[llvm] AMDGPU: Improve getShuffleCost accuracy for 8- and 16-bit shuffles (PR #168818)

[Nicolai Hähnle via llvm-commits]

https://github.com/nhaehnle updated https://github.com/llvm/llvm-project/pull/168818

>From d5944fb80d34c0e3d10a986764c47e4602680eda Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Nicolai=20H=C3=A4hnle?= <nicolai.haehnle at amd.com>
Date: Wed, 19 Nov 2025 18:34:08 -0800
Subject: [PATCH] AMDGPU: Improve getShuffleCost accuracy for 8- and 16-bit
 shuffles

These shuffles can always be implemented using v_perm_b32, and so this
rewrites the analysis from the perspective of "how many v_perm_b32s does
it take to assemble each register of the result?"

The test changes in Transforms/SLPVectorizer/reduction.ll are
reasonable: VI (gfx8) has native f16 math, but not packed math.

commit-id:8b76e888
---
 .../AMDGPU/AMDGPUTargetTransformInfo.cpp      | 135 ++-
 .../CostModel/AMDGPU/shufflevector.ll         | 986 +++++++++---------
 .../SLPVectorizer/AMDGPU/reduction.ll         | 127 ++-
 .../SLPVectorizer/AMDGPU/slp-v2f16.ll         |  51 +-
 .../extract-insert-chain-to-shuffles.ll       |  32 +-
 .../VectorCombine/AMDGPU/extract-insert-i8.ll |  97 +-
 6 files changed, 762 insertions(+), 666 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
index 03d16fdd54c42..dfa21515838ff 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -1241,46 +1241,123 @@ InstructionCost GCNTTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
       (ScalarSize == 16 || ScalarSize == 8)) {
     // Larger vector widths may require additional instructions, but are
     // typically cheaper than scalarized versions.
-    unsigned NumVectorElts = cast<FixedVectorType>(SrcTy)->getNumElements();
-    unsigned RequestedElts =
-        count_if(Mask, [](int MaskElt) { return MaskElt != -1; });
+    //
+    // We assume that shuffling at a register granularity can be done for free.
+    // This is not true for vectors fed into memory instructions, but it is
+    // effectively true for all other shuffling. The emphasis of the logic here
+    // is to assist generic transform in cleaning up / canonicalizing those
+    // shuffles.
+
+    // With op_sel VOP3P instructions freely can access the low half or high
+    // half of a register, so any swizzle of two elements is free.
+    if (auto *SrcVecTy = dyn_cast<FixedVectorType>(SrcTy)) {
+      unsigned NumSrcElts = SrcVecTy->getNumElements();
+      if (ST->hasVOP3PInsts() && ScalarSize == 16 && NumSrcElts == 2 &&
+          (Kind == TTI::SK_Broadcast || Kind == TTI::SK_Reverse ||
+           Kind == TTI::SK_PermuteSingleSrc))
+        return 0;
+    }
+
     unsigned EltsPerReg = 32 / ScalarSize;
-    if (RequestedElts == 0)
-      return 0;
     switch (Kind) {
     case TTI::SK_Broadcast:
+      // A single v_perm_b32 can be re-used for all destination registers.
+      return 1;
     case TTI::SK_Reverse:
-    case TTI::SK_PermuteSingleSrc: {
-      // With op_sel VOP3P instructions freely can access the low half or high
-      // half of a register, so any swizzle of two elements is free.
-      if (ST->hasVOP3PInsts() && ScalarSize == 16 && NumVectorElts == 2)
-        return 0;
-      unsigned NumPerms = alignTo(RequestedElts, EltsPerReg) / EltsPerReg;
-      // SK_Broadcast just reuses the same mask
-      unsigned NumPermMasks = Kind == TTI::SK_Broadcast ? 1 : NumPerms;
-      return NumPerms + NumPermMasks;
-    }
+      // One instruction per register.
+      if (auto *DstVecTy = dyn_cast<FixedVectorType>(DstTy))
+        return divideCeil(DstVecTy->getNumElements(), EltsPerReg);
+      return InstructionCost::getInvalid();
     case TTI::SK_ExtractSubvector:
+      if (Index % EltsPerReg == 0)
+        return 0; // Shuffling at register granularity
+      if (auto *DstVecTy = dyn_cast<FixedVectorType>(DstTy))
+        return divideCeil(DstVecTy->getNumElements(), EltsPerReg);
+      return InstructionCost::getInvalid();
     case TTI::SK_InsertSubvector: {
-      // Even aligned accesses are free
-      if (!(Index % 2))
-        return 0;
-      // Insert/extract subvectors only require shifts / extract code to get the
-      // relevant bits
-      return alignTo(RequestedElts, EltsPerReg) / EltsPerReg;
+      auto *DstVecTy = dyn_cast<FixedVectorType>(DstTy);
+      if (!DstVecTy)
+        return InstructionCost::getInvalid();
+      unsigned NumDstElts = DstVecTy->getNumElements();
+      unsigned NumInsertElts = cast<FixedVectorType>(SubTp)->getNumElements();
+      unsigned EndIndex = Index + NumInsertElts;
+      unsigned BeginSubIdx = Index % EltsPerReg;
+      unsigned EndSubIdx = EndIndex % EltsPerReg;
+      unsigned Cost = 0;
+
+      if (BeginSubIdx != 0) {
+        // Need to shift the inserted vector into place. The cost is the number
+        // of destination registers overlapped by the inserted vector.
+        Cost = divideCeil(EndIndex, EltsPerReg) - (Index / EltsPerReg);
+      }
+
+      // If the last register overlap is partial, there may be three source
+      // registers feeding into it; that takes an extra instruction.
+      if (EndIndex < NumDstElts && BeginSubIdx < EndSubIdx)
+        Cost += 1;
+
+      return Cost;
     }
-    case TTI::SK_PermuteTwoSrc:
-    case TTI::SK_Splice:
-    case TTI::SK_Select: {
-      unsigned NumPerms = alignTo(RequestedElts, EltsPerReg) / EltsPerReg;
-      // SK_Select just reuses the same mask
-      unsigned NumPermMasks = Kind == TTI::SK_Select ? 1 : NumPerms;
-      return NumPerms + NumPermMasks;
+    case TTI::SK_Splice: {
+      auto *DstVecTy = dyn_cast<FixedVectorType>(DstTy);
+      if (!DstVecTy)
+        return InstructionCost::getInvalid();
+      unsigned NumElts = DstVecTy->getNumElements();
+      assert(NumElts == cast<FixedVectorType>(SrcTy)->getNumElements());
+      // Determine the sub-region of the result vector that requires
+      // sub-register shuffles / mixing.
+      unsigned EltsFromLHS = NumElts - Index;
+      bool LHSIsAligned = (Index % EltsPerReg) == 0;
+      bool RHSIsAligned = (EltsFromLHS % EltsPerReg) == 0;
+      if (LHSIsAligned && RHSIsAligned)
+        return 0;
+      if (LHSIsAligned && !RHSIsAligned)
+        return divideCeil(NumElts, EltsPerReg) - (EltsFromLHS / EltsPerReg);
+      if (!LHSIsAligned && RHSIsAligned)
+        return divideCeil(EltsFromLHS, EltsPerReg);
+      return divideCeil(NumElts, EltsPerReg);
     }
-
     default:
       break;
     }
+
+    if (!Mask.empty()) {
+      unsigned NumSrcElts = cast<FixedVectorType>(SrcTy)->getNumElements();
+
+      // Generically estimate the cost by assuming that each destination
+      // register is derived from sources via v_perm_b32 instructions if it
+      // can't be copied as-is.
+      //
+      // For each destination register, derive the cost of obtaining it based
+      // on the number of source registers that feed into it.
+      unsigned Cost = 0;
+      for (unsigned DstIdx = 0; DstIdx < Mask.size(); DstIdx += EltsPerReg) {
+        SmallVector<int, 4> Regs;
+        bool Aligned = true;
+        for (unsigned I = 0; I < EltsPerReg && DstIdx + I < Mask.size(); ++I) {
+          int SrcIdx = Mask[DstIdx + I];
+          if (SrcIdx == -1)
+            continue;
+          int Reg;
+          if (SrcIdx < (int)NumSrcElts) {
+            Reg = SrcIdx / EltsPerReg;
+            if (SrcIdx % EltsPerReg != I)
+              Aligned = false;
+          } else {
+            Reg = NumSrcElts + (SrcIdx - NumSrcElts) / EltsPerReg;
+            if ((SrcIdx - NumSrcElts) % EltsPerReg != I)
+              Aligned = false;
+          }
+          if (!llvm::is_contained(Regs, Reg))
+            Regs.push_back(Reg);
+        }
+        if (Regs.size() >= 2)
+          Cost += Regs.size() - 1;
+        else if (!Aligned)
+          Cost += 1;
+      }
+      return Cost;
+    }
   }
 
