[llvm] AMDGPU: Split sgpr reg-sequences that are copied to vgpr (PR #216073)

Petar Avramovic via llvm-commits llvm-commits at lists.llvm.org
Thu Aug 13 07:56:22 PDT 2026


https://github.com/petar-avramovic created https://github.com/llvm/llvm-project/pull/216073

None

>From d2fd51351f5d0c3f6c08478fcf6584c7e395709d Mon Sep 17 00:00:00 2001
From: Petar Avramovic <Petar.Avramovic at amd.com>
Date: Thu, 13 Aug 2026 16:38:23 +0200
Subject: [PATCH] AMDGPU: Split sgpr reg-sequences that are copied to vgpr

---
 .../Target/AMDGPU/GCNPreRAOptimizations.cpp   |  301 ++
 llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll    |    4 +-
 .../AMDGPU/GlobalISel/atomicrmw_udec_wrap.ll  |   24 +-
 .../AMDGPU/GlobalISel/atomicrmw_uinc_wrap.ll  |   28 +-
 .../AMDGPU/GlobalISel/bitcast_38_i16.ll       |    6 +-
 .../bug_shuffle_vector_to_scalar.ll           |   13 +-
 .../AMDGPU/GlobalISel/cvt_f32_ubyte.ll        |    4 +-
 .../AMDGPU/GlobalISel/extractelement.i128.ll  |   16 +-
 .../AMDGPU/GlobalISel/extractelement.i16.ll   |   32 +-
 llvm/test/CodeGen/AMDGPU/GlobalISel/fabs.ll   |   14 +-
 .../CodeGen/AMDGPU/GlobalISel/flat-scratch.ll |   93 +-
 llvm/test/CodeGen/AMDGPU/GlobalISel/fneg.ll   |   14 +-
 llvm/test/CodeGen/AMDGPU/GlobalISel/frem.ll   |  116 +-
 ...licit-kernarg-backend-usage-global-isel.ll |   10 +-
 .../AMDGPU/GlobalISel/insertelement.ll        | 3242 +++++----------
 .../AMDGPU/GlobalISel/llvm.amdgcn.div.fmas.ll |   12 +-
 .../GlobalISel/llvm.amdgcn.intersect_ray.ll   |  151 +-
 .../GlobalISel/llvm.amdgcn.mfma.gfx90a.ll     |   71 +-
 .../GlobalISel/llvm.amdgcn.trig.preop.ll      |    8 +-
 .../AMDGPU/GlobalISel/llvm.atomic.cmpxchg.ll  |   44 +-
 .../AMDGPU/GlobalISel/load-unaligned.ll       |   88 +-
 .../AMDGPU/GlobalISel/load-uniform-in-vgpr.ll |  202 +-
 llvm/test/CodeGen/AMDGPU/GlobalISel/mad.ll    |   14 +-
 llvm/test/CodeGen/AMDGPU/GlobalISel/mul.ll    |   66 +-
 llvm/test/CodeGen/AMDGPU/GlobalISel/or.ll     |   24 +-
 .../test/CodeGen/AMDGPU/GlobalISel/sdivrem.ll | 1382 +++----
 .../AMDGPU/GlobalISel/strict_fma.f32.ll       |    2 +-
 .../AMDGPU/GlobalISel/strict_fma.f64.ll       |   59 +-
 llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll    |    4 +-
 .../test/CodeGen/AMDGPU/GlobalISel/udivrem.ll |  554 +--
 .../GlobalISel/widen-i8-i16-scalar-loads.ll   |    4 +-
 .../AMDGPU/GlobalISel/wmma-gfx12-w32-imm.ll   |  216 +-
 .../AMDGPU/GlobalISel/wmma-gfx12-w64-imm.ll   |  159 +-
 llvm/test/CodeGen/AMDGPU/add.ll               |  118 +-
 llvm/test/CodeGen/AMDGPU/always-uniform.ll    |   12 +-
 .../test/CodeGen/AMDGPU/amdgpu-cs-chain-cc.ll |   22 +-
 .../AMDGPU/amdgpu-cs-chain-preserve-cc.ll     |   19 +-
 llvm/test/CodeGen/AMDGPU/and.ll               |   16 +-
 .../CodeGen/AMDGPU/any_extend_vector_inreg.ll |  126 +-
 .../CodeGen/AMDGPU/atomicrmw_usub_cond.ll     |   16 +-
 .../test/CodeGen/AMDGPU/carryout-selection.ll |    8 +-
 llvm/test/CodeGen/AMDGPU/cluster_stores.ll    |   34 +-
 llvm/test/CodeGen/AMDGPU/ctlz.ll              |    3 +-
 llvm/test/CodeGen/AMDGPU/ctlz_zero_poison.ll  |   36 +-
 llvm/test/CodeGen/AMDGPU/ctpop64.ll           |   16 +-
 llvm/test/CodeGen/AMDGPU/cttz.ll              |    3 +-
 llvm/test/CodeGen/AMDGPU/cttz_zero_poison.ll  |   39 +-
 .../CodeGen/AMDGPU/extract_vector_dynelt.ll   |  308 +-
 .../fast-unaligned-load-store.global.ll       |    4 +-
 llvm/test/CodeGen/AMDGPU/fcopysign.f64.ll     |    4 +-
 llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll   |   12 +-
 llvm/test/CodeGen/AMDGPU/flat-scratch.ll      |  613 +--
 llvm/test/CodeGen/AMDGPU/flat_atomics.ll      |   35 +-
 llvm/test/CodeGen/AMDGPU/flat_atomics_i64.ll  |   24 +-
 .../AMDGPU/flat_atomics_i64_noprivate.ll      |  140 +-
 llvm/test/CodeGen/AMDGPU/fmaximum.ll          |   12 +-
 llvm/test/CodeGen/AMDGPU/fmaxnum.ll           |  200 +-
 llvm/test/CodeGen/AMDGPU/fminimum.ll          |   12 +-
 llvm/test/CodeGen/AMDGPU/fminnum.ll           |  180 +-
 .../AMDGPU/fmul-2-combine-multi-use.ll        |   94 +-
 llvm/test/CodeGen/AMDGPU/fnearbyint.ll        |   28 +-
 llvm/test/CodeGen/AMDGPU/fneg-fabs.f64.ll     |   24 +-
 .../AMDGPU/fp-min-max-num-flat-atomics.ll     |  130 +-
 llvm/test/CodeGen/AMDGPU/fp_to_sint.ll        |   71 +-
 llvm/test/CodeGen/AMDGPU/fp_to_uint.ll        |   44 +-
 llvm/test/CodeGen/AMDGPU/frem.ll              |   12 +-
 llvm/test/CodeGen/AMDGPU/fsqrt.f32.ll         |   48 +-
 .../AMDGPU/gfx-callable-return-types.ll       |   42 +-
 llvm/test/CodeGen/AMDGPU/global-saddr-load.ll |    6 +-
 llvm/test/CodeGen/AMDGPU/global_atomics.ll    |    8 +-
 .../test/CodeGen/AMDGPU/global_atomics_i64.ll |  116 +-
 llvm/test/CodeGen/AMDGPU/half.ll              |  246 +-
 .../identical-subrange-spill-infloop.ll       |    6 +-
 .../CodeGen/AMDGPU/indirect-addressing-si.ll  |  274 +-
 .../CodeGen/AMDGPU/insert_vector_dynelt.ll    |   96 +-
 llvm/test/CodeGen/AMDGPU/kernel-args.ll       |  100 +-
 .../CodeGen/AMDGPU/llvm.amdgcn.atomic.dec.ll  |    4 +-
 .../CodeGen/AMDGPU/llvm.amdgcn.atomic.inc.ll  |    4 +-
 .../AMDGPU/llvm.amdgcn.av.load.b128.ll        |   72 +-
 .../AMDGPU/llvm.amdgcn.bvh8_intersect_ray.ll  |   27 +-
 .../AMDGPU/llvm.amdgcn.cvt.scale.pk.ll        |   32 +-
 .../llvm.amdgcn.cvt.scalef32.pk.gfx13.ll      |  176 +-
 .../AMDGPU/llvm.amdgcn.dual_intersect_ray.ll  |   31 +-
 .../llvm.amdgcn.global.load.async.to.lds.ll   |    4 +-
 .../AMDGPU/llvm.amdgcn.intersect_ray.ll       |  219 +-
 .../CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll |  884 ++--
 llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll  | 1046 ++---
 .../AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll        |  123 +-
 .../AMDGPU/llvm.amdgcn.ptr.s.buffer.load.ll   |   20 +-
 .../CodeGen/AMDGPU/llvm.amdgcn.readlane.ll    |   20 +-
 .../CodeGen/AMDGPU/llvm.amdgcn.reduce.add.ll  |   30 +-
 .../CodeGen/AMDGPU/llvm.amdgcn.reduce.sub.ll  |   90 +-
 .../CodeGen/AMDGPU/llvm.amdgcn.reduce.xor.ll  |   22 +-
 .../llvm.amdgcn.wmma.imm.gfx1250.w32.ll       | 1040 ++---
 .../llvm.amdgcn.wmma.imm.gfx1251.w32.ll       |   51 +-
 llvm/test/CodeGen/AMDGPU/llvm.exp.ll          |  288 +-
 llvm/test/CodeGen/AMDGPU/llvm.exp10.ll        |  288 +-
 llvm/test/CodeGen/AMDGPU/llvm.maximum.f64.ll  |   18 +-
 llvm/test/CodeGen/AMDGPU/llvm.minimum.f64.ll  |   18 +-
 llvm/test/CodeGen/AMDGPU/load-constant-i1.ll  | 2282 +++++------
 llvm/test/CodeGen/AMDGPU/load-constant-i16.ll | 2883 ++++++-------
 llvm/test/CodeGen/AMDGPU/load-constant-i32.ll |  500 +--
 llvm/test/CodeGen/AMDGPU/load-constant-i64.ll |   18 +-
 llvm/test/CodeGen/AMDGPU/load-constant-i8.ll  | 3593 +++++++++--------
 llvm/test/CodeGen/AMDGPU/load-global-f32.ll   |   44 +-
 llvm/test/CodeGen/AMDGPU/load-global-i16.ll   |  659 ++-
 llvm/test/CodeGen/AMDGPU/load-global-i32.ll   |  305 +-
 llvm/test/CodeGen/AMDGPU/load-global-i8.ll    |  428 +-
 .../local-stack-alloc-block-sp-reference.ll   |   66 +-
 .../lower-work-group-id-intrinsics-hsa.ll     |    6 +-
 .../lower-work-group-id-intrinsics-pal.ll     |    6 +-
 .../AMDGPU/lower-work-group-id-intrinsics.ll  |   25 +-
 llvm/test/CodeGen/AMDGPU/mad_64_32.ll         |   10 +-
 .../CodeGen/AMDGPU/max-hard-clause-length.ll  |  439 +-
 .../CodeGen/AMDGPU/memintrinsic-unroll.ll     |   36 +-
 llvm/test/CodeGen/AMDGPU/memset-pattern.ll    |   99 +-
 llvm/test/CodeGen/AMDGPU/min.ll               |   64 +-
 llvm/test/CodeGen/AMDGPU/packed-fp32.ll       |  123 +-
 llvm/test/CodeGen/AMDGPU/packed-fp64.ll       |  141 +-
 llvm/test/CodeGen/AMDGPU/packed-u64.ll        |   72 +-
 ...-alloca-vector-dynamic-idx-bitcasts-llc.ll |  255 +-
 .../AMDGPU/promote-constOffset-to-imm.ll      |   10 +-
 llvm/test/CodeGen/AMDGPU/ptradd-sdag.ll       |    8 +-
 llvm/test/CodeGen/AMDGPU/rotr.ll              |   10 +-
 llvm/test/CodeGen/AMDGPU/saddo.ll             |    2 +-
 .../CodeGen/AMDGPU/scale-offset-global.ll     |    4 +-
 .../AMDGPU/schedule-amdgpu-trackers.ll        |    2 +-
 llvm/test/CodeGen/AMDGPU/scmp.ll              |    6 +-
 llvm/test/CodeGen/AMDGPU/sdiv64.ll            |    8 +-
 .../AMDGPU/sext-divergence-driven-isel.ll     |   10 +-
 llvm/test/CodeGen/AMDGPU/shl.ll               |    2 +-
 llvm/test/CodeGen/AMDGPU/sra.ll               |   40 +-
 llvm/test/CodeGen/AMDGPU/srem.ll              |  186 +-
 llvm/test/CodeGen/AMDGPU/srem64.ll            |    8 +-
 llvm/test/CodeGen/AMDGPU/ssubo.ll             |    2 +-
 llvm/test/CodeGen/AMDGPU/store-weird-sizes.ll |   16 +-
 llvm/test/CodeGen/AMDGPU/sub.ll               |    4 +-
 llvm/test/CodeGen/AMDGPU/ucmp.ll              |    6 +-
 llvm/test/CodeGen/AMDGPU/udiv.ll              |   24 +-
 llvm/test/CodeGen/AMDGPU/uint_to_fp.f64.ll    |    8 +-
 .../CodeGen/AMDGPU/whole-wave-functions.ll    |   38 +-
 141 files changed, 12450 insertions(+), 14839 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/GCNPreRAOptimizations.cpp b/llvm/lib/Target/AMDGPU/GCNPreRAOptimizations.cpp
index 50514e819f412..96946c0fcfa72 100644
--- a/llvm/lib/Target/AMDGPU/GCNPreRAOptimizations.cpp
+++ b/llvm/lib/Target/AMDGPU/GCNPreRAOptimizations.cpp
@@ -47,6 +47,7 @@ namespace {
 
 class GCNPreRAOptimizationsImpl {
 private:
+  const GCNSubtarget *ST;
   const SIInstrInfo *TII;
   const SIRegisterInfo *TRI;
   MachineRegisterInfo *MRI;
@@ -55,6 +56,8 @@ class GCNPreRAOptimizationsImpl {
   bool processReg(Register Reg);
   void hintTrue16Copy(const MachineInstr &MI);
   bool optimizeBVHStack(MachineInstr &MI);
+  bool optimizeSGPRToVGPRCopy(MachineInstr &MI);
+  bool narrowTupleCopy(MachineInstr &MI);
 
 public:
   GCNPreRAOptimizationsImpl(LiveIntervals *LS) : LIS(LS) {}
@@ -299,8 +302,299 @@ bool GCNPreRAOptimizationsImpl::optimizeBVHStack(MachineInstr &MI) {
   return true;
 }
 
+// A scalar move of an immediate, which has a VALU equivalent of the same width.
+static bool isMovImm(const MachineInstr &Def) {
+  switch (Def.getOpcode()) {
+  case AMDGPU::S_MOV_B32:
+  case AMDGPU::S_MOV_B64:
+  case AMDGPU::S_MOV_B64_IMM_PSEUDO:
+    return Def.getNumOperands() == 2 && Def.getOperand(1).isImm();
+  default:
+    return false;
+  }
+}
+
+// A def of a tuple subregister that can be redone directly in a VGPR: an
+// immediate move or a copy of a 32 bit virtual register.
+static bool isRewritableTupleDef(const MachineInstr &Def, Register Tuple,
+                                 const SIRegisterInfo *TRI,
+                                 const MachineRegisterInfo *MRI) {
+  if (isMovImm(Def))
+    return true;
+  switch (Def.getOpcode()) {
+  case AMDGPU::COPY: {
+    const MachineOperand &CopySrc = Def.getOperand(1);
+    if (!CopySrc.isReg())
+      return false;
+    Register CopySrcReg = CopySrc.getReg();
+    if (CopySrcReg == Tuple)
+      return true;
+    if (!CopySrcReg.isVirtual())
+      return false;
+    return (CopySrc.getSubReg() ? TRI->getSubRegIdxSize(CopySrc.getSubReg())
+                                : TRI->getRegSizeInBits(CopySrcReg, *MRI)) == 32;
+  }
+  default:
+    return false;
+  }
+}
+
+// Build a register tuple directly in VGPRs instead of building it in SGPRs and
+// copying the whole tuple:
+//
+//   undef %0.sub0:sgpr_256 = S_MOV_B32 1
+//   %0.sub1:sgpr_256 = COPY %0.sub0
+//   ...
+//   %1:vreg_256 = COPY %0:sgpr_256
+// =>
+//   undef %1.sub0:vreg_256 = V_MOV_B32_e32 1, implicit $exec
+//   %1.sub1:vreg_256 = COPY %1.sub0
+//   ...
+//
+// Runs after SIFoldOperands and the coalescer, so any use that could have taken
+// the SGPR tuple has already folded or coalesced it.
+bool GCNPreRAOptimizationsImpl::optimizeSGPRToVGPRCopy(MachineInstr &MI) {
+  const MachineOperand &DstMO = MI.getOperand(0);
+  const MachineOperand &SrcMO = MI.getOperand(1);
+  if (DstMO.getSubReg() || SrcMO.getSubReg())
+    return false;
+
+  Register Dst = DstMO.getReg();
+  Register Src = SrcMO.getReg();
+  if (!Dst.isVirtual() || !Src.isVirtual())
+    return false;
+
+  const TargetRegisterClass *DstRC = MRI->getRegClass(Dst);
+  const TargetRegisterClass *SrcRC = MRI->getRegClass(Src);
+  if (!TRI->isVGPRClass(DstRC) || !TRI->isSGPRClass(SrcRC) ||
+      DstRC->getSizeInBits() != SrcRC->getSizeInBits() ||
+      DstRC->getSizeInBits() <= 32)
+    return false;
+
+  // The tuple must be dead after this copy, otherwise the SGPR value is still
+  // needed and rewriting the defs only adds copies. Uses inside the tuple
+  // initialization itself are rewritten together with the defs.
+  for (MachineInstr &Use : MRI->use_nodbg_instructions(Src)) {
+    if (&Use == &MI)
+      continue;
+    if (Use.isCopy() && Use.getOperand(0).getReg() == Src)
+      continue;
+    return false;
+  }
+
+  // Nothing of Dst may be live before the copy, the defs are rewritten in
+  // place and would clobber it.
+  if (!LIS->hasInterval(Dst) || !LIS->hasInterval(Src))
+    return false;
+
+  SlotIndex CopyIdx = LIS->getInstructionIndex(MI);
+  if (LIS->getInterval(Dst).beginIndex() != CopyIdx.getRegSlot())
+    return false;
+
+  // All defs must be 32 bit subregister defs in this block, before the copy.
+  // Immediates and copies are redone in the VGPR tuple, anything else stays
+  // where it is and its result is copied into the tuple.
+  // A wide copy between properly aligned register classes is expanded into 64
+  // bit moves, splitting it into 32 bit pieces would double the move count.
+  // Without that the copy already costs one move per element, so building the
+  // tuple in VGPRs is never worse and gets rid of the SGPR tuple.
+  bool CopyIsPaired = DstRC->getSizeInBits() % 64 == 0 &&
+                      TRI->isProperlyAlignedRC(*DstRC) &&
+                      (ST->hasVMovB64Inst() || ST->hasPkMovB32());
+
+  SmallVector<MachineInstr *, 8> Defs;
+  bool AnyRewritten = !CopyIsPaired;
+  for (MachineOperand &DefMO : MRI->def_operands(Src)) {
+    MachineInstr &Def = *DefMO.getParent();
+    if (Def.getParent() != MI.getParent() ||
+        !SlotIndex::isEarlierInstr(LIS->getInstructionIndex(Def), CopyIdx))
+      return false;
+
+    unsigned SubIdx = DefMO.getSubReg();
+    if (!SubIdx || DefMO.isTied() || is_contained(Defs, &Def))
+      return false;
+
+    // A 64 bit immediate move is redone as a single V_MOV_B64, and a 64 bit
+    // copy inside the tuple stays a single move, so both are worth splitting
+    // even though that half of the wide copy costs one move either way.
+    unsigned SubSize = TRI->getSubRegIdxSize(SubIdx);
+    bool IsMovImm = isMovImm(Def) &&
+                    (SubSize == 32 ||
+                     TII->isImmOperandLegal(TII->get(AMDGPU::V_MOV_B64_PSEUDO),
+                                            1, Def.getOperand(1)));
+    bool IsTupleCopy = Def.isCopy() && Def.getOperand(1).getReg() == Src;
+    if (SubSize != 32 &&
+        (SubSize != 64 || !(ST->hasVMovB64Inst() || ST->hasPkMovB32()) ||
+         !(IsMovImm || IsTupleCopy)))
+      return false;
+
+    // A def that is redone in the VGPR tuple goes away, so the wide copy pays
+    // for itself. Everything else at best trades it for a narrow one.
+    bool Rewritable = isRewritableTupleDef(Def, Src, TRI, MRI);
+    if (Rewritable)
+      AnyRewritten = true;
+    if (!Rewritable && !TRI->getSubRegisterClass(SrcRC, SubIdx))
+      return false;
+
+    Defs.push_back(&Def);
+  }
+
+  // Splitting the copy without getting rid of at least one of the defs just
+  // trades one wide copy for a series of narrow ones.
+  if (!AnyRewritten)
+    return false;
+
+  // The tuple is rebuilt at the copy, keep the defs in their original order.
+  sort(Defs, [this](const MachineInstr *A, const MachineInstr *B) {
+    return SlotIndex::isEarlierInstr(LIS->getInstructionIndex(*A),
+                                     LIS->getInstructionIndex(*B));
+  });
+
+  LLVM_DEBUG(dbgs() << "Building tuple in VGPRs for: " << MI);
+
+  SmallVector<Register, 8> NewVRegs;
+  SmallSetVector<Register, 8> Recompute;
+  for (MachineInstr *Def : Defs) {
+    MachineOperand &DefMO = *Def->findRegisterDefOperand(Src, TRI);
+    unsigned SubIdx = DefMO.getSubReg();
+    RegState UndefFlag = getUndefRegState(DefMO.isUndef());
+    bool KeepDef = !isRewritableTupleDef(*Def, Src, TRI, MRI);
+    unsigned Opc = AMDGPU::COPY;
+    if (!KeepDef) {
+      if (Def->getOpcode() == AMDGPU::S_MOV_B32)
+        Opc = AMDGPU::V_MOV_B32_e32;
+      else if (isMovImm(*Def))
+        Opc = AMDGPU::V_MOV_B64_PSEUDO;
+    }
+
+    // Anything that is not redone in the VGPR tuple stays at its original slot
+    // index: moving it could extend the live range of a physical register, and
+    // it may not be safe to move at all. Redefine it into a plain 32 bit
+    // virtual register in place and copy that into the tuple instead.
+    MachineOperand DefSrcMO = Def->getOperand(1);
+    DefSrcMO.clearParent();
+    if (KeepDef) {
+      Register Tmp =
+          MRI->createVirtualRegister(TRI->getSubRegisterClass(SrcRC, SubIdx));
+      DefMO.setReg(Tmp);
+      DefMO.setSubReg(0);
+      DefMO.setIsUndef(false);
+      NewVRegs.push_back(Tmp);
+      DefSrcMO = MachineOperand::CreateReg(Tmp, false);
+    } else if (DefSrcMO.isReg()) {
+      if (DefSrcMO.getReg() == Src)
+        DefSrcMO.setReg(Dst);
+      else if (DefSrcMO.getReg().isVirtual())
+        // Moving the def to the copy extends the live range of what it reads.
+        Recompute.insert(DefSrcMO.getReg());
+    }
+
+    auto *NewDef =
+        BuildMI(*MI.getParent(), MI, Def->getDebugLoc(), TII->get(Opc))
+            .addDef(Dst, UndefFlag, SubIdx)
+            .add(DefSrcMO)
+            .getInstr();
+    if (!KeepDef) {
+      LIS->RemoveMachineInstrFromMaps(*Def);
+      Def->eraseFromParent();
+    }
+    LIS->InsertMachineInstrInMaps(*NewDef);
+    LLVM_DEBUG(dbgs() << "  " << *NewDef);
+  }
+
+  LIS->RemoveMachineInstrFromMaps(MI);
+  MI.eraseFromParent();
+
+  LIS->removeInterval(Src);
+  LIS->removeInterval(Dst);
+  LIS->createAndComputeVirtRegInterval(Dst);
+  for (Register Tmp : NewVRegs)
+    LIS->createAndComputeVirtRegInterval(Tmp);
+  for (Register Reg : Recompute) {
+    LIS->removeInterval(Reg);
+    LIS->createAndComputeVirtRegInterval(Reg);
+  }
+
+  return true;
+}
+
+// Copying a whole tuple when the uses only read some of its subregisters
+// writes the rest for nothing. Copy only the subregisters that are read:
+//
+//   %1:vreg_64 = COPY %0:sreg_64
+//   ... = %1.sub1
+// =>
+//   %2:vgpr_32 = COPY %0.sub1:sreg_64
+//   ... = %2
+bool GCNPreRAOptimizationsImpl::narrowTupleCopy(MachineInstr &MI) {
+  const MachineOperand &DstMO = MI.getOperand(0);
+  const MachineOperand &SrcMO = MI.getOperand(1);
+  if (DstMO.getSubReg() || SrcMO.getSubReg())
+    return false;
+
+  Register Dst = DstMO.getReg();
+  Register Src = SrcMO.getReg();
+  if (!Dst.isVirtual() || !Src.isVirtual() || !MRI->hasOneDef(Dst))
+    return false;
+
+  const TargetRegisterClass *DstRC = MRI->getRegClass(Dst);
+  const TargetRegisterClass *SrcRC = MRI->getRegClass(Src);
+  unsigned Size = DstRC->getSizeInBits();
+  if (Size <= 32 || Size != SrcRC->getSizeInBits())
+    return false;
+
+  SmallSetVector<unsigned, 8> SubRegs;
+  for (const MachineOperand &UseMO : MRI->use_nodbg_operands(Dst)) {
+    unsigned SubIdx = UseMO.getSubReg();
+    if (!SubIdx || TRI->getSubRegIdxSize(SubIdx) != 32 ||
+        !TRI->getSubRegisterClass(DstRC, SubIdx) ||
+        !TRI->getSubRegisterClass(SrcRC, SubIdx))
+      return false;
+    SubRegs.insert(SubIdx);
+  }
+
+  if (SubRegs.empty() || SubRegs.size() * 32 >= Size)
+    return false;
+
+  LLVM_DEBUG(dbgs() << "Narrowing tuple copy: " << MI);
+
+  SmallVector<Register, 8> NewVRegs;
+  for (unsigned SubIdx : SubRegs) {
+    Register NewReg =
+        MRI->createVirtualRegister(TRI->getSubRegisterClass(DstRC, SubIdx));
+    auto *NewCopy = BuildMI(*MI.getParent(), MI, MI.getDebugLoc(),
+                            TII->get(AMDGPU::COPY), NewReg)
+                        .addReg(Src, RegState::NoFlags, SubIdx)
+                        .getInstr();
+    LIS->InsertMachineInstrInMaps(*NewCopy);
+    NewVRegs.push_back(NewReg);
+    LLVM_DEBUG(dbgs() << "  " << *NewCopy);
+
+    for (MachineOperand &UseMO :
+         make_early_inc_range(MRI->use_nodbg_operands(Dst))) {
+      if (UseMO.getSubReg() != SubIdx)
+        continue;
+      UseMO.setReg(NewReg);
+      UseMO.setSubReg(AMDGPU::NoSubRegister);
+    }
+  }
+
+  MRI->markUsesInDebugValueAsUndef(Dst);
+  LIS->RemoveMachineInstrFromMaps(MI);
+  MI.eraseFromParent();
+
+  LIS->removeInterval(Src);
+  LIS->removeInterval(Dst);
+  LIS->createAndComputeVirtRegInterval(Src);
+  for (Register NewReg : NewVRegs)
+    LIS->createAndComputeVirtRegInterval(NewReg);
+
+  return true;
+}
+
 bool GCNPreRAOptimizationsImpl::run(MachineFunction &MF) {
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
+  this->ST = &ST;
   TII = ST.getInstrInfo();
   MRI = &MF.getRegInfo();
   TRI = ST.getRegisterInfo();
@@ -319,6 +613,13 @@ bool GCNPreRAOptimizationsImpl::run(MachineFunction &MF) {
     Changed |= processReg(Reg);
   }
 
+  for (MachineBasicBlock &MBB : MF) {
+    for (MachineInstr &MI : make_early_inc_range(MBB)) {
+      if (MI.isCopy())
+        Changed |= optimizeSGPRToVGPRCopy(MI) || narrowTupleCopy(MI);
+    }
+  }
+
   const bool HasBVHStack = ST.hasBVHDualAndBVH8Insts();
   const bool HasRealTrue16 = ST.useRealTrue16Insts();
 
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll
index 763370705ebda..6f6b62409d460 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll
@@ -508,7 +508,7 @@ define void @s_uaddo_uadde(i64 inreg %a, i64 inreg %b, ptr addrspace(1) %res, pt
 ; GFX11-NEXT:    s_add_u32 s0, s0, s2
 ; GFX11-NEXT:    s_addc_u32 s1, s1, s3
 ; GFX11-NEXT:    s_cselect_b32 s2, 1, 0
-; GFX11-NEXT:    v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0
+; GFX11-NEXT:    v_dual_mov_b32 v4, s0 :: v_dual_mov_b32 v5, s1
 ; GFX11-NEXT:    v_mov_b32_e32 v6, s2
 ; GFX11-NEXT:    global_store_b64 v[0:1], v[4:5], off
 ; GFX11-NEXT:    global_store_b32 v[2:3], v6, off
@@ -525,7 +525,7 @@ define void @s_uaddo_uadde(i64 inreg %a, i64 inreg %b, ptr addrspace(1) %res, pt
 ; GFX12-NEXT:    s_add_co_ci_u32 s1, s1, s3
 ; GFX12-NEXT:    s_cselect_b32 s2, 1, 0
 ; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0
+; GFX12-NEXT:    v_dual_mov_b32 v4, s0 :: v_dual_mov_b32 v5, s1
 ; GFX12-NEXT:    v_mov_b32_e32 v6, s2
 ; GFX12-NEXT:    global_store_b64 v[0:1], v[4:5], off
 ; GFX12-NEXT:    global_store_b32 v[2:3], v6, off
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_udec_wrap.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_udec_wrap.ll
index 1a8ae1180d0aa..e74012ef6f689 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_udec_wrap.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_udec_wrap.ll
@@ -1730,8 +1730,8 @@ define amdgpu_kernel void @flat_atomic_dec_ret_i64(ptr %out, ptr %ptr) #1 {
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
 ; CI-NEXT:    s_add_u32 s0, s0, 4
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v5, s1
 ; CI-NEXT:    v_mov_b32_e32 v4, s0
+; CI-NEXT:    v_mov_b32_e32 v5, s1
 ; CI-NEXT:    flat_store_dword v[2:3], v0
 ; CI-NEXT:    flat_store_dword v[4:5], v1
 ; CI-NEXT:    s_endpgm
@@ -1754,8 +1754,8 @@ define amdgpu_kernel void @flat_atomic_dec_ret_i64(ptr %out, ptr %ptr) #1 {
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    s_add_u32 s0, s0, 4
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    flat_store_dword v[2:3], v0
 ; VI-NEXT:    flat_store_dword v[4:5], v1
 ; VI-NEXT:    s_endpgm
@@ -1838,8 +1838,8 @@ define amdgpu_kernel void @flat_atomic_dec_ret_i64_offset(ptr %out, ptr %ptr) #1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
 ; CI-NEXT:    s_add_u32 s0, s0, 4
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v5, s1
 ; CI-NEXT:    v_mov_b32_e32 v4, s0
+; CI-NEXT:    v_mov_b32_e32 v5, s1
 ; CI-NEXT:    flat_store_dword v[2:3], v0
 ; CI-NEXT:    flat_store_dword v[4:5], v1
 ; CI-NEXT:    s_endpgm
@@ -1864,8 +1864,8 @@ define amdgpu_kernel void @flat_atomic_dec_ret_i64_offset(ptr %out, ptr %ptr) #1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    s_add_u32 s0, s0, 4
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    flat_store_dword v[2:3], v0
 ; VI-NEXT:    flat_store_dword v[4:5], v1
 ; VI-NEXT:    s_endpgm
@@ -2025,8 +2025,8 @@ define amdgpu_kernel void @flat_atomic_dec_noret_i64_offset(ptr %ptr) #1 {
 ; CI-NEXT:    s_add_u32 s0, s0, 32
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; CI-NEXT:    buffer_wbinvl1_vol
@@ -2043,8 +2043,8 @@ define amdgpu_kernel void @flat_atomic_dec_noret_i64_offset(ptr %ptr) #1 {
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -2115,8 +2115,8 @@ define amdgpu_kernel void @flat_atomic_dec_noret_i64_offset_system(ptr %ptr) #1
 ; CI-NEXT:    s_add_u32 s0, s0, 32
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; CI-NEXT:    buffer_wbinvl1_vol
@@ -2133,8 +2133,8 @@ define amdgpu_kernel void @flat_atomic_dec_noret_i64_offset_system(ptr %ptr) #1
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -3187,8 +3187,8 @@ define amdgpu_kernel void @global_atomic_dec_noret_i64_offset(ptr addrspace(1) %
 ; CI-NEXT:    s_add_u32 s0, s0, 32
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; CI-NEXT:    s_waitcnt vmcnt(0)
 ; CI-NEXT:    buffer_wbinvl1_vol
@@ -3205,8 +3205,8 @@ define amdgpu_kernel void @global_atomic_dec_noret_i64_offset(ptr addrspace(1) %
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -3265,8 +3265,8 @@ define amdgpu_kernel void @global_atomic_dec_noret_i64_offset_system(ptr addrspa
 ; CI-NEXT:    s_add_u32 s0, s0, 32
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; CI-NEXT:    s_waitcnt vmcnt(0)
 ; CI-NEXT:    buffer_wbinvl1_vol
@@ -3283,8 +3283,8 @@ define amdgpu_kernel void @global_atomic_dec_noret_i64_offset_system(ptr addrspa
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_uinc_wrap.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_uinc_wrap.ll
index bb26e2131473f..9f4d4469af19c 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_uinc_wrap.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_uinc_wrap.ll
@@ -1928,8 +1928,8 @@ define amdgpu_kernel void @global_atomic_inc_noret_i64_offset(ptr addrspace(1) %
 ; CI-NEXT:    s_add_u32 s0, s0, 32
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; CI-NEXT:    s_waitcnt vmcnt(0)
 ; CI-NEXT:    buffer_wbinvl1_vol
@@ -1946,8 +1946,8 @@ define amdgpu_kernel void @global_atomic_inc_noret_i64_offset(ptr addrspace(1) %
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -2017,8 +2017,8 @@ define amdgpu_kernel void @global_atomic_inc_noret_i64_offset_system(ptr addrspa
 ; CI-NEXT:    s_add_u32 s0, s0, 32
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; CI-NEXT:    s_waitcnt vmcnt(0)
 ; CI-NEXT:    buffer_wbinvl1_vol
@@ -2035,8 +2035,8 @@ define amdgpu_kernel void @global_atomic_inc_noret_i64_offset_system(ptr addrspa
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -3345,8 +3345,8 @@ define amdgpu_kernel void @flat_atomic_inc_ret_i64(ptr %out, ptr %ptr) #1 {
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
 ; CI-NEXT:    s_add_u32 s0, s0, 4
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v5, s1
 ; CI-NEXT:    v_mov_b32_e32 v4, s0
+; CI-NEXT:    v_mov_b32_e32 v5, s1
 ; CI-NEXT:    flat_store_dword v[2:3], v0
 ; CI-NEXT:    flat_store_dword v[4:5], v1
 ; CI-NEXT:    s_endpgm
@@ -3369,8 +3369,8 @@ define amdgpu_kernel void @flat_atomic_inc_ret_i64(ptr %out, ptr %ptr) #1 {
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    s_add_u32 s0, s0, 4
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    flat_store_dword v[2:3], v0
 ; VI-NEXT:    flat_store_dword v[4:5], v1
 ; VI-NEXT:    s_endpgm
@@ -3466,8 +3466,8 @@ define amdgpu_kernel void @flat_atomic_inc_ret_i64_offset(ptr %out, ptr %ptr) #1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
 ; CI-NEXT:    s_add_u32 s0, s0, 4
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v5, s1
 ; CI-NEXT:    v_mov_b32_e32 v4, s0
