mirror of
https://github.com/RPCSX/llvm.git
synced 2025-01-19 19:03:50 +00:00
Fix an optimization involving EXTRACT_SUBVECTOR in DAGCombine so it behaves correctly. PR11494.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@145996 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
parent
c9040b3b13
commit
26323442d5
@ -7181,19 +7181,23 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode* N) {
|
|||||||
if (NVT != SmallVT || NVT.getSizeInBits()*2 != BigVT.getSizeInBits())
|
if (NVT != SmallVT || NVT.getSizeInBits()*2 != BigVT.getSizeInBits())
|
||||||
return SDValue();
|
return SDValue();
|
||||||
|
|
||||||
// Combine:
|
// Only handle cases where both indexes are constants with the same type.
|
||||||
// (extract_subvec (insert_subvec V1, V2, InsIdx), ExtIdx)
|
ConstantSDNode *InsIdx = dyn_cast<ConstantSDNode>(N->getOperand(1));
|
||||||
// Into:
|
ConstantSDNode *ExtIdx = dyn_cast<ConstantSDNode>(V->getOperand(2));
|
||||||
// indicies are equal => V1
|
|
||||||
// otherwise => (extract_subvec V1, ExtIdx)
|
|
||||||
//
|
|
||||||
SDValue InsIdx = N->getOperand(1);
|
|
||||||
SDValue ExtIdx = V->getOperand(2);
|
|
||||||
|
|
||||||
if (InsIdx == ExtIdx)
|
if (InsIdx && ExtIdx &&
|
||||||
return V->getOperand(1);
|
InsIdx->getValueType(0).getSizeInBits() <= 64 &&
|
||||||
return DAG.getNode(ISD::EXTRACT_SUBVECTOR, N->getDebugLoc(), NVT,
|
ExtIdx->getValueType(0).getSizeInBits() <= 64) {
|
||||||
V->getOperand(0), N->getOperand(1));
|
// Combine:
|
||||||
|
// (extract_subvec (insert_subvec V1, V2, InsIdx), ExtIdx)
|
||||||
|
// Into:
|
||||||
|
// indices are equal => V1
|
||||||
|
// otherwise => (extract_subvec V1, ExtIdx)
|
||||||
|
if (InsIdx->getZExtValue() == ExtIdx->getZExtValue())
|
||||||
|
return V->getOperand(1);
|
||||||
|
return DAG.getNode(ISD::EXTRACT_SUBVECTOR, N->getDebugLoc(), NVT,
|
||||||
|
V->getOperand(0), N->getOperand(1));
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
return SDValue();
|
return SDValue();
|
||||||
|
18
test/CodeGen/X86/2011-12-06-AVXVectorExtractCombine.ll
Normal file
18
test/CodeGen/X86/2011-12-06-AVXVectorExtractCombine.ll
Normal file
@ -0,0 +1,18 @@
|
|||||||
|
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx | FileCheck %s
|
||||||
|
; PR11494
|
||||||
|
|
||||||
|
define void @test(<4 x i32>* nocapture %p) nounwind {
|
||||||
|
; CHECK: test:
|
||||||
|
; CHECK: vpxor %xmm0, %xmm0, %xmm0
|
||||||
|
; CHECK-NEXT: vpmaxsd {{.*}}, %xmm0, %xmm0
|
||||||
|
; CHECK-NEXT: vmovdqu %xmm0, (%rdi)
|
||||||
|
; CHECK-NEXT: ret
|
||||||
|
%a = call <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32> <i32 -8, i32 -9, i32 -10, i32 -11>, <4 x i32> zeroinitializer) nounwind
|
||||||
|
%b = shufflevector <4 x i32> %a, <4 x i32> undef, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 0, i32 1, i32 2, i32 3>
|
||||||
|
%c = shufflevector <8 x i32> %b, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
|
||||||
|
store <4 x i32> %c, <4 x i32>* %p, align 1
|
||||||
|
ret void
|
||||||
|
}
|
||||||
|
|
||||||
|
declare <4 x i32> @llvm.x86.sse41.pminsd(<4 x i32>, <4 x i32>) nounwind readnone
|
||||||
|
declare <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32>, <4 x i32>) nounwind readnone
|
Loading…
x
Reference in New Issue
Block a user