[llvm] r182909 - Swizzle vector inputs if it helps us eliminate shuffles.
Evgeniy Stepanov
eugeni.stepanov at gmail.com
Thu May 30 01:58:28 PDT 2013
A simple reproducer:
class S {
public:
char x[1024];
};
S s[16];
S *p[16];
void f(void) {
for (unsigned i = 0; i != 16; ++i) {
p[i] = s + i;
}
}
With -O3:
%0 = getelementptr inbounds [16 x %class.S]* @s, i64 0, i64 %index
%1 = insertelement <2 x %class.S*> undef, %class.S* %0, i32 0
%2 = insertelement <2 x %class.S*> %1, %class.S* getelementptr ([16
x %class.S]* @s, i64 0, i64 undef), i32 1
An unrelated issue: with -O2 this loop is fully unrolled, with -O3 it
is only 2x unrolled. Sounds like a phase order issue?
On Thu, May 30, 2013 at 12:49 PM, Evgeniy Stepanov
<eugeni.stepanov at gmail.com> wrote:
> Hey,
>
> with this change the loop in
> PartialDiagnostic::StorageAllocator::StorageAllocator is vectorized
> as:
>
> vector.body: ; preds =
> %vector.body, %vector.ph
> %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
> %11 = getelementptr inbounds
> %"class.clang::PartialDiagnostic::StorageAllocator"* %this, i64 0, i32
> 0, i64 %index, !dbg !3218
> %12 = insertelement <2 x
> %"struct.clang::PartialDiagnostic::Storage"*> undef,
> %"struct.clang::PartialDiagnostic::Storage"* %11, i32 0
> %13 = getelementptr inbounds
> %"class.clang::PartialDiagnostic::StorageAllocator"* %this, i64 0, i32
> 0, i64 undef, !dbg !3218
> %14 = insertelement <2 x
> %"struct.clang::PartialDiagnostic::Storage"*> %12,
> %"struct.clang::PartialDiagnostic::Storage"* %13, i32 1
> %15 = getelementptr inbounds
> %"class.clang::PartialDiagnostic::StorageAllocator"* %this, i64 0, i32
> 1, i64 %index, !dbg !3218
>
>
> "undef" in "%13 =" line does not look correct.
>
>
> On Thu, May 30, 2013 at 8:33 AM, Nick Lewycky <nicholas at mxc.ca> wrote:
>> Author: nicholas
>> Date: Wed May 29 23:33:38 2013
>> New Revision: 182909
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=182909&view=rev
>> Log:
>> Swizzle vector inputs if it helps us eliminate shuffles.
>>
>> Modified:
>> llvm/trunk/lib/Transforms/InstCombine/InstCombine.h
>> llvm/trunk/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
>> llvm/trunk/test/Transforms/InstCombine/vec_shuffle.ll
>>
>> Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombine.h
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombine.h?rev=182909&r1=182908&r2=182909&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/Transforms/InstCombine/InstCombine.h (original)
>> +++ llvm/trunk/lib/Transforms/InstCombine/InstCombine.h Wed May 29 23:33:38 2013
>> @@ -234,6 +234,7 @@ private:
>> bool WillNotOverflowSignedAdd(Value *LHS, Value *RHS);
>> Value *EmitGEPOffset(User *GEP);
>> Instruction *scalarizePHI(ExtractElementInst &EI, PHINode *PN);
>> + Value *EvaluateInDifferentElementOrder(Value *V, ArrayRef<int> Mask);
>>
>> public:
>> // InsertNewInstBefore - insert an instruction New before instruction Old
>>
>> Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineVectorOps.cpp?rev=182909&r1=182908&r2=182909&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/Transforms/InstCombine/InstCombineVectorOps.cpp (original)
>> +++ llvm/trunk/lib/Transforms/InstCombine/InstCombineVectorOps.cpp Wed May 29 23:33:38 2013
>> @@ -494,6 +494,241 @@ Instruction *InstCombiner::visitInsertEl
>> return 0;
>> }
>>
>> +/// Return true if we can evaluate the specified expression tree if the vector
>> +/// elements were shuffled in a different order.