+; CI-NEXT:    v_mov_b32_e32 v5, s1
 ; CI-NEXT:    flat_store_dword v[2:3], v0
 ; CI-NEXT:    flat_store_dword v[4:5], v1
 ; CI-NEXT:    s_endpgm
@@ -3492,8 +3492,8 @@ define amdgpu_kernel void @flat_atomic_inc_ret_i64_offset(ptr %out, ptr %ptr) #1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    s_add_u32 s0, s0, 4
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    flat_store_dword v[2:3], v0
 ; VI-NEXT:    flat_store_dword v[4:5], v1
 ; VI-NEXT:    s_endpgm
@@ -3592,8 +3592,8 @@ define amdgpu_kernel void @flat_atomic_inc_ret_i64_offset_system(ptr %out, ptr %
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
 ; CI-NEXT:    s_add_u32 s0, s0, 4
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v5, s1
 ; CI-NEXT:    v_mov_b32_e32 v4, s0
+; CI-NEXT:    v_mov_b32_e32 v5, s1
 ; CI-NEXT:    flat_store_dword v[2:3], v0
 ; CI-NEXT:    flat_store_dword v[4:5], v1
 ; CI-NEXT:    s_endpgm
@@ -3618,8 +3618,8 @@ define amdgpu_kernel void @flat_atomic_inc_ret_i64_offset_system(ptr %out, ptr %
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    s_add_u32 s0, s0, 4
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    flat_store_dword v[2:3], v0
 ; VI-NEXT:    flat_store_dword v[4:5], v1
 ; VI-NEXT:    s_endpgm
@@ -3805,8 +3805,8 @@ define amdgpu_kernel void @flat_atomic_inc_noret_i64_offset(ptr %ptr) #1 {
 ; CI-NEXT:    s_add_u32 s0, s0, 32
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; CI-NEXT:    buffer_wbinvl1_vol
@@ -3823,8 +3823,8 @@ define amdgpu_kernel void @flat_atomic_inc_noret_i64_offset(ptr %ptr) #1 {
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -3906,8 +3906,8 @@ define amdgpu_kernel void @flat_atomic_inc_noret_i64_offset_system(ptr %ptr) #1
 ; CI-NEXT:    s_add_u32 s0, s0, 32
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; CI-NEXT:    buffer_wbinvl1_vol
@@ -3924,8 +3924,8 @@ define amdgpu_kernel void @flat_atomic_inc_noret_i64_offset_system(ptr %ptr) #1
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/bitcast_38_i16.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/bitcast_38_i16.ll
index 848d542d72501..30b6d79d4644e 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/bitcast_38_i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/bitcast_38_i16.ll
@@ -11,8 +11,6 @@ define void @main(<19 x i32> %arg) {
 ; GCN-NEXT:    s_mov_b32 s4, 0
 ; GCN-NEXT:    v_cmp_eq_u16_e32 vcc, 0, v0
 ; GCN-NEXT:    v_mov_b32_e32 v1, 0
-; GCN-NEXT:    s_mov_b32 s12, s4
-; GCN-NEXT:    s_mov_b32 s13, s4
 ; GCN-NEXT:    s_mov_b32 s5, s4
 ; GCN-NEXT:    s_mov_b32 s6, s4
 ; GCN-NEXT:    s_mov_b32 s7, s4
@@ -23,8 +21,8 @@ define void @main(<19 x i32> %arg) {
 ; GCN-NEXT:    v_cndmask_b32_e64 v0, 0, 1, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v2, v1
 ; GCN-NEXT:    v_mov_b32_e32 v3, v1
-; GCN-NEXT:    v_mov_b32_e32 v4, s12
-; GCN-NEXT:    v_mov_b32_e32 v5, s13
+; GCN-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-NEXT:    v_mov_b32_e32 v5, s4
 ; GCN-NEXT:    image_store v[0:3], v[4:5], s[4:11] unorm
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_setpc_b64 s[30:31]
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/bug_shuffle_vector_to_scalar.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/bug_shuffle_vector_to_scalar.ll
index 9d40f03d99670..b4a6a34c14335 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/bug_shuffle_vector_to_scalar.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/bug_shuffle_vector_to_scalar.ll
@@ -8,21 +8,18 @@ define amdgpu_gs <4 x float> @_amdgpu_gs_main() {
 ; CHECK-LABEL: _amdgpu_gs_main:
 ; CHECK:       ; %bb.0: ; %bb
 ; CHECK-NEXT:    v_mov_b32_e32 v0, 16
-; CHECK-NEXT:    ds_read2_b32 v[3:4], v0 offset1:1
+; CHECK-NEXT:    ds_read2_b32 v[0:1], v0 offset1:1
 ; CHECK-NEXT:    s_mov_b32 s0, 0
 ; CHECK-NEXT:    s_mov_b32 s1, s0
 ; CHECK-NEXT:    s_mov_b32 s2, s0
 ; CHECK-NEXT:    s_mov_b32 s3, s0
 ; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
-; CHECK-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-NEXT:    v_readfirstlane_b32 s4, v0
 ; CHECK-NEXT:    v_mov_b32_e32 v0, s4
-; CHECK-NEXT:    v_mov_b32_e32 v1, s5
-; CHECK-NEXT:    v_mov_b32_e32 v2, s6
-; CHECK-NEXT:    v_mov_b32_e32 v3, s7
-; CHECK-NEXT:    v_mov_b32_e32 v5, 0
-; CHECK-NEXT:    buffer_store_dwordx4 v[0:3], v5, s[0:3], 0 idxen
+; CHECK-NEXT:    v_mov_b32_e32 v2, 0
+; CHECK-NEXT:    buffer_store_dwordx4 v[0:3], v2, s[0:3], 0 idxen
 ; CHECK-NEXT:    s_nop 0
-; CHECK-NEXT:    v_mov_b32_e32 v0, v4
+; CHECK-NEXT:    v_mov_b32_e32 v0, v1
 ; CHECK-NEXT:    s_waitcnt vmcnt(0)
 ; CHECK-NEXT:    ; return to shader part epilog
 bb:
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/cvt_f32_ubyte.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/cvt_f32_ubyte.ll
index cb727f183556b..7bd7196c5241b 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/cvt_f32_ubyte.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/cvt_f32_ubyte.ll
@@ -888,8 +888,8 @@ define amdgpu_kernel void @load_v7i8_to_v7f32(ptr addrspace(1) noalias %out, ptr
 ; VI-NEXT:    v_mov_b32_e32 v7, s0
 ; VI-NEXT:    s_add_u32 s0, s0, 16
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v10, s1
 ; VI-NEXT:    v_mov_b32_e32 v9, s0
+; VI-NEXT:    v_mov_b32_e32 v10, s1
 ; VI-NEXT:    s_waitcnt vmcnt(6)
 ; VI-NEXT:    v_cvt_f32_ubyte0_e32 v0, v0
 ; VI-NEXT:    s_waitcnt vmcnt(5)
@@ -961,8 +961,8 @@ define amdgpu_kernel void @load_v8i8_to_v8f32(ptr addrspace(1) noalias %out, ptr
 ; VI-NEXT:    v_mov_b32_e32 v8, s0
 ; VI-NEXT:    s_add_u32 s0, s0, 16
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v11, s1
 ; VI-NEXT:    v_mov_b32_e32 v10, s0
+; VI-NEXT:    v_mov_b32_e32 v11, s1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_cvt_f32_ubyte0_sdwa v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0
 ; VI-NEXT:    v_cvt_f32_ubyte0_sdwa v1, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i128.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i128.ll
index 56fb1c61a3ea7..01417d7dbe468 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i128.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i128.ll
@@ -90,9 +90,8 @@ define amdgpu_ps i128 @extractelement_vgpr_v4i128_sgpr_idx(ptr addrspace(1) %ptr
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s0, s2, 3
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 4
-; GFX9-NEXT:    s_mov_b32 s1, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v2, s0
-; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v2
 ; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
@@ -107,9 +106,8 @@ define amdgpu_ps i128 @extractelement_vgpr_v4i128_sgpr_idx(ptr addrspace(1) %ptr
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s0, s2, 3
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 4
-; GFX8-NEXT:    s_mov_b32 s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
+; GFX8-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
 ; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX8-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
@@ -138,9 +136,8 @@ define amdgpu_ps i128 @extractelement_vgpr_v4i128_sgpr_idx(ptr addrspace(1) %ptr
 ; GFX10-LABEL: extractelement_vgpr_v4i128_sgpr_idx:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s0, s2, 3
-; GFX10-NEXT:    s_mov_b32 s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 4
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
 ; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
@@ -155,12 +152,11 @@ define amdgpu_ps i128 @extractelement_vgpr_v4i128_sgpr_idx(ptr addrspace(1) %ptr
 ; GFX11-LABEL: extractelement_vgpr_v4i128_sgpr_idx:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s0, s2, 3
-; GFX11-NEXT:    s_mov_b32 s1, 0
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 4
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX11-NEXT:    global_load_b128 v[0:3], v[0:1], off
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i16.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i16.ll
index 33823dcbe927e..89085bec4cf56 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i16.ll
@@ -75,9 +75,8 @@ define amdgpu_ps i16 @extractelement_vgpr_v4i16_sgpr_idx(ptr addrspace(1) %ptr,
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s0, s2, 3
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX9-NEXT:    s_mov_b32 s1, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v2, s0
-; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v2
 ; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX9-NEXT:    global_load_ushort v0, v[0:1], off
@@ -89,9 +88,8 @@ define amdgpu_ps i16 @extractelement_vgpr_v4i16_sgpr_idx(ptr addrspace(1) %ptr,
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s0, s2, 3
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX8-NEXT:    s_mov_b32 s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
+; GFX8-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
 ; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX8-NEXT:    flat_load_ushort v0, v[0:1]
@@ -114,9 +112,8 @@ define amdgpu_ps i16 @extractelement_vgpr_v4i16_sgpr_idx(ptr addrspace(1) %ptr,
 ; GFX10-LABEL: extractelement_vgpr_v4i16_sgpr_idx:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s0, s2, 3
-; GFX10-NEXT:    s_mov_b32 s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
 ; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
@@ -128,12 +125,11 @@ define amdgpu_ps i16 @extractelement_vgpr_v4i16_sgpr_idx(ptr addrspace(1) %ptr,
 ; GFX11-LABEL: extractelement_vgpr_v4i16_sgpr_idx:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s0, s2, 3
-; GFX11-NEXT:    s_mov_b32 s1, 0
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX11-NEXT:    global_load_u16 v0, v[0:1], off
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
@@ -719,9 +715,8 @@ define amdgpu_ps i16 @extractelement_vgpr_v8i16_sgpr_idx(ptr addrspace(1) %ptr,
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s0, s2, 7
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX9-NEXT:    s_mov_b32 s1, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v2, s0
-; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v2
 ; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX9-NEXT:    global_load_ushort v0, v[0:1], off
@@ -733,9 +728,8 @@ define amdgpu_ps i16 @extractelement_vgpr_v8i16_sgpr_idx(ptr addrspace(1) %ptr,
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s0, s2, 7
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX8-NEXT:    s_mov_b32 s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
+; GFX8-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
 ; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX8-NEXT:    flat_load_ushort v0, v[0:1]
@@ -758,9 +752,8 @@ define amdgpu_ps i16 @extractelement_vgpr_v8i16_sgpr_idx(ptr addrspace(1) %ptr,
 ; GFX10-LABEL: extractelement_vgpr_v8i16_sgpr_idx:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s0, s2, 7
-; GFX10-NEXT:    s_mov_b32 s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
 ; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
@@ -772,12 +765,11 @@ define amdgpu_ps i16 @extractelement_vgpr_v8i16_sgpr_idx(ptr addrspace(1) %ptr,
 ; GFX11-LABEL: extractelement_vgpr_v8i16_sgpr_idx:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s0, s2, 7
-; GFX11-NEXT:    s_mov_b32 s1, 0
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX11-NEXT:    global_load_u16 v0, v[0:1], off
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fabs.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fabs.ll
index fb81eba869caa..74b3504aef1f5 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/fabs.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fabs.ll
@@ -127,7 +127,7 @@ define amdgpu_ps void @v_fabs_f64(double %in, ptr addrspace(1) %out) {
 define amdgpu_ps void @s_fabs_f64(double inreg %in, ptr addrspace(1) %out) {
 ; GCN-LABEL: s_fabs_f64:
 ; GCN:       ; %bb.0:
-; GCN-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GCN-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GCN-NEXT:    v_and_b32_e32 v3, 0x7fffffff, v3
 ; GCN-NEXT:    global_store_b64 v[0:1], v[2:3], off
@@ -139,12 +139,12 @@ define amdgpu_ps void @s_fabs_f64(double inreg %in, ptr addrspace(1) %out) {
 define amdgpu_ps void @s_fabs_f64_salu_use(double inreg %in, i32 inreg %val, ptr addrspace(1) %out) {
 ; GFX11-LABEL: s_fabs_f64_salu_use:
 ; GFX11:       ; %bb.0:
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX11-NEXT:    s_cmp_eq_u32 s2, 0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT:    v_and_b32_e32 v3, 0x7fffffff, v3
 ; GFX11-NEXT:    v_readfirstlane_b32 s0, v2
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    v_and_b32_e32 v3, 0x7fffffff, v3
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX11-NEXT:    s_cselect_b64 s[0:1], s[0:1], 0
 ; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
@@ -153,12 +153,12 @@ define amdgpu_ps void @s_fabs_f64_salu_use(double inreg %in, i32 inreg %val, ptr
 ;
 ; GFX12-LABEL: s_fabs_f64_salu_use:
 ; GFX12:       ; %bb.0:
-; GFX12-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX12-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX12-NEXT:    s_cmp_eq_u32 s2, 0
 ; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX12-NEXT:    v_and_b32_e32 v3, 0x7fffffff, v3
 ; GFX12-NEXT:    v_readfirstlane_b32 s0, v2
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-NEXT:    v_and_b32_e32 v3, 0x7fffffff, v3
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX12-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX12-NEXT:    s_cselect_b64 s[0:1], s[0:1], 0
 ; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
index fbdaee61479bf..e2b08fbac7625 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
@@ -2777,12 +2777,9 @@ define void @store_load_v3i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX9-LABEL: store_load_v3i32_unaligned:
 ; GFX9:       ; %bb.0: ; %bb
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    s_mov_b32 s2, 3
-; GFX9-NEXT:    s_mov_b32 s1, 2
-; GFX9-NEXT:    s_mov_b32 s0, 1
-; GFX9-NEXT:    v_mov_b32_e32 v3, s2
-; GFX9-NEXT:    v_mov_b32_e32 v2, s1
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
+; GFX9-NEXT:    v_mov_b32_e32 v3, 3
+; GFX9-NEXT:    v_mov_b32_e32 v2, 2
+; GFX9-NEXT:    v_mov_b32_e32 v1, 1
 ; GFX9-NEXT:    scratch_store_dwordx3 v0, v[1:3], off
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    scratch_load_dwordx3 v[0:2], v0, off glc
@@ -2792,12 +2789,9 @@ define void @store_load_v3i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX10-LABEL: store_load_v3i32_unaligned:
 ; GFX10:       ; %bb.0: ; %bb
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX10-NEXT:    s_mov_b32 s2, 3
-; GFX10-NEXT:    s_mov_b32 s1, 2
-; GFX10-NEXT:    s_mov_b32 s0, 1
-; GFX10-NEXT:    v_mov_b32_e32 v3, s2
-; GFX10-NEXT:    v_mov_b32_e32 v2, s1
-; GFX10-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10-NEXT:    v_mov_b32_e32 v3, 3
+; GFX10-NEXT:    v_mov_b32_e32 v2, 2
+; GFX10-NEXT:    v_mov_b32_e32 v1, 1
 ; GFX10-NEXT:    scratch_store_dwordx3 v0, v[1:3], off
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    scratch_load_dwordx3 v[0:2], v0, off glc dlc
@@ -2807,12 +2801,9 @@ define void @store_load_v3i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX942-LABEL: store_load_v3i32_unaligned:
 ; GFX942:       ; %bb.0: ; %bb
 ; GFX942-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-NEXT:    s_mov_b32 s2, 3
-; GFX942-NEXT:    s_mov_b32 s1, 2
-; GFX942-NEXT:    s_mov_b32 s0, 1
-; GFX942-NEXT:    v_mov_b32_e32 v4, s2
-; GFX942-NEXT:    v_mov_b32_e32 v3, s1
-; GFX942-NEXT:    v_mov_b32_e32 v2, s0
+; GFX942-NEXT:    v_mov_b32_e32 v4, 3
+; GFX942-NEXT:    v_mov_b32_e32 v3, 2
+; GFX942-NEXT:    v_mov_b32_e32 v2, 1
 ; GFX942-NEXT:    scratch_store_dwordx3 v0, v[2:4], off sc0 sc1
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
 ; GFX942-NEXT:    scratch_load_dwordx3 v[0:2], v0, off sc0 sc1
@@ -2822,11 +2813,8 @@ define void @store_load_v3i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX11-LABEL: store_load_v3i32_unaligned:
 ; GFX11:       ; %bb.0: ; %bb
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT:    s_mov_b32 s2, 3
-; GFX11-NEXT:    s_mov_b32 s1, 2
-; GFX11-NEXT:    s_mov_b32 s0, 1
-; GFX11-NEXT:    v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v2, s1
-; GFX11-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11-NEXT:    v_dual_mov_b32 v3, 3 :: v_dual_mov_b32 v2, 2
+; GFX11-NEXT:    v_mov_b32_e32 v1, 1
 ; GFX11-NEXT:    scratch_store_b96 v0, v[1:3], off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    scratch_load_b96 v[0:2], v0, off glc dlc
@@ -2840,12 +2828,8 @@ define void @store_load_v3i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX12-NEXT:    s_wait_samplecnt 0x0
 ; GFX12-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    s_mov_b32 s2, 3
-; GFX12-NEXT:    s_mov_b32 s1, 2
-; GFX12-NEXT:    s_mov_b32 s0, 1
-; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v2, s1
-; GFX12-NEXT:    v_mov_b32_e32 v1, s0
+; GFX12-NEXT:    v_dual_mov_b32 v3, 3 :: v_dual_mov_b32 v2, 2
+; GFX12-NEXT:    v_mov_b32_e32 v1, 1
 ; GFX12-NEXT:    s_wait_storecnt 0x0
 ; GFX12-NEXT:    scratch_store_b96 v0, v[1:3], off scope:SCOPE_SYS
 ; GFX12-NEXT:    s_wait_storecnt 0x0
@@ -3154,14 +3138,10 @@ define void @store_load_v4i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX9-LABEL: store_load_v4i32_unaligned:
 ; GFX9:       ; %bb.0: ; %bb
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    s_mov_b32 s3, 4
-; GFX9-NEXT:    s_mov_b32 s2, 3
-; GFX9-NEXT:    s_mov_b32 s1, 2
-; GFX9-NEXT:    s_mov_b32 s0, 1
-; GFX9-NEXT:    v_mov_b32_e32 v4, s3
-; GFX9-NEXT:    v_mov_b32_e32 v3, s2
-; GFX9-NEXT:    v_mov_b32_e32 v2, s1
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
+; GFX9-NEXT:    v_mov_b32_e32 v4, 4
+; GFX9-NEXT:    v_mov_b32_e32 v3, 3
+; GFX9-NEXT:    v_mov_b32_e32 v2, 2
+; GFX9-NEXT:    v_mov_b32_e32 v1, 1
 ; GFX9-NEXT:    scratch_store_dwordx4 v0, v[1:4], off
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    scratch_load_dwordx4 v[0:3], v0, off glc
@@ -3171,14 +3151,10 @@ define void @store_load_v4i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX10-LABEL: store_load_v4i32_unaligned:
 ; GFX10:       ; %bb.0: ; %bb
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX10-NEXT:    s_mov_b32 s3, 4
-; GFX10-NEXT:    s_mov_b32 s2, 3
-; GFX10-NEXT:    s_mov_b32 s1, 2
-; GFX10-NEXT:    s_mov_b32 s0, 1
-; GFX10-NEXT:    v_mov_b32_e32 v4, s3
-; GFX10-NEXT:    v_mov_b32_e32 v3, s2
-; GFX10-NEXT:    v_mov_b32_e32 v2, s1
-; GFX10-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10-NEXT:    v_mov_b32_e32 v4, 4
+; GFX10-NEXT:    v_mov_b32_e32 v3, 3
+; GFX10-NEXT:    v_mov_b32_e32 v2, 2
+; GFX10-NEXT:    v_mov_b32_e32 v1, 1
 ; GFX10-NEXT:    scratch_store_dwordx4 v0, v[1:4], off
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    scratch_load_dwordx4 v[0:3], v0, off glc dlc
@@ -3188,12 +3164,10 @@ define void @store_load_v4i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX942-LABEL: store_load_v4i32_unaligned:
 ; GFX942:       ; %bb.0: ; %bb
 ; GFX942-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-NEXT:    s_mov_b32 s3, 4
-; GFX942-NEXT:    s_mov_b32 s2, 3
-; GFX942-NEXT:    s_mov_b32 s1, 2
-; GFX942-NEXT:    s_mov_b32 s0, 1
-; GFX942-NEXT:    v_mov_b64_e32 v[4:5], s[2:3]
-; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[0:1]
+; GFX942-NEXT:    v_mov_b32_e32 v5, 4
+; GFX942-NEXT:    v_mov_b32_e32 v4, 3
+; GFX942-NEXT:    v_mov_b32_e32 v3, 2
+; GFX942-NEXT:    v_mov_b32_e32 v2, 1
 ; GFX942-NEXT:    scratch_store_dwordx4 v0, v[2:5], off sc0 sc1
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
 ; GFX942-NEXT:    scratch_load_dwordx4 v[0:3], v0, off sc0 sc1
@@ -3203,12 +3177,8 @@ define void @store_load_v4i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX11-LABEL: store_load_v4i32_unaligned:
 ; GFX11:       ; %bb.0: ; %bb
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT:    s_mov_b32 s3, 4
-; GFX11-NEXT:    s_mov_b32 s2, 3
-; GFX11-NEXT:    s_mov_b32 s1, 2
-; GFX11-NEXT:    s_mov_b32 s0, 1
-; GFX11-NEXT:    v_dual_mov_b32 v4, s3 :: v_dual_mov_b32 v3, s2
-; GFX11-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX11-NEXT:    v_dual_mov_b32 v4, 4 :: v_dual_mov_b32 v3, 3
+; GFX11-NEXT:    v_dual_mov_b32 v2, 2 :: v_dual_mov_b32 v1, 1
 ; GFX11-NEXT:    scratch_store_b128 v0, v[1:4], off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    scratch_load_b128 v[0:3], v0, off glc dlc
@@ -3222,13 +3192,8 @@ define void @store_load_v4i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX12-NEXT:    s_wait_samplecnt 0x0
 ; GFX12-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    s_mov_b32 s3, 4
-; GFX12-NEXT:    s_mov_b32 s2, 3
-; GFX12-NEXT:    s_mov_b32 s1, 2
-; GFX12-NEXT:    s_mov_b32 s0, 1
-; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    v_dual_mov_b32 v4, s3 :: v_dual_mov_b32 v3, s2
-; GFX12-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX12-NEXT:    v_dual_mov_b32 v4, 4 :: v_dual_mov_b32 v3, 3
+; GFX12-NEXT:    v_dual_mov_b32 v2, 2 :: v_dual_mov_b32 v1, 1
 ; GFX12-NEXT:    s_wait_storecnt 0x0
 ; GFX12-NEXT:    scratch_store_b128 v0, v[1:4], off scope:SCOPE_SYS
 ; GFX12-NEXT:    s_wait_storecnt 0x0
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fneg.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fneg.ll
index 0720794f94102..4dff8cfcec830 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/fneg.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fneg.ll
@@ -127,7 +127,7 @@ define amdgpu_ps void @v_fneg_f64(double %in, ptr addrspace(1) %out) {
 define amdgpu_ps void @s_fneg_f64(double inreg %in, ptr addrspace(1) %out) {
 ; GCN-LABEL: s_fneg_f64:
 ; GCN:       ; %bb.0:
-; GCN-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GCN-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GCN-NEXT:    v_xor_b32_e32 v3, 0x80000000, v3
 ; GCN-NEXT:    global_store_b64 v[0:1], v[2:3], off
@@ -139,12 +139,12 @@ define amdgpu_ps void @s_fneg_f64(double inreg %in, ptr addrspace(1) %out) {
 define amdgpu_ps void @s_fneg_f64_salu_use(double inreg %in, i32 inreg %val, ptr addrspace(1) %out) {
 ; GFX11-LABEL: s_fneg_f64_salu_use:
 ; GFX11:       ; %bb.0:
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX11-NEXT:    s_cmp_eq_u32 s2, 0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT:    v_xor_b32_e32 v3, 0x80000000, v3
 ; GFX11-NEXT:    v_readfirstlane_b32 s0, v2
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    v_xor_b32_e32 v3, 0x80000000, v3
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX11-NEXT:    s_cselect_b64 s[0:1], s[0:1], 0
 ; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
@@ -153,12 +153,12 @@ define amdgpu_ps void @s_fneg_f64_salu_use(double inreg %in, i32 inreg %val, ptr
 ;
 ; GFX12-LABEL: s_fneg_f64_salu_use:
 ; GFX12:       ; %bb.0:
-; GFX12-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX12-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX12-NEXT:    s_cmp_eq_u32 s2, 0
 ; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX12-NEXT:    v_xor_b32_e32 v3, 0x80000000, v3
 ; GFX12-NEXT:    v_readfirstlane_b32 s0, v2
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-NEXT:    v_xor_b32_e32 v3, 0x80000000, v3
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX12-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX12-NEXT:    s_cselect_b64 s[0:1], s[0:1], 0
 ; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/frem.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/frem.ll
index 0487a73109141..16bb3f2998e00 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/frem.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/frem.ll
@@ -2048,34 +2048,34 @@ define amdgpu_kernel void @frem_v4f16(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; CI-NEXT:    s_cselect_b32 s12, 1, 0
 ; CI-NEXT:    v_cmp_nge_f32_e32 vcc, v0, v1
 ; CI-NEXT:    s_or_b64 s[2:3], vcc, vcc
-; CI-NEXT:    s_cselect_b32 s3, 1, 0
-; CI-NEXT:    s_bfe_u32 s11, s17, 0x100000
-; CI-NEXT:    s_bfe_u32 s2, s6, 0x100000
+; CI-NEXT:    s_cselect_b32 s2, 1, 0
+; CI-NEXT:    s_bfe_u32 s3, s17, 0x100000
+; CI-NEXT:    s_bfe_u32 s6, s6, 0x100000
 ; CI-NEXT:    s_cmp_lg_u32 s16, 0
-; CI-NEXT:    s_cselect_b32 s2, s11, s2
+; CI-NEXT:    s_cselect_b32 s6, s3, s6
 ; CI-NEXT:    s_cmp_lg_u32 s18, 0
-; CI-NEXT:    s_cselect_b32 s2, s2, s11
-; CI-NEXT:    s_bfe_u32 s6, s9, 0x100000
+; CI-NEXT:    s_cselect_b32 s6, s6, s3
+; CI-NEXT:    s_bfe_u32 s9, s9, 0x100000
 ; CI-NEXT:    s_cmp_lg_u32 s8, 0
-; CI-NEXT:    s_cselect_b32 s6, s11, s6
+; CI-NEXT:    s_cselect_b32 s8, s3, s9
 ; CI-NEXT:    s_cmp_lg_u32 s7, 0
-; CI-NEXT:    s_cselect_b32 s6, s6, s11
-; CI-NEXT:    s_lshl_b32 s6, s6, 16
-; CI-NEXT:    s_or_b32 s2, s2, s6
-; CI-NEXT:    s_bfe_u32 s6, s10, 0x100000
+; CI-NEXT:    s_cselect_b32 s7, s8, s3
+; CI-NEXT:    s_lshl_b32 s7, s7, 16
+; CI-NEXT:    s_or_b32 s6, s6, s7
+; CI-NEXT:    s_bfe_u32 s7, s10, 0x100000
 ; CI-NEXT:    s_cmp_lg_u32 s4, 0
-; CI-NEXT:    s_cselect_b32 s4, s11, s6
+; CI-NEXT:    s_cselect_b32 s4, s3, s7
 ; CI-NEXT:    s_cmp_lg_u32 s5, 0
-; CI-NEXT:    s_cselect_b32 s4, s4, s11
+; CI-NEXT:    s_cselect_b32 s4, s4, s3
 ; CI-NEXT:    s_bfe_u32 s5, s13, 0x100000
 ; CI-NEXT:    s_cmp_lg_u32 s12, 0
-; CI-NEXT:    s_cselect_b32 s5, s11, s5
-; CI-NEXT:    s_cmp_lg_u32 s3, 0
-; CI-NEXT:    s_cselect_b32 s3, s5, s11
-; CI-NEXT:    s_lshl_b32 s3, s3, 16
-; CI-NEXT:    s_or_b32 s3, s4, s3
-; CI-NEXT:    v_mov_b32_e32 v0, s2
-; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    s_cselect_b32 s5, s3, s5
+; CI-NEXT:    s_cmp_lg_u32 s2, 0
+; CI-NEXT:    s_cselect_b32 s2, s5, s3
+; CI-NEXT:    s_lshl_b32 s2, s2, 16
+; CI-NEXT:    s_or_b32 s2, s4, s2
+; CI-NEXT:    v_mov_b32_e32 v0, s6
+; CI-NEXT:    v_mov_b32_e32 v1, s2
 ; CI-NEXT:    s_mov_b32 s2, -1
 ; CI-NEXT:    s_mov_b32 s3, 0xf000
 ; CI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
@@ -2486,34 +2486,34 @@ define amdgpu_kernel void @frem_v4f16(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; VI-NEXT:    s_movk_i32 s17, 0x7e00
 ; VI-NEXT:    s_cselect_b32 s12, 1, 0
 ; VI-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; VI-NEXT:    s_cselect_b32 s3, 1, 0
-; VI-NEXT:    s_bfe_u32 s11, s17, 0x100000
-; VI-NEXT:    s_bfe_u32 s2, s6, 0x100000
+; VI-NEXT:    s_cselect_b32 s2, 1, 0
+; VI-NEXT:    s_bfe_u32 s3, s17, 0x100000
+; VI-NEXT:    s_bfe_u32 s6, s6, 0x100000
 ; VI-NEXT:    s_cmp_lg_u32 s16, 0
-; VI-NEXT:    s_cselect_b32 s2, s11, s2
+; VI-NEXT:    s_cselect_b32 s6, s3, s6
 ; VI-NEXT:    s_cmp_lg_u32 s18, 0
-; VI-NEXT:    s_cselect_b32 s2, s2, s11
-; VI-NEXT:    s_bfe_u32 s6, s9, 0x100000
+; VI-NEXT:    s_cselect_b32 s6, s6, s3
+; VI-NEXT:    s_bfe_u32 s9, s9, 0x100000
 ; VI-NEXT:    s_cmp_lg_u32 s8, 0
-; VI-NEXT:    s_cselect_b32 s6, s11, s6
+; VI-NEXT:    s_cselect_b32 s8, s3, s9
 ; VI-NEXT:    s_cmp_lg_u32 s7, 0
-; VI-NEXT:    s_cselect_b32 s6, s6, s11
-; VI-NEXT:    s_lshl_b32 s6, s6, 16
-; VI-NEXT:    s_or_b32 s2, s2, s6
-; VI-NEXT:    s_bfe_u32 s6, s10, 0x100000
+; VI-NEXT:    s_cselect_b32 s7, s8, s3
+; VI-NEXT:    s_lshl_b32 s7, s7, 16
+; VI-NEXT:    s_or_b32 s6, s6, s7
+; VI-NEXT:    s_bfe_u32 s7, s10, 0x100000
 ; VI-NEXT:    s_cmp_lg_u32 s4, 0
-; VI-NEXT:    s_cselect_b32 s4, s11, s6
+; VI-NEXT:    s_cselect_b32 s4, s3, s7
 ; VI-NEXT:    s_cmp_lg_u32 s5, 0
-; VI-NEXT:    s_cselect_b32 s4, s4, s11
+; VI-NEXT:    s_cselect_b32 s4, s4, s3
 ; VI-NEXT:    s_bfe_u32 s5, s13, 0x100000
 ; VI-NEXT:    s_cmp_lg_u32 s12, 0
-; VI-NEXT:    s_cselect_b32 s5, s11, s5
-; VI-NEXT:    s_cmp_lg_u32 s3, 0
-; VI-NEXT:    s_cselect_b32 s3, s5, s11
-; VI-NEXT:    s_lshl_b32 s3, s3, 16
-; VI-NEXT:    s_or_b32 s3, s4, s3
-; VI-NEXT:    v_mov_b32_e32 v0, s2
-; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    s_cselect_b32 s5, s3, s5
+; VI-NEXT:    s_cmp_lg_u32 s2, 0
+; VI-NEXT:    s_cselect_b32 s2, s5, s3
+; VI-NEXT:    s_lshl_b32 s2, s2, 16
+; VI-NEXT:    s_or_b32 s2, s4, s2
+; VI-NEXT:    v_mov_b32_e32 v0, s6
+; VI-NEXT:    v_mov_b32_e32 v1, s2
 ; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
