diff options
| author | Tom Stellard <tstellar@redhat.com> | 2018-10-19 19:56:57 +0000 |
|---|---|---|
| committer | Tom Stellard <tstellar@redhat.com> | 2018-10-19 19:56:57 +0000 |
| commit | cddf73ef269ca589993366e1f7d6662b8976d6ed (patch) | |
| tree | 0da816db4e00ab89578f82ec383c555fe3796057 /llvm/test/CodeGen/X86/known-signbits-vector.ll | |
| parent | 80adf409497be8b6a80c8e72ae0e208efd730893 (diff) | |
| download | bcm5719-llvm-cddf73ef269ca589993366e1f7d6662b8976d6ed.tar.gz bcm5719-llvm-cddf73ef269ca589993366e1f7d6662b8976d6ed.zip | |
Merging r343373:
------------------------------------------------------------------------
r343373 | rksimon | 2018-09-29 06:25:22 -0700 (Sat, 29 Sep 2018) | 3 lines
[X86][SSE] Fixed issue with v2i64 variable shifts on 32-bit targets
The shift amount might have peeked through a extract_subvector, altering the number of vector elements in the 'Amt' variable - so we were incorrectly calculating the ratio when peeking through bitcasts, resulting in incorrectly detecting splats.
------------------------------------------------------------------------
llvm-svn: 344810
Diffstat (limited to 'llvm/test/CodeGen/X86/known-signbits-vector.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/known-signbits-vector.ll | 29 |
1 files changed, 18 insertions, 11 deletions
diff --git a/llvm/test/CodeGen/X86/known-signbits-vector.ll b/llvm/test/CodeGen/X86/known-signbits-vector.ll index d243616fc45..1e48f8683c3 100644 --- a/llvm/test/CodeGen/X86/known-signbits-vector.ll +++ b/llvm/test/CodeGen/X86/known-signbits-vector.ll @@ -381,19 +381,26 @@ define <4 x float> @signbits_ashr_sext_select_shuffle_sitofp(<4 x i64> %a0, <4 x ; X32-NEXT: movl %esp, %ebp ; X32-NEXT: andl $-16, %esp ; X32-NEXT: subl $16, %esp +; X32-NEXT: vmovdqa {{.*#+}} xmm3 = [33,0,63,0] +; X32-NEXT: vmovdqa {{.*#+}} xmm4 = [0,2147483648,0,2147483648] +; X32-NEXT: vpsrlq %xmm3, %xmm4, %xmm5 +; X32-NEXT: vpshufd {{.*#+}} xmm6 = xmm3[2,3,0,1] +; X32-NEXT: vpsrlq %xmm6, %xmm4, %xmm4 +; X32-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2,3],xmm4[4,5,6,7] +; X32-NEXT: vextractf128 $1, %ymm2, %xmm5 +; X32-NEXT: vpsrlq %xmm6, %xmm5, %xmm7 +; X32-NEXT: vpsrlq %xmm3, %xmm5, %xmm5 +; X32-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3],xmm7[4,5,6,7] +; X32-NEXT: vpsrlq %xmm6, %xmm2, %xmm6 +; X32-NEXT: vpsrlq %xmm3, %xmm2, %xmm2 +; X32-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm6[4,5,6,7] ; X32-NEXT: vpmovsxdq 16(%ebp), %xmm3 +; X32-NEXT: vpxor %xmm4, %xmm5, %xmm5 +; X32-NEXT: vpsubq %xmm4, %xmm5, %xmm5 +; X32-NEXT: vpxor %xmm4, %xmm2, %xmm2 +; X32-NEXT: vpsubq %xmm4, %xmm2, %xmm2 ; X32-NEXT: vpmovsxdq 8(%ebp), %xmm4 -; X32-NEXT: vmovdqa {{.*#+}} xmm5 = [33,0,63,0] -; X32-NEXT: vmovdqa {{.*#+}} xmm6 = [0,2147483648,0,2147483648] -; X32-NEXT: vpsrlq %xmm5, %xmm6, %xmm6 -; X32-NEXT: vextractf128 $1, %ymm2, %xmm7 -; X32-NEXT: vpsrlq %xmm5, %xmm7, %xmm7 -; X32-NEXT: vpxor %xmm6, %xmm7, %xmm7 -; X32-NEXT: vpsubq %xmm6, %xmm7, %xmm7 -; X32-NEXT: vpsrlq %xmm5, %xmm2, %xmm2 -; X32-NEXT: vpxor %xmm6, %xmm2, %xmm2 -; X32-NEXT: vpsubq %xmm6, %xmm2, %xmm2 -; X32-NEXT: vinsertf128 $1, %xmm7, %ymm2, %ymm2 +; X32-NEXT: vinsertf128 $1, %xmm5, %ymm2, %ymm2 ; X32-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm3 ; X32-NEXT: vextractf128 $1, %ymm1, %xmm4 ; X32-NEXT: vextractf128 $1, %ymm0, %xmm5 |
