From 9dd83f5ee8697fdb41ba73bc70d845085715b01d Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Mon, 1 Mar 2021 10:41:57 +0000
Subject: [PATCH] [DAG] visitVECTOR_SHUFFLE - attempt to match commuted
 shuffles with MergeInnerShuffle.

Try to match "shuffle(C, shuffle(A, B, M0), M1) -> shuffle(A, B, M2)" etc. by using MergeInnerShuffle's commuted inner shuffle mode.
---
 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp |  35 +-
 llvm/test/CodeGen/Thumb2/mve-shuffle.ll       |  52 +-
 llvm/test/CodeGen/Thumb2/mve-vld3.ll          | 452 +++++++++---------
 llvm/test/CodeGen/X86/oddshuffles.ll          |  32 +-
 llvm/test/CodeGen/X86/pr34592.ll              |  39 +-
 5 files changed, 307 insertions(+), 303 deletions(-)

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index d896b8c0cdef..06b07f32fb39 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -21196,23 +21196,28 @@ SDValue DAGCombiner::visitVECTOR_SHUFFLE(SDNode *N) {
     //   shuffle(shuffle(A, B, M0), C, M1) -> shuffle(B, C, M2)
     // Don't try to fold shuffles with illegal type.
     // Only fold if this shuffle is the only user of the other shuffle.
-    if (N0.getOpcode() == ISD::VECTOR_SHUFFLE &&
-        N->isOnlyUserOf(N0.getNode())) {
-      // The incoming shuffle must be of the same type as the result of the
-      // current shuffle.
-      auto *OtherSV = cast<ShuffleVectorSDNode>(N0);
-      assert(OtherSV->getOperand(0).getValueType() == VT &&
-             "Shuffle types don't match");
+    // Try matching shuffle(C,shuffle(A,B)) commutted patterns as well.
+    for (int i = 0; i != 2; ++i) {
+      if (N->getOperand(i).getOpcode() == ISD::VECTOR_SHUFFLE &&
+          N->isOnlyUserOf(N->getOperand(i).getNode())) {
+        // The incoming shuffle must be of the same type as the result of the
+        // current shuffle.
+        auto *OtherSV = cast<ShuffleVectorSDNode>(N->getOperand(i));
+        assert(OtherSV->getOperand(0).getValueType() == VT &&
+               "Shuffle types don't match");
 
-      SDValue SV0, SV1;
-      SmallVector<int, 4> Mask;
-      if (MergeInnerShuffle(false, SVN, OtherSV, N1, TLI, SV0, SV1, Mask)) {
-        // Check if all indices in Mask are Undef. In case, propagate Undef.
-        if (llvm::all_of(Mask, [](int M) { return M < 0; }))
-          return DAG.getUNDEF(VT);
+        SDValue SV0, SV1;
+        SmallVector<int, 4> Mask;
+        if (MergeInnerShuffle(i != 0, SVN, OtherSV, N->getOperand(1 - i), TLI,
+                              SV0, SV1, Mask)) {
+          // Check if all indices in Mask are Undef. In case, propagate Undef.
+          if (llvm::all_of(Mask, [](int M) { return M < 0; }))
+            return DAG.getUNDEF(VT);
 
-        return DAG.getVectorShuffle(VT, SDLoc(N), SV0 ? SV0 : DAG.getUNDEF(VT),
-                                    SV1 ? SV1 : DAG.getUNDEF(VT), Mask);
+          return DAG.getVectorShuffle(VT, SDLoc(N),
+                                      SV0 ? SV0 : DAG.getUNDEF(VT),
+                                      SV1 ? SV1 : DAG.getUNDEF(VT), Mask);
+        }
       }
     }
 
diff --git a/llvm/test/CodeGen/Thumb2/mve-shuffle.ll b/llvm/test/CodeGen/Thumb2/mve-shuffle.ll
index 128cfd0c2e2b..db8f7018ba55 100644
--- a/llvm/test/CodeGen/Thumb2/mve-shuffle.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-shuffle.ll
@@ -135,19 +135,19 @@ define arm_aapcs_vfpcc <4 x i32> @shuffle3step_i32(<16 x i32> %src) {
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9}
 ; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    vmov.f32 s14, s8
-; CHECK-NEXT:    vmov.f32 s15, s11
-; CHECK-NEXT:    vmov.f32 s16, s1
-; CHECK-NEXT:    vmov.f32 s12, s2
-; CHECK-NEXT:    vmov.f32 s17, s4
-; CHECK-NEXT:    vmov.f32 s1, s3
-; CHECK-NEXT:    vmov.f32 s18, s7