@@ -2723,15 +2723,15 @@ define amdgpu_kernel void @frem_v2f32(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; CI-NEXT:    v_cmp_nge_f32_e64 s[8:9], |s2|, v0
 ; CI-NEXT:    s_cselect_b32 s4, 0x7fc00000, s6
 ; CI-NEXT:    s_or_b64 s[8:9], s[8:9], s[8:9]
-; CI-NEXT:    s_cselect_b32 s2, s4, 0x7fc00000
+; CI-NEXT:    s_cselect_b32 s6, s4, 0x7fc00000
 ; CI-NEXT:    v_cmp_nlg_f32_e64 s[4:5], s5, 0
 ; CI-NEXT:    s_or_b64 s[4:5], s[4:5], s[4:5]
-; CI-NEXT:    v_cmp_nge_f32_e64 s[4:5], |s3|, v0
-; CI-NEXT:    s_cselect_b32 s6, 0x7fc00000, s7
-; CI-NEXT:    s_or_b64 s[4:5], s[4:5], s[4:5]
-; CI-NEXT:    s_cselect_b32 s3, s6, 0x7fc00000
-; CI-NEXT:    v_mov_b32_e32 v0, s2
-; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_cmp_nge_f32_e64 s[2:3], |s3|, v0
+; CI-NEXT:    s_cselect_b32 s4, 0x7fc00000, s7
+; CI-NEXT:    s_or_b64 s[2:3], s[2:3], s[2:3]
+; CI-NEXT:    s_cselect_b32 s2, s4, 0x7fc00000
+; CI-NEXT:    v_mov_b32_e32 v0, s6
+; CI-NEXT:    v_mov_b32_e32 v1, s2
 ; CI-NEXT:    s_mov_b32 s2, -1
 ; CI-NEXT:    s_mov_b32 s3, 0xf000
 ; CI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
@@ -2933,15 +2933,15 @@ define amdgpu_kernel void @frem_v2f32(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; VI-NEXT:    v_cmp_nge_f32_e64 s[8:9], |s2|, v0
 ; VI-NEXT:    s_cselect_b32 s4, 0x7fc00000, s6
 ; VI-NEXT:    s_cmp_lg_u64 s[8:9], 0
-; VI-NEXT:    s_cselect_b32 s2, s4, 0x7fc00000
+; VI-NEXT:    s_cselect_b32 s6, s4, 0x7fc00000
 ; VI-NEXT:    v_cmp_nlg_f32_e64 s[4:5], s5, 0
 ; VI-NEXT:    s_cmp_lg_u64 s[4:5], 0
-; VI-NEXT:    v_cmp_nge_f32_e64 s[4:5], |s3|, v0
-; VI-NEXT:    s_cselect_b32 s6, 0x7fc00000, s7
-; VI-NEXT:    s_cmp_lg_u64 s[4:5], 0
-; VI-NEXT:    s_cselect_b32 s3, s6, 0x7fc00000
-; VI-NEXT:    v_mov_b32_e32 v0, s2
-; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_cmp_nge_f32_e64 s[2:3], |s3|, v0
+; VI-NEXT:    s_cselect_b32 s4, 0x7fc00000, s7
+; VI-NEXT:    s_cmp_lg_u64 s[2:3], 0
+; VI-NEXT:    s_cselect_b32 s2, s4, 0x7fc00000
+; VI-NEXT:    v_mov_b32_e32 v0, s6
+; VI-NEXT:    v_mov_b32_e32 v1, s2
 ; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
@@ -3351,11 +3351,11 @@ define amdgpu_kernel void @frem_v4f32(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; CI-NEXT:    v_cmp_nge_f32_e64 s[2:3], |s7|, v0
 ; CI-NEXT:    s_cselect_b32 s8, 0x7fc00000, s13
 ; CI-NEXT:    s_or_b64 s[2:3], s[2:3], s[2:3]
-; CI-NEXT:    s_cselect_b32 s7, s8, 0x7fc00000
+; CI-NEXT:    s_cselect_b32 s2, s8, 0x7fc00000
 ; CI-NEXT:    v_mov_b32_e32 v0, s4
 ; CI-NEXT:    v_mov_b32_e32 v1, s5
 ; CI-NEXT:    v_mov_b32_e32 v2, s6
-; CI-NEXT:    v_mov_b32_e32 v3, s7
+; CI-NEXT:    v_mov_b32_e32 v3, s2
 ; CI-NEXT:    s_mov_b32 s2, -1
 ; CI-NEXT:    s_mov_b32 s3, 0xf000
 ; CI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
@@ -3757,11 +3757,11 @@ define amdgpu_kernel void @frem_v4f32(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; VI-NEXT:    v_cmp_nge_f32_e64 s[2:3], |s7|, v0
 ; VI-NEXT:    s_cselect_b32 s8, 0x7fc00000, s13
 ; VI-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; VI-NEXT:    s_cselect_b32 s7, s8, 0x7fc00000
+; VI-NEXT:    s_cselect_b32 s2, s8, 0x7fc00000
 ; VI-NEXT:    v_mov_b32_e32 v0, s4
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    v_mov_b32_e32 v2, s6
-; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    v_mov_b32_e32 v3, s2
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/implicit-kernarg-backend-usage-global-isel.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/implicit-kernarg-backend-usage-global-isel.ll
index 110aee0d27646..155b9dd58f8d4 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/implicit-kernarg-backend-usage-global-isel.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/implicit-kernarg-backend-usage-global-isel.ll
@@ -23,15 +23,15 @@ define amdgpu_kernel void @addrspacecast(ptr addrspace(5) %ptr.private, ptr addr
 ; GFX8V4-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8V4-NEXT:    s_mov_b32 s2, s0
 ; GFX8V4-NEXT:    s_cmp_lg_u32 s0, -1
-; GFX8V4-NEXT:    s_mov_b32 s0, s1
 ; GFX8V4-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8V4-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX8V4-NEXT:    s_cselect_b64 s[2:3], s[2:3], 0
-; GFX8V4-NEXT:    s_add_u32 s4, s6, 64
-; GFX8V4-NEXT:    s_addc_u32 s5, s7, 0
-; GFX8V4-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8V4-NEXT:    v_mov_b32_e32 v1, s5
+; GFX8V4-NEXT:    s_add_u32 s0, s6, 64
+; GFX8V4-NEXT:    s_addc_u32 s4, s7, 0
+; GFX8V4-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8V4-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8V4-NEXT:    flat_load_dword v2, v[0:1]
+; GFX8V4-NEXT:    s_mov_b32 s0, s1
 ; GFX8V4-NEXT:    s_cmp_lg_u32 s1, -1
 ; GFX8V4-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8V4-NEXT:    v_mov_b32_e32 v1, s3
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll
index 933a75340d96b..a134e02dc9b51 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll
@@ -329,23 +329,15 @@ entry:
 define amdgpu_ps <8 x float> @dyn_insertelement_v8f32_s_v_s(<8 x float> inreg %vec, float %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v8f32_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
 ; GPRIDX-NEXT:    v_mov_b32_e32 v8, v0
-; GPRIDX-NEXT:    v_mov_b32_e32 v0, s0
-; GPRIDX-NEXT:    v_mov_b32_e32 v1, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v0, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v1, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s9
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s10, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v0, v8
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -353,65 +345,41 @@ define amdgpu_ps <8 x float> @dyn_insertelement_v8f32_s_v_s(<8 x float> inreg %v
 ;
 ; GFX10-LABEL: dyn_insertelement_v8f32_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    v_mov_b32_e32 v8, v0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    s_mov_b32 m0, s10
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s7
+; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
+; GFX10-NEXT:    v_mov_b32_e32 v4, s6
+; GFX10-NEXT:    v_mov_b32_e32 v5, s7
+; GFX10-NEXT:    v_mov_b32_e32 v6, s8
+; GFX10-NEXT:    v_mov_b32_e32 v7, s9
 ; GFX10-NEXT:    v_movreld_b32_e32 v0, v8
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: dyn_insertelement_v8f32_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    v_dual_mov_b32 v8, v0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_dual_mov_b32 v8, v0 :: v_dual_mov_b32 v1, s3
+; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v3, s5
 ; GFX11-NEXT:    s_mov_b32 m0, s10
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v5, s5
-; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v7, s7
-; GFX11-NEXT:    v_mov_b32_e32 v6, s6
+; GFX11-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v5, s7
+; GFX11-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v7, s9
+; GFX11-NEXT:    v_mov_b32_e32 v6, s8
 ; GFX11-NEXT:    v_movreld_b32_e32 v0, v8
 ; GFX11-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: dyn_insertelement_v8f32_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
 ; GFX8-NEXT:    v_mov_b32_e32 v8, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s7
+; GFX8-NEXT:    v_mov_b32_e32 v0, s2
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NEXT:    v_mov_b32_e32 v6, s8
+; GFX8-NEXT:    v_mov_b32_e32 v7, s9
 ; GFX8-NEXT:    s_mov_b32 m0, s10
 ; GFX8-NEXT:    v_movreld_b32_e32 v0, v8
 ; GFX8-NEXT:    ; return to shader part epilog
@@ -1134,38 +1102,18 @@ entry:
 define amdgpu_ps void @dyn_insertelement_v8f64_s_s_v(<8 x double> inreg %vec, double inreg %val, i32 %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v8f64_s_s_v:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
-; GPRIDX-NEXT:    s_mov_b32 s12, s14
-; GPRIDX-NEXT:    s_mov_b32 s13, s15
-; GPRIDX-NEXT:    s_mov_b32 s14, s16
-; GPRIDX-NEXT:    s_mov_b32 s15, s17
-; GPRIDX-NEXT:    v_mov_b32_e32 v16, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v15, s14
-; GPRIDX-NEXT:    v_mov_b32_e32 v14, s13
-; GPRIDX-NEXT:    v_mov_b32_e32 v13, s12
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v1, s0
+; GPRIDX-NEXT:    v_mov_b32_e32 v1, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s13
 ; GPRIDX-NEXT:    v_mov_b32_e32 v17, s18
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v0
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e64 s[0:1], 2, v0
@@ -1176,6 +1124,10 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_s_v(<8 x double> inreg %vec, do
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e64 s[10:11], 7, v0
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e64 s[12:13], 0, v0
 ; GPRIDX-NEXT:    v_mov_b32_e32 v0, s19
+; GPRIDX-NEXT:    v_mov_b32_e32 v13, s14
+; GPRIDX-NEXT:    v_mov_b32_e32 v14, s15
+; GPRIDX-NEXT:    v_mov_b32_e32 v15, s16
+; GPRIDX-NEXT:    v_mov_b32_e32 v16, s17
 ; GPRIDX-NEXT:    v_cndmask_b32_e64 v1, v1, v17, s[12:13]
 ; GPRIDX-NEXT:    v_cndmask_b32_e32 v3, v3, v17, vcc
 ; GPRIDX-NEXT:    v_cndmask_b32_e64 v2, v2, v0, s[12:13]
@@ -1204,62 +1156,46 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_s_v(<8 x double> inreg %vec, do
 ;
 ; GFX10-LABEL: dyn_insertelement_v8f64_s_s_v:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    s_mov_b32 s12, s14
-; GFX10-NEXT:    s_mov_b32 s13, s15
-; GFX10-NEXT:    s_mov_b32 s14, s16
-; GFX10-NEXT:    s_mov_b32 s15, s17
-; GFX10-NEXT:    v_mov_b32_e32 v2, s1
-; GFX10-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s2
+; GFX10-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10-NEXT:    v_mov_b32_e32 v16, s15
-; GFX10-NEXT:    v_mov_b32_e32 v15, s14
-; GFX10-NEXT:    v_mov_b32_e32 v14, s13
-; GFX10-NEXT:    v_mov_b32_e32 v13, s12
-; GFX10-NEXT:    v_mov_b32_e32 v12, s11
-; GFX10-NEXT:    v_mov_b32_e32 v11, s10
-; GFX10-NEXT:    v_mov_b32_e32 v10, s9
-; GFX10-NEXT:    v_mov_b32_e32 v9, s8
-; GFX10-NEXT:    v_mov_b32_e32 v8, s7
-; GFX10-NEXT:    v_mov_b32_e32 v7, s6
-; GFX10-NEXT:    v_mov_b32_e32 v6, s5
-; GFX10-NEXT:    v_mov_b32_e32 v5, s4
-; GFX10-NEXT:    v_mov_b32_e32 v4, s3
-; GFX10-NEXT:    v_mov_b32_e32 v3, s2
+; GFX10-NEXT:    v_mov_b32_e32 v3, s4
+; GFX10-NEXT:    v_mov_b32_e32 v4, s5
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 1, v0
+; GFX10-NEXT:    v_mov_b32_e32 v5, s6
+; GFX10-NEXT:    v_mov_b32_e32 v6, s7
 ; GFX10-NEXT:    v_cndmask_b32_e64 v1, v1, s18, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e64 v2, v2, s19, vcc_lo
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 2, v0
-; GFX10-NEXT:    v_cmp_eq_u32_e64 s1, 7, v0
+; GFX10-NEXT:    v_mov_b32_e32 v7, s8
+; GFX10-NEXT:    v_mov_b32_e32 v8, s9
 ; GFX10-NEXT:    v_cndmask_b32_e64 v3, v3, s18, s0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v4, v4, s19, s0
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 3, v0
+; GFX10-NEXT:    v_mov_b32_e32 v9, s10
+; GFX10-NEXT:    v_mov_b32_e32 v10, s11
 ; GFX10-NEXT:    v_cndmask_b32_e64 v5, v5, s18, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e64 v6, v6, s19, vcc_lo
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 4, v0
-; GFX10-NEXT:    v_cndmask_b32_e64 v15, v15, s18, s1
+; GFX10-NEXT:    v_mov_b32_e32 v11, s12
+; GFX10-NEXT:    v_mov_b32_e32 v12, s13
 ; GFX10-NEXT:    v_cndmask_b32_e64 v7, v7, s18, s0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v8, v8, s19, s0
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 5, v0
+; GFX10-NEXT:    v_mov_b32_e32 v13, s14
+; GFX10-NEXT:    v_mov_b32_e32 v14, s15
+; GFX10-NEXT:    v_mov_b32_e32 v15, s16
+; GFX10-NEXT:    v_mov_b32_e32 v16, s17
 ; GFX10-NEXT:    v_cndmask_b32_e64 v9, v9, s18, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e64 v10, v10, s19, vcc_lo
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 6, v0
-; GFX10-NEXT:    v_cndmask_b32_e64 v16, v16, s19, s1
+; GFX10-NEXT:    v_cmp_eq_u32_e64 s1, 7, v0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v11, v11, s18, s0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v12, v12, s19, s0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v13, v13, s18, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e64 v14, v14, s19, vcc_lo
+; GFX10-NEXT:    v_cndmask_b32_e64 v15, v15, s18, s1
+; GFX10-NEXT:    v_cndmask_b32_e64 v16, v16, s19, s1
 ; GFX10-NEXT:    global_store_dwordx4 v[0:1], v[1:4], off
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_store_dwordx4 v[0:1], v[5:8], off
@@ -1272,54 +1208,38 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_s_v(<8 x double> inreg %vec, do
 ;
 ; GFX11-LABEL: dyn_insertelement_v8f64_s_s_v:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    s_mov_b32 s12, s14
-; GFX11-NEXT:    s_mov_b32 s13, s15
-; GFX11-NEXT:    s_mov_b32 s14, s16
-; GFX11-NEXT:    s_mov_b32 s15, s17
-; GFX11-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX11-NEXT:    v_dual_mov_b32 v1, s2 :: v_dual_mov_b32 v2, s3
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-NEXT:    v_dual_mov_b32 v16, s15 :: v_dual_mov_b32 v15, s14
-; GFX11-NEXT:    v_dual_mov_b32 v14, s13 :: v_dual_mov_b32 v13, s12
-; GFX11-NEXT:    v_dual_mov_b32 v12, s11 :: v_dual_mov_b32 v11, s10
-; GFX11-NEXT:    v_dual_mov_b32 v10, s9 :: v_dual_mov_b32 v9, s8
-; GFX11-NEXT:    v_dual_mov_b32 v8, s7 :: v_dual_mov_b32 v7, s6
-; GFX11-NEXT:    v_dual_mov_b32 v6, s5 :: v_dual_mov_b32 v5, s4
-; GFX11-NEXT:    v_dual_mov_b32 v4, s3 :: v_dual_mov_b32 v3, s2
+; GFX11-NEXT:    v_dual_mov_b32 v3, s4 :: v_dual_mov_b32 v4, s5
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 1, v0
+; GFX11-NEXT:    v_dual_mov_b32 v5, s6 :: v_dual_mov_b32 v6, s7
 ; GFX11-NEXT:    v_cndmask_b32_e64 v1, v1, s18, vcc_lo
 ; GFX11-NEXT:    v_cndmask_b32_e64 v2, v2, s19, vcc_lo
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 2, v0
-; GFX11-NEXT:    v_cmp_eq_u32_e64 s1, 7, v0
+; GFX11-NEXT:    v_dual_mov_b32 v7, s8 :: v_dual_mov_b32 v8, s9
 ; GFX11-NEXT:    v_cndmask_b32_e64 v3, v3, s18, s0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v4, v4, s19, s0
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 3, v0
+; GFX11-NEXT:    v_dual_mov_b32 v9, s10 :: v_dual_mov_b32 v10, s11
 ; GFX11-NEXT:    v_cndmask_b32_e64 v5, v5, s18, vcc_lo
 ; GFX11-NEXT:    v_cndmask_b32_e64 v6, v6, s19, vcc_lo
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 4, v0
-; GFX11-NEXT:    v_cndmask_b32_e64 v15, v15, s18, s1
+; GFX11-NEXT:    v_dual_mov_b32 v11, s12 :: v_dual_mov_b32 v12, s13
 ; GFX11-NEXT:    v_cndmask_b32_e64 v7, v7, s18, s0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v8, v8, s19, s0
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 5, v0
+; GFX11-NEXT:    v_dual_mov_b32 v13, s14 :: v_dual_mov_b32 v14, s15
+; GFX11-NEXT:    v_dual_mov_b32 v15, s16 :: v_dual_mov_b32 v16, s17
 ; GFX11-NEXT:    v_cndmask_b32_e64 v9, v9, s18, vcc_lo
 ; GFX11-NEXT:    v_cndmask_b32_e64 v10, v10, s19, vcc_lo
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 6, v0
-; GFX11-NEXT:    v_cndmask_b32_e64 v16, v16, s19, s1
+; GFX11-NEXT:    v_cmp_eq_u32_e64 s1, 7, v0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v11, v11, s18, s0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v12, v12, s19, s0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v13, v13, s18, vcc_lo
 ; GFX11-NEXT:    v_cndmask_b32_e64 v14, v14, s19, vcc_lo
+; GFX11-NEXT:    v_cndmask_b32_e64 v15, v15, s18, s1
+; GFX11-NEXT:    v_cndmask_b32_e64 v16, v16, s19, s1
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[1:4], off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[5:8], off dlc
@@ -1332,38 +1252,18 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_s_v(<8 x double> inreg %vec, do
 ;
 ; GFX8-LABEL: dyn_insertelement_v8f64_s_s_v:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
-; GFX8-NEXT:    s_mov_b32 s12, s14
-; GFX8-NEXT:    s_mov_b32 s13, s15
-; GFX8-NEXT:    s_mov_b32 s14, s16
-; GFX8-NEXT:    s_mov_b32 s15, s17
-; GFX8-NEXT:    v_mov_b32_e32 v16, s15
-; GFX8-NEXT:    v_mov_b32_e32 v15, s14
-; GFX8-NEXT:    v_mov_b32_e32 v14, s13
-; GFX8-NEXT:    v_mov_b32_e32 v13, s12
-; GFX8-NEXT:    v_mov_b32_e32 v12, s11
-; GFX8-NEXT:    v_mov_b32_e32 v11, s10
-; GFX8-NEXT:    v_mov_b32_e32 v10, s9
-; GFX8-NEXT:    v_mov_b32_e32 v9, s8
-; GFX8-NEXT:    v_mov_b32_e32 v8, s7
-; GFX8-NEXT:    v_mov_b32_e32 v7, s6
-; GFX8-NEXT:    v_mov_b32_e32 v6, s5
-; GFX8-NEXT:    v_mov_b32_e32 v5, s4
-; GFX8-NEXT:    v_mov_b32_e32 v4, s3
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
-; GFX8-NEXT:    v_mov_b32_e32 v2, s1
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    v_mov_b32_e32 v2, s3
+; GFX8-NEXT:    v_mov_b32_e32 v3, s4
+; GFX8-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NEXT:    v_mov_b32_e32 v5, s6
+; GFX8-NEXT:    v_mov_b32_e32 v6, s7
+; GFX8-NEXT:    v_mov_b32_e32 v7, s8
+; GFX8-NEXT:    v_mov_b32_e32 v8, s9
+; GFX8-NEXT:    v_mov_b32_e32 v9, s10
+; GFX8-NEXT:    v_mov_b32_e32 v10, s11
+; GFX8-NEXT:    v_mov_b32_e32 v11, s12
+; GFX8-NEXT:    v_mov_b32_e32 v12, s13
 ; GFX8-NEXT:    v_mov_b32_e32 v17, s18
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e64 s[0:1], 2, v0
@@ -1374,6 +1274,10 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_s_v(<8 x double> inreg %vec, do
 ; GFX8-NEXT:    v_cmp_eq_u32_e64 s[10:11], 7, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e64 s[12:13], 0, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s19
+; GFX8-NEXT:    v_mov_b32_e32 v13, s14
+; GFX8-NEXT:    v_mov_b32_e32 v14, s15
+; GFX8-NEXT:    v_mov_b32_e32 v15, s16
+; GFX8-NEXT:    v_mov_b32_e32 v16, s17
 ; GFX8-NEXT:    v_cndmask_b32_e64 v1, v1, v17, s[12:13]
 ; GFX8-NEXT:    v_cndmask_b32_e32 v3, v3, v17, vcc
 ; GFX8-NEXT:    v_cndmask_b32_e64 v2, v2, v0, s[12:13]
@@ -1415,46 +1319,30 @@ entry:
 define amdgpu_ps void @dyn_insertelement_v8f64_s_v_s(<8 x double> inreg %vec, double %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v8f64_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
-; GPRIDX-NEXT:    s_mov_b32 s12, s14
-; GPRIDX-NEXT:    s_mov_b32 s13, s15
-; GPRIDX-NEXT:    s_mov_b32 s14, s16
-; GPRIDX-NEXT:    s_mov_b32 s15, s17
-; GPRIDX-NEXT:    v_mov_b32_e32 v17, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v16, s14
-; GPRIDX-NEXT:    v_mov_b32_e32 v15, s13
-; GPRIDX-NEXT:    v_mov_b32_e32 v14, s12
-; GPRIDX-NEXT:    v_mov_b32_e32 v13, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s0
-; GPRIDX-NEXT:    s_lshl_b32 s0, s18, 1
-; GPRIDX-NEXT:    s_set_gpr_idx_on s0, gpr_idx(DST)
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, v0
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, v1
-; GPRIDX-NEXT:    s_set_gpr_idx_off
-; GPRIDX-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
-; GPRIDX-NEXT:    s_waitcnt vmcnt(0)
-; GPRIDX-NEXT:    global_store_dwordx4 v[0:1], v[6:9], off
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v13, s13
+; GPRIDX-NEXT:    v_mov_b32_e32 v14, s14
+; GPRIDX-NEXT:    v_mov_b32_e32 v15, s15
+; GPRIDX-NEXT:    v_mov_b32_e32 v16, s16
+; GPRIDX-NEXT:    v_mov_b32_e32 v17, s17
+; GPRIDX-NEXT:    s_lshl_b32 s0, s18, 1
+; GPRIDX-NEXT:    s_set_gpr_idx_on s0, gpr_idx(DST)
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, v0
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, v1
+; GPRIDX-NEXT:    s_set_gpr_idx_off
+; GPRIDX-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GPRIDX-NEXT:    s_waitcnt vmcnt(0)
+; GPRIDX-NEXT:    global_store_dwordx4 v[0:1], v[6:9], off
 ; GPRIDX-NEXT:    s_waitcnt vmcnt(0)
 ; GPRIDX-NEXT:    global_store_dwordx4 v[0:1], v[10:13], off
 ; GPRIDX-NEXT:    s_waitcnt vmcnt(0)
@@ -1464,39 +1352,23 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_v_s(<8 x double> inreg %vec, do
 ;
 ; GFX10-LABEL: dyn_insertelement_v8f64_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    s_mov_b32 s12, s14
-; GFX10-NEXT:    s_mov_b32 s13, s15
-; GFX10-NEXT:    s_mov_b32 s14, s16
-; GFX10-NEXT:    s_mov_b32 s15, s17
-; GFX10-NEXT:    v_mov_b32_e32 v2, s0
+; GFX10-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX10-NEXT:    s_lshl_b32 m0, s18, 1
-; GFX10-NEXT:    v_mov_b32_e32 v17, s15
-; GFX10-NEXT:    v_mov_b32_e32 v16, s14
-; GFX10-NEXT:    v_mov_b32_e32 v15, s13
-; GFX10-NEXT:    v_mov_b32_e32 v14, s12
-; GFX10-NEXT:    v_mov_b32_e32 v13, s11
-; GFX10-NEXT:    v_mov_b32_e32 v12, s10
-; GFX10-NEXT:    v_mov_b32_e32 v11, s9
-; GFX10-NEXT:    v_mov_b32_e32 v10, s8
-; GFX10-NEXT:    v_mov_b32_e32 v9, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
+; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v4, s4
+; GFX10-NEXT:    v_mov_b32_e32 v5, s5
+; GFX10-NEXT:    v_mov_b32_e32 v6, s6
+; GFX10-NEXT:    v_mov_b32_e32 v7, s7
+; GFX10-NEXT:    v_mov_b32_e32 v8, s8
+; GFX10-NEXT:    v_mov_b32_e32 v9, s9
+; GFX10-NEXT:    v_mov_b32_e32 v10, s10
+; GFX10-NEXT:    v_mov_b32_e32 v11, s11
+; GFX10-NEXT:    v_mov_b32_e32 v12, s12
+; GFX10-NEXT:    v_mov_b32_e32 v13, s13
+; GFX10-NEXT:    v_mov_b32_e32 v14, s14
+; GFX10-NEXT:    v_mov_b32_e32 v15, s15
+; GFX10-NEXT:    v_mov_b32_e32 v16, s16
+; GFX10-NEXT:    v_mov_b32_e32 v17, s17
 ; GFX10-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX10-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX10-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
@@ -1511,31 +1383,15 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_v_s(<8 x double> inreg %vec, do
 ;
 ; GFX11-LABEL: dyn_insertelement_v8f64_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    s_mov_b32 s12, s14
-; GFX11-NEXT:    s_mov_b32 s13, s15
-; GFX11-NEXT:    s_mov_b32 s14, s16
-; GFX11-NEXT:    s_mov_b32 s15, s17
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
 ; GFX11-NEXT:    s_lshl_b32 m0, s18, 1
-; GFX11-NEXT:    v_dual_mov_b32 v17, s15 :: v_dual_mov_b32 v16, s14
-; GFX11-NEXT:    v_dual_mov_b32 v15, s13 :: v_dual_mov_b32 v14, s12
-; GFX11-NEXT:    v_dual_mov_b32 v13, s11 :: v_dual_mov_b32 v12, s10
-; GFX11-NEXT:    v_dual_mov_b32 v11, s9 :: v_dual_mov_b32 v10, s8
-; GFX11-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
-; GFX11-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
+; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v5, s5
+; GFX11-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
+; GFX11-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
+; GFX11-NEXT:    v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
+; GFX11-NEXT:    v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
+; GFX11-NEXT:    v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s15
+; GFX11-NEXT:    v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
 ; GFX11-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX11-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[2:5], off dlc
@@ -1550,38 +1406,22 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_v_s(<8 x double> inreg %vec, do
 ;
 ; GFX8-LABEL: dyn_insertelement_v8f64_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
-; GFX8-NEXT:    s_mov_b32 s12, s14
-; GFX8-NEXT:    s_mov_b32 s13, s15
-; GFX8-NEXT:    s_mov_b32 s14, s16
-; GFX8-NEXT:    s_mov_b32 s15, s17
-; GFX8-NEXT:    v_mov_b32_e32 v17, s15
-; GFX8-NEXT:    v_mov_b32_e32 v16, s14
-; GFX8-NEXT:    v_mov_b32_e32 v15, s13
-; GFX8-NEXT:    v_mov_b32_e32 v14, s12
-; GFX8-NEXT:    v_mov_b32_e32 v13, s11
-; GFX8-NEXT:    v_mov_b32_e32 v12, s10
-; GFX8-NEXT:    v_mov_b32_e32 v11, s9
-; GFX8-NEXT:    v_mov_b32_e32 v10, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s3
+; GFX8-NEXT:    v_mov_b32_e32 v4, s4
+; GFX8-NEXT:    v_mov_b32_e32 v5, s5
+; GFX8-NEXT:    v_mov_b32_e32 v6, s6
+; GFX8-NEXT:    v_mov_b32_e32 v7, s7
+; GFX8-NEXT:    v_mov_b32_e32 v8, s8
+; GFX8-NEXT:    v_mov_b32_e32 v9, s9
+; GFX8-NEXT:    v_mov_b32_e32 v10, s10
+; GFX8-NEXT:    v_mov_b32_e32 v11, s11
+; GFX8-NEXT:    v_mov_b32_e32 v12, s12
+; GFX8-NEXT:    v_mov_b32_e32 v13, s13
+; GFX8-NEXT:    v_mov_b32_e32 v14, s14
+; GFX8-NEXT:    v_mov_b32_e32 v15, s15
+; GFX8-NEXT:    v_mov_b32_e32 v16, s16
+; GFX8-NEXT:    v_mov_b32_e32 v17, s17
 ; GFX8-NEXT:    s_lshl_b32 m0, s18, 1
 ; GFX8-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX8-NEXT:    v_movreld_b32_e32 v3, v1
@@ -1691,40 +1531,24 @@ entry:
 define amdgpu_ps void @dyn_insertelement_v8f64_s_v_v(<8 x double> inreg %vec, double %val, i32 %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v8f64_s_v_v:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
-; GPRIDX-NEXT:    s_mov_b32 s12, s14
-; GPRIDX-NEXT:    s_mov_b32 s13, s15
-; GPRIDX-NEXT:    s_mov_b32 s14, s16
-; GPRIDX-NEXT:    s_mov_b32 s15, s17
-; GPRIDX-NEXT:    v_mov_b32_e32 v18, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v17, s14
-; GPRIDX-NEXT:    v_mov_b32_e32 v16, s13
-; GPRIDX-NEXT:    v_mov_b32_e32 v15, s12
-; GPRIDX-NEXT:    v_mov_b32_e32 v14, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v13, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s0
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v13, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v14, s13
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v2
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e64 s[12:13], 0, v2
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v15, s14
+; GPRIDX-NEXT:    v_mov_b32_e32 v16, s15
+; GPRIDX-NEXT:    v_mov_b32_e32 v17, s16
+; GPRIDX-NEXT:    v_mov_b32_e32 v18, s17
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e64 s[0:1], 2, v2
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e64 s[2:3], 3, v2
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e64 s[4:5], 4, v2
@@ -1759,62 +1583,46 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_v_v(<8 x double> inreg %vec, do
 ;
 ; GFX10-LABEL: dyn_insertelement_v8f64_s_v_v:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    s_mov_b32 s12, s14
-; GFX10-NEXT:    s_mov_b32 s13, s15
-; GFX10-NEXT:    s_mov_b32 s14, s16
-; GFX10-NEXT:    s_mov_b32 s15, s17
-; GFX10-NEXT:    v_mov_b32_e32 v4, s1
-; GFX10-NEXT:    v_mov_b32_e32 v3, s0
+; GFX10-NEXT:    v_mov_b32_e32 v3, s2
+; GFX10-NEXT:    v_mov_b32_e32 v4, s3
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX10-NEXT:    v_mov_b32_e32 v18, s15
-; GFX10-NEXT:    v_mov_b32_e32 v17, s14
-; GFX10-NEXT:    v_mov_b32_e32 v16, s13
-; GFX10-NEXT:    v_mov_b32_e32 v15, s12
-; GFX10-NEXT:    v_mov_b32_e32 v14, s11
-; GFX10-NEXT:    v_mov_b32_e32 v13, s10
-; GFX10-NEXT:    v_mov_b32_e32 v12, s9
-; GFX10-NEXT:    v_mov_b32_e32 v11, s8
-; GFX10-NEXT:    v_mov_b32_e32 v10, s7
-; GFX10-NEXT:    v_mov_b32_e32 v9, s6
-; GFX10-NEXT:    v_mov_b32_e32 v8, s5
-; GFX10-NEXT:    v_mov_b32_e32 v7, s4
-; GFX10-NEXT:    v_mov_b32_e32 v6, s3
-; GFX10-NEXT:    v_mov_b32_e32 v5, s2
+; GFX10-NEXT:    v_mov_b32_e32 v5, s4
+; GFX10-NEXT:    v_mov_b32_e32 v6, s5
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 1, v2
+; GFX10-NEXT:    v_mov_b32_e32 v7, s6
+; GFX10-NEXT:    v_mov_b32_e32 v8, s7
 ; GFX10-NEXT:    v_cndmask_b32_e32 v3, v3, v0, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e32 v4, v4, v1, vcc_lo
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 2, v2
-; GFX10-NEXT:    v_cmp_eq_u32_e64 s1, 7, v2
+; GFX10-NEXT:    v_mov_b32_e32 v9, s8
+; GFX10-NEXT:    v_mov_b32_e32 v10, s9
 ; GFX10-NEXT:    v_cndmask_b32_e64 v5, v5, v0, s0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v6, v6, v1, s0
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 3, v2
+; GFX10-NEXT:    v_mov_b32_e32 v11, s10
+; GFX10-NEXT:    v_mov_b32_e32 v12, s11
 ; GFX10-NEXT:    v_cndmask_b32_e32 v7, v7, v0, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e32 v8, v8, v1, vcc_lo
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 4, v2
-; GFX10-NEXT:    v_cndmask_b32_e64 v17, v17, v0, s1
+; GFX10-NEXT:    v_mov_b32_e32 v13, s12
+; GFX10-NEXT:    v_mov_b32_e32 v14, s13
 ; GFX10-NEXT:    v_cndmask_b32_e64 v9, v9, v0, s0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v10, v10, v1, s0
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 5, v2
+; GFX10-NEXT:    v_mov_b32_e32 v15, s14
+; GFX10-NEXT:    v_mov_b32_e32 v16, s15
+; GFX10-NEXT:    v_mov_b32_e32 v17, s16
