diff options
Diffstat (limited to 'test/CodeGen/X86/vector-popcnt-512.ll')
| -rw-r--r-- | test/CodeGen/X86/vector-popcnt-512.ll | 53 |
1 files changed, 53 insertions, 0 deletions
diff --git a/test/CodeGen/X86/vector-popcnt-512.ll b/test/CodeGen/X86/vector-popcnt-512.ll index cf4f21e62b61..aa50206e7a5e 100644 --- a/test/CodeGen/X86/vector-popcnt-512.ll +++ b/test/CodeGen/X86/vector-popcnt-512.ll @@ -1,6 +1,7 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512f | FileCheck %s --check-prefix=ALL --check-prefix=AVX512 --check-prefix=AVX512F ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX512 --check-prefix=AVX512BW +; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx512vpopcntdq | FileCheck %s --check-prefix=ALL --check-prefix=AVX512 --check-prefix=AVX512VPOPCNTDQ define <8 x i64> @testv8i64(<8 x i64> %in) nounwind { ; AVX512F-LABEL: testv8i64: @@ -39,6 +40,11 @@ define <8 x i64> @testv8i64(<8 x i64> %in) nounwind { ; AVX512BW-NEXT: vpxord %zmm1, %zmm1, %zmm1 ; AVX512BW-NEXT: vpsadbw %zmm1, %zmm0, %zmm0 ; AVX512BW-NEXT: retq +; +; AVX512VPOPCNTDQ-LABEL: testv8i64: +; AVX512VPOPCNTDQ: ## BB#0: +; AVX512VPOPCNTDQ-NEXT: vpopcntq %zmm0, %zmm0 +; AVX512VPOPCNTDQ-NEXT: retq %out = call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %in) ret <8 x i64> %out } @@ -92,6 +98,11 @@ define <16 x i32> @testv16i32(<16 x i32> %in) nounwind { ; AVX512BW-NEXT: vpsadbw %zmm1, %zmm0, %zmm0 ; AVX512BW-NEXT: vpackuswb %zmm2, %zmm0, %zmm0 ; AVX512BW-NEXT: retq +; +; AVX512VPOPCNTDQ-LABEL: testv16i32: +; AVX512VPOPCNTDQ: ## BB#0: +; AVX512VPOPCNTDQ-NEXT: vpopcntd %zmm0, %zmm0 +; AVX512VPOPCNTDQ-NEXT: retq %out = call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %in) ret <16 x i32> %out } @@ -135,6 +146,30 @@ define <32 x i16> @testv32i16(<32 x i16> %in) nounwind { ; AVX512BW-NEXT: vpaddb %zmm0, %zmm1, %zmm0 ; AVX512BW-NEXT: vpsrlw $8, %zmm0, %zmm0 ; AVX512BW-NEXT: retq +; +; AVX512VPOPCNTDQ-LABEL: testv32i16: +; AVX512VPOPCNTDQ: ## BB#0: +; AVX512VPOPCNTDQ-NEXT: vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] +; AVX512VPOPCNTDQ-NEXT: vpand %ymm2, %ymm0, %ymm3 +; AVX512VPOPCNTDQ-NEXT: vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4] +; AVX512VPOPCNTDQ-NEXT: vpshufb %ymm3, %ymm4, %ymm3 +; AVX512VPOPCNTDQ-NEXT: vpsrlw $4, %ymm0, %ymm0 +; AVX512VPOPCNTDQ-NEXT: vpand %ymm2, %ymm0, %ymm0 +; AVX512VPOPCNTDQ-NEXT: vpshufb %ymm0, %ymm4, %ymm0 +; AVX512VPOPCNTDQ-NEXT: vpaddb %ymm3, %ymm0, %ymm0 +; AVX512VPOPCNTDQ-NEXT: vpsllw $8, %ymm0, %ymm3 +; AVX512VPOPCNTDQ-NEXT: vpaddb %ymm0, %ymm3, %ymm0 +; AVX512VPOPCNTDQ-NEXT: vpsrlw $8, %ymm0, %ymm0 +; AVX512VPOPCNTDQ-NEXT: vpand %ymm2, %ymm1, %ymm3 +; AVX512VPOPCNTDQ-NEXT: vpshufb %ymm3, %ymm4, %ymm3 +; AVX512VPOPCNTDQ-NEXT: vpsrlw $4, %ymm1, %ymm1 +; AVX512VPOPCNTDQ-NEXT: vpand %ymm2, %ymm1, %ymm1 +; AVX512VPOPCNTDQ-NEXT: vpshufb %ymm1, %ymm4, %ymm1 +; AVX512VPOPCNTDQ-NEXT: vpaddb %ymm3, %ymm1, %ymm1 +; AVX512VPOPCNTDQ-NEXT: vpsllw $8, %ymm1, %ymm2 +; AVX512VPOPCNTDQ-NEXT: vpaddb %ymm1, %ymm2, %ymm1 +; AVX512VPOPCNTDQ-NEXT: vpsrlw $8, %ymm1, %ymm1 +; AVX512VPOPCNTDQ-NEXT: retq %out = call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %in) ret <32 x i16> %out } @@ -169,6 +204,24 @@ define <64 x i8> @testv64i8(<64 x i8> %in) nounwind { ; AVX512BW-NEXT: vpshufb %zmm0, %zmm3, %zmm0 ; AVX512BW-NEXT: vpaddb %zmm2, %zmm0, %zmm0 ; AVX512BW-NEXT: retq +; +; AVX512VPOPCNTDQ-LABEL: testv64i8: +; AVX512VPOPCNTDQ: ## BB#0: +; AVX512VPOPCNTDQ-NEXT: vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] +; AVX512VPOPCNTDQ-NEXT: vpand %ymm2, %ymm0, %ymm3 +; AVX512VPOPCNTDQ-NEXT: vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4] +; AVX512VPOPCNTDQ-NEXT: vpshufb %ymm3, %ymm4, %ymm3 +; AVX512VPOPCNTDQ-NEXT: vpsrlw $4, %ymm0, %ymm0 +; AVX512VPOPCNTDQ-NEXT: vpand %ymm2, %ymm0, %ymm0 +; AVX512VPOPCNTDQ-NEXT: vpshufb %ymm0, %ymm4, %ymm0 +; AVX512VPOPCNTDQ-NEXT: vpaddb %ymm3, %ymm0, %ymm0 +; AVX512VPOPCNTDQ-NEXT: vpand %ymm2, %ymm1, %ymm3 +; AVX512VPOPCNTDQ-NEXT: vpshufb %ymm3, %ymm4, %ymm3 +; AVX512VPOPCNTDQ-NEXT: vpsrlw $4, %ymm1, %ymm1 +; AVX512VPOPCNTDQ-NEXT: vpand %ymm2, %ymm1, %ymm1 +; AVX512VPOPCNTDQ-NEXT: vpshufb %ymm1, %ymm4, %ymm1 +; AVX512VPOPCNTDQ-NEXT: vpaddb %ymm3, %ymm1, %ymm1 +; AVX512VPOPCNTDQ-NEXT: retq %out = call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %in) ret <64 x i8> %out } |