[x86] Teach the x86 vector shuffle lowering to detect mergable 128-bit

lanes.

By special casing these we can often either reduce the total number of
shuffles significantly or reduce the number of (high latency on Haswell)
AVX2 shuffles that potentially cross 128-bit lanes. Even when these
don't actually cross lanes, they have much higher latency to support
that. Doing two of them and a blend is worse than doing a single insert
across the 128-bit lanes to blend and then doing a single interleaved
shuffle.

While this seems like a narrow case, it kept cropping up on me and the
difference is *huge* as you can see in many of the test cases. I first
hit this trying to perfectly fix the interleaving shuffle patterns used
by Halide for AVX2.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@222533 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index c9006000f262f53ce5ff2d55c79644ddf5a69072..36873cf6241fcb5dec64883e6fb8d353b6ef83a9 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -9984,6 +9984,104 @@ static SDValue lowerV2X128VectorShuffle(SDLoc DL, MVT VT, SDValue V1,
                      DAG.getConstant(PermMask, MVT::i8));
 }
 
+/// \brief Lower a vector shuffle by first fixing the 128-bit lanes and then
+/// shuffling each lane.
+///
+/// This will only succeed when the result of fixing the 128-bit lanes results
+/// in a single-input non-lane-crossing shuffle with a repeating shuffle mask in
+/// each 128-bit lanes. This handles many cases where we can quickly blend away
+/// the lane crosses early and then use simpler shuffles within each lane.
+///
+/// FIXME: It might be worthwhile at some point to support this without
+/// requiring the 128-bit lane-relative shuffles to be repeating, but currently
+/// in x86 only floating point has interesting non-repeating shuffles, and even
+/// those are still *marginally* more expensive.
+static SDValue lowerVectorShuffleByMerging128BitLanes(
+    SDLoc DL, MVT VT, SDValue V1, SDValue V2, ArrayRef<int> Mask,
+    const X86Subtarget *Subtarget, SelectionDAG &DAG) {
+  assert(is128BitLaneCrossingShuffleMask(VT, Mask) &&
+         "This is only useful when there are cross-128-bit-lane shuffles.");
+
+  int Size = Mask.size();
+  int LaneSize = 128 / VT.getScalarSizeInBits();
+  int NumLanes = Size / LaneSize;
+  assert(NumLanes > 1 && "Only handles 256-bit and wider shuffles.");
+
+  // See if we can build a hypothetical 128-bit lane-fixing shuffle mask. Also
+  // check whether the in-128-bit lane shuffles share a repeating pattern.
+  SmallVector<int, 4> Lanes;
+  Lanes.resize(NumLanes, -1);
+  SmallVector<int, 4> InLaneMask;
+  InLaneMask.resize(LaneSize, -1);
+  for (int i = 0; i < Size; ++i) {
+    if (Mask[i] < 0)
+      continue;
+
+    int j = i / LaneSize;
+
+    if (Lanes[j] < 0) {
+      // First entry we've seen for this lane.
+      Lanes[j] = Mask[i] / LaneSize;
+    } else if (Lanes[j] != Mask[i] / LaneSize) {
+      // This doesn't match the lane selected previously!
+      return SDValue();
+    }
+
+    // Check that within each lane we have a consistent shuffle mask.
+    int k = i % LaneSize;
+    if (InLaneMask[k] < 0) {
+      InLaneMask[k] = Mask[i] % LaneSize;
+    } else if (InLaneMask[k] != Mask[i] % LaneSize) {
+      // This doesn't fit a repeating in-lane mask.
+      return SDValue();
+    }
+  }
+
+  // First shuffle the lanes into place.
+  MVT LaneVT = MVT::getVectorVT(VT.isFloatingPoint() ? MVT::f64 : MVT::i64,
+                                VT.getSizeInBits() / 64);
+  SmallVector<int, 8> LaneMask;
+  LaneMask.resize(NumLanes * 2, -1);
+  for (int i = 0; i < NumLanes; ++i)
+    if (Lanes[i] >= 0) {
+      LaneMask[2 * i + 0] = 2*Lanes[i] + 0;
+      LaneMask[2 * i + 1] = 2*Lanes[i] + 1;
+    }
+
+  V1 = DAG.getNode(ISD::BITCAST, DL, LaneVT, V1);
+  V2 = DAG.getNode(ISD::BITCAST, DL, LaneVT, V2);
+  SDValue LaneShuffle = DAG.getVectorShuffle(LaneVT, DL, V1, V2, LaneMask);
+
+  // Cast it back to the type we actually want.
+  LaneShuffle = DAG.getNode(ISD::BITCAST, DL, VT, LaneShuffle);
+
+  // Now do a simple shuffle that isn't lane crossing.
+  SmallVector<int, 8> NewMask;
+  NewMask.resize(Size, -1);
+  for (int i = 0; i < Size; ++i)
+    if (Mask[i] >= 0)
+      NewMask[i] = (i / LaneSize) * LaneSize + Mask[i] % LaneSize;
+  assert(!is128BitLaneCrossingShuffleMask(VT, NewMask) &&
+         "Must not introduce lane crosses at this point!");
+
+  return DAG.getVectorShuffle(VT, DL, LaneShuffle, DAG.getUNDEF(VT), NewMask);
+}
+
+/// \brief Test whether the specified input (0 or 1) is in-place blended by the
+/// given mask.
+///
+/// This returns true if the elements from a particular input are already in the
+/// slot required by the given mask and require no permutation.
+static bool isShuffleMaskInputInPlace(int Input, ArrayRef<int> Mask) {
+  assert((Input == 0 || Input == 1) && "Only two inputs to shuffles.");
+  int Size = Mask.size();
+  for (int i = 0; i < Size; ++i)
+    if (Mask[i] >= 0 && Mask[i] / Size == Input && Mask[i] % Size != i)
+      return false;
+
+  return true;
+}
+
 /// \brief Handle lowering of 4-lane 64-bit floating point shuffles.
 ///
 /// Also ends up handling lowering of 4-lane 64-bit integer shuffles when AVX2
@@ -10068,6 +10166,17 @@ static SDValue lowerV4F64VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
                        DAG.getConstant(SHUFPDMask, MVT::i8));
   }
 