+; GFX10-NEXT:    v_mov_b32_e32 v18, s17
 ; GFX10-NEXT:    v_cndmask_b32_e32 v11, v11, v0, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e32 v12, v12, v1, vcc_lo
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 6, v2
-; GFX10-NEXT:    v_cndmask_b32_e64 v18, v18, v1, s1
+; GFX10-NEXT:    v_cmp_eq_u32_e64 s1, 7, v2
 ; GFX10-NEXT:    v_cndmask_b32_e64 v13, v13, v0, s0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v14, v14, v1, s0
 ; GFX10-NEXT:    v_cndmask_b32_e32 v15, v15, v0, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e32 v16, v16, v1, vcc_lo
+; GFX10-NEXT:    v_cndmask_b32_e64 v17, v17, v0, s1
+; GFX10-NEXT:    v_cndmask_b32_e64 v18, v18, v1, s1
 ; GFX10-NEXT:    global_store_dwordx4 v[0:1], v[3:6], off
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_store_dwordx4 v[0:1], v[7:10], off
@@ -1827,50 +1635,34 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_v_v(<8 x double> inreg %vec, do
 ;
 ; GFX11-LABEL: dyn_insertelement_v8f64_s_v_v:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    s_mov_b32 s12, s14
-; GFX11-NEXT:    s_mov_b32 s13, s15
-; GFX11-NEXT:    s_mov_b32 s14, s16
-; GFX11-NEXT:    s_mov_b32 s15, s17
-; GFX11-NEXT:    v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v3, s0
+; GFX11-NEXT:    v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX11-NEXT:    v_dual_mov_b32 v18, s15 :: v_dual_mov_b32 v17, s14
-; GFX11-NEXT:    v_dual_mov_b32 v16, s13 :: v_dual_mov_b32 v15, s12
-; GFX11-NEXT:    v_dual_mov_b32 v14, s11 :: v_dual_mov_b32 v13, s10
-; GFX11-NEXT:    v_dual_mov_b32 v12, s9 :: v_dual_mov_b32 v11, s8
-; GFX11-NEXT:    v_dual_mov_b32 v10, s7 :: v_dual_mov_b32 v9, s6
-; GFX11-NEXT:    v_dual_mov_b32 v8, s5 :: v_dual_mov_b32 v7, s4
-; GFX11-NEXT:    v_dual_mov_b32 v6, s3 :: v_dual_mov_b32 v5, s2
+; GFX11-NEXT:    v_dual_mov_b32 v5, s4 :: v_dual_mov_b32 v6, s5
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 1, v2
+; GFX11-NEXT:    v_dual_mov_b32 v7, s6 :: v_dual_mov_b32 v8, s7
 ; GFX11-NEXT:    v_dual_cndmask_b32 v3, v3, v0 :: v_dual_cndmask_b32 v4, v4, v1
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 2, v2
-; GFX11-NEXT:    v_cmp_eq_u32_e64 s1, 7, v2
+; GFX11-NEXT:    v_dual_mov_b32 v9, s8 :: v_dual_mov_b32 v10, s9
 ; GFX11-NEXT:    v_cndmask_b32_e64 v5, v5, v0, s0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v6, v6, v1, s0
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 3, v2
+; GFX11-NEXT:    v_dual_mov_b32 v11, s10 :: v_dual_mov_b32 v12, s11
 ; GFX11-NEXT:    v_dual_cndmask_b32 v7, v7, v0 :: v_dual_cndmask_b32 v8, v8, v1
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 4, v2
-; GFX11-NEXT:    v_cndmask_b32_e64 v17, v17, v0, s1
+; GFX11-NEXT:    v_dual_mov_b32 v13, s12 :: v_dual_mov_b32 v14, s13
 ; GFX11-NEXT:    v_cndmask_b32_e64 v9, v9, v0, s0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v10, v10, v1, s0
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 5, v2
+; GFX11-NEXT:    v_dual_mov_b32 v15, s14 :: v_dual_mov_b32 v16, s15
+; GFX11-NEXT:    v_dual_mov_b32 v17, s16 :: v_dual_mov_b32 v18, s17
 ; GFX11-NEXT:    v_dual_cndmask_b32 v11, v11, v0 :: v_dual_cndmask_b32 v12, v12, v1
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 6, v2
-; GFX11-NEXT:    v_cndmask_b32_e64 v18, v18, v1, s1
+; GFX11-NEXT:    v_cmp_eq_u32_e64 s1, 7, v2
 ; GFX11-NEXT:    v_cndmask_b32_e64 v13, v13, v0, s0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v14, v14, v1, s0
 ; GFX11-NEXT:    v_dual_cndmask_b32 v15, v15, v0 :: v_dual_cndmask_b32 v16, v16, v1
+; GFX11-NEXT:    v_cndmask_b32_e64 v17, v17, v0, s1
+; GFX11-NEXT:    v_cndmask_b32_e64 v18, v18, v1, s1
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[3:6], off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[7:10], off dlc
@@ -1883,40 +1675,24 @@ define amdgpu_ps void @dyn_insertelement_v8f64_s_v_v(<8 x double> inreg %vec, do
 ;
 ; GFX8-LABEL: dyn_insertelement_v8f64_s_v_v:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
-; GFX8-NEXT:    s_mov_b32 s12, s14
-; GFX8-NEXT:    s_mov_b32 s13, s15
-; GFX8-NEXT:    s_mov_b32 s14, s16
-; GFX8-NEXT:    s_mov_b32 s15, s17
-; GFX8-NEXT:    v_mov_b32_e32 v18, s15
-; GFX8-NEXT:    v_mov_b32_e32 v17, s14
-; GFX8-NEXT:    v_mov_b32_e32 v16, s13
-; GFX8-NEXT:    v_mov_b32_e32 v15, s12
-; GFX8-NEXT:    v_mov_b32_e32 v14, s11
-; GFX8-NEXT:    v_mov_b32_e32 v13, s10
-; GFX8-NEXT:    v_mov_b32_e32 v12, s9
-; GFX8-NEXT:    v_mov_b32_e32 v11, s8
-; GFX8-NEXT:    v_mov_b32_e32 v10, s7
-; GFX8-NEXT:    v_mov_b32_e32 v9, s6
-; GFX8-NEXT:    v_mov_b32_e32 v8, s5
-; GFX8-NEXT:    v_mov_b32_e32 v7, s4
-; GFX8-NEXT:    v_mov_b32_e32 v6, s3
-; GFX8-NEXT:    v_mov_b32_e32 v5, s2
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s3
+; GFX8-NEXT:    v_mov_b32_e32 v5, s4
+; GFX8-NEXT:    v_mov_b32_e32 v6, s5
+; GFX8-NEXT:    v_mov_b32_e32 v13, s12
+; GFX8-NEXT:    v_mov_b32_e32 v14, s13
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v2
 ; GFX8-NEXT:    v_cmp_eq_u32_e64 s[12:13], 0, v2
+; GFX8-NEXT:    v_mov_b32_e32 v7, s6
+; GFX8-NEXT:    v_mov_b32_e32 v8, s7
+; GFX8-NEXT:    v_mov_b32_e32 v9, s8
+; GFX8-NEXT:    v_mov_b32_e32 v10, s9
+; GFX8-NEXT:    v_mov_b32_e32 v11, s10
+; GFX8-NEXT:    v_mov_b32_e32 v12, s11
+; GFX8-NEXT:    v_mov_b32_e32 v15, s14
+; GFX8-NEXT:    v_mov_b32_e32 v16, s15
+; GFX8-NEXT:    v_mov_b32_e32 v17, s16
+; GFX8-NEXT:    v_mov_b32_e32 v18, s17
 ; GFX8-NEXT:    v_cmp_eq_u32_e64 s[0:1], 2, v2
 ; GFX8-NEXT:    v_cmp_eq_u32_e64 s[2:3], 3, v2
 ; GFX8-NEXT:    v_cmp_eq_u32_e64 s[4:5], 4, v2
@@ -3278,25 +3054,16 @@ entry:
 define amdgpu_ps <9 x float> @dyn_insertelement_v9f32_s_v_s(<9 x float> inreg %vec, float %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v9f32_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
 ; GPRIDX-NEXT:    v_mov_b32_e32 v9, v0
-; GPRIDX-NEXT:    v_mov_b32_e32 v0, s0
-; GPRIDX-NEXT:    v_mov_b32_e32 v1, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v0, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v1, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s10
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s11, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v0, v9
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -3304,70 +3071,43 @@ define amdgpu_ps <9 x float> @dyn_insertelement_v9f32_s_v_s(<9 x float> inreg %v
 ;
 ; GFX10-LABEL: dyn_insertelement_v9f32_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
 ; GFX10-NEXT:    v_mov_b32_e32 v9, v0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    s_mov_b32 m0, s11
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s8
+; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
+; GFX10-NEXT:    v_mov_b32_e32 v4, s6
+; GFX10-NEXT:    v_mov_b32_e32 v5, s7
+; GFX10-NEXT:    v_mov_b32_e32 v6, s8
+; GFX10-NEXT:    v_mov_b32_e32 v7, s9
+; GFX10-NEXT:    v_mov_b32_e32 v8, s10
 ; GFX10-NEXT:    v_movreld_b32_e32 v0, v9
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: dyn_insertelement_v9f32_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    v_dual_mov_b32 v9, v0 :: v_dual_mov_b32 v0, s0
+; GFX11-NEXT:    v_dual_mov_b32 v9, v0 :: v_dual_mov_b32 v0, s2
 ; GFX11-NEXT:    s_mov_b32 m0, s11
-; GFX11-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX11-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s4
-; GFX11-NEXT:    v_dual_mov_b32 v5, s5 :: v_dual_mov_b32 v6, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s7 :: v_dual_mov_b32 v8, s8
+; GFX11-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX11-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
+; GFX11-NEXT:    v_dual_mov_b32 v5, s7 :: v_dual_mov_b32 v6, s8
+; GFX11-NEXT:    v_dual_mov_b32 v7, s9 :: v_dual_mov_b32 v8, s10
 ; GFX11-NEXT:    v_movreld_b32_e32 v0, v9
 ; GFX11-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: dyn_insertelement_v9f32_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
 ; GFX8-NEXT:    v_mov_b32_e32 v9, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s8
+; GFX8-NEXT:    v_mov_b32_e32 v0, s2
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NEXT:    v_mov_b32_e32 v6, s8
+; GFX8-NEXT:    v_mov_b32_e32 v7, s9
+; GFX8-NEXT:    v_mov_b32_e32 v8, s10
 ; GFX8-NEXT:    s_mov_b32 m0, s11
 ; GFX8-NEXT:    v_movreld_b32_e32 v0, v9
 ; GFX8-NEXT:    ; return to shader part epilog
@@ -3592,27 +3332,17 @@ entry:
 define amdgpu_ps <10 x float> @dyn_insertelement_v10f32_s_v_s(<10 x float> inreg %vec, float %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v10f32_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
 ; GPRIDX-NEXT:    v_mov_b32_e32 v10, v0
-; GPRIDX-NEXT:    v_mov_b32_e32 v0, s0
-; GPRIDX-NEXT:    v_mov_b32_e32 v1, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v0, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v1, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s11
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s12, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v0, v10
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -3620,76 +3350,46 @@ define amdgpu_ps <10 x float> @dyn_insertelement_v10f32_s_v_s(<10 x float> inreg
 ;
 ; GFX10-LABEL: dyn_insertelement_v10f32_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
 ; GFX10-NEXT:    v_mov_b32_e32 v10, v0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    s_mov_b32 m0, s12
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s8
-; GFX10-NEXT:    v_mov_b32_e32 v9, s9
+; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
+; GFX10-NEXT:    v_mov_b32_e32 v4, s6
+; GFX10-NEXT:    v_mov_b32_e32 v5, s7
+; GFX10-NEXT:    v_mov_b32_e32 v6, s8
+; GFX10-NEXT:    v_mov_b32_e32 v7, s9
+; GFX10-NEXT:    v_mov_b32_e32 v8, s10
+; GFX10-NEXT:    v_mov_b32_e32 v9, s11
 ; GFX10-NEXT:    v_movreld_b32_e32 v0, v10
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: dyn_insertelement_v10f32_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    v_dual_mov_b32 v10, v0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_dual_mov_b32 v10, v0 :: v_dual_mov_b32 v1, s3
+; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v3, s5
 ; GFX11-NEXT:    s_mov_b32 m0, s12
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v5, s5
-; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v7, s7
-; GFX11-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v9, s9
-; GFX11-NEXT:    v_mov_b32_e32 v8, s8
+; GFX11-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v5, s7
+; GFX11-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v7, s9
+; GFX11-NEXT:    v_dual_mov_b32 v6, s8 :: v_dual_mov_b32 v9, s11
+; GFX11-NEXT:    v_mov_b32_e32 v8, s10
 ; GFX11-NEXT:    v_movreld_b32_e32 v0, v10
 ; GFX11-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: dyn_insertelement_v10f32_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
 ; GFX8-NEXT:    v_mov_b32_e32 v10, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s9
+; GFX8-NEXT:    v_mov_b32_e32 v0, s2
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NEXT:    v_mov_b32_e32 v6, s8
+; GFX8-NEXT:    v_mov_b32_e32 v7, s9
+; GFX8-NEXT:    v_mov_b32_e32 v8, s10
+; GFX8-NEXT:    v_mov_b32_e32 v9, s11
 ; GFX8-NEXT:    s_mov_b32 m0, s12
 ; GFX8-NEXT:    v_movreld_b32_e32 v0, v10
 ; GFX8-NEXT:    ; return to shader part epilog
@@ -3930,29 +3630,18 @@ entry:
 define amdgpu_ps <11 x float> @dyn_insertelement_v11f32_s_v_s(<11 x float> inreg %vec, float %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v11f32_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
 ; GPRIDX-NEXT:    v_mov_b32_e32 v11, v0
-; GPRIDX-NEXT:    v_mov_b32_e32 v0, s0
-; GPRIDX-NEXT:    v_mov_b32_e32 v1, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v0, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v1, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s12
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s13, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v0, v11
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -3960,81 +3649,48 @@ define amdgpu_ps <11 x float> @dyn_insertelement_v11f32_s_v_s(<11 x float> inreg
 ;
 ; GFX10-LABEL: dyn_insertelement_v11f32_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
 ; GFX10-NEXT:    v_mov_b32_e32 v11, v0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    s_mov_b32 m0, s13
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s8
-; GFX10-NEXT:    v_mov_b32_e32 v9, s9
-; GFX10-NEXT:    v_mov_b32_e32 v10, s10
+; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
+; GFX10-NEXT:    v_mov_b32_e32 v4, s6
+; GFX10-NEXT:    v_mov_b32_e32 v5, s7
+; GFX10-NEXT:    v_mov_b32_e32 v6, s8
+; GFX10-NEXT:    v_mov_b32_e32 v7, s9
+; GFX10-NEXT:    v_mov_b32_e32 v8, s10
+; GFX10-NEXT:    v_mov_b32_e32 v9, s11
+; GFX10-NEXT:    v_mov_b32_e32 v10, s12
 ; GFX10-NEXT:    v_movreld_b32_e32 v0, v11
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: dyn_insertelement_v11f32_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    v_dual_mov_b32 v11, v0 :: v_dual_mov_b32 v0, s0
+; GFX11-NEXT:    v_dual_mov_b32 v11, v0 :: v_dual_mov_b32 v0, s2
 ; GFX11-NEXT:    s_mov_b32 m0, s13
-; GFX11-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX11-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s4
-; GFX11-NEXT:    v_dual_mov_b32 v5, s5 :: v_dual_mov_b32 v6, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s7 :: v_dual_mov_b32 v8, s8
-; GFX11-NEXT:    v_dual_mov_b32 v9, s9 :: v_dual_mov_b32 v10, s10
+; GFX11-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX11-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
+; GFX11-NEXT:    v_dual_mov_b32 v5, s7 :: v_dual_mov_b32 v6, s8
+; GFX11-NEXT:    v_dual_mov_b32 v7, s9 :: v_dual_mov_b32 v8, s10
+; GFX11-NEXT:    v_dual_mov_b32 v9, s11 :: v_dual_mov_b32 v10, s12
 ; GFX11-NEXT:    v_movreld_b32_e32 v0, v11
 ; GFX11-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: dyn_insertelement_v11f32_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
 ; GFX8-NEXT:    v_mov_b32_e32 v11, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s9
-; GFX8-NEXT:    v_mov_b32_e32 v10, s10
+; GFX8-NEXT:    v_mov_b32_e32 v0, s2
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NEXT:    v_mov_b32_e32 v6, s8
+; GFX8-NEXT:    v_mov_b32_e32 v7, s9
+; GFX8-NEXT:    v_mov_b32_e32 v8, s10
+; GFX8-NEXT:    v_mov_b32_e32 v9, s11
+; GFX8-NEXT:    v_mov_b32_e32 v10, s12
 ; GFX8-NEXT:    s_mov_b32 m0, s13
 ; GFX8-NEXT:    v_movreld_b32_e32 v0, v11
 ; GFX8-NEXT:    ; return to shader part epilog
@@ -4291,31 +3947,19 @@ entry:
 define amdgpu_ps <12 x float> @dyn_insertelement_v12f32_s_v_s(<12 x float> inreg %vec, float %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v12f32_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
 ; GPRIDX-NEXT:    v_mov_b32_e32 v12, v0
-; GPRIDX-NEXT:    v_mov_b32_e32 v0, s0
-; GPRIDX-NEXT:    v_mov_b32_e32 v1, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v0, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v1, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s13
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s14, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v0, v12
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -4323,87 +3967,51 @@ define amdgpu_ps <12 x float> @dyn_insertelement_v12f32_s_v_s(<12 x float> inreg
 ;
 ; GFX10-LABEL: dyn_insertelement_v12f32_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
 ; GFX10-NEXT:    v_mov_b32_e32 v12, v0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    s_mov_b32 m0, s14
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s8
-; GFX10-NEXT:    v_mov_b32_e32 v9, s9
-; GFX10-NEXT:    v_mov_b32_e32 v10, s10
-; GFX10-NEXT:    v_mov_b32_e32 v11, s11
+; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
+; GFX10-NEXT:    v_mov_b32_e32 v4, s6
+; GFX10-NEXT:    v_mov_b32_e32 v5, s7
+; GFX10-NEXT:    v_mov_b32_e32 v6, s8
+; GFX10-NEXT:    v_mov_b32_e32 v7, s9
+; GFX10-NEXT:    v_mov_b32_e32 v8, s10
+; GFX10-NEXT:    v_mov_b32_e32 v9, s11
+; GFX10-NEXT:    v_mov_b32_e32 v10, s12
+; GFX10-NEXT:    v_mov_b32_e32 v11, s13
 ; GFX10-NEXT:    v_movreld_b32_e32 v0, v12
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: dyn_insertelement_v12f32_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    v_dual_mov_b32 v12, v0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_dual_mov_b32 v12, v0 :: v_dual_mov_b32 v1, s3
+; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v3, s5
 ; GFX11-NEXT:    s_mov_b32 m0, s14
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v5, s5
-; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v7, s7
-; GFX11-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v9, s9
-; GFX11-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v11, s11
-; GFX11-NEXT:    v_mov_b32_e32 v10, s10
+; GFX11-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v5, s7
+; GFX11-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v7, s9
+; GFX11-NEXT:    v_dual_mov_b32 v6, s8 :: v_dual_mov_b32 v9, s11
+; GFX11-NEXT:    v_dual_mov_b32 v8, s10 :: v_dual_mov_b32 v11, s13
+; GFX11-NEXT:    v_mov_b32_e32 v10, s12
 ; GFX11-NEXT:    v_movreld_b32_e32 v0, v12
 ; GFX11-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: dyn_insertelement_v12f32_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
 ; GFX8-NEXT:    v_mov_b32_e32 v12, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s9
-; GFX8-NEXT:    v_mov_b32_e32 v10, s10
-; GFX8-NEXT:    v_mov_b32_e32 v11, s11
+; GFX8-NEXT:    v_mov_b32_e32 v0, s2
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NEXT:    v_mov_b32_e32 v6, s8
+; GFX8-NEXT:    v_mov_b32_e32 v7, s9
+; GFX8-NEXT:    v_mov_b32_e32 v8, s10
+; GFX8-NEXT:    v_mov_b32_e32 v9, s11
+; GFX8-NEXT:    v_mov_b32_e32 v10, s12
+; GFX8-NEXT:    v_mov_b32_e32 v11, s13
 ; GFX8-NEXT:    s_mov_b32 m0, s14
 ; GFX8-NEXT:    v_movreld_b32_e32 v0, v12
 ; GFX8-NEXT:    ; return to shader part epilog
@@ -5406,38 +5014,22 @@ entry:
 define amdgpu_ps <16 x i32> @dyn_insertelement_v16i32_s_v_s(<16 x i32> inreg %vec, i32 %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v16i32_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
-; GPRIDX-NEXT:    s_mov_b32 s12, s14
-; GPRIDX-NEXT:    s_mov_b32 s13, s15
-; GPRIDX-NEXT:    s_mov_b32 s14, s16
-; GPRIDX-NEXT:    s_mov_b32 s15, s17
-; GPRIDX-NEXT:    v_mov_b32_e32 v16, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v15, s14
-; GPRIDX-NEXT:    v_mov_b32_e32 v14, s13
-; GPRIDX-NEXT:    v_mov_b32_e32 v13, s12
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v1, s0
+; GPRIDX-NEXT:    v_mov_b32_e32 v1, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s13
+; GPRIDX-NEXT:    v_mov_b32_e32 v13, s14
+; GPRIDX-NEXT:    v_mov_b32_e32 v14, s15
+; GPRIDX-NEXT:    v_mov_b32_e32 v15, s16
+; GPRIDX-NEXT:    v_mov_b32_e32 v16, s17
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s18, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v1, v0
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -5461,39 +5053,23 @@ define amdgpu_ps <16 x i32> @dyn_insertelement_v16i32_s_v_s(<16 x i32> inreg %ve
 ;
 ; GFX10-LABEL: dyn_insertelement_v16i32_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    s_mov_b32 s12, s14
-; GFX10-NEXT:    s_mov_b32 s13, s15
-; GFX10-NEXT:    s_mov_b32 s14, s16
-; GFX10-NEXT:    s_mov_b32 s15, s17
-; GFX10-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-NEXT:    s_mov_b32 m0, s18
-; GFX10-NEXT:    v_mov_b32_e32 v16, s15
-; GFX10-NEXT:    v_mov_b32_e32 v15, s14
-; GFX10-NEXT:    v_mov_b32_e32 v14, s13
-; GFX10-NEXT:    v_mov_b32_e32 v13, s12
-; GFX10-NEXT:    v_mov_b32_e32 v12, s11
-; GFX10-NEXT:    v_mov_b32_e32 v11, s10
-; GFX10-NEXT:    v_mov_b32_e32 v10, s9
-; GFX10-NEXT:    v_mov_b32_e32 v9, s8
-; GFX10-NEXT:    v_mov_b32_e32 v8, s7
-; GFX10-NEXT:    v_mov_b32_e32 v7, s6
-; GFX10-NEXT:    v_mov_b32_e32 v6, s5
-; GFX10-NEXT:    v_mov_b32_e32 v5, s4
-; GFX10-NEXT:    v_mov_b32_e32 v4, s3
-; GFX10-NEXT:    v_mov_b32_e32 v3, s2
-; GFX10-NEXT:    v_mov_b32_e32 v2, s1
+; GFX10-NEXT:    v_mov_b32_e32 v2, s3
+; GFX10-NEXT:    v_mov_b32_e32 v3, s4
+; GFX10-NEXT:    v_mov_b32_e32 v4, s5
+; GFX10-NEXT:    v_mov_b32_e32 v5, s6
+; GFX10-NEXT:    v_mov_b32_e32 v6, s7
+; GFX10-NEXT:    v_mov_b32_e32 v7, s8
+; GFX10-NEXT:    v_mov_b32_e32 v8, s9
+; GFX10-NEXT:    v_mov_b32_e32 v9, s10
+; GFX10-NEXT:    v_mov_b32_e32 v10, s11
+; GFX10-NEXT:    v_mov_b32_e32 v11, s12
+; GFX10-NEXT:    v_mov_b32_e32 v12, s13
+; GFX10-NEXT:    v_mov_b32_e32 v13, s14
+; GFX10-NEXT:    v_mov_b32_e32 v14, s15
+; GFX10-NEXT:    v_mov_b32_e32 v15, s16
+; GFX10-NEXT:    v_mov_b32_e32 v16, s17
 ; GFX10-NEXT:    v_movreld_b32_e32 v1, v0
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v1
 ; GFX10-NEXT:    v_readfirstlane_b32 s1, v2
@@ -5515,31 +5091,15 @@ define amdgpu_ps <16 x i32> @dyn_insertelement_v16i32_s_v_s(<16 x i32> inreg %ve
 ;
 ; GFX11-LABEL: dyn_insertelement_v16i32_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    s_mov_b32 s12, s14
-; GFX11-NEXT:    s_mov_b32 s13, s15
-; GFX11-NEXT:    s_mov_b32 s14, s16
-; GFX11-NEXT:    s_mov_b32 s15, s17
-; GFX11-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX11-NEXT:    v_dual_mov_b32 v1, s2 :: v_dual_mov_b32 v2, s3
 ; GFX11-NEXT:    s_mov_b32 m0, s18
-; GFX11-NEXT:    v_dual_mov_b32 v16, s15 :: v_dual_mov_b32 v15, s14
-; GFX11-NEXT:    v_dual_mov_b32 v14, s13 :: v_dual_mov_b32 v13, s12
-; GFX11-NEXT:    v_dual_mov_b32 v12, s11 :: v_dual_mov_b32 v11, s10
-; GFX11-NEXT:    v_dual_mov_b32 v10, s9 :: v_dual_mov_b32 v9, s8
-; GFX11-NEXT:    v_dual_mov_b32 v8, s7 :: v_dual_mov_b32 v7, s6
-; GFX11-NEXT:    v_dual_mov_b32 v6, s5 :: v_dual_mov_b32 v5, s4
-; GFX11-NEXT:    v_dual_mov_b32 v4, s3 :: v_dual_mov_b32 v3, s2
+; GFX11-NEXT:    v_dual_mov_b32 v3, s4 :: v_dual_mov_b32 v4, s5
+; GFX11-NEXT:    v_dual_mov_b32 v5, s6 :: v_dual_mov_b32 v6, s7
+; GFX11-NEXT:    v_dual_mov_b32 v7, s8 :: v_dual_mov_b32 v8, s9
+; GFX11-NEXT:    v_dual_mov_b32 v9, s10 :: v_dual_mov_b32 v10, s11
+; GFX11-NEXT:    v_dual_mov_b32 v11, s12 :: v_dual_mov_b32 v12, s13
+; GFX11-NEXT:    v_dual_mov_b32 v13, s14 :: v_dual_mov_b32 v14, s15
+; GFX11-NEXT:    v_dual_mov_b32 v15, s16 :: v_dual_mov_b32 v16, s17
 ; GFX11-NEXT:    v_movreld_b32_e32 v1, v0
 ; GFX11-NEXT:    v_readfirstlane_b32 s0, v1
 ; GFX11-NEXT:    v_readfirstlane_b32 s1, v2
@@ -5561,38 +5121,22 @@ define amdgpu_ps <16 x i32> @dyn_insertelement_v16i32_s_v_s(<16 x i32> inreg %ve
 ;
 ; GFX8-LABEL: dyn_insertelement_v16i32_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
-; GFX8-NEXT:    s_mov_b32 s12, s14
-; GFX8-NEXT:    s_mov_b32 s13, s15
-; GFX8-NEXT:    s_mov_b32 s14, s16
-; GFX8-NEXT:    s_mov_b32 s15, s17
-; GFX8-NEXT:    v_mov_b32_e32 v16, s15
-; GFX8-NEXT:    v_mov_b32_e32 v15, s14
-; GFX8-NEXT:    v_mov_b32_e32 v14, s13
-; GFX8-NEXT:    v_mov_b32_e32 v13, s12
-; GFX8-NEXT:    v_mov_b32_e32 v12, s11
-; GFX8-NEXT:    v_mov_b32_e32 v11, s10
-; GFX8-NEXT:    v_mov_b32_e32 v10, s9
-; GFX8-NEXT:    v_mov_b32_e32 v9, s8
-; GFX8-NEXT:    v_mov_b32_e32 v8, s7
-; GFX8-NEXT:    v_mov_b32_e32 v7, s6
-; GFX8-NEXT:    v_mov_b32_e32 v6, s5
-; GFX8-NEXT:    v_mov_b32_e32 v5, s4
-; GFX8-NEXT:    v_mov_b32_e32 v4, s3
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
-; GFX8-NEXT:    v_mov_b32_e32 v2, s1
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    v_mov_b32_e32 v2, s3
+; GFX8-NEXT:    v_mov_b32_e32 v3, s4
+; GFX8-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NEXT:    v_mov_b32_e32 v5, s6
+; GFX8-NEXT:    v_mov_b32_e32 v6, s7
+; GFX8-NEXT:    v_mov_b32_e32 v7, s8
+; GFX8-NEXT:    v_mov_b32_e32 v8, s9
+; GFX8-NEXT:    v_mov_b32_e32 v9, s10
+; GFX8-NEXT:    v_mov_b32_e32 v10, s11
+; GFX8-NEXT:    v_mov_b32_e32 v11, s12
+; GFX8-NEXT:    v_mov_b32_e32 v12, s13
+; GFX8-NEXT:    v_mov_b32_e32 v13, s14
+; GFX8-NEXT:    v_mov_b32_e32 v14, s15
+; GFX8-NEXT:    v_mov_b32_e32 v15, s16
+; GFX8-NEXT:    v_mov_b32_e32 v16, s17
 ; GFX8-NEXT:    s_mov_b32 m0, s18
 ; GFX8-NEXT:    v_movreld_b32_e32 v1, v0
 ; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
@@ -5620,39 +5164,23 @@ entry:
 define amdgpu_ps <16 x float> @dyn_insertelement_v16f32_s_v_s(<16 x float> inreg %vec, float %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v16f32_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
-; GPRIDX-NEXT:    s_mov_b32 s12, s14
-; GPRIDX-NEXT:    s_mov_b32 s13, s15
-; GPRIDX-NEXT:    s_mov_b32 s14, s16
-; GPRIDX-NEXT:    s_mov_b32 s15, s17
 ; GPRIDX-NEXT:    v_mov_b32_e32 v16, v0
-; GPRIDX-NEXT:    v_mov_b32_e32 v0, s0
-; GPRIDX-NEXT:    v_mov_b32_e32 v1, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s12
-; GPRIDX-NEXT:    v_mov_b32_e32 v13, s13
-; GPRIDX-NEXT:    v_mov_b32_e32 v14, s14
-; GPRIDX-NEXT:    v_mov_b32_e32 v15, s15
+; GPRIDX-NEXT:    v_mov_b32_e32 v0, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v1, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s13
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s14
+; GPRIDX-NEXT:    v_mov_b32_e32 v13, s15
+; GPRIDX-NEXT:    v_mov_b32_e32 v14, s16
+; GPRIDX-NEXT:    v_mov_b32_e32 v15, s17
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s18, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v0, v16
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -5660,109 +5188,61 @@ define amdgpu_ps <16 x float> @dyn_insertelement_v16f32_s_v_s(<16 x float> inreg
 ;
 ; GFX10-LABEL: dyn_insertelement_v16f32_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    s_mov_b32 s12, s14
-; GFX10-NEXT:    s_mov_b32 s13, s15
-; GFX10-NEXT:    s_mov_b32 s14, s16
-; GFX10-NEXT:    s_mov_b32 s15, s17
 ; GFX10-NEXT:    v_mov_b32_e32 v16, v0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    s_mov_b32 m0, s18
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s8
-; GFX10-NEXT:    v_mov_b32_e32 v9, s9
-; GFX10-NEXT:    v_mov_b32_e32 v10, s10
-; GFX10-NEXT:    v_mov_b32_e32 v11, s11
-; GFX10-NEXT:    v_mov_b32_e32 v12, s12
-; GFX10-NEXT:    v_mov_b32_e32 v13, s13
-; GFX10-NEXT:    v_mov_b32_e32 v14, s14
-; GFX10-NEXT:    v_mov_b32_e32 v15, s15
+; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
+; GFX10-NEXT:    v_mov_b32_e32 v4, s6
+; GFX10-NEXT:    v_mov_b32_e32 v5, s7
+; GFX10-NEXT:    v_mov_b32_e32 v6, s8
+; GFX10-NEXT:    v_mov_b32_e32 v7, s9
+; GFX10-NEXT:    v_mov_b32_e32 v8, s10
+; GFX10-NEXT:    v_mov_b32_e32 v9, s11
+; GFX10-NEXT:    v_mov_b32_e32 v10, s12
+; GFX10-NEXT:    v_mov_b32_e32 v11, s13
+; GFX10-NEXT:    v_mov_b32_e32 v12, s14
+; GFX10-NEXT:    v_mov_b32_e32 v13, s15
+; GFX10-NEXT:    v_mov_b32_e32 v14, s16
+; GFX10-NEXT:    v_mov_b32_e32 v15, s17
 ; GFX10-NEXT:    v_movreld_b32_e32 v0, v16
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: dyn_insertelement_v16f32_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    s_mov_b32 s12, s14
-; GFX11-NEXT:    s_mov_b32 s13, s15
-; GFX11-NEXT:    s_mov_b32 s14, s16
-; GFX11-NEXT:    s_mov_b32 s15, s17
-; GFX11-NEXT:    v_dual_mov_b32 v16, v0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_dual_mov_b32 v16, v0 :: v_dual_mov_b32 v1, s3
+; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v3, s5
 ; GFX11-NEXT:    s_mov_b32 m0, s18
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v5, s5
-; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v7, s7
-; GFX11-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v9, s9
-; GFX11-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v11, s11
-; GFX11-NEXT:    v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v13, s13
-; GFX11-NEXT:    v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v15, s15
-; GFX11-NEXT:    v_mov_b32_e32 v14, s14
+; GFX11-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v5, s7
+; GFX11-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v7, s9
+; GFX11-NEXT:    v_dual_mov_b32 v6, s8 :: v_dual_mov_b32 v9, s11
+; GFX11-NEXT:    v_dual_mov_b32 v8, s10 :: v_dual_mov_b32 v11, s13
+; GFX11-NEXT:    v_dual_mov_b32 v10, s12 :: v_dual_mov_b32 v13, s15
+; GFX11-NEXT:    v_dual_mov_b32 v12, s14 :: v_dual_mov_b32 v15, s17
+; GFX11-NEXT:    v_mov_b32_e32 v14, s16
