from a high element of a vector, so
// the offset must be positive.
if (Offset.isNegative())
return false;
// The offset must be a multiple of the scalar element to shuffle cleanly
// in the element's size.
uint64_t ScalarSizeInBytes = ScalarSize / 8;
if (Offset.urem(ScalarSizeInBytes) != 0)
return false;
// If we load MinVecNumElts, will our target element still be loaded?
OffsetEltIndex = Offset.udiv(ScalarSizeInBytes).getZExtValue();
if (OffsetEltIndex >= MinVecNumElts)
return false;
if (!isSafeToLoadUnconditionally(SrcPtr, MinVecTy, Align(1), *DL, Load, &AC,
&DT))
return false;
// Update alignment with offset value. Note that the offset could be negated
// to more accurately represent "(new) SrcPtr - Offset = (old) SrcPtr", but
// negation does not change the result of the alignment calculation.
Alignment = commonAlignment(Alignment, Offset.getZExtValue());
}
// Original pattern: insertelt undef, load [free casts of] PtrOp, 0
// Use the greater of the alignment on the load or its source pointer.
Alignment = std::max(SrcPtr->getPointerAlignment(*DL), Alignment);
Type *LoadTy = Load->getType();
unsigned AS = Load->getPointerAddressSpace();
InstructionCost OldCost =
TTI.getMemoryOpCost(Instruction::Load, LoadTy, Alignment, AS, CostKind);
APInt DemandedElts = APInt::getOneBitSet(MinVecNumElts, 0);
OldCost +=
TTI.getScalarizationOverhead(MinVecTy, DemandedElts,
/* Insert */ true, HasExtract, CostKind);
// New pattern: load VecPtr
InstructionCost NewCost =
TTI.getMemoryOpCost(Instruction::Load, MinVecTy, Alignment, AS, CostKind);
// Optionally, we are shuffling the loaded vector element(s) into place.
// For the mask set everything but element 0 to undef to prevent poison from
// propagating from the extra loaded memory. This will also optionally
// shrink/grow the vector from the loaded size to the output size.
// We assume this operation has no cost in codegen if there was no offset.
// Note that we could use freeze to avoid poison problems, but then we might
// still need a shuffle to change the vector size.
auto *Ty = cast<FixedVectorType>(I.getType());
unsigned OutputNumElts = Ty->getNumElements();
SmallVector<int, 16> Mask(OutputNumElts, PoisonMaskElem);
assert(OffsetEltIndex < MinVecNumElts && "Address offset too big");
Mask[0] = OffsetEltIndex;
if (OffsetEltIndex)
NewCost +=
TTI.getShuffleCost(TTI::SK_PermuteSingleSrc, MinVecTy, Mask, CostKind);
// We can aggressively convert to the vector form because the backend can
// invert this transform if it does not result in a performance win.
if (OldCost < NewCost || !NewCost.isValid())
return false;
// It is safe and potentially profitable to load a vector directly:
// inselt undef, load Scalar, 0 --> load VecPtr
IRBuilder<> Builder(Load);
Value *CastedPtr =
Builder.CreatePointerBitCastOrAddrSpaceCast(SrcPtr, Builder.getPtrTy(AS));
Value *VecLd = Builder.CreateAlignedLoad(MinVecTy, CastedPtr, Alignment);
VecLd = Builder.CreateShuffleVector(VecLd, Mask);
replaceValue(I, *VecLd);
++NumVecLoad;
return true;
}
/// If we are loading a vector and then inserting it into a larger vector with
/// undefined elements, try to load the larger vector and eliminate the insert.
/// This removes a shuffle in IR and may allow combining of other loaded values.
bool VectorCombine::widenSubvectorLoad(Instruction &I) {
// Match subvector insert of fixed vector.
auto *Shuf = cast<ShuffleVectorInst>(&I);
if (!Shuf->isIdentityWithPadding())
return false;
// Allow a non-canonical shuffle mask that is choosing elements from op1.
unsigned NumOpElts =
cast<FixedVectorType>(Shuf->getOperand(0)->getType())->getNumElements();
unsigned OpIndex = any_of(Shuf->getShuffleMask(), [&NumOpElts](int M) {
return M >= (int)(NumOpElts);
});
auto *Load = dyn_cast<LoadInst>(Shuf->getOperand(OpIndex));
if (!canWidenLoad(Load, TTI))
return false;
// We use minimal alignment (maximum flexibility) because we only care about
// the dereferenceable region. When calculating cost and creating a new op,
// we may use a larger value based on alignment attributes.
auto *Ty = cast<FixedVectorType>(I.getType());
Value *SrcPtr = Load->getPointerOperand()->stripPointerCasts();
assert(isa<PointerType>(SrcPtr->getType()) && "Expected a pointer type");
Align Alignment = Load->getAlign();
if (!isSafeToLoadUnconditionally(SrcPtr, Ty, Align(1), *DL, Load, &AC, &DT))
return false;
Alignment = std::max(SrcPtr->getPointerAlignment(*DL), Alignment);
Type *LoadTy = Load->getType();
unsigned AS = Load->getPointerAddressSpace();
// Original pattern: insert_subvector (load PtrOp)
// This conservatively assumes that the cost of a subvector insert into an
// undef value is 0. We could add that cost if the cost model accurately
// reflects the real cost of that operation.
InstructionCost OldCost =
TTI.getMemoryOpCost(Instruction::Load, LoadTy, Alignment, AS, CostKind);
// New pattern: load PtrOp
InstructionCost NewCost =
TTI.getMemoryOpCost(Instruction::Load, Ty, Alignment, AS, CostKind);
// We can aggressively convert to the vector form because the backend can
// invert this transform if it does not result in a performance win.
if (OldCost < NewCost || !NewCost.isValid())
return false;
IRBuilder<> Builder(Load);
Value *CastedPtr =
Builder.CreatePointerBitCastOrAddrSpaceCast(SrcPtr, Builder.getPtrTy(AS));
Value *VecLd = Builder.CreateAlignedLoad(Ty, CastedPtr, Alignment);
replaceValue(I, *VecLd);
++NumVecLoad;
return true;
}
/// Determine which, if any, of the inputs should be replaced by a shuffle
/// followed by extract from a different index.
ExtractElementInst *VectorCombine::getShuffleExtract(
ExtractElementInst *Ext0, ExtractElementInst *Ext1,
unsigned PreferredExtractIndex = InvalidIndex) const {
auto *Index0C = dyn_cast<ConstantInt>(Ext0->getIndexOperand());
auto *Index1C = dyn_cast<ConstantInt>(Ext1->getIndexOperand());
assert(Index0C && Index1C && "Expected constant extract indexes");
unsigned Index0 = Index0C->getZExtValue();
unsigned Index1 = Index1C->getZExtValue();
// If the extract indexes are identical, no shuffle is needed.
if (Index0 == Index1)
return nullptr;
Type *VecTy = Ext0->getVectorOperand()->getType();
assert(VecTy == Ext1->getVectorOperand()->getType() && "Need matching types");
InstructionCost Cost0 =
TTI.getVectorInstrCost(*Ext0, VecTy, CostKind, Index0);
InstructionCost Cost1 =
TTI.getVectorInstrCost(*Ext1, VecTy, CostKind, Index1);
// If both costs are invalid no shuffle is needed
if (!Cost0.isValid() && !Cost1.isValid())
return nullptr;
// We are extracting from 2 different indexes, so one operand must be shuffled
// before performing a vector operation and/or extract. The more expensive
// extract will be replaced by a shuffle.
if (Cost0 > Cost1)
return Ext0;
if (Cost1 > Cost0)
return Ext1;
// If the costs are equal and there is a preferred extract index, shuffle the
// opposite operand.
if (PreferredExtractIndex == Index0)
return Ext1;
if (PreferredExtractIndex == Index1)
return Ext0;
// Otherwise, replace the extract with the higher index.
return Index0 > Index1 ? Ext0 : Ext1;
}
/// Compare the relative costs of 2 extracts followed by scalar operation vs.
/// vector operation(s) followed by extract. Return true if the existing
/// instructions are cheaper than a vector alternative. Otherwise, return false
/// and if one of the extracts should be transformed to a shufflevector, set
/// \p ConvertToShuffle to that extract instruction.
bool VectorCombine::isExtractExtractCheap(ExtractElementInst *Ext0,
ExtractElementInst *Ext1,
const Instruction &I,
ExtractElementInst *&ConvertToShuffle,
unsigned PreferredExtractIndex) {
auto *Ext0IndexC = dyn_cast<ConstantInt>(Ext0->getIndexOperand());
auto *Ext1IndexC = dyn_cast<ConstantInt>(Ext1->getIndexOperand());
assert(Ext0IndexC && Ext1IndexC && "Expected constant extract indexes");
unsigned Opcode = I.getOpcode();
Value *Ext0Src = Ext0->getVectorOperand();
Value *Ext1Src = Ext1->getVectorOperand();
Type *ScalarTy = Ext0->getType();
auto *VecTy = cast<VectorType>(Ext0Src->getType());
InstructionCost ScalarOpCost, VectorOpCost;
// Get cost estimates for scalar and vector versions of the operation.
bool IsBinOp = Instruction::isBinaryOp(Opcode);
if (IsBinOp) {
ScalarOpCost = TTI.getArithmeticInstrCost(Opcode, ScalarTy, CostKind);
VectorOpCost = TTI.getArithmeticInstrCost(Opcode, VecTy, CostKind);
} else {
assert((Opcode == Instruction::ICmp || Opcode == Instruction::FCmp) &&
"Expected a compare");
CmpInst::Predicate Pred = cast<CmpInst>(I).getPredicate();
ScalarOpCost = TTI.getCmpSelInstrCost(
Opcode, ScalarTy, CmpInst::makeCmpResultType(ScalarTy), Pred, CostKind);
VectorOpCost = TTI.getCmpSelInstrCost(
Opcode, VecTy, CmpInst::makeCmpResultType(VecTy), Pred, CostKind);
}
// Get cost estimates for the extract elements. These costs will factor into
// both sequences.
unsigned Ext0Index = Ext0IndexC->getZExtValue();
unsigned Ext1Index = Ext1IndexC->getZExtValue();
InstructionCost Extract0Cost =
TTI.getVectorInstrCost(*Ext0, VecTy, CostKind, Ext0Index);
InstructionCost Extract1Cost =
TTI.getVectorInstrCost(*Ext1, VecTy, CostKind, Ext1Index);
// A more expensive extract will always be replaced by a splat shuffle.
// For example, if Ext0 is more expensive:
// opcode (extelt V0, Ext0), (ext V1, Ext1) -->
// extelt (opcode (splat V0, Ext0), V1), Ext1
// TODO: Evaluate whether that always results in lowest cost. Alternatively,
// check the cost of creating a broadcast shuffle and shuffling both
// operands to element 0.
unsigned BestExtIndex = Extract0Cost > Extract1Cost ? Ext0Index : Ext1Index;
unsigned BestInsIndex = Extract0Cost > Extract1Cost ? Ext1Index : Ext0Index;
InstructionCost CheapExtractCost = std::min(Extract0Cost, Extract1Cost);
// Extra uses of the extracts mean that we include those costs in the
// vector total because those instructions will not be eliminated.
InstructionCost OldCost, NewCost;
if (Ext0Src == Ext1Src && Ext0Index == Ext1Index) {
// Handle a special case. If the 2 extracts are identical, adjust the
// formulas to account for that. The extra use charge allows for either the
// CSE'd pattern or an unoptimized form with identical values:
// opcode (extelt V, C), (extelt V, C) --> extelt (opcode V, V), C
bool HasUseTax = Ext0 == Ext1 ? !Ext0->hasNUses(2)
: !Ext0->hasOneUse() || !Ext1->hasOneUse();
OldCost = CheapExtractCost + ScalarOpCost;
NewCost = VectorOpCost + CheapExtractCost + HasUseTax * CheapExtractCost;
} else {
// Handle the general case. Each extract is actually a different value:
// opcode (extelt V0, C0), (extelt V1, C1) --> extelt (opcode V0, V1), C
OldCost = Extract0Cost + Extract1Cost + ScalarOpCost;
NewCost = VectorOpCost + CheapExtractCost +
!Ext0->hasOneUse() * Extract0Cost +
!Ext1->hasOneUse() * Extract1Cost;
}
ConvertToShuffle = getShuffleExtract(Ext0, Ext1, PreferredExtractIndex);
if (ConvertToShuffle) {
if (IsBinOp && DisableBinopExtractShuffle)
return true;
// If we are extracting from 2 different indexes, then one operand must be
// shuffled before performing the vector operation. The shuffle mask is
// poison except for 1 lane that is being translated to the remaining
// extraction lane. Therefore, it is a splat shuffle. Ex:
// ShufMask = { poison, poison, 0, poison }
// TODO: The cost model has an option for a "broadcast" shuffle
// (splat-from-element-0), but no option for a more general splat.
if (auto *FixedVecTy = dyn_cast<FixedVectorType>(VecTy)) {
SmallVector<int> ShuffleMask(FixedVecTy->getNumElements(),
PoisonMaskElem);
ShuffleMask[BestInsIndex] = BestExtIndex;
NewCost += TTI.getShuffleCost(TargetTransformInfo::SK_PermuteSingleSrc,
VecTy, ShuffleMask, CostKind, 0, nullptr,
{ConvertToShuffle});
} else {
NewCost +=
TTI.getShuffleCost(TargetTransformInfo::SK_PermuteSingleSrc, VecTy,
{}, CostKind, 0, nullptr, {ConvertToShuffle});
}
}
// Aggressively form a vector op if the cost is equal because the transform
// may enable further optimization.
// Codegen can reverse this transform (scalarize) if it was not profitable.
return OldCost < NewCost;
}
/// Create a shuffle that translates (shifts) 1 element from the input vector
/// to a new element location.
static Value *createShiftShuffle(Value *Vec, unsigned OldIndex,
unsigned NewIndex, IRBuilder<> &Builder) {
// The shuffle mask is poison except for 1 lane that is being translated
// to the new element index. Example for OldIndex == 2 and NewIndex == 0:
// ShufMask = { 2, poison, poison, poison }
auto *VecTy = cast<FixedVectorType>(Vec->getType());
SmallVector<int, 32> ShufMask(VecTy->getNumElements(), PoisonMaskElem);
ShufMask[NewIndex] = OldIndex;
return Builder.CreateShuffleVector(Vec, ShufMask, "shift");
}
/// Given an extract element instruction with constant index operand, shuffle
/// the source vector (shift the scalar element) to a NewIndex for extraction.
/// Return null if the input can be constant folded, so that we are not creating
/// unnecessary instructions.
static ExtractElementInst *translateExtract(ExtractElementInst *ExtElt,
unsigned NewIndex,
IRBuilder<> &Builder) {
// Shufflevectors can only be created for fixed-width vectors.
Value *X = ExtElt->getVectorOperand();
if (!isa<FixedVectorType>(X->getType()))
return nullptr;
// If the extract can be constant-folded, this code is unsimplified. Defer
// to other passes to handle that.
Value *C = ExtElt->getIndexOperand();
assert(isa<ConstantInt>(C) && "Expected a constant index operand");
if (isa<Constant>(X))
return nullptr;
Value *Shuf = createShiftShuffle(X, cast<ConstantInt>(C)->getZExtValue(),
NewIndex, Builder);
return cast<ExtractElementInst>(Builder.CreateExtractElement(Shuf, NewIndex));
}
/// Try to reduce extract element costs by converting scalar compares to vector
/// compares followed by extract.
/// cmp (ext0 V0, C), (ext1 V1, C)
void VectorCombine::foldExtExtCmp(ExtractElementInst *Ext0,
ExtractElementInst *Ext1, Instruction &I) {
assert(isa<CmpInst>(&I) && "Expected a compare");
assert(cast<ConstantInt>(Ext0->getIndexOperand())->getZExtValue() ==
cast<ConstantInt>(Ext1->getIndexOperand())->getZExtValue() &&
"Expected matching constant extract indexes");
// cmp Pred (extelt V0, C), (extelt V1, C) --> extelt (cmp Pred V0, V1), C
++NumVecCmp;
CmpInst::Predicate Pred = cast<CmpInst>(&I)->getPredicate();
Value *V0 = Ext0->getVectorOperand(), *V1 = Ext1->getVectorOperand();
Value *VecCmp = Builder.CreateCmp(Pred, V0, V1);
Value *NewExt = Builder.CreateExtractElement(VecCmp, Ext0->getIndexOperand());
replaceValue(I, *NewExt);
}
/// Try to reduce extract element costs by converting scalar binops to vector
/// binops followed by extract.
/// bo (ext0 V0, C), (ext1 V1, C)
void VectorCombine::foldExtExtBinop(ExtractElementInst *Ext0,
ExtractElementInst *Ext1, Instruction &I) {
assert(isa<BinaryOperator>(&I) && "Expected a binary operator");
assert(cast<ConstantInt>(Ext0->getIndexOperand())->getZExtValue() ==
cast<ConstantInt>(Ext1->getIndexOperand())->getZExtValue() &&
"Expected matching constant extract indexes");
// bo (extelt V0, C), (extelt V1, C) --> extelt (bo V0, V1), C
++NumVecBO;
Value *V0 = Ext0->getVectorOperand(), *V1 = Ext1->getVectorOperand();
Value *VecBO =
Builder.CreateBinOp(cast<BinaryOperator>(&I)->getOpcode(), V0, V1);
// All IR flags are safe to back-propagate because any potential poison
// created in unused vector elements is discarded by the extract.
if (auto *VecBOInst = dyn_cast<Instruction>(VecBO))
VecBOInst->copyIRFlags(&I);
Value *NewExt = Builder.CreateExtractElement(VecBO, Ext0->getIndexOperand());
replaceValue(I, *NewExt);
}
/// Match an instruction with extracted vector operands.
bool VectorCombine::foldExtractExtract(Instruction &I) {
// It is not safe to transform things like div, urem, etc. because we may
// create undefined behavior when executing those on unknown vector elements.
if (!isSafeToSpeculativelyExecute(&I))
return false;
Instruction *I0, *I1;
CmpPredicate Pred = CmpInst::BAD_ICMP_PREDICATE;
if (!match(&I, m_Cmp(Pred, m_Instruction(I0), m_Instruction(I1))) &&
!match(&I, m_BinOp(m_Instruction(I0), m_Instruction(I1))))
return false;
Value *V0, *V1;
uint64_t C0, C1;
if (!match(I0, m_ExtractElt(m_Value(V0), m_ConstantInt(C0))) ||
!match(I1, m_ExtractElt(m_Value(V1), m_ConstantInt(C1))) ||
V0->getType() != V1->getType())
return false;
// If the scalar value 'I' is going to be re-inserted into a vector, then try
// to create an extract to that same element. The extract/insert can be
// reduced to a "select shuffle".