+  // Try to simplify this by merging 128-bit lanes to enable a lane-based
+  // shuffle. However, if we have AVX2 and either inputs are already in place,
+  // we will be able to shuffle even across lanes the other input in a single
+  // instruction so skip this pattern.
+  if (is128BitLaneCrossingShuffleMask(MVT::v4f64, Mask) &&
+      !(Subtarget->hasAVX2() && (isShuffleMaskInputInPlace(0, Mask) ||
+                                 isShuffleMaskInputInPlace(1, Mask))))
+    if (SDValue Result = lowerVectorShuffleByMerging128BitLanes(
+            DL, MVT::v4f64, V1, V2, Mask, Subtarget, DAG))
+      return Result;
+
   // If we have AVX2 then we always want to lower with a blend because an v4 we
   // can fully permute the elements.
   if (Subtarget->hasAVX2())
@@ -10138,6 +10247,17 @@ static SDValue lowerV4I64VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
     return DAG.getNode(X86ISD::VPERMI, DL, MVT::v4i64, V1,
                        getV4X86ShuffleImm8ForMask(Mask, DAG));
 
+  // Try to simplify this by merging 128-bit lanes to enable a lane-based
+  // shuffle. However, if we have AVX2 and either inputs are already in place,
+  // we will be able to shuffle even across lanes the other input in a single
+  // instruction so skip this pattern.
+  if (is128BitLaneCrossingShuffleMask(MVT::v4i64, Mask) &&
+      !(Subtarget->hasAVX2() && (isShuffleMaskInputInPlace(0, Mask) ||
+                                 isShuffleMaskInputInPlace(1, Mask))))
+    if (SDValue Result = lowerVectorShuffleByMerging128BitLanes(
+            DL, MVT::v4i64, V1, V2, Mask, Subtarget, DAG))
+      return Result;
+
   // Otherwise fall back on generic blend lowering.
   return lowerVectorShuffleAsDecomposedShuffleBlend(DL, MVT::v4i64, V1, V2,
                                                     Mask, DAG);
@@ -10215,6 +10335,13 @@ static SDValue lowerV8F32VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
                                                    DAG);
   }
 