 ; GFX11-NEXT:    v_movreld_b32_e32 v0, v16
 ; GFX11-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: dyn_insertelement_v16f32_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
-; GFX8-NEXT:    s_mov_b32 s12, s14
-; GFX8-NEXT:    s_mov_b32 s13, s15
-; GFX8-NEXT:    s_mov_b32 s14, s16
-; GFX8-NEXT:    s_mov_b32 s15, s17
 ; GFX8-NEXT:    v_mov_b32_e32 v16, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s9
-; GFX8-NEXT:    v_mov_b32_e32 v10, s10
-; GFX8-NEXT:    v_mov_b32_e32 v11, s11
-; GFX8-NEXT:    v_mov_b32_e32 v12, s12
-; GFX8-NEXT:    v_mov_b32_e32 v13, s13
-; GFX8-NEXT:    v_mov_b32_e32 v14, s14
-; GFX8-NEXT:    v_mov_b32_e32 v15, s15
+; GFX8-NEXT:    v_mov_b32_e32 v0, s2
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NEXT:    v_mov_b32_e32 v6, s8
+; GFX8-NEXT:    v_mov_b32_e32 v7, s9
+; GFX8-NEXT:    v_mov_b32_e32 v8, s10
+; GFX8-NEXT:    v_mov_b32_e32 v9, s11
+; GFX8-NEXT:    v_mov_b32_e32 v10, s12
+; GFX8-NEXT:    v_mov_b32_e32 v11, s13
+; GFX8-NEXT:    v_mov_b32_e32 v12, s14
+; GFX8-NEXT:    v_mov_b32_e32 v13, s15
+; GFX8-NEXT:    v_mov_b32_e32 v14, s16
+; GFX8-NEXT:    v_mov_b32_e32 v15, s17
 ; GFX8-NEXT:    s_mov_b32 m0, s18
 ; GFX8-NEXT:    v_movreld_b32_e32 v0, v16
 ; GFX8-NEXT:    ; return to shader part epilog
@@ -5774,71 +5254,39 @@ entry:
 define amdgpu_ps <32 x float> @dyn_insertelement_v32f32_s_v_s(<32 x float> inreg %vec, float %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v32f32_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
-; GPRIDX-NEXT:    s_mov_b32 s12, s14
-; GPRIDX-NEXT:    s_mov_b32 s13, s15
-; GPRIDX-NEXT:    s_mov_b32 s14, s16
-; GPRIDX-NEXT:    s_mov_b32 s15, s17
-; GPRIDX-NEXT:    s_mov_b32 s16, s18
-; GPRIDX-NEXT:    s_mov_b32 s17, s19
-; GPRIDX-NEXT:    s_mov_b32 s18, s20
-; GPRIDX-NEXT:    s_mov_b32 s19, s21
-; GPRIDX-NEXT:    s_mov_b32 s20, s22
-; GPRIDX-NEXT:    s_mov_b32 s21, s23
-; GPRIDX-NEXT:    s_mov_b32 s22, s24
-; GPRIDX-NEXT:    s_mov_b32 s23, s25
-; GPRIDX-NEXT:    s_mov_b32 s24, s26
-; GPRIDX-NEXT:    s_mov_b32 s25, s27
-; GPRIDX-NEXT:    s_mov_b32 s26, s28
-; GPRIDX-NEXT:    s_mov_b32 s27, s29
-; GPRIDX-NEXT:    s_mov_b32 s28, s30
-; GPRIDX-NEXT:    s_mov_b32 s29, s31
-; GPRIDX-NEXT:    s_mov_b32 s31, s33
-; GPRIDX-NEXT:    s_mov_b32 s30, s32
 ; GPRIDX-NEXT:    v_mov_b32_e32 v32, v0
-; GPRIDX-NEXT:    v_mov_b32_e32 v0, s0
-; GPRIDX-NEXT:    v_mov_b32_e32 v1, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s12
-; GPRIDX-NEXT:    v_mov_b32_e32 v13, s13
-; GPRIDX-NEXT:    v_mov_b32_e32 v14, s14
-; GPRIDX-NEXT:    v_mov_b32_e32 v15, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v16, s16
-; GPRIDX-NEXT:    v_mov_b32_e32 v17, s17
-; GPRIDX-NEXT:    v_mov_b32_e32 v18, s18
-; GPRIDX-NEXT:    v_mov_b32_e32 v19, s19
-; GPRIDX-NEXT:    v_mov_b32_e32 v20, s20
-; GPRIDX-NEXT:    v_mov_b32_e32 v21, s21
-; GPRIDX-NEXT:    v_mov_b32_e32 v22, s22
-; GPRIDX-NEXT:    v_mov_b32_e32 v23, s23
-; GPRIDX-NEXT:    v_mov_b32_e32 v24, s24
-; GPRIDX-NEXT:    v_mov_b32_e32 v25, s25
-; GPRIDX-NEXT:    v_mov_b32_e32 v26, s26
-; GPRIDX-NEXT:    v_mov_b32_e32 v27, s27
-; GPRIDX-NEXT:    v_mov_b32_e32 v28, s28
-; GPRIDX-NEXT:    v_mov_b32_e32 v29, s29
-; GPRIDX-NEXT:    v_mov_b32_e32 v30, s30
-; GPRIDX-NEXT:    v_mov_b32_e32 v31, s31
+; GPRIDX-NEXT:    v_mov_b32_e32 v0, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v1, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s13
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s14
+; GPRIDX-NEXT:    v_mov_b32_e32 v13, s15
+; GPRIDX-NEXT:    v_mov_b32_e32 v14, s16
+; GPRIDX-NEXT:    v_mov_b32_e32 v15, s17
+; GPRIDX-NEXT:    v_mov_b32_e32 v16, s18
+; GPRIDX-NEXT:    v_mov_b32_e32 v17, s19
+; GPRIDX-NEXT:    v_mov_b32_e32 v18, s20
+; GPRIDX-NEXT:    v_mov_b32_e32 v19, s21
+; GPRIDX-NEXT:    v_mov_b32_e32 v20, s22
+; GPRIDX-NEXT:    v_mov_b32_e32 v21, s23
+; GPRIDX-NEXT:    v_mov_b32_e32 v22, s24
+; GPRIDX-NEXT:    v_mov_b32_e32 v23, s25
+; GPRIDX-NEXT:    v_mov_b32_e32 v24, s26
+; GPRIDX-NEXT:    v_mov_b32_e32 v25, s27
+; GPRIDX-NEXT:    v_mov_b32_e32 v26, s28
+; GPRIDX-NEXT:    v_mov_b32_e32 v27, s29
+; GPRIDX-NEXT:    v_mov_b32_e32 v28, s30
+; GPRIDX-NEXT:    v_mov_b32_e32 v29, s31
+; GPRIDX-NEXT:    v_mov_b32_e32 v31, s33
+; GPRIDX-NEXT:    v_mov_b32_e32 v30, s32
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s34, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v0, v32
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -5846,197 +5294,101 @@ define amdgpu_ps <32 x float> @dyn_insertelement_v32f32_s_v_s(<32 x float> inreg
 ;
 ; GFX10-LABEL: dyn_insertelement_v32f32_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    s_mov_b32 s12, s14
-; GFX10-NEXT:    s_mov_b32 s13, s15
-; GFX10-NEXT:    s_mov_b32 s14, s16
-; GFX10-NEXT:    s_mov_b32 s15, s17
-; GFX10-NEXT:    s_mov_b32 s16, s18
-; GFX10-NEXT:    s_mov_b32 s17, s19
-; GFX10-NEXT:    s_mov_b32 s18, s20
-; GFX10-NEXT:    s_mov_b32 s19, s21
-; GFX10-NEXT:    s_mov_b32 s20, s22
-; GFX10-NEXT:    s_mov_b32 s21, s23
-; GFX10-NEXT:    s_mov_b32 s22, s24
-; GFX10-NEXT:    s_mov_b32 s23, s25
-; GFX10-NEXT:    s_mov_b32 s24, s26
-; GFX10-NEXT:    s_mov_b32 s25, s27
-; GFX10-NEXT:    s_mov_b32 s26, s28
-; GFX10-NEXT:    s_mov_b32 s27, s29
-; GFX10-NEXT:    s_mov_b32 s28, s30
-; GFX10-NEXT:    s_mov_b32 s29, s31
-; GFX10-NEXT:    s_mov_b32 s31, s33
-; GFX10-NEXT:    s_mov_b32 s30, s32
 ; GFX10-NEXT:    v_mov_b32_e32 v32, v0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    s_mov_b32 m0, s34
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s8
-; GFX10-NEXT:    v_mov_b32_e32 v9, s9
-; GFX10-NEXT:    v_mov_b32_e32 v10, s10
-; GFX10-NEXT:    v_mov_b32_e32 v11, s11
-; GFX10-NEXT:    v_mov_b32_e32 v12, s12
-; GFX10-NEXT:    v_mov_b32_e32 v13, s13
-; GFX10-NEXT:    v_mov_b32_e32 v14, s14
-; GFX10-NEXT:    v_mov_b32_e32 v15, s15
-; GFX10-NEXT:    v_mov_b32_e32 v16, s16
-; GFX10-NEXT:    v_mov_b32_e32 v17, s17
-; GFX10-NEXT:    v_mov_b32_e32 v18, s18
-; GFX10-NEXT:    v_mov_b32_e32 v19, s19
-; GFX10-NEXT:    v_mov_b32_e32 v20, s20
-; GFX10-NEXT:    v_mov_b32_e32 v21, s21
-; GFX10-NEXT:    v_mov_b32_e32 v22, s22
-; GFX10-NEXT:    v_mov_b32_e32 v23, s23
-; GFX10-NEXT:    v_mov_b32_e32 v24, s24
-; GFX10-NEXT:    v_mov_b32_e32 v25, s25
-; GFX10-NEXT:    v_mov_b32_e32 v26, s26
-; GFX10-NEXT:    v_mov_b32_e32 v27, s27
-; GFX10-NEXT:    v_mov_b32_e32 v28, s28
-; GFX10-NEXT:    v_mov_b32_e32 v29, s29
-; GFX10-NEXT:    v_mov_b32_e32 v30, s30
-; GFX10-NEXT:    v_mov_b32_e32 v31, s31
+; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
+; GFX10-NEXT:    v_mov_b32_e32 v4, s6
+; GFX10-NEXT:    v_mov_b32_e32 v5, s7
+; GFX10-NEXT:    v_mov_b32_e32 v6, s8
+; GFX10-NEXT:    v_mov_b32_e32 v7, s9
+; GFX10-NEXT:    v_mov_b32_e32 v8, s10
+; GFX10-NEXT:    v_mov_b32_e32 v9, s11
+; GFX10-NEXT:    v_mov_b32_e32 v10, s12
+; GFX10-NEXT:    v_mov_b32_e32 v11, s13
+; GFX10-NEXT:    v_mov_b32_e32 v12, s14
+; GFX10-NEXT:    v_mov_b32_e32 v13, s15
+; GFX10-NEXT:    v_mov_b32_e32 v14, s16
+; GFX10-NEXT:    v_mov_b32_e32 v15, s17
+; GFX10-NEXT:    v_mov_b32_e32 v16, s18
+; GFX10-NEXT:    v_mov_b32_e32 v17, s19
+; GFX10-NEXT:    v_mov_b32_e32 v18, s20
+; GFX10-NEXT:    v_mov_b32_e32 v19, s21
+; GFX10-NEXT:    v_mov_b32_e32 v20, s22
+; GFX10-NEXT:    v_mov_b32_e32 v21, s23
+; GFX10-NEXT:    v_mov_b32_e32 v22, s24
+; GFX10-NEXT:    v_mov_b32_e32 v23, s25
+; GFX10-NEXT:    v_mov_b32_e32 v24, s26
+; GFX10-NEXT:    v_mov_b32_e32 v25, s27
+; GFX10-NEXT:    v_mov_b32_e32 v26, s28
+; GFX10-NEXT:    v_mov_b32_e32 v27, s29
+; GFX10-NEXT:    v_mov_b32_e32 v28, s30
+; GFX10-NEXT:    v_mov_b32_e32 v29, s31
+; GFX10-NEXT:    v_mov_b32_e32 v31, s33
+; GFX10-NEXT:    v_mov_b32_e32 v30, s32
 ; GFX10-NEXT:    v_movreld_b32_e32 v0, v32
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: dyn_insertelement_v32f32_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    s_mov_b32 s12, s14
-; GFX11-NEXT:    s_mov_b32 s13, s15
-; GFX11-NEXT:    s_mov_b32 s14, s16
-; GFX11-NEXT:    s_mov_b32 s15, s17
-; GFX11-NEXT:    s_mov_b32 s16, s18
-; GFX11-NEXT:    s_mov_b32 s17, s19
-; GFX11-NEXT:    s_mov_b32 s18, s20
-; GFX11-NEXT:    s_mov_b32 s19, s21
-; GFX11-NEXT:    s_mov_b32 s20, s22
-; GFX11-NEXT:    s_mov_b32 s21, s23
-; GFX11-NEXT:    s_mov_b32 s22, s24
-; GFX11-NEXT:    s_mov_b32 s23, s25
-; GFX11-NEXT:    s_mov_b32 s24, s26
-; GFX11-NEXT:    s_mov_b32 s25, s27
-; GFX11-NEXT:    s_mov_b32 s26, s28
-; GFX11-NEXT:    s_mov_b32 s27, s29
-; GFX11-NEXT:    s_mov_b32 s28, s30
-; GFX11-NEXT:    s_mov_b32 s29, s31
-; GFX11-NEXT:    s_mov_b32 s31, s33
-; GFX11-NEXT:    s_mov_b32 s30, s32
-; GFX11-NEXT:    v_dual_mov_b32 v32, v0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_dual_mov_b32 v32, v0 :: v_dual_mov_b32 v1, s3
+; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v3, s5
 ; GFX11-NEXT:    s_mov_b32 m0, s34
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v5, s5
-; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v7, s7
-; GFX11-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v9, s9
-; GFX11-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v11, s11
-; GFX11-NEXT:    v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v13, s13
-; GFX11-NEXT:    v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v15, s15
-; GFX11-NEXT:    v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v17, s17
-; GFX11-NEXT:    v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v19, s19
-; GFX11-NEXT:    v_dual_mov_b32 v18, s18 :: v_dual_mov_b32 v21, s21
-; GFX11-NEXT:    v_dual_mov_b32 v20, s20 :: v_dual_mov_b32 v23, s23
-; GFX11-NEXT:    v_dual_mov_b32 v22, s22 :: v_dual_mov_b32 v25, s25
-; GFX11-NEXT:    v_dual_mov_b32 v24, s24 :: v_dual_mov_b32 v27, s27
-; GFX11-NEXT:    v_dual_mov_b32 v26, s26 :: v_dual_mov_b32 v29, s29
-; GFX11-NEXT:    v_dual_mov_b32 v28, s28 :: v_dual_mov_b32 v31, s31
-; GFX11-NEXT:    v_mov_b32_e32 v30, s30
+; GFX11-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v5, s7
+; GFX11-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v7, s9
+; GFX11-NEXT:    v_dual_mov_b32 v6, s8 :: v_dual_mov_b32 v9, s11
+; GFX11-NEXT:    v_dual_mov_b32 v8, s10 :: v_dual_mov_b32 v11, s13
+; GFX11-NEXT:    v_dual_mov_b32 v10, s12 :: v_dual_mov_b32 v13, s15
+; GFX11-NEXT:    v_dual_mov_b32 v12, s14 :: v_dual_mov_b32 v15, s17
+; GFX11-NEXT:    v_dual_mov_b32 v14, s16 :: v_dual_mov_b32 v17, s19
+; GFX11-NEXT:    v_dual_mov_b32 v16, s18 :: v_dual_mov_b32 v19, s21
+; GFX11-NEXT:    v_dual_mov_b32 v18, s20 :: v_dual_mov_b32 v21, s23
+; GFX11-NEXT:    v_dual_mov_b32 v20, s22 :: v_dual_mov_b32 v23, s25
+; GFX11-NEXT:    v_dual_mov_b32 v22, s24 :: v_dual_mov_b32 v25, s27
+; GFX11-NEXT:    v_dual_mov_b32 v24, s26 :: v_dual_mov_b32 v27, s29
+; GFX11-NEXT:    v_dual_mov_b32 v26, s28 :: v_dual_mov_b32 v29, s31
+; GFX11-NEXT:    v_dual_mov_b32 v28, s30 :: v_dual_mov_b32 v31, s33
+; GFX11-NEXT:    v_mov_b32_e32 v30, s32
 ; GFX11-NEXT:    v_movreld_b32_e32 v0, v32
 ; GFX11-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: dyn_insertelement_v32f32_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
-; GFX8-NEXT:    s_mov_b32 s12, s14
-; GFX8-NEXT:    s_mov_b32 s13, s15
-; GFX8-NEXT:    s_mov_b32 s14, s16
-; GFX8-NEXT:    s_mov_b32 s15, s17
-; GFX8-NEXT:    s_mov_b32 s16, s18
-; GFX8-NEXT:    s_mov_b32 s17, s19
-; GFX8-NEXT:    s_mov_b32 s18, s20
-; GFX8-NEXT:    s_mov_b32 s19, s21
-; GFX8-NEXT:    s_mov_b32 s20, s22
-; GFX8-NEXT:    s_mov_b32 s21, s23
-; GFX8-NEXT:    s_mov_b32 s22, s24
-; GFX8-NEXT:    s_mov_b32 s23, s25
-; GFX8-NEXT:    s_mov_b32 s24, s26
-; GFX8-NEXT:    s_mov_b32 s25, s27
-; GFX8-NEXT:    s_mov_b32 s26, s28
-; GFX8-NEXT:    s_mov_b32 s27, s29
-; GFX8-NEXT:    s_mov_b32 s28, s30
-; GFX8-NEXT:    s_mov_b32 s29, s31
-; GFX8-NEXT:    s_mov_b32 s31, s33
-; GFX8-NEXT:    s_mov_b32 s30, s32
 ; GFX8-NEXT:    v_mov_b32_e32 v32, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s9
-; GFX8-NEXT:    v_mov_b32_e32 v10, s10
-; GFX8-NEXT:    v_mov_b32_e32 v11, s11
-; GFX8-NEXT:    v_mov_b32_e32 v12, s12
-; GFX8-NEXT:    v_mov_b32_e32 v13, s13
-; GFX8-NEXT:    v_mov_b32_e32 v14, s14
-; GFX8-NEXT:    v_mov_b32_e32 v15, s15
-; GFX8-NEXT:    v_mov_b32_e32 v16, s16
-; GFX8-NEXT:    v_mov_b32_e32 v17, s17
-; GFX8-NEXT:    v_mov_b32_e32 v18, s18
-; GFX8-NEXT:    v_mov_b32_e32 v19, s19
-; GFX8-NEXT:    v_mov_b32_e32 v20, s20
-; GFX8-NEXT:    v_mov_b32_e32 v21, s21
-; GFX8-NEXT:    v_mov_b32_e32 v22, s22
-; GFX8-NEXT:    v_mov_b32_e32 v23, s23
-; GFX8-NEXT:    v_mov_b32_e32 v24, s24
-; GFX8-NEXT:    v_mov_b32_e32 v25, s25
-; GFX8-NEXT:    v_mov_b32_e32 v26, s26
-; GFX8-NEXT:    v_mov_b32_e32 v27, s27
-; GFX8-NEXT:    v_mov_b32_e32 v28, s28
-; GFX8-NEXT:    v_mov_b32_e32 v29, s29
-; GFX8-NEXT:    v_mov_b32_e32 v30, s30
-; GFX8-NEXT:    v_mov_b32_e32 v31, s31
+; GFX8-NEXT:    v_mov_b32_e32 v0, s2
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NEXT:    v_mov_b32_e32 v6, s8
+; GFX8-NEXT:    v_mov_b32_e32 v7, s9
+; GFX8-NEXT:    v_mov_b32_e32 v8, s10
+; GFX8-NEXT:    v_mov_b32_e32 v9, s11
+; GFX8-NEXT:    v_mov_b32_e32 v10, s12
+; GFX8-NEXT:    v_mov_b32_e32 v11, s13
+; GFX8-NEXT:    v_mov_b32_e32 v12, s14
+; GFX8-NEXT:    v_mov_b32_e32 v13, s15
+; GFX8-NEXT:    v_mov_b32_e32 v14, s16
+; GFX8-NEXT:    v_mov_b32_e32 v15, s17
+; GFX8-NEXT:    v_mov_b32_e32 v16, s18
+; GFX8-NEXT:    v_mov_b32_e32 v17, s19
+; GFX8-NEXT:    v_mov_b32_e32 v18, s20
+; GFX8-NEXT:    v_mov_b32_e32 v19, s21
+; GFX8-NEXT:    v_mov_b32_e32 v20, s22
+; GFX8-NEXT:    v_mov_b32_e32 v21, s23
+; GFX8-NEXT:    v_mov_b32_e32 v22, s24
+; GFX8-NEXT:    v_mov_b32_e32 v23, s25
+; GFX8-NEXT:    v_mov_b32_e32 v24, s26
+; GFX8-NEXT:    v_mov_b32_e32 v25, s27
+; GFX8-NEXT:    v_mov_b32_e32 v26, s28
+; GFX8-NEXT:    v_mov_b32_e32 v27, s29
+; GFX8-NEXT:    v_mov_b32_e32 v28, s30
+; GFX8-NEXT:    v_mov_b32_e32 v29, s31
+; GFX8-NEXT:    v_mov_b32_e32 v31, s33
+; GFX8-NEXT:    v_mov_b32_e32 v30, s32
 ; GFX8-NEXT:    s_mov_b32 m0, s34
 ; GFX8-NEXT:    v_movreld_b32_e32 v0, v32
 ; GFX8-NEXT:    ; return to shader part epilog
@@ -6048,72 +5400,40 @@ entry:
 define amdgpu_ps <16 x i64> @dyn_insertelement_v16i64_s_v_s(<16 x i64> inreg %vec, i64 %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v16i64_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
-; GPRIDX-NEXT:    s_mov_b32 s12, s14
-; GPRIDX-NEXT:    s_mov_b32 s13, s15
-; GPRIDX-NEXT:    s_mov_b32 s14, s16
-; GPRIDX-NEXT:    s_mov_b32 s15, s17
-; GPRIDX-NEXT:    s_mov_b32 s16, s18
-; GPRIDX-NEXT:    s_mov_b32 s17, s19
-; GPRIDX-NEXT:    s_mov_b32 s18, s20
-; GPRIDX-NEXT:    s_mov_b32 s19, s21
-; GPRIDX-NEXT:    s_mov_b32 s20, s22
-; GPRIDX-NEXT:    s_mov_b32 s21, s23
-; GPRIDX-NEXT:    s_mov_b32 s22, s24
-; GPRIDX-NEXT:    s_mov_b32 s23, s25
-; GPRIDX-NEXT:    s_mov_b32 s24, s26
-; GPRIDX-NEXT:    s_mov_b32 s25, s27
-; GPRIDX-NEXT:    s_mov_b32 s26, s28
-; GPRIDX-NEXT:    s_mov_b32 s27, s29
-; GPRIDX-NEXT:    s_mov_b32 s28, s30
-; GPRIDX-NEXT:    s_mov_b32 s29, s31
-; GPRIDX-NEXT:    s_mov_b32 s31, s33
-; GPRIDX-NEXT:    s_mov_b32 s30, s32
-; GPRIDX-NEXT:    s_lshl_b32 s33, s34, 1
-; GPRIDX-NEXT:    v_mov_b32_e32 v33, s31
-; GPRIDX-NEXT:    v_mov_b32_e32 v32, s30
-; GPRIDX-NEXT:    v_mov_b32_e32 v31, s29
-; GPRIDX-NEXT:    v_mov_b32_e32 v30, s28
-; GPRIDX-NEXT:    v_mov_b32_e32 v29, s27
-; GPRIDX-NEXT:    v_mov_b32_e32 v28, s26
-; GPRIDX-NEXT:    v_mov_b32_e32 v27, s25
-; GPRIDX-NEXT:    v_mov_b32_e32 v26, s24
-; GPRIDX-NEXT:    v_mov_b32_e32 v25, s23
-; GPRIDX-NEXT:    v_mov_b32_e32 v24, s22
-; GPRIDX-NEXT:    v_mov_b32_e32 v23, s21
-; GPRIDX-NEXT:    v_mov_b32_e32 v22, s20
-; GPRIDX-NEXT:    v_mov_b32_e32 v21, s19
-; GPRIDX-NEXT:    v_mov_b32_e32 v20, s18
-; GPRIDX-NEXT:    v_mov_b32_e32 v19, s17
-; GPRIDX-NEXT:    v_mov_b32_e32 v18, s16
-; GPRIDX-NEXT:    v_mov_b32_e32 v17, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v16, s14
-; GPRIDX-NEXT:    v_mov_b32_e32 v15, s13
-; GPRIDX-NEXT:    v_mov_b32_e32 v14, s12
-; GPRIDX-NEXT:    v_mov_b32_e32 v13, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s0
-; GPRIDX-NEXT:    s_set_gpr_idx_on s33, gpr_idx(DST)
+; GPRIDX-NEXT:    s_lshl_b32 s1, s34, 1
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v13, s13
+; GPRIDX-NEXT:    v_mov_b32_e32 v14, s14
+; GPRIDX-NEXT:    v_mov_b32_e32 v15, s15
+; GPRIDX-NEXT:    v_mov_b32_e32 v16, s16
+; GPRIDX-NEXT:    v_mov_b32_e32 v17, s17
+; GPRIDX-NEXT:    v_mov_b32_e32 v18, s18
+; GPRIDX-NEXT:    v_mov_b32_e32 v19, s19
+; GPRIDX-NEXT:    v_mov_b32_e32 v20, s20
+; GPRIDX-NEXT:    v_mov_b32_e32 v21, s21
+; GPRIDX-NEXT:    v_mov_b32_e32 v22, s22
+; GPRIDX-NEXT:    v_mov_b32_e32 v23, s23
+; GPRIDX-NEXT:    v_mov_b32_e32 v24, s24
+; GPRIDX-NEXT:    v_mov_b32_e32 v25, s25
+; GPRIDX-NEXT:    v_mov_b32_e32 v26, s26
+; GPRIDX-NEXT:    v_mov_b32_e32 v27, s27
+; GPRIDX-NEXT:    v_mov_b32_e32 v28, s28
+; GPRIDX-NEXT:    v_mov_b32_e32 v29, s29
+; GPRIDX-NEXT:    v_mov_b32_e32 v30, s30
+; GPRIDX-NEXT:    v_mov_b32_e32 v31, s31
+; GPRIDX-NEXT:    v_mov_b32_e32 v33, s33
+; GPRIDX-NEXT:    v_mov_b32_e32 v32, s32
+; GPRIDX-NEXT:    s_set_gpr_idx_on s1, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v2, v0
 ; GPRIDX-NEXT:    v_mov_b32_e32 v3, v1
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -6153,71 +5473,39 @@ define amdgpu_ps <16 x i64> @dyn_insertelement_v16i64_s_v_s(<16 x i64> inreg %ve
 ;
 ; GFX10-LABEL: dyn_insertelement_v16i64_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    s_mov_b32 s12, s14
-; GFX10-NEXT:    s_mov_b32 s13, s15
-; GFX10-NEXT:    s_mov_b32 s14, s16
-; GFX10-NEXT:    s_mov_b32 s15, s17
-; GFX10-NEXT:    s_mov_b32 s16, s18
-; GFX10-NEXT:    s_mov_b32 s17, s19
-; GFX10-NEXT:    s_mov_b32 s18, s20
-; GFX10-NEXT:    s_mov_b32 s19, s21
-; GFX10-NEXT:    s_mov_b32 s20, s22
-; GFX10-NEXT:    s_mov_b32 s21, s23
-; GFX10-NEXT:    s_mov_b32 s22, s24
-; GFX10-NEXT:    s_mov_b32 s23, s25
-; GFX10-NEXT:    s_mov_b32 s24, s26
-; GFX10-NEXT:    s_mov_b32 s25, s27
-; GFX10-NEXT:    s_mov_b32 s26, s28
-; GFX10-NEXT:    s_mov_b32 s27, s29
-; GFX10-NEXT:    s_mov_b32 s28, s30
-; GFX10-NEXT:    s_mov_b32 s29, s31
-; GFX10-NEXT:    s_mov_b32 s31, s33
-; GFX10-NEXT:    s_mov_b32 s30, s32
-; GFX10-NEXT:    v_mov_b32_e32 v2, s0
+; GFX10-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX10-NEXT:    s_lshl_b32 m0, s34, 1
-; GFX10-NEXT:    v_mov_b32_e32 v33, s31
-; GFX10-NEXT:    v_mov_b32_e32 v32, s30
-; GFX10-NEXT:    v_mov_b32_e32 v31, s29
-; GFX10-NEXT:    v_mov_b32_e32 v30, s28
-; GFX10-NEXT:    v_mov_b32_e32 v29, s27
-; GFX10-NEXT:    v_mov_b32_e32 v28, s26
-; GFX10-NEXT:    v_mov_b32_e32 v27, s25
-; GFX10-NEXT:    v_mov_b32_e32 v26, s24
-; GFX10-NEXT:    v_mov_b32_e32 v25, s23
-; GFX10-NEXT:    v_mov_b32_e32 v24, s22
-; GFX10-NEXT:    v_mov_b32_e32 v23, s21
-; GFX10-NEXT:    v_mov_b32_e32 v22, s20
-; GFX10-NEXT:    v_mov_b32_e32 v21, s19
-; GFX10-NEXT:    v_mov_b32_e32 v20, s18
-; GFX10-NEXT:    v_mov_b32_e32 v19, s17
-; GFX10-NEXT:    v_mov_b32_e32 v18, s16
-; GFX10-NEXT:    v_mov_b32_e32 v17, s15
-; GFX10-NEXT:    v_mov_b32_e32 v16, s14
-; GFX10-NEXT:    v_mov_b32_e32 v15, s13
-; GFX10-NEXT:    v_mov_b32_e32 v14, s12
-; GFX10-NEXT:    v_mov_b32_e32 v13, s11
-; GFX10-NEXT:    v_mov_b32_e32 v12, s10
-; GFX10-NEXT:    v_mov_b32_e32 v11, s9
-; GFX10-NEXT:    v_mov_b32_e32 v10, s8
-; GFX10-NEXT:    v_mov_b32_e32 v9, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
+; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v4, s4
+; GFX10-NEXT:    v_mov_b32_e32 v5, s5
+; GFX10-NEXT:    v_mov_b32_e32 v6, s6
+; GFX10-NEXT:    v_mov_b32_e32 v7, s7
+; GFX10-NEXT:    v_mov_b32_e32 v8, s8
+; GFX10-NEXT:    v_mov_b32_e32 v9, s9
+; GFX10-NEXT:    v_mov_b32_e32 v10, s10
+; GFX10-NEXT:    v_mov_b32_e32 v11, s11
+; GFX10-NEXT:    v_mov_b32_e32 v12, s12
+; GFX10-NEXT:    v_mov_b32_e32 v13, s13
+; GFX10-NEXT:    v_mov_b32_e32 v14, s14
+; GFX10-NEXT:    v_mov_b32_e32 v15, s15
+; GFX10-NEXT:    v_mov_b32_e32 v16, s16
+; GFX10-NEXT:    v_mov_b32_e32 v17, s17
+; GFX10-NEXT:    v_mov_b32_e32 v18, s18
+; GFX10-NEXT:    v_mov_b32_e32 v19, s19
+; GFX10-NEXT:    v_mov_b32_e32 v20, s20
+; GFX10-NEXT:    v_mov_b32_e32 v21, s21
+; GFX10-NEXT:    v_mov_b32_e32 v22, s22
+; GFX10-NEXT:    v_mov_b32_e32 v23, s23
+; GFX10-NEXT:    v_mov_b32_e32 v24, s24
+; GFX10-NEXT:    v_mov_b32_e32 v25, s25
+; GFX10-NEXT:    v_mov_b32_e32 v26, s26
+; GFX10-NEXT:    v_mov_b32_e32 v27, s27
+; GFX10-NEXT:    v_mov_b32_e32 v28, s28
+; GFX10-NEXT:    v_mov_b32_e32 v29, s29
+; GFX10-NEXT:    v_mov_b32_e32 v30, s30
+; GFX10-NEXT:    v_mov_b32_e32 v31, s31
+; GFX10-NEXT:    v_mov_b32_e32 v33, s33
+; GFX10-NEXT:    v_mov_b32_e32 v32, s32
 ; GFX10-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX10-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v2
@@ -6256,55 +5544,23 @@ define amdgpu_ps <16 x i64> @dyn_insertelement_v16i64_s_v_s(<16 x i64> inreg %ve
 ;
 ; GFX11-LABEL: dyn_insertelement_v16i64_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    s_mov_b32 s12, s14
-; GFX11-NEXT:    s_mov_b32 s13, s15
-; GFX11-NEXT:    s_mov_b32 s14, s16
-; GFX11-NEXT:    s_mov_b32 s15, s17
-; GFX11-NEXT:    s_mov_b32 s16, s18
-; GFX11-NEXT:    s_mov_b32 s17, s19
-; GFX11-NEXT:    s_mov_b32 s18, s20
-; GFX11-NEXT:    s_mov_b32 s19, s21
-; GFX11-NEXT:    s_mov_b32 s20, s22
-; GFX11-NEXT:    s_mov_b32 s21, s23
-; GFX11-NEXT:    s_mov_b32 s22, s24
-; GFX11-NEXT:    s_mov_b32 s23, s25
-; GFX11-NEXT:    s_mov_b32 s24, s26
-; GFX11-NEXT:    s_mov_b32 s25, s27
-; GFX11-NEXT:    s_mov_b32 s26, s28
-; GFX11-NEXT:    s_mov_b32 s27, s29
-; GFX11-NEXT:    s_mov_b32 s28, s30
-; GFX11-NEXT:    s_mov_b32 s29, s31
-; GFX11-NEXT:    s_mov_b32 s31, s33
-; GFX11-NEXT:    s_mov_b32 s30, s32
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
 ; GFX11-NEXT:    s_lshl_b32 m0, s34, 1
-; GFX11-NEXT:    v_dual_mov_b32 v33, s31 :: v_dual_mov_b32 v32, s30
-; GFX11-NEXT:    v_dual_mov_b32 v31, s29 :: v_dual_mov_b32 v30, s28
-; GFX11-NEXT:    v_dual_mov_b32 v29, s27 :: v_dual_mov_b32 v28, s26
-; GFX11-NEXT:    v_dual_mov_b32 v27, s25 :: v_dual_mov_b32 v26, s24
-; GFX11-NEXT:    v_dual_mov_b32 v25, s23 :: v_dual_mov_b32 v24, s22
-; GFX11-NEXT:    v_dual_mov_b32 v23, s21 :: v_dual_mov_b32 v22, s20
-; GFX11-NEXT:    v_dual_mov_b32 v21, s19 :: v_dual_mov_b32 v20, s18
-; GFX11-NEXT:    v_dual_mov_b32 v19, s17 :: v_dual_mov_b32 v18, s16
-; GFX11-NEXT:    v_dual_mov_b32 v17, s15 :: v_dual_mov_b32 v16, s14
-; GFX11-NEXT:    v_dual_mov_b32 v15, s13 :: v_dual_mov_b32 v14, s12
-; GFX11-NEXT:    v_dual_mov_b32 v13, s11 :: v_dual_mov_b32 v12, s10
-; GFX11-NEXT:    v_dual_mov_b32 v11, s9 :: v_dual_mov_b32 v10, s8
-; GFX11-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
-; GFX11-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
+; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v5, s5
+; GFX11-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
+; GFX11-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
+; GFX11-NEXT:    v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
+; GFX11-NEXT:    v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
+; GFX11-NEXT:    v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s15
+; GFX11-NEXT:    v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
+; GFX11-NEXT:    v_dual_mov_b32 v18, s18 :: v_dual_mov_b32 v19, s19
+; GFX11-NEXT:    v_dual_mov_b32 v20, s20 :: v_dual_mov_b32 v21, s21
+; GFX11-NEXT:    v_dual_mov_b32 v22, s22 :: v_dual_mov_b32 v23, s23
+; GFX11-NEXT:    v_dual_mov_b32 v24, s24 :: v_dual_mov_b32 v25, s25
+; GFX11-NEXT:    v_dual_mov_b32 v26, s26 :: v_dual_mov_b32 v27, s27
+; GFX11-NEXT:    v_dual_mov_b32 v28, s28 :: v_dual_mov_b32 v29, s29
+; GFX11-NEXT:    v_dual_mov_b32 v30, s30 :: v_dual_mov_b32 v31, s31
+; GFX11-NEXT:    v_dual_mov_b32 v33, s33 :: v_dual_mov_b32 v32, s32
 ; GFX11-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX11-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX11-NEXT:    v_readfirstlane_b32 s0, v2
@@ -6343,71 +5599,39 @@ define amdgpu_ps <16 x i64> @dyn_insertelement_v16i64_s_v_s(<16 x i64> inreg %ve
 ;
 ; GFX8-LABEL: dyn_insertelement_v16i64_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
-; GFX8-NEXT:    s_mov_b32 s12, s14
-; GFX8-NEXT:    s_mov_b32 s13, s15
-; GFX8-NEXT:    s_mov_b32 s14, s16
-; GFX8-NEXT:    s_mov_b32 s15, s17
-; GFX8-NEXT:    s_mov_b32 s16, s18
-; GFX8-NEXT:    s_mov_b32 s17, s19
-; GFX8-NEXT:    s_mov_b32 s18, s20
-; GFX8-NEXT:    s_mov_b32 s19, s21
-; GFX8-NEXT:    s_mov_b32 s20, s22
-; GFX8-NEXT:    s_mov_b32 s21, s23
-; GFX8-NEXT:    s_mov_b32 s22, s24
-; GFX8-NEXT:    s_mov_b32 s23, s25
-; GFX8-NEXT:    s_mov_b32 s24, s26
-; GFX8-NEXT:    s_mov_b32 s25, s27
-; GFX8-NEXT:    s_mov_b32 s26, s28
-; GFX8-NEXT:    s_mov_b32 s27, s29
-; GFX8-NEXT:    s_mov_b32 s28, s30
-; GFX8-NEXT:    s_mov_b32 s29, s31
-; GFX8-NEXT:    s_mov_b32 s31, s33
-; GFX8-NEXT:    s_mov_b32 s30, s32
 ; GFX8-NEXT:    s_lshl_b32 m0, s34, 1
-; GFX8-NEXT:    v_mov_b32_e32 v33, s31
-; GFX8-NEXT:    v_mov_b32_e32 v32, s30
-; GFX8-NEXT:    v_mov_b32_e32 v31, s29
-; GFX8-NEXT:    v_mov_b32_e32 v30, s28
-; GFX8-NEXT:    v_mov_b32_e32 v29, s27
-; GFX8-NEXT:    v_mov_b32_e32 v28, s26
-; GFX8-NEXT:    v_mov_b32_e32 v27, s25
-; GFX8-NEXT:    v_mov_b32_e32 v26, s24
-; GFX8-NEXT:    v_mov_b32_e32 v25, s23
-; GFX8-NEXT:    v_mov_b32_e32 v24, s22
-; GFX8-NEXT:    v_mov_b32_e32 v23, s21
-; GFX8-NEXT:    v_mov_b32_e32 v22, s20
-; GFX8-NEXT:    v_mov_b32_e32 v21, s19
-; GFX8-NEXT:    v_mov_b32_e32 v20, s18
-; GFX8-NEXT:    v_mov_b32_e32 v19, s17
-; GFX8-NEXT:    v_mov_b32_e32 v18, s16
-; GFX8-NEXT:    v_mov_b32_e32 v17, s15
-; GFX8-NEXT:    v_mov_b32_e32 v16, s14
-; GFX8-NEXT:    v_mov_b32_e32 v15, s13
-; GFX8-NEXT:    v_mov_b32_e32 v14, s12
-; GFX8-NEXT:    v_mov_b32_e32 v13, s11
-; GFX8-NEXT:    v_mov_b32_e32 v12, s10
-; GFX8-NEXT:    v_mov_b32_e32 v11, s9
-; GFX8-NEXT:    v_mov_b32_e32 v10, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s3
+; GFX8-NEXT:    v_mov_b32_e32 v4, s4
+; GFX8-NEXT:    v_mov_b32_e32 v5, s5
+; GFX8-NEXT:    v_mov_b32_e32 v6, s6
+; GFX8-NEXT:    v_mov_b32_e32 v7, s7
+; GFX8-NEXT:    v_mov_b32_e32 v8, s8
+; GFX8-NEXT:    v_mov_b32_e32 v9, s9
+; GFX8-NEXT:    v_mov_b32_e32 v10, s10
+; GFX8-NEXT:    v_mov_b32_e32 v11, s11
+; GFX8-NEXT:    v_mov_b32_e32 v12, s12
+; GFX8-NEXT:    v_mov_b32_e32 v13, s13
+; GFX8-NEXT:    v_mov_b32_e32 v14, s14
+; GFX8-NEXT:    v_mov_b32_e32 v15, s15
+; GFX8-NEXT:    v_mov_b32_e32 v16, s16
+; GFX8-NEXT:    v_mov_b32_e32 v17, s17
+; GFX8-NEXT:    v_mov_b32_e32 v18, s18
+; GFX8-NEXT:    v_mov_b32_e32 v19, s19
+; GFX8-NEXT:    v_mov_b32_e32 v20, s20
+; GFX8-NEXT:    v_mov_b32_e32 v21, s21
+; GFX8-NEXT:    v_mov_b32_e32 v22, s22
+; GFX8-NEXT:    v_mov_b32_e32 v23, s23
+; GFX8-NEXT:    v_mov_b32_e32 v24, s24
+; GFX8-NEXT:    v_mov_b32_e32 v25, s25
+; GFX8-NEXT:    v_mov_b32_e32 v26, s26
+; GFX8-NEXT:    v_mov_b32_e32 v27, s27
+; GFX8-NEXT:    v_mov_b32_e32 v28, s28
+; GFX8-NEXT:    v_mov_b32_e32 v29, s29
+; GFX8-NEXT:    v_mov_b32_e32 v30, s30
+; GFX8-NEXT:    v_mov_b32_e32 v31, s31
+; GFX8-NEXT:    v_mov_b32_e32 v33, s33
+; GFX8-NEXT:    v_mov_b32_e32 v32, s32
 ; GFX8-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX8-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX8-NEXT:    v_readfirstlane_b32 s0, v2
@@ -6451,72 +5675,40 @@ entry:
 define amdgpu_ps <16 x double> @dyn_insertelement_v16f64_s_v_s(<16 x double> inreg %vec, double %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v16f64_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
-; GPRIDX-NEXT:    s_mov_b32 s12, s14
-; GPRIDX-NEXT:    s_mov_b32 s13, s15
-; GPRIDX-NEXT:    s_mov_b32 s14, s16
-; GPRIDX-NEXT:    s_mov_b32 s15, s17
-; GPRIDX-NEXT:    s_mov_b32 s16, s18
-; GPRIDX-NEXT:    s_mov_b32 s17, s19
-; GPRIDX-NEXT:    s_mov_b32 s18, s20
-; GPRIDX-NEXT:    s_mov_b32 s19, s21
-; GPRIDX-NEXT:    s_mov_b32 s20, s22
-; GPRIDX-NEXT:    s_mov_b32 s21, s23
-; GPRIDX-NEXT:    s_mov_b32 s22, s24
-; GPRIDX-NEXT:    s_mov_b32 s23, s25
-; GPRIDX-NEXT:    s_mov_b32 s24, s26
