diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll')
-rw-r--r-- | llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll | 671 |
1 files changed, 335 insertions, 336 deletions
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll index d253dd1..f3ec442 100644 --- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll +++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll @@ -5100,7 +5100,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm0 = ymm6[0,1,2,3,5,4,6,7,8,9,10,11,13,12,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm3[2,2,2,2,6,6,6,6] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6,7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13,14,15] +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm11 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6,7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13,14,15] ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm0 = ymm6[0,1,2,3,7,6,6,7,8,9,10,11,15,14,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm3[3,3,3,3,7,7,7,7] @@ -5139,8 +5139,8 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm0 ; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm9, %xmm0 ; AVX512F-SLOW-NEXT: vmovdqa 32(%rdx), %xmm15 -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm15[1,1,2,2] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm11[0],xmm0[1],xmm11[2,3],xmm0[4],xmm11[5,6],xmm0[7] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm10 = xmm15[1,1,2,2] +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm10[0],xmm0[1],xmm10[2,3],xmm0[4],xmm10[5,6],xmm0[7] ; AVX512F-SLOW-NEXT: vmovdqa64 %ymm0, %ymm26 ; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm15[0],xmm9[0],xmm15[1],xmm9[1],xmm15[2],xmm9[2],xmm15[3],xmm9[3] ; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm9 = xmm15[4],xmm9[4],xmm15[5],xmm9[5],xmm15[6],xmm9[6],xmm15[7],xmm9[7] @@ -5165,13 +5165,13 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vmovdqa 32(%rax), %ymm6 ; AVX512F-SLOW-NEXT: vpermd %zmm6, %zmm3, %zmm3 ; AVX512F-SLOW-NEXT: vmovdqa64 %ymm28, %ymm7 -; AVX512F-SLOW-NEXT: vpshufb %ymm7, %ymm6, %ymm11 +; AVX512F-SLOW-NEXT: vpshufb %ymm7, %ymm6, %ymm10 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm6[0,1,1,3,4,5,5,7] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,2,2,3] ; AVX512F-SLOW-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm6, %ymm6 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm6, %zmm11, %zmm6 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm6, %zmm10, %zmm6 ; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm7 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11] -; AVX512F-SLOW-NEXT: vpshufb %xmm7, %xmm2, %xmm11 +; AVX512F-SLOW-NEXT: vpshufb %xmm7, %xmm2, %xmm10 ; AVX512F-SLOW-NEXT: vpshufb %xmm7, %xmm1, %xmm1 ; AVX512F-SLOW-NEXT: vmovdqa (%rdi), %xmm13 ; AVX512F-SLOW-NEXT: vmovdqa (%rsi), %xmm14 @@ -5195,10 +5195,10 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[2,1,2,3,4,5,6,7] ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,5,4] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,0,1,3] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,0,1,1] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,0,1,1] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm28 = ymm30[2,2,2,3] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm29 = ymm23[0,2,2,3] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,2,2,3] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,2,2,3] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm19[2,1,3,3] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm30 = ymm18[0,2,2,3] ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,2,4,5,6,7] @@ -5217,21 +5217,21 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm12 = ymm12[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6,7,8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14,15] ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm29, %zmm28, %zmm12 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm10, %zmm7 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm10 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm12, %zmm10, %zmm7 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm11, %zmm7 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm11 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535] +; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm12, %zmm11, %zmm7 ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,1,3] ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm30, %zmm0 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm1, %zmm31, %zmm1 -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm10, %zmm1 +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm11, %zmm1 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm0 # 32-byte Folded Reload ; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm15, %zmm9 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm10 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm10, %zmm9 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm11 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535] +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm11, %zmm9 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm11[0,1,2,3],zmm0[4,5,6,7] -; AVX512F-SLOW-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm11 = mem[2,1,3,2] +; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm10[0,1,2,3],zmm0[4,5,6,7] +; AVX512F-SLOW-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm10 = mem[2,1,3,2] ; AVX512F-SLOW-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload ; AVX512F-SLOW-NEXT: # ymm12 = mem[2,2,2,3] ; AVX512F-SLOW-NEXT: vpshufd $254, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload @@ -5258,7 +5258,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload ; AVX512F-SLOW-NEXT: vpternlogq $236, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm16, %zmm0 ; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm9, %zmm0 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm11, %zmm9 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm10, %zmm9 ; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm7, %zmm9 ; AVX512F-SLOW-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm9, %zmm3 ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm15[2,1,3,2] @@ -5269,7 +5269,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm1, %zmm7 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm8, %zmm17, %zmm1 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm13, %zmm18, %zmm8 -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm1, %zmm10, %zmm8 +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm1, %zmm11, %zmm8 ; AVX512F-SLOW-NEXT: vpbroadcastd 36(%rax), %ymm1 ; AVX512F-SLOW-NEXT: vpbroadcastd 40(%rax), %ymm9 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm9, %zmm1, %zmm1 @@ -5281,9 +5281,8 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm28, %zmm9, %zmm9 ; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm8, %zmm9 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm29, %zmm0, %zmm8 -; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload -; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm8 = zmm10[0,1,2,3],zmm8[4,5,6,7] +; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload +; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm29[0,1,2,3] ; AVX512F-SLOW-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm8, %zmm6 ; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm9, %zmm6 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm31, %zmm30, %zmm8 @@ -10685,14 +10684,14 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; ; AVX512F-SLOW-LABEL: store_i16_stride7_vf64: ; AVX512F-SLOW: # %bb.0: -; AVX512F-SLOW-NEXT: subq $2200, %rsp # imm = 0x898 +; AVX512F-SLOW-NEXT: subq $2168, %rsp # imm = 0x878 ; AVX512F-SLOW-NEXT: vmovdqa 96(%rcx), %ymm2 ; AVX512F-SLOW-NEXT: vmovdqa 96(%rdx), %ymm9 ; AVX512F-SLOW-NEXT: vmovdqa 96(%rdi), %ymm7 ; AVX512F-SLOW-NEXT: vmovdqa 96(%rsi), %ymm8 ; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm0 = [128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128] ; AVX512F-SLOW-NEXT: