Revert rGe82e17d4d4cac8b2df00094e80d5e1cb22795664 - [X86] Add lowerShuffleAsBitRotate...

As noted on PR44379, we didn't attempt to lower vector shuffles using bit rotations on XOP/AVX512F targets.

This patch lowers to uniform ISD:ROTL nodes - ROTR isn't supported by XOP and they are interchangeable for constant values anyway.

There might be cases where targets without ISD:ROTL support would benefit from this (expanding to SRL+SHL+OR), which I'll investigate in a future patch.

Also, non-AVX512BW targets fail to concatenate 256-bit rotations back to 512-bits (split during shuffle lowering as they don't have v32i16/v64i8 types).
---
Internal shuffle tests indicate theres a bug somewhere that I haven't been able to track down yet.

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 279c91b..98981b9 100644 (file)
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -11665,66 +11665,6 @@ static SDValue lowerShuffleAsDecomposedShuffleBlend(
   return DAG.getVectorShuffle(VT, DL, V1, V2, BlendMask);
 }
 
-/// Try to lower a vector shuffle as a bit rotation.
-///
-/// Look for a repeated rotation pattern in each sub group.
-/// Returns a ISD::ROTL element rotation amount or -1 if failed.
-static int matchShuffleAsBitRotate(ArrayRef<int> Mask, int NumSubElts) {
-  int NumElts = Mask.size();
-  assert((NumElts % NumSubElts) == 0 && "Illegal shuffle mask");
-
-  int RotateAmt = -1;
-  for (int i = 0; i != NumElts; i += NumSubElts) {
-    for (int j = 0; j != NumSubElts; ++j) {
-      int M = Mask[i + j];
-      if (M < 0)
-        continue;
-      if (!isInRange(M, i, i + NumSubElts))
-        return -1;
-      int Offset = ((M - i) + (NumSubElts - j)) % NumSubElts;
-      if (0 <= RotateAmt && Offset != RotateAmt)
-        return -1;
-      RotateAmt = Offset;
-    }
-  }
-  return RotateAmt;
-}
-
-/// Lower shuffle using ISD::ROTL rotations.
-static SDValue lowerShuffleAsBitRotate(const SDLoc &DL, MVT VT, SDValue V1,
-                                       ArrayRef<int> Mask,
-                                       const X86Subtarget &Subtarget,
-                                       SelectionDAG &DAG) {
-  assert(!isNoopShuffleMask(Mask) && "We shouldn't lower no-op shuffles!");
-
-  MVT SVT = VT.getScalarType();
-  int EltSizeInBits = SVT.getScalarSizeInBits();
-  assert(EltSizeInBits < 64 && "Can't rotate 64-bit integers");
-
-  // Only XOP + AVX512 targets have bit rotation instructions.
-  if (!((VT.is128BitVector() && Subtarget.hasXOP()) || Subtarget.hasAVX512()))
-    return SDValue();
-
-  // AVX512 only has vXi32/vXi64 rotates, so limit the rotation sub group size.
-  int MinSubElts = Subtarget.hasXOP() ? 2 : std::max(32 / EltSizeInBits, 2);
-  int MaxSubElts = 64 / EltSizeInBits;
-  for (int NumSubElts = MinSubElts; NumSubElts <= MaxSubElts; NumSubElts *= 2) {
-    int RotateAmt = matchShuffleAsBitRotate(Mask, NumSubElts);
-    if (RotateAmt < 0)
-      continue;
-    int NumElts = VT.getVectorNumElements();
-    MVT RotateSVT = MVT::getIntegerVT(EltSizeInBits * NumSubElts);
-    MVT RotateVT = MVT::getVectorVT(RotateSVT, NumElts / NumSubElts);
-    int RotateAmtInBits = RotateAmt * EltSizeInBits;
-    SDValue Rot =
-        DAG.getNode(ISD::ROTL, DL, RotateVT, DAG.getBitcast(RotateVT, V1),
-                    DAG.getConstant(RotateAmtInBits, DL, RotateVT));
-    return DAG.getBitcast(VT, Rot);
-  }
-
-  return SDValue();
-}
-
 /// Try to lower a vector shuffle as a byte rotation.
 ///
 /// This is used for support PALIGNR for SSSE3 or VALIGND/Q for AVX512.
@@ -14280,11 +14220,6 @@ static SDValue lowerV8I16Shuffle(const SDLoc &DL, ArrayRef<int> Mask,
                                                     Mask, Subtarget, DAG))
       return Broadcast;
 