-; CHECK-NEXT:    vmov.f32 s2, s6
-; CHECK-NEXT:    vmov.f32 s19, s10
-; CHECK-NEXT:    vmov.f32 s3, s9
-; CHECK-NEXT:    vmov.f32 s13, s5
-; CHECK-NEXT:    vadd.i32 q0, q0, q4
-; CHECK-NEXT:    vadd.i32 q0, q0, q3
+; CHECK-NEXT:    vmov.f32 s12, s1
+; CHECK-NEXT:    vmov.f32 s16, s0
+; CHECK-NEXT:    vmov.f32 s13, s4
+; CHECK-NEXT:    vmov.f32 s17, s3
+; CHECK-NEXT:    vmov.f32 s14, s7
+; CHECK-NEXT:    vmov.f32 s18, s6
+; CHECK-NEXT:    vmov.f32 s4, s2
+; CHECK-NEXT:    vmov.f32 s6, s8
+; CHECK-NEXT:    vmov.f32 s15, s10
+; CHECK-NEXT:    vmov.f32 s19, s9
+; CHECK-NEXT:    vadd.i32 q3, q4, q3
+; CHECK-NEXT:    vmov.f32 s7, s11
+; CHECK-NEXT:    vadd.i32 q0, q3, q1
 ; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
@@ -1164,19 +1164,19 @@ define arm_aapcs_vfpcc <4 x float> @shuffle3step_f32(<16 x float> %src) {
 ; CHECKFP:       @ %bb.0: @ %entry
 ; CHECKFP-NEXT:    .vsave {d8, d9}
 ; CHECKFP-NEXT:    vpush {d8, d9}
-; CHECKFP-NEXT:    vmov.f32 s14, s8
-; CHECKFP-NEXT:    vmov.f32 s15, s11
-; CHECKFP-NEXT:    vmov.f32 s16, s1
-; CHECKFP-NEXT:    vmov.f32 s12, s2
-; CHECKFP-NEXT:    vmov.f32 s17, s4
-; CHECKFP-NEXT:    vmov.f32 s1, s3
-; CHECKFP-NEXT:    vmov.f32 s18, s7
-; CHECKFP-NEXT:    vmov.f32 s2, s6
-; CHECKFP-NEXT:    vmov.f32 s19, s10
-; CHECKFP-NEXT:    vmov.f32 s3, s9
-; CHECKFP-NEXT:    vmov.f32 s13, s5
-; CHECKFP-NEXT:    vadd.f32 q0, q0, q4
-; CHECKFP-NEXT:    vadd.f32 q0, q0, q3
+; CHECKFP-NEXT:    vmov.f32 s12, s1
+; CHECKFP-NEXT:    vmov.f32 s16, s0
+; CHECKFP-NEXT:    vmov.f32 s13, s4
+; CHECKFP-NEXT:    vmov.f32 s17, s3
+; CHECKFP-NEXT:    vmov.f32 s14, s7
+; CHECKFP-NEXT:    vmov.f32 s18, s6
+; CHECKFP-NEXT:    vmov.f32 s4, s2
+; CHECKFP-NEXT:    vmov.f32 s6, s8
+; CHECKFP-NEXT:    vmov.f32 s15, s10
+; CHECKFP-NEXT:    vmov.f32 s19, s9
+; CHECKFP-NEXT:    vadd.f32 q3, q4, q3
+; CHECKFP-NEXT:    vmov.f32 s7, s11
+; CHECKFP-NEXT:    vadd.f32 q0, q3, q1
 ; CHECKFP-NEXT:    vpop {d8, d9}
 ; CHECKFP-NEXT:    bx lr
 entry:
diff --git a/llvm/test/CodeGen/Thumb2/mve-vld3.ll b/llvm/test/CodeGen/Thumb2/mve-vld3.ll
index 4cab1a4668af..06dbfe8debbb 100644
--- a/llvm/test/CodeGen/Thumb2/mve-vld3.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-vld3.ll
@@ -36,22 +36,22 @@ define void @vld3_v4i32(<12 x i32> *%src, <4 x i32> *%dst) {
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9}
 ; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d3, d0
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vmov.f32 s16, s9
-; CHECK-NEXT:    vmov.f32 s4, s10
-; CHECK-NEXT:    vmov.f32 s17, s12
-; CHECK-NEXT:    vmov.f32 s9, s11
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f32 s10, s14
-; CHECK-NEXT:    vmov.f32 s19, s2
-; CHECK-NEXT:    vmov.f32 s11, s1
-; CHECK-NEXT:    vmov.f32 s5, s13
-; CHECK-NEXT:    vadd.i32 q0, q2, q4
-; CHECK-NEXT:    vadd.i32 q0, q0, q1
+; CHECK-NEXT:    vldrw.u32 q1, [r0]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    vmov.f64 d6, d2
+; CHECK-NEXT:    vmov.f32 s16, s5
+; CHECK-NEXT:    vmov.f32 s13, s7
+; CHECK-NEXT:    vmov.f32 s17, s0
+; CHECK-NEXT:    vmov.f32 s14, s2
+; CHECK-NEXT:    vmov.f32 s18, s3
+; CHECK-NEXT:    vmov.f32 s0, s6
+; CHECK-NEXT:    vmov.f32 s2, s8
+; CHECK-NEXT:    vmov.f32 s19, s10
+; CHECK-NEXT:    vmov.f32 s15, s9
+; CHECK-NEXT:    vadd.i32 q3, q3, q4
+; CHECK-NEXT:    vmov.f32 s3, s11
+; CHECK-NEXT:    vadd.i32 q0, q3, q0
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
@@ -71,39 +71,39 @@ define void @vld3_v8i32(<24 x i32> *%src, <8 x i32> *%dst) {
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #64]
-; CHECK-NEXT:    vmov.f64 d3, d0
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vmov.f32 s16, s9
-; CHECK-NEXT:    vmov.f32 s4, s10
-; CHECK-NEXT:    vmov.f32 s17, s12
-; CHECK-NEXT:    vmov.f32 s9, s11
