Skip to content

Commit

Permalink
[CostModel][X86][XOP] Improve costs for XOP shuffles
Browse files Browse the repository at this point in the history 
VPPERM/VPERMIL2PD/VPERMIL2PS all provide more effective 2-input shuffles than regular AVX instructions

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@311005 91177308-0d34-0410-b5e6-96231b3b80d8
  • Loading branch information
@RKSimon
RKSimon committed Aug 16, 2017
1 parent 9c77038 commit b148872
Show file tree
Hide file tree
Showing 3 changed files with 68 additions and 0 deletions.
22 changes: 22 additions & 0 deletions lib/Target/X86/X86TargetTransformInfo.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -861,6 +861,28 @@ int X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind, Type *Tp, int Index,
if (const auto *Entry = CostTableLookup(AVX2ShuffleTbl, Kind, LT.second))
return LT.first * Entry->Cost;

static const CostTblEntry XOPShuffleTbl[] = {
{ TTI::SK_PermuteSingleSrc, MVT::v4f64, 2 }, // vperm2f128 + vpermil2pd
{ TTI::SK_PermuteSingleSrc, MVT::v8f32, 2 }, // vperm2f128 + vpermil2ps
{ TTI::SK_PermuteSingleSrc, MVT::v4i64, 2 }, // vperm2f128 + vpermil2pd
{ TTI::SK_PermuteSingleSrc, MVT::v8i32, 2 }, // vperm2f128 + vpermil2ps
{ TTI::SK_PermuteSingleSrc, MVT::v16i16, 4 }, // vextractf128 + 2*vpperm
// + vinsertf128
{ TTI::SK_PermuteSingleSrc, MVT::v32i8, 4 }, // vextractf128 + 2*vpperm
// + vinsertf128

{ TTI::SK_PermuteTwoSrc, MVT::v16i16, 9 }, // 2*vextractf128 + 6*vpperm
// + vinsertf128
{ TTI::SK_PermuteTwoSrc, MVT::v8i16, 1 }, // vpperm
{ TTI::SK_PermuteTwoSrc, MVT::v32i8, 9 }, // 2*vextractf128 + 6*vpperm
// + vinsertf128
{ TTI::SK_PermuteTwoSrc, MVT::v16i8, 1 }, // vpperm
};

if (ST->hasXOP())
if (const auto *Entry = CostTableLookup(XOPShuffleTbl, Kind, LT.second))
return LT.first * Entry->Cost;

