From e3bb4bb2d553003df7d069ee2504c7e9403aac57 Mon Sep 17 00:00:00 2001 From: Chandler Carruth Date: Thu, 25 Sep 2014 02:52:12 +0000 Subject: [PATCH] [x86] Implement AVX2 support for v32i8 in the new vector shuffle lowering. This completes the basic AVX2 feature support, but there are still some improvements I'd like to do to really get the last mile of performance here. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218440 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/X86/X86ISelLowering.cpp | 62 ++++- test/CodeGen/X86/vector-shuffle-256-v32.ll | 266 +++++---------------- 2 files changed, 112 insertions(+), 216 deletions(-) diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp index 3afc951c60c..6b92fb96456 100644 --- a/lib/Target/X86/X86ISelLowering.cpp +++ b/lib/Target/X86/X86ISelLowering.cpp @@ -7375,8 +7375,10 @@ static SDValue lowerVectorShuffleAsBlend(SDLoc DL, MVT VT, SDValue V1, return DAG.getNode(X86ISD::BLENDI, DL, MVT::v16i16, V1, V2, DAG.getConstant(BlendMask, MVT::i8)); } - - // Fall back to a fully general variable byte blend. + } + // FALLTHROUGH + case MVT::v32i8: { + assert(Subtarget->hasAVX2() && "256-bit integer blends require AVX2!"); SDValue PBLENDVMask[32]; // Scale the blend by the number of bytes per element. int Scale = VT.getScalarSizeInBits() / 8; @@ -9700,9 +9702,59 @@ static SDValue lowerV32I8VectorShuffle(SDValue Op, SDValue V1, SDValue V2, assert(Mask.size() == 32 && "Unexpected mask size for v32 shuffle!"); assert(Subtarget->hasAVX2() && "We can only lower v32i8 with AVX2!"); - // FIXME: Actually implement this using AVX2!!! - (void)Mask; - return splitAndLower256BitVectorShuffle(Op, V1, V2, Subtarget, DAG); + // There are no generalized cross-lane shuffle operations available on i16 + // element types. + // FIXME: We should teach the "split and lower" path to do something more + // clever, or do it ourselves here. The optimal lowering of cross-lane + // shuffles I am aware of is to swap the lanes into a copy, shuffle both the + // original and the copy, and then blend to pick up the cross-lane elements. + // This is four instructions with a tree height of three which is better than + // the worst case for a gather-cross-scatter approach such as used in SSE2 + // v8i16 lowering (where we don't have blends). While for cross-lane blends it + // results in a blend tree, blends are very cheap in AVX2 and newer chips. We + // might also want to special case situations where we can always do a single + // VPERMD to produce a non-lane-crossing shuffle. + if (is128BitLaneCrossingShuffleMask(MVT::v32i8, Mask)) + return splitAndLower256BitVectorShuffle(Op, V1, V2, Subtarget, DAG); + + if (SDValue Blend = lowerVectorShuffleAsBlend(DL, MVT::v32i8, V1, V2, Mask, + Subtarget, DAG)) + return Blend; + + // Use dedicated unpack instructions for masks that match their pattern. + // Note that these are repeated 128-bit lane unpacks, not unpacks across all + // 256-bit lanes. + if (isShuffleEquivalent( + Mask, + // First 128-bit lane: + 0, 32, 1, 33, 2, 34, 3, 35, 4, 36, 5, 37, 6, 38, 7, 39, + // Second 128-bit lane: + 16, 48, 17, 49, 18, 50, 19, 51, 20, 52, 21, 53, 22, 54, 23, 55)) + return DAG.getNode(X86ISD::UNPCKL, DL, MVT::v32i8, V1, V2); + if (isShuffleEquivalent( + Mask, + // First 128-bit lane: + 8, 40, 9, 41, 10, 42, 11, 43, 12, 44, 13, 45, 14, 46, 15, 47, + // Second 128-bit lane: + 24, 56, 25, 57, 26, 58, 27, 59, 28, 60, 29, 61, 30, 62, 31, 63)) + return DAG.getNode(X86ISD::UNPCKH, DL, MVT::v32i8, V1, V2); + + if (isSingleInputShuffleMask(Mask)) { + SDValue PSHUFBMask[32]; + for (int i = 0; i < 32; ++i) + PSHUFBMask[i] = + Mask[i] < 0 + ? DAG.getUNDEF(MVT::i8) + : DAG.getConstant(Mask[i] < 16 ? Mask[i] : Mask[i] - 16, MVT::i8); + + return DAG.getNode( + X86ISD::PSHUFB, DL, MVT::v32i8, V1, + DAG.getNode(ISD::BUILD_VECTOR, DL, MVT::v32i8, PSHUFBMask)); + } + + // Otherwise fall back on generic blend lowering. + return