   return BaseT::getShuffleCost(Kind, DstTy, SrcTy, Mask, CostKind, Index,
diff --git a/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll b/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll
index b66e19e8bc563..78d43e8949269 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll
@@ -13,153 +13,153 @@ define amdgpu_kernel void @shufflevector_i16(<2 x i16> %vec1, <2 x i16> %vec2) {
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 1>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 3>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 1>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 0>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 3>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; VI-LABEL: 'shufflevector_i16'
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> zeroinitializer
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> zeroinitializer
 ; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
 ; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX9-10-SIZE-LABEL: 'shufflevector_i16'
@@ -167,153 +167,153 @@ define amdgpu_kernel void @shufflevector_i16(<2 x i16> %vec1, <2 x i16> %vec2) {
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 1>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 3>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 1>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 0>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 3>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; VI-SIZE-LABEL: 'shufflevector_i16'
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> zeroinitializer
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> zeroinitializer
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %shuf00 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> zeroinitializer
@@ -396,157 +396,157 @@ define amdgpu_kernel void @shufflevector_i16(<2 x i16> %vec1, <2 x i16> %vec2) {
 ; Should not assert
 define amdgpu_kernel void @shufflevector_i8(<2 x i8> %vec1, <2 x i8> %vec2) {
 ; ALL-LABEL: 'shufflevector_i8'
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> zeroinitializer
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf02 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf10 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf11 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 2>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf13 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf33 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 1>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf000 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> zeroinitializer
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf010 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf011 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf100 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf110 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf111 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf002 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf020 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf022 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf200 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf202 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf220 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf222 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf112 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf122 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf212 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf32 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf001 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf010 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf101 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf121 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf211 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf221 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> zeroinitializer
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf02_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf10_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf11_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 2>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf13_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf33_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 1>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf000_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> zeroinitializer
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf010_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf011_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf100_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf110_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf111_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf002_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf020_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf022_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf200_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf202_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf220_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf222_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf112_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf122_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf212_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf32_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf001_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf010_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf101_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf121_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf211_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf221_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; ALL-SIZE-LABEL: 'shufflevector_i8'
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> zeroinitializer
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf02 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf10 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf11 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 2>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf13 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf33 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 1>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf000 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> zeroinitializer
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf010 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf011 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf100 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf110 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf111 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf002 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf020 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf022 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf200 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf202 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf220 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf222 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf112 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf122 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf212 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf32 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf001 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf010 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf101 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf121 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf211 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf221 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> zeroinitializer
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf02_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf10_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf11_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf02_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 2>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf13_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf03_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf30_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf33_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf12_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf21_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf13_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf31_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 1>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf000_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> zeroinitializer
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf001_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf010_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf011_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf100_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf101_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf110_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf111_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf002_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf020_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf022_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf200_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf202_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf220_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf222_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf112_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf121_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf122_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf211_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf212_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf221_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf32_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf000_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf001_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf010_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf011_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf100_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf101_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf110_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf111_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf002_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf020_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf022_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf200_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf202_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf220_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf222_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf112_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf121_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf122_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf211_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf212_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf221_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %shuf00 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> zeroinitializer
@@ -861,28 +861,30 @@ define amdgpu_kernel void @shufflevector_i32(<2 x i32> %vec1, <2 x i32> %vec2) {
 ; Other shuffle cases
 define void @shuffle(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8> %i8v4_2, <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i8> %i8v8, <8 x i8> %i8v8_2, <16 x i8> %i8v16, <16 x i8> %i8v16_2, <2 x i16> %i16v2, <2 x i16> %i16v2_2, <4 x i16> %i16v4, <4 x i16> %i16v4_2, <8 x i16> %i16v8, <8 x i16> %i16v8_2, <2 x i32> %i32v2, <2 x i32> %i32v2_2, <4 x i32> %i32v4, <4 x i32> %i32v4_2, <2 x float> %floatv2, <2 x float> %floatv2_2, <4 x float> %floatv4, <4 x float> %floatv4_2,<2 x i64> %i64v2, <2 x i64> %i64v2_2,<2 x double> %doublev2, <2 x double> %doublev2_2) {
 ; GFX9-10-LABEL: 'shuffle'
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i8_8_2_a = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8_2_b = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 0, i32 4, i32 poison, i32 12, i32 poison, i32 5, i32 9, i32 13>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
@@ -898,28 +900,30 @@ define void @shuffle(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; VI-LABEL: 'shuffle'
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i8_8_2_a = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8_2_b = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 0, i32 4, i32 poison, i32 12, i32 poison, i32 5, i32 9, i32 13>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <2 x i32> <i32 1, i32 0>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <2 x i32> <i32 1, i32 0>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
@@ -935,28 +939,30 @@ define void @shuffle(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX9-10-SIZE-LABEL: 'shuffle'
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i8_8_2_a = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8_2_b = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 0, i32 4, i32 poison, i32 12, i32 poison, i32 5, i32 9, i32 13>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
@@ -972,28 +978,30 @@ define void @shuffle(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; VI-SIZE-LABEL: 'shuffle'
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i8_8_2_a = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i8_8_2_b = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 0, i32 4, i32 poison, i32 12, i32 poison, i32 5, i32 9, i32 13>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <2 x i32> <i32 1, i32 0>
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <2 x i32> <i32 1, i32 0>
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
@@ -1022,6 +1030,8 @@ define void @shuffle(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8>
   %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
   %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
   %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v8i8_8_2_a = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13>
+  %v8i8_8_2_b = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 0, i32 4, i32 poison, i32 12, i32 poison, i32 5, i32 9, i32 13>
   %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
   %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
   %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
@@ -1047,7 +1057,7 @@ define void @shuffle(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8>
 
