diff options
| author | Roman Lebedev <lebedev.ri@gmail.com> | 2018-05-21 21:41:10 +0000 |
|---|---|---|
| committer | Roman Lebedev <lebedev.ri@gmail.com> | 2018-05-21 21:41:10 +0000 |
| commit | 9f65d16d5d06bbfc075cdd9c8635f5af88baaf8b (patch) | |
| tree | 2202debcdd70e71ae4d06e6cf159c4c8c4b6c05a /llvm/test/CodeGen/X86/unfold-masked-merge-vector-variablemask-const.ll | |
| parent | 7772de25d07c977e41f8faa3bbf327033cd81c20 (diff) | |
| download | bcm5719-llvm-9f65d16d5d06bbfc075cdd9c8635f5af88baaf8b.tar.gz bcm5719-llvm-9f65d16d5d06bbfc075cdd9c8635f5af88baaf8b.zip | |
[DAGCombiner] isAllOnesConstantOrAllOnesSplatConstant(): look through bitcasts
Summary:
As pointed out in D46528, we errneously transform cases like `xor X, -1`,
even though we use said function.
It's because the `-1` is actually a bitcast there.
So i think we can just look through it in the function.
Differential Revision: https://reviews.llvm.org/D47156
llvm-svn: 332905
Diffstat (limited to 'llvm/test/CodeGen/X86/unfold-masked-merge-vector-variablemask-const.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/unfold-masked-merge-vector-variablemask-const.ll | 31 |
1 files changed, 14 insertions, 17 deletions
diff --git a/llvm/test/CodeGen/X86/unfold-masked-merge-vector-variablemask-const.ll b/llvm/test/CodeGen/X86/unfold-masked-merge-vector-variablemask-const.ll index 20c3ef8ea88..4e50bf91a2c 100644 --- a/llvm/test/CodeGen/X86/unfold-masked-merge-vector-variablemask-const.ll +++ b/llvm/test/CodeGen/X86/unfold-masked-merge-vector-variablemask-const.ll @@ -58,20 +58,18 @@ define <4 x i32> @in_constant_varx_mone(<4 x i32> *%px, <4 x i32> *%py, <4 x i32 ; ; CHECK-SSE2-LABEL: in_constant_varx_mone: ; CHECK-SSE2: # %bb.0: -; CHECK-SSE2-NEXT: movdqa (%rdx), %xmm0 +; CHECK-SSE2-NEXT: movdqa (%rdi), %xmm0 ; CHECK-SSE2-NEXT: pcmpeqd %xmm1, %xmm1 -; CHECK-SSE2-NEXT: pxor %xmm0, %xmm1 -; CHECK-SSE2-NEXT: pand (%rdi), %xmm0 -; CHECK-SSE2-NEXT: por %xmm1, %xmm0 +; CHECK-SSE2-NEXT: pandn (%rdx), %xmm0 +; CHECK-SSE2-NEXT: pxor %xmm1, %xmm0 ; CHECK-SSE2-NEXT: retq ; ; CHECK-XOP-LABEL: in_constant_varx_mone: ; CHECK-XOP: # %bb.0: -; CHECK-XOP-NEXT: vmovdqa (%rdx), %xmm0 +; CHECK-XOP-NEXT: vmovdqa (%rdi), %xmm0 ; CHECK-XOP-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 -; CHECK-XOP-NEXT: vpxor %xmm1, %xmm0, %xmm1 -; CHECK-XOP-NEXT: vpand (%rdi), %xmm0, %xmm0 -; CHECK-XOP-NEXT: vpor %xmm1, %xmm0, %xmm0 +; CHECK-XOP-NEXT: vpandn (%rdx), %xmm0, %xmm0 +; CHECK-XOP-NEXT: vpxor %xmm1, %xmm0, %xmm0 ; CHECK-XOP-NEXT: retq %x = load <4 x i32>, <4 x i32> *%px, align 16 %y = load <4 x i32>, <4 x i32> *%py, align 16 @@ -134,22 +132,21 @@ define <4 x i32> @in_constant_varx_mone_invmask(<4 x i32> *%px, <4 x i32> *%py, ; ; CHECK-SSE2-LABEL: in_constant_varx_mone_invmask: ; CHECK-SSE2: # %bb.0: -; CHECK-SSE2-NEXT: movdqa (%rdx), %xmm1 -; CHECK-SSE2-NEXT: pcmpeqd %xmm0, %xmm0 -; CHECK-SSE2-NEXT: pxor %xmm0, %xmm0 +; CHECK-SSE2-NEXT: movdqa (%rdi), %xmm0 +; CHECK-SSE2-NEXT: pcmpeqd %xmm1, %xmm1 +; CHECK-SSE2-NEXT: movdqa (%rdx), %xmm2 +; CHECK-SSE2-NEXT: pxor %xmm1, %xmm2 +; CHECK-SSE2-NEXT: pandn %xmm2, %xmm0 ; CHECK-SSE2-NEXT: pxor %xmm1, %xmm0 -; CHECK-SSE2-NEXT: pandn (%rdi), %xmm1 -; CHECK-SSE2-NEXT: por %xmm1, %xmm0 ; CHECK-SSE2-NEXT: retq ; ; CHECK-XOP-LABEL: in_constant_varx_mone_invmask: ; CHECK-XOP: # %bb.0: -; CHECK-XOP-NEXT: vmovdqa (%rdx), %xmm0 +; CHECK-XOP-NEXT: vmovdqa (%rdi), %xmm0 ; CHECK-XOP-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 -; CHECK-XOP-NEXT: vpandn (%rdi), %xmm0, %xmm2 -; CHECK-XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1 +; CHECK-XOP-NEXT: vpxor (%rdx), %xmm1, %xmm2 +; CHECK-XOP-NEXT: vpandn %xmm2, %xmm0, %xmm0 ; CHECK-XOP-NEXT: vpxor %xmm1, %xmm0, %xmm0 -; CHECK-XOP-NEXT: vpor %xmm0, %xmm2, %xmm0 ; CHECK-XOP-NEXT: retq %x = load <4 x i32>, <4 x i32> *%px, align 16 %y = load <4 x i32>, <4 x i32> *%py, align 16 |