// TODO: If we add a larger pattern match that starts from an insert, this
// probably becomes unnecessary.
auto *Ext0 = cast<ExtractElementInst>(I0);
auto *Ext1 = cast<ExtractElementInst>(I1);
uint64_t InsertIndex = InvalidIndex;
if (I.hasOneUse())
match(I.user_back(),
m_InsertElt(m_Value(), m_Value(), m_ConstantInt(InsertIndex)));
ExtractElementInst *ExtractToChange;
if (isExtractExtractCheap(Ext0, Ext1, I, ExtractToChange, InsertIndex))
return false;
if (ExtractToChange) {
unsigned CheapExtractIdx = ExtractToChange == Ext0 ? C1 : C0;
ExtractElementInst *NewExtract =
translateExtract(ExtractToChange, CheapExtractIdx, Builder);
if (!NewExtract)
return false;
if (ExtractToChange == Ext0)
Ext0 = NewExtract;
else
Ext1 = NewExtract;
}
if (Pred != CmpInst::BAD_ICMP_PREDICATE)
foldExtExtCmp(Ext0, Ext1, I);
else
foldExtExtBinop(Ext0, Ext1, I);
Worklist.push(Ext0);
Worklist.push(Ext1);
return true;
}
/// Try to replace an extract + scalar fneg + insert with a vector fneg +
/// shuffle.
bool VectorCombine::foldInsExtFNeg(Instruction &I) {
// Match an insert (op (extract)) pattern.
Value *DestVec;
uint64_t Index;
Instruction *FNeg;
if (!match(&I, m_InsertElt(m_Value(DestVec), m_OneUse(m_Instruction(FNeg)),
m_ConstantInt(Index))))
return false;
// Note: This handles the canonical fneg instruction and "fsub -0.0, X".
Value *SrcVec;
Instruction *Extract;
if (!match(FNeg, m_FNeg(m_CombineAnd(
m_Instruction(Extract),
m_ExtractElt(m_Value(SrcVec), m_SpecificInt(Index))))))
return false;
auto *VecTy = cast<FixedVectorType>(I.getType());
auto *ScalarTy = VecTy->getScalarType();
auto *SrcVecTy = dyn_cast<FixedVectorType>(SrcVec->getType());
if (!SrcVecTy || ScalarTy != SrcVecTy->getScalarType())
return false;
// Ignore bogus insert/extract index.
unsigned NumElts = VecTy->getNumElements();
if (Index >= NumElts)
return false;
// We are inserting the negated element into the same lane that we extracted
// from. This is equivalent to a select-shuffle that chooses all but the
// negated element from the destination vector.
SmallVector<int> Mask(NumElts);
std::iota(Mask.begin(), Mask.end(), 0);
Mask[Index] = Index + NumElts;
InstructionCost OldCost =
TTI.getArithmeticInstrCost(Instruction::FNeg, ScalarTy, CostKind) +
TTI.getVectorInstrCost(I, VecTy, CostKind, Index);
// If the extract has one use, it will be eliminated, so count it in the
// original cost. If it has more than one use, ignore the cost because it will
// be the same before/after.
if (Extract->hasOneUse())
OldCost += TTI.getVectorInstrCost(*Extract, VecTy, CostKind, Index);
InstructionCost NewCost =
TTI.getArithmeticInstrCost(Instruction::FNeg, VecTy, CostKind) +
TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, VecTy, Mask,
CostKind);
bool NeedLenChg = SrcVecTy->getNumElements() != NumElts;
// If the lengths of the two vectors are not equal,
// we need to add a length-change vector. Add this cost.
SmallVector<int> SrcMask;
if (NeedLenChg) {
SrcMask.assign(NumElts, PoisonMaskElem);
SrcMask[Index] = Index;
NewCost += TTI.getShuffleCost(TargetTransformInfo::SK_PermuteSingleSrc,
SrcVecTy, SrcMask, CostKind);
}
if (NewCost > OldCost)
return false;
Value *NewShuf;
// insertelt DestVec, (fneg (extractelt SrcVec, Index)), Index
Value *VecFNeg = Builder.CreateFNegFMF(SrcVec, FNeg);
if (NeedLenChg) {
// shuffle DestVec, (shuffle (fneg SrcVec), poison, SrcMask), Mask
Value *LenChgShuf = Builder.CreateShuffleVector(VecFNeg, SrcMask);
NewShuf = Builder.CreateShuffleVector(DestVec, LenChgShuf, Mask);
} else {
// shuffle DestVec, (fneg SrcVec), Mask
NewShuf = Builder.CreateShuffleVector(DestVec, VecFNeg, Mask);
}
replaceValue(I, *NewShuf);
return true;
}
/// Try to fold insert(binop(x,y),binop(a,b),idx)
/// --> binop(insert(x,a,idx),insert(y,b,idx))
bool VectorCombine::foldInsExtBinop(Instruction &I) {
BinaryOperator *VecBinOp, *SclBinOp;
uint64_t Index;
if (!match(&I,
m_InsertElt(m_OneUse(m_BinOp(VecBinOp)),
m_OneUse(m_BinOp(SclBinOp)), m_ConstantInt(Index))))
return false;
// TODO: Add support for addlike etc.
Instruction::BinaryOps BinOpcode = VecBinOp->getOpcode();
if (BinOpcode != SclBinOp->getOpcode())
return false;
auto *ResultTy = dyn_cast<FixedVectorType>(I.getType());
if (!ResultTy)
return false;
// TODO: Attempt to detect m_ExtractElt for scalar operands and convert to
// shuffle?
InstructionCost OldCost = TTI.getInstructionCost(&I, CostKind) +
TTI.getInstructionCost(VecBinOp, CostKind) +
TTI.getInstructionCost(SclBinOp, CostKind);
InstructionCost NewCost =
TTI.getArithmeticInstrCost(BinOpcode, ResultTy, CostKind) +
TTI.getVectorInstrCost(Instruction::InsertElement, ResultTy, CostKind,
Index, VecBinOp->getOperand(0),
SclBinOp->getOperand(0)) +
TTI.getVectorInstrCost(Instruction::InsertElement, ResultTy, CostKind,
Index, VecBinOp->getOperand(1),
SclBinOp->getOperand(1));
LLVM_DEBUG(dbgs() << "Found an insertion of two binops: " << I
<< "\n OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
if (NewCost > OldCost)
return false;
Value *NewIns0 = Builder.CreateInsertElement(VecBinOp->getOperand(0),
SclBinOp->getOperand(0), Index);
Value *NewIns1 = Builder.CreateInsertElement(VecBinOp->getOperand(1),
SclBinOp->getOperand(1), Index);
Value *NewBO = Builder.CreateBinOp(BinOpcode, NewIns0, NewIns1);
// Intersect flags from the old binops.
if (auto *NewInst = dyn_cast<Instruction>(NewBO)) {
NewInst->copyIRFlags(VecBinOp);
NewInst->andIRFlags(SclBinOp);
}
Worklist.pushValue(NewIns0);
Worklist.pushValue(NewIns1);
replaceValue(I, *NewBO);
return true;
}
/// If this is a bitcast of a shuffle, try to bitcast the source vector to the
/// destination type followed by shuffle. This can enable further transforms by
/// moving bitcasts or shuffles together.
bool VectorCombine::foldBitcastShuffle(Instruction &I) {
Value *V0, *V1;
ArrayRef<int> Mask;
if (!match(&I, m_BitCast(m_OneUse(
m_Shuffle(m_Value(V0), m_Value(V1), m_Mask(Mask))))))
return false;
// 1) Do not fold bitcast shuffle for scalable type. First, shuffle cost for
// scalable type is unknown; Second, we cannot reason if the narrowed shuffle
// mask for scalable type is a splat or not.
// 2) Disallow non-vector casts.
// TODO: We could allow any shuffle.
auto *DestTy = dyn_cast<FixedVectorType>(I.getType());
auto *SrcTy = dyn_cast<FixedVectorType>(V0->getType());
if (!DestTy || !SrcTy)
return false;
unsigned DestEltSize = DestTy->getScalarSizeInBits();
unsigned SrcEltSize = SrcTy->getScalarSizeInBits();
if (SrcTy->getPrimitiveSizeInBits() % DestEltSize != 0)
return false;
bool IsUnary = isa<UndefValue>(V1);
// For binary shuffles, only fold bitcast(shuffle(X,Y))
// if it won't increase the number of bitcasts.
if (!IsUnary) {
auto *BCTy0 = dyn_cast<FixedVectorType>(peekThroughBitcasts(V0)->getType());
auto *BCTy1 = dyn_cast<FixedVectorType>(peekThroughBitcasts(V1)->getType());
if (!(BCTy0 && BCTy0->getElementType() == DestTy->getElementType()) &&
!(BCTy1 && BCTy1->getElementType() == DestTy->getElementType()))
return false;
}
SmallVector<int, 16> NewMask;
if (DestEltSize <= SrcEltSize) {
// The bitcast is from wide to narrow/equal elements. The shuffle mask can
// always be expanded to the equivalent form choosing narrower elements.
assert(SrcEltSize % DestEltSize == 0 && "Unexpected shuffle mask");
unsigned ScaleFactor = SrcEltSize / DestEltSize;
narrowShuffleMaskElts(ScaleFactor, Mask, NewMask);
} else {
// The bitcast is from narrow elements to wide elements. The shuffle mask
// must choose consecutive elements to allow casting first.
assert(DestEltSize % SrcEltSize == 0 && "Unexpected shuffle mask");
unsigned ScaleFactor = DestEltSize / SrcEltSize;
if (!widenShuffleMaskElts(ScaleFactor, Mask, NewMask))
return false;
}
// Bitcast the shuffle src - keep its original width but using the destination
// scalar type.
unsigned NumSrcElts = SrcTy->getPrimitiveSizeInBits() / DestEltSize;
auto *NewShuffleTy =
FixedVectorType::get(DestTy->getScalarType(), NumSrcElts);
auto *OldShuffleTy =
FixedVectorType::get(SrcTy->getScalarType(), Mask.size());
unsigned NumOps = IsUnary ? 1 : 2;
// The new shuffle must not cost more than the old shuffle.
TargetTransformInfo::ShuffleKind SK =
IsUnary ? TargetTransformInfo::SK_PermuteSingleSrc
: TargetTransformInfo::SK_PermuteTwoSrc;
InstructionCost NewCost =
TTI.getShuffleCost(SK, NewShuffleTy, NewMask, CostKind) +
(NumOps * TTI.getCastInstrCost(Instruction::BitCast, NewShuffleTy, SrcTy,
TargetTransformInfo::CastContextHint::None,
CostKind));
InstructionCost OldCost =
TTI.getShuffleCost(SK, SrcTy, Mask, CostKind) +
TTI.getCastInstrCost(Instruction::BitCast, DestTy, OldShuffleTy,
TargetTransformInfo::CastContextHint::None,
CostKind);
LLVM_DEBUG(dbgs() << "Found a bitcasted shuffle: " << I << "\n OldCost: "
<< OldCost << " vs NewCost: " << NewCost << "\n");
if (NewCost > OldCost || !NewCost.isValid())
return false;
// bitcast (shuf V0, V1, MaskC) --> shuf (bitcast V0), (bitcast V1), MaskC'
++NumShufOfBitcast;
Value *CastV0 = Builder.CreateBitCast(peekThroughBitcasts(V0), NewShuffleTy);
Value *CastV1 = Builder.CreateBitCast(peekThroughBitcasts(V1), NewShuffleTy);
Value *Shuf = Builder.CreateShuffleVector(CastV0, CastV1, NewMask);
replaceValue(I, *Shuf);
return true;
}
/// VP Intrinsics whose vector operands are both splat values may be simplified
/// into the scalar version of the operation and the result splatted. This
/// can lead to scalarization down the line.
bool VectorCombine::scalarizeVPIntrinsic(Instruction &I) {
if (!isa<VPIntrinsic>(I))
return false;
VPIntrinsic &VPI = cast<VPIntrinsic>(I);
Value *Op0 = VPI.getArgOperand(0);
Value *Op1 = VPI.getArgOperand(1);
if (!isSplatValue(Op0) || !isSplatValue(Op1))
return false;
// Check getSplatValue early in this function, to avoid doing unnecessary
// work.
Value *ScalarOp0 = getSplatValue(Op0);
Value *ScalarOp1 = getSplatValue(Op1);
if (!ScalarOp0 || !ScalarOp1)
return false;
// For the binary VP intrinsics supported here, the result on disabled lanes
// is a poison value. For now, only do this simplification if all lanes
// are active.
// TODO: Relax the condition that all lanes are active by using insertelement
// on inactive lanes.
auto IsAllTrueMask = [](Value *MaskVal) {
if (Value *SplattedVal = getSplatValue(MaskVal))
if (auto *ConstValue = dyn_cast<Constant>(SplattedVal))
return ConstValue->isAllOnesValue();
return false;
};
if (!IsAllTrueMask(VPI.getArgOperand(2)))
return false;
// Check to make sure we support scalarization of the intrinsic
Intrinsic::ID IntrID = VPI.getIntrinsicID();
if (!VPBinOpIntrinsic::isVPBinOp(IntrID))
return false;
// Calculate cost of splatting both operands into vectors and the vector
// intrinsic
VectorType *VecTy = cast<VectorType>(VPI.getType());
SmallVector<int> Mask;
if (auto *FVTy = dyn_cast<FixedVectorType>(VecTy))
Mask.resize(FVTy->getNumElements(), 0);
InstructionCost SplatCost =
TTI.getVectorInstrCost(Instruction::InsertElement, VecTy, CostKind, 0) +
TTI.getShuffleCost(TargetTransformInfo::SK_Broadcast, VecTy, Mask,
CostKind);
// Calculate the cost of the VP Intrinsic
SmallVector<Type *, 4> Args;
for (Value *V : VPI.args())
Args.push_back(V->getType());
IntrinsicCostAttributes Attrs(IntrID, VecTy, Args);
InstructionCost VectorOpCost = TTI.getIntrinsicInstrCost(Attrs, CostKind);
InstructionCost OldCost = 2 * SplatCost + VectorOpCost;
// Determine scalar opcode
std::optional<unsigned> FunctionalOpcode =
VPI.getFunctionalOpcode();
std::optional<Intrinsic::ID> ScalarIntrID = std::nullopt;
if (!FunctionalOpcode) {
ScalarIntrID = VPI.getFunctionalIntrinsicID();
if (!ScalarIntrID)
return false;
}
// Calculate cost of scalarizing
InstructionCost ScalarOpCost = 0;
if (ScalarIntrID) {
IntrinsicCostAttributes Attrs(*ScalarIntrID, VecTy->getScalarType(), Args);
ScalarOpCost = TTI.getIntrinsicInstrCost(Attrs, CostKind);
} else {
ScalarOpCost = TTI.getArithmeticInstrCost(*FunctionalOpcode,
VecTy->getScalarType(), CostKind);
}
// The existing splats may be kept around if other instructions use them.
InstructionCost CostToKeepSplats =
(SplatCost * !Op0->hasOneUse()) + (SplatCost * !Op1->hasOneUse());
InstructionCost NewCost = ScalarOpCost + SplatCost + CostToKeepSplats;
LLVM_DEBUG(dbgs() << "Found a VP Intrinsic to scalarize: " << VPI
<< "\n");
LLVM_DEBUG(dbgs() << "Cost of Intrinsic: " << OldCost
<< ", Cost of scalarizing:" << NewCost << "\n");
// We want to scalarize unless the vector variant actually has lower cost.
if (OldCost < NewCost || !NewCost.isValid())
return false;
// Scalarize the intrinsic
ElementCount EC = cast<VectorType>(Op0->getType())->getElementCount();
Value *EVL = VPI.getArgOperand(3);
// If the VP op might introduce UB or poison, we can scalarize it provided
// that we know the EVL > 0: If the EVL is zero, then the original VP op
// becomes a no-op and thus won't be UB, so make sure we don't introduce UB by
// scalarizing it.
bool SafeToSpeculate;
if (ScalarIntrID)
SafeToSpeculate = Intrinsic::getFnAttributes(I.getContext(), *ScalarIntrID)
.hasAttribute(Attribute::AttrKind::Speculatable);
else
SafeToSpeculate = isSafeToSpeculativelyExecuteWithOpcode(
*FunctionalOpcode, &VPI, nullptr, &AC, &DT);
if (!SafeToSpeculate &&
!isKnownNonZero(EVL, SimplifyQuery(*DL, &DT, &AC, &VPI)))
return false;
Value *ScalarVal =
ScalarIntrID
? Builder.CreateIntrinsic(VecTy->getScalarType(), *ScalarIntrID,
{ScalarOp0, ScalarOp1})
: Builder.CreateBinOp((Instruction::BinaryOps)(*FunctionalOpcode),
ScalarOp0, ScalarOp1);
replaceValue(VPI, *Builder.CreateVectorSplat(EC, ScalarVal));
return true;
}
/// Match a vector op/compare/intrinsic with at least one
/// inserted scalar operand and convert to scalar op/cmp/intrinsic followed
/// by insertelement.
bool VectorCombine::scalarizeOpOrCmp(Instruction &I) {
auto *UO = dyn_cast<UnaryOperator>(&I);
auto *BO = dyn_cast<BinaryOperator>(&I);
auto *CI = dyn_cast<CmpInst>(&I);
auto *II = dyn_cast<IntrinsicInst>(&I);
if (!UO && !BO && !CI && !II)
return false;
// TODO: Allow intrinsics with different argument types
// TODO: Allow intrinsics with scalar arguments
if (II && (!isTriviallyVectorizable(II->getIntrinsicID()) ||
!all_of(II->args(), [&II](Value *Arg) {
return Arg->getType() == II->getType();
})))
return false;
// Do not convert the vector condition of a vector select into a scalar
// condition. That may cause problems for codegen because of differences in
// boolean formats and register-file transfers.
// TODO: Can we account for that in the cost model?
if (CI)
for (User *U : I.users())
if (match(U, m_Select(m_Specific(&I), m_Value(), m_Value())))
return false;
// Match constant vectors or scalars being inserted into constant vectors:
// vec_op [VecC0 | (inselt VecC0, V0, Index)], ...
SmallVector<Constant *> VecCs;
SmallVector<Value *> ScalarOps;
std::optional<uint64_t> Index;
auto Ops = II ? II->args() : I.operands();
for (Value *Op : Ops) {
Constant *VecC;
Value *V;
uint64_t InsIdx = 0;
VectorType *OpTy = cast<VectorType>(Op->getType());
if (match(Op, m_InsertElt(m_Constant(VecC), m_Value(V),
m_ConstantInt(InsIdx)))) {
// Bail if any inserts are out of bounds.
if (OpTy->getElementCount().getKnownMinValue() <= InsIdx)
return false;
// All inserts must have the same index.