static const CostTblEntry AVX1ShuffleTbl[] = {
{ TTI::SK_Broadcast, MVT::v4f64, 2 }, // vperm2f128 + vpermilpd
{ TTI::SK_Broadcast, MVT::v8f32, 2 }, // vperm2f128 + vpermilps
Expand Down
22 changes: 22 additions & 0 deletions test/Analysis/CostModel/X86/shuffle-single-src.ll
View file
Original file line number Diff line number Diff line change
@@ -1,6 +1,7 @@
; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-unknown-linux-gnu -mattr=+sse2 | FileCheck %s -check-prefix=CHECK -check-prefix=SSE -check-prefix=SSE2
; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-unknown-linux-gnu -mattr=+ssse3 | FileCheck %s -check-prefix=CHECK -check-prefix=SSE -check-prefix=SSSE3
; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-unknown-linux-gnu -mattr=+sse4.2 | FileCheck %s -check-prefix=CHECK -check-prefix=SSE -check-prefix=SSE42
; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-unknown-linux-gnu -mattr=+xop | FileCheck %s -check-prefix=CHECK -check-prefix=XOP
; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-unknown-linux-gnu -mattr=+avx | FileCheck %s -check-prefix=CHECK -check-prefix=AVX -check-prefix=AVX1
; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-unknown-linux-gnu -mattr=+avx2 | FileCheck %s -check-prefix=CHECK -check-prefix=AVX -check-prefix=AVX2
; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
Expand All @@ -17,6 +18,7 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
; SSE2: cost of 1 {{.*}} %V128 = shufflevector
; SSSE3: cost of 1 {{.*}} %V128 = shufflevector
; SSE42: cost of 1 {{.*}} %V128 = shufflevector
; XOP: cost of 1 {{.*}} %V128 = shufflevector
; AVX1: cost of 1 {{.*}} %V128 = shufflevector
; AVX2: cost of 1 {{.*}} %V128 = shufflevector
; AVX512: cost of 1 {{.*}} %V128 = shufflevector
Expand All @@ -25,6 +27,7 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
; SSE2: cost of 2 {{.*}} %V256 = shufflevector
; SSSE3: cost of 2 {{.*}} %V256 = shufflevector
; SSE42: cost of 2 {{.*}} %V256 = shufflevector
; XOP: cost of 2 {{.*}} %V256 = shufflevector
; AVX1: cost of 3 {{.*}} %V256 = shufflevector
; AVX2: cost of 1 {{.*}} %V256 = shufflevector
; AVX512: cost of 1 {{.*}} %V256 = shufflevector
Expand All @@ -33,6 +36,7 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
; SSE2: cost of 12 {{.*}} %V512 = shufflevector
; SSSE3: cost of 12 {{.*}} %V512 = shufflevector
; SSE42: cost of 12 {{.*}} %V512 = shufflevector
; XOP: cost of 8 {{.*}} %V512 = shufflevector
; AVX1: cost of 8 {{.*}} %V512 = shufflevector
; AVX2: cost of 6 {{.*}} %V512 = shufflevector
; AVX512: cost of 1 {{.*}} %V512 = shufflevector
Expand All @@ -41,6 +45,7 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
; SSE2: cost of 56 {{.*}} %V1024 = shufflevector
; SSSE3: cost of 56 {{.*}} %V1024 = shufflevector
; SSE42: cost of 56 {{.*}} %V1024 = shufflevector
; XOP: cost of 48 {{.*}} %V1024 = shufflevector
; AVX1: cost of 48 {{.*}} %V1024 = shufflevector
; AVX2: cost of 36 {{.*}} %V1024 = shufflevector
; AVX512: cost of 2 {{.*}} %V1024 = shufflevector
Expand All @@ -55,6 +60,7 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
; SSE2: cost of 1 {{.*}} %V128 = shufflevector
; SSSE3: cost of 1 {{.*}} %V128 = shufflevector
; SSE42: cost of 1 {{.*}} %V128 = shufflevector
; XOP: cost of 1 {{.*}} %V128 = shufflevector
; AVX1: cost of 1 {{.*}} %V128 = shufflevector
; AVX2: cost of 1 {{.*}} %V128 = shufflevector
; AVX512: cost of 1 {{.*}} %V128 = shufflevector
Expand All @@ -63,6 +69,7 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
; SSE2: cost of 2 {{.*}} %V256 = shufflevector
