bbf4af3109
As detailed in PR43971/D70267, the use of XFormVExtractWithShuffleIntoLoad causes issues where we end up in infinite loops of extract(targetshuffle(vecload)) -> extract(shuffle(vecload)) -> extract(vecload) -> extract(targetshuffle(vecload)), there are just too many legalization checks at every stage that we can't guarantee that extract(shuffle(vecload)) -> scalarload can occur. At the moment we see a number of minor regressions as we don't fold extract(shuffle(vecload)) -> scalarload before legal ops, these can be addressed in future patches and extension of X86ISelLowering's combineExtractWithShuffle.
33 lines
1.1 KiB
LLVM
33 lines
1.1 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc < %s -mtriple=i686-- -mcpu=corei7 | FileCheck %s --check-prefixes=CHECK,X86
|
|
; RUN: llc < %s -mtriple=x86_64-- -mcpu=corei7 | FileCheck %s --check-prefixes=CHECK,X64
|
|
|
|
define float @test(<4 x float>* %A) nounwind {
|
|
; X86-LABEL: test:
|
|
; X86: # %bb.0: # %entry
|
|
; X86-NEXT: pushl %eax
|
|
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-NEXT: movaps (%eax), %xmm0
|
|
; X86-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3]
|
|
; X86-NEXT: xorps %xmm1, %xmm1
|
|
; X86-NEXT: movaps %xmm1, (%eax)
|
|
; X86-NEXT: movss %xmm0, (%esp)
|
|
; X86-NEXT: flds (%esp)
|
|
; X86-NEXT: popl %eax
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test:
|
|
; X64: # %bb.0: # %entry
|
|
; X64-NEXT: movaps (%rdi), %xmm0
|
|
; X64-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3]
|
|
; X64-NEXT: xorps %xmm1, %xmm1
|
|
; X64-NEXT: movaps %xmm1, (%rdi)
|
|
; X64-NEXT: retq
|
|
entry:
|
|
%T = load <4 x float>, <4 x float>* %A
|
|
%R = extractelement <4 x float> %T, i32 3
|
|
store <4 x float><float 0.0, float 0.0, float 0.0, float 0.0>, <4 x float>* %A
|
|
ret float %R
|
|
}
|
|
|