vpshufb %ymm0, %ymm2, %ymm1 -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm2, %ymm19 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm2, %ymm18 ; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm11 = <u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u> ; AVX512F-SLOW-NEXT: vpshufb %ymm11, %ymm9, %ymm2 ; AVX512F-SLOW-NEXT: vporq %ymm1, %ymm2, %ymm16 @@ -10705,16 +10704,16 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vmovdqa 64(%r9), %ymm2 ; AVX512F-SLOW-NEXT: vpshufb %ymm3, %ymm2, %ymm1 ; AVX512F-SLOW-NEXT: vmovdqa %ymm3, %ymm10 -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm2, %ymm20 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm2, %ymm21 ; AVX512F-SLOW-NEXT: vmovdqa 64(%r8), %ymm3 ; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u> ; AVX512F-SLOW-NEXT: vpshufb %ymm15, %ymm3, %ymm2 -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm3, %ymm24 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm3, %ymm20 ; AVX512F-SLOW-NEXT: vpor %ymm1, %ymm2, %ymm1 ; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX512F-SLOW-NEXT: vmovdqa 64(%rcx), %ymm2 ; AVX512F-SLOW-NEXT: vpshufb %ymm0, %ymm2, %ymm1 -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm2, %ymm18 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm2, %ymm23 ; AVX512F-SLOW-NEXT: vmovdqa 64(%rdx), %ymm6 ; AVX512F-SLOW-NEXT: vpshufb %ymm11, %ymm6, %ymm2 ; AVX512F-SLOW-NEXT: vpor %ymm1, %ymm2, %ymm1 @@ -10744,10 +10743,10 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX512F-SLOW-NEXT: vmovdqa (%rsi), %ymm2 ; AVX512F-SLOW-NEXT: vpshufb %ymm12, %ymm2, %ymm1 -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm2, %ymm23 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm2, %ymm28 ; AVX512F-SLOW-NEXT: vmovdqa (%rdi), %ymm10 ; AVX512F-SLOW-NEXT: vpshufb %ymm13, %ymm10, %ymm2 -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm10, %ymm25 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm10, %ymm22 ; AVX512F-SLOW-NEXT: vpor %ymm1, %ymm2, %ymm1 ; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX512F-SLOW-NEXT: vmovdqa 32(%rcx), %ymm2 @@ -10780,7 +10779,7 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vbroadcasti128 {{.*#+}} ymm12 = [22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27,22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27] ; AVX512F-SLOW-NEXT: # ymm12 = mem[0,1,0,1] -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm19, %ymm14 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm18, %ymm14 ; AVX512F-SLOW-NEXT: vpshufb %ymm12, %ymm14, %ymm10 ; AVX512F-SLOW-NEXT: vmovdqa64 %ymm12, %ymm19 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm12 = ymm9[2,2,2,2,6,6,6,6] @@ -10814,54 +10813,53 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm9, %zmm16, %zmm8 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm17, %zmm7 ; AVX512F-SLOW-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm8, %zmm7 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535] -; AVX512F-SLOW-NEXT: vmovdqa 96(%r8), %ymm12 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm26 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535] +; AVX512F-SLOW-NEXT: vmovdqa 96(%r8), %ymm10 +; AVX512F-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm8 = ymm10[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm10[14,15,u,u,u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm10[16,17,u,u,u,u],zero,zero +; AVX512F-SLOW-NEXT: vpternlogq $248, %ymm26, %ymm7, %ymm8 +; AVX512F-SLOW-NEXT: vmovdqa 96(%r9), %ymm12 ; AVX512F-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm8 = ymm12[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm12[14,15,u,u,u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm12[16,17,u,u,u,u],zero,zero -; AVX512F-SLOW-NEXT: vpternlogq $248, %ymm17, %ymm7, %ymm8 -; AVX512F-SLOW-NEXT: vmovdqa 96(%r9), %ymm14 -; AVX512F-SLOW-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpshufb %ymm3, %ymm14, %ymm9 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm10 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogq $248, %ymm10, %ymm8, %ymm9 +; AVX512F-SLOW-NEXT: vpshufb %ymm3, %ymm12, %ymm9 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm14 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535] +; AVX512F-SLOW-NEXT: vpternlogq $248, %ymm14, %ymm8, %ymm9 ; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm7, %ymm7 -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm8 = ymm12[0,0,2,1,4,4,6,5] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm8 = ymm10[0,0,2,1,4,4,6,5] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm8[2,1,3,3] ; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm10 = [65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0] ; AVX512F-SLOW-NEXT: vpternlogq $184, %ymm7, %ymm10, %ymm8 -; AVX512F-SLOW-NEXT: vprold $16, %ymm14, %ymm7 +; AVX512F-SLOW-NEXT: vprold $16, %ymm12, %ymm7 ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm7[2,2,2,2] -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm10 = [65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogq $184, %ymm8, %ymm10, %ymm7 -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm10, %zmm16 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm14 = [65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535] +; AVX512F-SLOW-NEXT: vpternlogq $184, %ymm8, %ymm14, %ymm7 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm0, %zmm7 ; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm7 = zmm9[0,1,2,3],zmm7[4,5,6,7] ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax -; AVX512F-SLOW-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [6,5,0,0,7,6,0,7,6,5,0,0,7,6,0,7] -; AVX512F-SLOW-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3] +; AVX512F-SLOW-NEXT: vbroadcasti64x4 {{.*#+}} zmm18 = [6,5,0,0,7,6,0,7,6,5,0,0,7,6,0,7] +; AVX512F-SLOW-NEXT: # zmm18 = mem[0,1,2,3,0,1,2,3] ; AVX512F-SLOW-NEXT: vmovdqa 96(%rax), %ymm7 -; AVX512F-SLOW-NEXT: vpermd %zmm7, %zmm8, %zmm9 +; AVX512F-SLOW-NEXT: vpermd %zmm7, %zmm18, %zmm9 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm9 = ymm7[0,1,1,3,4,5,5,7] ; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm10 = [12,13,128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128] ; AVX512F-SLOW-NEXT: vpshufb %ymm10, %ymm7, %ymm7 ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,2,2,3] -; AVX512F-SLOW-NEXT: vpandnq %ymm9, %ymm17, %ymm9 +; AVX512F-SLOW-NEXT: vpandnq %ymm9, %ymm26, %ymm9 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm9, %zmm7, %zmm7 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm14 = [65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535] +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm16 = [65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535] ; AVX512F-SLOW-NEXT: vpbroadcastd 72(%rax), %ymm7 -; AVX512F-SLOW-NEXT: vpandn %ymm7, %ymm14, %ymm9 +; AVX512F-SLOW-NEXT: vpandnq %ymm7, %ymm16, %ymm9 ; AVX512F-SLOW-NEXT: vmovdqa 64(%rax), %ymm7 ; AVX512F-SLOW-NEXT: vpshufb %ymm10, %ymm7, %ymm12 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm9, %zmm9 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vpbroadcastd 8(%rax), %ymm9 -; AVX512F-SLOW-NEXT: vpandn %ymm9, %ymm14, %ymm9 -; AVX512F-SLOW-NEXT: vmovdqa (%rax), %ymm14 -; AVX512F-SLOW-NEXT: vpshufb %ymm10, %ymm14, %ymm12 -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm14, %ymm22 +; AVX512F-SLOW-NEXT: vpandnq %ymm9, %ymm16, %ymm9 +; AVX512F-SLOW-NEXT: vmovdqa (%rax), %ymm8 +; AVX512F-SLOW-NEXT: vpshufb %ymm10, %ymm8, %ymm12 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm8, %ymm24 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm9, %zmm9 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm9 = ymm2[1,2,2,3,4,5,6,7,9,10,10,11,12,13,14,15] @@ -10878,35 +10876,35 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vpshufb %ymm10, %ymm9, %ymm10 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm12 = ymm9[0,1,1,3,4,5,5,7] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm12 = ymm12[0,2,2,3] -; AVX512F-SLOW-NEXT: vpandnq %ymm12, %ymm17, %ymm12 +; AVX512F-SLOW-NEXT: vpandnq %ymm12, %ymm26, %ymm12 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm10, %zmm10 