-    // Try to use bit rotation instructions.
-    if (SDValue Rotate = lowerShuffleAsBitRotate(DL, MVT::v8i16, V1, Mask,
-                                                 Subtarget, DAG))
-      return Rotate;
-
     // Use dedicated unpack instructions for masks that match their pattern.
     if (SDValue V = lowerShuffleWithUNPCK(DL, MVT::v8i16, Mask, V1, V2, DAG))
       return V;
@@ -14509,11 +14444,6 @@ static SDValue lowerV16I8Shuffle(const SDLoc &DL, ArrayRef<int> Mask,
                                                     Mask, Subtarget, DAG))
       return Broadcast;
 
-    // Try to use bit rotation instructions.
-    if (SDValue Rotate = lowerShuffleAsBitRotate(DL, MVT::v16i8, V1, Mask,
-                                                 Subtarget, DAG))
-      return Rotate;
-
     if (SDValue V = lowerShuffleWithUNPCK(DL, MVT::v16i8, Mask, V1, V2, DAG))
       return V;
 
@@ -16404,11 +16334,6 @@ static SDValue lowerV16I16Shuffle(const SDLoc &DL, ArrayRef<int> Mask,
     return V;
 
   if (V2.isUndef()) {
-    // Try to use bit rotation instructions.
-    if (SDValue Rotate =
-            lowerShuffleAsBitRotate(DL, MVT::v16i16, V1, Mask, Subtarget, DAG))
-      return Rotate;
-
     // Try to produce a fixed cross-128-bit lane permute followed by unpack
     // because that should be faster than the variable permute alternatives.
     if (SDValue V = lowerShuffleWithUNPCK256(DL, MVT::v16i16, Mask, V1, V2, DAG))
@@ -16507,12 +16432,6 @@ static SDValue lowerV32I8Shuffle(const SDLoc &DL, ArrayRef<int> Mask,
                                                 Subtarget, DAG))
     return Rotate;
 
-  // Try to use bit rotation instructions.
-  if (V2.isUndef())
-    if (SDValue Rotate =
-            lowerShuffleAsBitRotate(DL, MVT::v32i8, V1, Mask, Subtarget, DAG))
-      return Rotate;
-
   // Try to create an in-lane repeating shuffle mask and then shuffle the
   // results into the target lanes.
   if (SDValue V = lowerShuffleAsRepeatedMaskAndLanePermute(
@@ -17007,11 +16926,6 @@ static SDValue lowerV32I16Shuffle(const SDLoc &DL, ArrayRef<int> Mask,
     return Rotate;
 
   if (V2.isUndef()) {
-    // Try to use bit rotation instructions.
-    if (SDValue Rotate =
-            lowerShuffleAsBitRotate(DL, MVT::v32i16, V1, Mask, Subtarget, DAG))
-      return Rotate;
-
     SmallVector<int, 8> RepeatedMask;
     if (is128BitLaneRepeatedShuffleMask(MVT::v32i16, Mask, RepeatedMask)) {
       // As this is a single-input shuffle, the repeated mask should be
@@ -17069,12 +16983,6 @@ static SDValue lowerV64I8Shuffle(const SDLoc &DL, ArrayRef<int> Mask,
                                                 Subtarget, DAG))
     return Rotate;
 