// TODO: Deal with mismatched index constants and variable indexes?
if (!Index)
Index = InsIdx;
else if (InsIdx != *Index)
return false;
VecCs.push_back(VecC);
ScalarOps.push_back(V);
} else if (match(Op, m_Constant(VecC))) {
VecCs.push_back(VecC);
ScalarOps.push_back(nullptr);
} else {
return false;
}
}
// Bail if all operands are constant.
if (!Index.has_value())
return false;
VectorType *VecTy = cast<VectorType>(I.getType());
Type *ScalarTy = VecTy->getScalarType();
assert(VecTy->isVectorTy() &&
(ScalarTy->isIntegerTy() || ScalarTy->isFloatingPointTy() ||
ScalarTy->isPointerTy()) &&
"Unexpected types for insert element into binop or cmp");
unsigned Opcode = I.getOpcode();
InstructionCost ScalarOpCost, VectorOpCost;
if (CI) {
CmpInst::Predicate Pred = CI->getPredicate();
ScalarOpCost = TTI.getCmpSelInstrCost(
Opcode, ScalarTy, CmpInst::makeCmpResultType(ScalarTy), Pred, CostKind);
VectorOpCost = TTI.getCmpSelInstrCost(
Opcode, VecTy, CmpInst::makeCmpResultType(VecTy), Pred, CostKind);
} else if (UO || BO) {
ScalarOpCost = TTI.getArithmeticInstrCost(Opcode, ScalarTy, CostKind);
VectorOpCost = TTI.getArithmeticInstrCost(Opcode, VecTy, CostKind);
} else {
IntrinsicCostAttributes ScalarICA(
II->getIntrinsicID(), ScalarTy,
SmallVector<Type *>(II->arg_size(), ScalarTy));
ScalarOpCost = TTI.getIntrinsicInstrCost(ScalarICA, CostKind);
IntrinsicCostAttributes VectorICA(
II->getIntrinsicID(), VecTy,
SmallVector<Type *>(II->arg_size(), VecTy));
VectorOpCost = TTI.getIntrinsicInstrCost(VectorICA, CostKind);
}
// Fold the vector constants in the original vectors into a new base vector to
// get more accurate cost modelling.
Value *NewVecC = nullptr;
if (CI)
NewVecC = ConstantFoldCompareInstOperands(CI->getPredicate(), VecCs[0],
VecCs[1], *DL);
else if (UO)
NewVecC = ConstantFoldUnaryOpOperand(Opcode, VecCs[0], *DL);
else if (BO)
NewVecC = ConstantFoldBinaryOpOperands(Opcode, VecCs[0], VecCs[1], *DL);
else if (II->arg_size() == 2)
NewVecC = ConstantFoldBinaryIntrinsic(II->getIntrinsicID(), VecCs[0],
VecCs[1], II->getType(), II);
// Get cost estimate for the insert element. This cost will factor into
// both sequences.
InstructionCost OldCost = VectorOpCost;
InstructionCost NewCost =
ScalarOpCost + TTI.getVectorInstrCost(Instruction::InsertElement, VecTy,
CostKind, *Index, NewVecC);
for (auto [Op, VecC, Scalar] : zip(Ops, VecCs, ScalarOps)) {
if (!Scalar)
continue;
InstructionCost InsertCost = TTI.getVectorInstrCost(
Instruction::InsertElement, VecTy, CostKind, *Index, VecC, Scalar);
OldCost += InsertCost;
NewCost += !Op->hasOneUse() * InsertCost;
}
// We want to scalarize unless the vector variant actually has lower cost.
if (OldCost < NewCost || !NewCost.isValid())
return false;
// vec_op (inselt VecC0, V0, Index), (inselt VecC1, V1, Index) -->
// inselt NewVecC, (scalar_op V0, V1), Index
if (CI)
++NumScalarCmp;
else if (UO || BO)
++NumScalarOps;
else
++NumScalarIntrinsic;
// For constant cases, extract the scalar element, this should constant fold.
for (auto [OpIdx, Scalar, VecC] : enumerate(ScalarOps, VecCs))
if (!Scalar)
ScalarOps[OpIdx] = ConstantExpr::getExtractElement(
cast<Constant>(VecC), Builder.getInt64(*Index));
Value *Scalar;
if (CI)
Scalar = Builder.CreateCmp(CI->getPredicate(), ScalarOps[0], ScalarOps[1]);
else if (UO || BO)
Scalar = Builder.CreateNAryOp(Opcode, ScalarOps);
else
Scalar = Builder.CreateIntrinsic(ScalarTy, II->getIntrinsicID(), ScalarOps);
Scalar->setName(I.getName() + ".scalar");
// All IR flags are safe to back-propagate. There is no potential for extra
// poison to be created by the scalar instruction.
if (auto *ScalarInst = dyn_cast<Instruction>(Scalar))
ScalarInst->copyIRFlags(&I);
// Create a new base vector if the constant folding failed.
if (!NewVecC) {
SmallVector<Value *> VecCValues;
VecCValues.reserve(VecCs.size());
append_range(VecCValues, VecCs);
if (CI)
NewVecC = Builder.CreateCmp(CI->getPredicate(), VecCs[0], VecCs[1]);
else if (UO || BO)
NewVecC = Builder.CreateNAryOp(Opcode, VecCValues);
else
NewVecC =
Builder.CreateIntrinsic(VecTy, II->getIntrinsicID(), VecCValues);
}
Value *Insert = Builder.CreateInsertElement(NewVecC, Scalar, *Index);
replaceValue(I, *Insert);
return true;
}
/// Try to combine a scalar binop + 2 scalar compares of extracted elements of
/// a vector into vector operations followed by extract. Note: The SLP pass
/// may miss this pattern because of implementation problems.
bool VectorCombine::foldExtractedCmps(Instruction &I) {
auto *BI = dyn_cast<BinaryOperator>(&I);
// We are looking for a scalar binop of booleans.
// binop i1 (cmp Pred I0, C0), (cmp Pred I1, C1)
if (!BI || !I.getType()->isIntegerTy(1))
return false;
// The compare predicates should match, and each compare should have a
// constant operand.
Value *B0 = I.getOperand(0), *B1 = I.getOperand(1);
Instruction *I0, *I1;
Constant *C0, *C1;
CmpPredicate P0, P1;
if (!match(B0, m_Cmp(P0, m_Instruction(I0), m_Constant(C0))) ||
!match(B1, m_Cmp(P1, m_Instruction(I1), m_Constant(C1))))
return false;
auto MatchingPred = CmpPredicate::getMatching(P0, P1);
if (!MatchingPred)
return false;
// The compare operands must be extracts of the same vector with constant
// extract indexes.
Value *X;
uint64_t Index0, Index1;
if (!match(I0, m_ExtractElt(m_Value(X), m_ConstantInt(Index0))) ||
!match(I1, m_ExtractElt(m_Specific(X), m_ConstantInt(Index1))))
return false;
auto *Ext0 = cast<ExtractElementInst>(I0);
auto *Ext1 = cast<ExtractElementInst>(I1);
ExtractElementInst *ConvertToShuf = getShuffleExtract(Ext0, Ext1, CostKind);
if (!ConvertToShuf)
return false;
assert((ConvertToShuf == Ext0 || ConvertToShuf == Ext1) &&
"Unknown ExtractElementInst");
// The original scalar pattern is:
// binop i1 (cmp Pred (ext X, Index0), C0), (cmp Pred (ext X, Index1), C1)
CmpInst::Predicate Pred = *MatchingPred;
unsigned CmpOpcode =
CmpInst::isFPPredicate(Pred) ? Instruction::FCmp : Instruction::ICmp;
auto *VecTy = dyn_cast<FixedVectorType>(X->getType());
if (!VecTy)
return false;
InstructionCost Ext0Cost =
TTI.getVectorInstrCost(*Ext0, VecTy, CostKind, Index0);
InstructionCost Ext1Cost =
TTI.getVectorInstrCost(*Ext1, VecTy, CostKind, Index1);
InstructionCost CmpCost = TTI.getCmpSelInstrCost(
CmpOpcode, I0->getType(), CmpInst::makeCmpResultType(I0->getType()), Pred,
CostKind);
InstructionCost OldCost =
Ext0Cost + Ext1Cost + CmpCost * 2 +
TTI.getArithmeticInstrCost(I.getOpcode(), I.getType(), CostKind);
// The proposed vector pattern is:
// vcmp = cmp Pred X, VecC
// ext (binop vNi1 vcmp, (shuffle vcmp, Index1)), Index0
int CheapIndex = ConvertToShuf == Ext0 ? Index1 : Index0;
int ExpensiveIndex = ConvertToShuf == Ext0 ? Index0 : Index1;
auto *CmpTy = cast<FixedVectorType>(CmpInst::makeCmpResultType(X->getType()));
InstructionCost NewCost = TTI.getCmpSelInstrCost(
CmpOpcode, X->getType(), CmpInst::makeCmpResultType(X->getType()), Pred,
CostKind);
SmallVector<int, 32> ShufMask(VecTy->getNumElements(), PoisonMaskElem);
ShufMask[CheapIndex] = ExpensiveIndex;
NewCost += TTI.getShuffleCost(TargetTransformInfo::SK_PermuteSingleSrc, CmpTy,
ShufMask, CostKind);
NewCost += TTI.getArithmeticInstrCost(I.getOpcode(), CmpTy, CostKind);
NewCost += TTI.getVectorInstrCost(*Ext0, CmpTy, CostKind, CheapIndex);
NewCost += Ext0->hasOneUse() ? 0 : Ext0Cost;
NewCost += Ext1->hasOneUse() ? 0 : Ext1Cost;
// Aggressively form vector ops if the cost is equal because the transform
// may enable further optimization.
// Codegen can reverse this transform (scalarize) if it was not profitable.
if (OldCost < NewCost || !NewCost.isValid())
return false;
// Create a vector constant from the 2 scalar constants.
SmallVector<Constant *, 32> CmpC(VecTy->getNumElements(),
PoisonValue::get(VecTy->getElementType()));
CmpC[Index0] = C0;
CmpC[Index1] = C1;
Value *VCmp = Builder.CreateCmp(Pred, X, ConstantVector::get(CmpC));
Value *Shuf = createShiftShuffle(VCmp, ExpensiveIndex, CheapIndex, Builder);
Value *LHS = ConvertToShuf == Ext0 ? Shuf : VCmp;
Value *RHS = ConvertToShuf == Ext0 ? VCmp : Shuf;
Value *VecLogic = Builder.CreateBinOp(BI->getOpcode(), LHS, RHS);
Value *NewExt = Builder.CreateExtractElement(VecLogic, CheapIndex);
replaceValue(I, *NewExt);
++NumVecCmpBO;
return true;
}
static void analyzeCostOfVecReduction(const IntrinsicInst &II,
TTI::TargetCostKind CostKind,
const TargetTransformInfo &TTI,
InstructionCost &CostBeforeReduction,
InstructionCost &CostAfterReduction) {
Instruction *Op0, *Op1;
auto *RedOp = dyn_cast<Instruction>(II.getOperand(0));
auto *VecRedTy = cast<VectorType>(II.getOperand(0)->getType());
unsigned ReductionOpc =
getArithmeticReductionInstruction(II.getIntrinsicID());
if (RedOp && match(RedOp, m_ZExtOrSExt(m_Value()))) {
bool IsUnsigned = isa<ZExtInst>(RedOp);
auto *ExtType = cast<VectorType>(RedOp->getOperand(0)->getType());
CostBeforeReduction =
TTI.getCastInstrCost(RedOp->getOpcode(), VecRedTy, ExtType,
TTI::CastContextHint::None, CostKind, RedOp);
CostAfterReduction =
TTI.getExtendedReductionCost(ReductionOpc, IsUnsigned, II.getType(),
ExtType, FastMathFlags(), CostKind);
return;
}
if (RedOp && II.getIntrinsicID() == Intrinsic::vector_reduce_add &&
match(RedOp,
m_ZExtOrSExt(m_Mul(m_Instruction(Op0), m_Instruction(Op1)))) &&
match(Op0, m_ZExtOrSExt(m_Value())) &&
Op0->getOpcode() == Op1->getOpcode() &&
Op0->getOperand(0)->getType() == Op1->getOperand(0)->getType() &&
(Op0->getOpcode() == RedOp->getOpcode() || Op0 == Op1)) {
// Matched reduce.add(ext(mul(ext(A), ext(B)))
bool IsUnsigned = isa<ZExtInst>(Op0);
auto *ExtType = cast<VectorType>(Op0->getOperand(0)->getType());
VectorType *MulType = VectorType::get(Op0->getType(), VecRedTy);
InstructionCost ExtCost =
TTI.getCastInstrCost(Op0->getOpcode(), MulType, ExtType,
TTI::CastContextHint::None, CostKind, Op0);
InstructionCost MulCost =
TTI.getArithmeticInstrCost(Instruction::Mul, MulType, CostKind);
InstructionCost Ext2Cost =
TTI.getCastInstrCost(RedOp->getOpcode(), VecRedTy, MulType,
TTI::CastContextHint::None, CostKind, RedOp);
CostBeforeReduction = ExtCost * 2 + MulCost + Ext2Cost;
CostAfterReduction =
TTI.getMulAccReductionCost(IsUnsigned, II.getType(), ExtType, CostKind);
return;
}
CostAfterReduction = TTI.getArithmeticReductionCost(ReductionOpc, VecRedTy,
std::nullopt, CostKind);
}
bool VectorCombine::foldBinopOfReductions(Instruction &I) {
Instruction::BinaryOps BinOpOpc = cast<BinaryOperator>(&I)->getOpcode();
Intrinsic::ID ReductionIID = getReductionForBinop(BinOpOpc);
if (BinOpOpc == Instruction::Sub)
ReductionIID = Intrinsic::vector_reduce_add;
if (ReductionIID == Intrinsic::not_intrinsic)
return false;
auto checkIntrinsicAndGetItsArgument = [](Value *V,
Intrinsic::ID IID) -> Value * {
auto *II = dyn_cast<IntrinsicInst>(V);
if (!II)
return nullptr;
if (II->getIntrinsicID() == IID && II->hasOneUse())
return II->getArgOperand(0);
return nullptr;
};
Value *V0 = checkIntrinsicAndGetItsArgument(I.getOperand(0), ReductionIID);
if (!V0)
return false;
Value *V1 = checkIntrinsicAndGetItsArgument(I.getOperand(1), ReductionIID);
if (!V1)
return false;
auto *VTy = cast<VectorType>(V0->getType());
if (V1->getType() != VTy)
return false;
const auto &II0 = *cast<IntrinsicInst>(I.getOperand(0));
const auto &II1 = *cast<IntrinsicInst>(I.getOperand(1));
unsigned ReductionOpc =
getArithmeticReductionInstruction(II0.getIntrinsicID());
InstructionCost OldCost = 0;
InstructionCost NewCost = 0;
InstructionCost CostOfRedOperand0 = 0;
InstructionCost CostOfRed0 = 0;
InstructionCost CostOfRedOperand1 = 0;
InstructionCost CostOfRed1 = 0;
analyzeCostOfVecReduction(II0, CostKind, TTI, CostOfRedOperand0, CostOfRed0);
analyzeCostOfVecReduction(II1, CostKind, TTI, CostOfRedOperand1, CostOfRed1);
OldCost = CostOfRed0 + CostOfRed1 + TTI.getInstructionCost(&I, CostKind);
NewCost =
CostOfRedOperand0 + CostOfRedOperand1 +
TTI.getArithmeticInstrCost(BinOpOpc, VTy, CostKind) +
TTI.getArithmeticReductionCost(ReductionOpc, VTy, std::nullopt, CostKind);
if (NewCost >= OldCost || !NewCost.isValid())
return false;
LLVM_DEBUG(dbgs() << "Found two mergeable reductions: " << I
<< "\n OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
Value *VectorBO = Builder.CreateBinOp(BinOpOpc, V0, V1);
if (auto *PDInst = dyn_cast<PossiblyDisjointInst>(&I))
if (auto *PDVectorBO = dyn_cast<PossiblyDisjointInst>(VectorBO))
PDVectorBO->setIsDisjoint(PDInst->isDisjoint());
Instruction *Rdx = Builder.CreateIntrinsic(ReductionIID, {VTy}, {VectorBO});
replaceValue(I, *Rdx);
return true;
}
// Check if memory loc modified between two instrs in the same BB
static bool isMemModifiedBetween(BasicBlock::iterator Begin,
BasicBlock::iterator End,
const MemoryLocation &Loc, AAResults &AA) {
unsigned NumScanned = 0;
return std::any_of(Begin, End, [&](const Instruction &Instr) {
return isModSet(AA.getModRefInfo(&Instr, Loc)) ||
++NumScanned > MaxInstrsToScan;
});
}
namespace {
/// Helper class to indicate whether a vector index can be safely scalarized and
/// if a freeze needs to be inserted.
class ScalarizationResult {
enum class StatusTy { Unsafe, Safe, SafeWithFreeze };
StatusTy Status;
Value *ToFreeze;
ScalarizationResult(StatusTy Status, Value *ToFreeze = nullptr)
: Status(Status), ToFreeze(ToFreeze) {}
public:
ScalarizationResult(const ScalarizationResult &Other) = default;
~ScalarizationResult() {
assert(!ToFreeze && "freeze() not called with ToFreeze being set");
}
static ScalarizationResult unsafe() { return {StatusTy::Unsafe}; }
static ScalarizationResult safe() { return {StatusTy::Safe}; }
static ScalarizationResult safeWithFreeze(Value *ToFreeze) {
return {StatusTy::SafeWithFreeze, ToFreeze};
}
/// Returns true if the index can be scalarize without requiring a freeze.
bool isSafe() const { return Status == StatusTy::Safe; }
/// Returns true if the index cannot be scalarized.
bool isUnsafe() const { return Status == StatusTy::Unsafe; }
/// Returns true if the index can be scalarize, but requires inserting a
/// freeze.
bool isSafeWithFreeze() const { return Status == StatusTy::SafeWithFreeze; }
/// Reset the state of Unsafe and clear ToFreze if set.
void discard() {
ToFreeze = nullptr;
Status = StatusTy::Unsafe;
}
/// Freeze the ToFreeze and update the use in \p User to use it.
void freeze(IRBuilder<> &Builder, Instruction &UserI) {
assert(isSafeWithFreeze() &&
"should only be used when freezing is required");
assert(is_contained(ToFreeze->users(), &UserI) &&
"UserI must be a user of ToFreeze");
IRBuilder<>::InsertPointGuard Guard(Builder);
Builder.SetInsertPoint(cast<Instruction>(&UserI));
Value *Frozen =
Builder.CreateFreeze(ToFreeze, ToFreeze->getName() + ".frozen");
for (Use &U : make_early_inc_range((UserI.operands())))
if (U.get() == ToFreeze)
U.set(Frozen);
ToFreeze = nullptr;
}
};
} // namespace
/// Check if it is legal to scalarize a memory access to \p VecTy at index \p
/// Idx. \p Idx must access a valid vector element.
static ScalarizationResult canScalarizeAccess(VectorType *VecTy, Value *Idx,
Instruction *CtxI,
AssumptionCache &AC,
const DominatorTree &DT) {
// We do checks for both fixed vector types and scalable vector types.