lowerVectorShuffleAsDecomposedShuffleBlend(DL, MVT::v32i8, V1, V2, + Mask, DAG); } /// \brief High-level routine to lower various 256-bit x86 vector shuffles. diff --git a/test/CodeGen/X86/vector-shuffle-256-v32.ll b/test/CodeGen/X86/vector-shuffle-256-v32.ll index e2a25d451f7..127159db2c4 100644 --- a/test/CodeGen/X86/vector-shuffle-256-v32.ll +++ b/test/CodeGen/X86/vector-shuffle-256-v32.ll @@ -757,11 +757,8 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2 -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpxor %ymm1, %ymm1, %ymm1 +; AVX2-NEXT: vpshufb %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -779,11 +776,7 @@ define <32 x i8> @shuffle_v32i8_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_ ; ; AVX2-LABEL: @shuffle_v32i8_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_31_31_31_31_31_31_31_31_31_31_31_31_31_31_31_31 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -801,11 +794,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_16_16_16_16_16_16_16_16_24_24_24_24_24_24_24_24 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [0,0,0,0,0,0,0,0,8,8,8,8,8,8,8,8] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,0,0,0,0,8,8,8,8,8,8,8,8,16,16,16,16,16,16,16,16,24,24,24,24,24,24,24,24] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -823,11 +812,7 @@ define <32 x i8> @shuffle_v32i8_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15_ ; ; AVX2-LABEL: @shuffle_v32i8_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15_23_23_23_23_23_23_23_23_31_31_31_31_31_31_31_31 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [7,7,7,7,7,7,7,7,15,15,15,15,15,15,15,15] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[7,7,7,7,7,7,7,7,15,15,15,15,15,15,15,15,23,23,23,23,23,23,23,23,31,31,31,31,31,31,31,31] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -845,11 +830,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_20_24_24_24_24_28_28_28_28 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [0,0,0,0,4,4,4,4,8,8,8,8,12,12,12,12] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,4,4,4,4,8,8,8,8,12,12,12,12,16,16,16,16,20,20,20,20,24,24,24,24,28,28,28,28] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -867,11 +848,7 @@ define <32 x i8> @shuffle_v32i8_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_15_ ; ; AVX2-LABEL: @shuffle_v32i8_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_15_19_19_19_19_23_23_23_23_27_27_27_27_31_31_31_31 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15,19,19,19,19,23,23,23,23,27,27,27,27,31,31,31,31] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -889,11 +866,7 @@ define <32 x i8> @shuffle_v32i8_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_14_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_14_16_16_18_18_20_20_22_22_24_24_26_26_28_28_30_30 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -911,11 +884,7 @@ define <32 x i8> @shuffle_v32i8_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_15_ ; ; AVX2-LABEL: @shuffle_v32i8_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_15_17_17_19_19_21_21_23_23_25_25_27_27_29_29_31_31 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15,17,17,19,19,21,21,23,23,25,25,27,27,29,29,31,31] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1039,17 +1008,8 @@ define <32 x i8> @shuffle_v32i8_00_33_02_35_04_37_06_39_08_41_10_43_12_45_14_47_ ; ; AVX2-LABEL: @shuffle_v32i8_00_33_02_35_04_37_06_39_08_41_10_43_12_45_14_47_16_49_18_51_20_53_22_55_24_57_26_59_28_61_30_63 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2 -; AVX2-NEXT: vmovdqa {{.*}} # xmm3 = <1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u> -; AVX2-NEXT: vpshufb %xmm3, %xmm2, %xmm2 -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm4 -; AVX2-NEXT: vmovdqa {{.