diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll | 433 |
1 files changed, 338 insertions, 95 deletions
diff --git a/llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll b/llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll index 710f5bae92b..20e3432f99b 100644 --- a/llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll +++ b/llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll @@ -44,12 +44,32 @@ define void @shuffle_v16i8_to_v8i8_1(<16 x i8>* %L, <8 x i8>* %S) nounwind { ; AVX-NEXT: vmovq %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v8i8_1: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vmovq %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v8i8_1: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vmovq %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v8i8_1: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u] +; AVX512VL-NEXT: vmovq %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v8i8_1: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vmovq %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v8i8_1: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrlw $8, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovwb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> store <8 x i8> %strided.vec, <8 x i8>* %S @@ -80,12 +100,31 @@ define void @shuffle_v8i16_to_v4i16_1(<8 x i16>* %L, <4 x i16>* %S) nounwind { ; AVX-NEXT: vmovq %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v8i16_to_v4i16_1: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15] -; AVX512-NEXT: vmovq %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v8i16_to_v4i16_1: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15] +; AVX512F-NEXT: vmovq %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v8i16_to_v4i16_1: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vpsrld $16, (%rdi), %xmm0 +; AVX512VL-NEXT: vpmovdw %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v8i16_to_v4i16_1: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15] +; AVX512BW-NEXT: vmovq %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v8i16_to_v4i16_1: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrld $16, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovdw %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <8 x i16>, <8 x i16>* %L %strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7> store <4 x i16> %strided.vec, <4 x i16>* %S @@ -105,11 +144,29 @@ define void @shuffle_v4i32_to_v2i32_1(<4 x i32>* %L, <2 x i32>* %S) nounwind { ; AVX-NEXT: vmovlps %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v4i32_to_v2i32_1: -; AVX512: # %bb.0: -; AVX512-NEXT: vpermilps {{.*#+}} xmm0 = mem[1,3,2,3] -; AVX512-NEXT: vmovlps %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v4i32_to_v2i32_1: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vpermilps {{.*#+}} xmm0 = mem[1,3,2,3] +; AVX512F-NEXT: vmovlps %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v4i32_to_v2i32_1: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vpshufd {{.*#+}} xmm0 = mem[1,1,3,3] +; AVX512VL-NEXT: vpmovqd %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v4i32_to_v2i32_1: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vpermilps {{.*#+}} xmm0 = mem[1,3,2,3] +; AVX512BW-NEXT: vmovlps %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v4i32_to_v2i32_1: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpshufd {{.*#+}} xmm0 = mem[1,1,3,3] +; AVX512BWVL-NEXT: vpmovqd %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <4 x i32>, <4 x i32>* %L %strided.vec = shufflevector <4 x i32> %vec, <4 x i32> undef, <2 x i32> <i32 1, i32 3> store <2 x i32> %strided.vec, <2 x i32>* %S @@ -147,12 +204,32 @@ define void @shuffle_v16i8_to_v4i8_1(<16 x i8>* %L, <4 x i8>* %S) nounwind { ; AVX-NEXT: vmovd %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v4i8_1: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vmovd %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v4i8_1: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vmovd %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v4i8_1: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512VL-NEXT: vpsrlw $8, %xmm0, %xmm0 +; AVX512VL-NEXT: vpmovdb %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v4i8_1: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vmovd %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v4i8_1: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrlw $8, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovdb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13> store <4 x i8> %strided.vec, <4 x i8>* %S @@ -186,12 +263,31 @@ define void @shuffle_v16i8_to_v4i8_2(<16 x i8>* %L, <4 x i8>* %S) nounwind { ; AVX-NEXT: vmovd %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v4i8_2: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vmovd %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v4i8_2: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vmovd %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v4i8_2: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vpsrld $16, (%rdi), %xmm0 +; AVX512VL-NEXT: vpmovdb %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v4i8_2: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vmovd %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v4i8_2: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrld $16, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovdb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14> store <4 x i8> %strided.vec, <4 x i8>* %S @@ -229,12 +325,31 @@ define void @shuffle_v16i8_to_v4i8_3(<16 x i8>* %L, <4 x i8>* %S) nounwind { ; AVX-NEXT: vmovd %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v4i8_3: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vmovd %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v4i8_3: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vmovd %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v4i8_3: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vpsrld $24, (%rdi), %xmm0 +; AVX512VL-NEXT: vpmovdb %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v4i8_3: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vmovd %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v4i8_3: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrld $24, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovdb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15> store <4 x i8> %strided.vec, <4 x i8>* %S @@ -279,9 +394,8 @@ define void @shuffle_v8i16_to_v2i16_1(<8 x i16>* %L, <2 x i16>* %S) nounwind { ; ; AVX512VL-LABEL: shuffle_v8i16_to_v2i16_1: ; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15] -; AVX512VL-NEXT: vmovd %xmm0, (%rsi) +; AVX512VL-NEXT: vpsrld $16, (%rdi), %xmm0 +; AVX512VL-NEXT: vpmovqw %xmm0, (%rsi) ; AVX512VL-NEXT: retq ; ; AVX512BW-LABEL: shuffle_v8i16_to_v2i16_1: @@ -293,9 +407,8 @@ define void @shuffle_v8i16_to_v2i16_1(<8 x i16>* %L, <2 x i16>* %S) nounwind { ; ; AVX512BWVL-LABEL: shuffle_v8i16_to_v2i16_1: ; AVX512BWVL: # %bb.0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BWVL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15] -; AVX512BWVL-NEXT: vmovd %xmm0, (%rsi) +; AVX512BWVL-NEXT: vpsrld $16, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovqw %xmm0, (%rsi) ; AVX512BWVL-NEXT: retq %vec = load <8 x i16>, <8 x i16>* %L %strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <2 x i32> <i32 1, i32 5> @@ -341,9 +454,8 @@ define void @shuffle_v8i16_to_v2i16_2(<8 x i16>* %L, <2 x i16>* %S) nounwind { ; ; AVX512VL-LABEL: shuffle_v8i16_to_v2i16_2: ; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,12,13,4,5,6,7,8,9,10,11,12,13,14,15] -; AVX512VL-NEXT: vmovd %xmm0, (%rsi) +; AVX512VL-NEXT: vpshufd {{.*#+}} xmm0 = mem[1,1,3,3] +; AVX512VL-NEXT: vpmovqw %xmm0, (%rsi) ; AVX512VL-NEXT: retq ; ; AVX512BW-LABEL: shuffle_v8i16_to_v2i16_2: @@ -355,9 +467,8 @@ define void @shuffle_v8i16_to_v2i16_2(<8 x i16>* %L, <2 x i16>* %S) nounwind { ; ; AVX512BWVL-LABEL: shuffle_v8i16_to_v2i16_2: ; AVX512BWVL: # %bb.0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BWVL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,12,13,4,5,6,7,8,9,10,11,12,13,14,15] -; AVX512BWVL-NEXT: vmovd %xmm0, (%rsi) +; AVX512BWVL-NEXT: vpshufd {{.*#+}} xmm0 = mem[1,1,3,3] +; AVX512BWVL-NEXT: vpmovqw %xmm0, (%rsi) ; AVX512BWVL-NEXT: retq %vec = load <8 x i16>, <8 x i16>* %L %strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <2 x i32> <i32 2, i32 6> @@ -403,9 +514,8 @@ define void @shuffle_v8i16_to_v2i16_3(<8 x i16>* %L, <2 x i16>* %S) nounwind { ; ; AVX512VL-LABEL: shuffle_v8i16_to_v2i16_3: ; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,7,14,15,4,5,6,7,8,9,10,11,12,13,14,15] -; AVX512VL-NEXT: vmovd %xmm0, (%rsi) +; AVX512VL-NEXT: vpsrlq $48, (%rdi), %xmm0 +; AVX512VL-NEXT: vpmovqw %xmm0, (%rsi) ; AVX512VL-NEXT: retq ; ; AVX512BW-LABEL: shuffle_v8i16_to_v2i16_3: @@ -417,9 +527,8 @@ define void @shuffle_v8i16_to_v2i16_3(<8 x i16>* %L, <2 x i16>* %S) nounwind { ; ; AVX512BWVL-LABEL: shuffle_v8i16_to_v2i16_3: ; AVX512BWVL: # %bb.0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BWVL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,7,14,15,4,5,6,7,8,9,10,11,12,13,14,15] -; AVX512BWVL-NEXT: vmovd %xmm0, (%rsi) +; AVX512BWVL-NEXT: vpsrlq $48, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovqw %xmm0, (%rsi) ; AVX512BWVL-NEXT: retq %vec = load <8 x i16>, <8 x i16>* %L %strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <2 x i32> <i32 3, i32 7> @@ -456,12 +565,32 @@ define void @shuffle_v16i8_to_v2i8_1(<16 x i8>* %L, <2 x i8>* %S) nounwind { ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v2i8_1: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v2i8_1: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v2i8_1: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512VL-NEXT: vpsrlw $8, %xmm0, %xmm0 +; AVX512VL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v2i8_1: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v2i8_1: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrlw $8, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 1, i32 9> store <2 x i8> %strided.vec, <2 x i8>* %S @@ -494,12 +623,31 @@ define void @shuffle_v16i8_to_v2i8_2(<16 x i8>* %L, <2 x i8>* %S) nounwind { ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v2i8_2: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v2i8_2: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v2i8_2: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vpsrld $16, (%rdi), %xmm0 +; AVX512VL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v2i8_2: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v2i8_2: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrld $16, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 2, i32 10> store <2 x i8> %strided.vec, <2 x i8>* %S @@ -535,12 +683,31 @@ define void @shuffle_v16i8_to_v2i8_3(<16 x i8>* %L, <2 x i8>* %S) nounwind { ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v2i8_3: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v2i8_3: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v2i8_3: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vpsrld $24, (%rdi), %xmm0 +; AVX512VL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v2i8_3: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v2i8_3: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrld $24, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 3, i32 11> store <2 x i8> %strided.vec, <2 x i8>* %S @@ -573,12 +740,31 @@ define void @shuffle_v16i8_to_v2i8_4(<16 x i8>* %L, <2 x i8>* %S) nounwind { ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v2i8_4: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v2i8_4: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v2i8_4: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vpshufd {{.*#+}} xmm0 = mem[1,1,3,3] +; AVX512VL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v2i8_4: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v2i8_4: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpshufd {{.*#+}} xmm0 = mem[1,1,3,3] +; AVX512BWVL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 4, i32 12> store <2 x i8> %strided.vec, <2 x i8>* %S @@ -614,12 +800,31 @@ define void @shuffle_v16i8_to_v2i8_5(<16 x i8>* %L, <2 x i8>* %S) nounwind { ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v2i8_5: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v2i8_5: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v2i8_5: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vpsrlq $40, (%rdi), %xmm0 +; AVX512VL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v2i8_5: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v2i8_5: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrlq $40, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 5, i32 13> store <2 x i8> %strided.vec, <2 x i8>* %S @@ -652,12 +857,31 @@ define void @shuffle_v16i8_to_v2i8_6(<16 x i8>* %L, <2 x i8>* %S) nounwind { ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v2i8_6: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v2i8_6: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v2i8_6: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vpsrlq $48, (%rdi), %xmm0 +; AVX512VL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v2i8_6: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v2i8_6: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrlq $48, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 6, i32 14> store <2 x i8> %strided.vec, <2 x i8>* %S @@ -693,12 +917,31 @@ define void @shuffle_v16i8_to_v2i8_7(<16 x i8>* %L, <2 x i8>* %S) nounwind { ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512-LABEL: shuffle_v16i8_to_v2i8_7: -; AVX512: # %bb.0: -; AVX512-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi) -; AVX512-NEXT: retq +; AVX512F-LABEL: shuffle_v16i8_to_v2i8_7: +; AVX512F: # %bb.0: +; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512F-NEXT: retq +; +; AVX512VL-LABEL: shuffle_v16i8_to_v2i8_7: +; AVX512VL: # %bb.0: +; AVX512VL-NEXT: vpsrlq $56, (%rdi), %xmm0 +; AVX512VL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512VL-NEXT: retq +; +; AVX512BW-LABEL: shuffle_v16i8_to_v2i8_7: +; AVX512BW: # %bb.0: +; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512BW-NEXT: retq +; +; AVX512BWVL-LABEL: shuffle_v16i8_to_v2i8_7: +; AVX512BWVL: # %bb.0: +; AVX512BWVL-NEXT: vpsrlq $56, (%rdi), %xmm0 +; AVX512BWVL-NEXT: vpmovqb %xmm0, (%rsi) +; AVX512BWVL-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 7, i32 15> store <2 x i8> %strided.vec, <2 x i8>* %S |