// This is the number of elements of fixed vector types,
// or the minimum number of elements of scalable vector types.
uint64_t NumElements = VecTy->getElementCount().getKnownMinValue();
unsigned IntWidth = Idx->getType()->getScalarSizeInBits();
if (auto *C = dyn_cast<ConstantInt>(Idx)) {
if (C->getValue().ult(NumElements))
return ScalarizationResult::safe();
return ScalarizationResult::unsafe();
}
// Always unsafe if the index type can't handle all inbound values.
if (!llvm::isUIntN(IntWidth, NumElements))
return ScalarizationResult::unsafe();
APInt Zero(IntWidth, 0);
APInt MaxElts(IntWidth, NumElements);
ConstantRange ValidIndices(Zero, MaxElts);
ConstantRange IdxRange(IntWidth, true);
if (isGuaranteedNotToBePoison(Idx, &AC)) {
if (ValidIndices.contains(computeConstantRange(Idx, /* ForSigned */ false,
true, &AC, CtxI, &DT)))
return ScalarizationResult::safe();
return ScalarizationResult::unsafe();
}
// If the index may be poison, check if we can insert a freeze before the
// range of the index is restricted.
Value *IdxBase;
ConstantInt *CI;
if (match(Idx, m_And(m_Value(IdxBase), m_ConstantInt(CI)))) {
IdxRange = IdxRange.binaryAnd(CI->getValue());
} else if (match(Idx, m_URem(m_Value(IdxBase), m_ConstantInt(CI)))) {
IdxRange = IdxRange.urem(CI->getValue());
}
if (ValidIndices.contains(IdxRange))
return ScalarizationResult::safeWithFreeze(IdxBase);
return ScalarizationResult::unsafe();
}
/// The memory operation on a vector of \p ScalarType had alignment of
/// \p VectorAlignment. Compute the maximal, but conservatively correct,
/// alignment that will be valid for the memory operation on a single scalar
/// element of the same type with index \p Idx.
static Align computeAlignmentAfterScalarization(Align VectorAlignment,
Type *ScalarType, Value *Idx,
const DataLayout &DL) {
if (auto *C = dyn_cast<ConstantInt>(Idx))
return commonAlignment(VectorAlignment,
C->getZExtValue() * DL.getTypeStoreSize(ScalarType));
return commonAlignment(VectorAlignment, DL.getTypeStoreSize(ScalarType));
}
// Combine patterns like:
// %0 = load <4 x i32>, <4 x i32>* %a
// %1 = insertelement <4 x i32> %0, i32 %b, i32 1
// store <4 x i32> %1, <4 x i32>* %a
// to:
// %0 = bitcast <4 x i32>* %a to i32*
// %1 = getelementptr inbounds i32, i32* %0, i64 0, i64 1
// store i32 %b, i32* %1
bool VectorCombine::foldSingleElementStore(Instruction &I) {
auto *SI = cast<StoreInst>(&I);
if (!SI->isSimple() || !isa<VectorType>(SI->getValueOperand()->getType()))
return false;
// TODO: Combine more complicated patterns (multiple insert) by referencing
// TargetTransformInfo.
Instruction *Source;
Value *NewElement;
Value *Idx;
if (!match(SI->getValueOperand(),
m_InsertElt(m_Instruction(Source), m_Value(NewElement),
m_Value(Idx))))
return false;
if (auto *Load = dyn_cast<LoadInst>(Source)) {
auto VecTy = cast<VectorType>(SI->getValueOperand()->getType());
Value *SrcAddr = Load->getPointerOperand()->stripPointerCasts();
// Don't optimize for atomic/volatile load or store. Ensure memory is not
// modified between, vector type matches store size, and index is inbounds.
if (!Load->isSimple() || Load->getParent() != SI->getParent() ||
!DL->typeSizeEqualsStoreSize(Load->getType()->getScalarType()) ||
SrcAddr != SI->getPointerOperand()->stripPointerCasts())
return false;
auto ScalarizableIdx = canScalarizeAccess(VecTy, Idx, Load, AC, DT);
if (ScalarizableIdx.isUnsafe() ||
isMemModifiedBetween(Load->getIterator(), SI->getIterator(),
MemoryLocation::get(SI), AA))
return false;
// Ensure we add the load back to the worklist BEFORE its users so they can
// erased in the correct order.
Worklist.push(Load);
if (ScalarizableIdx.isSafeWithFreeze())
ScalarizableIdx.freeze(Builder, *cast<Instruction>(Idx));
Value *GEP = Builder.CreateInBoundsGEP(
SI->getValueOperand()->getType(), SI->getPointerOperand(),
{ConstantInt::get(Idx->getType(), 0), Idx});
StoreInst *NSI = Builder.CreateStore(NewElement, GEP);
NSI->copyMetadata(*SI);
Align ScalarOpAlignment = computeAlignmentAfterScalarization(
std::max(SI->getAlign(), Load->getAlign()), NewElement->getType(), Idx,
*DL);
NSI->setAlignment(ScalarOpAlignment);
replaceValue(I, *NSI);
eraseInstruction(I);
return true;
}
return false;
}
/// Try to scalarize vector loads feeding extractelement instructions.
bool VectorCombine::scalarizeLoadExtract(Instruction &I) {
Value *Ptr;
if (!match(&I, m_Load(m_Value(Ptr))))
return false;
auto *LI = cast<LoadInst>(&I);
auto *VecTy = cast<VectorType>(LI->getType());
if (LI->isVolatile() || !DL->typeSizeEqualsStoreSize(VecTy->getScalarType()))
return false;
InstructionCost OriginalCost =
TTI.getMemoryOpCost(Instruction::Load, VecTy, LI->getAlign(),
LI->getPointerAddressSpace(), CostKind);
InstructionCost ScalarizedCost = 0;
Instruction *LastCheckedInst = LI;
unsigned NumInstChecked = 0;
DenseMap<ExtractElementInst *, ScalarizationResult> NeedFreeze;
auto FailureGuard = make_scope_exit([&]() {
// If the transform is aborted, discard the ScalarizationResults.
for (auto &Pair : NeedFreeze)
Pair.second.discard();
});
// Check if all users of the load are extracts with no memory modifications
// between the load and the extract. Compute the cost of both the original
// code and the scalarized version.
for (User *U : LI->users()) {
auto *UI = dyn_cast<ExtractElementInst>(U);
if (!UI || UI->getParent() != LI->getParent())
return false;
// If any extract is waiting to be erased, then bail out as this will
// distort the cost calculation and possibly lead to infinite loops.
if (UI->use_empty())
return false;
// Check if any instruction between the load and the extract may modify
// memory.
if (LastCheckedInst->comesBefore(UI)) {
for (Instruction &I :
make_range(std::next(LI->getIterator()), UI->getIterator())) {
// Bail out if we reached the check limit or the instruction may write
// to memory.
if (NumInstChecked == MaxInstrsToScan || I.mayWriteToMemory())
return false;
NumInstChecked++;
}
LastCheckedInst = UI;
}
auto ScalarIdx =
canScalarizeAccess(VecTy, UI->getIndexOperand(), LI, AC, DT);
if (ScalarIdx.isUnsafe())
return false;
if (ScalarIdx.isSafeWithFreeze()) {
NeedFreeze.try_emplace(UI, ScalarIdx);
ScalarIdx.discard();
}
auto *Index = dyn_cast<ConstantInt>(UI->getIndexOperand());
OriginalCost +=
TTI.getVectorInstrCost(Instruction::ExtractElement, VecTy, CostKind,
Index ? Index->getZExtValue() : -1);
ScalarizedCost +=
TTI.getMemoryOpCost(Instruction::Load, VecTy->getElementType(),
Align(1), LI->getPointerAddressSpace(), CostKind);
ScalarizedCost += TTI.getAddressComputationCost(VecTy->getElementType());
}
LLVM_DEBUG(dbgs() << "Found all extractions of a vector load: " << I
<< "\n LoadExtractCost: " << OriginalCost
<< " vs ScalarizedCost: " << ScalarizedCost << "\n");
if (ScalarizedCost >= OriginalCost)
return false;
// Ensure we add the load back to the worklist BEFORE its users so they can
// erased in the correct order.
Worklist.push(LI);
// Replace extracts with narrow scalar loads.
for (User *U : LI->users()) {
auto *EI = cast<ExtractElementInst>(U);
Value *Idx = EI->getIndexOperand();
// Insert 'freeze' for poison indexes.
auto It = NeedFreeze.find(EI);
if (It != NeedFreeze.end())
It->second.freeze(Builder, *cast<Instruction>(Idx));
Builder.SetInsertPoint(EI);
Value *GEP =
Builder.CreateInBoundsGEP(VecTy, Ptr, {Builder.getInt32(0), Idx});
auto *NewLoad = cast<LoadInst>(Builder.CreateLoad(
VecTy->getElementType(), GEP, EI->getName() + ".scalar"));
Align ScalarOpAlignment = computeAlignmentAfterScalarization(
LI->getAlign(), VecTy->getElementType(), Idx, *DL);
NewLoad->setAlignment(ScalarOpAlignment);
replaceValue(*EI, *NewLoad);
}
FailureGuard.release();
return true;
}
/// Try to fold "(or (zext (bitcast X)), (shl (zext (bitcast Y)), C))"
/// to "(bitcast (concat X, Y))"
/// where X/Y are bitcasted from i1 mask vectors.
bool VectorCombine::foldConcatOfBoolMasks(Instruction &I) {
Type *Ty = I.getType();
if (!Ty->isIntegerTy())
return false;
// TODO: Add big endian test coverage
if (DL->isBigEndian())
return false;
// Restrict to disjoint cases so the mask vectors aren't overlapping.
Instruction *X, *Y;
if (!match(&I, m_DisjointOr(m_Instruction(X), m_Instruction(Y))))
return false;
// Allow both sources to contain shl, to handle more generic pattern:
// "(or (shl (zext (bitcast X)), C1), (shl (zext (bitcast Y)), C2))"
Value *SrcX;
uint64_t ShAmtX = 0;
if (!match(X, m_OneUse(m_ZExt(m_OneUse(m_BitCast(m_Value(SrcX)))))) &&
!match(X, m_OneUse(
m_Shl(m_OneUse(m_ZExt(m_OneUse(m_BitCast(m_Value(SrcX))))),
m_ConstantInt(ShAmtX)))))
return false;
Value *SrcY;
uint64_t ShAmtY = 0;
if (!match(Y, m_OneUse(m_ZExt(m_OneUse(m_BitCast(m_Value(SrcY)))))) &&
!match(Y, m_OneUse(
m_Shl(m_OneUse(m_ZExt(m_OneUse(m_BitCast(m_Value(SrcY))))),
m_ConstantInt(ShAmtY)))))
return false;
// Canonicalize larger shift to the RHS.
if (ShAmtX > ShAmtY) {
std::swap(X, Y);
std::swap(SrcX, SrcY);
std::swap(ShAmtX, ShAmtY);
}
// Ensure both sources are matching vXi1 bool mask types, and that the shift
// difference is the mask width so they can be easily concatenated together.
uint64_t ShAmtDiff = ShAmtY - ShAmtX;
unsigned NumSHL = (ShAmtX > 0) + (ShAmtY > 0);
unsigned BitWidth = Ty->getPrimitiveSizeInBits();
auto *MaskTy = dyn_cast<FixedVectorType>(SrcX->getType());
if (!MaskTy || SrcX->getType() != SrcY->getType() ||
!MaskTy->getElementType()->isIntegerTy(1) ||
MaskTy->getNumElements() != ShAmtDiff ||
MaskTy->getNumElements() > (BitWidth / 2))
return false;
auto *ConcatTy = FixedVectorType::getDoubleElementsVectorType(MaskTy);
auto *ConcatIntTy =
Type::getIntNTy(Ty->getContext(), ConcatTy->getNumElements());
auto *MaskIntTy = Type::getIntNTy(Ty->getContext(), ShAmtDiff);
SmallVector<int, 32> ConcatMask(ConcatTy->getNumElements());
std::iota(ConcatMask.begin(), ConcatMask.end(), 0);
// TODO: Is it worth supporting multi use cases?
InstructionCost OldCost = 0;
OldCost += TTI.getArithmeticInstrCost(Instruction::Or, Ty, CostKind);
OldCost +=
NumSHL * TTI.getArithmeticInstrCost(Instruction::Shl, Ty, CostKind);
OldCost += 2 * TTI.getCastInstrCost(Instruction::ZExt, Ty, MaskIntTy,
TTI::CastContextHint::None, CostKind);
OldCost += 2 * TTI.getCastInstrCost(Instruction::BitCast, MaskIntTy, MaskTy,
TTI::CastContextHint::None, CostKind);
InstructionCost NewCost = 0;
NewCost += TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, MaskTy,
ConcatMask, CostKind);
NewCost += TTI.getCastInstrCost(Instruction::BitCast, ConcatIntTy, ConcatTy,
TTI::CastContextHint::None, CostKind);
if (Ty != ConcatIntTy)
NewCost += TTI.getCastInstrCost(Instruction::ZExt, Ty, ConcatIntTy,
TTI::CastContextHint::None, CostKind);
if (ShAmtX > 0)
NewCost += TTI.getArithmeticInstrCost(Instruction::Shl, Ty, CostKind);
LLVM_DEBUG(dbgs() << "Found a concatenation of bitcasted bool masks: " << I
<< "\n OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
if (NewCost > OldCost)
return false;
// Build bool mask concatenation, bitcast back to scalar integer, and perform
// any residual zero-extension or shifting.
Value *Concat = Builder.CreateShuffleVector(SrcX, SrcY, ConcatMask);
Worklist.pushValue(Concat);
Value *Result = Builder.CreateBitCast(Concat, ConcatIntTy);
if (Ty != ConcatIntTy) {
Worklist.pushValue(Result);
Result = Builder.CreateZExt(Result, Ty);
}
if (ShAmtX > 0) {
Worklist.pushValue(Result);
Result = Builder.CreateShl(Result, ShAmtX);
}
replaceValue(I, *Result);
return true;
}
/// Try to convert "shuffle (binop (shuffle, shuffle)), undef"
/// --> "binop (shuffle), (shuffle)".
bool VectorCombine::foldPermuteOfBinops(Instruction &I) {
BinaryOperator *BinOp;
ArrayRef<int> OuterMask;
if (!match(&I,
m_Shuffle(m_OneUse(m_BinOp(BinOp)), m_Undef(), m_Mask(OuterMask))))
return false;
// Don't introduce poison into div/rem.
if (BinOp->isIntDivRem() && llvm::is_contained(OuterMask, PoisonMaskElem))
return false;
Value *Op00, *Op01, *Op10, *Op11;
ArrayRef<int> Mask0, Mask1;
bool Match0 =
match(BinOp->getOperand(0),
m_OneUse(m_Shuffle(m_Value(Op00), m_Value(Op01), m_Mask(Mask0))));
bool Match1 =
match(BinOp->getOperand(1),
m_OneUse(m_Shuffle(m_Value(Op10), m_Value(Op11), m_Mask(Mask1))));
if (!Match0 && !Match1)
return false;
Op00 = Match0 ? Op00 : BinOp->getOperand(0);
Op01 = Match0 ? Op01 : BinOp->getOperand(0);
Op10 = Match1 ? Op10 : BinOp->getOperand(1);
Op11 = Match1 ? Op11 : BinOp->getOperand(1);
Instruction::BinaryOps Opcode = BinOp->getOpcode();
auto *ShuffleDstTy = dyn_cast<FixedVectorType>(I.getType());
auto *BinOpTy = dyn_cast<FixedVectorType>(BinOp->getType());
auto *Op0Ty = dyn_cast<FixedVectorType>(Op00->getType());
auto *Op1Ty = dyn_cast<FixedVectorType>(Op10->getType());
if (!ShuffleDstTy || !BinOpTy || !Op0Ty || !Op1Ty)
return false;
unsigned NumSrcElts = BinOpTy->getNumElements();
// Don't accept shuffles that reference the second operand in
// div/rem or if its an undef arg.
if ((BinOp->isIntDivRem() || !isa<PoisonValue>(I.getOperand(1))) &&
any_of(OuterMask, [NumSrcElts](int M) { return M >= (int)NumSrcElts; }))
return false;
// Merge outer / inner (or identity if no match) shuffles.
SmallVector<int> NewMask0, NewMask1;
for (int M : OuterMask) {
if (M < 0 || M >= (int)NumSrcElts) {
NewMask0.push_back(PoisonMaskElem);
NewMask1.push_back(PoisonMaskElem);
} else {
NewMask0.push_back(Match0 ? Mask0[M] : M);
NewMask1.push_back(Match1 ? Mask1[M] : M);
}
}
unsigned NumOpElts = Op0Ty->getNumElements();
bool IsIdentity0 = ShuffleDstTy == Op0Ty &&
all_of(NewMask0, [NumOpElts](int M) { return M < (int)NumOpElts; }) &&
ShuffleVectorInst::isIdentityMask(NewMask0, NumOpElts);
bool IsIdentity1 = ShuffleDstTy == Op1Ty &&
all_of(NewMask1, [NumOpElts](int M) { return M < (int)NumOpElts; }) &&
ShuffleVectorInst::isIdentityMask(NewMask1, NumOpElts);
// Try to merge shuffles across the binop if the new shuffles are not costly.
InstructionCost OldCost =
TTI.getArithmeticInstrCost(Opcode, BinOpTy, CostKind) +
TTI.getShuffleCost(TargetTransformInfo::SK_PermuteSingleSrc, BinOpTy,
OuterMask, CostKind, 0, nullptr, {BinOp}, &I);
if (Match0)
OldCost += TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, Op0Ty,
Mask0, CostKind, 0, nullptr, {Op00, Op01},
cast<Instruction>(BinOp->getOperand(0)));
if (Match1)
OldCost += TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, Op1Ty,
Mask1, CostKind, 0, nullptr, {Op10, Op11},
cast<Instruction>(BinOp->getOperand(1)));
InstructionCost NewCost =
TTI.getArithmeticInstrCost(Opcode, ShuffleDstTy, CostKind);
if (!IsIdentity0)
NewCost += TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, Op0Ty,
NewMask0, CostKind, 0, nullptr, {Op00, Op01});
if (!IsIdentity1)
NewCost += TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, Op1Ty,
NewMask1, CostKind, 0, nullptr, {Op10, Op11});
LLVM_DEBUG(dbgs() << "Found a shuffle feeding a shuffled binop: " << I
<< "\n OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
// If costs are equal, still fold as we reduce instruction count.
if (NewCost > OldCost)
return false;
Value *LHS =
IsIdentity0 ? Op00 : Builder.CreateShuffleVector(Op00, Op01, NewMask0);
Value *RHS =
IsIdentity1 ? Op10 : Builder.CreateShuffleVector(Op10, Op11, NewMask1);
Value *NewBO = Builder.CreateBinOp(Opcode, LHS, RHS);
// Intersect flags from the old binops.
if (auto *NewInst = dyn_cast<Instruction>(NewBO))
NewInst->copyIRFlags(BinOp);
Worklist.pushValue(LHS);
Worklist.pushValue(RHS);
replaceValue(I, *NewBO);
return true;
}
/// Try to convert "shuffle (binop), (binop)" into "binop (shuffle), (shuffle)".
