diff options
| author | Simon Pilgrim <llvm-dev@redking.me.uk> | 2016-03-17 15:59:36 +0000 |
|---|---|---|
| committer | Simon Pilgrim <llvm-dev@redking.me.uk> | 2016-03-17 15:59:36 +0000 |
| commit | 0f37fbac51743051a5bc739778fc92da05efdc70 (patch) | |
| tree | 7ec2d4bb63f016e8f052cd12ce81a0781e542e18 /llvm/test/CodeGen/X86/vector-zext.ll | |
| parent | 9e23fedaf0f84d4a020514df4beefe65fe6da50b (diff) | |
| download | bcm5719-llvm-0f37fbac51743051a5bc739778fc92da05efdc70.tar.gz bcm5719-llvm-0f37fbac51743051a5bc739778fc92da05efdc70.zip | |
[X86][SSE] Simplified blend-with-zero combining
We were being too aggressive in trying to combine a shuffle into a blend-with-zero pattern, often resulting in a endless loop of contrasting combines
This patch stops the combine if we already have a blend in place (means we miss some domain corrections)
llvm-svn: 263717
Diffstat (limited to 'llvm/test/CodeGen/X86/vector-zext.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/vector-zext.ll | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/llvm/test/CodeGen/X86/vector-zext.ll b/llvm/test/CodeGen/X86/vector-zext.ll index ec1a4ecdc55..803244c7848 100644 --- a/llvm/test/CodeGen/X86/vector-zext.ll +++ b/llvm/test/CodeGen/X86/vector-zext.ll @@ -1151,8 +1151,8 @@ define <4 x i64> @shuf_zext_4i32_to_4i64(<4 x i32> %A) nounwind uwtable readnone ; AVX1-LABEL: shuf_zext_4i32_to_4i64: ; AVX1: # BB#0: # %entry ; AVX1-NEXT: vinsertps {{.*#+}} xmm1 = xmm0[0],zero,xmm0[1],zero -; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2 -; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1,2,3],xmm0[4,5,6,7] +; AVX1-NEXT: vxorpd %xmm2, %xmm2, %xmm2 +; AVX1-NEXT: vblendpd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,0,3,0] ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0 ; AVX1-NEXT: retq @@ -1581,8 +1581,8 @@ define <4 x i64> @shuf_zext_4i32_to_4i64_offset1(<4 x i32> %A) nounwind uwtable ; AVX1-LABEL: shuf_zext_4i32_to_4i64_offset1: ; AVX1: # BB#0: # %entry ; AVX1-NEXT: vinsertps {{.*#+}} xmm1 = xmm0[3],zero,zero,zero -; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2 -; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1,2,3],xmm0[4,5],xmm2[6,7] +; AVX1-NEXT: vxorps %xmm2, %xmm2, %xmm2 +; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm2[0,1],xmm0[2],xmm2[3] ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 ; AVX1-NEXT: retq ; |