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm10 = ymm13[0,0,2,1,4,4,6,5] ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm12 = ymm15[1,2,2,3,4,5,6,7,9,10,10,11,12,13,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm12 = ymm12[0,0,0,0,4,4,4,4] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm12[0,1,2],ymm10[3],ymm12[4,5],ymm10[6],ymm12[7,8,9,10],ymm10[11],ymm12[12,13],ymm10[14],ymm12[15] -; AVX512F-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm8 = ymm12[0,1,2],ymm10[3],ymm12[4,5],ymm10[6],ymm12[7,8,9,10],ymm10[11],ymm12[12,13],ymm10[14],ymm12[15] +; AVX512F-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX512F-SLOW-NEXT: vmovdqa64 %ymm19, %ymm13 ; AVX512F-SLOW-NEXT: vpshufb %ymm13, %ymm2, %ymm10 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm12 = ymm1[2,2,2,2,6,6,6,6] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm12[2],ymm10[3,4],ymm12[5],ymm10[6,7,8,9],ymm12[10],ymm10[11,12],ymm12[13],ymm10[14,15] -; AVX512F-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm8 = ymm10[0,1],ymm12[2],ymm10[3,4],ymm12[5],ymm10[6,7,8,9],ymm12[10],ymm10[11,12],ymm12[13],ymm10[14,15] +; AVX512F-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[3,3,3,3,7,7,7,7] ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,7,6,6,7,8,9,10,11,15,14,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6,7,8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14,15] -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm31 +; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm0[0,1,2,3,5,4,6,7,8,9,10,11,13,12,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm11[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15] -; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm31 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm11[3,3,3,3,7,7,7,7] ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,6,7,8,9,10,11,15,14,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8,9,10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15] ; AVX512F-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpermd %zmm9, %zmm8, %zmm0 +; AVX512F-SLOW-NEXT: vpermd %zmm9, %zmm18, %zmm0 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm0 = ymm5[1,2,3,3,4,5,6,7,9,10,11,11,12,13,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,1,4,4,6,5] @@ -10918,31 +10916,31 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm4[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6,7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13,14,15] ; AVX512F-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm18, %ymm2 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm23, %ymm2 ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm0 = ymm2[1,2,2,3,4,5,6,7,9,10,10,11,12,13,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm6[0,1,1,3,4,5,5,7] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7,8,9],ymm0[10],ymm1[11,12],ymm0[13],ymm1[14,15] -; AVX512F-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm0, %ymm30 ; AVX512F-SLOW-NEXT: vpshufb %ymm13, %ymm2, %ymm0 -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm18, %ymm3 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm23, %ymm3 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm6[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7,8,9],ymm1[10],ymm0[11,12],ymm1[13],ymm0[14,15] ; AVX512F-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm20, %ymm8 -; AVX512F-SLOW-NEXT: vprold $16, %ymm20, %ymm0 -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm24[1,2,2,3,5,6,6,7] +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm21, %ymm8 +; AVX512F-SLOW-NEXT: vprold $16, %ymm21, %ymm0 +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm20[1,2,2,3,5,6,6,7] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7,8,9],ymm0[10],ymm1[11,12],ymm0[13],ymm1[14,15] ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm1 = ymm8[1,2,2,3,4,5,6,7,9,10,10,11,12,13,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm24[0,0,2,1,4,4,6,5] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm20[0,0,2,1,4,4,6,5] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7,8,9,10],ymm2[11],ymm1[12,13],ymm2[14],ymm1[15] ; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm10 = [2,2,3,3,10,9,11,10] ; AVX512F-SLOW-NEXT: vpermt2q %zmm0, %zmm10, %zmm1 ; AVX512F-SLOW-NEXT: vbroadcasti64x4 {{.*#+}} zmm11 = [0,5,4,0,0,6,5,0,0,5,4,0,0,6,5,0] ; AVX512F-SLOW-NEXT: # zmm11 = mem[0,1,2,3,0,1,2,3] ; AVX512F-SLOW-NEXT: vpermd 64(%rax), %zmm11, %zmm0 -; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm1, %zmm17, %zmm0 +; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm1, %zmm26, %zmm0 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm6[3,3,3,3,7,7,7,7] ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm3[0,1,2,3,7,6,6,7,8,9,10,11,15,14,14,15] @@ -10954,61 +10952,61 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5],ymm1[6],ymm0[7,8,9,10],ymm1[11],ymm0[12,13],ymm1[14],ymm0[15] ; AVX512F-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm24[3,3,3,3,7,7,7,7] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm20[3,3,3,3,7,7,7,7] ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm8[0,1,2,3,5,6,7,7,8,9,10,11,13,14,15,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,2,2,3,6,6,6,7] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7,8,9],ymm0[10],ymm1[11,12],ymm0[13],ymm1[14,15] +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7,8,9],ymm0[10],ymm1[11,12],ymm0[13],ymm1[14,15] ; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm12 = [4,5,4,5,4,5,6,7,16,17,16,17,16,17,17,19] ; AVX512F-SLOW-NEXT: vmovdqa 96(%r9), %xmm0 ; AVX512F-SLOW-NEXT: vmovdqa 96(%r8), %xmm1 -; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3] -; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm3[0,1,3,2,4,5,6,7] -; AVX512F-SLOW-NEXT: vpermt2d %zmm4, %zmm12, %zmm2 +; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3] +; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[0,1,3,2,4,5,6,7] +; AVX512F-SLOW-NEXT: vpermt2d %zmm4, %zmm12, %zmm3 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm7[2,3,3,3,6,7,7,7] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,1,3,2] ; AVX512F-SLOW-NEXT: vpbroadcastd 96(%rax), %ymm5 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm4, %zmm4 ; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm5 = [65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm2, %zmm5, %zmm4 +; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm3, %zmm5, %zmm4 ; AVX512F-SLOW-NEXT: vmovdqa64 %zmm5, %zmm15 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill -; AVX512F-SLOW-NEXT: vmovdqa 96(%rsi), %xmm2 +; AVX512F-SLOW-NEXT: vmovdqa 96(%rsi), %xmm3 ; AVX512F-SLOW-NEXT: vmovdqa 96(%rdi), %xmm4 -; AVX512F-SLOW-NEXT: vprold $16, %xmm2, %xmm5 +; AVX512F-SLOW-NEXT: vprold $16, %xmm3, %xmm5 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm4[1,1,2,3] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1],xmm5[2],xmm6[3,4],xmm5[5],xmm6[6,7] ; AVX512F-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3] -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm18 -; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7] -; AVX512F-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill -; AVX512F-SLOW-NEXT: vmovdqa 96(%rcx), %xmm2 +; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3] +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm23 +; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7] +; AVX512F-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; AVX512F-SLOW-NEXT: vmovdqa 96(%rcx), %xmm3 ; AVX512F-SLOW-NEXT: vmovdqa 96(%rdx), %xmm4 ; AVX512F-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm6 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9] -; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm2, %xmm5 +; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm3, %xmm5 ; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm7 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm4[1,1,2,2] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2,3],xmm5[4],xmm6[5,6],xmm5[7] ; AVX512F-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3] +; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3] ; AVX512F-SLOW-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill -; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7] -; AVX512F-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7] +; AVX512F-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] -; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,5,7,6] +; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,6] ; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11] ; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm1, %xmm1 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm28 = <16,18,19,19,19,19,u,u,0,1,0,1,2,3,2,3> -; AVX512F-SLOW-NEXT: vpermt2d %zmm2, %zmm28, %zmm1 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm29 = <16,18,19,19,19,19,u,u,0,1,0,1,2,3,2,3> +; AVX512F-SLOW-NEXT: vpermt2d %zmm2, %zmm29, %zmm1 ; AVX512F-SLOW-NEXT: vpbroadcastd 100(%rax), %ymm2 ; AVX512F-SLOW-NEXT: vpbroadcastd 104(%rax), %ymm3 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 -; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm1, %zmm16, %zmm2 +; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm1, %zmm14, %zmm2 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vmovdqa 64(%rcx), %xmm1 ; AVX512F-SLOW-NEXT: vmovdqa 64(%rdx), %xmm2 ; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7] -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm3, %xmm30 +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm3, %xmm27 ; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3] ; AVX512F-SLOW-NEXT: vpshufb %xmm7, %xmm1, %xmm1 ; AVX512F-SLOW-NEXT: vmovdqa %xmm7, %xmm8 @@ -11023,37 +11021,37 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm5 ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,0,1,1] ; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7] -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm26 +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm25 ; AVX512F-SLOW-NEXT: vprold $16, %xmm4, %xmm4 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,2,3] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm4[2],xmm2[3,4],xmm4[5],xmm2[6,7] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,0,2,1] -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm5, %zmm4 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0] -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm1, %zmm2, %zmm4 -; AVX512F-SLOW-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm5, %zmm2 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0] +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm1, %zmm4, %zmm2 +; AVX512F-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vmovdqa 64(%r9), %xmm1 -; AVX512F-SLOW-NEXT: vmovdqa 64(%r8), %xmm4 -; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7] -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm16 -; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3] -; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,5,7,6] +; AVX512F-SLOW-NEXT: vmovdqa 64(%r8), %xmm2 +; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7] +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm17 +; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3] +; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,5,7,6] ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,3,2,4,5,6,7] ; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm6 = <0,1,0,1,0,1,1,3,16,18,19,19,19,19,u,u> -; AVX512F-SLOW-NEXT: vpermt2d %zmm4, %zmm6, %zmm1 -; AVX512F-SLOW-NEXT: vpbroadcastd 64(%rax), %ymm4 +; AVX512F-SLOW-NEXT: vpermt2d %zmm2, %zmm6, %zmm1 +; AVX512F-SLOW-NEXT: vpbroadcastd 64(%rax), %ymm2 ; AVX512F-SLOW-NEXT: vpbroadcastd 68(%rax), %ymm5 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm4, %zmm4 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm2, %zmm2 ; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm9 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm1, %zmm9, %zmm4 -; AVX512F-SLOW-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill +; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm1, %zmm9, %zmm2 +; AVX512F-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vmovdqa (%rcx), %xmm1 ; AVX512F-SLOW-NEXT: vmovdqa (%rdx), %xmm5 -; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7] -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm4, %xmm19 +; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7] +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm2, %xmm20 ; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3] ; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm1, %xmm1 -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm8, %xmm29 +; AVX512F-SLOW-NEXT: vmovdqa %xmm8, %xmm14 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[1,1,2,2] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm5[0],xmm1[1],xmm5[2,3],xmm1[4],xmm5[5,6],xmm1[7] ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm7[0,1,3,2,4,5,6,7] @@ -11063,150 +11061,153 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3] ; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm7, %xmm7 ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm7[0,0,1,1] -; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm14 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7] +; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7] +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm2, %xmm19 ; AVX512F-SLOW-NEXT: vprold $16, %xmm5, %xmm5 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[1,1,2,3] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm5[2],xmm3[3,4],xmm5[5],xmm3[6,7] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,0,2,1] -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm3, %zmm7, %zmm3 -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm1, %zmm2, %zmm3 -; AVX512F-SLOW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm3, %zmm7, %zmm2 +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm1, %zmm4, %zmm2 +; AVX512F-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vmovdqa (%r9), %xmm1 -; AVX512F-SLOW-NEXT: vmovdqa (%r8), %xmm2 -; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7] -; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3] -; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,5,7,6] +; AVX512F-SLOW-NEXT: vmovdqa (%r8), %xmm3 +; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7] +; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3] +; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,5,7,6] ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,3,2,4,5,6,7] -; AVX512F-SLOW-NEXT: vpermt2d %zmm2, %zmm6, %zmm1 -; AVX512F-SLOW-NEXT: vpbroadcastd (%rax), %ymm2 -; AVX512F-SLOW-NEXT: vpbroadcastd 4(%rax), %ymm3 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-SLOW-NEXT: vpermt2d %zmm3, %zmm6, %zmm1 +; AVX512F-SLOW-NEXT: vpbroadcastd (%rax), %ymm3 +; AVX512F-SLOW-NEXT: vpbroadcastd 4(%rax), %ymm4 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm3, %zmm2 ; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm1, %zmm9, %zmm2 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm23, %ymm7 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm28, %ymm7 ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm1 = ymm7[1,2,3,3,4,5,6,7,9,10,11,11,12,13,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,0,2,1,4,4,6,5] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm25[1,1,1,1,5,5,5,5] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15] -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm20 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm22, %ymm2 +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm22[1,1,1,1,5,5,5,5] +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15] +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm22 ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm7[0,1,2,3,5,4,6,7,8,9,10,11,13,12,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm25[2,2,2,2,6,6,6,6] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15] -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm27 +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm2[2,2,2,2,6,6,6,6] +; AVX512F-SLOW-NEXT: vmovdqa %ymm2, %ymm9 +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0],ymm1[1],ymm4[2,3],ymm1[4],ymm4[5,6,7,8],ymm1[9],ymm4[10,11],ymm1[12],ymm4[13,14,15] +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm21 +; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload +; AVX512F-SLOW-NEXT: vprold $16, %ymm2, %ymm1 ; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload -; AVX512F-SLOW-NEXT: vprold $16, %ymm3, %ymm1 -; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm5[1,2,2,3,5,6,6,7] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7,8,9],ymm1[10],ymm2[11,12],ymm1[13],ymm2[14,15] -; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm2 = ymm3[1,2,2,3,4,5,6,7,9,10,10,11,12,13,14,15] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[0,0,0,0,4,4,4,4] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm5[0,0,2,1,4,4,6,5] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2],ymm6[3],ymm2[4,5],ymm6[6],ymm2[7,8,9,10],ymm6[11],ymm2[12,13],ymm6[14],ymm2[15] -; AVX512F-SLOW-NEXT: vpermt2q %zmm1, %zmm10, %zmm2 +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm3[1,2,2,3,5,6,6,7] +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0,1],ymm1[2],ymm4[3,4],ymm1[5],ymm4[6,7,8,9],ymm1[10],ymm4[11,12],ymm1[13],ymm4[14,15] +; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm4 = ymm2[1,2,2,3,4,5,6,7,9,10,10,11,12,13,14,15] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[0,0,0,0,4,4,4,4] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm3[0,0,2,1,4,4,6,5] +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3],ymm4[4,5],ymm6[6],ymm4[7,8,9,10],ymm6[11],ymm4[12,13],ymm6[14],ymm4[15] +; AVX512F-SLOW-NEXT: vpermt2q %zmm1, %zmm10, %zmm4 ; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm1 = ymm8[1,2,2,3,4,5,6,7,9,10,10,11,12,13,14,15] ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4] ; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm10[0,1,1,3,4,5,5,7] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm6[0,1],ymm1[2],ymm6[3,4],ymm1[5],ymm6[6,7,8,9],ymm1[10],ymm6[11,12],ymm1[13],ymm6[14,15] -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm24 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm16 ; AVX512F-SLOW-NEXT: vpshufb %ymm13, %ymm8, %ymm1 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm10[2,2,2,2,6,6,6,6] ; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm6[2],ymm1[3,4],ymm6[5],ymm1[6,7,8,9],ymm6[10],ymm1[11,12],ymm6[13],ymm1[14,15] -; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm21 +; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm18 ; AVX512F-SLOW-NEXT: vpermd (%rax), %zmm11, %zmm1 -; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm2, %zmm17, %zmm1 +; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm4, %zmm26, %zmm1 ; AVX512F-SLOW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm10[3,3,3,3,7,7,7,7] -; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm8[0,1,2,3,7,6,6,7,8,9,10,11,15,14,14,15] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,2,2,2,6,6,6,6] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6,7,8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14,15] +; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm4 = ymm8[0,1,2,3,7,6,6,7,8,9,10,11,15,14,14,15] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[2,2,2,2,6,6,6,6] +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3],ymm4[4],ymm1[5,6,7,8],ymm4[9],ymm1[10,11],ymm4[12],ymm1[13,14,15] ; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm25[3,3,3,3,7,7,7,7] -; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm7[0,1,2,3,7,6,6,7,8,9,10,11,15,14,14,15] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,2,2,2,6,6,6,6] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7,8,9,10],ymm2[11],ymm1[12,13],ymm2[14],ymm1[15] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm9[3,3,3,3,7,7,7,7] +; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm4 = ymm7[0,1,2,3,7,6,6,7,8,9,10,11,15,14,14,15] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[2,2,2,2,6,6,6,6] +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3],ymm1[4,5],ymm4[6],ymm1[7,8,9,10],ymm4[11],ymm1[12,13],ymm4[14],ymm1[15] ; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm5[3,3,3,3,7,7,7,7] -; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm3[0,1,2,3,5,6,7,7,8,9,10,11,13,14,15,15] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,2,2,3,6,6,6,7] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7,8,9],ymm1[10],ymm2[11,12],ymm1[13],ymm2[14,15] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm3[3,3,3,3,7,7,7,7] +; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm4 = ymm2[0,1,2,3,5,6,7,7,8,9,10,11,13,14,15,15] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[2,2,2,3,6,6,6,7] +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0,1],ymm1[2],ymm4[3,4],ymm1[5],ymm4[6,7,8,9],ymm1[10],ymm4[11,12],ymm1[13],ymm4[14,15] ; AVX512F-SLOW-NEXT: vmovdqa 32(%r9), %xmm7 -; AVX512F-SLOW-NEXT: vmovdqa 32(%r8), %xmm2 -; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm2[0],xmm7[0],xmm2[1],xmm7[1],xmm2[2],xmm7[2],xmm2[3],xmm7[3] +; AVX512F-SLOW-NEXT: vmovdqa 32(%r8), %xmm4 +; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3] ; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm11 = xmm6[0,1,3,2,4,5,6,7] ; AVX512F-SLOW-NEXT: vpermt2d %zmm11, %zmm12, %zmm1 -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm11 = ymm22[2,3,3,3,6,7,7,7] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm11 = ymm24[2,3,3,3,6,7,7,7] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,1,3,2] ; AVX512F-SLOW-NEXT: vpbroadcastd 32(%rax), %ymm12 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm11, %zmm25 -; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm1, %zmm15, %zmm25 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm11, %zmm26 +; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm1, %zmm15, %zmm26 ; AVX512F-SLOW-NEXT: vmovdqa 32(%rdi), %xmm1 ; AVX512F-SLOW-NEXT: vmovdqa 32(%rsi), %xmm12 ; AVX512F-SLOW-NEXT: vprold $16, %xmm12, %xmm15 ; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm13 = xmm1[1,1,2,3] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm13[0,1],xmm15[2],xmm13[3,4],xmm15[5],xmm13[6,7] -; AVX512F-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm13[0,1],xmm15[2],xmm13[3,4],xmm15[5],xmm13[6,7] +; AVX512F-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm13 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3] ; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm12[4],xmm1[4],xmm12[5],xmm1[5],xmm12[6],xmm1[6],xmm12[7],xmm1[7] ; AVX512F-SLOW-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill -; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm7[4],xmm2[5],xmm7[5],xmm2[6],xmm7[6],xmm2[7],xmm7[7] -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm16, %xmm1 +; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm7[4],xmm4[5],xmm7[5],xmm4[6],xmm7[6],xmm4[7],xmm7[7] +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm1 +; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm1, %xmm2 +; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm3 +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm1 ; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm1, %xmm1 -; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm3 -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm4 +; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm13, %xmm1 +; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm4 -; AVX512F-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm13, %xmm4 -; AVX512F-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm2, %xmm2 -; AVX512F-SLOW-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm23 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm23 = mem[2,2,2,3] -; AVX512F-SLOW-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm22 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm22 = mem[0,2,2,3] -; AVX512F-SLOW-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm18 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm18 = mem[0,2,2,3] +; AVX512F-SLOW-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm24 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm24 = mem[2,2,2,3] +; AVX512F-SLOW-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm23 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm23 = mem[0,2,2,3] +; AVX512F-SLOW-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm7 = mem[0,2,2,3] ; AVX512F-SLOW-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm17 # 32-byte Folded Reload ; AVX512F-SLOW-NEXT: # ymm17 = mem[2,1,3,3] ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,4,5,7,6] -; AVX512F-SLOW-NEXT: vpermt2d %zmm0, %zmm28, %zmm2 +; AVX512F-SLOW-NEXT: vpermt2d %zmm0, %zmm29, %zmm4 ; AVX512F-SLOW-NEXT: vpbroadcastd 36(%rax), %ymm0 ; AVX512F-SLOW-NEXT: vpbroadcastd 40(%rax), %ymm6 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm28 -; AVX512F-SLOW-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm28 +; AVX512F-SLOW-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm28 ; AVX512F-SLOW-NEXT: vmovdqa 32(%rcx), %xmm6 -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm0 -; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm2 +; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm6, %xmm4 ; AVX512F-SLOW-NEXT: vmovdqa 32(%rdx), %xmm0 -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm0[1,1,2,2] -; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm7[0],xmm2[1],xmm7[2,3],xmm2[4],xmm7[5,6],xmm2[7] -; AVX512F-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill -; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3] -; AVX512F-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm12 = xmm0[1,1,2,2] +; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm12[0],xmm4[1],xmm12[2,3],xmm4[4],xmm12[5,6],xmm4[7] +; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3] +; AVX512F-SLOW-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX512F-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7] ; AVX512F-SLOW-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill -; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload -; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm0 = ymm4[0,1,2,3,5,5,7,6,8,9,10,11,13,13,15,14] +; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload +; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} ymm0 = ymm1[0,1,2,3,5,5,7,6,8,9,10,11,13,13,15,14] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm29 = ymm0[3,3,3,3] -; AVX512F-SLOW-NEXT: vpshufd $233, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm2 = mem[1,2,2,3,5,6,6,7] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm10 = ymm4[2,1,2,3,6,5,6,7] -; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm10 = ymm10[0,0,3,3,4,5,6,7,8,8,11,11,12,13,14,15] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm10[2,2,2,2] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,1,1] -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm4 -; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm10 = xmm4[0,2,3,3,4,5,6,7] +; AVX512F-SLOW-NEXT: vpshufd $233, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm0 = mem[1,2,2,3,5,6,6,7] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} ymm14 = ymm1[2,1,2,3,6,5,6,7] +; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} ymm14 = ymm14[0,0,3,3,4,5,6,7,8,8,11,11,12,13,14,15] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm14 = ymm14[2,2,2,2] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,0,1,1] +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm27, %xmm1 +; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm10 = xmm1[0,2,3,3,4,5,6,7] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,0,2,1] -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm4 -; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm8 = xmm4[2,1,2,3,4,5,6,7] +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm1 +; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm8 = xmm1[2,1,2,3,4,5,6,7] ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,5,4] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm8[0,0,1,3] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,0,1,1] -; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm4 -; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm15 = xmm4[0,2,3,3,4,5,6,7] +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm1 +; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm15 = xmm1[0,2,3,3,4,5,6,7] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,0,2,1] -; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm12 = xmm14[2,1,2,3,4,5,6,7] +; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm1 +; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm12 = xmm1[2,1,2,3,4,5,6,7] ; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,5,5,4] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm12 = ymm12[0,0,1,3] ; AVX512F-SLOW-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload @@ -11215,176 +11216,174 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve ; AVX512F-SLOW-NEXT: # ymm5 = mem[2,2,2,3] ; AVX512F-SLOW-NEXT: vpermq $250, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload ; AVX512F-SLOW-NEXT: # ymm6 = mem[2,2,3,3] -; AVX512F-SLOW-NEXT: vpermpd $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm7 = mem[2,2,2,3] -; AVX512F-SLOW-NEXT: vmovups %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill +; AVX512F-SLOW-NEXT: vpermpd $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm1 = mem[2,2,2,3] +; AVX512F-SLOW-NEXT: vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill +; AVX512F-SLOW-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm1 = mem[0,2,2,3] +; AVX512F-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm31 = ymm31[0,2,2,3] -; AVX512F-SLOW-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm30 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm30 = mem[0,2,2,3] -; AVX512F-SLOW-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm26 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm26 = mem[2,1,3,3] -; AVX512F-SLOW-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm19 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm19 = mem[2,2,2,3] +; AVX512F-SLOW-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm27 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm27 = mem[2,1,3,3] +; AVX512F-SLOW-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm20 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm20 = mem[2,2,2,3] ; AVX512F-SLOW-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload ; AVX512F-SLOW-NEXT: # ymm11 = mem[0,2,2,3] -; AVX512F-SLOW-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm16 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm16 = mem[2,1,3,2] -; AVX512F-SLOW-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm7 = mem[2,2,2,3] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm20 = ymm20[2,2,2,3] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm27[0,2,2,3] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm14 = ymm24[2,1,3,2] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm21[2,2,2,3] -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm22, %zmm23, %zmm22 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm17, %zmm18, %zmm23 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm27 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm22, %zmm27, %zmm23 -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,1,3,2] -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogq $184, %ymm23, %ymm17, %ymm2 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535] -; AVX512F-SLOW-NEXT: vpternlogq $184, %ymm2, %ymm17, %ymm0 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm1[0,1,2,3],zmm2[4,5,6,7] -; AVX512F-SLOW-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm17, %zmm2 # 64-byte Folded Reload -; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm1[4,5,6,7] -; AVX512F-SLOW-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm17, %zmm3 # 64-byte Folded Reload +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm19 = ymm30[2,1,3,2] +; AVX512F-SLOW-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm1 = mem[2,2,2,3] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm25 = ymm22[2,2,2,3] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm21[0,2,2,3] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm16 = ymm16[2,1,3,2] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm18[2,2,2,3] +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm23, %zmm24, %zmm23 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm17, %zmm7, %zmm24 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm30 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535] +; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm23, %zmm30, %zmm24 +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,1,3,2] +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm7 = [65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535] +; AVX512F-SLOW-NEXT: vpternlogq $184, %ymm24, %ymm7, %ymm0 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm18 = [65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535] +; AVX512F-SLOW-NEXT: vpternlogq $184, %ymm0, %ymm18, %ymm14 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm0[4,5,6,7] +; AVX512F-SLOW-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm18, %zmm2 # 64-byte Folded Reload +; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm0[4,5,6,7] +; AVX512F-SLOW-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm18, %zmm3 # 64-byte Folded Reload ; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm10 # 32-byte Folded Reload ; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm8 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm22 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm10, %zmm22, %zmm8 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm23 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535] +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm10, %zmm23, %zmm8 ; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm17 = [0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535] ; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm2, %zmm17, %zmm8 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm15, %zmm2 # 32-byte Folded Reload ; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm12, %zmm10 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm2, %zmm22, %zmm10 +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm2, %zmm23, %zmm10 ; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm3, %zmm17, %zmm10 -; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm1, %zmm2 -; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm18 -; AVX512F-SLOW-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm18 -; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm23, %ymm2 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogd $226, 124(%r8){1to8}, %ymm1, %ymm2 +; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm2 +; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm0, %zmm21 +; AVX512F-SLOW-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm21 +; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm24, %ymm2 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535] +; AVX512F-SLOW-NEXT: vpternlogd $226, 124(%r8){1to8}, %ymm0, %ymm2 ; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535] ; AVX512F-SLOW-NEXT: vpternlogq $184, %ymm2, %ymm3, %ymm29 -; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload -; AVX512F-SLOW-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm21 # 64-byte Folded Reload -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm2 -; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload -; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm24 = zmm1[0,1,2,3],zmm2[4,5,6,7] +; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload +; AVX512F-SLOW-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm22 # 64-byte Folded Reload +; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm24 = zmm0[0,1,2,3],zmm6[0,1,2,3] ; AVX512F-SLOW-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm24 # 64-byte Folded Reload ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm29, %zmm0, %zmm2 -; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7] -; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm23 # 64-byte Reload -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0] -; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm0, %zmm1, %zmm23 +; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm14[0,1,2,3],zmm2[4,5,6,7] ; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload -; AVX512F-SLOW-NEXT: vpternlogd $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm29 # 64-byte Folded Reload -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535] +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0] +; AVX512F-SLOW-NEXT: vpternlogd $184, %zmm2, %zmm0, %zmm29 +; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload +; AVX512F-SLOW-NEXT: vpternlogd $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm7 # 64-byte Folded Reload +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535] +; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload +; AVX512F-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm18 # 64-byte Folded Reload ; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload -; AVX512F-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm17 # 64-byte Folded Reload -; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload -; AVX512F-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm6 # 64-byte Folded Reload -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm11, %zmm19, %zmm0 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm16, %zmm2 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535] -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm3, %zmm2 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm13, %zmm20, %zmm0 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm9, %zmm14, %zmm4 -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm3, %zmm4 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0] -; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload -; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm2, %zmm0, %zmm19 +; AVX512F-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm17 # 64-byte Folded Reload +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm11, %zmm20, %zmm2 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm1, %zmm19, %zmm3 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535] +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm2, %zmm4, %zmm3 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm13, %zmm25, %zmm2 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm9, %zmm16, %zmm5 +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm2, %zmm4, %zmm5 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0] ; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload -; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm4, %zmm0, %zmm20 +; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm3, %zmm2, %zmm20 +; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload +; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm5, %zmm2, %zmm25 ; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm31, %zmm0, %zmm0 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm26, %zmm30, %zmm2 -; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm0, %zmm27, %zmm2 -; AVX512F-SLOW-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm0 = mem[0,2,2,3] -; AVX512F-SLOW-NEXT: vpshuflw $180, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload -; AVX512F-SLOW-NEXT: # xmm3 = mem[0,1,3,2,4,5,6,7] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,0,1,1] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,1,3] -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0 -; AVX512F-SLOW-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm3 = mem[2,1,3,3] -; AVX512F-SLOW-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm4 = mem[0,0,1,1] -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm3, %zmm3 -; AVX512F-SLOW-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm4 = mem[0,0,2,1] -; AVX512F-SLOW-NEXT: vpshuflw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload -; AVX512F-SLOW-NEXT: # xmm5 = mem[2,1,2,3,4,5,6,7] -; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,5,4] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,0,1,3] -; AVX512F-SLOW-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm26 