/// Try to convert "shuffle (cmpop), (cmpop)" into "cmpop (shuffle), (shuffle)".
bool VectorCombine::foldShuffleOfBinops(Instruction &I) {
ArrayRef<int> OldMask;
Instruction *LHS, *RHS;
if (!match(&I, m_Shuffle(m_OneUse(m_Instruction(LHS)),
m_OneUse(m_Instruction(RHS)), m_Mask(OldMask))))
return false;
// TODO: Add support for addlike etc.
if (LHS->getOpcode() != RHS->getOpcode())
return false;
Value *X, *Y, *Z, *W;
bool IsCommutative = false;
CmpPredicate PredLHS = CmpInst::BAD_ICMP_PREDICATE;
CmpPredicate PredRHS = CmpInst::BAD_ICMP_PREDICATE;
if (match(LHS, m_BinOp(m_Value(X), m_Value(Y))) &&
match(RHS, m_BinOp(m_Value(Z), m_Value(W)))) {
auto *BO = cast<BinaryOperator>(LHS);
// Don't introduce poison into div/rem.
if (llvm::is_contained(OldMask, PoisonMaskElem) && BO->isIntDivRem())
return false;
IsCommutative = BinaryOperator::isCommutative(BO->getOpcode());
} else if (match(LHS, m_Cmp(PredLHS, m_Value(X), m_Value(Y))) &&
match(RHS, m_Cmp(PredRHS, m_Value(Z), m_Value(W))) &&
(CmpInst::Predicate)PredLHS == (CmpInst::Predicate)PredRHS) {
IsCommutative = cast<CmpInst>(LHS)->isCommutative();
} else
return false;
auto *ShuffleDstTy = dyn_cast<FixedVectorType>(I.getType());
auto *BinResTy = dyn_cast<FixedVectorType>(LHS->getType());
auto *BinOpTy = dyn_cast<FixedVectorType>(X->getType());
if (!ShuffleDstTy || !BinResTy || !BinOpTy || X->getType() != Z->getType())
return false;
unsigned NumSrcElts = BinOpTy->getNumElements();
// If we have something like "add X, Y" and "add Z, X", swap ops to match.
if (IsCommutative && X != Z && Y != W && (X == W || Y == Z))
std::swap(X, Y);
auto ConvertToUnary = [NumSrcElts](int &M) {
if (M >= (int)NumSrcElts)
M -= NumSrcElts;
};
SmallVector<int> NewMask0(OldMask);
TargetTransformInfo::ShuffleKind SK0 = TargetTransformInfo::SK_PermuteTwoSrc;
if (X == Z) {
llvm::for_each(NewMask0, ConvertToUnary);
SK0 = TargetTransformInfo::SK_PermuteSingleSrc;
Z = PoisonValue::get(BinOpTy);
}
SmallVector<int> NewMask1(OldMask);
TargetTransformInfo::ShuffleKind SK1 = TargetTransformInfo::SK_PermuteTwoSrc;
if (Y == W) {
llvm::for_each(NewMask1, ConvertToUnary);
SK1 = TargetTransformInfo::SK_PermuteSingleSrc;
W = PoisonValue::get(BinOpTy);
}
// Try to replace a binop with a shuffle if the shuffle is not costly.
InstructionCost OldCost =
TTI.getInstructionCost(LHS, CostKind) +
TTI.getInstructionCost(RHS, CostKind) +
TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, BinResTy,
OldMask, CostKind, 0, nullptr, {LHS, RHS}, &I);
// Handle shuffle(binop(shuffle(x),y),binop(z,shuffle(w))) style patterns
// where one use shuffles have gotten split across the binop/cmp. These
// often allow a major reduction in total cost that wouldn't happen as
// individual folds.
auto MergeInner = [&](Value *&Op, int Offset, MutableArrayRef<int> Mask,
TTI::TargetCostKind CostKind) -> bool {
Value *InnerOp;
ArrayRef<int> InnerMask;
if (match(Op, m_OneUse(m_Shuffle(m_Value(InnerOp), m_Undef(),
m_Mask(InnerMask)))) &&
InnerOp->getType() == Op->getType() &&
all_of(InnerMask,
[NumSrcElts](int M) { return M < (int)NumSrcElts; })) {
for (int &M : Mask)
if (Offset <= M && M < (int)(Offset + NumSrcElts)) {
M = InnerMask[M - Offset];
M = 0 <= M ? M + Offset : M;
}
OldCost += TTI.getInstructionCost(cast<Instruction>(Op), CostKind);
Op = InnerOp;
return true;
}
return false;
};
bool ReducedInstCount = false;
ReducedInstCount |= MergeInner(X, 0, NewMask0, CostKind);
ReducedInstCount |= MergeInner(Y, 0, NewMask1, CostKind);
ReducedInstCount |= MergeInner(Z, NumSrcElts, NewMask0, CostKind);
ReducedInstCount |= MergeInner(W, NumSrcElts, NewMask1, CostKind);
InstructionCost NewCost =
TTI.getShuffleCost(SK0, BinOpTy, NewMask0, CostKind, 0, nullptr, {X, Z}) +
TTI.getShuffleCost(SK1, BinOpTy, NewMask1, CostKind, 0, nullptr, {Y, W});
if (PredLHS == CmpInst::BAD_ICMP_PREDICATE) {
NewCost +=
TTI.getArithmeticInstrCost(LHS->getOpcode(), ShuffleDstTy, CostKind);
} else {
auto *ShuffleCmpTy =
FixedVectorType::get(BinOpTy->getElementType(), ShuffleDstTy);
NewCost += TTI.getCmpSelInstrCost(LHS->getOpcode(), ShuffleCmpTy,
ShuffleDstTy, PredLHS, CostKind);
}
LLVM_DEBUG(dbgs() << "Found a shuffle feeding two binops: " << I
<< "\n OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
// If either shuffle will constant fold away, then fold for the same cost as
// we will reduce the instruction count.
ReducedInstCount |= (isa<Constant>(X) && isa<Constant>(Z)) ||
(isa<Constant>(Y) && isa<Constant>(W));
if (ReducedInstCount ? (NewCost > OldCost) : (NewCost >= OldCost))
return false;
Value *Shuf0 = Builder.CreateShuffleVector(X, Z, NewMask0);
Value *Shuf1 = Builder.CreateShuffleVector(Y, W, NewMask1);
Value *NewBO = PredLHS == CmpInst::BAD_ICMP_PREDICATE
? Builder.CreateBinOp(
cast<BinaryOperator>(LHS)->getOpcode(), Shuf0, Shuf1)
: Builder.CreateCmp(PredLHS, Shuf0, Shuf1);
// Intersect flags from the old binops.
if (auto *NewInst = dyn_cast<Instruction>(NewBO)) {
NewInst->copyIRFlags(LHS);
NewInst->andIRFlags(RHS);
}
Worklist.pushValue(Shuf0);
Worklist.pushValue(Shuf1);
replaceValue(I, *NewBO);
return true;
}
/// Try to convert,
/// (shuffle(select(c1,t1,f1)), (select(c2,t2,f2)), m) into
/// (select (shuffle c1,c2,m), (shuffle t1,t2,m), (shuffle f1,f2,m))
bool VectorCombine::foldShuffleOfSelects(Instruction &I) {
ArrayRef<int> Mask;
Value *C1, *T1, *F1, *C2, *T2, *F2;
if (!match(&I, m_Shuffle(
m_OneUse(m_Select(m_Value(C1), m_Value(T1), m_Value(F1))),
m_OneUse(m_Select(m_Value(C2), m_Value(T2), m_Value(F2))),
m_Mask(Mask))))
return false;
auto *C1VecTy = dyn_cast<FixedVectorType>(C1->getType());
auto *C2VecTy = dyn_cast<FixedVectorType>(C2->getType());
if (!C1VecTy || !C2VecTy || C1VecTy != C2VecTy)
return false;
auto *SI0FOp = dyn_cast<FPMathOperator>(I.getOperand(0));
auto *SI1FOp = dyn_cast<FPMathOperator>(I.getOperand(1));
// SelectInsts must have the same FMF.
if (((SI0FOp == nullptr) != (SI1FOp == nullptr)) ||
((SI0FOp != nullptr) &&
(SI0FOp->getFastMathFlags() != SI1FOp->getFastMathFlags())))
return false;
auto *SrcVecTy = cast<FixedVectorType>(T1->getType());
auto *DstVecTy = cast<FixedVectorType>(I.getType());
auto SK = TargetTransformInfo::SK_PermuteTwoSrc;
auto SelOp = Instruction::Select;
InstructionCost OldCost = TTI.getCmpSelInstrCost(
SelOp, SrcVecTy, C1VecTy, CmpInst::BAD_ICMP_PREDICATE, CostKind);
OldCost += TTI.getCmpSelInstrCost(SelOp, SrcVecTy, C2VecTy,
CmpInst::BAD_ICMP_PREDICATE, CostKind);
OldCost += TTI.getShuffleCost(SK, SrcVecTy, Mask, CostKind, 0, nullptr,
{I.getOperand(0), I.getOperand(1)}, &I);
InstructionCost NewCost =
TTI.getShuffleCost(SK, C1VecTy, Mask, CostKind, 0, nullptr, {C1, C2});
NewCost +=
TTI.getShuffleCost(SK, SrcVecTy, Mask, CostKind, 0, nullptr, {T1, T2});
NewCost +=
TTI.getShuffleCost(SK, SrcVecTy, Mask, CostKind, 0, nullptr, {F1, F2});
auto *C1C2ShuffledVecTy = cast<FixedVectorType>(
toVectorTy(Type::getInt1Ty(I.getContext()), DstVecTy->getNumElements()));
NewCost += TTI.getCmpSelInstrCost(SelOp, DstVecTy, C1C2ShuffledVecTy,
CmpInst::BAD_ICMP_PREDICATE, CostKind);
LLVM_DEBUG(dbgs() << "Found a shuffle feeding two selects: " << I
<< "\n OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
if (NewCost > OldCost)
return false;
Value *ShuffleCmp = Builder.CreateShuffleVector(C1, C2, Mask);
Value *ShuffleTrue = Builder.CreateShuffleVector(T1, T2, Mask);
Value *ShuffleFalse = Builder.CreateShuffleVector(F1, F2, Mask);
Value *NewSel;
// We presuppose that the SelectInsts have the same FMF.
if (SI0FOp)
NewSel = Builder.CreateSelectFMF(ShuffleCmp, ShuffleTrue, ShuffleFalse,
SI0FOp->getFastMathFlags());
else
NewSel = Builder.CreateSelect(ShuffleCmp, ShuffleTrue, ShuffleFalse);
Worklist.pushValue(ShuffleCmp);
Worklist.pushValue(ShuffleTrue);
Worklist.pushValue(ShuffleFalse);
replaceValue(I, *NewSel);
return true;
}
/// Try to convert "shuffle (castop), (castop)" with a shared castop operand
/// into "castop (shuffle)".
bool VectorCombine::foldShuffleOfCastops(Instruction &I) {
Value *V0, *V1;
ArrayRef<int> OldMask;
if (!match(&I, m_Shuffle(m_Value(V0), m_Value(V1), m_Mask(OldMask))))
return false;
auto *C0 = dyn_cast<CastInst>(V0);
auto *C1 = dyn_cast<CastInst>(V1);
if (!C0 || !C1)
return false;
Instruction::CastOps Opcode = C0->getOpcode();
if (C0->getSrcTy() != C1->getSrcTy())
return false;
// Handle shuffle(zext_nneg(x), sext(y)) -> sext(shuffle(x,y)) folds.
if (Opcode != C1->getOpcode()) {
if (match(C0, m_SExtLike(m_Value())) && match(C1, m_SExtLike(m_Value())))
Opcode = Instruction::SExt;
else
return false;
}
auto *ShuffleDstTy = dyn_cast<FixedVectorType>(I.getType());
auto *CastDstTy = dyn_cast<FixedVectorType>(C0->getDestTy());
auto *CastSrcTy = dyn_cast<FixedVectorType>(C0->getSrcTy());
if (!ShuffleDstTy || !CastDstTy || !CastSrcTy)
return false;
unsigned NumSrcElts = CastSrcTy->getNumElements();
unsigned NumDstElts = CastDstTy->getNumElements();
assert((NumDstElts == NumSrcElts || Opcode == Instruction::BitCast) &&
"Only bitcasts expected to alter src/dst element counts");
// Check for bitcasting of unscalable vector types.
// e.g. <32 x i40> -> <40 x i32>
if (NumDstElts != NumSrcElts && (NumSrcElts % NumDstElts) != 0 &&
(NumDstElts % NumSrcElts) != 0)
return false;
SmallVector<int, 16> NewMask;
if (NumSrcElts >= NumDstElts) {
// The bitcast is from wide to narrow/equal elements. The shuffle mask can
// always be expanded to the equivalent form choosing narrower elements.
assert(NumSrcElts % NumDstElts == 0 && "Unexpected shuffle mask");
unsigned ScaleFactor = NumSrcElts / NumDstElts;
narrowShuffleMaskElts(ScaleFactor, OldMask, NewMask);
} else {
// The bitcast is from narrow elements to wide elements. The shuffle mask
// must choose consecutive elements to allow casting first.
assert(NumDstElts % NumSrcElts == 0 && "Unexpected shuffle mask");
unsigned ScaleFactor = NumDstElts / NumSrcElts;
if (!widenShuffleMaskElts(ScaleFactor, OldMask, NewMask))
return false;
}
auto *NewShuffleDstTy =
FixedVectorType::get(CastSrcTy->getScalarType(), NewMask.size());
// Try to replace a castop with a shuffle if the shuffle is not costly.
InstructionCost CostC0 =
TTI.getCastInstrCost(C0->getOpcode(), CastDstTy, CastSrcTy,
TTI::CastContextHint::None, CostKind);
InstructionCost CostC1 =
TTI.getCastInstrCost(C1->getOpcode(), CastDstTy, CastSrcTy,
TTI::CastContextHint::None, CostKind);
InstructionCost OldCost = CostC0 + CostC1;
OldCost +=
TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, CastDstTy,
OldMask, CostKind, 0, nullptr, {}, &I);
InstructionCost NewCost = TTI.getShuffleCost(
TargetTransformInfo::SK_PermuteTwoSrc, CastSrcTy, NewMask, CostKind);
NewCost += TTI.getCastInstrCost(Opcode, ShuffleDstTy, NewShuffleDstTy,
TTI::CastContextHint::None, CostKind);
if (!C0->hasOneUse())
NewCost += CostC0;
if (!C1->hasOneUse())
NewCost += CostC1;
LLVM_DEBUG(dbgs() << "Found a shuffle feeding two casts: " << I
<< "\n OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
if (NewCost > OldCost)
return false;
Value *Shuf = Builder.CreateShuffleVector(C0->getOperand(0),
C1->getOperand(0), NewMask);
Value *Cast = Builder.CreateCast(Opcode, Shuf, ShuffleDstTy);
// Intersect flags from the old casts.
if (auto *NewInst = dyn_cast<Instruction>(Cast)) {
NewInst->copyIRFlags(C0);
NewInst->andIRFlags(C1);
}
Worklist.pushValue(Shuf);
replaceValue(I, *Cast);
return true;
}
/// Try to convert any of:
/// "shuffle (shuffle x, y), (shuffle y, x)"
/// "shuffle (shuffle x, undef), (shuffle y, undef)"
/// "shuffle (shuffle x, undef), y"
/// "shuffle x, (shuffle y, undef)"
/// into "shuffle x, y".
bool VectorCombine::foldShuffleOfShuffles(Instruction &I) {
ArrayRef<int> OuterMask;
Value *OuterV0, *OuterV1;
if (!match(&I,
m_Shuffle(m_Value(OuterV0), m_Value(OuterV1), m_Mask(OuterMask))))
return false;
ArrayRef<int> InnerMask0, InnerMask1;
Value *X0, *X1, *Y0, *Y1;
bool Match0 =
match(OuterV0, m_Shuffle(m_Value(X0), m_Value(Y0), m_Mask(InnerMask0)));
bool Match1 =
match(OuterV1, m_Shuffle(m_Value(X1), m_Value(Y1), m_Mask(InnerMask1)));
if (!Match0 && !Match1)
return false;
X0 = Match0 ? X0 : OuterV0;
Y0 = Match0 ? Y0 : OuterV0;
X1 = Match1 ? X1 : OuterV1;
Y1 = Match1 ? Y1 : OuterV1;
auto *ShuffleDstTy = dyn_cast<FixedVectorType>(I.getType());
auto *ShuffleSrcTy = dyn_cast<FixedVectorType>(X0->getType());
auto *ShuffleImmTy = dyn_cast<FixedVectorType>(OuterV0->getType());
if (!ShuffleDstTy || !ShuffleSrcTy || !ShuffleImmTy ||
X0->getType() != X1->getType())
return false;
unsigned NumSrcElts = ShuffleSrcTy->getNumElements();
unsigned NumImmElts = ShuffleImmTy->getNumElements();
// Attempt to merge shuffles, matching upto 2 source operands.
// Replace index to a poison arg with PoisonMaskElem.
// Bail if either inner masks reference an undef arg.
SmallVector<int, 16> NewMask(OuterMask);
Value *NewX = nullptr, *NewY = nullptr;
for (int &M : NewMask) {
Value *Src = nullptr;
if (0 <= M && M < (int)NumImmElts) {
Src = OuterV0;
if (Match0) {
M = InnerMask0[M];
Src = M >= (int)NumSrcElts ? Y0 : X0;
M = M >= (int)NumSrcElts ? (M - NumSrcElts) : M;
}
} else if (M >= (int)NumImmElts) {
Src = OuterV1;
M -= NumImmElts;
if (Match1) {
M = InnerMask1[M];
Src = M >= (int)NumSrcElts ? Y1 : X1;
M = M >= (int)NumSrcElts ? (M - NumSrcElts) : M;
}
}
if (Src && M != PoisonMaskElem) {
assert(0 <= M && M < (int)NumSrcElts && "Unexpected shuffle mask index");
if (isa<UndefValue>(Src)) {
// We've referenced an undef element - if its poison, update the shuffle
// mask, else bail.
if (!isa<PoisonValue>(Src))
return false;
M = PoisonMaskElem;
continue;
}
if (!NewX || NewX == Src) {
NewX = Src;
continue;
}
if (!NewY || NewY == Src) {
M += NumSrcElts;
NewY = Src;
continue;
}
return false;
}
}
if (!NewX)
return PoisonValue::get(ShuffleDstTy);
if (!NewY)
NewY = PoisonValue::get(ShuffleSrcTy);
// Have we folded to an Identity shuffle?
if (ShuffleVectorInst::isIdentityMask(NewMask, NumSrcElts)) {
replaceValue(I, *NewX);
return true;
}
// Try to merge the shuffles if the new shuffle is not costly.