 define void @concat(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i8> %i8v8, <8 x i8> %i8v8_2, <2 x half> %halfv2, <2 x half> %halfv2_2, <4 x half> %halfv4, <4 x half> %halfv4_2, <8 x half> %halfv8, <8 x half> %halfv8_2, <2 x i16> %i16v2, <2 x i16> %i16v2_2, <4 x i16> %i16v4, <4 x i16> %i16v4_2, <8 x i16> %i16v8, <8 x i16> %i16v8_2, <2 x i32> %i32v2, <2 x i32> %i32v2_2, <4 x i32> %i32v4, <4 x i32> %i32v4_2, <2 x float> %floatv2, <2 x float> %floatv2_2, <4 x float> %floatv4, <4 x float> %floatv4_2,<2 x i64> %i64v2, <2 x i64> %i64v2_2,<2 x double> %doublev2, <2 x double> %doublev2_2) {
 ; ALL-LABEL: 'concat'
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i16 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
@@ -1062,7 +1072,7 @@ define void @concat(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8> %
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
@@ -1080,7 +1090,7 @@ define void @concat(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8> %
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; ALL-SIZE-LABEL: 'concat'
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i16 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
@@ -1095,7 +1105,7 @@ define void @concat(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8> %
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
@@ -1147,25 +1157,25 @@ define void @concat(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8> %
 
 define void @insert(<16 x i8> %i8v16, <16 x i8> %i8v16_2) {
 ; ALL-LABEL: 'insert'
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %test0 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %test1 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %test2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %test0 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %test1 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %test2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test3 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test4 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %test4 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test5 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test6 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test7 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 1, i32 16, i32 17, i32 18, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %test6 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %test7 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 1, i32 16, i32 17, i32 18, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; ALL-SIZE-LABEL: 'insert'
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %test0 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %test1 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %test2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %test0 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %test1 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %test2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test3 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test4 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %test4 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test5 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test6 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test7 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 1, i32 16, i32 17, i32 18, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %test6 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %test7 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 1, i32 16, i32 17, i32 18, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %test0 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 0, i32 16, i32 17, i32 18, i32 19, i32 20, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
@@ -1181,13 +1191,13 @@ define void @insert(<16 x i8> %i8v16, <16 x i8> %i8v16_2) {
 
 define void @splice(<16 x i8> %i8v16, <16 x i8> %i8v16_2) {
 ; ALL-LABEL: 'splice'
-; ALL-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %test0 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %test1 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test0 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %test1 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; ALL-SIZE-LABEL: 'splice'
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %test0 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %test1 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %test0 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %test1 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %test0 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27>
diff --git a/llvm/test/Transforms/SLPVectorizer/AMDGPU/reduction.ll b/llvm/test/Transforms/SLPVectorizer/AMDGPU/reduction.ll
index b2246e4f9c6c4..126d195314772 100644
--- a/llvm/test/Transforms/SLPVectorizer/AMDGPU/reduction.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AMDGPU/reduction.ll
@@ -3,10 +3,21 @@
 ; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -passes=slp-vectorizer,dce < %s | FileCheck -check-prefixes=GCN,VI %s
 
 define half @reduction_half4(<4 x half> %a) {
-; GCN-LABEL: @reduction_half4(
-; GCN-NEXT:  entry:
-; GCN-NEXT:    [[TMP0:%.*]] = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> [[A:%.*]])
-; GCN-NEXT:    ret half [[TMP0]]
+; GFX9-LABEL: @reduction_half4(
+; GFX9-NEXT:  entry:
+; GFX9-NEXT:    [[TMP0:%.*]] = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> [[A:%.*]])
+; GFX9-NEXT:    ret half [[TMP0]]
+;
+; VI-LABEL: @reduction_half4(
+; VI-NEXT:  entry:
+; VI-NEXT:    [[ELT0:%.*]] = extractelement <4 x half> [[A:%.*]], i64 0
+; VI-NEXT:    [[ELT1:%.*]] = extractelement <4 x half> [[A]], i64 1
+; VI-NEXT:    [[ELT2:%.*]] = extractelement <4 x half> [[A]], i64 2
+; VI-NEXT:    [[ELT3:%.*]] = extractelement <4 x half> [[A]], i64 3
+; VI-NEXT:    [[ADD1:%.*]] = fadd fast half [[ELT1]], [[ELT0]]
+; VI-NEXT:    [[ADD2:%.*]] = fadd fast half [[ELT2]], [[ADD1]]
+; VI-NEXT:    [[ADD3:%.*]] = fadd fast half [[ELT3]], [[ADD2]]
+; VI-NEXT:    ret half [[ADD3]]
 ;
 entry:
   %elt0 = extractelement <4 x half> %a, i64 0
@@ -22,10 +33,29 @@ entry:
 }
 