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f32 s10, s14
-; CHECK-NEXT:    vmov.f32 s19, s2
-; CHECK-NEXT:    vmov.f32 s11, s1
-; CHECK-NEXT:    vmov.f32 s5, s13
-; CHECK-NEXT:    vadd.i32 q0, q2, q4
-; CHECK-NEXT:    vadd.i32 q0, q0, q1
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
+; CHECK-NEXT:    vmov.f64 d6, d2
+; CHECK-NEXT:    vmov.f32 s16, s5
+; CHECK-NEXT:    vmov.f32 s13, s7
+; CHECK-NEXT:    vmov.f32 s17, s0
+; CHECK-NEXT:    vmov.f32 s14, s2
+; CHECK-NEXT:    vmov.f32 s18, s3
+; CHECK-NEXT:    vmov.f32 s0, s6
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vmov.f32 s2, s8
+; CHECK-NEXT:    vmov.f32 s19, s10
+; CHECK-NEXT:    vmov.f32 s15, s9
+; CHECK-NEXT:    vmov.f32 s3, s11
+; CHECK-NEXT:    vadd.i32 q3, q3, q4
+; CHECK-NEXT:    vadd.i32 q0, q3, q0
 ; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d5, d2
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
 ; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s11, s7
-; CHECK-NEXT:    vmov.f32 s20, s13
-; CHECK-NEXT:    vmov.f32 s8, s14
-; CHECK-NEXT:    vmov.f32 s21, s16
-; CHECK-NEXT:    vmov.f32 s13, s15
-; CHECK-NEXT:    vmov.f32 s22, s19
-; CHECK-NEXT:    vmov.f32 s14, s18
-; CHECK-NEXT:    vmov.f32 s23, s6
-; CHECK-NEXT:    vmov.f32 s15, s5
-; CHECK-NEXT:    vmov.f32 s9, s17
-; CHECK-NEXT:    vadd.i32 q1, q3, q5
-; CHECK-NEXT:    vadd.i32 q1, q1, q2
+; CHECK-NEXT:    vmov.f32 s16, s13
+; CHECK-NEXT:    vmov.f64 d10, d6
+; CHECK-NEXT:    vmov.f32 s17, s4
+; CHECK-NEXT:    vmov.f32 s21, s15
+; CHECK-NEXT:    vmov.f32 s18, s7
+; CHECK-NEXT:    vmov.f32 s22, s6
+; CHECK-NEXT:    vmov.f32 s4, s14
+; CHECK-NEXT:    vmov.f32 s6, s8
+; CHECK-NEXT:    vmov.f32 s19, s10
+; CHECK-NEXT:    vmov.f32 s23, s9
+; CHECK-NEXT:    vadd.i32 q4, q5, q4
+; CHECK-NEXT:    vmov.f32 s7, s11
+; CHECK-NEXT:    vadd.i32 q1, q4, q1
 ; CHECK-NEXT:    vstrw.32 q1, [r1]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11}
 ; CHECK-NEXT:    bx lr
@@ -123,73 +123,73 @@ define void @vld3_v16i32(<48 x i32> *%src, <16 x i32> *%dst) {
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #64]
-; CHECK-NEXT:    vmov.f64 d3, d0
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vmov.f32 s16, s9
-; CHECK-NEXT:    vmov.f32 s4, s10
-; CHECK-NEXT:    vmov.f32 s17, s12
-; CHECK-NEXT:    vmov.f32 s9, s11
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f32 s10, s14
-; CHECK-NEXT:    vmov.f32 s19, s2
-; CHECK-NEXT:    vmov.f32 s11, s1
-; CHECK-NEXT:    vmov.f32 s5, s13
-; CHECK-NEXT:    vadd.i32 q0, q2, q4
-; CHECK-NEXT:    vadd.i32 q0, q0, q1
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d5, d2
-; CHECK-NEXT:    vmov.f32 s11, s7
-; CHECK-NEXT:    vmov.f32 s20, s13
-; CHECK-NEXT:    vmov.f32 s8, s14
-; CHECK-NEXT:    vmov.f32 s21, s16
-; CHECK-NEXT:    vmov.f32 s13, s15
-; CHECK-NEXT:    vmov.f32 s22, s19
-; CHECK-NEXT:    vmov.f32 s14, s18
-; CHECK-NEXT:    vmov.f32 s23, s6
-; CHECK-NEXT:    vmov.f32 s15, s5
-; CHECK-NEXT:    vmov.f32 s9, s17
-; CHECK-NEXT:    vadd.i32 q1, q3, q5
-; CHECK-NEXT:    vadd.i32 q1, q1, q2
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #176]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #144]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #160]
-; CHECK-NEXT:    vmov.f64 d7, d4
-; CHECK-NEXT:    vmov.f32 s15, s11
-; CHECK-NEXT:    vmov.f32 s24, s17
-; CHECK-NEXT:    vmov.f32 s12, s18
-; CHECK-NEXT:    vmov.f32 s25, s20
-; CHECK-NEXT:    vmov.f32 s17, s19
-; CHECK-NEXT:    vmov.f32 s26, s23
-; CHECK-NEXT:    vmov.f32 s18, s22
-; CHECK-NEXT:    vmov.f32 s27, s10
-; CHECK-NEXT:    vmov.f32 s19, s9
-; CHECK-NEXT:    vmov.f32 s13, s21
-; CHECK-NEXT:    vadd.i32 q2, q4, q6
-; CHECK-NEXT:    vadd.i32 q2, q2, q3
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #128]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #96]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #112]