InstructionCost InnerCost0 = 0;
if (Match0)
InnerCost0 = TTI.getInstructionCost(cast<Instruction>(OuterV0), CostKind);
InstructionCost InnerCost1 = 0;
if (Match1)
InnerCost1 = TTI.getInstructionCost(cast<Instruction>(OuterV1), CostKind);
InstructionCost OuterCost = TTI.getInstructionCost(&I, CostKind);
InstructionCost OldCost = InnerCost0 + InnerCost1 + OuterCost;
bool IsUnary = all_of(NewMask, [&](int M) { return M < (int)NumSrcElts; });
TargetTransformInfo::ShuffleKind SK =
IsUnary ? TargetTransformInfo::SK_PermuteSingleSrc
: TargetTransformInfo::SK_PermuteTwoSrc;
InstructionCost NewCost = TTI.getShuffleCost(
SK, ShuffleSrcTy, NewMask, CostKind, 0, nullptr, {NewX, NewY});
if (!OuterV0->hasOneUse())
NewCost += InnerCost0;
if (!OuterV1->hasOneUse())
NewCost += InnerCost1;
LLVM_DEBUG(dbgs() << "Found a shuffle feeding two shuffles: " << I
<< "\n OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
if (NewCost > OldCost)
return false;
Value *Shuf = Builder.CreateShuffleVector(NewX, NewY, NewMask);
replaceValue(I, *Shuf);
return true;
}
/// Try to convert
/// "shuffle (intrinsic), (intrinsic)" into "intrinsic (shuffle), (shuffle)".
bool VectorCombine::foldShuffleOfIntrinsics(Instruction &I) {
Value *V0, *V1;
ArrayRef<int> OldMask;
if (!match(&I, m_Shuffle(m_OneUse(m_Value(V0)), m_OneUse(m_Value(V1)),
m_Mask(OldMask))))
return false;
auto *II0 = dyn_cast<IntrinsicInst>(V0);
auto *II1 = dyn_cast<IntrinsicInst>(V1);
if (!II0 || !II1)
return false;
Intrinsic::ID IID = II0->getIntrinsicID();
if (IID != II1->getIntrinsicID())
return false;
auto *ShuffleDstTy = dyn_cast<FixedVectorType>(I.getType());
auto *II0Ty = dyn_cast<FixedVectorType>(II0->getType());
if (!ShuffleDstTy || !II0Ty)
return false;
if (!isTriviallyVectorizable(IID))
return false;
for (unsigned I = 0, E = II0->arg_size(); I != E; ++I)
if (isVectorIntrinsicWithScalarOpAtArg(IID, I, &TTI) &&
II0->getArgOperand(I) != II1->getArgOperand(I))
return false;
InstructionCost OldCost =
TTI.getIntrinsicInstrCost(IntrinsicCostAttributes(IID, *II0), CostKind) +
TTI.getIntrinsicInstrCost(IntrinsicCostAttributes(IID, *II1), CostKind) +
TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, II0Ty, OldMask,
CostKind, 0, nullptr, {II0, II1}, &I);
SmallVector<Type *> NewArgsTy;
InstructionCost NewCost = 0;
for (unsigned I = 0, E = II0->arg_size(); I != E; ++I)
if (isVectorIntrinsicWithScalarOpAtArg(IID, I, &TTI)) {
NewArgsTy.push_back(II0->getArgOperand(I)->getType());
} else {
auto *VecTy = cast<FixedVectorType>(II0->getArgOperand(I)->getType());
NewArgsTy.push_back(FixedVectorType::get(VecTy->getElementType(),
ShuffleDstTy->getNumElements()));
NewCost += TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc,
VecTy, OldMask, CostKind);
}
IntrinsicCostAttributes NewAttr(IID, ShuffleDstTy, NewArgsTy);
NewCost += TTI.getIntrinsicInstrCost(NewAttr, CostKind);
LLVM_DEBUG(dbgs() << "Found a shuffle feeding two intrinsics: " << I
<< "\n OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
if (NewCost > OldCost)
return false;
SmallVector<Value *> NewArgs;
for (unsigned I = 0, E = II0->arg_size(); I != E; ++I)
if (isVectorIntrinsicWithScalarOpAtArg(IID, I, &TTI)) {
NewArgs.push_back(II0->getArgOperand(I));
} else {
Value *Shuf = Builder.CreateShuffleVector(II0->getArgOperand(I),
II1->getArgOperand(I), OldMask);
NewArgs.push_back(Shuf);
Worklist.pushValue(Shuf);
}
Value *NewIntrinsic = Builder.CreateIntrinsic(ShuffleDstTy, IID, NewArgs);
// Intersect flags from the old intrinsics.
if (auto *NewInst = dyn_cast<Instruction>(NewIntrinsic)) {
NewInst->copyIRFlags(II0);
NewInst->andIRFlags(II1);
}
replaceValue(I, *NewIntrinsic);
return true;
}
using InstLane = std::pair<Use *, int>;
static InstLane lookThroughShuffles(Use *U, int Lane) {
while (auto *SV = dyn_cast<ShuffleVectorInst>(U->get())) {
unsigned NumElts =
cast<FixedVectorType>(SV->getOperand(0)->getType())->getNumElements();
int M = SV->getMaskValue(Lane);
if (M < 0)
return {nullptr, PoisonMaskElem};
if (static_cast<unsigned>(M) < NumElts) {
U = &SV->getOperandUse(0);
Lane = M;
} else {
U = &SV->getOperandUse(1);
Lane = M - NumElts;
}
}
return InstLane{U, Lane};
}
static SmallVector<InstLane>
generateInstLaneVectorFromOperand(ArrayRef<InstLane> Item, int Op) {
SmallVector<InstLane> NItem;
for (InstLane IL : Item) {
auto [U, Lane] = IL;
InstLane OpLane =
U ? lookThroughShuffles(&cast<Instruction>(U->get())->getOperandUse(Op),
Lane)
: InstLane{nullptr, PoisonMaskElem};
NItem.emplace_back(OpLane);
}
return NItem;
}
/// Detect concat of multiple values into a vector
static bool isFreeConcat(ArrayRef<InstLane> Item, TTI::TargetCostKind CostKind,
const TargetTransformInfo &TTI) {
auto *Ty = cast<FixedVectorType>(Item.front().first->get()->getType());
unsigned NumElts = Ty->getNumElements();
if (Item.size() == NumElts || NumElts == 1 || Item.size() % NumElts != 0)
return false;
// Check that the concat is free, usually meaning that the type will be split
// during legalization.
SmallVector<int, 16> ConcatMask(NumElts * 2);
std::iota(ConcatMask.begin(), ConcatMask.end(), 0);
if (TTI.getShuffleCost(TTI::SK_PermuteTwoSrc, Ty, ConcatMask, CostKind) != 0)
return false;
unsigned NumSlices = Item.size() / NumElts;
// Currently we generate a tree of shuffles for the concats, which limits us
// to a power2.
if (!isPowerOf2_32(NumSlices))
return false;
for (unsigned Slice = 0; Slice < NumSlices; ++Slice) {
Use *SliceV = Item[Slice * NumElts].first;
if (!SliceV || SliceV->get()->getType() != Ty)
return false;
for (unsigned Elt = 0; Elt < NumElts; ++Elt) {
auto [V, Lane] = Item[Slice * NumElts + Elt];
if (Lane != static_cast<int>(Elt) || SliceV->get() != V->get())
return false;
}
}
return true;
}
static Value *generateNewInstTree(ArrayRef<InstLane> Item, FixedVectorType *Ty,
const SmallPtrSet<Use *, 4> &IdentityLeafs,
const SmallPtrSet<Use *, 4> &SplatLeafs,
const SmallPtrSet<Use *, 4> &ConcatLeafs,
IRBuilder<> &Builder,
const TargetTransformInfo *TTI) {
auto [FrontU, FrontLane] = Item.front();
if (IdentityLeafs.contains(FrontU)) {
return FrontU->get();
}
if (SplatLeafs.contains(FrontU)) {
SmallVector<int, 16> Mask(Ty->getNumElements(), FrontLane);
return Builder.CreateShuffleVector(FrontU->get(), Mask);
}
if (ConcatLeafs.contains(FrontU)) {
unsigned NumElts =
cast<FixedVectorType>(FrontU->get()->getType())->getNumElements();
SmallVector<Value *> Values(Item.size() / NumElts, nullptr);
for (unsigned S = 0; S < Values.size(); ++S)
Values[S] = Item[S * NumElts].first->get();
while (Values.size() > 1) {
NumElts *= 2;
SmallVector<int, 16> Mask(NumElts, 0);
std::iota(Mask.begin(), Mask.end(), 0);
SmallVector<Value *> NewValues(Values.size() / 2, nullptr);
for (unsigned S = 0; S < NewValues.size(); ++S)
NewValues[S] =
Builder.CreateShuffleVector(Values[S * 2], Values[S * 2 + 1], Mask);
Values = NewValues;
}
return Values[0];
}
auto *I = cast<Instruction>(FrontU->get());
auto *II = dyn_cast<IntrinsicInst>(I);
unsigned NumOps = I->getNumOperands() - (II ? 1 : 0);
SmallVector<Value *> Ops(NumOps);
for (unsigned Idx = 0; Idx < NumOps; Idx++) {
if (II &&
isVectorIntrinsicWithScalarOpAtArg(II->getIntrinsicID(), Idx, TTI)) {
Ops[Idx] = II->getOperand(Idx);
continue;
}
Ops[Idx] = generateNewInstTree(generateInstLaneVectorFromOperand(Item, Idx),
Ty, IdentityLeafs, SplatLeafs, ConcatLeafs,
Builder, TTI);
}
SmallVector<Value *, 8> ValueList;
for (const auto &Lane : Item)
if (Lane.first)
ValueList.push_back(Lane.first->get());
Type *DstTy =
FixedVectorType::get(I->getType()->getScalarType(), Ty->getNumElements());
if (auto *BI = dyn_cast<BinaryOperator>(I)) {
auto *Value = Builder.CreateBinOp((Instruction::BinaryOps)BI->getOpcode(),
Ops[0], Ops[1]);
propagateIRFlags(Value, ValueList);
return Value;
}
if (auto *CI = dyn_cast<CmpInst>(I)) {
auto *Value = Builder.CreateCmp(CI->getPredicate(), Ops[0], Ops[1]);
propagateIRFlags(Value, ValueList);
return Value;
}
if (auto *SI = dyn_cast<SelectInst>(I)) {
auto *Value = Builder.CreateSelect(Ops[0], Ops[1], Ops[2], "", SI);
propagateIRFlags(Value, ValueList);
return Value;
}
if (auto *CI = dyn_cast<CastInst>(I)) {
auto *Value = Builder.CreateCast((Instruction::CastOps)CI->getOpcode(),
Ops[0], DstTy);
propagateIRFlags(Value, ValueList);
return Value;
}
if (II) {
auto *Value = Builder.CreateIntrinsic(DstTy, II->getIntrinsicID(), Ops);
propagateIRFlags(Value, ValueList);
return Value;
}
assert(isa<UnaryInstruction>(I) && "Unexpected instruction type in Generate");
auto *Value =
Builder.CreateUnOp((Instruction::UnaryOps)I->getOpcode(), Ops[0]);
propagateIRFlags(Value, ValueList);
return Value;
}
// Starting from a shuffle, look up through operands tracking the shuffled index
// of each lane. If we can simplify away the shuffles to identities then
// do so.
bool VectorCombine::foldShuffleToIdentity(Instruction &I) {
auto *Ty = dyn_cast<FixedVectorType>(I.getType());
if (!Ty || I.use_empty())
return false;
SmallVector<InstLane> Start(Ty->getNumElements());
for (unsigned M = 0, E = Ty->getNumElements(); M < E; ++M)
Start[M] = lookThroughShuffles(&*I.use_begin(), M);
SmallVector<SmallVector<InstLane>> Worklist;
Worklist.push_back(Start);
SmallPtrSet<Use *, 4> IdentityLeafs, SplatLeafs, ConcatLeafs;
unsigned NumVisited = 0;
while (!Worklist.empty()) {
if (++NumVisited > MaxInstrsToScan)
return false;
SmallVector<InstLane> Item = Worklist.pop_back_val();
auto [FrontU, FrontLane] = Item.front();
// If we found an undef first lane then bail out to keep things simple.
if (!FrontU)
return false;
// Helper to peek through bitcasts to the same value.
auto IsEquiv = [&](Value *X, Value *Y) {
return X->getType() == Y->getType() &&
peekThroughBitcasts(X) == peekThroughBitcasts(Y);
};
// Look for an identity value.
if (FrontLane == 0 &&
cast<FixedVectorType>(FrontU->get()->getType())->getNumElements() ==
Ty->getNumElements() &&
all_of(drop_begin(enumerate(Item)), [IsEquiv, Item](const auto &E) {
Value *FrontV = Item.front().first->get();
return !E.value().first || (IsEquiv(E.value().first->get(), FrontV) &&
E.value().second == (int)E.index());
})) {
IdentityLeafs.insert(FrontU);
continue;
}
// Look for constants, for the moment only supporting constant splats.
if (auto *C = dyn_cast<Constant>(FrontU);
C && C->getSplatValue() &&
all_of(drop_begin(Item), [Item](InstLane &IL) {
Value *FrontV = Item.front().first->get();
Use *U = IL.first;
return !U || (isa<Constant>(U->get()) &&
cast<Constant>(U->get())->getSplatValue() ==
cast<Constant>(FrontV)->getSplatValue());
})) {
SplatLeafs.insert(FrontU);
continue;
}
// Look for a splat value.
if (all_of(drop_begin(Item), [Item](InstLane &IL) {
auto [FrontU, FrontLane] = Item.front();
auto [U, Lane] = IL;
return !U || (U->get() == FrontU->get() && Lane == FrontLane);
})) {
SplatLeafs.insert(FrontU);
continue;
}
// We need each element to be the same type of value, and check that each
// element has a single use.
auto CheckLaneIsEquivalentToFirst = [Item](InstLane IL) {
Value *FrontV = Item.front().first->get();
if (!IL.first)
return true;
Value *V = IL.first->get();
if (auto *I = dyn_cast<Instruction>(V); I && !I->hasOneUse())
return false;
if (V->getValueID() != FrontV->getValueID())
return false;
if (auto *CI = dyn_cast<CmpInst>(V))
if (CI->getPredicate() != cast<CmpInst>(FrontV)->getPredicate())
return false;
if (auto *CI = dyn_cast<CastInst>(V))
if (CI->getSrcTy()->getScalarType() !=
cast<CastInst>(FrontV)->getSrcTy()->getScalarType())
return false;
if (auto *SI = dyn_cast<SelectInst>(V))
if (!isa<VectorType>(SI->getOperand(0)->getType()) ||
SI->getOperand(0)->getType() !=
cast<SelectInst>(FrontV)->getOperand(0)->getType())
return false;
if (isa<CallInst>(V) && !isa<IntrinsicInst>(V))
return false;
auto *II = dyn_cast<IntrinsicInst>(V);
return !II || (isa<IntrinsicInst>(FrontV) &&
II->getIntrinsicID() ==
cast<IntrinsicInst>(FrontV)->getIntrinsicID() &&
!II->hasOperandBundles());
};
if (all_of(drop_begin(Item), CheckLaneIsEquivalentToFirst)) {
// Check the operator is one that we support.
if (isa<BinaryOperator, CmpInst>(FrontU)) {
// We exclude div/rem in case they hit UB from poison lanes.
if (auto *BO = dyn_cast<BinaryOperator>(FrontU);
BO && BO->isIntDivRem())
return false;
Worklist.push_back(generateInstLaneVectorFromOperand(Item, 0));
Worklist.push_back(generateInstLaneVectorFromOperand(Item, 1));
continue;
} else if (isa<UnaryOperator, TruncInst, ZExtInst, SExtInst, FPToSIInst,
FPToUIInst, SIToFPInst, UIToFPInst>(FrontU)) {
Worklist.push_back(generateInstLaneVectorFromOperand(Item, 0));
continue;
} else if (auto *BitCast = dyn_cast<BitCastInst>(FrontU)) {
// TODO: Handle vector widening/narrowing bitcasts.
auto *DstTy = dyn_cast<FixedVectorType>(BitCast->getDestTy());
auto *SrcTy = dyn_cast<FixedVectorType>(BitCast->getSrcTy());
if (DstTy && SrcTy &&
SrcTy->getNumElements() == DstTy->getNumElements()) {
Worklist.push_back(generateInstLaneVectorFromOperand(Item, 0));
continue;
}
} else if (isa<SelectInst>(FrontU)) {
Worklist.push_back(generateInstLaneVectorFromOperand(Item, 0));
Worklist.push_back(generateInstLaneVectorFromOperand(Item, 1));
Worklist.push_back(generateInstLaneVectorFromOperand(Item, 2));
continue;
} else if (auto *II = dyn_cast<IntrinsicInst>(FrontU);
II && isTriviallyVectorizable(II->getIntrinsicID()) &&
!II->hasOperandBundles()) {
for (unsigned Op = 0, E = II->getNumOperands() - 1; Op < E; Op++) {
if (isVectorIntrinsicWithScalarOpAtArg(II->getIntrinsicID(), Op,
&TTI)) {
if (!all_of(drop_begin(Item), [Item, Op](InstLane &IL) {
Value *FrontV = Item.front().first->get();
Use *U = IL.first;
return !U || (cast<Instruction>(U->get())->getOperand(Op) ==
cast<Instruction>(FrontV)->getOperand(Op));
}))
return false;
continue;
}
Worklist.push_back(generateInstLaneVectorFromOperand(Item, Op));
}
continue;
}
}
if (isFreeConcat(Item, CostKind, TTI)) {
ConcatLeafs.insert(FrontU);
continue;
}
return false;
}
if (NumVisited <= 1)
return false;
LLVM_DEBUG(dbgs() << "Found a superfluous identity shuffle: " << I << "\n");
// If we got this far, we know the shuffles are superfluous and can be
// removed. Scan through again and generate the new tree of instructions.
