[llvm] r255137 - [InstCombine] fold bitcasts around an extractelement (2nd try)
Sanjay Patel via llvm-commits
llvm-commits at lists.llvm.org
Thu Dec 10 07:42:30 PST 2015
Thanks, Akira.
The first bitcast isn't necessary; I missed handling the case where the
source type of the bitcast feeding the extract isn't a vector itself.
I'll add a check + test case and try again!
On Thu, Dec 10, 2015 at 1:07 AM, Akira Hatanaka <ahatanak at gmail.com> wrote:
> Hi Sanjay,
>
> I reverted this in r255227.
>
> This causes an assert when the following IR is compiled with "opt
> -instcombine":
>
> define internal double @foo1(<2 x double> %a) {
>
> %1 = bitcast <2 x double> %a to i128
>
> %2 = bitcast i128 %1 to <2 x i64>
>
> %3 = extractelement <2 x i64> %2, i32 0
>
> %4 = bitcast i64 %3 to double
>
> ret double %4
>
> }
>
> On Wed, Dec 9, 2015 at 10:57 AM, Sanjay Patel via llvm-commits <
> llvm-commits at lists.llvm.org> wrote:
>
>> Author: spatel
>> Date: Wed Dec 9 12:57:16 2015
>> New Revision: 255137
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=255137&view=rev
>> Log:
>> [InstCombine] fold bitcasts around an extractelement (2nd try)
>>
>> This is a redo of r255124 (reverted at r255126) with an added check for a
>> scalar destination type and an added test for the failure seen in Clang's
>> test/CodeGen/vector.c. The extra test shows a different missing
>> optimization.
>>
>> Original commit message:
>>
>> Example:
>> bitcast (extractelement (bitcast <2 x float> %X to <2 x i32>), 1) to
>> float
>> --->
>> extractelement <2 x float> %X, i32 1
>>
>> This is part of fixing PR25543:
>> https://llvm.org/bugs/show_bug.cgi?id=25543
>>
>> The next step will be to generalize this fold:
>> trunc ( lshr ( bitcast X) ) -> extractelement (X)
>>
>> Ie, I'm hoping to replace the existing transform of:
>> bitcast ( trunc ( lshr ( bitcast X)))
>> added by:
>> http://reviews.llvm.org/rL112232
>>
>> with 2 less specific transforms to catch the case in the bug report.
>>
>> Differential Revision: http://reviews.llvm.org/D14879
>>
>>
>> Modified:
>> llvm/trunk/lib/Transforms/InstCombine/InstCombineCasts.cpp
>> llvm/trunk/test/Transforms/InstCombine/bitcast.ll
>>
>> Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCasts.cpp
>> URL:
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCasts.cpp?rev=255137&r1=255136&r2=255137&view=diff
>>
>> ==============================================================================
>> --- llvm/trunk/lib/Transforms/InstCombine/InstCombineCasts.cpp (original)
>> +++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCasts.cpp Wed Dec 9
>> 12:57:16 2015
>> @@ -1715,6 +1715,42 @@ static Value *optimizeIntegerToVectorIns
>> return Result;
>> }
>>
>> +/// Given a bitcasted vector fed into an extract element instruction and
>> then
>> +/// bitcasted again to a scalar type, eliminate at least one bitcast by
>> changing
>> +/// the vector type of the extractelement instruction.
>> +/// Example:
>> +/// bitcast (extractelement (bitcast <2 x float> %X to <2 x i32>), 1)
>> to float
>> +/// --->
>> +/// extractelement <2 x float> %X, i32 1
>> +static Instruction *foldBitCastExtElt(BitCastInst &BitCast, InstCombiner
>> &IC,
>> + const DataLayout &DL) {
>> + Type *DestType = BitCast.getType();
>> + if (DestType->isVectorTy())
>> + return nullptr;
>> +
>> + // TODO: Create and use a pattern matcher for ExtractElementInst.
>> + auto *ExtElt = dyn_cast<ExtractElementInst>(BitCast.getOperand(0));
>> + if (!ExtElt || !ExtElt->hasOneUse())
>> + return nullptr;
>> +
>> + Value *InnerBitCast = nullptr;
>> + if (!match(ExtElt->getOperand(0), m_BitCast(m_Value(InnerBitCast))))
>> + return nullptr;
>> +
>> + // If the element type of the vector doesn't match the result type,
>> + // bitcast it to a vector type that we can extract from.
