diff options
| author | Simon Pilgrim <llvm-dev@redking.me.uk> | 2017-05-14 18:52:15 +0000 |
|---|---|---|
| committer | Simon Pilgrim <llvm-dev@redking.me.uk> | 2017-05-14 18:52:15 +0000 |
| commit | d3f0d03cc5f35c5a5816878a8aa70163d6a0d987 (patch) | |
| tree | a9cb4210315605af9ac781f4e76ee79bf08214cd /llvm/test | |
| parent | c1c78617e7949e54472306b125ff7952e0de2e65 (diff) | |
| download | bcm5719-llvm-d3f0d03cc5f35c5a5816878a8aa70163d6a0d987.tar.gz bcm5719-llvm-d3f0d03cc5f35c5a5816878a8aa70163d6a0d987.zip | |
[X86][AVX1] Account for cost of extract/insert of 256-bit SDIV/UDIV by mul sequences
llvm-svn: 303017
Diffstat (limited to 'llvm/test')
| -rw-r--r-- | llvm/test/Analysis/CostModel/X86/div.ll | 32 |
1 files changed, 16 insertions, 16 deletions
diff --git a/llvm/test/Analysis/CostModel/X86/div.ll b/llvm/test/Analysis/CostModel/X86/div.ll index 0ac06ff75eb..dabaaef3596 100644 --- a/llvm/test/Analysis/CostModel/X86/div.ll +++ b/llvm/test/Analysis/CostModel/X86/div.ll @@ -139,14 +139,14 @@ define i32 @sdiv_uniformconst() { ; SSE2: cost of 38 {{.*}} %V8i32 = sdiv ; SSSE3: cost of 38 {{.*}} %V8i32 = sdiv ; SSE42: cost of 30 {{.*}} %V8i32 = sdiv - ; AVX1: cost of 30 {{.*}} %V8i32 = sdiv + ; AVX1: cost of 32 {{.*}} %V8i32 = sdiv ; AVX2: cost of 15 {{.*}} %V8i32 = sdiv ; AVX512: cost of 15 {{.*}} %V8i32 = sdiv %V8i32 = sdiv <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7> ; SSE2: cost of 76 {{.*}} %V16i32 = sdiv ; SSSE3: cost of 76 {{.*}} %V16i32 = sdiv ; SSE42: cost of 60 {{.*}} %V16i32 = sdiv - ; AVX1: cost of 60 {{.*}} %V16i32 = sdiv + ; AVX1: cost of 64 {{.*}} %V16i32 = sdiv ; AVX2: cost of 30 {{.*}} %V16i32 = sdiv ; AVX512: cost of 15 {{.*}} %V16i32 = sdiv %V16i32 = sdiv <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7> @@ -157,12 +157,12 @@ define i32 @sdiv_uniformconst() { ; AVX: cost of 6 {{.*}} %V8i16 = sdiv %V8i16 = sdiv <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7> ; SSE: cost of 12 {{.*}} %V16i16 = sdiv - ; AVX1: cost of 12 {{.*}} %V16i16 = sdiv + ; AVX1: cost of 14 {{.*}} %V16i16 = sdiv ; AVX2: cost of 6 {{.*}} %V16i16 = sdiv ; AVX512: cost of 6 {{.*}} %V16i16 = sdiv %V16i16 = sdiv <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7> ; SSE: cost of 24 {{.*}} %V32i16 = sdiv - ; AVX1: cost of 24 {{.*}} %V32i16 = sdiv + ; AVX1: cost of 28 {{.*}} %V32i16 = sdiv ; AVX2: cost of 12 {{.*}} %V32i16 = sdiv ; AVX512F: cost of 12 {{.*}} %V32i16 = sdiv ; AVX512BW: cost of 6 {{.*}} %V32i16 = sdiv @@ -203,12 +203,12 @@ define i32 @udiv_uniformconst() { ; AVX: cost of 15 {{.*}} %V4i32 = udiv %V4i32 = udiv <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7> ; SSE: cost of 30 {{.*}} %V8i32 = udiv - ; AVX1: cost of 30 {{.*}} %V8i32 = udiv + ; AVX1: cost of 32 {{.*}} %V8i32 = udiv ; AVX2: cost of 15 {{.*}} %V8i32 = udiv ; AVX512: cost of 15 {{.*}} %V8i32 = udiv %V8i32 = udiv <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7> ; SSE: cost of 60 {{.*}} %V16i32 = udiv - ; AVX1: cost of 60 {{.*}} %V16i32 = udiv + ; AVX1: cost of 64 {{.*}} %V16i32 = udiv ; AVX2: cost of 30 {{.*}} %V16i32 = udiv ; AVX512: cost of 15 {{.*}} %V16i32 = udiv %V16i32 = udiv <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7> @@ -219,12 +219,12 @@ define i32 @udiv_uniformconst() { ; AVX: cost of 6 {{.*}} %V8i16 = udiv %V8i16 = udiv <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7> ; SSE: cost of 12 {{.*}} %V16i16 = udiv - ; AVX1: cost of 12 {{.