diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/shuffle-vs-trunc-256.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/shuffle-vs-trunc-256.ll | 272 |
1 files changed, 153 insertions, 119 deletions
diff --git a/llvm/test/CodeGen/X86/shuffle-vs-trunc-256.ll b/llvm/test/CodeGen/X86/shuffle-vs-trunc-256.ll index 0bca230f14e..c8c2abc570c 100644 --- a/llvm/test/CodeGen/X86/shuffle-vs-trunc-256.ll +++ b/llvm/test/CodeGen/X86/shuffle-vs-trunc-256.ll @@ -35,36 +35,17 @@ define void @shuffle_v32i8_to_v16i8(<32 x i8>* %L, <16 x i8>* %S) nounwind { ; AVX2-NEXT: vzeroupper ; AVX2-NEXT: retq ; -; AVX512F-LABEL: shuffle_v32i8_to_v16i8: -; AVX512F: # BB#0: -; AVX512F-NEXT: vpmovsxwd (%rdi), %zmm0 -; AVX512F-NEXT: vpmovdb %zmm0, %xmm0 -; AVX512F-NEXT: vmovdqa %xmm0, (%rsi) -; AVX512F-NEXT: vzeroupper -; AVX512F-NEXT: retq -; -; AVX512VL-LABEL: shuffle_v32i8_to_v16i8: -; AVX512VL: # BB#0: -; AVX512VL-NEXT: vpmovsxwd (%rdi), %zmm0 -; AVX512VL-NEXT: vpmovdb %zmm0, %xmm0 -; AVX512VL-NEXT: vmovdqa %xmm0, (%rsi) -; AVX512VL-NEXT: vzeroupper -; AVX512VL-NEXT: retq -; -; AVX512BW-LABEL: shuffle_v32i8_to_v16i8: -; AVX512BW: # BB#0: -; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BW-NEXT: vpmovwb %zmm0, %ymm0 -; AVX512BW-NEXT: vmovdqa %xmm0, (%rsi) -; AVX512BW-NEXT: vzeroupper -; AVX512BW-NEXT: retq -; -; AVX512BWVL-LABEL: shuffle_v32i8_to_v16i8: -; AVX512BWVL: # BB#0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BWVL-NEXT: vpmovwb %ymm0, (%rsi) -; AVX512BWVL-NEXT: vzeroupper -; AVX512BWVL-NEXT: retq +; AVX512-LABEL: shuffle_v32i8_to_v16i8: +; AVX512: # BB#0: +; AVX512-NEXT: vmovdqa (%rdi), %ymm0 +; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512-NEXT: vmovdqa {{.*#+}} xmm2 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u> +; AVX512-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX512-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX512-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] +; AVX512-NEXT: vmovdqa %xmm0, (%rsi) +; AVX512-NEXT: vzeroupper +; AVX512-NEXT: retq %vec = load <32 x i8>, <32 x i8>* %L %strided.vec = shufflevector <32 x i8> %vec, <32 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30> store <16 x i8> %strided.vec, <16 x i8>* %S @@ -149,8 +130,11 @@ define void @shuffle_v16i16_to_v8i16(<16 x i16>* %L, <8 x i16>* %S) nounwind { ; AVX2-LABEL: shuffle_v16i16_to_v8i16: ; AVX2: # BB#0: ; AVX2-NEXT: vmovdqa (%rdi), %ymm0 -; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31] -; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3] +; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] +; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] ; AVX2-NEXT: vmovdqa %xmm0, (%rsi) ; AVX2-NEXT: vzeroupper ; AVX2-NEXT: retq @@ -158,7 +142,11 @@ define void @shuffle_v16i16_to_v8i16(<16 x i16>* %L, <8 x i16>* %S) nounwind { ; AVX512F-LABEL: shuffle_v16i16_to_v8i16: ; AVX512F: # BB#0: ; AVX512F-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512F-NEXT: vpmovdw %zmm0, %ymm0 +; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] +; AVX512F-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX512F-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX512F-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] ; AVX512F-NEXT: vmovdqa %xmm0, (%rsi) ; AVX512F-NEXT: vzeroupper ; AVX512F-NEXT: retq @@ -166,14 +154,26 @@ define void @shuffle_v16i16_to_v8i16(<16 x i16>* %L, <8 x i16>* %S) nounwind { ; AVX512VL-LABEL: shuffle_v16i16_to_v8i16: ; AVX512VL: # BB#0: ; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512VL-NEXT: vpmovdw %ymm0, (%rsi) +; AVX512VL-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512VL-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7] +; AVX512VL-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7] +; AVX512VL-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3] +; AVX512VL-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7] +; AVX512VL-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7] +; AVX512VL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3] +; AVX512VL-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] +; AVX512VL-NEXT: vmovdqa %xmm0, (%rsi) ; AVX512VL-NEXT: vzeroupper ; AVX512VL-NEXT: retq ; ; AVX512BW-LABEL: shuffle_v16i16_to_v8i16: ; AVX512BW: # BB#0: ; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BW-NEXT: vpmovdw %zmm0, %ymm0 +; AVX512BW-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] +; AVX512BW-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX512BW-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX512BW-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] ; AVX512BW-NEXT: vmovdqa %xmm0, (%rsi) ; AVX512BW-NEXT: vzeroupper ; AVX512BW-NEXT: retq @@ -181,7 +181,15 @@ define void @shuffle_v16i16_to_v8i16(<16 x i16>* %L, <8 x i16>* %S) nounwind { ; AVX512BWVL-LABEL: shuffle_v16i16_to_v8i16: ; AVX512BWVL: # BB#0: ; AVX512BWVL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BWVL-NEXT: vpmovdw %ymm0, (%rsi) +; AVX512BWVL-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512BWVL-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7] +; AVX512BWVL-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7] +; AVX512BWVL-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3] +; AVX512BWVL-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7] +; AVX512BWVL-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7] +; AVX512BWVL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3] +; AVX512BWVL-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] +; AVX512BWVL-NEXT: vmovdqa %xmm0, (%rsi) ; AVX512BWVL-NEXT: vzeroupper ; AVX512BWVL-NEXT: retq %vec = load <16 x i16>, <16 x i16>* %L @@ -249,52 +257,23 @@ define void @trunc_v8i32_to_v8i16(<16 x i16>* %L, <8 x i16>* %S) nounwind { } define void @shuffle_v8i32_to_v4i32(<8 x i32>* %L, <4 x i32>* %S) nounwind { -; AVX1-LABEL: shuffle_v8i32_to_v4i32: -; AVX1: # BB#0: -; AVX1-NEXT: vmovaps (%rdi), %ymm0 -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 -; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2] -; AVX1-NEXT: vmovaps %xmm0, (%rsi) -; AVX1-NEXT: vzeroupper -; AVX1-NEXT: retq -; -; AVX2-LABEL: shuffle_v8i32_to_v4i32: -; AVX2: # BB#0: -; AVX2-NEXT: vpermilps {{.*#+}} ymm0 = mem[0,2,2,3,4,6,6,7] -; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3] -; AVX2-NEXT: vmovaps %xmm0, (%rsi) -; AVX2-NEXT: vzeroupper -; AVX2-NEXT: retq -; -; AVX512F-LABEL: shuffle_v8i32_to_v4i32: -; AVX512F: # BB#0: -; AVX512F-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512F-NEXT: vpmovqd %zmm0, %ymm0 -; AVX512F-NEXT: vmovdqa %xmm0, (%rsi) -; AVX512F-NEXT: vzeroupper -; AVX512F-NEXT: retq -; -; AVX512VL-LABEL: shuffle_v8i32_to_v4i32: -; AVX512VL: # BB#0: -; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512VL-NEXT: vpmovqd %ymm0, (%rsi) -; AVX512VL-NEXT: vzeroupper -; AVX512VL-NEXT: retq -; -; AVX512BW-LABEL: shuffle_v8i32_to_v4i32: -; AVX512BW: # BB#0: -; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BW-NEXT: vpmovqd %zmm0, %ymm0 -; AVX512BW-NEXT: vmovdqa %xmm0, (%rsi) -; AVX512BW-NEXT: vzeroupper -; AVX512BW-NEXT: retq -; -; AVX512BWVL-LABEL: shuffle_v8i32_to_v4i32: -; AVX512BWVL: # BB#0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BWVL-NEXT: vpmovqd %ymm0, (%rsi) -; AVX512BWVL-NEXT: vzeroupper -; AVX512BWVL-NEXT: retq +; AVX-LABEL: shuffle_v8i32_to_v4i32: +; AVX: # BB#0: +; AVX-NEXT: vmovaps (%rdi), %ymm0 +; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1 +; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2] +; AVX-NEXT: vmovaps %xmm0, (%rsi) +; AVX-NEXT: vzeroupper +; AVX-NEXT: retq +; +; AVX512-LABEL: shuffle_v8i32_to_v4i32: +; AVX512: # BB#0: +; AVX512-NEXT: vmovaps (%rdi), %ymm0 +; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1 +; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2] +; AVX512-NEXT: vmovaps %xmm0, (%rsi) +; AVX512-NEXT: vzeroupper +; AVX512-NEXT: retq %vec = load <8 x i32>, <8 x i32>* %L %strided.vec = shufflevector <8 x i32> %vec, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6> store <4 x i32> %strided.vec, <4 x i32>* %S @@ -371,9 +350,11 @@ define void @shuffle_v32i8_to_v8i8(<32 x i8>* %L, <8 x i8>* %S) nounwind { ; AVX2-LABEL: shuffle_v32i8_to_v8i8: ; AVX2: # BB#0: ; AVX2-NEXT: vmovdqa (%rdi), %ymm0 -; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31] -; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3] -; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] +; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; AVX2-NEXT: vmovq %xmm0, (%rsi) ; AVX2-NEXT: vzeroupper ; AVX2-NEXT: retq @@ -381,8 +362,11 @@ define void @shuffle_v32i8_to_v8i8(<32 x i8>* %L, <8 x i8>* %S) nounwind { ; AVX512F-LABEL: shuffle_v32i8_to_v8i8: ; AVX512F: # BB#0: ; AVX512F-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512F-NEXT: vpmovdw %zmm0, %ymm0 -; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm2 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX512F-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX512F-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; AVX512F-NEXT: vmovq %xmm0, (%rsi) ; AVX512F-NEXT: vzeroupper ; AVX512F-NEXT: retq @@ -390,15 +374,23 @@ define void @shuffle_v32i8_to_v8i8(<32 x i8>* %L, <8 x i8>* %S) nounwind { ; AVX512VL-LABEL: shuffle_v32i8_to_v8i8: ; AVX512VL: # BB#0: ; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512VL-NEXT: vpmovdb %ymm0, (%rsi) +; AVX512VL-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm2 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512VL-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX512VL-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX512VL-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] +; AVX512VL-NEXT: vmovq %xmm0, (%rsi) ; AVX512VL-NEXT: vzeroupper ; AVX512VL-NEXT: retq ; ; AVX512BW-LABEL: shuffle_v32i8_to_v8i8: ; AVX512BW: # BB#0: ; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BW-NEXT: vpmovdw %zmm0, %ymm0 -; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm2 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX512BW-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX512BW-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; AVX512BW-NEXT: vmovq %xmm0, (%rsi) ; AVX512BW-NEXT: vzeroupper ; AVX512BW-NEXT: retq @@ -406,7 +398,15 @@ define void @shuffle_v32i8_to_v8i8(<32 x i8>* %L, <8 x i8>* %S) nounwind { ; AVX512BWVL-LABEL: shuffle_v32i8_to_v8i8: ; AVX512BWVL: # BB#0: ; AVX512BWVL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BWVL-NEXT: vpmovdb %ymm0, (%rsi) +; AVX512BWVL-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512BWVL-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7] +; AVX512BWVL-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7] +; AVX512BWVL-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3] +; AVX512BWVL-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7] +; AVX512BWVL-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7] +; AVX512BWVL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3] +; AVX512BWVL-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] +; AVX512BWVL-NEXT: vpmovwb %xmm0, (%rsi) ; AVX512BWVL-NEXT: vzeroupper ; AVX512BWVL-NEXT: retq %vec = load <32 x i8>, <32 x i8>* %L @@ -479,19 +479,26 @@ define void @trunc_v8i32_to_v8i8(<32 x i8>* %L, <8 x i8>* %S) nounwind { define void @shuffle_v16i16_to_v4i16(<16 x i16>* %L, <4 x i16>* %S) nounwind { ; AVX1-LABEL: shuffle_v16i16_to_v4i16: ; AVX1: # BB#0: -; AVX1-NEXT: vmovaps (%rdi), %ymm0 +; AVX1-NEXT: vmovdqa (%rdi), %ymm0 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 -; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2] -; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] +; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3] +; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7] +; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3] +; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7] +; AVX1-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; AVX1-NEXT: vmovq %xmm0, (%rsi) ; AVX1-NEXT: vzeroupper ; AVX1-NEXT: retq ; ; AVX2-LABEL: shuffle_v16i16_to_v4i16: ; AVX2: # BB#0: -; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = mem[0,2,2,3,4,6,6,7] -; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3] -; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] +; AVX2-NEXT: vmovdqa (%rdi), %ymm0 +; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3] +; AVX2-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7] +; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3] +; AVX2-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7] +; AVX2-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; AVX2-NEXT: vmovq %xmm0, (%rsi) ; AVX2-NEXT: vzeroupper ; AVX2-NEXT: retq @@ -499,32 +506,44 @@ define void @shuffle_v16i16_to_v4i16(<16 x i16>* %L, <4 x i16>* %S) nounwind { ; AVX512F-LABEL: shuffle_v16i16_to_v4i16: ; AVX512F: # BB#0: ; AVX512F-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512F-NEXT: vpmovqd %zmm0, %ymm0 -; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] +; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512F-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3] +; AVX512F-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7] +; AVX512F-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3] +; AVX512F-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7] +; AVX512F-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; AVX512F-NEXT: vmovq %xmm0, (%rsi) ; AVX512F-NEXT: vzeroupper ; AVX512F-NEXT: retq ; ; AVX512VL-LABEL: shuffle_v16i16_to_v4i16: ; AVX512VL: # BB#0: -; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512VL-NEXT: vpmovqw %ymm0, (%rsi) +; AVX512VL-NEXT: vmovaps (%rdi), %ymm0 +; AVX512VL-NEXT: vextractf128 $1, %ymm0, %xmm1 +; AVX512VL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2] +; AVX512VL-NEXT: vpmovdw %xmm0, (%rsi) ; AVX512VL-NEXT: vzeroupper ; AVX512VL-NEXT: retq ; ; AVX512BW-LABEL: shuffle_v16i16_to_v4i16: ; AVX512BW: # BB#0: ; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BW-NEXT: vpmovqd %zmm0, %ymm0 -; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] +; AVX512BW-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512BW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3] +; AVX512BW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7] +; AVX512BW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3] +; AVX512BW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7] +; AVX512BW-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; AVX512BW-NEXT: vmovq %xmm0, (%rsi) ; AVX512BW-NEXT: vzeroupper ; AVX512BW-NEXT: retq ; ; AVX512BWVL-LABEL: shuffle_v16i16_to_v4i16: ; AVX512BWVL: # BB#0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BWVL-NEXT: vpmovqw %ymm0, (%rsi) +; AVX512BWVL-NEXT: vmovaps (%rdi), %ymm0 +; AVX512BWVL-NEXT: vextractf128 $1, %ymm0, %xmm1 +; AVX512BWVL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2] +; AVX512BWVL-NEXT: vpmovdw %xmm0, (%rsi) ; AVX512BWVL-NEXT: vzeroupper ; AVX512BWVL-NEXT: retq %vec = load <16 x i16>, <16 x i16>* %L @@ -594,19 +613,24 @@ define void @trunc_v4i64_to_v4i16(<16 x i16>* %L, <4 x i16>* %S) nounwind { define void @shuffle_v32i8_to_v4i8(<32 x i8>* %L, <4 x i8>* %S) nounwind { ; AVX1-LABEL: shuffle_v32i8_to_v4i8: ; AVX1: # BB#0: -; AVX1-NEXT: vmovaps (%rdi), %ymm0 +; AVX1-NEXT: vmovdqa (%rdi), %ymm0 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 -; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2] -; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX1-NEXT: vmovdqa {{.*#+}} xmm2 = <0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX1-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] ; AVX1-NEXT: vmovd %xmm0, (%rsi) ; AVX1-NEXT: vzeroupper ; AVX1-NEXT: retq ; ; AVX2-LABEL: shuffle_v32i8_to_v4i8: ; AVX2: # BB#0: -; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = mem[0,2,2,3,4,6,6,7] -; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3] -; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX2-NEXT: vmovdqa (%rdi), %ymm0 +; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = <0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] ; AVX2-NEXT: vmovd %xmm0, (%rsi) ; AVX2-NEXT: vzeroupper ; AVX2-NEXT: retq @@ -614,32 +638,42 @@ define void @shuffle_v32i8_to_v4i8(<32 x i8>* %L, <4 x i8>* %S) nounwind { ; AVX512F-LABEL: shuffle_v32i8_to_v4i8: ; AVX512F: # BB#0: ; AVX512F-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512F-NEXT: vpmovqd %zmm0, %ymm0 -; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512F-NEXT: vmovdqa {{.*#+}} xmm2 = <0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512F-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX512F-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] ; AVX512F-NEXT: vmovd %xmm0, (%rsi) ; AVX512F-NEXT: vzeroupper ; AVX512F-NEXT: retq ; ; AVX512VL-LABEL: shuffle_v32i8_to_v4i8: ; AVX512VL: # BB#0: -; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512VL-NEXT: vpmovqb %ymm0, (%rsi) +; AVX512VL-NEXT: vmovaps (%rdi), %ymm0 +; AVX512VL-NEXT: vextractf128 $1, %ymm0, %xmm1 +; AVX512VL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2] +; AVX512VL-NEXT: vpmovdb %xmm0, (%rsi) ; AVX512VL-NEXT: vzeroupper ; AVX512VL-NEXT: retq ; ; AVX512BW-LABEL: shuffle_v32i8_to_v4i8: ; AVX512BW: # BB#0: ; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BW-NEXT: vpmovqd %zmm0, %ymm0 -; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512BW-NEXT: vextracti128 $1, %ymm0, %xmm1 +; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm2 = <0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u> +; AVX512BW-NEXT: vpshufb %xmm2, %xmm1, %xmm1 +; AVX512BW-NEXT: vpshufb %xmm2, %xmm0, %xmm0 +; AVX512BW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] ; AVX512BW-NEXT: vmovd %xmm0, (%rsi) ; AVX512BW-NEXT: vzeroupper ; AVX512BW-NEXT: retq ; ; AVX512BWVL-LABEL: shuffle_v32i8_to_v4i8: ; AVX512BWVL: # BB#0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %ymm0 -; AVX512BWVL-NEXT: vpmovqb %ymm0, (%rsi) +; AVX512BWVL-NEXT: vmovaps (%rdi), %ymm0 +; AVX512BWVL-NEXT: vextractf128 $1, %ymm0, %xmm1 +; AVX512BWVL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2] +; AVX512BWVL-NEXT: vpmovdb %xmm0, (%rsi) ; AVX512BWVL-NEXT: vzeroupper ; AVX512BWVL-NEXT: retq %vec = load <32 x i8>, <32 x i8>* %L |