-; CHECK-NEXT:    vmov.f64 d9, d6
-; CHECK-NEXT:    vstrw.32 q2, [r1, #48]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vstrw.32 q1, [r1]
-; CHECK-NEXT:    vmov.f32 s19, s15
-; CHECK-NEXT:    vmov.f32 s28, s21
-; CHECK-NEXT:    vmov.f32 s16, s22
-; CHECK-NEXT:    vmov.f32 s29, s24
-; CHECK-NEXT:    vmov.f32 s21, s23
-; CHECK-NEXT:    vmov.f32 s30, s27
-; CHECK-NEXT:    vmov.f32 s22, s26
-; CHECK-NEXT:    vmov.f32 s31, s14
-; CHECK-NEXT:    vmov.f32 s23, s13
-; CHECK-NEXT:    vmov.f32 s17, s25
-; CHECK-NEXT:    vadd.i32 q3, q5, q7
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
+; CHECK-NEXT:    vmov.f64 d6, d2
+; CHECK-NEXT:    vmov.f32 s16, s5
+; CHECK-NEXT:    vmov.f32 s13, s7
+; CHECK-NEXT:    vmov.f32 s17, s0
+; CHECK-NEXT:    vmov.f32 s14, s2
+; CHECK-NEXT:    vmov.f32 s18, s3
+; CHECK-NEXT:    vmov.f32 s0, s6
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vmov.f32 s2, s8
+; CHECK-NEXT:    vmov.f32 s19, s10
+; CHECK-NEXT:    vmov.f32 s15, s9
+; CHECK-NEXT:    vmov.f32 s3, s11
 ; CHECK-NEXT:    vadd.i32 q3, q3, q4
+; CHECK-NEXT:    vadd.i32 q0, q3, q0
+; CHECK-NEXT:    vldrw.u32 q3, [r0]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    vmov.f32 s16, s13
+; CHECK-NEXT:    vmov.f64 d10, d6
+; CHECK-NEXT:    vmov.f32 s17, s4
+; CHECK-NEXT:    vmov.f32 s21, s15
+; CHECK-NEXT:    vmov.f32 s18, s7
+; CHECK-NEXT:    vmov.f32 s22, s6
+; CHECK-NEXT:    vmov.f32 s4, s14
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #144]
+; CHECK-NEXT:    vmov.f32 s6, s8
+; CHECK-NEXT:    vmov.f32 s19, s10
+; CHECK-NEXT:    vmov.f32 s23, s9
+; CHECK-NEXT:    vmov.f32 s7, s11
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #160]
+; CHECK-NEXT:    vadd.i32 q4, q5, q4
+; CHECK-NEXT:    vmov.f64 d10, d6
+; CHECK-NEXT:    vadd.i32 q1, q4, q1
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #176]
+; CHECK-NEXT:    vmov.f32 s24, s13
+; CHECK-NEXT:    vmov.f32 s21, s15
+; CHECK-NEXT:    vmov.f32 s25, s8
+; CHECK-NEXT:    vmov.f32 s22, s10
+; CHECK-NEXT:    vmov.f32 s26, s11
+; CHECK-NEXT:    vmov.f32 s8, s14
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #112]
+; CHECK-NEXT:    vmov.f32 s10, s16
+; CHECK-NEXT:    vmov.f32 s27, s18
+; CHECK-NEXT:    vmov.f32 s23, s17
+; CHECK-NEXT:    vmov.f32 s11, s19
+; CHECK-NEXT:    vadd.i32 q5, q5, q6
+; CHECK-NEXT:    vadd.i32 q2, q5, q2
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #96]
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #128]
+; CHECK-NEXT:    vstrw.32 q2, [r1, #48]
+; CHECK-NEXT:    vmov.f32 s24, s21
+; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
+; CHECK-NEXT:    vmov.f64 d14, d10
+; CHECK-NEXT:    vstrw.32 q1, [r1]
+; CHECK-NEXT:    vmov.f32 s25, s12
+; CHECK-NEXT:    vmov.f32 s29, s23
+; CHECK-NEXT:    vmov.f32 s26, s15
+; CHECK-NEXT:    vmov.f32 s30, s14
+; CHECK-NEXT:    vmov.f32 s12, s22
+; CHECK-NEXT:    vmov.f32 s14, s16
+; CHECK-NEXT:    vmov.f32 s27, s18
+; CHECK-NEXT:    vmov.f32 s31, s17
+; CHECK-NEXT:    vadd.i32 q6, q7, q6
+; CHECK-NEXT:    vmov.f32 s15, s19
+; CHECK-NEXT:    vadd.i32 q3, q6, q3
 ; CHECK-NEXT:    vstrw.32 q3, [r1, #32]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
@@ -929,22 +929,22 @@ define void @vld3_v4f32(<12 x float> *%src, <4 x float> *%dst) {
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9}
 ; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d3, d0
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vmov.f32 s16, s9
-; CHECK-NEXT:    vmov.f32 s4, s10
-; CHECK-NEXT:    vmov.f32 s17, s12
-; CHECK-NEXT:    vmov.f32 s9, s11
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f32 s10, s14
-; CHECK-NEXT:    vmov.f32 s19, s2
-; CHECK-NEXT:    vmov.f32 s11, s1
-; CHECK-NEXT:    vmov.f32 s5, s13
-; CHECK-NEXT:    vadd.f32 q0, q2, q4