 define half @reduction_half8(<8 x half> %vec8) {
-; GCN-LABEL: @reduction_half8(
-; GCN-NEXT:  entry:
-; GCN-NEXT:    [[TMP0:%.*]] = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> [[VEC8:%.*]])
-; GCN-NEXT:    ret half [[TMP0]]
+; GFX9-LABEL: @reduction_half8(
+; GFX9-NEXT:  entry:
+; GFX9-NEXT:    [[TMP0:%.*]] = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> [[VEC8:%.*]])
+; GFX9-NEXT:    ret half [[TMP0]]
+;
+; VI-LABEL: @reduction_half8(
+; VI-NEXT:  entry:
+; VI-NEXT:    [[ELT0:%.*]] = extractelement <8 x half> [[VEC8:%.*]], i64 0
+; VI-NEXT:    [[ELT1:%.*]] = extractelement <8 x half> [[VEC8]], i64 1
+; VI-NEXT:    [[ELT2:%.*]] = extractelement <8 x half> [[VEC8]], i64 2
+; VI-NEXT:    [[ELT3:%.*]] = extractelement <8 x half> [[VEC8]], i64 3
+; VI-NEXT:    [[ELT4:%.*]] = extractelement <8 x half> [[VEC8]], i64 4
+; VI-NEXT:    [[ELT5:%.*]] = extractelement <8 x half> [[VEC8]], i64 5
+; VI-NEXT:    [[ELT6:%.*]] = extractelement <8 x half> [[VEC8]], i64 6
+; VI-NEXT:    [[ELT7:%.*]] = extractelement <8 x half> [[VEC8]], i64 7
+; VI-NEXT:    [[ADD1:%.*]] = fadd fast half [[ELT1]], [[ELT0]]
+; VI-NEXT:    [[ADD2:%.*]] = fadd fast half [[ELT2]], [[ADD1]]
+; VI-NEXT:    [[ADD3:%.*]] = fadd fast half [[ELT3]], [[ADD2]]
+; VI-NEXT:    [[ADD4:%.*]] = fadd fast half [[ELT4]], [[ADD3]]
+; VI-NEXT:    [[ADD5:%.*]] = fadd fast half [[ELT5]], [[ADD4]]
+; VI-NEXT:    [[ADD6:%.*]] = fadd fast half [[ELT6]], [[ADD5]]
+; VI-NEXT:    [[ADD7:%.*]] = fadd fast half [[ELT7]], [[ADD6]]
+; VI-NEXT:    ret half [[ADD7]]
 ;
 entry:
   %elt0 = extractelement <8 x half> %vec8, i64 0
@@ -56,10 +86,37 @@ define half @reduction_half16(<16 x half> %vec16) {
 ;
 ; VI-LABEL: @reduction_half16(
 ; VI-NEXT:  entry:
-; VI-NEXT:    [[TMP0:%.*]] = shufflevector <16 x half> [[VEC16:%.*]], <16 x half> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; VI-NEXT:    [[TMP2:%.*]] = shufflevector <16 x half> [[VEC16]], <16 x half> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; VI-NEXT:    [[RDX_OP:%.*]] = fadd fast <8 x half> [[TMP0]], [[TMP2]]
-; VI-NEXT:    [[OP_RDX:%.*]] = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> [[RDX_OP]])
+; VI-NEXT:    [[ELT0:%.*]] = extractelement <16 x half> [[VEC16:%.*]], i64 0
+; VI-NEXT:    [[ELT1:%.*]] = extractelement <16 x half> [[VEC16]], i64 1
+; VI-NEXT:    [[ELT2:%.*]] = extractelement <16 x half> [[VEC16]], i64 2
+; VI-NEXT:    [[ELT3:%.*]] = extractelement <16 x half> [[VEC16]], i64 3
+; VI-NEXT:    [[ELT4:%.*]] = extractelement <16 x half> [[VEC16]], i64 4
+; VI-NEXT:    [[ELT5:%.*]] = extractelement <16 x half> [[VEC16]], i64 5
+; VI-NEXT:    [[ELT6:%.*]] = extractelement <16 x half> [[VEC16]], i64 6
+; VI-NEXT:    [[ELT7:%.*]] = extractelement <16 x half> [[VEC16]], i64 7
+; VI-NEXT:    [[ELT8:%.*]] = extractelement <16 x half> [[VEC16]], i64 8
+; VI-NEXT:    [[ELT9:%.*]] = extractelement <16 x half> [[VEC16]], i64 9
+; VI-NEXT:    [[ELT10:%.*]] = extractelement <16 x half> [[VEC16]], i64 10
+; VI-NEXT:    [[ELT11:%.*]] = extractelement <16 x half> [[VEC16]], i64 11
+; VI-NEXT:    [[ELT12:%.*]] = extractelement <16 x half> [[VEC16]], i64 12
+; VI-NEXT:    [[ELT13:%.*]] = extractelement <16 x half> [[VEC16]], i64 13
+; VI-NEXT:    [[ELT14:%.*]] = extractelement <16 x half> [[VEC16]], i64 14
+; VI-NEXT:    [[ELT15:%.*]] = extractelement <16 x half> [[VEC16]], i64 15
+; VI-NEXT:    [[ADD1:%.*]] = fadd fast half [[ELT1]], [[ELT0]]
+; VI-NEXT:    [[ADD2:%.*]] = fadd fast half [[ELT2]], [[ADD1]]
+; VI-NEXT:    [[ADD3:%.*]] = fadd fast half [[ELT3]], [[ADD2]]
+; VI-NEXT:    [[ADD4:%.*]] = fadd fast half [[ELT4]], [[ADD3]]
+; VI-NEXT:    [[ADD5:%.*]] = fadd fast half [[ELT5]], [[ADD4]]
+; VI-NEXT:    [[ADD6:%.*]] = fadd fast half [[ELT6]], [[ADD5]]
+; VI-NEXT:    [[ADD7:%.*]] = fadd fast half [[ELT7]], [[ADD6]]
+; VI-NEXT:    [[ADD8:%.*]] = fadd fast half [[ELT8]], [[ADD7]]
+; VI-NEXT:    [[ADD9:%.*]] = fadd fast half [[ELT9]], [[ADD8]]
+; VI-NEXT:    [[ADD10:%.*]] = fadd fast half [[ELT10]], [[ADD9]]
+; VI-NEXT:    [[ADD11:%.*]] = fadd fast half [[ELT11]], [[ADD10]]
+; VI-NEXT:    [[ADD12:%.*]] = fadd fast half [[ELT12]], [[ADD11]]
+; VI-NEXT:    [[ADD13:%.*]] = fadd fast half [[ELT13]], [[ADD12]]
+; VI-NEXT:    [[ADD14:%.*]] = fadd fast half [[ELT14]], [[ADD13]]
+; VI-NEXT:    [[OP_RDX:%.*]] = fadd fast half [[ELT15]], [[ADD14]]
 ; VI-NEXT:    ret half [[OP_RDX]]
 ;
 entry:
@@ -126,10 +183,21 @@ entry:
 }
 
 define i16 @reduction_v4i16(<4 x i16> %a) {
-; GCN-LABEL: @reduction_v4i16(
-; GCN-NEXT:  entry:
-; GCN-NEXT:    [[TMP0:%.*]] = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> [[A:%.*]])
-; GCN-NEXT:    ret i16 [[TMP0]]
+; GFX9-LABEL: @reduction_v4i16(
+; GFX9-NEXT:  entry:
+; GFX9-NEXT:    [[TMP0:%.*]] = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> [[A:%.*]])
+; GFX9-NEXT:    ret i16 [[TMP0]]
+;
+; VI-LABEL: @reduction_v4i16(
+; VI-NEXT:  entry:
+; VI-NEXT:    [[ELT0:%.*]] = extractelement <4 x i16> [[A:%.*]], i64 0
+; VI-NEXT:    [[ELT1:%.*]] = extractelement <4 x i16> [[A]], i64 1
+; VI-NEXT:    [[ELT2:%.*]] = extractelement <4 x i16> [[A]], i64 2
+; VI-NEXT:    [[ELT3:%.*]] = extractelement <4 x i16> [[A]], i64 3
+; VI-NEXT:    [[ADD1:%.*]] = add i16 [[ELT1]], [[ELT0]]
+; VI-NEXT:    [[ADD2:%.*]] = add i16 [[ELT2]], [[ADD1]]
+; VI-NEXT:    [[ADD3:%.*]] = add i16 [[ELT3]], [[ADD2]]
+; VI-NEXT:    ret i16 [[ADD3]]
 ;
 entry:
   %elt0 = extractelement <4 x i16> %a, i64 0
@@ -145,10 +213,29 @@ entry:
 }
 
 define i16 @reduction_v8i16(<8 x i16> %vec8) {
-; GCN-LABEL: @reduction_v8i16(
-; GCN-NEXT:  entry:
-; GCN-NEXT:    [[TMP0:%.*]] = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> [[VEC8:%.*]])
-; GCN-NEXT:    ret i16 [[TMP0]]
+; GFX9-LABEL: @reduction_v8i16(
+; GFX9-NEXT:  entry:
+; GFX9-NEXT:    [[TMP0:%.*]] = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> [[VEC8:%.*]])
+; GFX9-NEXT:    ret i16 [[TMP0]]
+;
+; VI-LABEL: @reduction_v8i16(
+; VI-NEXT:  entry:
+; VI-NEXT:    [[ELT0:%.*]] = extractelement <8 x i16> [[VEC8:%.*]], i64 0
+; VI-NEXT:    [[ELT1:%.*]] = extractelement <8 x i16> [[VEC8]], i64 1
+; VI-NEXT:    [[ELT2:%.*]] = extractelement <8 x i16> [[VEC8]], i64 2
+; VI-NEXT:    [[ELT3:%.*]] = extractelement <8 x i16> [[VEC8]], i64 3
+; VI-NEXT:    [[ELT4:%.*]] = extractelement <8 x i16> [[VEC8]], i64 4
+; VI-NEXT:    [[ELT5:%.*]] = extractelement <8 x i16> [[VEC8]], i64 5
+; VI-NEXT:    [[ELT6:%.*]] = extractelement <8 x i16> [[VEC8]], i64 6
+; VI-NEXT:    [[ELT7:%.*]] = extractelement <8 x i16> [[VEC8]], i64 7
+; VI-NEXT:    [[ADD1:%.*]] = add i16 [[ELT1]], [[ELT0]]
+; VI-NEXT:    [[ADD2:%.*]] = add i16 [[ELT2]], [[ADD1]]
+; VI-NEXT:    [[ADD3:%.*]] = add i16 [[ELT3]], [[ADD2]]
+; VI-NEXT:    [[ADD4:%.*]] = add i16 [[ELT4]], [[ADD3]]
+; VI-NEXT:    [[ADD5:%.*]] = add i16 [[ELT5]], [[ADD4]]
+; VI-NEXT:    [[ADD6:%.*]] = add i16 [[ELT6]], [[ADD5]]
+; VI-NEXT:    [[ADD7:%.*]] = add i16 [[ELT7]], [[ADD6]]
+; VI-NEXT:    ret i16 [[ADD7]]
 ;
 entry:
   %elt0 = extractelement <8 x i16> %vec8, i64 0
diff --git a/llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll b/llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll
index c9a013bd58322..d8d4460accf7c 100644
--- a/llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll
@@ -434,34 +434,18 @@ bb:
 }
 