*}} %V16i16 = udiv + ; AVX1: cost of 14 {{.*}} %V16i16 = udiv ; AVX2: cost of 6 {{.*}} %V16i16 = udiv ; AVX512: cost of 6 {{.*}} %V16i16 = udiv %V16i16 = udiv <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7> ; SSE: cost of 24 {{.*}} %V32i16 = udiv - ; AVX1: cost of 24 {{.*}} %V32i16 = udiv + ; AVX1: cost of 28 {{.*}} %V32i16 = udiv ; AVX2: cost of 12 {{.*}} %V32i16 = udiv ; AVX512F: cost of 12 {{.*}} %V32i16 = udiv ; AVX512BW: cost of 6 {{.*}} %V32i16 = udiv @@ -269,14 +269,14 @@ define i32 @sdiv_uniformconstpow2() { ; SSE2: cost of 38 {{.*}} %V8i32 = sdiv ; SSSE3: cost of 38 {{.*}} %V8i32 = sdiv ; SSE42: cost of 30 {{.*}} %V8i32 = sdiv - ; AVX1: cost of 30 {{.*}} %V8i32 = sdiv + ; AVX1: cost of 32 {{.*}} %V8i32 = sdiv ; AVX2: cost of 15 {{.*}} %V8i32 = sdiv ; AVX512: cost of 15 {{.*}} %V8i32 = sdiv %V8i32 = sdiv <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16> ; SSE2: cost of 76 {{.*}} %V16i32 = sdiv ; SSSE3: cost of 76 {{.*}} %V16i32 = sdiv ; SSE42: cost of 60 {{.*}} %V16i32 = sdiv - ; AVX1: cost of 60 {{.*}} %V16i32 = sdiv + ; AVX1: cost of 64 {{.*}} %V16i32 = sdiv ; AVX2: cost of 30 {{.*}} %V16i32 = sdiv ; AVX512: cost of 15 {{.*}} %V16i32 = sdiv %V16i32 = sdiv <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16> @@ -287,12 +287,12 @@ define i32 @sdiv_uniformconstpow2() { ; AVX: cost of 6 {{.*}} %V8i16 = sdiv %V8i16 = sdiv <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16> ; SSE: cost of 12 {{.*}} %V16i16 = sdiv - ; AVX1: cost of 12 {{.*}} %V16i16 = sdiv + ; AVX1: cost of 14 {{.*}} %V16i16 = sdiv ; AVX2: cost of 6 {{.*}} %V16i16 = sdiv ; AVX512: cost of 6 {{.*}} %V16i16 = sdiv %V16i16 = sdiv <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16> ; SSE: cost of 24 {{.*}} %V32i16 = sdiv - ; AVX1: cost of 24 {{.*}} %V32i16 = sdiv + ; AVX1: cost of 28 {{.*}} %V32i16 = sdiv ; AVX2: cost of 12 {{.*}} %V32i16 = sdiv ; AVX512F: cost of 12 {{.*}} %V32i16 = sdiv ; AVX512BW: cost of 6 {{.*}} %V32i16 = sdiv @@ -333,12 +333,12 @@ define i32 @udiv_uniformconstpow2() { ; AVX: cost of 15 {{.*}} %V4i32 = udiv %V4i32 = udiv <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16> ; SSE: cost of 30 {{.*}} %V8i32 = udiv - ; AVX1: cost of 30 {{.*}} %V8i32 = udiv + ; AVX1: cost of 32 {{.*}} %V8i32 = udiv ; AVX2: cost of 15 {{.*}} %V8i32 = udiv ; AVX512: cost of 15 {{.*}} %V8i32 = udiv %V8i32 = udiv <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16> ; SSE: cost of 60 {{.*}} %V16i32 = udiv - ; AVX1: cost of 60 {{.*}} %V16i32 = udiv + ; AVX1: cost of 64 {{.*}} %V16i32 = udiv ; AVX2: cost of 30 {{.*}} %V16i32 = udiv ; AVX512: cost of 15 {{.*}} %V16i32 = udiv %V16i32 = udiv <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16> @@ -349,12 +349,12 @@ define i32 @udiv_uniformconstpow2() { ; AVX: cost of 6 {{.*}} %V8i16 = udiv %V8i16 = udiv <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16> ; SSE: cost of 12 {{.*}} %V16i16 = udiv - ; AVX1: cost of 12 {{.*}} %V16i16 = udiv + ; AVX1: cost of 14 {{.*}} %V16i16 = udiv ; AVX2: cost of 6 {{.*}} %V16i16 = udiv ; AVX512: cost of 6 {{.*}} %V16i16 = udiv %V16i16 = udiv <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16> ; SSE: cost of 24 {{.*}} %V32i16 = udiv - ; AVX1: cost of 24 {{.*}} %V32i16 = udiv + ; AVX1: cost of 28 {{.*}} %V32i16 = udiv ; AVX2: cost of 12 {{.*}} %V32i16 = udiv ; AVX512F: cost of 12 {{.*}} %V32i16 = udiv ; AVX512BW: cost of 6 {{.*}} %V32i16 = udiv |