-  // Try to use bit rotation instructions.
-  if (V2.isUndef())
-    if (SDValue Rotate =
-            lowerShuffleAsBitRotate(DL, MVT::v64i8, V1, Mask, Subtarget, DAG))
-      return Rotate;
-
   if (SDValue PSHUFB = lowerShuffleWithPSHUFB(DL, MVT::v64i8, Mask, V1, V2,
                                               Zeroable, Subtarget, DAG))
     return PSHUFB;

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-128-v16.ll b/llvm/test/CodeGen/X86/vector-shuffle-128-v16.ll
index 8b3843d..a44fd17 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shuffle-128-v16.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-128-v16.ll
@@ -479,20 +479,10 @@ define <16 x i8> @shuffle_v16i8_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14(
 ; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
 ; SSE41-NEXT:    retq
 ;
-; AVX1-LABEL: shuffle_v16i8_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
-; AVX1-NEXT:    retq
-;
-; AVX2OR512VL-LABEL: shuffle_v16i8_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
-; AVX2OR512VL-NEXT:    retq
-;
-; XOP-LABEL: shuffle_v16i8_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14:
-; XOP:       # %bb.0:
-; XOP-NEXT:    vprotw $8, %xmm0, %xmm0
-; XOP-NEXT:    retq
+; AVX-LABEL: shuffle_v16i8_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; AVX-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
   ret <16 x i8> %shuffle
 }
@@ -1912,25 +1902,10 @@ define <16 x i8> @shuffle_v16i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14(
 ; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[3,0,1,2,7,4,5,6,11,8,9,10,15,12,13,14]
 ; SSE41-NEXT:    retq
 ;
-; AVX1-LABEL: shuffle_v16i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[3,0,1,2,7,4,5,6,11,8,9,10,15,12,13,14]
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: shuffle_v16i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[3,0,1,2,7,4,5,6,11,8,9,10,15,12,13,14]
-; AVX2-NEXT:    retq
-;
-; AVX512VL-LABEL: shuffle_v16i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vprold $24, %xmm0, %xmm0
-; AVX512VL-NEXT:    retq
-;
-; XOP-LABEL: shuffle_v16i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14:
-; XOP:       # %bb.0:
-; XOP-NEXT:    vprotd $24, %xmm0, %xmm0
-; XOP-NEXT:    retq
+; AVX-LABEL: shuffle_v16i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[3,0,1,2,7,4,5,6,11,8,9,10,15,12,13,14]
+; AVX-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> undef, <16 x i32> <i32 3, i32 0, i32 1, i32 2, i32 7, i32 4, i32 5, i32 6, i32 11, i32 8, i32 9, i32 10, i32 15, i32 12, i32 13, i32 14>
   ret <16 x i8> %shuffle
 }
@@ -1962,12 +1937,13 @@ define <16 x i8> @shuffle_v16i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09(
 ;
 ; AVX512VL-LABEL: shuffle_v16i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vprolq $16, %xmm0, %xmm0
+; AVX512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,0,1,10,11,12,13,14,15,8,9]
 ; AVX512VL-NEXT:    retq
 ;
 ; XOP-LABEL: shuffle_v16i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09:
 ; XOP:       # %bb.0:
-; XOP-NEXT:    vprotq $16, %xmm0, %xmm0
+; XOP-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
+; XOP-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; XOP-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 8, i32 9>
   ret <16 x i8> %shuffle