+  // Try to simplify this by merging 128-bit lanes to enable a lane-based
+  // shuffle.
+  if (is128BitLaneCrossingShuffleMask(MVT::v8f32, Mask))
+    if (SDValue Result = lowerVectorShuffleByMerging128BitLanes(
+            DL, MVT::v8f32, V1, V2, Mask, Subtarget, DAG))
+      return Result;
+
   // If we have AVX2 then we always want to lower with a blend because at v8 we
   // can fully permute the elements.
   if (Subtarget->hasAVX2())
@@ -10278,6 +10405,13 @@ static SDValue lowerV8I32VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
         DAG.getNode(ISD::BUILD_VECTOR, DL, MVT::v8i32, VPermMask), V1);
   }
 
+  // Try to simplify this by merging 128-bit lanes to enable a lane-based
+  // shuffle.
+  if (is128BitLaneCrossingShuffleMask(MVT::v8i32, Mask))
+    if (SDValue Result = lowerVectorShuffleByMerging128BitLanes(
+            DL, MVT::v8i32, V1, V2, Mask, Subtarget, DAG))
+      return Result;
+
   // Otherwise fall back on generic blend lowering.
   return lowerVectorShuffleAsDecomposedShuffleBlend(DL, MVT::v8i32, V1, V2,
                                                     Mask, DAG);
@@ -10305,9 +10439,17 @@ static SDValue lowerV16I16VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
 
   // There are no generalized cross-lane shuffle operations available on i16
   // element types.
-  if (is128BitLaneCrossingShuffleMask(MVT::v16i16, Mask))
+  if (is128BitLaneCrossingShuffleMask(MVT::v16i16, Mask)) {
+    // Try to simplify this by merging 128-bit lanes to enable a lane-based
+    // shuffle.
+    if (!isSingleInputShuffleMask(Mask))
+      if (SDValue Result = lowerVectorShuffleByMerging128BitLanes(
+              DL, MVT::v16i16, V1, V2, Mask, Subtarget, DAG))
+        return Result;
+
     return lowerVectorShuffleAsLanePermuteAndBlend(DL, MVT::v16i16, V1, V2,
                                                    Mask, DAG);
+  }
 
   if (SDValue Blend = lowerVectorShuffleAsBlend(DL, MVT::v16i16, V1, V2, Mask,
                                                 Subtarget, DAG))
@@ -10374,9 +10516,17 @@ static SDValue lowerV32I8VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
 
   // There are no generalized cross-lane shuffle operations available on i8
   // element types.
-  if (is128BitLaneCrossingShuffleMask(MVT::v32i8, Mask))
-    return lowerVectorShuffleAsLanePermuteAndBlend(DL, MVT::v32i8, V1, V2,
-                                                   Mask, DAG);
+  if (is128BitLaneCrossingShuffleMask(MVT::v32i8, Mask)) {
+    // Try to simplify this by merging 128-bit lanes to enable a lane-based
+    // shuffle.
+    if (!isSingleInputShuffleMask(Mask))
+      if (SDValue Result = lowerVectorShuffleByMerging128BitLanes(
+              DL, MVT::v32i8, V1, V2, Mask, Subtarget, DAG))
+        return Result;
+
+    return lowerVectorShuffleAsLanePermuteAndBlend(DL, MVT::v32i8, V1, V2, Mask,
+                                                   DAG);
+  }
 
   if (SDValue Blend = lowerVectorShuffleAsBlend(DL, MVT::v32i8, V1, V2, Mask,
                                                 Subtarget, DAG))

diff --git a/test/CodeGen/X86/avx-vperm2x128.ll b/test/CodeGen/X86/avx-vperm2x128.ll
index a103405a85deac84347a98eacc13f8b4061dd10e..43303ca57c4fc70afb5b06bb8786ca2adfc8abc7 100644 (file)
--- a/test/CodeGen/X86/avx-vperm2x128.ll
+++ b/test/CodeGen/X86/avx-vperm2x128.ll
@@ -182,20 +182,11 @@ entry:
 ;;;; Cases we must not select vperm2f128
 
 define <8 x float> @G(<8 x float> %a, <8 x float> %b) nounwind uwtable readnone ssp {
-; AVX1-LABEL: G:
-; AVX1:       ## BB#0: ## %entry
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,2,3]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: G:
-; AVX2:       ## BB#0: ## %entry
-; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,2,3,4,4,6,7]
-; AVX2-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
-; AVX2-NEXT:    retq
+; ALL-LABEL: G:
+; ALL:       ## BB#0: ## %entry
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
+; ALL-NEXT:    retq
 entry:
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 undef, i32 undef, i32 6, i32 7, i32 undef, i32 12, i32 undef, i32 15>
   ret <8 x float> %shuffle