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm26 = mem[0,0,1,1] -; AVX512F-SLOW-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload -; AVX512F-SLOW-NEXT: # xmm7 = mem[0,2,3,3,4,5,6,7] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm7[0,0,2,1] -; AVX512F-SLOW-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm9 = mem[0,2,2,3] -; AVX512F-SLOW-NEXT: vpshuflw $180, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload -; AVX512F-SLOW-NEXT: # xmm11 = mem[0,1,3,2,4,5,6,7] -; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[0,0,1,1] -; AVX512F-SLOW-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm12 = mem[2,1,3,3] -; AVX512F-SLOW-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm13 = mem[0,0,1,1] -; AVX512F-SLOW-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload -; AVX512F-SLOW-NEXT: # ymm14 = mem[0,0,2,1] -; AVX512F-SLOW-NEXT: vpshuflw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload -; AVX512F-SLOW-NEXT: # xmm15 = mem[2,1,2,3,4,5,6,7] -; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,5,4] -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,0,1,3] +; AVX512F-SLOW-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm27, %zmm31, %zmm3 +; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm2, %zmm30, %zmm3 +; AVX512F-SLOW-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm2 = mem[0,2,2,3] +; AVX512F-SLOW-NEXT: vpshuflw $180, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload +; AVX512F-SLOW-NEXT: # xmm4 = mem[0,1,3,2,4,5,6,7] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,0,1,1] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,1,3] +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm2, %zmm2 +; AVX512F-SLOW-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm4 = mem[2,1,3,3] +; AVX512F-SLOW-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm5 = mem[0,0,1,1] +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm4, %zmm4 +; AVX512F-SLOW-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm5 = mem[0,0,2,1] +; AVX512F-SLOW-NEXT: vpshuflw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload +; AVX512F-SLOW-NEXT: # xmm6 = mem[2,1,2,3,4,5,6,7] +; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,5,4] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,0,1,3] +; AVX512F-SLOW-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm19 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm19 = mem[0,0,1,1] +; AVX512F-SLOW-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload +; AVX512F-SLOW-NEXT: # xmm9 = mem[0,2,3,3,4,5,6,7] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,0,2,1] +; AVX512F-SLOW-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm11 = mem[0,2,2,3] +; AVX512F-SLOW-NEXT: vpshuflw $180, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload +; AVX512F-SLOW-NEXT: # xmm12 = mem[0,1,3,2,4,5,6,7] +; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm12 = xmm12[0,0,1,1] +; AVX512F-SLOW-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm13 = mem[2,1,3,3] +; AVX512F-SLOW-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm14 = mem[0,0,1,1] +; AVX512F-SLOW-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload +; AVX512F-SLOW-NEXT: # ymm15 = mem[0,0,2,1] +; AVX512F-SLOW-NEXT: vpshuflw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; AVX512F-SLOW-NEXT: # xmm0 = mem[2,1,2,3,4,5,6,7] +; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,4] +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,1,3] ; AVX512F-SLOW-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm16 # 32-byte Folded Reload ; AVX512F-SLOW-NEXT: # ymm16 = mem[0,0,1,1] ; AVX512F-SLOW-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; AVX512F-SLOW-NEXT: # xmm1 = mem[0,2,3,3,4,5,6,7] ; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1] -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm27, %zmm3 -; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm11[0,1,1,3] -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm9, %zmm0 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm13, %zmm12, %zmm9 -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm27, %zmm9 -; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535] -; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload -; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm3, %zmm0, %zmm11 -; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm9, %zmm0, %zmm25 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm4, %zmm0 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm26, %zmm3 -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm22, %zmm3 -; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm15, %zmm14, %zmm0 +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm2, %zmm30, %zmm4 +; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm12[0,1,1,3] +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm11, %zmm2 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm14, %zmm13, %zmm11 +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm2, %zmm30, %zmm11 +; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535] +; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload +; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm4, %zmm2, %zmm12 +; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm11, %zmm2, %zmm26 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm6, %zmm5, %zmm2 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm9, %zmm19, %zmm4 +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm2, %zmm23, %zmm4 +; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm15, %zmm0 ; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm1, %zmm16, %zmm1 -; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm22, %zmm1 +; AVX512F-SLOW-NEXT: vpternlogq $226, %zmm0, %zmm23, %zmm1 ; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0] -; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload -; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm3, %zmm0, %zmm4 +; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload +; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm4, %zmm0, %zmm2 ; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm1, %zmm0, %zmm28 -; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm18, %zmm24 -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm29, %zmm0 -; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm0 +; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm21, %zmm24 +; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm3, %zmm7 ; AVX512F-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512F-SLOW-NEXT: vmovdqa64 %zmm24, 320(%rax) ; AVX512F-SLOW-NEXT: vmovdqa64 %zmm28, 256(%rax) -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm25, 192(%rax) -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm20, 128(%rax) +; AVX512F-SLOW-NEXT: vmovdqa64 %zmm26, 192(%rax) +; AVX512F-SLOW-NEXT: vmovdqa64 %zmm25, 128(%rax) ; AVX512F-SLOW-NEXT: vmovdqa64 %zmm10, 64(%rax) -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm6, (%rax) -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm17, 448(%rax) -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm4, 704(%rax) -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm11, 640(%rax) -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm19, 576(%rax) +; AVX512F-SLOW-NEXT: vmovdqa64 %zmm17, (%rax) +; AVX512F-SLOW-NEXT: vmovdqa64 %zmm18, 448(%rax) +; AVX512F-SLOW-NEXT: vmovdqa64 %zmm2, 704(%rax) +; AVX512F-SLOW-NEXT: vmovdqa64 %zmm12, 640(%rax) +; AVX512F-SLOW-NEXT: vmovdqa64 %zmm20, 576(%rax) ; AVX512F-SLOW-NEXT: vmovdqa64 %zmm8, 512(%rax) -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm0, 384(%rax) -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm21, 768(%rax) -; AVX512F-SLOW-NEXT: vmovdqa64 %zmm23, 832(%rax) -; AVX512F-SLOW-NEXT: addq $2200, %rsp # imm = 0x898 +; AVX512F-SLOW-NEXT: vmovdqa64 %zmm7, 384(%rax) +; AVX512F-SLOW-NEXT: vmovdqa64 %zmm22, 768(%rax) +; AVX512F-SLOW-NEXT: vmovdqa64 %zmm29, 832(%rax) +; AVX512F-SLOW-NEXT: addq $2168, %rsp # imm = 0x878 ; AVX512F-SLOW-NEXT: vzeroupper ; AVX512F-SLOW-NEXT: retq ; |