*}} # xmm5 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u> -; AVX2-NEXT: vpshufb %xmm5, %xmm4, %xmm4 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3],xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7] -; AVX2-NEXT: vpshufb %xmm3, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm5, %xmm0, %xmm0 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] -; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0 +; AVX2-NEXT: vmovdqa {{.*}} # ymm2 = [0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128] +; AVX2-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1073,17 +1033,8 @@ define <32 x i8> @shuffle_v32i8_32_01_34_03_36_05_38_07_40_09_42_11_44_13_46_15_ ; ; AVX2-LABEL: @shuffle_v32i8_32_01_34_03_36_05_38_07_40_09_42_11_44_13_46_15_48_17_50_19_52_21_54_23_56_25_58_27_60_29_62_31 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2 -; AVX2-NEXT: vmovdqa {{.*}} # xmm3 = <1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u> -; AVX2-NEXT: vpshufb %xmm3, %xmm2, %xmm2 -; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm4 -; AVX2-NEXT: vmovdqa {{.*}} # xmm5 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u> -; AVX2-NEXT: vpshufb %xmm5, %xmm4, %xmm4 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3],xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7] -; AVX2-NEXT: vpshufb %xmm3, %xmm0, %xmm0 -; AVX2-NEXT: vpshufb %xmm5, %xmm1, %xmm1 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] -; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0 +; AVX2-NEXT: vmovdqa {{.*}} # ymm2 = [0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128] +; AVX2-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1123,13 +1074,11 @@ define <32 x i8> @shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_ ; ; AVX2-LABEL: @shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_16_48_16_48_16_48_16_48_16_48_16_48_16_48_16_48 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2 -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpxor %ymm2, %ymm2, %ymm2 +; AVX2-NEXT: vpshufb %ymm2, %ymm1, %ymm1 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17] +; AVX2-NEXT: vmovdqa {{.*}} # ymm2 = [0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128] +; AVX2-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1151,15 +1100,9 @@ define <32 x i8> @shuffle_v32i8_32_32_32_32_32_32_32_32_08_09_10_11_12_13_14_15_ ; ; AVX2-LABEL: @shuffle_v32i8_32_32_32_32_32_32_32_32_08_09_10_11_12_13_14_15_48_48_48_48_48_48_48_48_24_25_26_27_28_29_30_31 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2 -; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm3 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] -; AVX2-NEXT: vpshuflw {{.*}} # xmm3 = xmm3[0,0,0,0,4,5,6,7] -; AVX2-NEXT: vpblendd {{.*}} # xmm2 = xmm3[0,1],xmm2[2,3] -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] -; AVX2-NEXT: vpshuflw {{.*}} # xmm1 = xmm1[0,0,0,0,4,5,6,7] -; AVX2-NEXT: vpblendd {{.*}} # xmm0 = xmm1[0,1],xmm0[2,3] -; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0 +; AVX2-NEXT: vpxor %ymm2, %ymm2, %ymm2 +; AVX2-NEXT: vpshufb %ymm2, %ymm1, %ymm1 +; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1183,17 +1126,9 @@ define <32 x i8> @shuffle_v32i8_39_38_37_36_35_34_33_32_15_14_13_12_11_10_09_08_ ; ; AVX2-LABEL: @shuffle_v32i8_39_38_37_36_35_34_33_32_15_14_13_12_11_10_09_08_55_54_53_52_51_50_49_48_31_30_29_28_27_26_25_24 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2 -; AVX2-NEXT: vmovdqa {{.*}} # xmm3 = -; AVX2-NEXT: vpshufb %xmm3, %xmm2, %xmm2 -; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm4 -; AVX2-NEXT: vmovdqa {{.*}} # xmm5 = <7,6,5,4,3,2,1,0,u,u,u,u,u,u,u,u> -; AVX2-NEXT: vpshufb %xmm5, %xmm4, %xmm4 -; AVX2-NEXT: vpblendd {{.*}} # xmm2 = xmm4[0,1],xmm2[2,3] -; AVX2-NEXT: vpshufb %xmm3, %xmm0, %xmm0 -; AVX2-NEXT: vpshufb %xmm5, %xmm1, %xmm1 -; AVX2-NEXT: vpblendd {{.*}} # xmm0 = xmm1[0,1],xmm0[2,3] -; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[u,u,u,u,u,u,u,u,15,14,13,12,11,10,9,8,u,u,u,u,u,u,u,u,31,30,29,28,27,26,25,24] +; AVX2-NEXT: vpshufb {{.