-; CHECK-NEXT:    vadd.f32 q0, q0, q1
+; CHECK-NEXT:    vldrw.u32 q1, [r0]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    vmov.f64 d6, d2
+; CHECK-NEXT:    vmov.f32 s16, s5
+; CHECK-NEXT:    vmov.f32 s13, s7
+; CHECK-NEXT:    vmov.f32 s17, s0
+; CHECK-NEXT:    vmov.f32 s14, s2
+; CHECK-NEXT:    vmov.f32 s18, s3
+; CHECK-NEXT:    vmov.f32 s0, s6
+; CHECK-NEXT:    vmov.f32 s2, s8
+; CHECK-NEXT:    vmov.f32 s19, s10
+; CHECK-NEXT:    vmov.f32 s15, s9
+; CHECK-NEXT:    vadd.f32 q3, q3, q4
+; CHECK-NEXT:    vmov.f32 s3, s11
+; CHECK-NEXT:    vadd.f32 q0, q3, q0
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
@@ -964,39 +964,39 @@ define void @vld3_v8f32(<24 x float> *%src, <8 x float> *%dst) {
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #64]
-; CHECK-NEXT:    vmov.f64 d3, d0
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vmov.f32 s16, s9
-; CHECK-NEXT:    vmov.f32 s4, s10
-; CHECK-NEXT:    vmov.f32 s17, s12
-; CHECK-NEXT:    vmov.f32 s9, s11
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f32 s10, s14
-; CHECK-NEXT:    vmov.f32 s19, s2
-; CHECK-NEXT:    vmov.f32 s11, s1
-; CHECK-NEXT:    vmov.f32 s5, s13
-; CHECK-NEXT:    vadd.f32 q0, q2, q4
-; CHECK-NEXT:    vadd.f32 q0, q0, q1
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
+; CHECK-NEXT:    vmov.f64 d6, d2
+; CHECK-NEXT:    vmov.f32 s16, s5
+; CHECK-NEXT:    vmov.f32 s13, s7
+; CHECK-NEXT:    vmov.f32 s17, s0
+; CHECK-NEXT:    vmov.f32 s14, s2
+; CHECK-NEXT:    vmov.f32 s18, s3
+; CHECK-NEXT:    vmov.f32 s0, s6
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vmov.f32 s2, s8
+; CHECK-NEXT:    vmov.f32 s19, s10
+; CHECK-NEXT:    vmov.f32 s15, s9
+; CHECK-NEXT:    vmov.f32 s3, s11
+; CHECK-NEXT:    vadd.f32 q3, q3, q4
+; CHECK-NEXT:    vadd.f32 q0, q3, q0
 ; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d5, d2
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
 ; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s11, s7
-; CHECK-NEXT:    vmov.f32 s20, s13
-; CHECK-NEXT:    vmov.f32 s8, s14
-; CHECK-NEXT:    vmov.f32 s21, s16
-; CHECK-NEXT:    vmov.f32 s13, s15
-; CHECK-NEXT:    vmov.f32 s22, s19
-; CHECK-NEXT:    vmov.f32 s14, s18
-; CHECK-NEXT:    vmov.f32 s23, s6
-; CHECK-NEXT:    vmov.f32 s15, s5
-; CHECK-NEXT:    vmov.f32 s9, s17
-; CHECK-NEXT:    vadd.f32 q1, q3, q5
-; CHECK-NEXT:    vadd.f32 q1, q1, q2
+; CHECK-NEXT:    vmov.f32 s16, s13
+; CHECK-NEXT:    vmov.f64 d10, d6
+; CHECK-NEXT:    vmov.f32 s17, s4
+; CHECK-NEXT:    vmov.f32 s21, s15
+; CHECK-NEXT:    vmov.f32 s18, s7
+; CHECK-NEXT:    vmov.f32 s22, s6
+; CHECK-NEXT:    vmov.f32 s4, s14
+; CHECK-NEXT:    vmov.f32 s6, s8
+; CHECK-NEXT:    vmov.f32 s19, s10
+; CHECK-NEXT:    vmov.f32 s23, s9
+; CHECK-NEXT:    vadd.f32 q4, q5, q4
+; CHECK-NEXT:    vmov.f32 s7, s11
+; CHECK-NEXT:    vadd.f32 q1, q4, q1
 ; CHECK-NEXT:    vstrw.32 q1, [r1]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11}
 ; CHECK-NEXT:    bx lr
@@ -1016,73 +1016,73 @@ define void @vld3_v16f32(<48 x float> *%src, <16 x float> *%dst) {
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #64]
-; CHECK-NEXT:    vmov.f64 d3, d0
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vmov.f32 s16, s9
-; CHECK-NEXT:    vmov.f32 s4, s10
-; CHECK-NEXT:    vmov.f32 s17, s12
-; CHECK-NEXT:    vmov.f32 s9, s11
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f32 s10, s14
-; CHECK-NEXT:    vmov.f32 s19, s2
-; CHECK-NEXT:    vmov.f32 s11, s1
-; CHECK-NEXT:    vmov.f32 s5, s13
-; CHECK-NEXT:    vadd.f32 q0, q2, q4
-; CHECK-NEXT:    vadd.f32 q0, q0, q1
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d5, d2
-; CHECK-NEXT:    vmov.f32 s11, s7
-; CHECK-NEXT:    vmov.f32 s20, s13
-; CHECK-NEXT:    vmov.f32 s8, s14
-; CHECK-NEXT:    vmov.f32 s21, s16
-; CHECK-NEXT:    vmov.f32 s13, s15