-; GPRIDX-NEXT:    s_mov_b32 s25, s27
-; GPRIDX-NEXT:    s_mov_b32 s26, s28
-; GPRIDX-NEXT:    s_mov_b32 s27, s29
-; GPRIDX-NEXT:    s_mov_b32 s28, s30
-; GPRIDX-NEXT:    s_mov_b32 s29, s31
-; GPRIDX-NEXT:    s_mov_b32 s31, s33
-; GPRIDX-NEXT:    s_mov_b32 s30, s32
-; GPRIDX-NEXT:    s_lshl_b32 s33, s34, 1
-; GPRIDX-NEXT:    v_mov_b32_e32 v33, s31
-; GPRIDX-NEXT:    v_mov_b32_e32 v32, s30
-; GPRIDX-NEXT:    v_mov_b32_e32 v31, s29
-; GPRIDX-NEXT:    v_mov_b32_e32 v30, s28
-; GPRIDX-NEXT:    v_mov_b32_e32 v29, s27
-; GPRIDX-NEXT:    v_mov_b32_e32 v28, s26
-; GPRIDX-NEXT:    v_mov_b32_e32 v27, s25
-; GPRIDX-NEXT:    v_mov_b32_e32 v26, s24
-; GPRIDX-NEXT:    v_mov_b32_e32 v25, s23
-; GPRIDX-NEXT:    v_mov_b32_e32 v24, s22
-; GPRIDX-NEXT:    v_mov_b32_e32 v23, s21
-; GPRIDX-NEXT:    v_mov_b32_e32 v22, s20
-; GPRIDX-NEXT:    v_mov_b32_e32 v21, s19
-; GPRIDX-NEXT:    v_mov_b32_e32 v20, s18
-; GPRIDX-NEXT:    v_mov_b32_e32 v19, s17
-; GPRIDX-NEXT:    v_mov_b32_e32 v18, s16
-; GPRIDX-NEXT:    v_mov_b32_e32 v17, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v16, s14
-; GPRIDX-NEXT:    v_mov_b32_e32 v15, s13
-; GPRIDX-NEXT:    v_mov_b32_e32 v14, s12
-; GPRIDX-NEXT:    v_mov_b32_e32 v13, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s0
-; GPRIDX-NEXT:    s_set_gpr_idx_on s33, gpr_idx(DST)
+; GPRIDX-NEXT:    s_lshl_b32 s1, s34, 1
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v13, s13
+; GPRIDX-NEXT:    v_mov_b32_e32 v14, s14
+; GPRIDX-NEXT:    v_mov_b32_e32 v15, s15
+; GPRIDX-NEXT:    v_mov_b32_e32 v16, s16
+; GPRIDX-NEXT:    v_mov_b32_e32 v17, s17
+; GPRIDX-NEXT:    v_mov_b32_e32 v18, s18
+; GPRIDX-NEXT:    v_mov_b32_e32 v19, s19
+; GPRIDX-NEXT:    v_mov_b32_e32 v20, s20
+; GPRIDX-NEXT:    v_mov_b32_e32 v21, s21
+; GPRIDX-NEXT:    v_mov_b32_e32 v22, s22
+; GPRIDX-NEXT:    v_mov_b32_e32 v23, s23
+; GPRIDX-NEXT:    v_mov_b32_e32 v24, s24
+; GPRIDX-NEXT:    v_mov_b32_e32 v25, s25
+; GPRIDX-NEXT:    v_mov_b32_e32 v26, s26
+; GPRIDX-NEXT:    v_mov_b32_e32 v27, s27
+; GPRIDX-NEXT:    v_mov_b32_e32 v28, s28
+; GPRIDX-NEXT:    v_mov_b32_e32 v29, s29
+; GPRIDX-NEXT:    v_mov_b32_e32 v30, s30
+; GPRIDX-NEXT:    v_mov_b32_e32 v31, s31
+; GPRIDX-NEXT:    v_mov_b32_e32 v33, s33
+; GPRIDX-NEXT:    v_mov_b32_e32 v32, s32
+; GPRIDX-NEXT:    s_set_gpr_idx_on s1, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v2, v0
 ; GPRIDX-NEXT:    v_mov_b32_e32 v3, v1
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -6556,71 +5748,39 @@ define amdgpu_ps <16 x double> @dyn_insertelement_v16f64_s_v_s(<16 x double> inr
 ;
 ; GFX10-LABEL: dyn_insertelement_v16f64_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    s_mov_b32 s12, s14
-; GFX10-NEXT:    s_mov_b32 s13, s15
-; GFX10-NEXT:    s_mov_b32 s14, s16
-; GFX10-NEXT:    s_mov_b32 s15, s17
-; GFX10-NEXT:    s_mov_b32 s16, s18
-; GFX10-NEXT:    s_mov_b32 s17, s19
-; GFX10-NEXT:    s_mov_b32 s18, s20
-; GFX10-NEXT:    s_mov_b32 s19, s21
-; GFX10-NEXT:    s_mov_b32 s20, s22
-; GFX10-NEXT:    s_mov_b32 s21, s23
-; GFX10-NEXT:    s_mov_b32 s22, s24
-; GFX10-NEXT:    s_mov_b32 s23, s25
-; GFX10-NEXT:    s_mov_b32 s24, s26
-; GFX10-NEXT:    s_mov_b32 s25, s27
-; GFX10-NEXT:    s_mov_b32 s26, s28
-; GFX10-NEXT:    s_mov_b32 s27, s29
-; GFX10-NEXT:    s_mov_b32 s28, s30
-; GFX10-NEXT:    s_mov_b32 s29, s31
-; GFX10-NEXT:    s_mov_b32 s31, s33
-; GFX10-NEXT:    s_mov_b32 s30, s32
-; GFX10-NEXT:    v_mov_b32_e32 v2, s0
+; GFX10-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX10-NEXT:    s_lshl_b32 m0, s34, 1
-; GFX10-NEXT:    v_mov_b32_e32 v33, s31
-; GFX10-NEXT:    v_mov_b32_e32 v32, s30
-; GFX10-NEXT:    v_mov_b32_e32 v31, s29
-; GFX10-NEXT:    v_mov_b32_e32 v30, s28
-; GFX10-NEXT:    v_mov_b32_e32 v29, s27
-; GFX10-NEXT:    v_mov_b32_e32 v28, s26
-; GFX10-NEXT:    v_mov_b32_e32 v27, s25
-; GFX10-NEXT:    v_mov_b32_e32 v26, s24
-; GFX10-NEXT:    v_mov_b32_e32 v25, s23
-; GFX10-NEXT:    v_mov_b32_e32 v24, s22
-; GFX10-NEXT:    v_mov_b32_e32 v23, s21
-; GFX10-NEXT:    v_mov_b32_e32 v22, s20
-; GFX10-NEXT:    v_mov_b32_e32 v21, s19
-; GFX10-NEXT:    v_mov_b32_e32 v20, s18
-; GFX10-NEXT:    v_mov_b32_e32 v19, s17
-; GFX10-NEXT:    v_mov_b32_e32 v18, s16
-; GFX10-NEXT:    v_mov_b32_e32 v17, s15
-; GFX10-NEXT:    v_mov_b32_e32 v16, s14
-; GFX10-NEXT:    v_mov_b32_e32 v15, s13
-; GFX10-NEXT:    v_mov_b32_e32 v14, s12
-; GFX10-NEXT:    v_mov_b32_e32 v13, s11
-; GFX10-NEXT:    v_mov_b32_e32 v12, s10
-; GFX10-NEXT:    v_mov_b32_e32 v11, s9
-; GFX10-NEXT:    v_mov_b32_e32 v10, s8
-; GFX10-NEXT:    v_mov_b32_e32 v9, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
+; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v4, s4
+; GFX10-NEXT:    v_mov_b32_e32 v5, s5
+; GFX10-NEXT:    v_mov_b32_e32 v6, s6
+; GFX10-NEXT:    v_mov_b32_e32 v7, s7
+; GFX10-NEXT:    v_mov_b32_e32 v8, s8
+; GFX10-NEXT:    v_mov_b32_e32 v9, s9
+; GFX10-NEXT:    v_mov_b32_e32 v10, s10
+; GFX10-NEXT:    v_mov_b32_e32 v11, s11
+; GFX10-NEXT:    v_mov_b32_e32 v12, s12
+; GFX10-NEXT:    v_mov_b32_e32 v13, s13
+; GFX10-NEXT:    v_mov_b32_e32 v14, s14
+; GFX10-NEXT:    v_mov_b32_e32 v15, s15
+; GFX10-NEXT:    v_mov_b32_e32 v16, s16
+; GFX10-NEXT:    v_mov_b32_e32 v17, s17
+; GFX10-NEXT:    v_mov_b32_e32 v18, s18
+; GFX10-NEXT:    v_mov_b32_e32 v19, s19
+; GFX10-NEXT:    v_mov_b32_e32 v20, s20
+; GFX10-NEXT:    v_mov_b32_e32 v21, s21
+; GFX10-NEXT:    v_mov_b32_e32 v22, s22
+; GFX10-NEXT:    v_mov_b32_e32 v23, s23
+; GFX10-NEXT:    v_mov_b32_e32 v24, s24
+; GFX10-NEXT:    v_mov_b32_e32 v25, s25
+; GFX10-NEXT:    v_mov_b32_e32 v26, s26
+; GFX10-NEXT:    v_mov_b32_e32 v27, s27
+; GFX10-NEXT:    v_mov_b32_e32 v28, s28
+; GFX10-NEXT:    v_mov_b32_e32 v29, s29
+; GFX10-NEXT:    v_mov_b32_e32 v30, s30
+; GFX10-NEXT:    v_mov_b32_e32 v31, s31
+; GFX10-NEXT:    v_mov_b32_e32 v33, s33
+; GFX10-NEXT:    v_mov_b32_e32 v32, s32
 ; GFX10-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX10-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v2
@@ -6659,55 +5819,23 @@ define amdgpu_ps <16 x double> @dyn_insertelement_v16f64_s_v_s(<16 x double> inr
 ;
 ; GFX11-LABEL: dyn_insertelement_v16f64_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    s_mov_b32 s12, s14
-; GFX11-NEXT:    s_mov_b32 s13, s15
-; GFX11-NEXT:    s_mov_b32 s14, s16
-; GFX11-NEXT:    s_mov_b32 s15, s17
-; GFX11-NEXT:    s_mov_b32 s16, s18
-; GFX11-NEXT:    s_mov_b32 s17, s19
-; GFX11-NEXT:    s_mov_b32 s18, s20
-; GFX11-NEXT:    s_mov_b32 s19, s21
-; GFX11-NEXT:    s_mov_b32 s20, s22
-; GFX11-NEXT:    s_mov_b32 s21, s23
-; GFX11-NEXT:    s_mov_b32 s22, s24
-; GFX11-NEXT:    s_mov_b32 s23, s25
-; GFX11-NEXT:    s_mov_b32 s24, s26
-; GFX11-NEXT:    s_mov_b32 s25, s27
-; GFX11-NEXT:    s_mov_b32 s26, s28
-; GFX11-NEXT:    s_mov_b32 s27, s29
-; GFX11-NEXT:    s_mov_b32 s28, s30
-; GFX11-NEXT:    s_mov_b32 s29, s31
-; GFX11-NEXT:    s_mov_b32 s31, s33
-; GFX11-NEXT:    s_mov_b32 s30, s32
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
 ; GFX11-NEXT:    s_lshl_b32 m0, s34, 1
-; GFX11-NEXT:    v_dual_mov_b32 v33, s31 :: v_dual_mov_b32 v32, s30
-; GFX11-NEXT:    v_dual_mov_b32 v31, s29 :: v_dual_mov_b32 v30, s28
-; GFX11-NEXT:    v_dual_mov_b32 v29, s27 :: v_dual_mov_b32 v28, s26
-; GFX11-NEXT:    v_dual_mov_b32 v27, s25 :: v_dual_mov_b32 v26, s24
-; GFX11-NEXT:    v_dual_mov_b32 v25, s23 :: v_dual_mov_b32 v24, s22
-; GFX11-NEXT:    v_dual_mov_b32 v23, s21 :: v_dual_mov_b32 v22, s20
-; GFX11-NEXT:    v_dual_mov_b32 v21, s19 :: v_dual_mov_b32 v20, s18
-; GFX11-NEXT:    v_dual_mov_b32 v19, s17 :: v_dual_mov_b32 v18, s16
-; GFX11-NEXT:    v_dual_mov_b32 v17, s15 :: v_dual_mov_b32 v16, s14
-; GFX11-NEXT:    v_dual_mov_b32 v15, s13 :: v_dual_mov_b32 v14, s12
-; GFX11-NEXT:    v_dual_mov_b32 v13, s11 :: v_dual_mov_b32 v12, s10
-; GFX11-NEXT:    v_dual_mov_b32 v11, s9 :: v_dual_mov_b32 v10, s8
-; GFX11-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
-; GFX11-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
+; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v5, s5
+; GFX11-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
+; GFX11-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
+; GFX11-NEXT:    v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
+; GFX11-NEXT:    v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
+; GFX11-NEXT:    v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s15
+; GFX11-NEXT:    v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
+; GFX11-NEXT:    v_dual_mov_b32 v18, s18 :: v_dual_mov_b32 v19, s19
+; GFX11-NEXT:    v_dual_mov_b32 v20, s20 :: v_dual_mov_b32 v21, s21
+; GFX11-NEXT:    v_dual_mov_b32 v22, s22 :: v_dual_mov_b32 v23, s23
+; GFX11-NEXT:    v_dual_mov_b32 v24, s24 :: v_dual_mov_b32 v25, s25
+; GFX11-NEXT:    v_dual_mov_b32 v26, s26 :: v_dual_mov_b32 v27, s27
+; GFX11-NEXT:    v_dual_mov_b32 v28, s28 :: v_dual_mov_b32 v29, s29
+; GFX11-NEXT:    v_dual_mov_b32 v30, s30 :: v_dual_mov_b32 v31, s31
+; GFX11-NEXT:    v_dual_mov_b32 v33, s33 :: v_dual_mov_b32 v32, s32
 ; GFX11-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX11-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX11-NEXT:    v_readfirstlane_b32 s0, v2
@@ -6716,101 +5844,69 @@ define amdgpu_ps <16 x double> @dyn_insertelement_v16f64_s_v_s(<16 x double> inr
 ; GFX11-NEXT:    v_readfirstlane_b32 s3, v5
 ; GFX11-NEXT:    v_readfirstlane_b32 s4, v6
 ; GFX11-NEXT:    v_readfirstlane_b32 s5, v7
-; GFX11-NEXT:    v_readfirstlane_b32 s6, v8
-; GFX11-NEXT:    v_readfirstlane_b32 s7, v9
-; GFX11-NEXT:    v_readfirstlane_b32 s8, v10
-; GFX11-NEXT:    v_readfirstlane_b32 s9, v11
-; GFX11-NEXT:    v_readfirstlane_b32 s10, v12
-; GFX11-NEXT:    v_readfirstlane_b32 s11, v13
-; GFX11-NEXT:    v_readfirstlane_b32 s12, v14
-; GFX11-NEXT:    v_readfirstlane_b32 s13, v15
-; GFX11-NEXT:    v_readfirstlane_b32 s14, v16
-; GFX11-NEXT:    v_readfirstlane_b32 s15, v17
-; GFX11-NEXT:    v_readfirstlane_b32 s16, v18
-; GFX11-NEXT:    v_readfirstlane_b32 s17, v19
-; GFX11-NEXT:    v_readfirstlane_b32 s18, v20
-; GFX11-NEXT:    v_readfirstlane_b32 s19, v21
-; GFX11-NEXT:    v_readfirstlane_b32 s20, v22
-; GFX11-NEXT:    v_readfirstlane_b32 s21, v23
-; GFX11-NEXT:    v_readfirstlane_b32 s22, v24
-; GFX11-NEXT:    v_readfirstlane_b32 s23, v25
-; GFX11-NEXT:    v_readfirstlane_b32 s24, v26
-; GFX11-NEXT:    v_readfirstlane_b32 s25, v27
-; GFX11-NEXT:    v_readfirstlane_b32 s26, v28
-; GFX11-NEXT:    v_readfirstlane_b32 s27, v29
-; GFX11-NEXT:    v_readfirstlane_b32 s28, v30
-; GFX11-NEXT:    v_readfirstlane_b32 s29, v31
-; GFX11-NEXT:    v_readfirstlane_b32 s30, v32
-; GFX11-NEXT:    v_readfirstlane_b32 s31, v33
-; GFX11-NEXT:    ; return to shader part epilog
-;
-; GFX8-LABEL: dyn_insertelement_v16f64_s_v_s:
-; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
-; GFX8-NEXT:    s_mov_b32 s12, s14
-; GFX8-NEXT:    s_mov_b32 s13, s15
-; GFX8-NEXT:    s_mov_b32 s14, s16
-; GFX8-NEXT:    s_mov_b32 s15, s17
-; GFX8-NEXT:    s_mov_b32 s16, s18
-; GFX8-NEXT:    s_mov_b32 s17, s19
-; GFX8-NEXT:    s_mov_b32 s18, s20
-; GFX8-NEXT:    s_mov_b32 s19, s21
-; GFX8-NEXT:    s_mov_b32 s20, s22
-; GFX8-NEXT:    s_mov_b32 s21, s23
-; GFX8-NEXT:    s_mov_b32 s22, s24
-; GFX8-NEXT:    s_mov_b32 s23, s25
-; GFX8-NEXT:    s_mov_b32 s24, s26
-; GFX8-NEXT:    s_mov_b32 s25, s27
-; GFX8-NEXT:    s_mov_b32 s26, s28
-; GFX8-NEXT:    s_mov_b32 s27, s29
-; GFX8-NEXT:    s_mov_b32 s28, s30
-; GFX8-NEXT:    s_mov_b32 s29, s31
-; GFX8-NEXT:    s_mov_b32 s31, s33
-; GFX8-NEXT:    s_mov_b32 s30, s32
+; GFX11-NEXT:    v_readfirstlane_b32 s6, v8
+; GFX11-NEXT:    v_readfirstlane_b32 s7, v9
+; GFX11-NEXT:    v_readfirstlane_b32 s8, v10
+; GFX11-NEXT:    v_readfirstlane_b32 s9, v11
+; GFX11-NEXT:    v_readfirstlane_b32 s10, v12
+; GFX11-NEXT:    v_readfirstlane_b32 s11, v13
+; GFX11-NEXT:    v_readfirstlane_b32 s12, v14
+; GFX11-NEXT:    v_readfirstlane_b32 s13, v15
+; GFX11-NEXT:    v_readfirstlane_b32 s14, v16
+; GFX11-NEXT:    v_readfirstlane_b32 s15, v17
+; GFX11-NEXT:    v_readfirstlane_b32 s16, v18
+; GFX11-NEXT:    v_readfirstlane_b32 s17, v19
+; GFX11-NEXT:    v_readfirstlane_b32 s18, v20
+; GFX11-NEXT:    v_readfirstlane_b32 s19, v21
+; GFX11-NEXT:    v_readfirstlane_b32 s20, v22
+; GFX11-NEXT:    v_readfirstlane_b32 s21, v23
+; GFX11-NEXT:    v_readfirstlane_b32 s22, v24
+; GFX11-NEXT:    v_readfirstlane_b32 s23, v25
+; GFX11-NEXT:    v_readfirstlane_b32 s24, v26
+; GFX11-NEXT:    v_readfirstlane_b32 s25, v27
+; GFX11-NEXT:    v_readfirstlane_b32 s26, v28
+; GFX11-NEXT:    v_readfirstlane_b32 s27, v29
+; GFX11-NEXT:    v_readfirstlane_b32 s28, v30
+; GFX11-NEXT:    v_readfirstlane_b32 s29, v31
+; GFX11-NEXT:    v_readfirstlane_b32 s30, v32
+; GFX11-NEXT:    v_readfirstlane_b32 s31, v33
+; GFX11-NEXT:    ; return to shader part epilog
+;
+; GFX8-LABEL: dyn_insertelement_v16f64_s_v_s:
+; GFX8:       ; %bb.0: ; %entry
 ; GFX8-NEXT:    s_lshl_b32 m0, s34, 1
-; GFX8-NEXT:    v_mov_b32_e32 v33, s31
-; GFX8-NEXT:    v_mov_b32_e32 v32, s30
-; GFX8-NEXT:    v_mov_b32_e32 v31, s29
-; GFX8-NEXT:    v_mov_b32_e32 v30, s28
-; GFX8-NEXT:    v_mov_b32_e32 v29, s27
-; GFX8-NEXT:    v_mov_b32_e32 v28, s26
-; GFX8-NEXT:    v_mov_b32_e32 v27, s25
-; GFX8-NEXT:    v_mov_b32_e32 v26, s24
-; GFX8-NEXT:    v_mov_b32_e32 v25, s23
-; GFX8-NEXT:    v_mov_b32_e32 v24, s22
-; GFX8-NEXT:    v_mov_b32_e32 v23, s21
-; GFX8-NEXT:    v_mov_b32_e32 v22, s20
-; GFX8-NEXT:    v_mov_b32_e32 v21, s19
-; GFX8-NEXT:    v_mov_b32_e32 v20, s18
-; GFX8-NEXT:    v_mov_b32_e32 v19, s17
-; GFX8-NEXT:    v_mov_b32_e32 v18, s16
-; GFX8-NEXT:    v_mov_b32_e32 v17, s15
-; GFX8-NEXT:    v_mov_b32_e32 v16, s14
-; GFX8-NEXT:    v_mov_b32_e32 v15, s13
-; GFX8-NEXT:    v_mov_b32_e32 v14, s12
-; GFX8-NEXT:    v_mov_b32_e32 v13, s11
-; GFX8-NEXT:    v_mov_b32_e32 v12, s10
-; GFX8-NEXT:    v_mov_b32_e32 v11, s9
-; GFX8-NEXT:    v_mov_b32_e32 v10, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s3
+; GFX8-NEXT:    v_mov_b32_e32 v4, s4
+; GFX8-NEXT:    v_mov_b32_e32 v5, s5
+; GFX8-NEXT:    v_mov_b32_e32 v6, s6
+; GFX8-NEXT:    v_mov_b32_e32 v7, s7
+; GFX8-NEXT:    v_mov_b32_e32 v8, s8
+; GFX8-NEXT:    v_mov_b32_e32 v9, s9
+; GFX8-NEXT:    v_mov_b32_e32 v10, s10
+; GFX8-NEXT:    v_mov_b32_e32 v11, s11
+; GFX8-NEXT:    v_mov_b32_e32 v12, s12
+; GFX8-NEXT:    v_mov_b32_e32 v13, s13
+; GFX8-NEXT:    v_mov_b32_e32 v14, s14
+; GFX8-NEXT:    v_mov_b32_e32 v15, s15
+; GFX8-NEXT:    v_mov_b32_e32 v16, s16
+; GFX8-NEXT:    v_mov_b32_e32 v17, s17
+; GFX8-NEXT:    v_mov_b32_e32 v18, s18
+; GFX8-NEXT:    v_mov_b32_e32 v19, s19
+; GFX8-NEXT:    v_mov_b32_e32 v20, s20
+; GFX8-NEXT:    v_mov_b32_e32 v21, s21
+; GFX8-NEXT:    v_mov_b32_e32 v22, s22
+; GFX8-NEXT:    v_mov_b32_e32 v23, s23
+; GFX8-NEXT:    v_mov_b32_e32 v24, s24
+; GFX8-NEXT:    v_mov_b32_e32 v25, s25
+; GFX8-NEXT:    v_mov_b32_e32 v26, s26
+; GFX8-NEXT:    v_mov_b32_e32 v27, s27
+; GFX8-NEXT:    v_mov_b32_e32 v28, s28
+; GFX8-NEXT:    v_mov_b32_e32 v29, s29
+; GFX8-NEXT:    v_mov_b32_e32 v30, s30
+; GFX8-NEXT:    v_mov_b32_e32 v31, s31
+; GFX8-NEXT:    v_mov_b32_e32 v33, s33
+; GFX8-NEXT:    v_mov_b32_e32 v32, s32
 ; GFX8-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX8-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX8-NEXT:    v_readfirstlane_b32 s0, v2
@@ -6944,21 +6040,14 @@ entry:
 define amdgpu_ps <7 x float> @dyn_insertelement_v7f32_s_v_s(<7 x float> inreg %vec, float %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v7f32_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
 ; GPRIDX-NEXT:    v_mov_b32_e32 v7, v0
-; GPRIDX-NEXT:    v_mov_b32_e32 v0, s0
-; GPRIDX-NEXT:    v_mov_b32_e32 v1, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v0, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v1, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s8
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s9, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v0, v7
 ; GPRIDX-NEXT:    s_set_gpr_idx_off
@@ -6966,59 +6055,38 @@ define amdgpu_ps <7 x float> @dyn_insertelement_v7f32_s_v_s(<7 x float> inreg %v
 ;
 ; GFX10-LABEL: dyn_insertelement_v7f32_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
 ; GFX10-NEXT:    v_mov_b32_e32 v7, v0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    s_mov_b32 m0, s9
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s6
+; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
+; GFX10-NEXT:    v_mov_b32_e32 v4, s6
+; GFX10-NEXT:    v_mov_b32_e32 v5, s7
+; GFX10-NEXT:    v_mov_b32_e32 v6, s8
 ; GFX10-NEXT:    v_movreld_b32_e32 v0, v7
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: dyn_insertelement_v7f32_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    v_dual_mov_b32 v7, v0 :: v_dual_mov_b32 v0, s0
+; GFX11-NEXT:    v_dual_mov_b32 v7, v0 :: v_dual_mov_b32 v0, s2
 ; GFX11-NEXT:    s_mov_b32 m0, s9
-; GFX11-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX11-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s4
-; GFX11-NEXT:    v_dual_mov_b32 v5, s5 :: v_dual_mov_b32 v6, s6
+; GFX11-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX11-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
+; GFX11-NEXT:    v_dual_mov_b32 v5, s7 :: v_dual_mov_b32 v6, s8
 ; GFX11-NEXT:    v_movreld_b32_e32 v0, v7
 ; GFX11-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: dyn_insertelement_v7f32_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
 ; GFX8-NEXT:    v_mov_b32_e32 v7, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s6
+; GFX8-NEXT:    v_mov_b32_e32 v0, s2
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NEXT:    v_mov_b32_e32 v6, s8
 ; GFX8-NEXT:    s_mov_b32 m0, s9
 ; GFX8-NEXT:    v_movreld_b32_e32 v0, v7
 ; GFX8-NEXT:    ; return to shader part epilog
@@ -7270,36 +6338,20 @@ entry:
 define amdgpu_ps <7 x double> @dyn_insertelement_v7f64_s_v_s(<7 x double> inreg %vec, double %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v7f64_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
-; GPRIDX-NEXT:    s_mov_b32 s12, s14
-; GPRIDX-NEXT:    s_mov_b32 s13, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v17, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v16, s14
-; GPRIDX-NEXT:    v_mov_b32_e32 v15, s13
-; GPRIDX-NEXT:    v_mov_b32_e32 v14, s12
-; GPRIDX-NEXT:    v_mov_b32_e32 v13, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s0
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v13, s13
+; GPRIDX-NEXT:    v_mov_b32_e32 v14, s14
+; GPRIDX-NEXT:    v_mov_b32_e32 v15, s15
 ; GPRIDX-NEXT:    s_lshl_b32 s0, s16, 1
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s0, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v2, v0
@@ -7323,37 +6375,21 @@ define amdgpu_ps <7 x double> @dyn_insertelement_v7f64_s_v_s(<7 x double> inreg
 ;
 ; GFX10-LABEL: dyn_insertelement_v7f64_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    s_mov_b32 s12, s14
-; GFX10-NEXT:    s_mov_b32 s13, s15
-; GFX10-NEXT:    v_mov_b32_e32 v2, s0
+; GFX10-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX10-NEXT:    s_lshl_b32 m0, s16, 1
-; GFX10-NEXT:    v_mov_b32_e32 v17, s15
-; GFX10-NEXT:    v_mov_b32_e32 v16, s14
-; GFX10-NEXT:    v_mov_b32_e32 v15, s13
-; GFX10-NEXT:    v_mov_b32_e32 v14, s12
-; GFX10-NEXT:    v_mov_b32_e32 v13, s11
-; GFX10-NEXT:    v_mov_b32_e32 v12, s10
-; GFX10-NEXT:    v_mov_b32_e32 v11, s9
-; GFX10-NEXT:    v_mov_b32_e32 v10, s8
-; GFX10-NEXT:    v_mov_b32_e32 v9, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
+; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v4, s4
+; GFX10-NEXT:    v_mov_b32_e32 v5, s5
+; GFX10-NEXT:    v_mov_b32_e32 v6, s6
+; GFX10-NEXT:    v_mov_b32_e32 v7, s7
+; GFX10-NEXT:    v_mov_b32_e32 v8, s8
+; GFX10-NEXT:    v_mov_b32_e32 v9, s9
+; GFX10-NEXT:    v_mov_b32_e32 v10, s10
+; GFX10-NEXT:    v_mov_b32_e32 v11, s11
+; GFX10-NEXT:    v_mov_b32_e32 v12, s12
+; GFX10-NEXT:    v_mov_b32_e32 v13, s13
+; GFX10-NEXT:    v_mov_b32_e32 v14, s14
+; GFX10-NEXT:    v_mov_b32_e32 v15, s15
 ; GFX10-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX10-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v2
@@ -7374,29 +6410,14 @@ define amdgpu_ps <7 x double> @dyn_insertelement_v7f64_s_v_s(<7 x double> inreg
 ;
 ; GFX11-LABEL: dyn_insertelement_v7f64_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    s_mov_b32 s12, s14
-; GFX11-NEXT:    s_mov_b32 s13, s15
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
 ; GFX11-NEXT:    s_lshl_b32 m0, s16, 1
-; GFX11-NEXT:    v_dual_mov_b32 v17, s15 :: v_dual_mov_b32 v16, s14
-; GFX11-NEXT:    v_dual_mov_b32 v15, s13 :: v_dual_mov_b32 v14, s12
-; GFX11-NEXT:    v_dual_mov_b32 v13, s11 :: v_dual_mov_b32 v12, s10
-; GFX11-NEXT:    v_dual_mov_b32 v11, s9 :: v_dual_mov_b32 v10, s8
-; GFX11-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
-; GFX11-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
+; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v5, s5
+; GFX11-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
+; GFX11-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
+; GFX11-NEXT:    v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
+; GFX11-NEXT:    v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
+; GFX11-NEXT:    v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s15
 ; GFX11-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX11-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX11-NEXT:    v_readfirstlane_b32 s0, v2
@@ -7417,36 +6438,20 @@ define amdgpu_ps <7 x double> @dyn_insertelement_v7f64_s_v_s(<7 x double> inreg
 ;
 ; GFX8-LABEL: dyn_insertelement_v7f64_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
-; GFX8-NEXT:    s_mov_b32 s12, s14
-; GFX8-NEXT:    s_mov_b32 s13, s15
-; GFX8-NEXT:    v_mov_b32_e32 v17, s15
-; GFX8-NEXT:    v_mov_b32_e32 v16, s14
-; GFX8-NEXT:    v_mov_b32_e32 v15, s13
-; GFX8-NEXT:    v_mov_b32_e32 v14, s12
-; GFX8-NEXT:    v_mov_b32_e32 v13, s11
-; GFX8-NEXT:    v_mov_b32_e32 v12, s10
-; GFX8-NEXT:    v_mov_b32_e32 v11, s9
-; GFX8-NEXT:    v_mov_b32_e32 v10, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s3
+; GFX8-NEXT:    v_mov_b32_e32 v4, s4
+; GFX8-NEXT:    v_mov_b32_e32 v5, s5
+; GFX8-NEXT:    v_mov_b32_e32 v6, s6
+; GFX8-NEXT:    v_mov_b32_e32 v7, s7
+; GFX8-NEXT:    v_mov_b32_e32 v8, s8
+; GFX8-NEXT:    v_mov_b32_e32 v9, s9
+; GFX8-NEXT:    v_mov_b32_e32 v10, s10
+; GFX8-NEXT:    v_mov_b32_e32 v11, s11
+; GFX8-NEXT:    v_mov_b32_e32 v12, s12
+; GFX8-NEXT:    v_mov_b32_e32 v13, s13
+; GFX8-NEXT:    v_mov_b32_e32 v14, s14
+; GFX8-NEXT:    v_mov_b32_e32 v15, s15
 ; GFX8-NEXT:    s_lshl_b32 m0, s16, 1
 ; GFX8-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX8-NEXT:    v_movreld_b32_e32 v3, v1
@@ -7473,36 +6478,20 @@ entry:
 define amdgpu_ps <7 x double> @dyn_insertelement_v7f64_s_v_v(<7 x double> inreg %vec, double %val, i32 %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v7f64_s_v_v:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    s_mov_b32 s10, s12
-; GPRIDX-NEXT:    s_mov_b32 s11, s13
-; GPRIDX-NEXT:    s_mov_b32 s12, s14
-; GPRIDX-NEXT:    s_mov_b32 s13, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v18, s15
-; GPRIDX-NEXT:    v_mov_b32_e32 v17, s14
-; GPRIDX-NEXT:    v_mov_b32_e32 v16, s13
-; GPRIDX-NEXT:    v_mov_b32_e32 v15, s12
-; GPRIDX-NEXT:    v_mov_b32_e32 v14, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v13, s10
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s0
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s11
+; GPRIDX-NEXT:    v_mov_b32_e32 v13, s12
+; GPRIDX-NEXT:    v_mov_b32_e32 v14, s13
+; GPRIDX-NEXT:    v_mov_b32_e32 v15, s14
+; GPRIDX-NEXT:    v_mov_b32_e32 v16, s15
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e64 s[0:1], 2, v2
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e64 s[2:3], 3, v2
@@ -7542,61 +6531,45 @@ define amdgpu_ps <7 x double> @dyn_insertelement_v7f64_s_v_v(<7 x double> inreg
 ;
 ; GFX10-LABEL: dyn_insertelement_v7f64_s_v_v:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    s_mov_b32 s10, s12
-; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    s_mov_b32 s12, s14
-; GFX10-NEXT:    s_mov_b32 s13, s15
-; GFX10-NEXT:    v_mov_b32_e32 v18, s15
-; GFX10-NEXT:    v_mov_b32_e32 v17, s14
-; GFX10-NEXT:    v_mov_b32_e32 v16, s13
-; GFX10-NEXT:    v_mov_b32_e32 v15, s12
-; GFX10-NEXT:    v_mov_b32_e32 v14, s11
-; GFX10-NEXT:    v_mov_b32_e32 v13, s10
-; GFX10-NEXT:    v_mov_b32_e32 v12, s9
-; GFX10-NEXT:    v_mov_b32_e32 v11, s8
-; GFX10-NEXT:    v_mov_b32_e32 v10, s7
-; GFX10-NEXT:    v_mov_b32_e32 v9, s6
-; GFX10-NEXT:    v_mov_b32_e32 v8, s5
-; GFX10-NEXT:    v_mov_b32_e32 v7, s4
-; GFX10-NEXT:    v_mov_b32_e32 v6, s3
-; GFX10-NEXT:    v_mov_b32_e32 v5, s2
-; GFX10-NEXT:    v_mov_b32_e32 v4, s1
-; GFX10-NEXT:    v_mov_b32_e32 v3, s0
+; GFX10-NEXT:    v_mov_b32_e32 v3, s2
+; GFX10-NEXT:    v_mov_b32_e32 v4, s3
+; GFX10-NEXT:    v_mov_b32_e32 v5, s4
+; GFX10-NEXT:    v_mov_b32_e32 v6, s5
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v2
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 1, v2
-; GFX10-NEXT:    v_cmp_eq_u32_e64 s1, 6, v2
+; GFX10-NEXT:    v_mov_b32_e32 v7, s6
+; GFX10-NEXT:    v_mov_b32_e32 v8, s7
+; GFX10-NEXT:    v_mov_b32_e32 v9, s8
+; GFX10-NEXT:    v_mov_b32_e32 v10, s9
 ; GFX10-NEXT:    v_cndmask_b32_e32 v3, v3, v0, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e32 v4, v4, v1, vcc_lo
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 2, v2
 ; GFX10-NEXT:    v_cndmask_b32_e64 v5, v5, v0, s0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v6, v6, v1, s0
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 3, v2
+; GFX10-NEXT:    v_mov_b32_e32 v11, s10
+; GFX10-NEXT:    v_mov_b32_e32 v12, s11
+; GFX10-NEXT:    v_mov_b32_e32 v13, s12
+; GFX10-NEXT:    v_mov_b32_e32 v14, s13
+; GFX10-NEXT:    v_mov_b32_e32 v15, s14
+; GFX10-NEXT:    v_mov_b32_e32 v16, s15
 ; GFX10-NEXT:    v_cndmask_b32_e32 v7, v7, v0, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e32 v8, v8, v1, vcc_lo
-; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 4, v2
 ; GFX10-NEXT:    v_cndmask_b32_e64 v9, v9, v0, s0
+; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 4, v2
 ; GFX10-NEXT:    v_cndmask_b32_e64 v10, v10, v1, s0
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 5, v2
+; GFX10-NEXT:    v_cmp_eq_u32_e64 s1, 6, v2
 ; GFX10-NEXT:    v_readfirstlane_b32 s2, v5
 ; GFX10-NEXT:    v_cndmask_b32_e32 v11, v11, v0, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e32 v2, v12, v1, vcc_lo
-; GFX10-NEXT:    v_readfirstlane_b32 s3, v6
 ; GFX10-NEXT:    v_cndmask_b32_e64 v12, v13, v0, s0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v13, v14, v1, s0
 ; GFX10-NEXT:    v_cndmask_b32_e64 v0, v15, v0, s1
 ; GFX10-NEXT:    v_cndmask_b32_e64 v1, v16, v1, s1
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v3
 ; GFX10-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX10-NEXT:    v_readfirstlane_b32 s3, v6
 ; GFX10-NEXT:    v_readfirstlane_b32 s4, v7
 ; GFX10-NEXT:    v_readfirstlane_b32 s5, v8
 ; GFX10-NEXT:    v_readfirstlane_b32 s6, v9
@@ -7611,50 +6584,35 @@ define amdgpu_ps <7 x double> @dyn_insertelement_v7f64_s_v_v(<7 x double> inreg
 ;
 ; GFX11-LABEL: dyn_insertelement_v7f64_s_v_v:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    s_mov_b32 s10, s12
-; GFX11-NEXT:    s_mov_b32 s11, s13
-; GFX11-NEXT:    s_mov_b32 s12, s14
-; GFX11-NEXT:    s_mov_b32 s13, s15
-; GFX11-NEXT:    v_dual_mov_b32 v18, s15 :: v_dual_mov_b32 v17, s14
-; GFX11-NEXT:    v_dual_mov_b32 v16, s13 :: v_dual_mov_b32 v15, s12
-; GFX11-NEXT:    v_dual_mov_b32 v14, s11 :: v_dual_mov_b32 v13, s10
-; GFX11-NEXT:    v_dual_mov_b32 v12, s9 :: v_dual_mov_b32 v11, s8
-; GFX11-NEXT:    v_dual_mov_b32 v10, s7 :: v_dual_mov_b32 v9, s6
-; GFX11-NEXT:    v_dual_mov_b32 v8, s5 :: v_dual_mov_b32 v7, s4
-; GFX11-NEXT:    v_dual_mov_b32 v6, s3 :: v_dual_mov_b32 v5, s2
-; GFX11-NEXT:    v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v3, s0
+; GFX11-NEXT:    v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