diff --git a/test/CodeGen/X86/vector-shuffle-256-v16.ll b/test/CodeGen/X86/vector-shuffle-256-v16.ll
index b3fba4f349bad35e9fa9d31a0d85a7677b606f7a..2a6efda0dfae40f0e2aa35534f7b2f4b08742d62 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -1278,11 +1278,8 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_04_04_04_04_16_16_16_16_20_20_20_2
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_16_16_16_16_20_20_20_20:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; AVX2-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; AVX2-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; AVX2-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
 ; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4, i32 16, i32 16, i32 16, i32 16, i32 20, i32 20, i32 20, i32 20>
   ret <16 x i16> %shuffle
@@ -1301,13 +1298,8 @@ define <16 x i16> @shuffle_v16i16_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_2
 ;
 ; AVX2-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_20:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; AVX2-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
-; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; AVX2-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 8, i32 8, i32 8, i32 8, i32 12, i32 12, i32 12, i32 12, i32 16, i32 16, i32 16, i32 16, i32 20, i32 20, i32 20, i32 20>
   ret <16 x i16> %shuffle
@@ -1327,13 +1319,8 @@ define <16 x i16> @shuffle_v16i16_08_08_08_08_12_12_12_12_24_24_24_24_28_28_28_2
 ;
 ; AVX2-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_24_24_24_24_28_28_28_28:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; AVX2-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
-; AVX2-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; AVX2-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
-; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 8, i32 8, i32 8, i32 8, i32 12, i32 12, i32 12, i32 12, i32 24, i32 24, i32 24, i32 24, i32 28, i32 28, i32 28, i32 28>
   ret <16 x i16> %shuffle

diff --git a/test/CodeGen/X86/vector-shuffle-256-v32.ll b/test/CodeGen/X86/vector-shuffle-256-v32.ll
index 1c09ba0861dfda32ab2e3008324a01e4cdca2d08..d8390fdfb351cdbc4ca5665fa281d412e2bc8bb5 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-256-v32.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v32.ll
@@ -1572,10 +1572,8 @@ 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_32_32_32_32_32_32_32_32_40_40_40_40_40_40_40_40:
 ; AVX2:       # BB#0:
-; 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> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 40, i32 40, i32 40, i32 40, i32 40, i32 40, i32 40, i32 40>
   ret <32 x i8> %shuffle
@@ -1593,12 +1591,8 @@ define <32 x i8> @shuffle_v32i8_16_16_16_16_16_16_16_16_24_24_24_24_24_24_24_24_
 ;
 ; AVX2-LABEL: shuffle_v32i8_16_16_16_16_16_16_16_16_24_24_24_24_24_24_24_24_32_32_32_32_32_32_32_32_40_40_40_40_40_40_40_40:
 ; AVX2:       # BB#0:
-; 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:    vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
+; 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> <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 40, i32 40, i32 40, i32 40, i32 40, i32 40, i32 40, i32 40>
   ret <32 x i8> %shuffle
@@ -1617,12 +1611,8 @@ define <32 x i8> @shuffle_v32i8_16_16_16_16_16_16_16_16_24_24_24_24_24_24_24_24_
 ;
 ; AVX2-LABEL: shuffle_v32i8_16_16_16_16_16_16_16_16_24_24_24_24_24_24_24_24_48_48_48_48_48_48_48_48_56_56_56_56_56_56_56_56:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %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:    vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; 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> <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 56, i32 56, i32 56, i32 56, i32 56, i32 56, i32 56, i32 56>
   ret <32 x i8> %shuffle

diff --git a/test/CodeGen/X86/vector-shuffle-256-v4.ll b/test/CodeGen/X86/vector-shuffle-256-v4.ll
index 18b9970f0701a0fa91771879d343697293a85102..19f3c7743ca9120fa65dc5a4869390bc5aed23de 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-256-v4.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v4.ll
@@ -301,58 +301,31 @@ define <4 x double> @shuffle_v4f64_0167(<4 x double> %a, <4 x double> %b) {
 }
 