-; CHECK-NEXT:    vmov.f32 s22, s19
-; CHECK-NEXT:    vmov.f32 s14, s18
-; CHECK-NEXT:    vmov.f32 s23, s6
-; CHECK-NEXT:    vmov.f32 s15, s5
-; CHECK-NEXT:    vmov.f32 s9, s17
-; CHECK-NEXT:    vadd.f32 q1, q3, q5
-; CHECK-NEXT:    vadd.f32 q1, q1, q2
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #176]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #144]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #160]
-; CHECK-NEXT:    vmov.f64 d7, d4
-; CHECK-NEXT:    vmov.f32 s15, s11
-; CHECK-NEXT:    vmov.f32 s24, s17
-; CHECK-NEXT:    vmov.f32 s12, s18
-; CHECK-NEXT:    vmov.f32 s25, s20
-; CHECK-NEXT:    vmov.f32 s17, s19
-; CHECK-NEXT:    vmov.f32 s26, s23
-; CHECK-NEXT:    vmov.f32 s18, s22
-; CHECK-NEXT:    vmov.f32 s27, s10
-; CHECK-NEXT:    vmov.f32 s19, s9
-; CHECK-NEXT:    vmov.f32 s13, s21
-; CHECK-NEXT:    vadd.f32 q2, q4, q6
-; CHECK-NEXT:    vadd.f32 q2, q2, q3
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #128]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #96]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #112]
-; CHECK-NEXT:    vmov.f64 d9, d6
-; CHECK-NEXT:    vstrw.32 q2, [r1, #48]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vstrw.32 q1, [r1]
-; CHECK-NEXT:    vmov.f32 s19, s15
-; CHECK-NEXT:    vmov.f32 s28, s21
-; CHECK-NEXT:    vmov.f32 s16, s22
-; CHECK-NEXT:    vmov.f32 s29, s24
-; CHECK-NEXT:    vmov.f32 s21, s23
-; CHECK-NEXT:    vmov.f32 s30, s27
-; CHECK-NEXT:    vmov.f32 s22, s26
-; CHECK-NEXT:    vmov.f32 s31, s14
-; CHECK-NEXT:    vmov.f32 s23, s13
-; CHECK-NEXT:    vmov.f32 s17, s25
-; CHECK-NEXT:    vadd.f32 q3, q5, q7
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
+; CHECK-NEXT:    vmov.f64 d6, d2
+; CHECK-NEXT:    vmov.f32 s16, s5
+; CHECK-NEXT:    vmov.f32 s13, s7
+; CHECK-NEXT:    vmov.f32 s17, s0
+; CHECK-NEXT:    vmov.f32 s14, s2
+; CHECK-NEXT:    vmov.f32 s18, s3
+; CHECK-NEXT:    vmov.f32 s0, s6
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vmov.f32 s2, s8
+; CHECK-NEXT:    vmov.f32 s19, s10
+; CHECK-NEXT:    vmov.f32 s15, s9
+; CHECK-NEXT:    vmov.f32 s3, s11
 ; CHECK-NEXT:    vadd.f32 q3, q3, q4
+; CHECK-NEXT:    vadd.f32 q0, q3, q0
+; CHECK-NEXT:    vldrw.u32 q3, [r0]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    vmov.f32 s16, s13
+; CHECK-NEXT:    vmov.f64 d10, d6
+; CHECK-NEXT:    vmov.f32 s17, s4
+; CHECK-NEXT:    vmov.f32 s21, s15
+; CHECK-NEXT:    vmov.f32 s18, s7
+; CHECK-NEXT:    vmov.f32 s22, s6
+; CHECK-NEXT:    vmov.f32 s4, s14
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #144]
+; CHECK-NEXT:    vmov.f32 s6, s8
+; CHECK-NEXT:    vmov.f32 s19, s10
+; CHECK-NEXT:    vmov.f32 s23, s9
+; CHECK-NEXT:    vmov.f32 s7, s11
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #160]
+; CHECK-NEXT:    vadd.f32 q4, q5, q4
+; CHECK-NEXT:    vmov.f64 d10, d6
+; CHECK-NEXT:    vadd.f32 q1, q4, q1
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #176]
+; CHECK-NEXT:    vmov.f32 s24, s13
+; CHECK-NEXT:    vmov.f32 s21, s15
+; CHECK-NEXT:    vmov.f32 s25, s8
+; CHECK-NEXT:    vmov.f32 s22, s10
+; CHECK-NEXT:    vmov.f32 s26, s11
+; CHECK-NEXT:    vmov.f32 s8, s14
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #112]
+; CHECK-NEXT:    vmov.f32 s10, s16
+; CHECK-NEXT:    vmov.f32 s27, s18
+; CHECK-NEXT:    vmov.f32 s23, s17
+; CHECK-NEXT:    vmov.f32 s11, s19
+; CHECK-NEXT:    vadd.f32 q5, q5, q6
+; CHECK-NEXT:    vadd.f32 q2, q5, q2
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #96]
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #128]
+; CHECK-NEXT:    vstrw.32 q2, [r1, #48]
+; CHECK-NEXT:    vmov.f32 s24, s21
+; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
+; CHECK-NEXT:    vmov.f64 d14, d10
+; CHECK-NEXT:    vstrw.32 q1, [r1]
+; CHECK-NEXT:    vmov.f32 s25, s12