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll b/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll
index 6b9bb14..a0584ff 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll
@@ -2462,14 +2462,21 @@ define <8 x i16> @shuffle_v8i16_10325476(<8 x i16> %a) {
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,0,1,6,7,4,5,10,11,8,9,14,15,12,13]
 ; AVX2-FAST-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v8i16_10325476:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vprold $16, %xmm0, %xmm0
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_10325476:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,6]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_10325476:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,0,1,6,7,4,5,10,11,8,9,14,15,12,13]
+; AVX512VL-FAST-NEXT:    retq
 ;
 ; XOP-LABEL: shuffle_v8i16_10325476:
 ; XOP:       # %bb.0:
-; XOP-NEXT:    vprotd $16, %xmm0, %xmm0
+; XOP-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
+; XOP-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,6]
 ; XOP-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> zeroinitializer, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
   ret <8 x i16> %shuffle
@@ -2499,14 +2506,21 @@ define <8 x i16> @shuffle_v8i16_12305674(<8 x i16> %a) {
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,0,1,10,11,12,13,14,15,8,9]
 ; AVX2-FAST-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v8i16_12305674:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vprolq $16, %xmm0, %xmm0
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_12305674:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_12305674:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,0,1,10,11,12,13,14,15,8,9]
+; AVX512VL-FAST-NEXT:    retq
 ;
 ; XOP-LABEL: shuffle_v8i16_12305674:
 ; XOP:       # %bb.0:
-; XOP-NEXT:    vprotq $16, %xmm0, %xmm0
+; XOP-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
+; XOP-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; XOP-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> zeroinitializer, <8 x i32> <i32 1, i32 2, i32 3, i32 0, i32 5, i32 6, i32 7, i32 4>
   ret <8 x i16> %shuffle

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
index 52b710b..1c2b52e 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -5385,7 +5385,7 @@ define <16 x i16> @shuffle_v16i16_01_00_17_16_03_02_19_26_09_08_25_24_11_10_27_2
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; XOPAVX1-NEXT:    vpperm {{.*#+}} xmm3 = xmm3[2,3,0,1],xmm2[2,3,0,1],xmm3[6,7,4,5],xmm2[6,7,4,5]
 ; XOPAVX1-NEXT:    vpperm {{.*#+}} xmm1 = xmm1[2,3,0,1,6,7],xmm2[4,5],xmm1[4,5],xmm2[4,5],xmm1[6,7],xmm2[4,5]
-; XOPAVX1-NEXT:    vprotd $16, %xmm0, %xmm0
+; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
 ; XOPAVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; XOPAVX1-NEXT:    retq
@@ -7054,16 +7054,24 @@ define <16 x i16> @shuffle_v16i16_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_1
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5,10,11,8,9,14,15,12,13,18,19,16,17,22,23,20,21,26,27,24,25,30,31,28,29]
 ; AVX2-FAST-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v16i16_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vprold $16, %ymm0, %ymm0
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v16i16_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v16i16_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5,10,11,8,9,14,15,12,13,18,19,16,17,22,23,20,21,26,27,24,25,30,31,28,29]
+; AVX512VL-FAST-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: shuffle_v16i16_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vprotd $16, %xmm0, %xmm1
+; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[1,0,3,2,4,5,6,7]
+; XOPAVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,4,7,6]
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; XOPAVX1-NEXT:    vprotd $16, %xmm0, %xmm0
+; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
+; XOPAVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,6]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; XOPAVX1-NEXT:    retq
 ;
@@ -7098,16 +7106,24 @@ define <16 x i16> @shuffle_v16i16_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_1
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,0,1,2,3,4,5,14,15,8,9,10,11,12,13,22,23,16,17,18,19,20,21,30,31,24,25,26,27,28,29]
 ; AVX2-FAST-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v16i16_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vprolq $48, %ymm0, %ymm0
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v16i16_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,0,1,2,4,5,6,7,11,8,9,10,12,13,14,15]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,4,5,6,8,9,10,11,15,12,13,14]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v16i16_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,0,1,2,3,4,5,14,15,8,9,10,11,12,13,22,23,16,17,18,19,20,21,30,31,24,25,26,27,28,29]
+; AVX512VL-FAST-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: shuffle_v16i16_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vprotq $48, %xmm0, %xmm1
+; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[3,0,1,2,4,5,6,7]
+; XOPAVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,5,6]
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; XOPAVX1-NEXT:    vprotq $48, %xmm0, %xmm0
+; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,0,1,2,4,5,6,7]
+; XOPAVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,4,5,6]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; XOPAVX1-NEXT:    retq
 ;