>> +static bool CanEvaluateShuffled(Value *V, ArrayRef<int> Mask,
>> + unsigned Depth = 100) {
>> + // We can always reorder the elements of a constant.
>> + if (isa<Constant>(V))
>> + return true;
>> +
>> + // We won't reorder vector arguments. No IPO here.
>> + Instruction *I = dyn_cast<Instruction>(V);
>> + if (!I) return false;
>> +
>> + // Two users may expect different orders of the elements. Don't try it.
>> + if (!I->hasOneUse())
>> + return false;
>> +
>> + if (Depth == 0) return false;
>> +
>> + switch (I->getOpcode()) {
>> + case Instruction::Add:
>> + case Instruction::FAdd:
>> + case Instruction::Sub:
>> + case Instruction::FSub:
>> + case Instruction::Mul:
>> + case Instruction::FMul:
>> + case Instruction::UDiv:
>> + case Instruction::SDiv:
>> + case Instruction::FDiv:
>> + case Instruction::URem:
>> + case Instruction::SRem:
>> + case Instruction::FRem:
>> + case Instruction::Shl:
>> + case Instruction::LShr:
>> + case Instruction::AShr:
>> + case Instruction::And:
>> + case Instruction::Or:
>> + case Instruction::Xor:
>> + case Instruction::ICmp:
>> + case Instruction::FCmp:
>> + case Instruction::Trunc:
>> + case Instruction::ZExt:
>> + case Instruction::SExt:
>> + case Instruction::FPToUI:
>> + case Instruction::FPToSI:
>> + case Instruction::UIToFP:
>> + case Instruction::SIToFP:
>> + case Instruction::FPTrunc:
>> + case Instruction::FPExt:
>> + case Instruction::GetElementPtr: {
>> + for (int i = 0, e = I->getNumOperands(); i != e; ++i) {
>> + if (!CanEvaluateShuffled(I->getOperand(i), Mask, Depth-1))
>> + return false;
>> + }
>> + return true;
>> + }
>> + case Instruction::InsertElement: {
>> + ConstantInt *CI = dyn_cast<ConstantInt>(I->getOperand(2));
>> + if (!CI) return false;
>> + int ElementNumber = CI->getLimitedValue();
>> +
>> + // Verify that 'CI' does not occur twice in Mask. A single 'insertelement'
>> + // can't put an element into multiple indices.
>> + bool SeenOnce = false;
>> + for (int i = 0, e = Mask.size(); i != e; ++i) {
>> + if (Mask[i] == ElementNumber) {
>> + if (SeenOnce)
>> + return false;
>> + SeenOnce = true;
>> + }
>> + }
>> + return CanEvaluateShuffled(I->getOperand(0), Mask, Depth-1);
>> + }
>> + }
>> + return false;
>> +}
>> +
>> +/// Rebuild a new instruction just like 'I' but with the new operands given.
>> +/// In the event of type mismatch, the type of the operands is correct.
>> +static Value *BuildNew(Instruction *I, ArrayRef<Value*> NewOps) {
>> + // We don't want to use the IRBuilder here because we want the replacement
>> + // instructions to appear next to 'I', not the builder's insertion point.