+; CHECK-NEXT:    vmov.f32 s29, s23
+; CHECK-NEXT:    vmov.f32 s26, s15
+; CHECK-NEXT:    vmov.f32 s30, s14
+; CHECK-NEXT:    vmov.f32 s12, s22
+; CHECK-NEXT:    vmov.f32 s14, s16
+; CHECK-NEXT:    vmov.f32 s27, s18
+; CHECK-NEXT:    vmov.f32 s31, s17
+; CHECK-NEXT:    vadd.f32 q6, q7, q6
+; CHECK-NEXT:    vmov.f32 s15, s19
+; CHECK-NEXT:    vadd.f32 q3, q6, q3
 ; CHECK-NEXT:    vstrw.32 q3, [r1, #32]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
diff --git a/llvm/test/CodeGen/X86/oddshuffles.ll b/llvm/test/CodeGen/X86/oddshuffles.ll
index 6bbfe5c699c4..dad3251db0e7 100644
--- a/llvm/test/CodeGen/X86/oddshuffles.ll
+++ b/llvm/test/CodeGen/X86/oddshuffles.ll
@@ -1247,38 +1247,36 @@ define void @interleave_24i32_out(<24 x i32>* %p, <8 x i32>* %q1, <8 x i32>* %q2
 ; SSE42:       # %bb.0:
 ; SSE42-NEXT:    movups 80(%rdi), %xmm8
 ; SSE42-NEXT:    movdqu 64(%rdi), %xmm9
-; SSE42-NEXT:    movdqu (%rdi), %xmm4
+; SSE42-NEXT:    movdqu (%rdi), %xmm3
 ; SSE42-NEXT:    movdqu 16(%rdi), %xmm2
 ; SSE42-NEXT:    movups 32(%rdi), %xmm10
 ; SSE42-NEXT:    movdqu 48(%rdi), %xmm5
 ; SSE42-NEXT:    movdqa %xmm2, %xmm6
-; SSE42-NEXT:    pblendw {{.*#+}} xmm6 = xmm6[0,1],xmm4[2,3],xmm6[4,5,6,7]
-; SSE42-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[2,3,2,3]
-; SSE42-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,3],xmm2[2,3]
-; SSE42-NEXT:    insertps {{.*#+}} xmm4 = xmm4[0,1,2],xmm10[1]
+; SSE42-NEXT:    pblendw {{.*#+}} xmm6 = xmm6[0,1],xmm3[2,3],xmm6[4,5,6,7]
+; SSE42-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[2,3,2,3]
+; SSE42-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm2[2,3]
+; SSE42-NEXT:    insertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm10[1]
 ; SSE42-NEXT:    movdqa %xmm9, %xmm1
 ; SSE42-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3],xmm1[4,5,6,7]
 ; SSE42-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
 ; SSE42-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm9[2,3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm5 = xmm5[0,1,2],xmm8[1]
-; SSE42-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[2,2,2,2]
+; SSE42-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[2,2,2,2]
 ; SSE42-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,0,3,3]
-; SSE42-NEXT:    pblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5],xmm3[6,7]
+; SSE42-NEXT:    pblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5],xmm4[6,7]
 ; SSE42-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,0,3,3]
-; SSE42-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[2,2,2,2]
-; SSE42-NEXT:    pblendw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,5],xmm3[6,7]
+; SSE42-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[2,2,2,2]
+; SSE42-NEXT:    pblendw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,5],xmm4[6,7]
 ; SSE42-NEXT:    pblendw {{.*#+}} xmm7 = xmm7[0,1],xmm2[2,3],xmm7[4,5,6,7]
-; SSE42-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,1,0,3]
-; SSE42-NEXT:    pblendw {{.*#+}} xmm1 = xmm7[0,1,2,3],xmm1[4,5,6,7]
+; SSE42-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm10[0,3]
 ; SSE42-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm9[2,3],xmm0[4,5,6,7]
-; SSE42-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,1,0,3]
-; SSE42-NEXT:    pblendw {{.*#+}} xmm2 = xmm0[0,1,2,3],xmm2[4,5,6,7]
+; SSE42-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm8[0,3]
 ; SSE42-NEXT:    movups %xmm5, 16(%rsi)
-; SSE42-NEXT:    movups %xmm4, (%rsi)
-; SSE42-NEXT:    movdqu %xmm3, 16(%rdx)
+; SSE42-NEXT:    movups %xmm3, (%rsi)
+; SSE42-NEXT:    movdqu %xmm4, 16(%rdx)
 ; SSE42-NEXT:    movdqu %xmm6, (%rdx)
-; SSE42-NEXT:    movdqu %xmm2, 16(%rcx)
-; SSE42-NEXT:    movdqu %xmm1, (%rcx)
+; SSE42-NEXT:    movups %xmm0, 16(%rcx)
+; SSE42-NEXT:    movups %xmm7, (%rcx)
