; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mattr=+sse2 | FileCheck %s -check-prefixes=SSE,SSE2 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mattr=+ssse3 | FileCheck %s -check-prefixes=SSE,SSSE3 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mattr=+sse4.2 | FileCheck %s -check-prefixes=SSE,SSE42 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mattr=+xop | FileCheck %s -check-prefixes=XOP ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mattr=+avx | FileCheck %s -check-prefixes=AVX,AVX1 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mattr=+avx2 | FileCheck %s -check-prefixes=AVX,AVX2 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mattr=+avx512f | FileCheck %s --check-prefixes=AVX512 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX512 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mattr=+avx512f,+avx512bw,+avx512vbmi | FileCheck %s --check-prefixes=AVX512 ; ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mcpu=slm | FileCheck %s --check-prefixes=SSE,SSE42 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mcpu=goldmont | FileCheck %s --check-prefixes=SSE,SSE42 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print" 2>&1 -disable-output -cost-kind=all -mcpu=btver2 | FileCheck %s --check-prefixes=AVX,AVX1 ; ; Verify the cost model for select shuffles ; define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double> %src512, <16 x double> %src1024, <2 x double> %src128_1, <4 x double> %src256_1, <8 x double> %src512_1, <16 x double> %src1024_1) { ; SSE-LABEL: 'test_vXf64' ; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> ; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> ; SSE-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> ; SSE-NEXT: Cost Model: Found costs of 0 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> ; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; XOP-LABEL: 'test_vXf64' ; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> ; XOP-NEXT: Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> ; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX-LABEL: 'test_vXf64' ; AVX-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> ; AVX-NEXT: Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> ; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX512-LABEL: 'test_vXf64' ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> ; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> ret void } define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512, <16 x i64> %src1024, <2 x i64> %src128_1, <4 x i64> %src256_1, <8 x i64> %src512_1, <16 x i64> %src1024_1) { ; SSE-LABEL: 'test_vXi64' ; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> ; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> ; SSE-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> ; SSE-NEXT: Cost Model: Found costs of 0 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> ; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; XOP-LABEL: 'test_vXi64' ; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> ; XOP-NEXT: Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> ; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX-LABEL: 'test_vXi64' ; AVX-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> ; AVX-NEXT: Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> ; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX512-LABEL: 'test_vXi64' ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> ; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> ret void } define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %src256, <16 x float> %src512, <32 x float> %src1024, <2 x float> %src64_1, <4 x float> %src128_1, <8 x float> %src256_1, <16 x float> %src512_1, <32 x float> %src1024_1) { ; SSE-LABEL: 'test_vXf32' ; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> ; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> ; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> ; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> ; SSE-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> ; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; XOP-LABEL: 'test_vXf32' ; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> ; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX-LABEL: 'test_vXf32' ; AVX-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> ; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX512-LABEL: 'test_vXf32' ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> ; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> ret void } define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512, <32 x i32> %src1024, <2 x i32> %src64_1, <4 x i32> %src128_1, <8 x i32> %src256_1, <16 x i32> %src512_1, <32 x i32> %src1024_1) { ; SSE2-LABEL: 'test_vXi32' ; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> ; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> ; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> ; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> ; SSE2-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> ; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; SSSE3-LABEL: 'test_vXi32' ; SSSE3-NEXT: Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> ; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; SSE42-LABEL: 'test_vXi32' ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> ; SSE42-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> ; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; XOP-LABEL: 'test_vXi32' ; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> ; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX-LABEL: 'test_vXi32' ; AVX-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> ; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX512-LABEL: 'test_vXi32' ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> ; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> ret void } define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src512, <2 x i16> %src32_1, <4 x i16> %src64_1, <64 x i16> %src1024, <8 x i16> %src128_1, <16 x i16> %src256_1, <32 x i16> %src512_1, <64 x i16> %src1024_1) { ; SSE2-LABEL: 'test_vXi16' ; SSE2-NEXT: Cost Model: Found costs of 3 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> ; SSE2-NEXT: Cost Model: Found costs of 3 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> ; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> ; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> ; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> ; SSE2-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> ; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; SSSE3-LABEL: 'test_vXi16' ; SSSE3-NEXT: Cost Model: Found costs of 3 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 3 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> ; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; SSE42-LABEL: 'test_vXi16' ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> ; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; XOP-LABEL: 'test_vXi16' ; XOP-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> ; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX-LABEL: 'test_vXi16' ; AVX-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> ; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX512-LABEL: 'test_vXi16' ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> ; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> ret void } define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <64 x i8> %src512, <2 x i8> %src16_1, <4 x i8> %src32_1, <8 x i8> %src64_1, <16 x i8> %src128_1, <32 x i8> %src256_1, <64 x i8> %src512_1) { ; SSE2-LABEL: 'test_vXi8' ; SSE2-NEXT: Cost Model: Found costs of 3 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> %src16_1, <2 x i32> ; SSE2-NEXT: Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> ; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> ; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> ; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> ; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> ; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; SSSE3-LABEL: 'test_vXi8' ; SSSE3-NEXT: Cost Model: Found costs of 3 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> %src16_1, <2 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> ; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> ; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; SSE42-LABEL: 'test_vXi8' ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> %src16_1, <2 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> ; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> ; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; XOP-LABEL: 'test_vXi8' ; XOP-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> %src16_1, <2 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> ; XOP-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> ; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX-LABEL: 'test_vXi8' ; AVX-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> %src16_1, <2 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> ; AVX-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> ; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; ; AVX512-LABEL: 'test_vXi8' ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> %src16_1, <2 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> ; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> ; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void ; %V16 = shufflevector <2 x i8> %src16, <2 x i8> %src16_1, <2 x i32> %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> ret void } ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: ; AVX1: {{.*}} ; AVX2: {{.*}}