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v32.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v32.ll
index e85b120..7d71316 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v32.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v32.ll
@@ -4678,22 +4678,18 @@ define <32 x i8> @shuffle_v32i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v32i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_19_16_17_18_23_20_21_22_27_24_25_26_31_28_29_30:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[3,0,1,2,7,4,5,6,11,8,9,10,15,12,13,14,19,16,17,18,23,20,21,22,27,24,25,26,31,28,29,30]
-; AVX2-NEXT:    retq
-;
-; AVX512VL-LABEL: shuffle_v32i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_19_16_17_18_23_20_21_22_27_24_25_26_31_28_29_30:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vprold $24, %ymm0, %ymm0
-; AVX512VL-NEXT:    retq
+; AVX2OR512VL-LABEL: shuffle_v32i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_19_16_17_18_23_20_21_22_27_24_25_26_31_28_29_30:
+; AVX2OR512VL:       # %bb.0:
+; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[3,0,1,2,7,4,5,6,11,8,9,10,15,12,13,14,19,16,17,18,23,20,21,22,27,24,25,26,31,28,29,30]
+; AVX2OR512VL-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: shuffle_v32i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_19_16_17_18_23_20_21_22_27_24_25_26_31_28_29_30:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vprotd $24, %xmm0, %xmm1
-; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; XOPAVX1-NEXT:    vprotd $24, %xmm0, %xmm0
-; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; XOPAVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [3,0,1,2,7,4,5,6,11,8,9,10,15,12,13,14]
+; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v32i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_19_16_17_18_23_20_21_22_27_24_25_26_31_28_29_30:
@@ -4727,16 +4723,35 @@ define <32 x i8> @shuffle_v32i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,4,5,6,7,0,1,10,11,12,13,14,15,8,9,18,19,20,21,22,23,16,17,26,27,28,29,30,31,24,25]
 ; AVX2-FAST-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v32i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vprolq $16, %ymm0, %ymm0
-; AVX512VL-NEXT:    retq
+; AVX512VLBW-SLOW-LABEL: shuffle_v32i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25:
+; AVX512VLBW-SLOW:       # %bb.0:
+; AVX512VLBW-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,2,3,0,4,5,6,7,9,10,11,8,12,13,14,15]
+; AVX512VLBW-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,6,7,4,8,9,10,11,13,14,15,12]
+; AVX512VLBW-SLOW-NEXT:    retq
+;
+; AVX512VLBW-FAST-LABEL: shuffle_v32i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25:
+; AVX512VLBW-FAST:       # %bb.0:
+; AVX512VLBW-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,4,5,6,7,0,1,10,11,12,13,14,15,8,9,18,19,20,21,22,23,16,17,26,27,28,29,30,31,24,25]
+; AVX512VLBW-FAST-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,2,3,0,4,5,6,7,9,10,11,8,12,13,14,15]
+; AVX512VLVBMI-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,6,7,4,8,9,10,11,13,14,15,12]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,4,5,6,7,0,1,10,11,12,13,14,15,8,9,18,19,20,21,22,23,16,17,26,27,28,29,30,31,24,25]
+; AVX512VLVBMI-FAST-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: shuffle_v32i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vprotq $16, %xmm0, %xmm1
+; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[1,2,3,0,4,5,6,7]
+; XOPAVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; XOPAVX1-NEXT:    vprotq $16, %xmm0, %xmm0
+; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
+; XOPAVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; XOPAVX1-NEXT:    retq
 ;