 define void @copysign_combine_v2f16(ptr addrspace(1) %arg, half %sign) {
-; GFX8-LABEL: define void @copysign_combine_v2f16(
-; GFX8-SAME: ptr addrspace(1) [[ARG:%.*]], half [[SIGN:%.*]]) #[[ATTR0]] {
-; GFX8-NEXT:  [[BB:.*:]]
-; GFX8-NEXT:    [[TMP:%.*]] = tail call i32 @llvm.amdgcn.workitem.id.x()
-; GFX8-NEXT:    [[ITMP1:%.*]] = zext i32 [[TMP]] to i64
-; GFX8-NEXT:    [[ITMP2:%.*]] = getelementptr inbounds half, ptr addrspace(1) [[ARG]], i64 [[ITMP1]]
-; GFX8-NEXT:    [[ITMP3:%.*]] = load half, ptr addrspace(1) [[ITMP2]], align 2
-; GFX8-NEXT:    [[ITMP4:%.*]] = call half @llvm.copysign.f16(half [[ITMP3]], half [[SIGN]])
-; GFX8-NEXT:    store half [[ITMP4]], ptr addrspace(1) [[ITMP2]], align 2
-; GFX8-NEXT:    [[ITMP5:%.*]] = add nuw nsw i64 [[ITMP1]], 1
-; GFX8-NEXT:    [[ITMP6:%.*]] = getelementptr inbounds half, ptr addrspace(1) [[ARG]], i64 [[ITMP5]]
-; GFX8-NEXT:    [[ITMP7:%.*]] = load half, ptr addrspace(1) [[ITMP6]], align 2
-; GFX8-NEXT:    [[ITMP8:%.*]] = call half @llvm.copysign.f16(half [[ITMP7]], half [[SIGN]])
-; GFX8-NEXT:    store half [[ITMP8]], ptr addrspace(1) [[ITMP6]], align 2
-; GFX8-NEXT:    ret void
-;
-; GFX9-LABEL: define void @copysign_combine_v2f16(
-; GFX9-SAME: ptr addrspace(1) [[ARG:%.*]], half [[SIGN:%.*]]) #[[ATTR0]] {
-; GFX9-NEXT:  [[BB:.*:]]
-; GFX9-NEXT:    [[TMP:%.*]] = tail call i32 @llvm.amdgcn.workitem.id.x()
-; GFX9-NEXT:    [[ITMP1:%.*]] = zext i32 [[TMP]] to i64
-; GFX9-NEXT:    [[ITMP2:%.*]] = getelementptr inbounds half, ptr addrspace(1) [[ARG]], i64 [[ITMP1]]
-; GFX9-NEXT:    [[TMP0:%.*]] = load <2 x half>, ptr addrspace(1) [[ITMP2]], align 2
-; GFX9-NEXT:    [[TMP1:%.*]] = insertelement <2 x half> poison, half [[SIGN]], i32 0
-; GFX9-NEXT:    [[TMP2:%.*]] = shufflevector <2 x half> [[TMP1]], <2 x half> poison, <2 x i32> zeroinitializer
-; GFX9-NEXT:    [[TMP3:%.*]] = call <2 x half> @llvm.copysign.v2f16(<2 x half> [[TMP0]], <2 x half> [[TMP2]])
-; GFX9-NEXT:    store <2 x half> [[TMP3]], ptr addrspace(1) [[ITMP2]], align 2
-; GFX9-NEXT:    ret void
+; GCN-LABEL: define void @copysign_combine_v2f16(
+; GCN-SAME: ptr addrspace(1) [[ARG:%.*]], half [[SIGN:%.*]]) #[[ATTR0]] {
+; GCN-NEXT:  [[BB:.*:]]
+; GCN-NEXT:    [[TMP:%.*]] = tail call i32 @llvm.amdgcn.workitem.id.x()
+; GCN-NEXT:    [[ITMP1:%.*]] = zext i32 [[TMP]] to i64
+; GCN-NEXT:    [[ITMP2:%.*]] = getelementptr inbounds half, ptr addrspace(1) [[ARG]], i64 [[ITMP1]]
+; GCN-NEXT:    [[TMP0:%.*]] = load <2 x half>, ptr addrspace(1) [[ITMP2]], align 2
+; GCN-NEXT:    [[TMP1:%.*]] = insertelement <2 x half> poison, half [[SIGN]], i32 0
+; GCN-NEXT:    [[TMP2:%.*]] = shufflevector <2 x half> [[TMP1]], <2 x half> poison, <2 x i32> zeroinitializer
+; GCN-NEXT:    [[TMP3:%.*]] = call <2 x half> @llvm.copysign.v2f16(<2 x half> [[TMP0]], <2 x half> [[TMP2]])
+; GCN-NEXT:    store <2 x half> [[TMP3]], ptr addrspace(1) [[ITMP2]], align 2
+; GCN-NEXT:    ret void
 ;
 bb:
   %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
@@ -487,14 +471,13 @@ define void @copysign_combine_v4f16(ptr addrspace(1) %arg, half %sign) {
 ; GFX8-NEXT:    [[TMP:%.*]] = tail call i32 @llvm.amdgcn.workitem.id.x()
 ; GFX8-NEXT:    [[ITMP1:%.*]] = zext i32 [[TMP]] to i64
 ; GFX8-NEXT:    [[ITMP2:%.*]] = getelementptr inbounds half, ptr addrspace(1) [[ARG]], i64 [[ITMP1]]
-; GFX8-NEXT:    [[ITMP3:%.*]] = load half, ptr addrspace(1) [[ITMP2]], align 2
-; GFX8-NEXT:    [[ITMP4:%.*]] = call half @llvm.copysign.f16(half [[ITMP3]], half [[SIGN]])
-; GFX8-NEXT:    store half [[ITMP4]], ptr addrspace(1) [[ITMP2]], align 2
 ; GFX8-NEXT:    [[ITMP5:%.*]] = add nuw nsw i64 [[ITMP1]], 1
 ; GFX8-NEXT:    [[ITMP6:%.*]] = getelementptr inbounds half, ptr addrspace(1) [[ARG]], i64 [[ITMP5]]
-; GFX8-NEXT:    [[ITMP7:%.*]] = load half, ptr addrspace(1) [[ITMP6]], align 2
-; GFX8-NEXT:    [[ITMP8:%.*]] = call half @llvm.copysign.f16(half [[ITMP7]], half [[SIGN]])
-; GFX8-NEXT:    store half [[ITMP8]], ptr addrspace(1) [[ITMP6]], align 2
+; GFX8-NEXT:    [[TMP0:%.*]] = load <2 x half>, ptr addrspace(1) [[ITMP2]], align 2
+; GFX8-NEXT:    [[TMP1:%.*]] = insertelement <2 x half> poison, half [[SIGN]], i32 0
+; GFX8-NEXT:    [[TMP2:%.*]] = shufflevector <2 x half> [[TMP1]], <2 x half> poison, <2 x i32> zeroinitializer
+; GFX8-NEXT:    [[TMP3:%.*]] = call <2 x half> @llvm.copysign.v2f16(<2 x half> [[TMP0]], <2 x half> [[TMP2]])
+; GFX8-NEXT:    store <2 x half> [[TMP3]], ptr addrspace(1) [[ITMP2]], align 2
 ; GFX8-NEXT:    [[ITMP9:%.*]] = add nuw nsw i64 [[ITMP1]], 2
 ; GFX8-NEXT:    [[ITMP10:%.*]] = getelementptr inbounds half, ptr addrspace(1) [[ARG]], i64 [[ITMP9]]
 ; GFX8-NEXT:    [[ITMP11:%.*]] = load half, ptr addrspace(1) [[ITMP6]], align 2
diff --git a/llvm/test/Transforms/VectorCombine/AMDGPU/extract-insert-chain-to-shuffles.ll b/llvm/test/Transforms/VectorCombine/AMDGPU/extract-insert-chain-to-shuffles.ll
index ca22e44d1c97f..b6ba6eb984c85 100644
--- a/llvm/test/Transforms/VectorCombine/AMDGPU/extract-insert-chain-to-shuffles.ll
+++ b/llvm/test/Transforms/VectorCombine/AMDGPU/extract-insert-chain-to-shuffles.ll
@@ -8,37 +8,7 @@ define amdgpu_kernel void @extract_insert_chain_to_shuffles(<16 x i8> %in, <16 x
 ; OPT-LABEL: define amdgpu_kernel void @extract_insert_chain_to_shuffles(
 ; OPT-SAME: <16 x i8> [[IN:%.*]], <16 x i8> [[ADD:%.*]], ptr addrspace(3) [[OUT:%.*]]) #[[ATTR0:[0-9]+]] {