 define <4 x double> @shuffle_v4f64_1054(<4 x double> %a, <4 x double> %b) {
-; AVX1-LABEL: shuffle_v4f64_1054:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: shuffle_v4f64_1054:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,0]
-; AVX2-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
-; AVX2-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX2-NEXT:    retq
+; ALL-LABEL: shuffle_v4f64_1054:
+; ALL:       # BB#0:
+; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; ALL-NEXT:    retq
   %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 5, i32 4>
   ret <4 x double> %shuffle
 }
 
 define <4 x double> @shuffle_v4f64_3254(<4 x double> %a, <4 x double> %b) {
-; AVX1-LABEL: shuffle_v4f64_3254:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: shuffle_v4f64_3254:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,0]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3]
-; AVX2-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX2-NEXT:    retq
+; ALL-LABEL: shuffle_v4f64_3254:
+; ALL:       # BB#0:
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
+; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; ALL-NEXT:    retq
   %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 3, i32 2, i32 5, i32 4>
   ret <4 x double> %shuffle
 }
 
 define <4 x double> @shuffle_v4f64_3276(<4 x double> %a, <4 x double> %b) {
-; AVX1-LABEL: shuffle_v4f64_3276:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: shuffle_v4f64_3276:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3]
-; AVX2-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX2-NEXT:    retq
+; ALL-LABEL: shuffle_v4f64_3276:
+; ALL:       # BB#0:
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; ALL-NEXT:    retq
   %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 3, i32 2, i32 7, i32 6>
   ret <4 x double> %shuffle
 }
@@ -687,16 +660,14 @@ define <4 x i64> @shuffle_v4i64_1251(<4 x i64> %a, <4 x i64> %b) {
 define <4 x i64> @shuffle_v4i64_1054(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: shuffle_v4i64_1054:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v4i64_1054:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,1,0]
+; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 0, i32 5, i32 4>
   ret <4 x i64> %shuffle
@@ -705,17 +676,14 @@ define <4 x i64> @shuffle_v4i64_1054(<4 x i64> %a, <4 x i64> %b) {
 define <4 x i64> @shuffle_v4i64_3254(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: shuffle_v4i64_3254:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
+; AVX1-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v4i64_3254:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,1,0]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 3, i32 2, i32 5, i32 4>
   ret <4 x i64> %shuffle
@@ -724,18 +692,14 @@ define <4 x i64> @shuffle_v4i64_3254(<4 x i64> %a, <4 x i64> %b) {
 define <4 x i64> @shuffle_v4i64_3276(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: shuffle_v4i64_3276:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX1-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v4i64_3276:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 3, i32 2, i32 7, i32 6>
   ret <4 x i64> %shuffle

diff --git a/test/CodeGen/X86/vector-shuffle-256-v8.ll b/test/CodeGen/X86/vector-shuffle-256-v8.ll
index ded8232dad2d091db419e68877341cb18eccd413..d04adf9f7e872dfe45ac4330fa8665e024df29e0 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-256-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v8.ll
@@ -736,20 +736,11 @@ define <8 x float> @shuffle_v8f32_76543210(<8 x float> %a, <8 x float> %b) {
 }
 
 define <8 x float> @shuffle_v8f32_3210ba98(<8 x float> %a, <8 x float> %b) {
-; AVX1-LABEL: shuffle_v8f32_3210ba98:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: shuffle_v8f32_3210ba98:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vmovaps {{.*#+}} ymm2 = <u,u,u,u,3,2,1,0>
-; AVX2-NEXT:    vpermps %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; AVX2-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX2-NEXT:    retq
+; ALL-LABEL: shuffle_v8f32_3210ba98:
+; ALL:       # BB#0:
+; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 11, i32 10, i32 9, i32 8>
   ret <8 x float> %shuffle
 }
@@ -774,43 +765,21 @@ define <8 x float> @shuffle_v8f32_3210fedc(<8 x float> %a, <8 x float> %b) {
 }
 
 define <8 x float> @shuffle_v8f32_7654fedc(<8 x float> %a, <8 x float> %b) {
-; AVX1-LABEL: shuffle_v8f32_7654fedc:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: shuffle_v8f32_7654fedc:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vmovaps {{.*#+}} ymm2 = <7,6,5,4,u,u,u,u>
-; AVX2-NEXT:    vpermps %ymm0, %ymm2, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,2,1,0,7,6,5,4]
-; AVX2-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX2-NEXT:    retq
+; ALL-LABEL: shuffle_v8f32_7654fedc:
+; ALL:       # BB#0:
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 15, i32 14, i32 13, i32 12>
   ret <8 x float> %shuffle
 }
 