>> + switch (I->getOpcode()) {
>> + case Instruction::Add:
>> + case Instruction::FAdd:
>> + case Instruction::Sub:
>> + case Instruction::FSub:
>> + case Instruction::Mul:
>> + case Instruction::FMul:
>> + case Instruction::UDiv:
>> + case Instruction::SDiv:
>> + case Instruction::FDiv:
>> + case Instruction::URem:
>> + case Instruction::SRem:
>> + case Instruction::FRem:
>> + case Instruction::Shl:
>> + case Instruction::LShr:
>> + case Instruction::AShr:
>> + case Instruction::And:
>> + case Instruction::Or:
>> + case Instruction::Xor: {
>> + BinaryOperator *BO = cast<BinaryOperator>(I);
>> + assert(NewOps.size() == 2 && "binary operator with #ops != 2");
>> + BinaryOperator *New =
>> + BinaryOperator::Create(cast<BinaryOperator>(I)->getOpcode(),
>> + NewOps[0], NewOps[1], "", BO);
>> + if (isa<OverflowingBinaryOperator>(BO)) {
>> + New->setHasNoUnsignedWrap(BO->hasNoUnsignedWrap());
>> + New->setHasNoSignedWrap(BO->hasNoSignedWrap());
>> + }
>> + if (isa<PossiblyExactOperator>(BO)) {
>> + New->setIsExact(BO->isExact());
>> + }
>> + return New;
>> + }
>> + case Instruction::ICmp:
>> + assert(NewOps.size() == 2 && "icmp with #ops != 2");
>> + return new ICmpInst(I, cast<ICmpInst>(I)->getPredicate(),
>> + NewOps[0], NewOps[1]);
>> + case Instruction::FCmp:
>> + assert(NewOps.size() == 2 && "fcmp with #ops != 2");
>> + return new FCmpInst(I, cast<FCmpInst>(I)->getPredicate(),
>> + NewOps[0], NewOps[1]);
>> + case Instruction::Trunc:
>> + case Instruction::ZExt:
>> + case Instruction::SExt:
>> + case Instruction::FPToUI:
>> + case Instruction::FPToSI:
>> + case Instruction::UIToFP:
>> + case Instruction::SIToFP:
>> + case Instruction::FPTrunc:
>> + case Instruction::FPExt: {
>> + // It's possible that the mask has a different number of elements from
>> + // the original cast. We recompute the destination type to match the mask.
>> + Type *DestTy =
>> + VectorType::get(I->getType()->getScalarType(),
>> + NewOps[0]->getType()->getVectorNumElements());
>> + assert(NewOps.size() == 1 && "cast with #ops != 1");
>> + return CastInst::Create(cast<CastInst>(I)->getOpcode(), NewOps[0], DestTy,
>> + "", I);
>> + }
>> + case Instruction::GetElementPtr: {
>> + Value *Ptr = NewOps[0];
>> + ArrayRef<Value*> Idx = NewOps.slice(1);
>> + GetElementPtrInst *GEP = GetElementPtrInst::Create(Ptr, Idx, "", I);
>> + GEP->setIsInBounds(cast<GetElementPtrInst>(I)->isInBounds());
>> + return GEP;
>> + }
>> + }
>> + llvm_unreachable("failed to rebuild vector instructions");
>> +}
>> +
>> +Value *
>> +InstCombiner::EvaluateInDifferentElementOrder(Value *V, ArrayRef<int> Mask) {
>> + // Mask.size() does not need to be equal to the number of vector elements.
>> +
>> + assert(V->getType()->isVectorTy() && "can't reorder non-vector elements");
>> + if (isa<UndefValue>(V)) {
>> + return UndefValue::get(VectorType::get(V->getType()->getScalarType(),
>> + Mask.size()));
>> + }
>> + if (isa<ConstantAggregateZero>(V)) {
>> + return ConstantAggregateZero::get(
>> + VectorType::get(V->getType()->getScalarType(),
>> + Mask.size()));
>> + }
>> + if (Constant *C = dyn_cast<Constant>(V)) {
>> + SmallVector<Constant *, 16> MaskValues;
>> + for (int i = 0, e = Mask.size(); i != e; ++i) {
>> + if (Mask[i] == -1)
>> + MaskValues.push_back(UndefValue::get(Builder->getInt32Ty()));
>> + else
>> + MaskValues.push_back(Builder->getInt32(Mask[i]));
>> + }
>> + return ConstantExpr::getShuffleVector(C, UndefValue::get(C->getType()),
>> + ConstantVector::get(MaskValues));
>> + }
>> +
>> + Instruction *I = cast<Instruction>(V);
>> + switch (I->getOpcode()) {
>> + case Instruction::Add:
>> + case Instruction::FAdd:
>> + case Instruction::Sub:
>> + case Instruction::FSub:
>> + case Instruction::Mul:
>> + case Instruction::FMul:
>> + case Instruction::UDiv:
>> + case Instruction::SDiv:
>> + case Instruction::FDiv:
>> + case Instruction::URem:
>> + case Instruction::SRem:
>> + case Instruction::FRem:
>> + case Instruction::Shl:
>> + case Instruction::LShr:
>> + case Instruction::AShr:
>> + case Instruction::And:
>> + case Instruction::Or:
>> + case Instruction::Xor:
>> + case Instruction::ICmp:
>> + case Instruction::FCmp:
>> + case Instruction::Trunc:
>> + case Instruction::ZExt:
>> + case Instruction::SExt:
>> + case Instruction::FPToUI:
>> + case Instruction::FPToSI:
>> + case Instruction::UIToFP:
>> + case Instruction::SIToFP:
>> + case Instruction::FPTrunc:
>> + case Instruction::FPExt:
>> + case Instruction::Select:
>> + case Instruction::GetElementPtr: {
>> + SmallVector<Value*, 8> NewOps;
>> + bool NeedsRebuild = (Mask.size() != I->getType()->getVectorNumElements());
>> + for (int i = 0, e = I->getNumOperands(); i != e; ++i) {
>> + Value *V = EvaluateInDifferentElementOrder(I->getOperand(i), Mask);
>> + NewOps.push_back(V);
>> + NeedsRebuild |= (V != I->getOperand(i));
>> + }
>> + if (NeedsRebuild) {
>> + return BuildNew(I, NewOps);
>> + }
>> + return I;
>> + }
>> + case Instruction::InsertElement: {
>> + uint32_t Element = cast<ConstantInt>(I->getOperand(2))->getLimitedValue();
>> + if (Element >= Mask.size()) {
>> + // Such instructions are valid and exhibit undefined behaviour.
>> + return UndefValue::get(I->getType());
>> + }
>> + Value *V = EvaluateInDifferentElementOrder(I->getOperand(0), Mask);
>> + return InsertElementInst::Create(V, I->getOperand(1),
>> + Builder->getInt32(Mask[Element]), "", I);
>> + }
>> + }
>> + llvm_unreachable("failed to reorder elements of vector instruction!");
>> +}
>>
>> Instruction *InstCombiner::visitShuffleVectorInst(ShuffleVectorInst &SVI) {
>> Value *LHS = SVI.getOperand(0);
>> @@ -574,6 +809,16 @@ Instruction *InstCombiner::visitShuffleV
>> if (isRHSID) return ReplaceInstUsesWith(SVI, RHS);
>> }
>>
>> + if (isa<UndefValue>(RHS) &&
>> + // This isn't necessary for correctness, but the comment block below
>> + // claims that there are cases where folding two shuffles into one would
>> + // cause worse codegen on some targets.
>> + !isa<ShuffleVectorInst>(LHS) &&
>> + CanEvaluateShuffled(LHS, Mask)) {
>> + Value *V = EvaluateInDifferentElementOrder(LHS, Mask);
>> + return ReplaceInstUsesWith(SVI, V);
>> + }
>> +
>> // If the LHS is a shufflevector itself, see if we can combine it with this
>> // one without producing an unusual shuffle.
>> // Cases that might be simplified:
>>
>> Modified: llvm/trunk/test/Transforms/InstCombine/vec_shuffle.ll
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/vec_shuffle.ll?rev=182909&r1=182908&r2=182909&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/Transforms/InstCombine/vec_shuffle.ll (original)
>> +++ llvm/trunk/test/Transforms/InstCombine/vec_shuffle.ll Wed May 29 23:33:38 2013
>> @@ -153,3 +153,15 @@ define <8 x i8> @test12a(<8 x i8> %tmp6,
>> ret <8 x i8> %tmp3
>> }
>>
>> +define <2 x i8> @test13(i8 %x1, i8 %x2) {
>> +; CHECK: @test13
>> +; CHECK-NEXT: insertelement {{.*}} undef, i8 %x1, i32 1
>> +; CHECK-NEXT: insertelement {{.*}} i8 %x2, i32 0
>> +; CHECK-NEXT: add {{.*}} <i8 7, i8 5>
>> +; CHECK-NEXT: ret
>> + %A = insertelement <2 x i8> undef, i8 %x1, i32 0
>> + %B = insertelement <2 x i8> %A, i8 %x2, i32 1
>> + %C = add <2 x i8> %B, <i8 5, i8 7>
>> + %D = shufflevector <2 x i8> %C, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
>> + ret <2 x i8> %D
>> +}
>>
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
More information about the llvm-commits
mailing list