*}} # ymm1 = ymm1[7,6,5,4,3,2,1,0,u,u,u,u,u,u,u,u,23,22,21,20,19,18,17,16,u,u,u,u,u,u,u,u] +; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1217,17 +1152,9 @@ define <32 x i8> @shuffle_v32i8_39_38_37_36_35_34_33_32_07_06_05_04_03_02_01_00_ ; ; AVX2-LABEL: @shuffle_v32i8_39_38_37_36_35_34_33_32_07_06_05_04_03_02_01_00_55_54_53_52_51_50_49_48_23_22_21_20_19_18_17_16 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2 -; AVX2-NEXT: vmovdqa {{.*}} # xmm3 = -; AVX2-NEXT: vpshufb %xmm3, %xmm2, %xmm2 -; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm4 -; AVX2-NEXT: vmovdqa {{.*}} # xmm5 = <7,6,5,4,3,2,1,0,u,u,u,u,u,u,u,u> -; AVX2-NEXT: vpshufb %xmm5, %xmm4, %xmm4 -; AVX2-NEXT: vpblendd {{.*}} # xmm2 = xmm4[0,1],xmm2[2,3] -; AVX2-NEXT: vpshufb %xmm3, %xmm0, %xmm0 -; AVX2-NEXT: vpshufb %xmm5, %xmm1, %xmm1 -; AVX2-NEXT: vpblendd {{.*}} # xmm0 = xmm1[0,1],xmm0[2,3] -; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[u,u,u,u,u,u,u,u,7,6,5,4,3,2,1,0,u,u,u,u,u,u,u,u,23,22,21,20,19,18,17,16] +; AVX2-NEXT: vpshufb {{.*}} # ymm1 = ymm1[7,6,5,4,3,2,1,0,u,u,u,u,u,u,u,u,23,22,21,20,19,18,17,16,u,u,u,u,u,u,u,u] +; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1245,11 +1172,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00_16_16_16_16_16_16_16_16_16_16_16_16_16_16_17_16 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,17,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1267,11 +1190,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00_16_16_16_16_16_16_16_16_16_16_16_16_16_18_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,18,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1289,11 +1208,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00_16_16_16_16_16_16_16_16_23_16_16_16_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0,16,16,16,16,16,16,16,16,23,16,16,16,16,16,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1311,11 +1226,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00_16_16_16_16_16_16_16_24_16_16_16_16_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0,16,16,16,16,16,16,16,24,16,16,16,16,16,16,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1333,11 +1244,7 @@ define <32 x i8> @shuffle_v32i8_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00_16_30_16_16_16_16_16_16_16_16_16_16_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = [0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0] -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0,16,30,16,16,16,16,16,16,16,16,16,16,16,16,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1356,12 +1263,7 @@ define <32 x i8> @shuffle_v32i8_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_31_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: movl $15, %eax -; AVX2-NEXT: vmovd %eax, %xmm1 -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2 -; AVX2-NEXT: vpshufb %xmm1, %xmm2, %xmm2 -; AVX2-NEXT: vpshufb %xmm1, %xmm0, %xmm0 -; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,31,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1383,15 +1285,7 @@ define <32 x i8> @shuffle_v32i8_00_32_01_33_02_34_03_35_04_36_05_37_06_38_07_39_ ; ; AVX2-LABEL: @shuffle_v32i8_00_32_01_33_02_34_03_35_04_36_05_37_06_38_07_39_16_48_17_49_18_50_19_51_20_52_21_53_22_54_23_55 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vpmovzxbw %xmm1, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u> -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] -; AVX2-NEXT: vpmovzxbw %xmm0, %xmm0 -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpunpcklbw {{.*}} # ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1411,13 +1305,7 @@ define <32 x i8> @shuffle_v32i8_08_40_09_41_10_42_11_43_12_44_13_45_14_46_15_47_ ; ; AVX2-LABEL: @shuffle_v32i8_08_40_09_41_10_42_11_43_12_44_13_45_14_46_15_47_24_56_25_57_26_58_27_59_28_60_29_61_30_62_31_63 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vmovdqa {{.*}} # xmm2 = <8,9,10,11,12,13,14,15,u,u,u,u,u,u,u,u> -; AVX2-NEXT: vpshufb %xmm2, %xmm1, %xmm1 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] -; AVX2-NEXT: vpshufb %xmm2, %xmm0, %xmm0 -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpunpckhbw {{.