 ; OPT-NEXT:  [[ENTRY:.*:]]
-; OPT-NEXT:    [[TMP226:%.*]] = extractelement <16 x i8> [[IN]], i64 1
-; OPT-NEXT:    [[TMP228:%.*]] = extractelement <16 x i8> [[IN]], i64 2
-; OPT-NEXT:    [[TMP230:%.*]] = extractelement <16 x i8> [[IN]], i64 3
-; OPT-NEXT:    [[TMP232:%.*]] = extractelement <16 x i8> [[IN]], i64 4
-; OPT-NEXT:    [[TMP234:%.*]] = extractelement <16 x i8> [[IN]], i64 5
-; OPT-NEXT:    [[TMP236:%.*]] = extractelement <16 x i8> [[IN]], i64 6
-; OPT-NEXT:    [[TMP238:%.*]] = extractelement <16 x i8> [[IN]], i64 7
-; OPT-NEXT:    [[TMP240:%.*]] = extractelement <16 x i8> [[IN]], i64 8
-; OPT-NEXT:    [[TMP242:%.*]] = extractelement <16 x i8> [[IN]], i64 9
-; OPT-NEXT:    [[TMP244:%.*]] = extractelement <16 x i8> [[IN]], i64 10
-; OPT-NEXT:    [[TMP246:%.*]] = extractelement <16 x i8> [[IN]], i64 11
-; OPT-NEXT:    [[TMP248:%.*]] = extractelement <16 x i8> [[IN]], i64 12
-; OPT-NEXT:    [[TMP250:%.*]] = extractelement <16 x i8> [[IN]], i64 13
-; OPT-NEXT:    [[TMP252:%.*]] = extractelement <16 x i8> [[IN]], i64 14
-; OPT-NEXT:    [[TMP256:%.*]] = shufflevector <16 x i8> [[IN]], <16 x i8> poison, <16 x i32> <i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; OPT-NEXT:    [[I257:%.*]] = insertelement <16 x i8> [[TMP256]], i8 [[TMP226]], i64 1
-; OPT-NEXT:    [[I258:%.*]] = insertelement <16 x i8> [[I257]], i8 [[TMP228]], i64 2
-; OPT-NEXT:    [[I259:%.*]] = insertelement <16 x i8> [[I258]], i8 [[TMP230]], i64 3
-; OPT-NEXT:    [[I260:%.*]] = insertelement <16 x i8> [[I259]], i8 [[TMP232]], i64 4
-; OPT-NEXT:    [[I261:%.*]] = insertelement <16 x i8> [[I260]], i8 [[TMP234]], i64 5
-; OPT-NEXT:    [[I262:%.*]] = insertelement <16 x i8> [[I261]], i8 [[TMP236]], i64 6
-; OPT-NEXT:    [[I263:%.*]] = insertelement <16 x i8> [[I262]], i8 [[TMP238]], i64 7
-; OPT-NEXT:    [[I264:%.*]] = insertelement <16 x i8> [[I263]], i8 [[TMP240]], i64 8
-; OPT-NEXT:    [[I265:%.*]] = insertelement <16 x i8> [[I264]], i8 [[TMP242]], i64 9
-; OPT-NEXT:    [[I266:%.*]] = insertelement <16 x i8> [[I265]], i8 [[TMP244]], i64 10
-; OPT-NEXT:    [[I267:%.*]] = insertelement <16 x i8> [[I266]], i8 [[TMP246]], i64 11
-; OPT-NEXT:    [[I268:%.*]] = insertelement <16 x i8> [[I267]], i8 [[TMP248]], i64 12
-; OPT-NEXT:    [[I269:%.*]] = insertelement <16 x i8> [[I268]], i8 [[TMP250]], i64 13
-; OPT-NEXT:    [[TMP270:%.*]] = insertelement <16 x i8> [[I269]], i8 [[TMP252]], i64 14
-; OPT-NEXT:    [[TMP271:%.*]] = shufflevector <16 x i8> [[TMP270]], <16 x i8> [[IN]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 31>
-; OPT-NEXT:    [[SUM:%.*]] = add <16 x i8> [[TMP271]], [[ADD]]
+; OPT-NEXT:    [[SUM:%.*]] = add <16 x i8> [[IN]], [[ADD]]
 ; OPT-NEXT:    store <16 x i8> [[SUM]], ptr addrspace(3) [[OUT]], align 16
 ; OPT-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/VectorCombine/AMDGPU/extract-insert-i8.ll b/llvm/test/Transforms/VectorCombine/AMDGPU/extract-insert-i8.ll
index f847a90e91e9e..7a415f4cb71d0 100644
--- a/llvm/test/Transforms/VectorCombine/AMDGPU/extract-insert-i8.ll
+++ b/llvm/test/Transforms/VectorCombine/AMDGPU/extract-insert-i8.ll
@@ -5,70 +5,39 @@ define <32 x i8> @extract_insert_chain(<8 x i8> %in0, <8 x i8> %in1, <8 x i8> %i
 ; OPT-LABEL: define <32 x i8> @extract_insert_chain(
 ; OPT-SAME: <8 x i8> [[IN0:%.*]], <8 x i8> [[IN1:%.*]], <8 x i8> [[IN2:%.*]], <8 x i8> [[IN3:%.*]]) #[[ATTR0:[0-9]+]] {
 ; OPT-NEXT:  [[ENTRY:.*:]]
-; OPT-NEXT:    [[I_0_0:%.*]] = extractelement <8 x i8> [[IN0]], i64 0
-; OPT-NEXT:    [[I_0_1:%.*]] = extractelement <8 x i8> [[IN0]], i64 1
-; OPT-NEXT:    [[I_0_2:%.*]] = extractelement <8 x i8> [[IN0]], i64 2
-; OPT-NEXT:    [[I_0_3:%.*]] = extractelement <8 x i8> [[IN0]], i64 3
-; OPT-NEXT:    [[I_0_4:%.*]] = extractelement <8 x i8> [[IN0]], i64 4
-; OPT-NEXT:    [[I_0_5:%.*]] = extractelement <8 x i8> [[IN0]], i64 5
-; OPT-NEXT:    [[I_0_6:%.*]] = extractelement <8 x i8> [[IN0]], i64 6
-; OPT-NEXT:    [[I_0_7:%.*]] = extractelement <8 x i8> [[IN0]], i64 7
-; OPT-NEXT:    [[I_1_1:%.*]] = extractelement <8 x i8> [[IN1]], i64 1
-; OPT-NEXT:    [[I_1_2:%.*]] = extractelement <8 x i8> [[IN1]], i64 2
-; OPT-NEXT:    [[I_1_3:%.*]] = extractelement <8 x i8> [[IN1]], i64 3
-; OPT-NEXT:    [[I_1_4:%.*]] = extractelement <8 x i8> [[IN1]], i64 4
-; OPT-NEXT:    [[I_1_5:%.*]] = extractelement <8 x i8> [[IN1]], i64 5
-; OPT-NEXT:    [[I_1_6:%.*]] = extractelement <8 x i8> [[IN1]], i64 6
-; OPT-NEXT:    [[I_1_7:%.*]] = extractelement <8 x i8> [[IN1]], i64 7
-; OPT-NEXT:    [[I_2_1:%.*]] = extractelement <8 x i8> [[IN2]], i64 1
-; OPT-NEXT:    [[I_2_2:%.*]] = extractelement <8 x i8> [[IN2]], i64 2
-; OPT-NEXT:    [[I_2_3:%.*]] = extractelement <8 x i8> [[IN2]], i64 3
-; OPT-NEXT:    [[I_2_4:%.*]] = extractelement <8 x i8> [[IN2]], i64 4
-; OPT-NEXT:    [[I_2_5:%.*]] = extractelement <8 x i8> [[IN2]], i64 5
-; OPT-NEXT:    [[I_2_6:%.*]] = extractelement <8 x i8> [[IN2]], i64 6
-; OPT-NEXT:    [[I_2_7:%.*]] = extractelement <8 x i8> [[IN2]], i64 7
-; OPT-NEXT:    [[I_3_1:%.*]] = extractelement <8 x i8> [[IN3]], i64 1
-; OPT-NEXT:    [[I_3_2:%.*]] = extractelement <8 x i8> [[IN3]], i64 2
-; OPT-NEXT:    [[I_3_3:%.*]] = extractelement <8 x i8> [[IN3]], i64 3
-; OPT-NEXT:    [[I_3_4:%.*]] = extractelement <8 x i8> [[IN3]], i64 4