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll b/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll
index c513a57..99742b6 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll
@@ -198,19 +198,39 @@ define <32 x i16> @shuffle_v32i16_1_1_0_0_5_5_4_4_9_9_11_11_13_13_12_12_17_17_19
 }
 
 define <32 x i16> @shuffle_v32i16_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14_17_16_19_18_21_20_23_22_25_24_27_26_29_28_31_30(<32 x i16> %a) {
-; ALL-LABEL: shuffle_v32i16_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14_17_16_19_18_21_20_23_22_25_24_27_26_29_28_31_30:
-; ALL:       ## %bb.0:
-; ALL-NEXT:    vprold $16, %zmm0, %zmm0
-; ALL-NEXT:    retq
+; KNL-LABEL: shuffle_v32i16_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14_17_16_19_18_21_20_23_22_25_24_27_26_29_28_31_30:
+; KNL:       ## %bb.0:
+; KNL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; KNL-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15]
+; KNL-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14]
+; KNL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15]
+; KNL-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14]
+; KNL-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: shuffle_v32i16_01_00_03_02_05_04_07_06_09_08_11_10_13_12_15_14_17_16_19_18_21_20_23_22_25_24_27_26_29_28_31_30:
+; SKX:       ## %bb.0:
+; SKX-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[2,3,0,1,6,7,4,5,10,11,8,9,14,15,12,13,18,19,16,17,22,23,20,21,26,27,24,25,30,31,28,29,34,35,32,33,38,39,36,37,42,43,40,41,46,47,44,45,50,51,48,49,54,55,52,53,58,59,56,57,62,63,60,61]
+; SKX-NEXT:    retq
   %shuffle = shufflevector <32 x i16> %a, <32 x i16> undef, <32 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14, i32 17, i32 16, i32 19, i32 18, i32 21, i32 20, i32 23, i32 22, i32 25, i32 24, i32 27, i32 26, i32 29, i32 28, i32 31, i32 30>
   ret <32 x i16> %shuffle
 }
 
 define <32 x i16> @shuffle_v32i16_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_19_16_17_18_23_20_21_22_27_24_25_26_31_28_29_30(<32 x i16> %a) {
-; ALL-LABEL: shuffle_v32i16_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_19_16_17_18_23_20_21_22_27_24_25_26_31_28_29_30:
-; ALL:       ## %bb.0:
-; ALL-NEXT:    vprolq $48, %zmm0, %zmm0
-; ALL-NEXT:    retq
+; KNL-LABEL: shuffle_v32i16_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_19_16_17_18_23_20_21_22_27_24_25_26_31_28_29_30:
+; KNL:       ## %bb.0:
+; KNL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; KNL-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[3,0,1,2,4,5,6,7,11,8,9,10,12,13,14,15]
+; KNL-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,7,4,5,6,8,9,10,11,15,12,13,14]
+; KNL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,0,1,2,4,5,6,7,11,8,9,10,12,13,14,15]
+; KNL-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,4,5,6,8,9,10,11,15,12,13,14]
+; KNL-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: shuffle_v32i16_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_19_16_17_18_23_20_21_22_27_24_25_26_31_28_29_30:
+; SKX:       ## %bb.0:
+; SKX-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[6,7,0,1,2,3,4,5,14,15,8,9,10,11,12,13,22,23,16,17,18,19,20,21,30,31,24,25,26,27,28,29,38,39,32,33,34,35,36,37,46,47,40,41,42,43,44,45,54,55,48,49,50,51,52,53,62,63,56,57,58,59,60,61]
+; SKX-NEXT:    retq
   %shuffle = shufflevector <32 x i16> %a, <32 x i16> undef, <32 x i32> <i32 3, i32 0, i32 1, i32 2, i32 7, i32 4, i32 5, i32 6, i32 11, i32 8, i32 9, i32 10, i32 15, i32 12, i32 13, i32 14, i32 19, i32 16, i32 17, i32 18, i32 23, i32 20, i32 21, i32 22, i32 27, i32 24, i32 25, i32 26, i32 31, i32 28, i32 29, i32 30>
   ret <32 x i16> %shuffle
 }

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-512-v64.ll b/llvm/test/CodeGen/X86/vector-shuffle-512-v64.ll
index 8cbac72..f00673f 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shuffle-512-v64.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-512-v64.ll
@@ -200,10 +200,37 @@ define <64 x i8> @shuffle_v64i8_63_62_61_60_59_58_57_56_55_54_53_52_51_50_49_48_
 
