diff options
| author | Craig Topper <craig.topper@gmail.com> | 2016-07-25 07:20:35 +0000 |
|---|---|---|
| committer | Craig Topper <craig.topper@gmail.com> | 2016-07-25 07:20:35 +0000 |
| commit | ce415ff9c5a469aa75f6471ead4318d99f457ac9 (patch) | |
| tree | b672e498ea6ecc1f6d47d831188b033ca464d012 /llvm/test/CodeGen | |
| parent | 318e40b6f71559f97f604951cebcfbaf7ee86e6a (diff) | |
| download | bcm5719-llvm-ce415ff9c5a469aa75f6471ead4318d99f457ac9.tar.gz bcm5719-llvm-ce415ff9c5a469aa75f6471ead4318d99f457ac9.zip | |
[AVX512] Add load folding support for the unmasked forms of the FMA instructions.
llvm-svn: 276615
Diffstat (limited to 'llvm/test/CodeGen')
| -rw-r--r-- | llvm/test/CodeGen/X86/fma_patterns.ll | 6 | ||||
| -rw-r--r-- | llvm/test/CodeGen/X86/fma_patterns_wide.ll | 6 |
2 files changed, 4 insertions, 8 deletions
diff --git a/llvm/test/CodeGen/X86/fma_patterns.ll b/llvm/test/CodeGen/X86/fma_patterns.ll index b724da20425..371361ddeef 100644 --- a/llvm/test/CodeGen/X86/fma_patterns.ll +++ b/llvm/test/CodeGen/X86/fma_patterns.ll @@ -524,8 +524,7 @@ define <4 x float> @test_4f32_fmadd_load(<4 x float>* %a0, <4 x float> %a1, <4 x ; ; AVX512-LABEL: test_4f32_fmadd_load: ; AVX512: # BB#0: -; AVX512-NEXT: vmovaps (%rdi), %xmm2 -; AVX512-NEXT: vfmadd213ps %xmm1, %xmm2, %xmm0 +; AVX512-NEXT: vfmadd132ps (%rdi), %xmm1, %xmm0 ; AVX512-NEXT: retq %x = load <4 x float>, <4 x float>* %a0 %y = fmul <4 x float> %x, %a1 @@ -546,8 +545,7 @@ define <2 x double> @test_2f64_fmsub_load(<2 x double>* %a0, <2 x double> %a1, < ; ; AVX512-LABEL: test_2f64_fmsub_load: ; AVX512: # BB#0: -; AVX512-NEXT: vmovapd (%rdi), %xmm2 -; AVX512-NEXT: vfmsub213pd %xmm1, %xmm2, %xmm0 +; AVX512-NEXT: vfmsub132pd (%rdi), %xmm1, %xmm0 ; AVX512-NEXT: retq %x = load <2 x double>, <2 x double>* %a0 %y = fmul <2 x double> %x, %a1 diff --git a/llvm/test/CodeGen/X86/fma_patterns_wide.ll b/llvm/test/CodeGen/X86/fma_patterns_wide.ll index 98b6c496e08..2c49bb851c3 100644 --- a/llvm/test/CodeGen/X86/fma_patterns_wide.ll +++ b/llvm/test/CodeGen/X86/fma_patterns_wide.ll @@ -217,8 +217,7 @@ define <16 x float> @test_16f32_fmadd_load(<16 x float>* %a0, <16 x float> %a1, ; ; AVX512-LABEL: test_16f32_fmadd_load: ; AVX512: # BB#0: -; AVX512-NEXT: vmovaps (%rdi), %zmm2 -; AVX512-NEXT: vfmadd213ps %zmm1, %zmm2, %zmm0 +; AVX512-NEXT: vfmadd132ps (%rdi), %zmm1, %zmm0 ; AVX512-NEXT: retq %x = load <16 x float>, <16 x float>* %a0 %y = fmul <16 x float> %x, %a1 @@ -241,8 +240,7 @@ define <8 x double> @test_8f64_fmsub_load(<8 x double>* %a0, <8 x double> %a1, < ; ; AVX512-LABEL: test_8f64_fmsub_load: ; AVX512: # BB#0: -; AVX512-NEXT: vmovapd (%rdi), %zmm2 -; AVX512-NEXT: vfmsub213pd %zmm1, %zmm2, %zmm0 +; AVX512-NEXT: vfmsub132pd (%rdi), %zmm1, %zmm0 ; AVX512-NEXT: retq %x = load <8 x double>, <8 x double>* %a0 %y = fmul <8 x double> %x, %a1 |
