diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll | 589 |
1 files changed, 470 insertions, 119 deletions
diff --git a/llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll b/llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll index b73e50e091d..2837974eec8 100644 --- a/llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll +++ b/llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll @@ -425,23 +425,77 @@ define void @shuffle_v32i16_to_v8i16_3(<32 x i16>* %L, <8 x i16>* %S) nounwind { } define void @shuffle_v64i8_to_v8i8_1(<64 x i8>* %L, <8 x i8>* %S) nounwind { -; AVX512-LABEL: shuffle_v64i8_to_v8i8_1: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vmovdqa 16(%rdi), %xmm1 -; AVX512-NEXT: vmovdqa 32(%rdi), %xmm2 -; AVX512-NEXT: vmovdqa 48(%rdi), %xmm3 -; AVX512-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,1,9,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm4, %xmm3, %xmm3 -; AVX512-NEXT: vpshufb %xmm4, %xmm2, %xmm2 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] -; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = <1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm3, %xmm1, %xmm1 -; AVX512-NEXT: vpshufb %xmm3, %xmm0, %xmm0 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] -; AVX512-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] -; AVX512-NEXT: vmovq %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v64i8_to_v8i8_1: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512F-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512F-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,1,9,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512F-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = <1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512F-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512F-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512F-NEXT: vmovq %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_1: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512VL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512VL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512VL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,1,9,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512VL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = <1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512VL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512VL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512VL-NEXT: vmovq %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_1: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BW-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BW-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,1,9,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BW-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = <1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BW-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512BW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512BW-NEXT: vmovq %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_1: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BWVL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BWVL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BWVL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm4 = [0,0,1,1,1,1,9,9,8,8,9,9,10,10,11,11] +; AVX512BWVL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BWVL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BWVL-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1] +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm3 = [1,1,9,9,8,8,9,9,8,8,9,9,10,10,11,11] +; AVX512BWVL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BWVL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BWVL-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] +; AVX512BWVL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3] +; AVX512BWVL-NEXT: vpmovwb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <64 x i8>, <64 x i8>* %L %strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57> store <8 x i8> %strided.vec, <8 x i8>* %S @@ -449,23 +503,68 @@ define void @shuffle_v64i8_to_v8i8_1(<64 x i8>* %L, <8 x i8>* %S) nounwind { } define void @shuffle_v64i8_to_v8i8_2(<64 x i8>* %L, <8 x i8>* %S) nounwind { -; AVX512-LABEL: shuffle_v64i8_to_v8i8_2: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vmovdqa 16(%rdi), %xmm1 -; AVX512-NEXT: vmovdqa 32(%rdi), %xmm2 -; AVX512-NEXT: vmovdqa 48(%rdi), %xmm3 -; AVX512-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,2,10,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm4, %xmm3, %xmm3 -; AVX512-NEXT: vpshufb %xmm4, %xmm2, %xmm2 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] -; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = <2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm3, %xmm1, %xmm1 -; AVX512-NEXT: vpshufb %xmm3, %xmm0, %xmm0 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] -; AVX512-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] -; AVX512-NEXT: vmovq %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v64i8_to_v8i8_2: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512F-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512F-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,2,10,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512F-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = <2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512F-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512F-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512F-NEXT: vmovq %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_2: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512VL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512VL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512VL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,2,10,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512VL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = <2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512VL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512VL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512VL-NEXT: vmovq %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_2: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BW-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BW-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,2,10,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BW-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = <2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BW-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512BW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512BW-NEXT: vmovq %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_2: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm0 = [1,5,9,13,17,21,25,29] +; AVX512BWVL-NEXT: vmovdqa (%rdi), %ymm1 +; AVX512BWVL-NEXT: vpermt2w 32(%rdi), %ymm0, %ymm1 +; AVX512BWVL-NEXT: vpmovwb %xmm1, (%rsi) +; AVX512BWVL-NEXT: vzeroupper +; AVX512BWVL-NEXT: retq %vec = load <64 x i8>, <64 x i8>* %L %strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58> store <8 x i8> %strided.vec, <8 x i8>* %S @@ -473,23 +572,77 @@ define void @shuffle_v64i8_to_v8i8_2(<64 x i8>* %L, <8 x i8>* %S) nounwind { } define void @shuffle_v64i8_to_v8i8_3(<64 x i8>* %L, <8 x i8>* %S) nounwind { -; AVX512-LABEL: shuffle_v64i8_to_v8i8_3: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vmovdqa 16(%rdi), %xmm1 -; AVX512-NEXT: vmovdqa 32(%rdi), %xmm2 -; AVX512-NEXT: vmovdqa 48(%rdi), %xmm3 -; AVX512-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,3,11,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm4, %xmm3, %xmm3 -; AVX512-NEXT: vpshufb %xmm4, %xmm2, %xmm2 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] -; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = <3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm3, %xmm1, %xmm1 -; AVX512-NEXT: vpshufb %xmm3, %xmm0, %xmm0 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] -; AVX512-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] -; AVX512-NEXT: vmovq %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v64i8_to_v8i8_3: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512F-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512F-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,3,11,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512F-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = <3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512F-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512F-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512F-NEXT: vmovq %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_3: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512VL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512VL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512VL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,3,11,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512VL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = <3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512VL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512VL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512VL-NEXT: vmovq %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_3: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BW-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BW-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,3,11,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BW-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = <3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BW-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512BW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512BW-NEXT: vmovq %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_3: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BWVL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BWVL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BWVL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm4 = [10,10,11,11,3,3,11,11,8,8,9,9,10,10,11,11] +; AVX512BWVL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BWVL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BWVL-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1] +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm3 = [3,3,11,11,2,2,3,3,8,8,9,9,10,10,11,11] +; AVX512BWVL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BWVL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BWVL-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] +; AVX512BWVL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3] +; AVX512BWVL-NEXT: vpmovwb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <64 x i8>, <64 x i8>* %L %strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59> store <8 x i8> %strided.vec, <8 x i8>* %S @@ -497,23 +650,68 @@ define void @shuffle_v64i8_to_v8i8_3(<64 x i8>* %L, <8 x i8>* %S) nounwind { } define void @shuffle_v64i8_to_v8i8_4(<64 x i8>* %L, <8 x i8>* %S) nounwind { -; AVX512-LABEL: shuffle_v64i8_to_v8i8_4: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vmovdqa 16(%rdi), %xmm1 -; AVX512-NEXT: vmovdqa 32(%rdi), %xmm2 -; AVX512-NEXT: vmovdqa 48(%rdi), %xmm3 -; AVX512-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,4,12,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm4, %xmm3, %xmm3 -; AVX512-NEXT: vpshufb %xmm4, %xmm2, %xmm2 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] -; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = <4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm3, %xmm1, %xmm1 -; AVX512-NEXT: vpshufb %xmm3, %xmm0, %xmm0 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] -; AVX512-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] -; AVX512-NEXT: vmovq %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v64i8_to_v8i8_4: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512F-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512F-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,4,12,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512F-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = <4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512F-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512F-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512F-NEXT: vmovq %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_4: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512VL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512VL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512VL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,4,12,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512VL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = <4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512VL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512VL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512VL-NEXT: vmovq %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_4: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BW-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BW-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,4,12,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BW-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = <4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BW-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512BW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512BW-NEXT: vmovq %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_4: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm0 = [2,6,10,14,18,22,26,30] +; AVX512BWVL-NEXT: vmovdqa (%rdi), %ymm1 +; AVX512BWVL-NEXT: vpermt2w 32(%rdi), %ymm0, %ymm1 +; AVX512BWVL-NEXT: vpmovwb %xmm1, (%rsi) +; AVX512BWVL-NEXT: vzeroupper +; AVX512BWVL-NEXT: retq %vec = load <64 x i8>, <64 x i8>* %L %strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60> store <8 x i8> %strided.vec, <8 x i8>* %S @@ -521,23 +719,77 @@ define void @shuffle_v64i8_to_v8i8_4(<64 x i8>* %L, <8 x i8>* %S) nounwind { } define void @shuffle_v64i8_to_v8i8_5(<64 x i8>* %L, <8 x i8>* %S) nounwind { -; AVX512-LABEL: shuffle_v64i8_to_v8i8_5: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vmovdqa 16(%rdi), %xmm1 -; AVX512-NEXT: vmovdqa 32(%rdi), %xmm2 -; AVX512-NEXT: vmovdqa 48(%rdi), %xmm3 -; AVX512-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,5,13,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm4, %xmm3, %xmm3 -; AVX512-NEXT: vpshufb %xmm4, %xmm2, %xmm2 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] -; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = <5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm3, %xmm1, %xmm1 -; AVX512-NEXT: vpshufb %xmm3, %xmm0, %xmm0 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] -; AVX512-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] -; AVX512-NEXT: vmovq %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v64i8_to_v8i8_5: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512F-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512F-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,5,13,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512F-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = <5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512F-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512F-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512F-NEXT: vmovq %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_5: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512VL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512VL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512VL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,5,13,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512VL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = <5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512VL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512VL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512VL-NEXT: vmovq %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_5: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BW-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BW-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,5,13,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BW-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = <5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BW-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512BW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512BW-NEXT: vmovq %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_5: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BWVL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BWVL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BWVL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm4 = [12,12,13,13,5,5,13,13,4,4,5,5,6,6,7,7] +; AVX512BWVL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BWVL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BWVL-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1] +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm3 = [5,5,13,13,4,4,5,5,4,4,5,5,6,6,7,7] +; AVX512BWVL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BWVL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BWVL-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] +; AVX512BWVL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3] +; AVX512BWVL-NEXT: vpmovwb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <64 x i8>, <64 x i8>* %L %strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61> store <8 x i8> %strided.vec, <8 x i8>* %S @@ -545,23 +797,68 @@ define void @shuffle_v64i8_to_v8i8_5(<64 x i8>* %L, <8 x i8>* %S) nounwind { } define void @shuffle_v64i8_to_v8i8_6(<64 x i8>* %L, <8 x i8>* %S) nounwind { -; AVX512-LABEL: shuffle_v64i8_to_v8i8_6: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vmovdqa 16(%rdi), %xmm1 -; AVX512-NEXT: vmovdqa 32(%rdi), %xmm2 -; AVX512-NEXT: vmovdqa 48(%rdi), %xmm3 -; AVX512-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,6,14,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm4, %xmm3, %xmm3 -; AVX512-NEXT: vpshufb %xmm4, %xmm2, %xmm2 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] -; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = <6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm3, %xmm1, %xmm1 -; AVX512-NEXT: vpshufb %xmm3, %xmm0, %xmm0 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] -; AVX512-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] -; AVX512-NEXT: vmovq %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v64i8_to_v8i8_6: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512F-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512F-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,6,14,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512F-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = <6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512F-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512F-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512F-NEXT: vmovq %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_6: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512VL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512VL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512VL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,6,14,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512VL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = <6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512VL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512VL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512VL-NEXT: vmovq %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_6: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BW-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BW-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,6,14,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BW-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = <6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BW-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512BW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512BW-NEXT: vmovq %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_6: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm0 = [3,7,11,15,19,23,27,31] +; AVX512BWVL-NEXT: vmovdqa (%rdi), %ymm1 +; AVX512BWVL-NEXT: vpermt2w 32(%rdi), %ymm0, %ymm1 +; AVX512BWVL-NEXT: vpmovwb %xmm1, (%rsi) +; AVX512BWVL-NEXT: vzeroupper +; AVX512BWVL-NEXT: retq %vec = load <64 x i8>, <64 x i8>* %L %strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62> store <8 x i8> %strided.vec, <8 x i8>* %S @@ -569,23 +866,77 @@ define void @shuffle_v64i8_to_v8i8_6(<64 x i8>* %L, <8 x i8>* %S) nounwind { } define void @shuffle_v64i8_to_v8i8_7(<64 x i8>* %L, <8 x i8>* %S) nounwind { -; AVX512-LABEL: shuffle_v64i8_to_v8i8_7: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vmovdqa 16(%rdi), %xmm1 -; AVX512-NEXT: vmovdqa 32(%rdi), %xmm2 -; AVX512-NEXT: vmovdqa 48(%rdi), %xmm3 -; AVX512-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,7,15,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm4, %xmm3, %xmm3 -; AVX512-NEXT: vpshufb %xmm4, %xmm2, %xmm2 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] -; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = <7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u> -; AVX512-NEXT: vpshufb %xmm3, %xmm1, %xmm1 -; AVX512-NEXT: vpshufb %xmm3, %xmm0, %xmm0 -; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] -; AVX512-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] -; AVX512-NEXT: vmovq %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v64i8_to_v8i8_7: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512F-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512F-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,7,15,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512F-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = <7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512F-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512F-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512F-NEXT: vmovq %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_7: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512VL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512VL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512VL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,7,15,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512VL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = <7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512VL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512VL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512VL-NEXT: vmovq %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_7: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BW-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BW-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,7,15,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BW-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3] +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = <7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BW-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] +; AVX512BW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] +; AVX512BW-NEXT: vmovq %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_7: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BWVL-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX512BWVL-NEXT: vmovdqa 32(%rdi), %xmm2 +; AVX512BWVL-NEXT: vmovdqa 48(%rdi), %xmm3 +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm4 = [14,14,15,15,7,7,15,15,4,4,5,5,6,6,7,7] +; AVX512BWVL-NEXT: vpshufb %xmm4, %xmm3, %xmm3 +; AVX512BWVL-NEXT: vpshufb %xmm4, %xmm2, %xmm2 +; AVX512BWVL-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1] +; AVX512BWVL-NEXT: vmovdqa {{.*#+}} xmm3 = [7,7,15,15,6,6,7,7,4,4,5,5,6,6,7,7] +; AVX512BWVL-NEXT: vpshufb %xmm3, %xmm1, %xmm1 +; AVX512BWVL-NEXT: vpshufb %xmm3, %xmm0, %xmm0 +; AVX512BWVL-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] +; AVX512BWVL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3] +; AVX512BWVL-NEXT: vpmovwb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <64 x i8>, <64 x i8>* %L %strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63> store <8 x i8> %strided.vec, <8 x i8>* %S |