>> + VectorType *VecType = cast<VectorType>(InnerBitCast->getType());
>> + if (VecType->getElementType() != DestType) {
>> + unsigned VecWidth = VecType->getPrimitiveSizeInBits();
>> + unsigned DestWidth = DestType->getPrimitiveSizeInBits();
>> + unsigned NumElts = VecWidth / DestWidth;
>> + VecType = VectorType::get(DestType, NumElts);
>> + InnerBitCast = IC.Builder->CreateBitCast(InnerBitCast, VecType,
>> "bc");
>> + }
>> +
>> + return ExtractElementInst::Create(InnerBitCast, ExtElt->getOperand(1));
>> +}
>> +
>> static Instruction *foldVecTruncToExtElt(Value *VecInput, Type *DestTy,
>> unsigned ShiftAmt, InstCombiner
>> &IC,
>> const DataLayout &DL) {
>> @@ -1886,6 +1922,9 @@ Instruction *InstCombiner::visitBitCast(
>> }
>> }
>>
>> + if (Instruction *I = foldBitCastExtElt(CI, *this, DL))
>> + return I;
>> +
>> if (SrcTy->isPointerTy())
>> return commonPointerCastTransforms(CI);
>> return commonCastTransforms(CI);
>>
>> Modified: llvm/trunk/test/Transforms/InstCombine/bitcast.ll
>> URL:
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/bitcast.ll?rev=255137&r1=255136&r2=255137&view=diff
>>
>> ==============================================================================
>> --- llvm/trunk/test/Transforms/InstCombine/bitcast.ll (original)
>> +++ llvm/trunk/test/Transforms/InstCombine/bitcast.ll Wed Dec 9 12:57:16
>> 2015
>> @@ -64,7 +64,7 @@ define float @test3(<2 x float> %A, <2 x
>> ; CHECK-NEXT: ret float %add
>> }
>>
>> -; TODO: Both bitcasts are unnecessary; change the extractelement.
>> +; Both bitcasts are unnecessary; change the extractelement.
>>
>> define float @bitcast_extelt1(<2 x float> %A) {
>> %bc1 = bitcast <2 x float> %A to <2 x i32>
>> @@ -73,13 +73,11 @@ define float @bitcast_extelt1(<2 x float
>> ret float %bc2
>>
>> ; CHECK-LABEL: @bitcast_extelt1(
>> -; CHECK-NEXT: %bc1 = bitcast <2 x float> %A to <2 x i32>
>> -; CHECK-NEXT: %ext = extractelement <2 x i32> %bc1, i32 0
>> -; CHECK-NEXT: %bc2 = bitcast i32 %ext to float
>> +; CHECK-NEXT: %bc2 = extractelement <2 x float> %A, i32 0
>> ; CHECK-NEXT: ret float %bc2
>> }
>>
>> -; TODO: Second bitcast can be folded into the first.
>> +; Second bitcast can be folded into the first.
>>
>> define i64 @bitcast_extelt2(<4 x float> %A) {
>> %bc1 = bitcast <4 x float> %A to <2 x double>
>> @@ -88,12 +86,26 @@ define i64 @bitcast_extelt2(<4 x float>
>> ret i64 %bc2
>>
>> ; CHECK-LABEL: @bitcast_extelt2(
>> -; CHECK-NEXT: %bc1 = bitcast <4 x float> %A to <2 x double>
>> -; CHECK-NEXT: %ext = extractelement <2 x double> %bc1, i32 1
>> -; CHECK-NEXT: %bc2 = bitcast double %ext to i64
>> +; CHECK-NEXT: %bc = bitcast <4 x float> %A to <2 x i64>
>> +; CHECK-NEXT: %bc2 = extractelement <2 x i64> %bc, i32 1
>> ; CHECK-NEXT: ret i64 %bc2
>> }
>>
>> +; TODO: This should return %A.
>> +
>> +define <2 x i32> @bitcast_extelt3(<2 x i32> %A) {
>> + %bc1 = bitcast <2 x i32> %A to <1 x i64>
>> + %ext = extractelement <1 x i64> %bc1, i32 0
>> + %bc2 = bitcast i64 %ext to <2 x i32>
>> + ret <2 x i32> %bc2
>> +
>> +; CHECK-LABEL: @bitcast_extelt3(
>> +; CHECK-NEXT: %bc1 = bitcast <2 x i32> %A to <1 x i64>
>> +; CHECK-NEXT: %ext = extractelement <1 x i64> %bc1, i32 0
>> +; CHECK-NEXT: %bc2 = bitcast i64 %ext to <2 x i32>
>> +; CHECK-NEXT: ret <2 x i32> %bc2
>> +}
>> +
>> define <2 x i32> @test4(i32 %A, i32 %B){
>> %tmp38 = zext i32 %A to i64
>> %tmp32 = zext i32 %B to i64
>>
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151210/947c8f96/attachment.html>
More information about the llvm-commits
mailing list