; SSSE3: cost of 2 {{.*}} %V256 = shufflevector
; SSE42: cost of 2 {{.*}} %V256 = shufflevector
; XOP: cost of 2 {{.*}} %V256 = shufflevector
; AVX1: cost of 3 {{.*}} %V256 = shufflevector
; AVX2: cost of 1 {{.*}} %V256 = shufflevector
; AVX512: cost of 1 {{.*}} %V256 = shufflevector
Expand All @@ -71,6 +78,7 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
; SSE2: cost of 12 {{.*}} %V512 = shufflevector
; SSSE3: cost of 12 {{.*}} %V512 = shufflevector
; SSE42: cost of 12 {{.*}} %V512 = shufflevector
; XOP: cost of 8 {{.*}} %V512 = shufflevector
; AVX1: cost of 8 {{.*}} %V512 = shufflevector
; AVX2: cost of 6 {{.*}} %V512 = shufflevector
; AVX512: cost of 1 {{.*}} %V512 = shufflevector
Expand All @@ -85,6 +93,7 @@ define void @test_vXf32(<4 x float> %src128, <8 x float> %src256, <16 x float> %
; SSE2: cost of 1 {{.*}} %V128 = shufflevector
; SSSE3: cost of 1 {{.*}} %V128 = shufflevector
; SSE42: cost of 1 {{.*}} %V128 = shufflevector
; XOP: cost of 1 {{.*}} %V128 = shufflevector
; AVX1: cost of 1 {{.*}} %V128 = shufflevector
; AVX2: cost of 1 {{.*}} %V128 = shufflevector
; AVX512: cost of 1 {{.*}} %V128 = shufflevector
Expand All @@ -93,6 +102,7 @@ define void @test_vXf32(<4 x float> %src128, <8 x float> %src256, <16 x float> %
; SSE2: cost of 4 {{.*}} %V256 = shufflevector
; SSSE3: cost of 4 {{.*}} %V256 = shufflevector
; SSE42: cost of 4 {{.*}} %V256 = shufflevector
; XOP: cost of 2 {{.*}} %V256 = shufflevector
; AVX1: cost of 4 {{.*}} %V256 = shufflevector
; AVX2: cost of 1 {{.*}} %V256 = shufflevector
; AVX512: cost of 1 {{.*}} %V256 = shufflevector
Expand All @@ -101,6 +111,7 @@ define void @test_vXf32(<4 x float> %src128, <8 x float> %src256, <16 x float> %
; SSE2: cost of 24 {{.*}} %V512 = shufflevector
; SSSE3: cost of 24 {{.*}} %V512 = shufflevector
; SSE42: cost of 24 {{.*}} %V512 = shufflevector
; XOP: cost of 8 {{.*}} %V512 = shufflevector
; AVX1: cost of 8 {{.*}} %V512 = shufflevector
; AVX2: cost of 6 {{.*}} %V512 = shufflevector
; AVX512: cost of 1 {{.*}} %V512 = shufflevector
Expand All @@ -115,6 +126,7 @@ define void @test_vXi32(<4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512
; SSE2: cost of 1 {{.*}} %V128 = shufflevector
; SSSE3: cost of 1 {{.*}} %V128 = shufflevector
; SSE42: cost of 1 {{.*}} %V128 = shufflevector
; XOP: cost of 1 {{.*}} %V128 = shufflevector
; AVX1: cost of 1 {{.*}} %V128 = shufflevector
; AVX2: cost of 1 {{.*}} %V128 = shufflevector
; AVX512: cost of 1 {{.*}} %V128 = shufflevector
Expand All @@ -123,6 +135,7 @@ define void @test_vXi32(<4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512
; SSE2: cost of 4 {{.*}} %V256 = shufflevector
; SSSE3: cost of 4 {{.*}} %V256 = shufflevector
; SSE42: cost of 4 {{.*}} %V256 = shufflevector
; XOP: cost of 2 {{.*}} %V256 = shufflevector
; AVX1: cost of 4 {{.*}} %V256 = shufflevector
; AVX2: cost of 1 {{.*}} %V256 = shufflevector
; AVX512: cost of 1 {{.*}} %V256 = shufflevector
Expand All @@ -131,6 +144,7 @@ define void @test_vXi32(<4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512
; SSE2: cost of 24 {{.*}} %V512 = shufflevector
; SSSE3: cost of 24 {{.*}} %V512 = shufflevector
; SSE42: cost of 24 {{.*}} %V512 = shufflevector
; XOP: cost of 8 {{.*}} %V512 = shufflevector
; AVX1: cost of 8 {{.*}} %V512 = shufflevector
; AVX2: cost of 6 {{.*}} %V512 = shufflevector
; AVX512: cost of 1 {{.*}} %V512 = shufflevector
Expand All @@ -139,6 +153,7 @@ define void @test_vXi32(<4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512
; SSE2: cost of 112 {{.*}} %V1024 = shufflevector