+; GFX11-NEXT:    v_dual_mov_b32 v5, s4 :: v_dual_mov_b32 v6, s5
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v2
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 1, v2
-; GFX11-NEXT:    v_cmp_eq_u32_e64 s1, 6, v2
+; GFX11-NEXT:    v_dual_mov_b32 v7, s6 :: v_dual_mov_b32 v8, s7
+; GFX11-NEXT:    v_dual_mov_b32 v9, s8 :: v_dual_mov_b32 v10, s9
 ; GFX11-NEXT:    v_dual_cndmask_b32 v3, v3, v0 :: v_dual_cndmask_b32 v4, v4, v1
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 2, v2
 ; GFX11-NEXT:    v_cndmask_b32_e64 v5, v5, v0, s0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v6, v6, v1, s0
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 3, v2
+; GFX11-NEXT:    v_dual_mov_b32 v11, s10 :: v_dual_mov_b32 v12, s11
+; GFX11-NEXT:    v_dual_mov_b32 v13, s12 :: v_dual_mov_b32 v14, s13
+; GFX11-NEXT:    v_dual_mov_b32 v15, s14 :: v_dual_mov_b32 v16, s15
 ; GFX11-NEXT:    v_dual_cndmask_b32 v7, v7, v0 :: v_dual_cndmask_b32 v8, v8, v1
-; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 4, v2
 ; GFX11-NEXT:    v_cndmask_b32_e64 v9, v9, v0, s0
+; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 4, v2
 ; GFX11-NEXT:    v_cndmask_b32_e64 v10, v10, v1, s0
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 5, v2
+; GFX11-NEXT:    v_cmp_eq_u32_e64 s1, 6, v2
 ; GFX11-NEXT:    v_readfirstlane_b32 s2, v5
 ; GFX11-NEXT:    v_dual_cndmask_b32 v11, v11, v0 :: v_dual_cndmask_b32 v2, v12, v1
-; GFX11-NEXT:    v_readfirstlane_b32 s3, v6
 ; GFX11-NEXT:    v_cndmask_b32_e64 v12, v13, v0, s0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v13, v14, v1, s0
 ; GFX11-NEXT:    v_cndmask_b32_e64 v0, v15, v0, s1
 ; GFX11-NEXT:    v_cndmask_b32_e64 v1, v16, v1, s1
 ; GFX11-NEXT:    v_readfirstlane_b32 s0, v3
 ; GFX11-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-NEXT:    v_readfirstlane_b32 s3, v6
 ; GFX11-NEXT:    v_readfirstlane_b32 s4, v7
 ; GFX11-NEXT:    v_readfirstlane_b32 s5, v8
 ; GFX11-NEXT:    v_readfirstlane_b32 s6, v9
@@ -7669,36 +6627,20 @@ define amdgpu_ps <7 x double> @dyn_insertelement_v7f64_s_v_v(<7 x double> inreg
 ;
 ; GFX8-LABEL: dyn_insertelement_v7f64_s_v_v:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    s_mov_b32 s10, s12
-; GFX8-NEXT:    s_mov_b32 s11, s13
-; GFX8-NEXT:    s_mov_b32 s12, s14
-; GFX8-NEXT:    s_mov_b32 s13, s15
-; GFX8-NEXT:    v_mov_b32_e32 v18, s15
-; GFX8-NEXT:    v_mov_b32_e32 v17, s14
-; GFX8-NEXT:    v_mov_b32_e32 v16, s13
-; GFX8-NEXT:    v_mov_b32_e32 v15, s12
-; GFX8-NEXT:    v_mov_b32_e32 v14, s11
-; GFX8-NEXT:    v_mov_b32_e32 v13, s10
-; GFX8-NEXT:    v_mov_b32_e32 v12, s9
-; GFX8-NEXT:    v_mov_b32_e32 v11, s8
-; GFX8-NEXT:    v_mov_b32_e32 v10, s7
-; GFX8-NEXT:    v_mov_b32_e32 v9, s6
-; GFX8-NEXT:    v_mov_b32_e32 v8, s5
-; GFX8-NEXT:    v_mov_b32_e32 v7, s4
-; GFX8-NEXT:    v_mov_b32_e32 v6, s3
-; GFX8-NEXT:    v_mov_b32_e32 v5, s2
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s3
+; GFX8-NEXT:    v_mov_b32_e32 v5, s4
+; GFX8-NEXT:    v_mov_b32_e32 v6, s5
+; GFX8-NEXT:    v_mov_b32_e32 v7, s6
+; GFX8-NEXT:    v_mov_b32_e32 v8, s7
+; GFX8-NEXT:    v_mov_b32_e32 v9, s8
+; GFX8-NEXT:    v_mov_b32_e32 v10, s9
+; GFX8-NEXT:    v_mov_b32_e32 v11, s10
+; GFX8-NEXT:    v_mov_b32_e32 v12, s11
+; GFX8-NEXT:    v_mov_b32_e32 v13, s12
+; GFX8-NEXT:    v_mov_b32_e32 v14, s13
+; GFX8-NEXT:    v_mov_b32_e32 v15, s14
+; GFX8-NEXT:    v_mov_b32_e32 v16, s15
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
 ; GFX8-NEXT:    v_cmp_eq_u32_e64 s[0:1], 2, v2
 ; GFX8-NEXT:    v_cmp_eq_u32_e64 s[2:3], 3, v2
@@ -8021,26 +6963,16 @@ entry:
 define amdgpu_ps <5 x double> @dyn_insertelement_v5f64_s_v_s(<5 x double> inreg %vec, double %val, i32 inreg %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v5f64_s_v_s:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s6
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s4
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s2
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v2, s0
+; GPRIDX-NEXT:    v_mov_b32_e32 v2, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s3
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s5
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s7
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s11
 ; GPRIDX-NEXT:    s_lshl_b32 s0, s12, 1
 ; GPRIDX-NEXT:    s_set_gpr_idx_on s0, gpr_idx(DST)
 ; GPRIDX-NEXT:    v_mov_b32_e32 v2, v0
@@ -8060,27 +6992,17 @@ define amdgpu_ps <5 x double> @dyn_insertelement_v5f64_s_v_s(<5 x double> inreg
 ;
 ; GFX10-LABEL: dyn_insertelement_v5f64_s_v_s:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    v_mov_b32_e32 v2, s0
+; GFX10-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX10-NEXT:    s_lshl_b32 m0, s12, 1
-; GFX10-NEXT:    v_mov_b32_e32 v11, s9
-; GFX10-NEXT:    v_mov_b32_e32 v10, s8
-; GFX10-NEXT:    v_mov_b32_e32 v9, s7
-; GFX10-NEXT:    v_mov_b32_e32 v8, s6
-; GFX10-NEXT:    v_mov_b32_e32 v7, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
+; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v4, s4
+; GFX10-NEXT:    v_mov_b32_e32 v5, s5
+; GFX10-NEXT:    v_mov_b32_e32 v6, s6
+; GFX10-NEXT:    v_mov_b32_e32 v7, s7
+; GFX10-NEXT:    v_mov_b32_e32 v8, s8
+; GFX10-NEXT:    v_mov_b32_e32 v9, s9
+; GFX10-NEXT:    v_mov_b32_e32 v10, s10
+; GFX10-NEXT:    v_mov_b32_e32 v11, s11
 ; GFX10-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX10-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v2
@@ -8097,22 +7019,12 @@ define amdgpu_ps <5 x double> @dyn_insertelement_v5f64_s_v_s(<5 x double> inreg
 ;
 ; GFX11-LABEL: dyn_insertelement_v5f64_s_v_s:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
 ; GFX11-NEXT:    s_lshl_b32 m0, s12, 1
-; GFX11-NEXT:    v_dual_mov_b32 v11, s9 :: v_dual_mov_b32 v10, s8
-; GFX11-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
-; GFX11-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
+; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v5, s5
+; GFX11-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
+; GFX11-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
+; GFX11-NEXT:    v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
 ; GFX11-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX11-NEXT:    v_movreld_b32_e32 v3, v1
 ; GFX11-NEXT:    v_readfirstlane_b32 s0, v2
@@ -8129,26 +7041,16 @@ define amdgpu_ps <5 x double> @dyn_insertelement_v5f64_s_v_s(<5 x double> inreg
 ;
 ; GFX8-LABEL: dyn_insertelement_v5f64_s_v_s:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    v_mov_b32_e32 v11, s9
-; GFX8-NEXT:    v_mov_b32_e32 v10, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s7
-; GFX8-NEXT:    v_mov_b32_e32 v8, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s3
+; GFX8-NEXT:    v_mov_b32_e32 v4, s4
+; GFX8-NEXT:    v_mov_b32_e32 v5, s5
+; GFX8-NEXT:    v_mov_b32_e32 v6, s6
+; GFX8-NEXT:    v_mov_b32_e32 v7, s7
+; GFX8-NEXT:    v_mov_b32_e32 v8, s8
+; GFX8-NEXT:    v_mov_b32_e32 v9, s9
+; GFX8-NEXT:    v_mov_b32_e32 v10, s10
+; GFX8-NEXT:    v_mov_b32_e32 v11, s11
 ; GFX8-NEXT:    s_lshl_b32 m0, s12, 1
 ; GFX8-NEXT:    v_movreld_b32_e32 v2, v0
 ; GFX8-NEXT:    v_movreld_b32_e32 v3, v1
@@ -8171,33 +7073,23 @@ entry:
 define amdgpu_ps <5 x double> @dyn_insertelement_v5f64_s_v_v(<5 x double> inreg %vec, double %val, i32 %idx) {
 ; GPRIDX-LABEL: dyn_insertelement_v5f64_s_v_v:
 ; GPRIDX:       ; %bb.0: ; %entry
-; GPRIDX-NEXT:    s_mov_b32 s0, s2
-; GPRIDX-NEXT:    s_mov_b32 s1, s3
-; GPRIDX-NEXT:    s_mov_b32 s2, s4
-; GPRIDX-NEXT:    s_mov_b32 s3, s5
-; GPRIDX-NEXT:    s_mov_b32 s4, s6
-; GPRIDX-NEXT:    s_mov_b32 s5, s7
-; GPRIDX-NEXT:    s_mov_b32 s6, s8
-; GPRIDX-NEXT:    s_mov_b32 s7, s9
-; GPRIDX-NEXT:    s_mov_b32 s8, s10
-; GPRIDX-NEXT:    s_mov_b32 s9, s11
-; GPRIDX-NEXT:    v_mov_b32_e32 v4, s1
-; GPRIDX-NEXT:    v_mov_b32_e32 v3, s0
+; GPRIDX-NEXT:    v_mov_b32_e32 v3, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v4, s3
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
-; GPRIDX-NEXT:    v_mov_b32_e32 v6, s3
-; GPRIDX-NEXT:    v_mov_b32_e32 v5, s2
+; GPRIDX-NEXT:    v_mov_b32_e32 v5, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v6, s5
 ; GPRIDX-NEXT:    v_cndmask_b32_e32 v3, v3, v0, vcc
 ; GPRIDX-NEXT:    v_cndmask_b32_e32 v4, v4, v1, vcc
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v2
-; GPRIDX-NEXT:    v_mov_b32_e32 v8, s5
-; GPRIDX-NEXT:    v_mov_b32_e32 v7, s4
+; GPRIDX-NEXT:    v_mov_b32_e32 v7, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v8, s7
 ; GPRIDX-NEXT:    v_cndmask_b32_e32 v5, v5, v0, vcc
 ; GPRIDX-NEXT:    v_cndmask_b32_e32 v6, v6, v1, vcc
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e32 vcc, 2, v2
-; GPRIDX-NEXT:    v_mov_b32_e32 v12, s9
-; GPRIDX-NEXT:    v_mov_b32_e32 v11, s8
-; GPRIDX-NEXT:    v_mov_b32_e32 v10, s7
-; GPRIDX-NEXT:    v_mov_b32_e32 v9, s6
+; GPRIDX-NEXT:    v_mov_b32_e32 v9, s8
+; GPRIDX-NEXT:    v_mov_b32_e32 v10, s9
+; GPRIDX-NEXT:    v_mov_b32_e32 v11, s10
+; GPRIDX-NEXT:    v_mov_b32_e32 v12, s11
 ; GPRIDX-NEXT:    v_cndmask_b32_e32 v7, v7, v0, vcc
 ; GPRIDX-NEXT:    v_cndmask_b32_e32 v8, v8, v1, vcc
 ; GPRIDX-NEXT:    v_cmp_eq_u32_e32 vcc, 3, v2
@@ -8220,35 +7112,25 @@ define amdgpu_ps <5 x double> @dyn_insertelement_v5f64_s_v_v(<5 x double> inreg
 ;
 ; GFX10-LABEL: dyn_insertelement_v5f64_s_v_v:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    s_mov_b32 s8, s10
-; GFX10-NEXT:    s_mov_b32 s9, s11
-; GFX10-NEXT:    v_mov_b32_e32 v11, s8
-; GFX10-NEXT:    v_mov_b32_e32 v12, s9
-; GFX10-NEXT:    v_mov_b32_e32 v10, s7
-; GFX10-NEXT:    v_mov_b32_e32 v9, s6
-; GFX10-NEXT:    v_mov_b32_e32 v8, s5
-; GFX10-NEXT:    v_mov_b32_e32 v7, s4
-; GFX10-NEXT:    v_mov_b32_e32 v6, s3
-; GFX10-NEXT:    v_mov_b32_e32 v5, s2
-; GFX10-NEXT:    v_mov_b32_e32 v4, s1
-; GFX10-NEXT:    v_mov_b32_e32 v3, s0
+; GFX10-NEXT:    v_mov_b32_e32 v3, s2
+; GFX10-NEXT:    v_mov_b32_e32 v4, s3
+; GFX10-NEXT:    v_mov_b32_e32 v5, s4
+; GFX10-NEXT:    v_mov_b32_e32 v6, s5
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v2
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 1, v2
-; GFX10-NEXT:    v_cmp_eq_u32_e64 s1, 4, v2
+; GFX10-NEXT:    v_mov_b32_e32 v7, s6
+; GFX10-NEXT:    v_mov_b32_e32 v8, s7
+; GFX10-NEXT:    v_mov_b32_e32 v9, s8
+; GFX10-NEXT:    v_mov_b32_e32 v10, s9
+; GFX10-NEXT:    v_mov_b32_e32 v11, s10
+; GFX10-NEXT:    v_mov_b32_e32 v12, s11
 ; GFX10-NEXT:    v_cndmask_b32_e32 v3, v3, v0, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e32 v4, v4, v1, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e64 v5, v5, v0, s0
 ; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 2, v2
 ; GFX10-NEXT:    v_cndmask_b32_e64 v6, v6, v1, s0
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s0, 3, v2
+; GFX10-NEXT:    v_cmp_eq_u32_e64 s1, 4, v2
 ; GFX10-NEXT:    v_readfirstlane_b32 s2, v5
 ; GFX10-NEXT:    v_cndmask_b32_e32 v7, v7, v0, vcc_lo
 ; GFX10-NEXT:    v_cndmask_b32_e32 v2, v8, v1, vcc_lo
@@ -8269,29 +7151,19 @@ define amdgpu_ps <5 x double> @dyn_insertelement_v5f64_s_v_v(<5 x double> inreg
 ;
 ; GFX11-LABEL: dyn_insertelement_v5f64_s_v_v:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_mov_b32 s0, s2
-; GFX11-NEXT:    s_mov_b32 s1, s3
-; GFX11-NEXT:    s_mov_b32 s2, s4
-; GFX11-NEXT:    s_mov_b32 s3, s5
-; GFX11-NEXT:    s_mov_b32 s4, s6
-; GFX11-NEXT:    s_mov_b32 s5, s7
-; GFX11-NEXT:    s_mov_b32 s6, s8
-; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    s_mov_b32 s8, s10
-; GFX11-NEXT:    s_mov_b32 s9, s11
-; GFX11-NEXT:    v_dual_mov_b32 v11, s8 :: v_dual_mov_b32 v12, s9
-; GFX11-NEXT:    v_dual_mov_b32 v10, s7 :: v_dual_mov_b32 v9, s6
-; GFX11-NEXT:    v_dual_mov_b32 v8, s5 :: v_dual_mov_b32 v7, s4
-; GFX11-NEXT:    v_dual_mov_b32 v6, s3 :: v_dual_mov_b32 v5, s2
-; GFX11-NEXT:    v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v3, s0
+; GFX11-NEXT:    v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
+; GFX11-NEXT:    v_dual_mov_b32 v5, s4 :: v_dual_mov_b32 v6, s5
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v2
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 1, v2
-; GFX11-NEXT:    v_cmp_eq_u32_e64 s1, 4, v2
+; GFX11-NEXT:    v_dual_mov_b32 v7, s6 :: v_dual_mov_b32 v8, s7
+; GFX11-NEXT:    v_dual_mov_b32 v9, s8 :: v_dual_mov_b32 v10, s9
+; GFX11-NEXT:    v_dual_mov_b32 v11, s10 :: v_dual_mov_b32 v12, s11
 ; GFX11-NEXT:    v_dual_cndmask_b32 v3, v3, v0 :: v_dual_cndmask_b32 v4, v4, v1
 ; GFX11-NEXT:    v_cndmask_b32_e64 v5, v5, v0, s0
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 2, v2
 ; GFX11-NEXT:    v_cndmask_b32_e64 v6, v6, v1, s0
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 3, v2
+; GFX11-NEXT:    v_cmp_eq_u32_e64 s1, 4, v2
 ; GFX11-NEXT:    v_readfirstlane_b32 s2, v5
 ; GFX11-NEXT:    v_dual_cndmask_b32 v7, v7, v0 :: v_dual_cndmask_b32 v2, v8, v1
 ; GFX11-NEXT:    v_cndmask_b32_e64 v8, v9, v0, s0
@@ -8311,33 +7183,23 @@ define amdgpu_ps <5 x double> @dyn_insertelement_v5f64_s_v_v(<5 x double> inreg
 ;
 ; GFX8-LABEL: dyn_insertelement_v5f64_s_v_v:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_mov_b32 s0, s2
-; GFX8-NEXT:    s_mov_b32 s1, s3
-; GFX8-NEXT:    s_mov_b32 s2, s4
-; GFX8-NEXT:    s_mov_b32 s3, s5
-; GFX8-NEXT:    s_mov_b32 s4, s6
-; GFX8-NEXT:    s_mov_b32 s5, s7
-; GFX8-NEXT:    s_mov_b32 s6, s8
-; GFX8-NEXT:    s_mov_b32 s7, s9
-; GFX8-NEXT:    s_mov_b32 s8, s10
-; GFX8-NEXT:    s_mov_b32 s9, s11
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s3
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
-; GFX8-NEXT:    v_mov_b32_e32 v6, s3
-; GFX8-NEXT:    v_mov_b32_e32 v5, s2
+; GFX8-NEXT:    v_mov_b32_e32 v5, s4
+; GFX8-NEXT:    v_mov_b32_e32 v6, s5
 ; GFX8-NEXT:    v_cndmask_b32_e32 v3, v3, v0, vcc
 ; GFX8-NEXT:    v_cndmask_b32_e32 v4, v4, v1, vcc
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v8, s5
-; GFX8-NEXT:    v_mov_b32_e32 v7, s4
+; GFX8-NEXT:    v_mov_b32_e32 v7, s6
+; GFX8-NEXT:    v_mov_b32_e32 v8, s7
 ; GFX8-NEXT:    v_cndmask_b32_e32 v5, v5, v0, vcc
 ; GFX8-NEXT:    v_cndmask_b32_e32 v6, v6, v1, vcc
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 2, v2
-; GFX8-NEXT:    v_mov_b32_e32 v12, s9
-; GFX8-NEXT:    v_mov_b32_e32 v11, s8
-; GFX8-NEXT:    v_mov_b32_e32 v10, s7
-; GFX8-NEXT:    v_mov_b32_e32 v9, s6
+; GFX8-NEXT:    v_mov_b32_e32 v9, s8
+; GFX8-NEXT:    v_mov_b32_e32 v10, s9
+; GFX8-NEXT:    v_mov_b32_e32 v11, s10
+; GFX8-NEXT:    v_mov_b32_e32 v12, s11
 ; GFX8-NEXT:    v_cndmask_b32_e32 v7, v7, v0, vcc
 ; GFX8-NEXT:    v_cndmask_b32_e32 v8, v8, v1, vcc
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 3, v2
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.div.fmas.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.div.fmas.ll
index e7f327bc7a80f..f5a88e87225b3 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.div.fmas.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.div.fmas.ll
@@ -1177,16 +1177,16 @@ define amdgpu_kernel void @test_div_fmas_f32_logical_cond_to_vcc(ptr addrspace(1
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    flat_load_dword v3, v[5:6] glc
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    s_add_u32 s0, s0, 8
-; GFX8-NEXT:    s_addc_u32 s1, s1, 0
+; GFX8-NEXT:    s_add_u32 s2, s0, 8
+; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    s_cmp_lg_u32 s4, 0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_cselect_b64 s[2:3], exec, 0
-; GFX8-NEXT:    s_and_b64 vcc, vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_cselect_b64 s[0:1], exec, 0
+; GFX8-NEXT:    s_and_b64 vcc, vcc, s[0:1]
+; GFX8-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_div_fmas_f32 v2, v1, v2, v3
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.intersect_ray.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.intersect_ray.ll
index ee93f413566e5..17f3364422ee1 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.intersect_ray.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.intersect_ray.ll
@@ -769,33 +769,25 @@ define amdgpu_kernel void @image_bvh_intersect_ray_nsa_reassign(ptr %p_node_ptr,
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
 ; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX11-NEXT:    s_mov_b32 s10, 0x40a00000
-; GFX11-NEXT:    s_mov_b32 s9, 4.0
-; GFX11-NEXT:    s_mov_b32 s8, 0x40400000
-; GFX11-NEXT:    s_mov_b32 s14, 0x41000000
-; GFX11-NEXT:    s_mov_b32 s13, 0x40e00000
-; GFX11-NEXT:    s_mov_b32 s12, 0x40c00000
-; GFX11-NEXT:    v_mov_b32_e32 v8, s14
-; GFX11-NEXT:    v_dual_mov_b32 v6, s12 :: v_dual_mov_b32 v5, s10
-; GFX11-NEXT:    v_mov_b32_e32 v7, s13
+; GFX11-NEXT:    v_mov_b32_e32 v8, 0x41000000
+; GFX11-NEXT:    v_mov_b32_e32 v6, 0x40c00000
+; GFX11-NEXT:    v_mov_b32_e32 v5, 0x40a00000
+; GFX11-NEXT:    v_mov_b32_e32 v7, 0x40e00000
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_lshlrev_b32 v4, 2, v0
-; GFX11-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v2, s2
-; GFX11-NEXT:    s_mov_b32 s2, 2.0
-; GFX11-NEXT:    v_mov_b32_e32 v0, s0
-; GFX11-NEXT:    s_mov_b32 s1, 1.0
-; GFX11-NEXT:    s_mov_b32 s0, 0
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v0, s0
+; GFX11-NEXT:    v_mov_b32_e32 v2, s2
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
 ; GFX11-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
 ; GFX11-NEXT:    flat_load_b32 v9, v[0:1]
 ; GFX11-NEXT:    flat_load_b32 v10, v[2:3]
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s8
-; GFX11-NEXT:    v_mov_b32_e32 v4, s9
+; GFX11-NEXT:    v_dual_mov_b32 v2, 2.0 :: v_dual_mov_b32 v3, 0x40400000
+; GFX11-NEXT:    v_dual_mov_b32 v1, 1.0 :: v_dual_mov_b32 v0, 0
+; GFX11-NEXT:    v_mov_b32_e32 v4, 4.0
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    image_bvh_intersect_ray v[0:3], [v9, v10, v[0:2], v[3:5], v[6:8]], s[4:7]
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
@@ -878,30 +870,29 @@ define amdgpu_kernel void @image_bvh_intersect_ray_a16_nsa_reassign(ptr %p_node_
 ; GFX11-TRUE16-LABEL: image_bvh_intersect_ray_a16_nsa_reassign:
 ; GFX11-TRUE16:       ; %bb.0:
 ; GFX11-TRUE16-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
-; GFX11-TRUE16-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX11-TRUE16-NEXT:    s_movk_i32 s8, 0x4600
-; GFX11-TRUE16-NEXT:    s_movk_i32 s9, 0x4700
-; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s8, s8, 0x4200
-; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s9, s9, 0x4400
 ; GFX11-TRUE16-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_lshlrev_b32 v4, 2, v0
-; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v2, s2
-; GFX11-TRUE16-NEXT:    s_mov_b32 s2, 2.0
+; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_and_b32 v0, 0x3ff, v0
+; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_lshlrev_b32 v4, 2, v0
 ; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v0, s0
-; GFX11-TRUE16-NEXT:    s_movk_i32 s3, 0x4800
-; GFX11-TRUE16-NEXT:    s_mov_b32 s1, 1.0
-; GFX11-TRUE16-NEXT:    s_mov_b32 s0, 0
-; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s10, s3, 0x4500
+; GFX11-TRUE16-NEXT:    s_movk_i32 s1, 0x4600
+; GFX11-TRUE16-NEXT:    s_movk_i32 s0, 0x4800
+; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s1, s1, 0x4200
+; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX11-TRUE16-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
-; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_3)
 ; GFX11-TRUE16-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
 ; GFX11-TRUE16-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
+; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-TRUE16-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
 ; GFX11-TRUE16-NEXT:    flat_load_b32 v6, v[0:1]
 ; GFX11-TRUE16-NEXT:    flat_load_b32 v7, v[2:3]
-; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v5, s10
-; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v3, s8 :: v_dual_mov_b32 v4, s9
+; GFX11-TRUE16-NEXT:    s_movk_i32 s2, 0x4700
+; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s0, s0, 0x4500
+; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s2, s2, 0x4400
+; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v2, 2.0 :: v_dual_mov_b32 v5, s0
+; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v1, 1.0 :: v_dual_mov_b32 v0, 0
+; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v4, s2
 ; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-TRUE16-NEXT:    image_bvh_intersect_ray v[0:3], [v6, v7, v[0:2], v[3:5]], s[4:7] a16
 ; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0)
@@ -912,29 +903,23 @@ define amdgpu_kernel void @image_bvh_intersect_ray_a16_nsa_reassign(ptr %p_node_
 ; GFX11-FAKE16:       ; %bb.0:
 ; GFX11-FAKE16-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
 ; GFX11-FAKE16-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX11-FAKE16-NEXT:    s_mov_b32 s8, 0x42004600
-; GFX11-FAKE16-NEXT:    s_mov_b32 s9, 0x44004700
-; GFX11-FAKE16-NEXT:    s_mov_b32 s10, 0x45004800
-; GFX11-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v5, s10
+; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v5, 0x45004800
 ; GFX11-FAKE16-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_lshlrev_b32 v4, 2, v0
-; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v2, s2
-; GFX11-FAKE16-NEXT:    s_mov_b32 s2, 2.0
-; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v0, s0
-; GFX11-FAKE16-NEXT:    s_mov_b32 s1, 1.0
-; GFX11-FAKE16-NEXT:    s_mov_b32 s0, 0
-; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v0, s0
+; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX11-FAKE16-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
+; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_3)
 ; GFX11-FAKE16-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
 ; GFX11-FAKE16-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
 ; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-FAKE16-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
 ; GFX11-FAKE16-NEXT:    flat_load_b32 v6, v[0:1]
 ; GFX11-FAKE16-NEXT:    flat_load_b32 v7, v[2:3]
-; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s8
-; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v4, s9
+; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v2, 2.0 :: v_dual_mov_b32 v1, 1.0
+; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v3, 0x42004600
+; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v4, 0x44004700
 ; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-FAKE16-NEXT:    image_bvh_intersect_ray v[0:3], [v6, v7, v[0:2], v[3:5]], s[4:7] a16
 ; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0)
@@ -1024,30 +1009,22 @@ define amdgpu_kernel void @image_bvh64_intersect_ray_nsa_reassign(ptr %p_ray, <4
 ; GFX11-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24
 ; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x34
 ; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX11-NEXT:    s_mov_b32 s5, 1.0
-; GFX11-NEXT:    s_mov_b32 s4, 0
-; GFX11-NEXT:    s_mov_b32 s10, 0x40a00000
-; GFX11-NEXT:    s_mov_b32 s9, 4.0
-; GFX11-NEXT:    s_mov_b32 s8, 0x40400000
-; GFX11-NEXT:    s_mov_b32 s14, 0x41000000
-; GFX11-NEXT:    s_mov_b32 s13, 0x40e00000
-; GFX11-NEXT:    s_mov_b32 s12, 0x40c00000
 ; GFX11-NEXT:    v_bfrev_b32_e32 v10, 4.0
-; GFX11-NEXT:    v_dual_mov_b32 v4, s9 :: v_dual_mov_b32 v7, s13
-; GFX11-NEXT:    v_mov_b32_e32 v6, s12
-; GFX11-NEXT:    v_dual_mov_b32 v8, s14 :: v_dual_mov_b32 v9, 0xb36211c7
-; GFX11-NEXT:    v_mov_b32_e32 v5, s10
-; GFX11-NEXT:    v_mov_b32_e32 v3, s8
+; GFX11-NEXT:    v_dual_mov_b32 v4, 4.0 :: v_dual_mov_b32 v7, 0x40e00000
+; GFX11-NEXT:    v_mov_b32_e32 v8, 0x41000000
+; GFX11-NEXT:    v_mov_b32_e32 v6, 0x40c00000
+; GFX11-NEXT:    v_mov_b32_e32 v9, 0xb36211c7
+; GFX11-NEXT:    v_mov_b32_e32 v5, 0x40a00000
+; GFX11-NEXT:    v_mov_b32_e32 v3, 0x40400000
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_lshlrev_b32 v2, 2, v0
 ; GFX11-NEXT:    v_mov_b32_e32 v0, s6
-; GFX11-NEXT:    s_mov_b32 s6, 2.0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
 ; GFX11-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
-; GFX11-NEXT:    v_mov_b32_e32 v2, s6
+; GFX11-NEXT:    v_mov_b32_e32 v2, 2.0
 ; GFX11-NEXT:    flat_load_b32 v11, v[0:1]
-; GFX11-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
+; GFX11-NEXT:    v_dual_mov_b32 v1, 1.0 :: v_dual_mov_b32 v0, 0
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    image_bvh64_intersect_ray v[0:3], [v[9:10], v11, v[0:2], v[3:5], v[6:8]], s[0:3]
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
@@ -1129,28 +1106,27 @@ define amdgpu_kernel void @image_bvh64_intersect_ray_a16_nsa_reassign(ptr %p_ray
 ; GFX11-TRUE16-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24
 ; GFX11-TRUE16-NEXT:    s_load_b128 s[0:3], s[4:5], 0x34
 ; GFX11-TRUE16-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX11-TRUE16-NEXT:    s_movk_i32 s8, 0x4600
-; GFX11-TRUE16-NEXT:    s_movk_i32 s9, 0x4700
-; GFX11-TRUE16-NEXT:    s_mov_b32 s5, 1.0
-; GFX11-TRUE16-NEXT:    s_mov_b32 s4, 0
-; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s8, s8, 0x4200
-; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s9, s9, 0x4400
+; GFX11-TRUE16-NEXT:    s_movk_i32 s5, 0x4600
+; GFX11-TRUE16-NEXT:    s_movk_i32 s4, 0x4800
+; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s5, s5, 0x4200
+; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s4, s4, 0x4500
+; GFX11-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v6, 0xb36211c6 :: v_dual_mov_b32 v5, s4
 ; GFX11-TRUE16-NEXT:    v_bfrev_b32_e32 v7, 4.0
+; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX11-TRUE16-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_lshlrev_b32 v2, 2, v0
 ; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v0, s6
-; GFX11-TRUE16-NEXT:    s_movk_i32 s7, 0x4800
-; GFX11-TRUE16-NEXT:    s_mov_b32 s6, 2.0
-; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s10, s7, 0x4500
-; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v3, s8
+; GFX11-TRUE16-NEXT:    s_movk_i32 s6, 0x4700
+; GFX11-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT:    s_pack_ll_b32_b16 s6, s6, 0x4400
 ; GFX11-TRUE16-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
 ; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-TRUE16-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
-; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v6, 0xb36211c6 :: v_dual_mov_b32 v5, s10
-; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v2, s6
+; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v2, 2.0
+; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX11-TRUE16-NEXT:    flat_load_b32 v8, v[0:1]
-; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
-; GFX11-TRUE16-NEXT:    v_mov_b32_e32 v4, s9
+; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v1, 1.0 :: v_dual_mov_b32 v0, 0
 ; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-TRUE16-NEXT:    image_bvh64_intersect_ray v[0:3], [v[6:7], v8, v[0:2], v[3:5]], s[0:3] a16
 ; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0)
@@ -1163,25 +1139,20 @@ define amdgpu_kernel void @image_bvh64_intersect_ray_a16_nsa_reassign(ptr %p_ray
 ; GFX11-FAKE16-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24
 ; GFX11-FAKE16-NEXT:    s_load_b128 s[0:3], s[4:5], 0x34
 ; GFX11-FAKE16-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX11-FAKE16-NEXT:    s_mov_b32 s5, 1.0
-; GFX11-FAKE16-NEXT:    s_mov_b32 s4, 0
-; GFX11-FAKE16-NEXT:    s_mov_b32 s8, 0x42004600
-; GFX11-FAKE16-NEXT:    s_mov_b32 s9, 0x44004700
-; GFX11-FAKE16-NEXT:    s_mov_b32 s10, 0x45004800
-; GFX11-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v6, 0xb36211c6 :: v_dual_mov_b32 v5, s10
+; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v6, 0xb36211c6
 ; GFX11-FAKE16-NEXT:    v_bfrev_b32_e32 v7, 4.0
-; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v4, s9 :: v_dual_mov_b32 v3, s8
+; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v4, 0x44004700
+; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v3, 0x42004600
+; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v5, 0x45004800