-; OPT-NEXT:    [[I_3_5:%.*]] = extractelement <8 x i8> [[IN3]], i64 5
-; OPT-NEXT:    [[I_3_6:%.*]] = extractelement <8 x i8> [[IN3]], i64 6
-; OPT-NEXT:    [[I_3_7:%.*]] = extractelement <8 x i8> [[IN3]], i64 7
-; OPT-NEXT:    [[O_0_0:%.*]] = insertelement <32 x i8> poison, i8 [[I_0_0]], i32 0
-; OPT-NEXT:    [[O_0_1:%.*]] = insertelement <32 x i8> [[O_0_0]], i8 [[I_0_1]], i32 1
-; OPT-NEXT:    [[O_0_2:%.*]] = insertelement <32 x i8> [[O_0_1]], i8 [[I_0_2]], i32 2
-; OPT-NEXT:    [[O_0_3:%.*]] = insertelement <32 x i8> [[O_0_2]], i8 [[I_0_3]], i32 3
-; OPT-NEXT:    [[O_0_4:%.*]] = insertelement <32 x i8> [[O_0_3]], i8 [[I_0_4]], i32 4
-; OPT-NEXT:    [[O_0_5:%.*]] = insertelement <32 x i8> [[O_0_4]], i8 [[I_0_5]], i32 5
-; OPT-NEXT:    [[O_0_6:%.*]] = insertelement <32 x i8> [[O_0_5]], i8 [[I_0_6]], i32 6
-; OPT-NEXT:    [[O_0_7:%.*]] = insertelement <32 x i8> [[O_0_6]], i8 [[I_0_7]], i32 7
-; OPT-NEXT:    [[TMP0:%.*]] = shufflevector <8 x i8> [[IN1]], <8 x i8> poison, <32 x i32> <i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; OPT-NEXT:    [[O_1_0:%.*]] = shufflevector <32 x i8> [[O_0_7]], <32 x i8> [[TMP0]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 32, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; OPT-NEXT:    [[O_1_1:%.*]] = insertelement <32 x i8> [[O_1_0]], i8 [[I_1_1]], i32 9
-; OPT-NEXT:    [[O_1_2:%.*]] = insertelement <32 x i8> [[O_1_1]], i8 [[I_1_2]], i32 10
-; OPT-NEXT:    [[O_1_3:%.*]] = insertelement <32 x i8> [[O_1_2]], i8 [[I_1_3]], i32 11
-; OPT-NEXT:    [[O_1_4:%.*]] = insertelement <32 x i8> [[O_1_3]], i8 [[I_1_4]], i32 12
-; OPT-NEXT:    [[O_1_5:%.*]] = insertelement <32 x i8> [[O_1_4]], i8 [[I_1_5]], i32 13
-; OPT-NEXT:    [[O_1_6:%.*]] = insertelement <32 x i8> [[O_1_5]], i8 [[I_1_6]], i32 14
-; OPT-NEXT:    [[O_1_7:%.*]] = insertelement <32 x i8> [[O_1_6]], i8 [[I_1_7]], i32 15
+; OPT-NEXT:    [[O_1_7:%.*]] = shufflevector <8 x i8> [[IN0]], <8 x i8> [[IN1]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
 ; OPT-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i8> [[IN2]], <8 x i8> poison, <32 x i32> <i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
 ; OPT-NEXT:    [[O_2_0:%.*]] = shufflevector <32 x i8> [[O_1_7]], <32 x i8> [[TMP1]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 32, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; OPT-NEXT:    [[O_2_1:%.*]] = insertelement <32 x i8> [[O_2_0]], i8 [[I_2_1]], i32 17
-; OPT-NEXT:    [[O_2_2:%.*]] = insertelement <32 x i8> [[O_2_1]], i8 [[I_2_2]], i32 18
-; OPT-NEXT:    [[O_2_3:%.*]] = insertelement <32 x i8> [[O_2_2]], i8 [[I_2_3]], i32 19
-; OPT-NEXT:    [[O_2_4:%.*]] = insertelement <32 x i8> [[O_2_3]], i8 [[I_2_4]], i32 20
-; OPT-NEXT:    [[O_2_5:%.*]] = insertelement <32 x i8> [[O_2_4]], i8 [[I_2_5]], i32 21
-; OPT-NEXT:    [[O_2_6:%.*]] = insertelement <32 x i8> [[O_2_5]], i8 [[I_2_6]], i32 22
-; OPT-NEXT:    [[O_2_7:%.*]] = insertelement <32 x i8> [[O_2_6]], i8 [[I_2_7]], i32 23
+; OPT-NEXT:    [[TMP8:%.*]] = shufflevector <8 x i8> [[IN2]], <8 x i8> poison, <32 x i32> <i32 poison, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_2_1:%.*]] = shufflevector <32 x i8> [[O_2_0]], <32 x i8> [[TMP8]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 33, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; OPT-NEXT:    [[TMP16:%.*]] = shufflevector <8 x i8> [[IN2]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 2, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_2_2:%.*]] = shufflevector <32 x i8> [[O_2_1]], <32 x i8> [[TMP16]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 34, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; OPT-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i8> [[IN2]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_2_3:%.*]] = shufflevector <32 x i8> [[O_2_2]], <32 x i8> [[TMP3]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 35, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; OPT-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i8> [[IN2]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 4, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_2_4:%.*]] = shufflevector <32 x i8> [[O_2_3]], <32 x i8> [[TMP4]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 36, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; OPT-NEXT:    [[TMP5:%.*]] = shufflevector <8 x i8> [[IN2]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 5, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_2_5:%.*]] = shufflevector <32 x i8> [[O_2_4]], <32 x i8> [[TMP5]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 37, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; OPT-NEXT:    [[TMP6:%.*]] = shufflevector <8 x i8> [[IN2]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 6, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_2_6:%.*]] = shufflevector <32 x i8> [[O_2_5]], <32 x i8> [[TMP6]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 38, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; OPT-NEXT:    [[TMP7:%.*]] = shufflevector <8 x i8> [[IN2]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_2_7:%.*]] = shufflevector <32 x i8> [[O_2_6]], <32 x i8> [[TMP7]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 39, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