 ; PR44379
 define <64 x i8> @shuffle_v64i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25_34_35_36_37_38_39_32_33_42_43_44_45_46_47_40_41_50_51_52_53_54_55_48_49_58_59_60_61_62_63_56_57(<64 x i8> %a) {
-; ALL-LABEL: shuffle_v64i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25_34_35_36_37_38_39_32_33_42_43_44_45_46_47_40_41_50_51_52_53_54_55_48_49_58_59_60_61_62_63_56_57:
-; ALL:       # %bb.0:
-; ALL-NEXT:    vprolq $16, %zmm0, %zmm0
-; ALL-NEXT:    retq
+; AVX512F-LABEL: shuffle_v64i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25_34_35_36_37_38_39_32_33_42_43_44_45_46_47_40_41_50_51_52_53_54_55_48_49_58_59_60_61_62_63_56_57:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[1,2,3,0,4,5,6,7,9,10,11,8,12,13,14,15]
+; AVX512F-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,5,6,7,4,8,9,10,11,13,14,15,12]
+; AVX512F-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,2,3,0,4,5,6,7,9,10,11,8,12,13,14,15]
+; AVX512F-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,6,7,4,8,9,10,11,13,14,15,12]
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: shuffle_v64i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25_34_35_36_37_38_39_32_33_42_43_44_45_46_47_40_41_50_51_52_53_54_55_48_49_58_59_60_61_62_63_56_57:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpshuflw {{.*#+}} zmm0 = zmm0[1,2,3,0,4,5,6,7,9,10,11,8,12,13,14,15,17,18,19,16,20,21,22,23,25,26,27,24,28,29,30,31]
+; AVX512BW-NEXT:    vpshufhw {{.*#+}} zmm0 = zmm0[0,1,2,3,5,6,7,4,8,9,10,11,13,14,15,12,16,17,18,19,21,22,23,20,24,25,26,27,29,30,31,28]
+; AVX512BW-NEXT:    retq
+;
+; AVX512DQ-LABEL: shuffle_v64i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25_34_35_36_37_38_39_32_33_42_43_44_45_46_47_40_41_50_51_52_53_54_55_48_49_58_59_60_61_62_63_56_57:
+; AVX512DQ:       # %bb.0:
+; AVX512DQ-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512DQ-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[1,2,3,0,4,5,6,7,9,10,11,8,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,5,6,7,4,8,9,10,11,13,14,15,12]
+; AVX512DQ-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,2,3,0,4,5,6,7,9,10,11,8,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,6,7,4,8,9,10,11,13,14,15,12]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512DQ-NEXT:    retq
+;
+; AVX512VBMI-LABEL: shuffle_v64i8_02_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_18_19_20_21_22_23_16_17_26_27_28_29_30_31_24_25_34_35_36_37_38_39_32_33_42_43_44_45_46_47_40_41_50_51_52_53_54_55_48_49_58_59_60_61_62_63_56_57:
+; AVX512VBMI:       # %bb.0:
+; AVX512VBMI-NEXT:    vpshuflw {{.*#+}} zmm0 = zmm0[1,2,3,0,4,5,6,7,9,10,11,8,12,13,14,15,17,18,19,16,20,21,22,23,25,26,27,24,28,29,30,31]
+; AVX512VBMI-NEXT:    vpshufhw {{.*#+}} zmm0 = zmm0[0,1,2,3,5,6,7,4,8,9,10,11,13,14,15,12,16,17,18,19,21,22,23,20,24,25,26,27,29,30,31,28]
+; AVX512VBMI-NEXT:    retq
   %shuffle = shufflevector <64 x i8> %a, <64 x i8> undef, <64 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 8, i32 9, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 16, i32 17, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 24, i32 25, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 32, i32 33, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 40, i32 41, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 48, i32 49, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 56, i32 57>
   ret <64 x i8> %shuffle
 }