Builder.SetInsertPoint(&I);
Value *V = generateNewInstTree(Start, Ty, IdentityLeafs, SplatLeafs,
ConcatLeafs, Builder, &TTI);
replaceValue(I, *V);
return true;
}
/// Given a commutative reduction, the order of the input lanes does not alter
/// the results. We can use this to remove certain shuffles feeding the
/// reduction, removing the need to shuffle at all.
bool VectorCombine::foldShuffleFromReductions(Instruction &I) {
auto *II = dyn_cast<IntrinsicInst>(&I);
if (!II)
return false;
switch (II->getIntrinsicID()) {
case Intrinsic::vector_reduce_add:
case Intrinsic::vector_reduce_mul:
case Intrinsic::vector_reduce_and:
case Intrinsic::vector_reduce_or:
case Intrinsic::vector_reduce_xor:
case Intrinsic::vector_reduce_smin:
case Intrinsic::vector_reduce_smax:
case Intrinsic::vector_reduce_umin:
case Intrinsic::vector_reduce_umax:
break;
default:
return false;
}
// Find all the inputs when looking through operations that do not alter the
// lane order (binops, for example). Currently we look for a single shuffle,
// and can ignore splat values.
std::queue<Value *> Worklist;
SmallPtrSet<Value *, 4> Visited;
ShuffleVectorInst *Shuffle = nullptr;
if (auto *Op = dyn_cast<Instruction>(I.getOperand(0)))
Worklist.push(Op);
while (!Worklist.empty()) {
Value *CV = Worklist.front();
Worklist.pop();
if (Visited.contains(CV))
continue;
// Splats don't change the order, so can be safely ignored.
if (isSplatValue(CV))
continue;
Visited.insert(CV);
if (auto *CI = dyn_cast<Instruction>(CV)) {
if (CI->isBinaryOp()) {
for (auto *Op : CI->operand_values())
Worklist.push(Op);
continue;
} else if (auto *SV = dyn_cast<ShuffleVectorInst>(CI)) {
if (Shuffle && Shuffle != SV)
return false;
Shuffle = SV;
continue;
}
}
// Anything else is currently an unknown node.
return false;
}
if (!Shuffle)
return false;
// Check all uses of the binary ops and shuffles are also included in the
// lane-invariant operations (Visited should be the list of lanewise
// instructions, including the shuffle that we found).
for (auto *V : Visited)
for (auto *U : V->users())
if (!Visited.contains(U) && U != &I)
return false;
FixedVectorType *VecType =
dyn_cast<FixedVectorType>(II->getOperand(0)->getType());
if (!VecType)
return false;
FixedVectorType *ShuffleInputType =
dyn_cast<FixedVectorType>(Shuffle->getOperand(0)->getType());
if (!ShuffleInputType)
return false;
unsigned NumInputElts = ShuffleInputType->getNumElements();
// Find the mask from sorting the lanes into order. This is most likely to
// become a identity or concat mask. Undef elements are pushed to the end.
SmallVector<int> ConcatMask;
Shuffle->getShuffleMask(ConcatMask);
sort(ConcatMask, [](int X, int Y) { return (unsigned)X < (unsigned)Y; });
// In the case of a truncating shuffle it's possible for the mask
// to have an index greater than the size of the resulting vector.
// This requires special handling.
bool IsTruncatingShuffle = VecType->getNumElements() < NumInputElts;
bool UsesSecondVec =
any_of(ConcatMask, [&](int M) { return M >= (int)NumInputElts; });
FixedVectorType *VecTyForCost =
(UsesSecondVec && !IsTruncatingShuffle) ? VecType : ShuffleInputType;
InstructionCost OldCost = TTI.getShuffleCost(
UsesSecondVec ? TTI::SK_PermuteTwoSrc : TTI::SK_PermuteSingleSrc,
VecTyForCost, Shuffle->getShuffleMask(), CostKind);
InstructionCost NewCost = TTI.getShuffleCost(
UsesSecondVec ? TTI::SK_PermuteTwoSrc : TTI::SK_PermuteSingleSrc,
VecTyForCost, ConcatMask, CostKind);
LLVM_DEBUG(dbgs() << "Found a reduction feeding from a shuffle: " << *Shuffle
<< "\n");
LLVM_DEBUG(dbgs() << " OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
if (NewCost < OldCost) {
Builder.SetInsertPoint(Shuffle);
Value *NewShuffle = Builder.CreateShuffleVector(
Shuffle->getOperand(0), Shuffle->getOperand(1), ConcatMask);
LLVM_DEBUG(dbgs() << "Created new shuffle: " << *NewShuffle << "\n");
replaceValue(*Shuffle, *NewShuffle);
}
// See if we can re-use foldSelectShuffle, getting it to reduce the size of
// the shuffle into a nicer order, as it can ignore the order of the shuffles.
return foldSelectShuffle(*Shuffle, true);
}
/// Determine if its more efficient to fold:
/// reduce(trunc(x)) -> trunc(reduce(x)).
/// reduce(sext(x)) -> sext(reduce(x)).
/// reduce(zext(x)) -> zext(reduce(x)).
bool VectorCombine::foldCastFromReductions(Instruction &I) {
auto *II = dyn_cast<IntrinsicInst>(&I);
if (!II)
return false;
bool TruncOnly = false;
Intrinsic::ID IID = II->getIntrinsicID();
switch (IID) {
case Intrinsic::vector_reduce_add:
case Intrinsic::vector_reduce_mul:
TruncOnly = true;
break;
case Intrinsic::vector_reduce_and:
case Intrinsic::vector_reduce_or:
case Intrinsic::vector_reduce_xor:
break;
default:
return false;
}
unsigned ReductionOpc = getArithmeticReductionInstruction(IID);
Value *ReductionSrc = I.getOperand(0);
Value *Src;
if (!match(ReductionSrc, m_OneUse(m_Trunc(m_Value(Src)))) &&
(TruncOnly || !match(ReductionSrc, m_OneUse(m_ZExtOrSExt(m_Value(Src))))))
return false;
auto CastOpc =
(Instruction::CastOps)cast<Instruction>(ReductionSrc)->getOpcode();
auto *SrcTy = cast<VectorType>(Src->getType());
auto *ReductionSrcTy = cast<VectorType>(ReductionSrc->getType());
Type *ResultTy = I.getType();
InstructionCost OldCost = TTI.getArithmeticReductionCost(
ReductionOpc, ReductionSrcTy, std::nullopt, CostKind);
OldCost += TTI.getCastInstrCost(CastOpc, ReductionSrcTy, SrcTy,
TTI::CastContextHint::None, CostKind,
cast<CastInst>(ReductionSrc));
InstructionCost NewCost =
TTI.getArithmeticReductionCost(ReductionOpc, SrcTy, std::nullopt,
CostKind) +
TTI.getCastInstrCost(CastOpc, ResultTy, ReductionSrcTy->getScalarType(),
TTI::CastContextHint::None, CostKind);
if (OldCost <= NewCost || !NewCost.isValid())
return false;
Value *NewReduction = Builder.CreateIntrinsic(SrcTy->getScalarType(),
II->getIntrinsicID(), {Src});
Value *NewCast = Builder.CreateCast(CastOpc, NewReduction, ResultTy);
replaceValue(I, *NewCast);
return true;
}
/// This method looks for groups of shuffles acting on binops, of the form:
/// %x = shuffle ...
/// %y = shuffle ...
/// %a = binop %x, %y
/// %b = binop %x, %y
/// shuffle %a, %b, selectmask
/// We may, especially if the shuffle is wider than legal, be able to convert
/// the shuffle to a form where only parts of a and b need to be computed. On
/// architectures with no obvious "select" shuffle, this can reduce the total
/// number of operations if the target reports them as cheaper.
bool VectorCombine::foldSelectShuffle(Instruction &I, bool FromReduction) {
auto *SVI = cast<ShuffleVectorInst>(&I);
auto *VT = cast<FixedVectorType>(I.getType());
auto *Op0 = dyn_cast<Instruction>(SVI->getOperand(0));
auto *Op1 = dyn_cast<Instruction>(SVI->getOperand(1));
if (!Op0 || !Op1 || Op0 == Op1 || !Op0->isBinaryOp() || !Op1->isBinaryOp() ||
VT != Op0->getType())
return false;
auto *SVI0A = dyn_cast<Instruction>(Op0->getOperand(0));
auto *SVI0B = dyn_cast<Instruction>(Op0->getOperand(1));
auto *SVI1A = dyn_cast<Instruction>(Op1->getOperand(0));
auto *SVI1B = dyn_cast<Instruction>(Op1->getOperand(1));
SmallPtrSet<Instruction *, 4> InputShuffles({SVI0A, SVI0B, SVI1A, SVI1B});
auto checkSVNonOpUses = [&](Instruction *I) {
if (!I || I->getOperand(0)->getType() != VT)
return true;
return any_of(I->users(), [&](User *U) {
return U != Op0 && U != Op1 &&
!(isa<ShuffleVectorInst>(U) &&
(InputShuffles.contains(cast<Instruction>(U)) ||
isInstructionTriviallyDead(cast<Instruction>(U))));
});
};
if (checkSVNonOpUses(SVI0A) || checkSVNonOpUses(SVI0B) ||
checkSVNonOpUses(SVI1A) || checkSVNonOpUses(SVI1B))
return false;
// Collect all the uses that are shuffles that we can transform together. We
// may not have a single shuffle, but a group that can all be transformed
// together profitably.
SmallVector<ShuffleVectorInst *> Shuffles;
auto collectShuffles = [&](Instruction *I) {
for (auto *U : I->users()) {
auto *SV = dyn_cast<ShuffleVectorInst>(U);
if (!SV || SV->getType() != VT)
return false;
if ((SV->getOperand(0) != Op0 && SV->getOperand(0) != Op1) ||
(SV->getOperand(1) != Op0 && SV->getOperand(1) != Op1))
return false;
if (!llvm::is_contained(Shuffles, SV))
Shuffles.push_back(SV);
}
return true;
};
if (!collectShuffles(Op0) || !collectShuffles(Op1))
return false;
// From a reduction, we need to be processing a single shuffle, otherwise the
// other uses will not be lane-invariant.
if (FromReduction && Shuffles.size() > 1)
return false;
// Add any shuffle uses for the shuffles we have found, to include them in our
// cost calculations.
if (!FromReduction) {
for (ShuffleVectorInst *SV : Shuffles) {
for (auto *U : SV->users()) {
ShuffleVectorInst *SSV = dyn_cast<ShuffleVectorInst>(U);
if (SSV && isa<UndefValue>(SSV->getOperand(1)) && SSV->getType() == VT)
Shuffles.push_back(SSV);
}
}
}
// For each of the output shuffles, we try to sort all the first vector
// elements to the beginning, followed by the second array elements at the
// end. If the binops are legalized to smaller vectors, this may reduce total
// number of binops. We compute the ReconstructMask mask needed to convert
// back to the original lane order.
SmallVector<std::pair<int, int>> V1, V2;
SmallVector<SmallVector<int>> OrigReconstructMasks;
int MaxV1Elt = 0, MaxV2Elt = 0;
unsigned NumElts = VT->getNumElements();
for (ShuffleVectorInst *SVN : Shuffles) {
SmallVector<int> Mask;
SVN->getShuffleMask(Mask);
// Check the operands are the same as the original, or reversed (in which
// case we need to commute the mask).
Value *SVOp0 = SVN->getOperand(0);
Value *SVOp1 = SVN->getOperand(1);
if (isa<UndefValue>(SVOp1)) {
auto *SSV = cast<ShuffleVectorInst>(SVOp0);
SVOp0 = SSV->getOperand(0);
SVOp1 = SSV->getOperand(1);
for (unsigned I = 0, E = Mask.size(); I != E; I++) {
if (Mask[I] >= static_cast<int>(SSV->getShuffleMask().size()))
return false;
Mask[I] = Mask[I] < 0 ? Mask[I] : SSV->getMaskValue(Mask[I]);
}
}
if (SVOp0 == Op1 && SVOp1 == Op0) {
std::swap(SVOp0, SVOp1);
ShuffleVectorInst::commuteShuffleMask(Mask, NumElts);
}
if (SVOp0 != Op0 || SVOp1 != Op1)
return false;
// Calculate the reconstruction mask for this shuffle, as the mask needed to
// take the packed values from Op0/Op1 and reconstructing to the original
// order.
SmallVector<int> ReconstructMask;
for (unsigned I = 0; I < Mask.size(); I++) {
if (Mask[I] < 0) {
ReconstructMask.push_back(-1);
} else if (Mask[I] < static_cast<int>(NumElts)) {
MaxV1Elt = std::max(MaxV1Elt, Mask[I]);
auto It = find_if(V1, [&](const std::pair<int, int> &A) {
return Mask[I] == A.first;
});
if (It != V1.end())
ReconstructMask.push_back(It - V1.begin());
else {
ReconstructMask.push_back(V1.size());
V1.emplace_back(Mask[I], V1.size());
}
} else {
MaxV2Elt = std::max<int>(MaxV2Elt, Mask[I] - NumElts);
auto It = find_if(V2, [&](const std::pair<int, int> &A) {
return Mask[I] - static_cast<int>(NumElts) == A.first;
});
if (It != V2.end())
ReconstructMask.push_back(NumElts + It - V2.begin());
else {
ReconstructMask.push_back(NumElts + V2.size());
V2.emplace_back(Mask[I] - NumElts, NumElts + V2.size());
}
}
}
// For reductions, we know that the lane ordering out doesn't alter the
// result. In-order can help simplify the shuffle away.
if (FromReduction)
sort(ReconstructMask);
OrigReconstructMasks.push_back(std::move(ReconstructMask));
}
// If the Maximum element used from V1 and V2 are not larger than the new
// vectors, the vectors are already packes and performing the optimization
// again will likely not help any further. This also prevents us from getting
// stuck in a cycle in case the costs do not also rule it out.
if (V1.empty() || V2.empty() ||
(MaxV1Elt == static_cast<int>(V1.size()) - 1 &&
MaxV2Elt == static_cast<int>(V2.size()) - 1))
return false;
// GetBaseMaskValue takes one of the inputs, which may either be a shuffle, a
// shuffle of another shuffle, or not a shuffle (that is treated like a
// identity shuffle).
auto GetBaseMaskValue = [&](Instruction *I, int M) {
auto *SV = dyn_cast<ShuffleVectorInst>(I);
if (!SV)
return M;
if (isa<UndefValue>(SV->getOperand(1)))
if (auto *SSV = dyn_cast<ShuffleVectorInst>(SV->getOperand(0)))
if (InputShuffles.contains(SSV))
return SSV->getMaskValue(SV->getMaskValue(M));
return SV->getMaskValue(M);
};
// Attempt to sort the inputs my ascending mask values to make simpler input
// shuffles and push complex shuffles down to the uses. We sort on the first
// of the two input shuffle orders, to try and get at least one input into a
// nice order.
auto SortBase = [&](Instruction *A, std::pair<int, int> X,
std::pair<int, int> Y) {
int MXA = GetBaseMaskValue(A, X.first);
int MYA = GetBaseMaskValue(A, Y.first);
return MXA < MYA;
};
stable_sort(V1, [&](std::pair<int, int> A, std::pair<int, int> B) {
return SortBase(SVI0A, A, B);
});
stable_sort(V2, [&](std::pair<int, int> A, std::pair<int, int> B) {
return SortBase(SVI1A, A, B);
});
// Calculate our ReconstructMasks from the OrigReconstructMasks and the
// modified order of the input shuffles.
SmallVector<SmallVector<int>> ReconstructMasks;
for (const auto &Mask : OrigReconstructMasks) {
SmallVector<int> ReconstructMask;
for (int M : Mask) {
auto FindIndex = [](const SmallVector<std::pair<int, int>> &V, int M) {
auto It = find_if(V, [M](auto A) { return A.second == M; });
assert(It != V.end() && "Expected all entries in Mask");
return std::distance(V.begin(), It);
};
if (M < 0)
ReconstructMask.push_back(-1);
else if (M < static_cast<int>(NumElts)) {
ReconstructMask.push_back(FindIndex(V1, M));
} else {
ReconstructMask.push_back(NumElts + FindIndex(V2, M));
}
}
ReconstructMasks.push_back(std::move(ReconstructMask));
}
// Calculate the masks needed for the new input shuffles, which get padded
// with undef
SmallVector<int> V1A, V1B, V2A, V2B;
for (unsigned I = 0; I < V1.size(); I++) {
V1A.push_back(GetBaseMaskValue(SVI0A, V1[I].first));
V1B.push_back(GetBaseMaskValue(SVI0B, V1[I].first));
}
for (unsigned I = 0; I < V2.size(); I++) {
V2A.push_back(GetBaseMaskValue(SVI1A, V2[I].first));
V2B.push_back(GetBaseMaskValue(SVI1B, V2[I].first));
}
while (V1A.size() < NumElts) {
V1A.push_back(PoisonMaskElem);
V1B.push_back(PoisonMaskElem);
}
while (V2A.size() < NumElts) {
V2A.push_back(PoisonMaskElem);
V2B.push_back(PoisonMaskElem);
}
auto AddShuffleCost = [&](InstructionCost C, Instruction *I) {
auto *SV = dyn_cast<ShuffleVectorInst>(I);
if (!SV)
return C;
return C + TTI.getShuffleCost(isa<UndefValue>(SV->getOperand(1))
? TTI::SK_PermuteSingleSrc
: TTI::SK_PermuteTwoSrc,
VT, SV->getShuffleMask(), CostKind);
};
auto AddShuffleMaskCost = [&](InstructionCost C, ArrayRef<int> Mask) {
return C + TTI.getShuffleCost(TTI::SK_PermuteTwoSrc, VT, Mask, CostKind);
};
// Get the costs of the shuffles + binops before and after with the new
// shuffle masks.
InstructionCost CostBefore =
TTI.getArithmeticInstrCost(Op0->getOpcode(), VT, CostKind) +
TTI.getArithmeticInstrCost(Op1->getOpcode(), VT, CostKind);
CostBefore += std::accumulate(Shuffles.begin(), Shuffles.end(),
InstructionCost(0), AddShuffleCost);
CostBefore += std::accumulate(InputShuffles.begin(), InputShuffles.end(),
InstructionCost(0), AddShuffleCost);
// The new binops will be unused for lanes past the used shuffle lengths.
// These types attempt to get the correct cost for that from the target.