 ; OPT-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i8> [[IN3]], <8 x i8> poison, <32 x i32> <i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
 ; OPT-NEXT:    [[O_3_0:%.*]] = shufflevector <32 x i8> [[O_2_7]], <32 x i8> [[TMP2]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 32, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; OPT-NEXT:    [[O_3_1:%.*]] = insertelement <32 x i8> [[O_3_0]], i8 [[I_3_1]], i32 25
-; OPT-NEXT:    [[O_3_2:%.*]] = insertelement <32 x i8> [[O_3_1]], i8 [[I_3_2]], i32 26
-; OPT-NEXT:    [[O_3_3:%.*]] = insertelement <32 x i8> [[O_3_2]], i8 [[I_3_3]], i32 27
-; OPT-NEXT:    [[O_3_4:%.*]] = insertelement <32 x i8> [[O_3_3]], i8 [[I_3_4]], i32 28
-; OPT-NEXT:    [[O_3_5:%.*]] = insertelement <32 x i8> [[O_3_4]], i8 [[I_3_5]], i32 29
-; OPT-NEXT:    [[O_3_6:%.*]] = insertelement <32 x i8> [[O_3_5]], i8 [[I_3_6]], i32 30
-; OPT-NEXT:    [[O_3_7:%.*]] = insertelement <32 x i8> [[O_3_6]], i8 [[I_3_7]], i32 31
+; OPT-NEXT:    [[TMP9:%.*]] = shufflevector <8 x i8> [[IN3]], <8 x i8> poison, <32 x i32> <i32 poison, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_3_1:%.*]] = shufflevector <32 x i8> [[O_3_0]], <32 x i8> [[TMP9]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 33, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; OPT-NEXT:    [[TMP10:%.*]] = shufflevector <8 x i8> [[IN3]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 2, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_3_2:%.*]] = shufflevector <32 x i8> [[O_3_1]], <32 x i8> [[TMP10]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 34, i32 27, i32 28, i32 29, i32 30, i32 31>
+; OPT-NEXT:    [[TMP11:%.*]] = shufflevector <8 x i8> [[IN3]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_3_3:%.*]] = shufflevector <32 x i8> [[O_3_2]], <32 x i8> [[TMP11]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 35, i32 28, i32 29, i32 30, i32 31>
+; OPT-NEXT:    [[TMP12:%.*]] = shufflevector <8 x i8> [[IN3]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 4, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_3_4:%.*]] = shufflevector <32 x i8> [[O_3_3]], <32 x i8> [[TMP12]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 36, i32 29, i32 30, i32 31>
+; OPT-NEXT:    [[TMP13:%.*]] = shufflevector <8 x i8> [[IN3]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 5, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_3_5:%.*]] = shufflevector <32 x i8> [[O_3_4]], <32 x i8> [[TMP13]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 37, i32 30, i32 31>
+; OPT-NEXT:    [[TMP14:%.*]] = shufflevector <8 x i8> [[IN3]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 6, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_3_6:%.*]] = shufflevector <32 x i8> [[O_3_5]], <32 x i8> [[TMP14]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 38, i32 31>
+; OPT-NEXT:    [[TMP15:%.*]] = shufflevector <8 x i8> [[IN3]], <8 x i8> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; OPT-NEXT:    [[O_3_7:%.*]] = shufflevector <32 x i8> [[O_3_6]], <32 x i8> [[TMP15]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 39>
 ; OPT-NEXT:    ret <32 x i8> [[O_3_7]]
 ;
 entry:
@@ -148,19 +117,19 @@ define <8 x i8> @extract_insert_chain_shortening(<32 x i8> %in) {
 ; OPT-LABEL: define <8 x i8> @extract_insert_chain_shortening(
 ; OPT-SAME: <32 x i8> [[IN:%.*]]) #[[ATTR0]] {
 ; OPT-NEXT:    [[I_1:%.*]] = extractelement <32 x i8> [[IN]], i64 17
+; OPT-NEXT:    [[I_2:%.*]] = extractelement <32 x i8> [[IN]], i64 18
 ; OPT-NEXT:    [[I_3:%.*]] = extractelement <32 x i8> [[IN]], i64 19
 ; OPT-NEXT:    [[I_5:%.*]] = extractelement <32 x i8> [[IN]], i64 21
+; OPT-NEXT:    [[I_6:%.*]] = extractelement <32 x i8> [[IN]], i64 22
 ; OPT-NEXT:    [[I_7:%.*]] = extractelement <32 x i8> [[IN]], i64 23
 ; OPT-NEXT:    [[O_0:%.*]] = shufflevector <32 x i8> [[IN]], <32 x i8> poison, <8 x i32> <i32 16, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
 ; OPT-NEXT:    [[O_1:%.*]] = insertelement <8 x i8> [[O_0]], i8 [[I_1]], i32 1
-; OPT-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i8> [[IN]], <32 x i8> poison, <8 x i32> <i32 18, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; OPT-NEXT:    [[O_2:%.*]] = shufflevector <8 x i8> [[O_1]], <8 x i8> [[TMP2]], <8 x i32> <i32 0, i32 1, i32 8, i32 3, i32 4, i32 5, i32 6, i32 7>
+; OPT-NEXT:    [[O_2:%.*]] = insertelement <8 x i8> [[O_1]], i8 [[I_2]], i32 2
 ; OPT-NEXT:    [[O_3:%.*]] = insertelement <8 x i8> [[O_2]], i8 [[I_3]], i32 3
 ; OPT-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i8> [[IN]], <32 x i8> poison, <8 x i32> <i32 20, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
 ; OPT-NEXT:    [[O_4:%.*]] = shufflevector <8 x i8> [[O_3]], <8 x i8> [[TMP1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 5, i32 6, i32 7>
 ; OPT-NEXT:    [[O_5:%.*]] = insertelement <8 x i8> [[O_4]], i8 [[I_5]], i32 5
-; OPT-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i8> [[IN]], <32 x i8> poison, <8 x i32> <i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; OPT-NEXT:    [[O_6:%.*]] = shufflevector <8 x i8> [[O_5]], <8 x i8> [[TMP3]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 8, i32 7>
+; OPT-NEXT:    [[O_6:%.*]] = insertelement <8 x i8> [[O_5]], i8 [[I_6]], i32 6
 ; OPT-NEXT:    [[O_7:%.*]] = insertelement <8 x i8> [[O_6]], i8 [[I_7]], i32 7
 ; OPT-NEXT:    ret <8 x i8> [[O_7]]
 ;