 define <8 x float> @shuffle_v8f32_fedc7654(<8 x float> %a, <8 x float> %b) {
-; AVX1-LABEL: shuffle_v8f32_fedc7654:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,1,0]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: shuffle_v8f32_fedc7654:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vmovaps {{.*#+}} ymm2 = <7,6,5,4,u,u,u,u>
-; AVX2-NEXT:    vpermps %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; AVX2-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; AVX2-NEXT:    retq
+; ALL-LABEL: shuffle_v8f32_fedc7654:
+; ALL:       # BB#0:
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 7, i32 6, i32 5, i32 4>
   ret <8 x float> %shuffle
 }
@@ -1789,17 +1758,14 @@ define <8 x i32> @shuffle_v8i32_76543210(<8 x i32> %a, <8 x i32> %b) {
 define <8 x i32> @shuffle_v8i32_3210ba98(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_3210ba98:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v8i32_3210ba98:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <u,u,u,u,3,2,1,0>
-; AVX2-NEXT:    vpermd %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 11, i32 10, i32 9, i32 8>
   ret <8 x i32> %shuffle
@@ -1827,19 +1793,14 @@ define <8 x i32> @shuffle_v8i32_3210fedc(<8 x i32> %a, <8 x i32> %b) {
 define <8 x i32> @shuffle_v8i32_7654fedc(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_7654fedc:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v8i32_7654fedc:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <7,6,5,4,u,u,u,u>
-; AVX2-NEXT:    vpermd %ymm0, %ymm2, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[3,2,1,0,7,6,5,4]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 15, i32 14, i32 13, i32 12>
   ret <8 x i32> %shuffle
@@ -1848,19 +1809,14 @@ define <8 x i32> @shuffle_v8i32_7654fedc(<8 x i32> %a, <8 x i32> %b) {
 define <8 x i32> @shuffle_v8i32_fedc7654(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_fedc7654:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,1,0]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v8i32_fedc7654:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <7,6,5,4,u,u,u,u>
-; AVX2-NEXT:    vpermd %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 7, i32 6, i32 5, i32 4>
   ret <8 x i32> %shuffle

diff --git a/test/CodeGen/X86/vector-shuffle-512-v8.ll b/test/CodeGen/X86/vector-shuffle-512-v8.ll
index 31cabbcaf5fe4cf4c7923a848cf3165515b852e3..6671735248af913d2c279fd16fdc300123cdf818 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-512-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-512-v8.ll
@@ -208,15 +208,13 @@ define <8 x double> @shuffle_v8f64_9832dc76(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_9810dc54(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_9810dc54:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2 
-; ALL-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,1,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3 
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm3[1,0,2,2]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,0]
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; ALL-NEXT:    vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm2
+; ALL-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[1,0,3,2]
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm1
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
+; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm2, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 9, i32 8, i32 1, i32 0, i32 13, i32 12, i32 5, i32 4>
   ret <8 x double> %shuffle
@@ -930,15 +928,13 @@ define <8 x i64> @shuffle_v8i64_9832dc76(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_9810dc54(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_9810dc54:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2 
-; ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,1,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3 
-; ALL-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,1,0]
-; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; ALL-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm2
+; ALL-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; ALL-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
+; ALL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm2, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 9, i32 8, i32 1, i32 0, i32 13, i32 12, i32 5, i32 4>
   ret <8 x i64> %shuffle

diff --git a/test/CodeGen/X86/vector-shuffle-combining.ll b/test/CodeGen/X86/vector-shuffle-combining.ll
index 22a67492110faed92a93173ebfa645f21ef01d0a..5f6e5c60022259befca0b8884d08bc5597b9cd55 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-combining.ll
+++ b/test/CodeGen/X86/vector-shuffle-combining.ll
@@ -2463,19 +2463,16 @@ define <8 x i32> @combine_unneeded_subvector2(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpaddd {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,1,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: combine_unneeded_subvector2:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpaddd {{.*}}(%rip), %ymm0, %ymm0
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <7,6,5,4,u,u,u,u>
-; AVX2-NEXT:    vpermd %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-NEXT:    retq
   %c = add <8 x i32> %a, <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
   %d = shufflevector <8 x i32> %b, <8 x i32> %c, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 15, i32 14, i32 13, i32 12>