FixedVectorType *Op0SmallVT =
FixedVectorType::get(VT->getScalarType(), V1.size());
FixedVectorType *Op1SmallVT =
FixedVectorType::get(VT->getScalarType(), V2.size());
InstructionCost CostAfter =
TTI.getArithmeticInstrCost(Op0->getOpcode(), Op0SmallVT, CostKind) +
TTI.getArithmeticInstrCost(Op1->getOpcode(), Op1SmallVT, CostKind);
CostAfter += std::accumulate(ReconstructMasks.begin(), ReconstructMasks.end(),
InstructionCost(0), AddShuffleMaskCost);
std::set<SmallVector<int>> OutputShuffleMasks({V1A, V1B, V2A, V2B});
CostAfter +=
std::accumulate(OutputShuffleMasks.begin(), OutputShuffleMasks.end(),
InstructionCost(0), AddShuffleMaskCost);
LLVM_DEBUG(dbgs() << "Found a binop select shuffle pattern: " << I << "\n");
LLVM_DEBUG(dbgs() << " CostBefore: " << CostBefore
<< " vs CostAfter: " << CostAfter << "\n");
if (CostBefore <= CostAfter)
return false;
// The cost model has passed, create the new instructions.
auto GetShuffleOperand = [&](Instruction *I, unsigned Op) -> Value * {
auto *SV = dyn_cast<ShuffleVectorInst>(I);
if (!SV)
return I;
if (isa<UndefValue>(SV->getOperand(1)))
if (auto *SSV = dyn_cast<ShuffleVectorInst>(SV->getOperand(0)))
if (InputShuffles.contains(SSV))
return SSV->getOperand(Op);
return SV->getOperand(Op);
};
Builder.SetInsertPoint(*SVI0A->getInsertionPointAfterDef());
Value *NSV0A = Builder.CreateShuffleVector(GetShuffleOperand(SVI0A, 0),
GetShuffleOperand(SVI0A, 1), V1A);
Builder.SetInsertPoint(*SVI0B->getInsertionPointAfterDef());
Value *NSV0B = Builder.CreateShuffleVector(GetShuffleOperand(SVI0B, 0),
GetShuffleOperand(SVI0B, 1), V1B);
Builder.SetInsertPoint(*SVI1A->getInsertionPointAfterDef());
Value *NSV1A = Builder.CreateShuffleVector(GetShuffleOperand(SVI1A, 0),
GetShuffleOperand(SVI1A, 1), V2A);
Builder.SetInsertPoint(*SVI1B->getInsertionPointAfterDef());
Value *NSV1B = Builder.CreateShuffleVector(GetShuffleOperand(SVI1B, 0),
GetShuffleOperand(SVI1B, 1), V2B);
Builder.SetInsertPoint(Op0);
Value *NOp0 = Builder.CreateBinOp((Instruction::BinaryOps)Op0->getOpcode(),
NSV0A, NSV0B);
if (auto *I = dyn_cast<Instruction>(NOp0))
I->copyIRFlags(Op0, true);
Builder.SetInsertPoint(Op1);
Value *NOp1 = Builder.CreateBinOp((Instruction::BinaryOps)Op1->getOpcode(),
NSV1A, NSV1B);
if (auto *I = dyn_cast<Instruction>(NOp1))
I->copyIRFlags(Op1, true);
for (int S = 0, E = ReconstructMasks.size(); S != E; S++) {
Builder.SetInsertPoint(Shuffles[S]);
Value *NSV = Builder.CreateShuffleVector(NOp0, NOp1, ReconstructMasks[S]);
replaceValue(*Shuffles[S], *NSV);
}
Worklist.pushValue(NSV0A);
Worklist.pushValue(NSV0B);
Worklist.pushValue(NSV1A);
Worklist.pushValue(NSV1B);
return true;
}
/// Check if instruction depends on ZExt and this ZExt can be moved after the
/// instruction. Move ZExt if it is profitable. For example:
/// logic(zext(x),y) -> zext(logic(x,trunc(y)))
/// lshr((zext(x),y) -> zext(lshr(x,trunc(y)))
/// Cost model calculations takes into account if zext(x) has other users and
/// whether it can be propagated through them too.
bool VectorCombine::shrinkType(Instruction &I) {
Value *ZExted, *OtherOperand;
if (!match(&I, m_c_BitwiseLogic(m_ZExt(m_Value(ZExted)),
m_Value(OtherOperand))) &&
!match(&I, m_LShr(m_ZExt(m_Value(ZExted)), m_Value(OtherOperand))))
return false;
Value *ZExtOperand = I.getOperand(I.getOperand(0) == OtherOperand ? 1 : 0);
auto *BigTy = cast<FixedVectorType>(I.getType());
auto *SmallTy = cast<FixedVectorType>(ZExted->getType());
unsigned BW = SmallTy->getElementType()->getPrimitiveSizeInBits();
if (I.getOpcode() == Instruction::LShr) {
// Check that the shift amount is less than the number of bits in the
// smaller type. Otherwise, the smaller lshr will return a poison value.
KnownBits ShAmtKB = computeKnownBits(I.getOperand(1), *DL);
if (ShAmtKB.getMaxValue().uge(BW))
return false;
} else {
// Check that the expression overall uses at most the same number of bits as
// ZExted
KnownBits KB = computeKnownBits(&I, *DL);
if (KB.countMaxActiveBits() > BW)
return false;
}
// Calculate costs of leaving current IR as it is and moving ZExt operation
// later, along with adding truncates if needed
InstructionCost ZExtCost = TTI.getCastInstrCost(
Instruction::ZExt, BigTy, SmallTy,
TargetTransformInfo::CastContextHint::None, CostKind);
InstructionCost CurrentCost = ZExtCost;
InstructionCost ShrinkCost = 0;
// Calculate total cost and check that we can propagate through all ZExt users
for (User *U : ZExtOperand->users()) {
auto *UI = cast<Instruction>(U);
if (UI == &I) {
CurrentCost +=
TTI.getArithmeticInstrCost(UI->getOpcode(), BigTy, CostKind);
ShrinkCost +=
TTI.getArithmeticInstrCost(UI->getOpcode(), SmallTy, CostKind);
ShrinkCost += ZExtCost;
continue;
}
if (!Instruction::isBinaryOp(UI->getOpcode()))
return false;
// Check if we can propagate ZExt through its other users
KnownBits KB = computeKnownBits(UI, *DL);
if (KB.countMaxActiveBits() > BW)
return false;
CurrentCost += TTI.getArithmeticInstrCost(UI->getOpcode(), BigTy, CostKind);
ShrinkCost +=
TTI.getArithmeticInstrCost(UI->getOpcode(), SmallTy, CostKind);
ShrinkCost += ZExtCost;
}
// If the other instruction operand is not a constant, we'll need to
// generate a truncate instruction. So we have to adjust cost
if (!isa<Constant>(OtherOperand))
ShrinkCost += TTI.getCastInstrCost(
Instruction::Trunc, SmallTy, BigTy,
TargetTransformInfo::CastContextHint::None, CostKind);
// If the cost of shrinking types and leaving the IR is the same, we'll lean
// towards modifying the IR because shrinking opens opportunities for other
// shrinking optimisations.
if (ShrinkCost > CurrentCost)
return false;
Builder.SetInsertPoint(&I);
Value *Op0 = ZExted;
Value *Op1 = Builder.CreateTrunc(OtherOperand, SmallTy);
// Keep the order of operands the same
if (I.getOperand(0) == OtherOperand)
std::swap(Op0, Op1);
Value *NewBinOp =
Builder.CreateBinOp((Instruction::BinaryOps)I.getOpcode(), Op0, Op1);
cast<Instruction>(NewBinOp)->copyIRFlags(&I);
cast<Instruction>(NewBinOp)->copyMetadata(I);
Value *NewZExtr = Builder.CreateZExt(NewBinOp, BigTy);
replaceValue(I, *NewZExtr);
return true;
}
/// insert (DstVec, (extract SrcVec, ExtIdx), InsIdx) -->
/// shuffle (DstVec, SrcVec, Mask)
bool VectorCombine::foldInsExtVectorToShuffle(Instruction &I) {
Value *DstVec, *SrcVec;
uint64_t ExtIdx, InsIdx;
if (!match(&I,
m_InsertElt(m_Value(DstVec),
m_ExtractElt(m_Value(SrcVec), m_ConstantInt(ExtIdx)),
m_ConstantInt(InsIdx))))
return false;
auto *DstVecTy = dyn_cast<FixedVectorType>(I.getType());
auto *SrcVecTy = dyn_cast<FixedVectorType>(SrcVec->getType());
// We can try combining vectors with different element sizes.
if (!DstVecTy || !SrcVecTy ||
SrcVecTy->getElementType() != DstVecTy->getElementType())
return false;
unsigned NumDstElts = DstVecTy->getNumElements();
unsigned NumSrcElts = SrcVecTy->getNumElements();
if (InsIdx >= NumDstElts || ExtIdx >= NumSrcElts || NumDstElts == 1)
return false;
// Insertion into poison is a cheaper single operand shuffle.
TargetTransformInfo::ShuffleKind SK;
SmallVector<int> Mask(NumDstElts, PoisonMaskElem);
bool NeedExpOrNarrow = NumSrcElts != NumDstElts;
bool IsExtIdxInBounds = ExtIdx < NumDstElts;
bool NeedDstSrcSwap = isa<PoisonValue>(DstVec) && !isa<UndefValue>(SrcVec);
if (NeedDstSrcSwap) {
SK = TargetTransformInfo::SK_PermuteSingleSrc;
if (!IsExtIdxInBounds && NeedExpOrNarrow)
Mask[InsIdx] = 0;
else
Mask[InsIdx] = ExtIdx;
std::swap(DstVec, SrcVec);
} else {
SK = TargetTransformInfo::SK_PermuteTwoSrc;
std::iota(Mask.begin(), Mask.end(), 0);
if (!IsExtIdxInBounds && NeedExpOrNarrow)
Mask[InsIdx] = NumDstElts;
else
Mask[InsIdx] = ExtIdx + NumDstElts;
}
// Cost
auto *Ins = cast<InsertElementInst>(&I);
auto *Ext = cast<ExtractElementInst>(I.getOperand(1));
InstructionCost InsCost =
TTI.getVectorInstrCost(*Ins, DstVecTy, CostKind, InsIdx);
InstructionCost ExtCost =
TTI.getVectorInstrCost(*Ext, DstVecTy, CostKind, ExtIdx);
InstructionCost OldCost = ExtCost + InsCost;
InstructionCost NewCost = 0;
SmallVector<int> ExtToVecMask;
if (!NeedExpOrNarrow) {
// Ignore 'free' identity insertion shuffle.
// TODO: getShuffleCost should return TCC_Free for Identity shuffles.
if (!ShuffleVectorInst::isIdentityMask(Mask, NumSrcElts))
NewCost += TTI.getShuffleCost(SK, DstVecTy, Mask, CostKind, 0, nullptr,
{DstVec, SrcVec});
} else {
// When creating length-changing-vector, always create with a Mask whose
// first element has an ExtIdx, so that the first element of the vector
// being created is always the target to be extracted.
ExtToVecMask.assign(NumDstElts, PoisonMaskElem);
if (IsExtIdxInBounds)
ExtToVecMask[ExtIdx] = ExtIdx;
else
ExtToVecMask[0] = ExtIdx;
// Add cost for expanding or narrowing
NewCost = TTI.getShuffleCost(TargetTransformInfo::SK_PermuteSingleSrc,
SrcVecTy, ExtToVecMask, CostKind);
NewCost += TTI.getShuffleCost(SK, DstVecTy, Mask, CostKind);
}
if (!Ext->hasOneUse())
NewCost += ExtCost;
LLVM_DEBUG(dbgs() << "Found a insert/extract shuffle-like pair: " << I
<< "\n OldCost: " << OldCost << " vs NewCost: " << NewCost
<< "\n");
if (OldCost < NewCost)
return false;
if (NeedExpOrNarrow) {
if (!NeedDstSrcSwap)
SrcVec = Builder.CreateShuffleVector(SrcVec, ExtToVecMask);
else
DstVec = Builder.CreateShuffleVector(DstVec, ExtToVecMask);
}
// Canonicalize undef param to RHS to help further folds.
if (isa<UndefValue>(DstVec) && !isa<UndefValue>(SrcVec)) {
ShuffleVectorInst::commuteShuffleMask(Mask, NumDstElts);
std::swap(DstVec, SrcVec);
}
Value *Shuf = Builder.CreateShuffleVector(DstVec, SrcVec, Mask);
replaceValue(I, *Shuf);
return true;
}
/// If we're interleaving 2 constant splats, for instance `<vscale x 8 x i32>
/// <splat of 666>` and `<vscale x 8 x i32> <splat of 777>`, we can create a
/// larger splat `<vscale x 8 x i64> <splat of ((777 << 32) | 666)>` first
/// before casting it back into `<vscale x 16 x i32>`.
bool VectorCombine::foldInterleaveIntrinsics(Instruction &I) {
const APInt *SplatVal0, *SplatVal1;
if (!match(&I, m_Intrinsic<Intrinsic::vector_interleave2>(
m_APInt(SplatVal0), m_APInt(SplatVal1))))
return false;
LLVM_DEBUG(dbgs() << "VC: Folding interleave2 with two splats: " << I
<< "\n");
auto *VTy =
cast<VectorType>(cast<IntrinsicInst>(I).getArgOperand(0)->getType());
auto *ExtVTy = VectorType::getExtendedElementVectorType(VTy);
unsigned Width = VTy->getElementType()->getIntegerBitWidth();
// Just in case the cost of interleave2 intrinsic and bitcast are both
// invalid, in which case we want to bail out, we use <= rather
// than < here. Even they both have valid and equal costs, it's probably
// not a good idea to emit a high-cost constant splat.
if (TTI.getInstructionCost(&I, CostKind) <=
TTI.getCastInstrCost(Instruction::BitCast, I.getType(), ExtVTy,
TTI::CastContextHint::None, CostKind)) {
LLVM_DEBUG(dbgs() << "VC: The cost to cast from " << *ExtVTy << " to "
<< *I.getType() << " is too high.\n");
return false;
}
APInt NewSplatVal = SplatVal1->zext(Width * 2);
NewSplatVal <<= Width;
NewSplatVal |= SplatVal0->zext(Width * 2);
auto *NewSplat = ConstantVector::getSplat(
ExtVTy->getElementCount(), ConstantInt::get(F.getContext(), NewSplatVal));
IRBuilder<> Builder(&I);
replaceValue(I, *Builder.CreateBitCast(NewSplat, I.getType()));
return true;
}
/// This is the entry point for all transforms. Pass manager differences are
/// handled in the callers of this function.
bool VectorCombine::run() {
if (DisableVectorCombine)
return false;
// Don't attempt vectorization if the target does not support vectors.
if (!TTI.getNumberOfRegisters(TTI.getRegisterClassForType(/*Vector*/ true)))
return false;
LLVM_DEBUG(dbgs() << "\n\nVECTORCOMBINE on " << F.getName() << "\n");
bool MadeChange = false;
auto FoldInst = [this, &MadeChange](Instruction &I) {
Builder.SetInsertPoint(&I);
bool IsVectorType = isa<VectorType>(I.getType());
bool IsFixedVectorType = isa<FixedVectorType>(I.getType());
auto Opcode = I.getOpcode();
LLVM_DEBUG(dbgs() << "VC: Visiting: " << I << '\n');
// These folds should be beneficial regardless of when this pass is run
// in the optimization pipeline.
// The type checking is for run-time efficiency. We can avoid wasting time
// dispatching to folding functions if there's no chance of matching.
if (IsFixedVectorType) {
switch (Opcode) {
case Instruction::InsertElement:
MadeChange |= vectorizeLoadInsert(I);
break;
case Instruction::ShuffleVector:
MadeChange |= widenSubvectorLoad(I);
break;
default:
break;
}
}
// This transform works with scalable and fixed vectors
// TODO: Identify and allow other scalable transforms
if (IsVectorType) {
MadeChange |= scalarizeOpOrCmp(I);
MadeChange |= scalarizeLoadExtract(I);
MadeChange |= scalarizeVPIntrinsic(I);
MadeChange |= foldInterleaveIntrinsics(I);
}
if (Opcode == Instruction::Store)
MadeChange |= foldSingleElementStore(I);
// If this is an early pipeline invocation of this pass, we are done.
if (TryEarlyFoldsOnly)
return;
// Otherwise, try folds that improve codegen but may interfere with
// early IR canonicalizations.
// The type checking is for run-time efficiency. We can avoid wasting time
// dispatching to folding functions if there's no chance of matching.
if (IsFixedVectorType) {
switch (Opcode) {
case Instruction::InsertElement:
MadeChange |= foldInsExtFNeg(I);
MadeChange |= foldInsExtBinop(I);
MadeChange |= foldInsExtVectorToShuffle(I);
break;
case Instruction::ShuffleVector:
MadeChange |= foldPermuteOfBinops(I);
MadeChange |= foldShuffleOfBinops(I);
MadeChange |= foldShuffleOfSelects(I);
MadeChange |= foldShuffleOfCastops(I);
MadeChange |= foldShuffleOfShuffles(I);
MadeChange |= foldShuffleOfIntrinsics(I);
MadeChange |= foldSelectShuffle(I);
MadeChange |= foldShuffleToIdentity(I);
break;
case Instruction::BitCast:
MadeChange |= foldBitcastShuffle(I);
break;
default:
MadeChange |= shrinkType(I);
break;
}
} else {
switch (Opcode) {
case Instruction::Call:
MadeChange |= foldShuffleFromReductions(I);
MadeChange |= foldCastFromReductions(I);
break;
case Instruction::ICmp:
case Instruction::FCmp:
MadeChange |= foldExtractExtract(I);
break;
case Instruction::Or:
MadeChange |= foldConcatOfBoolMasks(I);
[[fallthrough]];
default:
if (Instruction::isBinaryOp(Opcode)) {
MadeChange |= foldExtractExtract(I);
MadeChange |= foldExtractedCmps(I);
MadeChange |= foldBinopOfReductions(I);
}
break;
}
}
};
for (BasicBlock &BB : F) {
// Ignore unreachable basic blocks.
if (!DT.isReachableFromEntry(&BB))
continue;
// Use early increment range so that we can erase instructions in loop.
for (Instruction &I : make_early_inc_range(BB)) {
if (I.isDebugOrPseudoInst())
continue;
FoldInst(I);
}
}
while (!Worklist.isEmpty()) {
Instruction *I = Worklist.removeOne();
if (!I)
continue;
if (isInstructionTriviallyDead(I)) {
eraseInstruction(*I);
continue;
}
FoldInst(*I);
}
return MadeChange;
}
PreservedAnalyses VectorCombinePass::run(Function &F,
FunctionAnalysisManager &FAM) {
auto &AC = FAM.getResult<AssumptionAnalysis>(F);
TargetTransformInfo &TTI = FAM.getResult<TargetIRAnalysis>(F);
DominatorTree &DT = FAM.getResult<DominatorTreeAnalysis>(F);
AAResults &AA = FAM.getResult<AAManager>(F);
const DataLayout *DL = &F.getDataLayout();
VectorCombine Combiner(F, TTI, DT, AA, AC, DL, TTI::TCK_RecipThroughput,
TryEarlyFoldsOnly);
if (!Combiner.run())
return PreservedAnalyses::all();
PreservedAnalyses PA;
PA.preserveSet<CFGAnalyses>();
return PA;
}