; SSSE3: cost of 112 {{.*}} %V1024 = shufflevector
; SSE42: cost of 112 {{.*}} %V1024 = shufflevector
; XOP: cost of 48 {{.*}} %V1024 = shufflevector
; AVX1: cost of 48 {{.*}} %V1024 = shufflevector
; AVX2: cost of 36 {{.*}} %V1024 = shufflevector
; AVX512: cost of 2 {{.*}} %V1024 = shufflevector
Expand All @@ -152,6 +167,7 @@ define void @test_vXi16(<8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src51
; SSE2: cost of 5 {{.*}} %V128 = shufflevector
; SSSE3: cost of 1 {{.*}} %V128 = shufflevector
; SSE42: cost of 1 {{.*}} %V128 = shufflevector
; XOP: cost of 1 {{.*}} %V128 = shufflevector
; AVX1: cost of 1 {{.*}} %V128 = shufflevector
; AVX2: cost of 1 {{.*}} %V128 = shufflevector
; AVX512F: cost of 1 {{.*}} %V128 = shufflevector
Expand All @@ -162,6 +178,7 @@ define void @test_vXi16(<8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src51
; SSE2: cost of 16 {{.*}} %V256 = shufflevector
; SSSE3: cost of 6 {{.*}} %V256 = shufflevector
; SSE42: cost of 6 {{.*}} %V256 = shufflevector
; XOP: cost of 4 {{.*}} %V256 = shufflevector
; AVX1: cost of 8 {{.*}} %V256 = shufflevector
; AVX2: cost of 4 {{.*}} %V256 = shufflevector
; AVX512F: cost of 4 {{.*}} %V256 = shufflevector
Expand All @@ -172,6 +189,7 @@ define void @test_vXi16(<8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src51
; SSE2: cost of 96 {{.*}} %V512 = shufflevector
; SSSE3: cost of 36 {{.*}} %V512 = shufflevector
; SSE42: cost of 36 {{.*}} %V512 = shufflevector
; XOP: cost of 18 {{.*}} %V512 = shufflevector
; AVX1: cost of 30 {{.*}} %V512 = shufflevector
; AVX2: cost of 14 {{.*}} %V512 = shufflevector
; AVX512F: cost of 14 {{.*}} %V512 = shufflevector
Expand All @@ -182,6 +200,7 @@ define void @test_vXi16(<8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src51
; SSE2: cost of 448 {{.*}} %V1024 = shufflevector
; SSSE3: cost of 168 {{.*}} %V1024 = shufflevector
; SSE42: cost of 168 {{.*}} %V1024 = shufflevector
; XOP: cost of 108 {{.*}} %V1024 = shufflevector
; AVX1: cost of 180 {{.*}} %V1024 = shufflevector
; AVX2: cost of 84 {{.*}} %V1024 = shufflevector
; AVX512F: cost of 84 {{.*}} %V1024 = shufflevector
Expand All @@ -196,6 +215,7 @@ define void @test_vXi8(<16 x i8> %src128, <32 x i8> %src256, <64 x i8> %src512)
; SSE2: cost of 10 {{.*}} %V128 = shufflevector
; SSSE3: cost of 1 {{.*}} %V128 = shufflevector
; SSE42: cost of 1 {{.*}} %V128 = shufflevector
; XOP: cost of 1 {{.*}} %V128 = shufflevector
; AVX1: cost of 1 {{.*}} %V128 = shufflevector
; AVX2: cost of 1 {{.*}} %V128 = shufflevector
; AVX512: cost of 1 {{.*}} %V128 = shufflevector
Expand All @@ -204,6 +224,7 @@ define void @test_vXi8(<16 x i8> %src128, <32 x i8> %src256, <64 x i8> %src512)
; SSE2: cost of 26 {{.*}} %V256 = shufflevector
; SSSE3: cost of 6 {{.*}} %V256 = shufflevector
; SSE42: cost of 6 {{.*}} %V256 = shufflevector
; XOP: cost of 4 {{.*}} %V256 = shufflevector
; AVX1: cost of 8 {{.*}} %V256 = shufflevector
; AVX2: cost of 4 {{.*}} %V256 = shufflevector
; AVX512F: cost of 4 {{.*}} %V256 = shufflevector
Expand All @@ -214,6 +235,7 @@ define void @test_vXi8(<16 x i8> %src128, <32 x i8> %src256, <64 x i8> %src512)
; SSE2: cost of 156 {{.*}} %V512 = shufflevector
; SSSE3: cost of 36 {{.*}} %V512 = shufflevector
; SSE42: cost of 36 {{.*}} %V512 = shufflevector
; XOP: cost of 18 {{.*}} %V512 = shufflevector
; AVX1: cost of 30 {{.*}} %V512 = shufflevector
; AVX2: cost of 14 {{.*}} %V512 = shufflevector
; AVX512F: cost of 14 {{.*}} %V512 = shufflevector
Expand Down
Loading

0 comments on commit b148872

Please sign in to comment.