*}} # ymm0 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1437,13 +1325,10 @@ define <32 x i8> @shuffle_v32i8_00_32_01_33_02_34_03_35_04_36_05_37_06_38_07_39_ ; ; AVX2-LABEL: @shuffle_v32i8_00_32_01_33_02_34_03_35_04_36_05_37_06_38_07_39_24_56_25_57_26_58_27_59_28_60_29_61_30_62_31_63 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm1[8,9,10,11,12,13,14,15,u,u,u,u,u,u,u,u] -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] -; AVX2-NEXT: vpmovzxbw %xmm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,u,1,u,2,u,3,u,4,u,5,u,6,u,7,u,24,u,25,u,26,u,27,u,28,u,29,u,30,u,31,u] +; AVX2-NEXT: vpshufb {{.*}} # ymm1 = ymm1[u,0,u,1,u,2,u,3,u,4,u,5,u,6,u,7,u,24,u,25,u,26,u,27,u,28,u,29,u,30,u,31] +; AVX2-NEXT: vmovdqa {{.*}} # ymm2 = [0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128] +; AVX2-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1463,13 +1348,10 @@ define <32 x i8> @shuffle_v32i8_08_40_09_41_10_42_11_43_12_44_13_45_14_46_15_47_ ; ; AVX2-LABEL: @shuffle_v32i8_08_40_09_41_10_42_11_43_12_44_13_45_14_46_15_47_16_48_17_49_18_50_19_51_20_52_21_53_22_54_23_55 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vpmovzxbw %xmm1, %xmm1 -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm1[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[8,9,10,11,12,13,14,15,u,u,u,u,u,u,u,u] -; AVX2-NEXT: vpunpcklbw {{.*}} # xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[8,u,9,u,10,u,11,u,12,u,13,u,14,u,15,u,16,u,17,u,18,u,19,u,20,u,21,u,22,u,23,u] +; AVX2-NEXT: vpshufb {{.*}} # ymm1 = ymm1[u,8,u,9,u,10,u,11,u,12,u,13,u,14,u,15,u,16,u,17,u,18,u,19,u,20,u,21,u,22,u,23] +; AVX2-NEXT: vmovdqa {{.*}} # ymm2 = [0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128,0,128] +; AVX2-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1486,10 +1368,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00_16_17_16_16_16_16_16_16_16_16_16_16_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0] -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,16,17,16,16,16,16,16,16,16,16,16,16,16,16,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1506,10 +1385,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00_16_16_18_16_16_16_16_16_16_16_16_16_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0] -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[0,0,2,0,0,0,0,0,0,0,0,0,0,0,0,0] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0,16,16,18,16,16,16,16,16,16,16,16,16,16,16,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1526,10 +1402,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00_16_16_16_16_16_16_16_23_16_16_16_16_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0] -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[0,0,0,0,0,0,0,7,0,0,0,0,0,0,0,0] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0,16,16,16,16,16,16,16,23,16,16,16,16,16,16,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1546,10 +1419,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00_16_16_16_16_16_16_16_16_24_16_16_16_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm0[0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0] -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[0,0,0,0,0,0,0,0,8,0,0,0,0,0,0,0] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0,16,16,16,16,16,16,16,16,24,16,16,16,16,16,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1566,10 +1436,7 @@ define <32 x i8> @shuffle_v32i8_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00_16_16_16_16_16_16_16_16_16_16_16_16_16_16_30_16 ; AVX2: # BB#0: -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm0[0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0] -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,14,0] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,30,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1588,12 +1455,7 @@ define <32 x i8> @shuffle_v32i8_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_31 ; AVX2: # BB#0: -; AVX2-NEXT: movl $15, %eax -; AVX2-NEXT: vmovd %eax, %xmm1 -; AVX2-NEXT: vpshufb %xmm1, %xmm0, %xmm1 -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,15] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,31] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1610,10 +1472,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12_28_28_28_28_24_24_24_24_20_20_20_20_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm0[0,0,0,0,4,4,4,4,8,8,8,8,12,12,12,12] -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[12,12,12,12,8,8,8,8,4,4,4,4,0,0,0,0] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,0,4,4,4,4,8,8,8,8,12,12,12,12,28,28,28,28,24,24,24,24,20,20,20,20,16,16,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1630,10 +1489,7 @@ define <32 x i8> @shuffle_v32i8_08_08_08_08_08_08_08_08_00_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_08_08_08_08_08_08_08_08_00_00_00_00_00_00_00_00_16_16_16_16_16_16_16_16_24_24_24_24_24_24_24_24 ; AVX2: # BB#0: -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm0[8,8,8,8,8,8,8,8,0,0,0,0,0,0,0,0] -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[0,0,0,0,0,0,0,0,8,8,8,8,8,8,8,8] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[8,8,8,8,8,8,8,8,0,0,0,0,0,0,0,0,16,16,16,16,16,16,16,16,24,24,24,24,24,24,24,24] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1649,9 +1505,7 @@ define <32 x i8> @shuffle_v32i8_00_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_ ; ; AVX2-LABEL: @shuffle_v32i8_00_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_16_16_16_16_uu_uu_uu_uu_uu_16_16_16_16_16_30_16 ; AVX2: # BB#0: -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm1[0,0,0,0,u,u,u,u,u,0,0,0,0,0,14,0] -; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,16,16,16,u,u,u,u,u,16,16,16,16,16,30,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1668,10 +1522,7 @@ define <32 x i8> @shuffle_v32i8_uu_14_uu_uu_00_00_00_00_00_00_00_00_00_00_00_00_ ; ; AVX2-LABEL: @shuffle_v32i8_uu_14_uu_uu_00_00_00_00_00_00_00_00_00_00_00_00_16_16_uu_16_uu_uu_uu_uu_16_16_16_16_16_16_30_16 ; AVX2: # BB#0: -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm0[0,0,1,1,0,0,0,0,0,0,0,0,0,0,0,0] -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[0,0,u,0,u,u,u,u,0,0,0,0,0,0,14,0] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[u,14,u,u,0,0,0,0,0,0,0,0,0,0,0,0,16,16,u,16,u,u,u,u,16,16,16,16,16,16,30,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1688,10 +1539,7 @@ define <32 x i8> @shuffle_v32i8_00_00_00_uu_uu_uu_04_uu_08_08_08_08_uu_uu_12_uu_ ; ; AVX2-LABEL: @shuffle_v32i8_00_00_00_uu_uu_uu_04_uu_08_08_08_08_uu_uu_12_uu_28_28_28_28_uu_uu_uu_24_20_20_20_20_16_16_16_16 ; AVX2: # BB#0: -; AVX2-NEXT: vpshufb {{.*}} # xmm1 = xmm0[0,0,0,0,4,4,4,4,8,8,8,8,12,12,12,12] -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[12,12,12,12,12,12,13,13,4,4,4,4,0,0,0,0] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[0,0,0,u,u,u,4,u,8,8,8,8,u,u,12,u,28,28,28,28,u,u,u,24,20,20,20,20,16,16,16,16] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle @@ -1709,11 +1557,7 @@ define <32 x i8> @shuffle_v32i8_08_08_08_08_08_08_08_08_uu_uu_uu_uu_uu_uu_uu_uu_ ; ; AVX2-LABEL: @shuffle_v32i8_08_08_08_08_08_08_08_08_uu_uu_uu_uu_uu_uu_uu_uu_16_16_16_uu_uu_uu_uu_uu_uu_uu_24_24_24_24_24_24 ; AVX2: # BB#0: -; AVX2-NEXT: vpunpckhbw {{.*}} # xmm1 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX2-NEXT: vpshuflw {{.*}} # xmm1 = xmm1[0,0,0,0,4,5,6,7] -; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX2-NEXT: vpshufb {{.*}} # xmm0 = xmm0[0,0,0,0,8,8,9,9,8,8,8,8,8,8,8,8] -; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vpshufb {{.*}} # ymm0 = ymm0[8,8,8,8,8,8,8,8,u,u,u,u,u,u,u,u,16,16,16,u,u,u,u,u,u,u,24,24,24,24,24,24] ; AVX2-NEXT: retq %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> ret <32 x i8> %shuffle -- 2.34.1