diff options
| author | Nadav Rotem <nrotem@apple.com> | 2012-12-07 19:01:13 +0000 |
|---|---|---|
| committer | Nadav Rotem <nrotem@apple.com> | 2012-12-07 19:01:13 +0000 |
| commit | 481e50efe092e7dc4527a54a802121404bb7a58c (patch) | |
| tree | cbd06aecb25530a6781cb9adbbfb32031e1a5fc7 /llvm/test/CodeGen | |
| parent | 889037d754b29d34501c35b0c127feb53344ed62 (diff) | |
| download | bcm5719-llvm-481e50efe092e7dc4527a54a802121404bb7a58c.tar.gz bcm5719-llvm-481e50efe092e7dc4527a54a802121404bb7a58c.zip | |
X86: Prefer using VPSHUFD over VPERMIL because it has better throughput.
llvm-svn: 169624
Diffstat (limited to 'llvm/test/CodeGen')
| -rw-r--r-- | llvm/test/CodeGen/X86/avx-intrinsics-x86.ll | 2 | ||||
| -rw-r--r-- | llvm/test/CodeGen/X86/avx-shuffle.ll | 4 | ||||
| -rw-r--r-- | llvm/test/CodeGen/X86/avx-splat.ll | 4 |
3 files changed, 5 insertions, 5 deletions
diff --git a/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll b/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll index fa90ae7ef1a..0be83f648d1 100644 --- a/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll +++ b/llvm/test/CodeGen/X86/avx-intrinsics-x86.ll @@ -2348,7 +2348,7 @@ declare <4 x double> @llvm.x86.avx.vpermil.pd.256(<4 x double>, i8) nounwind rea define <4 x float> @test_x86_avx_vpermil_ps(<4 x float> %a0) { - ; CHECK: vpermilps + ; CHECK: vpshufd %res = call <4 x float> @llvm.x86.avx.vpermil.ps(<4 x float> %a0, i8 7) ; <<4 x float>> [#uses=1] ret <4 x float> %res } diff --git a/llvm/test/CodeGen/X86/avx-shuffle.ll b/llvm/test/CodeGen/X86/avx-shuffle.ll index 904f048d1e9..65685a3224f 100644 --- a/llvm/test/CodeGen/X86/avx-shuffle.ll +++ b/llvm/test/CodeGen/X86/avx-shuffle.ll @@ -6,7 +6,7 @@ define <4 x float> @test1(<4 x float> %a) nounwind { ret <4 x float> %b ; CHECK: test1: ; CHECK: vshufps -; CHECK: vpermilps +; CHECK: vpshufd } ; rdar://10538417 @@ -106,7 +106,7 @@ define <4 x float> @test11(<4 x float> %a) nounwind { define <4 x float> @test12(<4 x float>* %a) nounwind { ; CHECK: test12 -; CHECK: vpermilps $27, ( +; CHECK: vpshufd %tmp0 = load <4 x float>* %a %tmp1 = shufflevector <4 x float> %tmp0, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> ret <4 x float> %tmp1 diff --git a/llvm/test/CodeGen/X86/avx-splat.ll b/llvm/test/CodeGen/X86/avx-splat.ll index 5ad75236e11..67e4b40810e 100644 --- a/llvm/test/CodeGen/X86/avx-splat.ll +++ b/llvm/test/CodeGen/X86/avx-splat.ll @@ -84,7 +84,7 @@ define <8 x float> @funcF(i32 %val) nounwind { ret <8 x float> %tmp } -; CHECK: vpermilps $0 +; CHECK: vpshufd $0 ; CHECK-NEXT: vinsertf128 $1 define <8 x float> @funcG(<8 x float> %a) nounwind uwtable readnone ssp { entry: @@ -93,7 +93,7 @@ entry: } ; CHECK: vextractf128 $1 -; CHECK-NEXT: vpermilps $85 +; CHECK-NEXT: vpshufd ; CHECK-NEXT: vinsertf128 $1 define <8 x float> @funcH(<8 x float> %a) nounwind uwtable readnone ssp { entry: |