 ; SSE42-NEXT:    retq
 ;
 ; AVX1-LABEL: interleave_24i32_out:
diff --git a/llvm/test/CodeGen/X86/pr34592.ll b/llvm/test/CodeGen/X86/pr34592.ll
index 3c5345bf3411..224a3dd5d207 100644
--- a/llvm/test/CodeGen/X86/pr34592.ll
+++ b/llvm/test/CodeGen/X86/pr34592.ll
@@ -14,38 +14,39 @@ define <16 x i64> @pluto(<16 x i64> %arg, <16 x i64> %arg1, <16 x i64> %arg2, <1
 ; CHECK-NEXT:    vmovaps %ymm4, %ymm10
 ; CHECK-NEXT:    vmovaps %ymm3, %ymm9
 ; CHECK-NEXT:    vmovaps %ymm1, %ymm8
-; CHECK-NEXT:    vmovaps %ymm0, %ymm4
+; CHECK-NEXT:    vmovaps %ymm0, %ymm3
 ; CHECK-NEXT:    vmovaps 240(%rbp), %ymm1
-; CHECK-NEXT:    vmovaps 208(%rbp), %ymm3
+; CHECK-NEXT:    vmovaps 208(%rbp), %ymm4
 ; CHECK-NEXT:    vmovaps 176(%rbp), %ymm0
 ; CHECK-NEXT:    vmovaps 144(%rbp), %ymm0
 ; CHECK-NEXT:    vmovaps 112(%rbp), %ymm11
 ; CHECK-NEXT:    vmovaps 80(%rbp), %ymm11
 ; CHECK-NEXT:    vmovaps 48(%rbp), %ymm11
 ; CHECK-NEXT:    vmovaps 16(%rbp), %ymm11
-; CHECK-NEXT:    vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm2[6,7]
-; CHECK-NEXT:    vmovaps %xmm3, %xmm8
+; CHECK-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5],ymm2[6,7]
+; CHECK-NEXT:    vmovaps %xmm4, %xmm6
 ; CHECK-NEXT:    # implicit-def: $ymm2
-; CHECK-NEXT:    vinserti128 $1, %xmm8, %ymm2, %ymm2
-; CHECK-NEXT:    vpalignr {{.*#+}} ymm0 = ymm4[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; CHECK-NEXT:    vinserti128 $1, %xmm6, %ymm2, %ymm2
+; CHECK-NEXT:    vpalignr {{.*#+}} ymm0 = ymm3[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm3[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
 ; CHECK-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,0]
 ; CHECK-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; CHECK-NEXT:    vmovaps %xmm7, %xmm2
-; CHECK-NEXT:    vpslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6,7]
+; CHECK-NEXT:    vextracti128 $1, %ymm7, %xmm2
+; CHECK-NEXT:    vmovq {{.*#+}} xmm6 = xmm2[0],zero
 ; CHECK-NEXT:    # implicit-def: $ymm2
-; CHECK-NEXT:    vmovaps %xmm4, %xmm2
-; CHECK-NEXT:    vpalignr {{.*#+}} ymm3 = ymm3[8,9,10,11,12,13,14,15],ymm5[0,1,2,3,4,5,6,7],ymm3[24,25,26,27,28,29,30,31],ymm5[16,17,18,19,20,21,22,23]
-; CHECK-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; CHECK-NEXT:    vpblendd {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; CHECK-NEXT:    vmovaps %xmm6, %xmm2
+; CHECK-NEXT:    # kill: def $xmm3 killed $xmm3 killed $ymm3
+; CHECK-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; CHECK-NEXT:    vmovaps %xmm7, %xmm3
+; CHECK-NEXT:    vpslldq {{.*#+}} xmm6 = zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0,1,2,3,4,5,6,7]
+; CHECK-NEXT:    # implicit-def: $ymm3
+; CHECK-NEXT:    vmovaps %xmm6, %xmm3
+; CHECK-NEXT:    vpalignr {{.*#+}} ymm4 = ymm4[8,9,10,11,12,13,14,15],ymm5[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm5[16,17,18,19,20,21,22,23]
+; CHECK-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; CHECK-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
 ; CHECK-NEXT:    vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3],ymm7[4,5,6,7]
 ; CHECK-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,1,1,3]
-; CHECK-NEXT:    vpshufd {{.*#+}} ymm2 = ymm5[0,1,0,1,4,5,4,5]
-; CHECK-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; CHECK-NEXT:    vextracti128 $1, %ymm7, %xmm2
-; CHECK-NEXT:    vmovq {{.*#+}} xmm4 = xmm2[0],zero
-; CHECK-NEXT:    # implicit-def: $ymm2
-; CHECK-NEXT:    vmovaps %xmm4, %xmm2
-; CHECK-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm2[0,1],ymm6[0,1]
+; CHECK-NEXT:    vpshufd {{.*#+}} ymm4 = ymm5[0,1,0,1,4,5,4,5]
+; CHECK-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
 ; CHECK-NEXT:    movq %rbp, %rsp
 ; CHECK-NEXT:    popq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa %rsp, 8