 ; GFX11-FAKE16-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_lshlrev_b32 v2, 2, v0
 ; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v0, s6
-; GFX11-FAKE16-NEXT:    s_mov_b32 s6, 2.0
 ; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-FAKE16-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
 ; GFX11-FAKE16-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
-; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v2, s6
+; GFX11-FAKE16-NEXT:    v_mov_b32_e32 v2, 2.0
 ; GFX11-FAKE16-NEXT:    flat_load_b32 v8, v[0:1]
-; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
+; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v1, 1.0 :: v_dual_mov_b32 v0, 0
 ; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-FAKE16-NEXT:    image_bvh64_intersect_ray v[0:3], [v[6:7], v8, v[0:2], v[3:5]], s[0:3] a16
 ; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mfma.gfx90a.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mfma.gfx90a.ll
index 8a6a3df315723..24018f5bdc429 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mfma.gfx90a.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mfma.gfx90a.ll
@@ -16,8 +16,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4bf16_1k(ptr addrspace(1) %arg) #
 ; GCN-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
 ; GCN-NEXT:    s_mov_b64 s[36:37], 1
 ; GCN-NEXT:    v_pk_mov_b32 v[32:33], s[36:37], s[36:37] op_sel:[0,1]
-; GCN-NEXT:    s_mov_b32 s36, 2
-; GCN-NEXT:    v_pk_mov_b32 v[34:35], s[36:37], s[36:37] op_sel:[0,1]
+; GCN-NEXT:    v_mov_b32_e32 v34, 2
+; GCN-NEXT:    v_mov_b32_e32 v35, s37
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x0
 ; GCN-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x40
@@ -67,8 +67,8 @@ define amdgpu_kernel void @test_mfma_f32_16x16x4bf16_1k(ptr addrspace(1) %arg) #
 ; GCN-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
 ; GCN-NEXT:    s_mov_b64 s[18:19], 1
 ; GCN-NEXT:    v_pk_mov_b32 v[16:17], s[18:19], s[18:19] op_sel:[0,1]
-; GCN-NEXT:    s_mov_b32 s18, 2
-; GCN-NEXT:    v_pk_mov_b32 v[18:19], s[18:19], s[18:19] op_sel:[0,1]
+; GCN-NEXT:    v_mov_b32_e32 v18, 2
+; GCN-NEXT:    v_mov_b32_e32 v19, s19
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
@@ -104,8 +104,8 @@ define amdgpu_kernel void @test_mfma_f32_4x4x4bf16_1k(ptr addrspace(1) %arg) #0
 ; GCN-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; GCN-NEXT:    s_mov_b64 s[4:5], 1
 ; GCN-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GCN-NEXT:    s_mov_b32 s4, 2
-; GCN-NEXT:    v_pk_mov_b32 v[6:7], s[4:5], s[4:5] op_sel:[0,1]
+; GCN-NEXT:    v_mov_b32_e32 v6, 2
+; GCN-NEXT:    v_mov_b32_e32 v7, s5
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
@@ -132,8 +132,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8bf16_1k(ptr addrspace(1) %arg) #
 ; GCN-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
 ; GCN-NEXT:    s_mov_b64 s[18:19], 1
 ; GCN-NEXT:    v_pk_mov_b32 v[16:17], s[18:19], s[18:19] op_sel:[0,1]
-; GCN-NEXT:    s_mov_b32 s18, 2
-; GCN-NEXT:    v_pk_mov_b32 v[18:19], s[18:19], s[18:19] op_sel:[0,1]
+; GCN-NEXT:    v_mov_b32_e32 v18, 2
+; GCN-NEXT:    v_mov_b32_e32 v19, s19
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
@@ -170,8 +170,8 @@ define amdgpu_kernel void @test_mfma_f32_16x16x16bf16_1k(ptr addrspace(1) %arg)
 ; GCN-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; GCN-NEXT:    s_mov_b64 s[4:5], 1
 ; GCN-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GCN-NEXT:    s_mov_b32 s4, 2
-; GCN-NEXT:    v_pk_mov_b32 v[6:7], s[4:5], s[4:5] op_sel:[0,1]
+; GCN-NEXT:    v_mov_b32_e32 v6, 2
+; GCN-NEXT:    v_mov_b32_e32 v7, s5
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
@@ -271,26 +271,23 @@ bb:
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_imm(ptr addrspace(1) %arg, double %a, double %b) #0 {
 ; GCN-LABEL: test_mfma_f64_16x16x4f64_imm:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GCN-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GCN-NEXT:    s_mov_b64 s[0:1], 0
-; GCN-NEXT:    s_mov_b64 s[6:7], 1.0
-; GCN-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GCN-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GCN-NEXT:    v_pk_mov_b32 v[0:1], 0, 0
+; GCN-NEXT:    v_mov_b32_e32 v6, 0
+; GCN-NEXT:    v_mov_b32_e32 v7, 0x3ff00000
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GCN-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GCN-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GCN-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GCN-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GCN-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GCN-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GCN-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GCN-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
+; GCN-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GCN-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GCN-NEXT:    s_nop 1
 ; GCN-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GCN-NEXT:    v_mov_b32_e32 v8, 0
 ; GCN-NEXT:    s_nop 15
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GCN-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GCN-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GCN-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GCN-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> <double 0.0, double 0.0, double 0.0, double 1.0>, i32 0, i32 0, i32 0)
@@ -301,27 +298,23 @@ bb:
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_lit(ptr addrspace(1) %arg, double %a, double %b) #0 {
 ; GCN-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GCN-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GCN-NEXT:    s_mov_b32 s0, 0
-; GCN-NEXT:    s_mov_b32 s1, 0x405ec000
-; GCN-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GCN-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GCN-NEXT:    v_mov_b32_e32 v0, 0
+; GCN-NEXT:    v_mov_b32_e32 v1, 0x405ec000
+; GCN-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GCN-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GCN-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GCN-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GCN-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GCN-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GCN-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GCN-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GCN-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GCN-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GCN-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GCN-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GCN-NEXT:    s_nop 1
 ; GCN-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GCN-NEXT:    v_mov_b32_e32 v8, 0
 ; GCN-NEXT:    s_nop 15
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GCN-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GCN-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GCN-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GCN-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> <double 123.0, double 123.0, double 123.0, double 123.0>, i32 0, i32 0, i32 0)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.trig.preop.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.trig.preop.ll
index 10441ce81149c..49910ab0e2b14 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.trig.preop.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.trig.preop.ll
@@ -50,8 +50,8 @@ define amdgpu_kernel void @s_trig_preop_f64(double %a, i32 %b) {
 ; CI-NEXT:    v_trig_preop_f64 v[0:1], s[0:1], v0
 ; CI-NEXT:    s_add_u32 s0, s0, 4
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_store_dword v[0:1], v0
 ; CI-NEXT:    s_waitcnt vmcnt(0)
 ; CI-NEXT:    flat_store_dword v[2:3], v1
@@ -70,8 +70,8 @@ define amdgpu_kernel void @s_trig_preop_f64(double %a, i32 %b) {
 ; VI-NEXT:    v_trig_preop_f64 v[0:1], s[0:1], v0
 ; VI-NEXT:    s_add_u32 s0, s0, 4
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dword v[0:1], v0
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    flat_store_dword v[2:3], v1
@@ -132,8 +132,8 @@ define amdgpu_kernel void @s_trig_preop_f64_imm(double %a, i32 %b) {
 ; CI-NEXT:    v_trig_preop_f64 v[0:1], s[0:1], 7
 ; CI-NEXT:    s_add_u32 s0, s0, 4
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_store_dword v[0:1], v0
 ; CI-NEXT:    s_waitcnt vmcnt(0)
 ; CI-NEXT:    flat_store_dword v[2:3], v1
@@ -150,8 +150,8 @@ define amdgpu_kernel void @s_trig_preop_f64_imm(double %a, i32 %b) {
 ; VI-NEXT:    v_trig_preop_f64 v[0:1], s[0:1], 7
 ; VI-NEXT:    s_add_u32 s0, s0, 4
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dword v[0:1], v0
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    flat_store_dword v[2:3], v1
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.atomic.cmpxchg.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.atomic.cmpxchg.ll
index e63109ecfaed0..95bc50ca5a55d 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.atomic.cmpxchg.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.atomic.cmpxchg.ll
@@ -4,10 +4,8 @@
 define amdgpu_ps i32 @test_global_p1_s32_uniform(ptr addrspace(1) inreg %ptr, i32 inreg %cmp, i32 inreg %new) {
 ; GFX9-LABEL: test_global_p1_s32_uniform:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s5, s2
-; GFX9-NEXT:    s_mov_b32 s4, s3
-; GFX9-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
@@ -36,12 +34,10 @@ define amdgpu_ps i32 @test_global_p1_s32_divergent(ptr addrspace(1) %ptr, i32 %c
 define amdgpu_ps i64 @test_global_p1_s64_uniform(ptr addrspace(1) inreg %ptr, i64 inreg %cmp, i64 inreg %new) {
 ; GFX9-LABEL: test_global_p1_s64_uniform:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s6, s2
-; GFX9-NEXT:    s_mov_b32 s7, s3
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s5
-; GFX9-NEXT:    v_mov_b32_e32 v2, s6
-; GFX9-NEXT:    v_mov_b32_e32 v3, s7
 ; GFX9-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX9-NEXT:    global_atomic_cmpswap_x2 v[0:1], v4, v[0:3], s[0:1] glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
@@ -75,12 +71,10 @@ define amdgpu_ps i64 @test_global_p1_s64_divergent(ptr addrspace(1) %ptr, i64 %c
 define amdgpu_ps i32 @test_flat_p0_s32_uniform(ptr inreg %ptr, i32 inreg %cmp, i32 inreg %new) {
 ; GFX9-LABEL: test_flat_p0_s32_uniform:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s5, s2
-; GFX9-NEXT:    s_mov_b32 s4, s3
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-NEXT:    v_mov_b32_e32 v2, s4
-; GFX9-NEXT:    v_mov_b32_e32 v3, s5
+; GFX9-NEXT:    v_mov_b32_e32 v3, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX9-NEXT:    flat_atomic_cmpswap v0, v[0:1], v[2:3] glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
@@ -108,12 +102,10 @@ define amdgpu_ps i32 @test_flat_p0_s32_divergent(ptr %ptr, i32 %cmp, i32 %new) {
 define amdgpu_ps i32 @test_flat_p0_s32_uniform_noalias_private(ptr inreg %ptr, i32 inreg %cmp, i32 inreg %new) {
 ; GFX9-LABEL: test_flat_p0_s32_uniform_noalias_private:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s5, s2
-; GFX9-NEXT:    s_mov_b32 s4, s3
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-NEXT:    v_mov_b32_e32 v2, s4
-; GFX9-NEXT:    v_mov_b32_e32 v3, s5
+; GFX9-NEXT:    v_mov_b32_e32 v3, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX9-NEXT:    flat_atomic_cmpswap v0, v[0:1], v[2:3] glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
@@ -154,14 +146,12 @@ define amdgpu_ps i32 @test_flat_p0_s32_divergent_noalias_private(ptr %ptr, i32 %
 define amdgpu_ps i64 @test_flat_p0_s64_uniform_noalias_private(ptr inreg %ptr, i64 inreg %cmp, i64 inreg %new) {
 ; GFX9-LABEL: test_flat_p0_s64_uniform_noalias_private:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s6, s2
-; GFX9-NEXT:    s_mov_b32 s7, s3
-; GFX9-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX9-NEXT:    v_mov_b32_e32 v4, s0
+; GFX9-NEXT:    v_mov_b32_e32 v5, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s5
-; GFX9-NEXT:    v_mov_b32_e32 v2, s6
-; GFX9-NEXT:    v_mov_b32_e32 v3, s7
 ; GFX9-NEXT:    flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1_vol
@@ -299,15 +289,11 @@ define amdgpu_ps i32 @test_local_p3_s32_divergent(ptr addrspace(3) %ptr, i32 %cm
 define amdgpu_ps i64 @test_local_p3_s64_uniform(ptr addrspace(3) inreg %ptr, i64 inreg %cmp, i64 inreg %new) {
 ; GFX9-LABEL: test_local_p3_s64_uniform:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s6, s1
-; GFX9-NEXT:    s_mov_b32 s7, s2
-; GFX9-NEXT:    s_mov_b32 s2, s3
-; GFX9-NEXT:    s_mov_b32 s3, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v4, s0
-; GFX9-NEXT:    v_mov_b32_e32 v0, s6
-; GFX9-NEXT:    v_mov_b32_e32 v1, s7
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    v_mov_b32_e32 v0, s1
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s3
+; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    ds_cmpst_rtn_b64 v[0:1], v4, v[0:1], v[2:3]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_readfirstlane_b32 s0, v0
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/load-unaligned.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/load-unaligned.ll
index 7d3237e792c24..255e0fb3fb0ea 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/load-unaligned.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/load-unaligned.ll
@@ -285,23 +285,23 @@ define amdgpu_ps void @test_s_load_constant_v8i32_align1(ptr addrspace(4) inreg
 ; GFX9-NEXT:    global_load_dwordx4 v[0:3], v8, s[0:1]
 ; GFX9-NEXT:    global_load_dwordx4 v[4:7], v8, s[0:1] offset:16
 ; GFX9-NEXT:    s_waitcnt vmcnt(1)
-; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
-; GFX9-NEXT:    v_readfirstlane_b32 s5, v1
-; GFX9-NEXT:    v_readfirstlane_b32 s6, v2
-; GFX9-NEXT:    v_readfirstlane_b32 s7, v3
+; GFX9-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s1, v1
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX9-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s8, v4
-; GFX9-NEXT:    v_readfirstlane_b32 s9, v5
-; GFX9-NEXT:    v_readfirstlane_b32 s10, v6
-; GFX9-NEXT:    v_readfirstlane_b32 s11, v7
-; GFX9-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9-NEXT:    v_mov_b32_e32 v1, s5
-; GFX9-NEXT:    v_mov_b32_e32 v2, s6
-; GFX9-NEXT:    v_mov_b32_e32 v3, s7
-; GFX9-NEXT:    v_mov_b32_e32 v4, s8
-; GFX9-NEXT:    v_mov_b32_e32 v5, s9
-; GFX9-NEXT:    v_mov_b32_e32 v6, s10
-; GFX9-NEXT:    v_mov_b32_e32 v7, s11
+; GFX9-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX9-NEXT:    v_readfirstlane_b32 s7, v5
+; GFX9-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX9-NEXT:    v_readfirstlane_b32 s9, v7
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s4
+; GFX9-NEXT:    v_mov_b32_e32 v3, s5
+; GFX9-NEXT:    v_mov_b32_e32 v4, s6
+; GFX9-NEXT:    v_mov_b32_e32 v5, s7
+; GFX9-NEXT:    v_mov_b32_e32 v6, s8
+; GFX9-NEXT:    v_mov_b32_e32 v7, s9
 ; GFX9-NEXT:    global_store_dwordx4 v8, v[0:3], s[2:3]
 ; GFX9-NEXT:    global_store_dwordx4 v8, v[4:7], s[2:3] offset:16
 ; GFX9-NEXT:    s_endpgm
@@ -344,23 +344,23 @@ define amdgpu_ps void @test_s_load_constant_v8i32_align1(ptr addrspace(4) inreg
 ; GFX10-NEXT:    global_load_dwordx4 v[0:3], v8, s[0:1]
 ; GFX10-NEXT:    global_load_dwordx4 v[4:7], v8, s[0:1] offset:16
 ; GFX10-NEXT:    s_waitcnt vmcnt(1)
-; GFX10-NEXT:    v_readfirstlane_b32 s4, v0
-; GFX10-NEXT:    v_readfirstlane_b32 s5, v1
-; GFX10-NEXT:    v_readfirstlane_b32 s6, v2
-; GFX10-NEXT:    v_readfirstlane_b32 s7, v3
+; GFX10-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-NEXT:    v_readfirstlane_b32 s1, v1
+; GFX10-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX10-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    v_readfirstlane_b32 s8, v4
-; GFX10-NEXT:    v_readfirstlane_b32 s9, v5
-; GFX10-NEXT:    v_readfirstlane_b32 s10, v6
-; GFX10-NEXT:    v_readfirstlane_b32 s11, v7
-; GFX10-NEXT:    v_mov_b32_e32 v0, s4
-; GFX10-NEXT:    v_mov_b32_e32 v1, s5
-; GFX10-NEXT:    v_mov_b32_e32 v2, s6
-; GFX10-NEXT:    v_mov_b32_e32 v3, s7
-; GFX10-NEXT:    v_mov_b32_e32 v4, s8
-; GFX10-NEXT:    v_mov_b32_e32 v5, s9
-; GFX10-NEXT:    v_mov_b32_e32 v6, s10
-; GFX10-NEXT:    v_mov_b32_e32 v7, s11
+; GFX10-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX10-NEXT:    v_readfirstlane_b32 s7, v5
+; GFX10-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX10-NEXT:    v_readfirstlane_b32 s9, v7
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
+; GFX10-NEXT:    v_mov_b32_e32 v4, s6
+; GFX10-NEXT:    v_mov_b32_e32 v5, s7
+; GFX10-NEXT:    v_mov_b32_e32 v6, s8
+; GFX10-NEXT:    v_mov_b32_e32 v7, s9
 ; GFX10-NEXT:    global_store_dwordx4 v8, v[0:3], s[2:3]
 ; GFX10-NEXT:    global_store_dwordx4 v8, v[4:7], s[2:3] offset:16
 ; GFX10-NEXT:    s_endpgm
@@ -372,20 +372,20 @@ define amdgpu_ps void @test_s_load_constant_v8i32_align1(ptr addrspace(4) inreg
 ; GFX11-NEXT:    global_load_b128 v[0:3], v8, s[0:1]
 ; GFX11-NEXT:    global_load_b128 v[4:7], v8, s[0:1] offset:16
 ; GFX11-NEXT:    s_waitcnt vmcnt(1)
-; GFX11-NEXT:    v_readfirstlane_b32 s4, v0
-; GFX11-NEXT:    v_readfirstlane_b32 s5, v1
-; GFX11-NEXT:    v_readfirstlane_b32 s6, v2
-; GFX11-NEXT:    v_readfirstlane_b32 s7, v3
+; GFX11-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11-NEXT:    v_readfirstlane_b32 s1, v1
+; GFX11-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX11-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-NEXT:    v_readfirstlane_b32 s8, v4
-; GFX11-NEXT:    v_readfirstlane_b32 s9, v5
-; GFX11-NEXT:    v_readfirstlane_b32 s10, v6
-; GFX11-NEXT:    v_readfirstlane_b32 s11, v7
-; GFX11-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
-; GFX11-NEXT:    v_dual_mov_b32 v2, s6 :: v_dual_mov_b32 v3, s7
-; GFX11-NEXT:    v_dual_mov_b32 v4, s8 :: v_dual_mov_b32 v5, s9
+; GFX11-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX11-NEXT:    v_readfirstlane_b32 s7, v5
+; GFX11-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX11-NEXT:    v_readfirstlane_b32 s9, v7
+; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX11-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s5
+; GFX11-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v5, s7
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_4)
-; GFX11-NEXT:    v_dual_mov_b32 v6, s10 :: v_dual_mov_b32 v7, s11
+; GFX11-NEXT:    v_dual_mov_b32 v6, s8 :: v_dual_mov_b32 v7, s9
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    global_store_b128 v8, v[0:3], s[2:3]
 ; GFX11-NEXT:    global_store_b128 v8, v[4:7], s[2:3] offset:16
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/load-uniform-in-vgpr.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/load-uniform-in-vgpr.ll
index 730443a56b003..56ca5fdf02caf 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/load-uniform-in-vgpr.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/load-uniform-in-vgpr.ll
@@ -412,8 +412,8 @@ define amdgpu_ps void @load_uniform_P1_v2i32(ptr addrspace(1) inreg %ptra, ptr a
 ; GFX7-NEXT:    v_readfirstlane_b32 s5, v5
 ; GFX7-NEXT:    s_add_i32 s0, s0, s4
 ; GFX7-NEXT:    s_add_i32 s1, s1, s5
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX7-NEXT:    buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64
 ; GFX7-NEXT:    s_endpgm
@@ -472,16 +472,16 @@ define amdgpu_ps void @load_uniform_P1_v3i32_gfx12(ptr addrspace(1) inreg %ptra,
 ; GFX7-NEXT:    s_mov_b32 s2, 0
 ; GFX7-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX7-NEXT:    v_readfirstlane_b32 s1, v3
-; GFX7-NEXT:    v_readfirstlane_b32 s6, v4
-; GFX7-NEXT:    v_readfirstlane_b32 s4, v5
-; GFX7-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX7-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX7-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX7-NEXT:    v_readfirstlane_b32 s6, v6
 ; GFX7-NEXT:    v_readfirstlane_b32 s7, v7
-; GFX7-NEXT:    s_add_i32 s4, s0, s4
-; GFX7-NEXT:    s_add_i32 s5, s1, s5
-; GFX7-NEXT:    s_add_i32 s6, s6, s7
-; GFX7-NEXT:    v_mov_b32_e32 v2, s4
-; GFX7-NEXT:    v_mov_b32_e32 v3, s5
-; GFX7-NEXT:    v_mov_b32_e32 v4, s6
+; GFX7-NEXT:    s_add_i32 s0, s0, s5
+; GFX7-NEXT:    s_add_i32 s1, s1, s6
+; GFX7-NEXT:    s_add_i32 s4, s4, s7
+; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
+; GFX7-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX7-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX7-NEXT:    buffer_store_dwordx3 v[2:4], v[0:1], s[0:3], 0 addr64
 ; GFX7-NEXT:    s_endpgm
@@ -502,9 +502,8 @@ define amdgpu_ps void @load_uniform_P1_v3i32_gfx12(ptr addrspace(1) inreg %ptra,
 ; GFX11-NEXT:    s_add_i32 s0, s0, s3
 ; GFX11-NEXT:    s_add_i32 s1, s1, s4
 ; GFX11-NEXT:    s_add_i32 s2, s2, s5
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v4, s2
-; GFX11-NEXT:    v_mov_b32_e32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX11-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX11-NEXT:    global_store_b96 v[0:1], v[2:4], off
 ; GFX11-NEXT:    s_endpgm
 ;
@@ -524,9 +523,8 @@ define amdgpu_ps void @load_uniform_P1_v3i32_gfx12(ptr addrspace(1) inreg %ptra,
 ; GFX12-NEXT:    s_add_co_i32 s0, s0, s3
 ; GFX12-NEXT:    s_add_co_i32 s1, s1, s4
 ; GFX12-NEXT:    s_add_co_i32 s2, s2, s5
-; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX12-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v4, s2
-; GFX12-NEXT:    v_mov_b32_e32 v2, s0
+; GFX12-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX12-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX12-NEXT:    global_store_b96 v[0:1], v[2:4], off
 ; GFX12-NEXT:    s_endpgm
   %a = load <3 x i32>, ptr addrspace(1) %ptra, align 2
@@ -548,20 +546,20 @@ define amdgpu_ps void @load_uniform_P1_v4i32(ptr addrspace(1) inreg %ptra, ptr a
 ; GFX7-NEXT:    s_mov_b32 s2, 0
 ; GFX7-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX7-NEXT:    v_readfirstlane_b32 s1, v3
-; GFX7-NEXT:    v_readfirstlane_b32 s6, v4
-; GFX7-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX7-NEXT:    v_readfirstlane_b32 s4, v6
-; GFX7-NEXT:    v_readfirstlane_b32 s5, v7
+; GFX7-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX7-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX7-NEXT:    v_readfirstlane_b32 s6, v6
+; GFX7-NEXT:    v_readfirstlane_b32 s7, v7
 ; GFX7-NEXT:    v_readfirstlane_b32 s8, v8
 ; GFX7-NEXT:    v_readfirstlane_b32 s9, v9
-; GFX7-NEXT:    s_add_i32 s4, s0, s4
-; GFX7-NEXT:    s_add_i32 s5, s1, s5
-; GFX7-NEXT:    s_add_i32 s6, s6, s8
-; GFX7-NEXT:    s_add_i32 s7, s7, s9
-; GFX7-NEXT:    v_mov_b32_e32 v2, s4
-; GFX7-NEXT:    v_mov_b32_e32 v3, s5
-; GFX7-NEXT:    v_mov_b32_e32 v4, s6
-; GFX7-NEXT:    v_mov_b32_e32 v5, s7
+; GFX7-NEXT:    s_add_i32 s0, s0, s6
+; GFX7-NEXT:    s_add_i32 s1, s1, s7
+; GFX7-NEXT:    s_add_i32 s4, s4, s8
+; GFX7-NEXT:    s_add_i32 s5, s5, s9
+; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
+; GFX7-NEXT:    v_mov_b32_e32 v4, s4
+; GFX7-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX7-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX7-NEXT:    buffer_store_dwordx4 v[2:5], v[0:1], s[0:3], 0 addr64
 ; GFX7-NEXT:    s_endpgm
@@ -585,9 +583,8 @@ define amdgpu_ps void @load_uniform_P1_v4i32(ptr addrspace(1) inreg %ptra, ptr a
 ; GFX11-NEXT:    s_add_i32 s1, s1, s5
 ; GFX11-NEXT:    s_add_i32 s2, s2, s6
 ; GFX11-NEXT:    s_add_i32 s3, s3, s7
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX11-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX11-NEXT:    s_endpgm
 ;
@@ -610,9 +607,8 @@ define amdgpu_ps void @load_uniform_P1_v4i32(ptr addrspace(1) inreg %ptra, ptr a
 ; GFX12-NEXT:    s_add_co_i32 s1, s1, s5
 ; GFX12-NEXT:    s_add_co_i32 s2, s2, s6
 ; GFX12-NEXT:    s_add_co_i32 s3, s3, s7
-; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX12-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX12-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
-; GFX12-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
 ; GFX12-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX12-NEXT:    s_endpgm
   %a = load <4 x i32>, ptr addrspace(1) %ptra, align 2
@@ -706,10 +702,10 @@ define amdgpu_ps void @load_uniform_P1_v8i32(ptr addrspace(1) inreg %ptra, ptr a
 ; GFX11-NEXT:    s_add_i32 s5, s5, s13
 ; GFX11-NEXT:    s_add_i32 s6, s6, s14
 ; GFX11-NEXT:    s_add_i32 s7, s7, s15
-; GFX11-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
-; GFX11-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX11-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
+; GFX11-NEXT:    v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s5
+; GFX11-NEXT:    v_dual_mov_b32 v8, s6 :: v_dual_mov_b32 v9, s7
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[6:9], off offset:16
@@ -749,10 +745,10 @@ define amdgpu_ps void @load_uniform_P1_v8i32(ptr addrspace(1) inreg %ptra, ptr a
 ; GFX12-NEXT:    s_add_co_i32 s5, s5, s13
 ; GFX12-NEXT:    s_add_co_i32 s6, s6, s14
 ; GFX12-NEXT:    s_add_co_i32 s7, s7, s15
-; GFX12-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
-; GFX12-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
-; GFX12-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX12-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
+; GFX12-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX12-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
+; GFX12-NEXT:    v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s5
+; GFX12-NEXT:    v_dual_mov_b32 v8, s6 :: v_dual_mov_b32 v9, s7
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX12-NEXT:    global_store_b128 v[0:1], v[6:9], off offset:16
@@ -915,18 +911,18 @@ define amdgpu_ps void @load_uniform_P1_v16i32(ptr addrspace(1) inreg %ptra, ptr
 ; GFX11-NEXT:    s_add_i32 s9, s9, s25
 ; GFX11-NEXT:    s_add_i32 s10, s10, s26
 ; GFX11-NEXT:    s_add_i32 s11, s11, s27
-; GFX11-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX11-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
 ; GFX11-NEXT:    s_add_i32 s12, s12, s28
 ; GFX11-NEXT:    s_add_i32 s13, s13, s29
 ; GFX11-NEXT:    s_add_i32 s14, s14, s30
 ; GFX11-NEXT:    s_add_i32 s15, s15, s31
-; GFX11-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
-; GFX11-NEXT:    v_dual_mov_b32 v13, s11 :: v_dual_mov_b32 v12, s10
-; GFX11-NEXT:    v_dual_mov_b32 v11, s9 :: v_dual_mov_b32 v10, s8
-; GFX11-NEXT:    v_dual_mov_b32 v17, s15 :: v_dual_mov_b32 v16, s14
-; GFX11-NEXT:    v_dual_mov_b32 v15, s13 :: v_dual_mov_b32 v14, s12
+; GFX11-NEXT:    v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s5
+; GFX11-NEXT:    v_dual_mov_b32 v8, s6 :: v_dual_mov_b32 v9, s7
+; GFX11-NEXT:    v_dual_mov_b32 v10, s8 :: v_dual_mov_b32 v11, s9
+; GFX11-NEXT:    v_dual_mov_b32 v12, s10 :: v_dual_mov_b32 v13, s11
+; GFX11-NEXT:    v_dual_mov_b32 v14, s12 :: v_dual_mov_b32 v15, s13
+; GFX11-NEXT:    v_dual_mov_b32 v16, s14 :: v_dual_mov_b32 v17, s15
 ; GFX11-NEXT:    s_clause 0x3
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[6:9], off offset:16
@@ -994,18 +990,18 @@ define amdgpu_ps void @load_uniform_P1_v16i32(ptr addrspace(1) inreg %ptra, ptr
 ; GFX12-NEXT:    s_add_co_i32 s9, s9, s25
 ; GFX12-NEXT:    s_add_co_i32 s10, s10, s26
 ; GFX12-NEXT:    s_add_co_i32 s11, s11, s27
-; GFX12-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
-; GFX12-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX12-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX12-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
 ; GFX12-NEXT:    s_add_co_i32 s12, s12, s28
 ; GFX12-NEXT:    s_add_co_i32 s13, s13, s29
 ; GFX12-NEXT:    s_add_co_i32 s14, s14, s30
 ; GFX12-NEXT:    s_add_co_i32 s15, s15, s31
-; GFX12-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX12-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
-; GFX12-NEXT:    v_dual_mov_b32 v13, s11 :: v_dual_mov_b32 v12, s10
-; GFX12-NEXT:    v_dual_mov_b32 v11, s9 :: v_dual_mov_b32 v10, s8
-; GFX12-NEXT:    v_dual_mov_b32 v17, s15 :: v_dual_mov_b32 v16, s14
-; GFX12-NEXT:    v_dual_mov_b32 v15, s13 :: v_dual_mov_b32 v14, s12
+; GFX12-NEXT:    v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s5
+; GFX12-NEXT:    v_dual_mov_b32 v8, s6 :: v_dual_mov_b32 v9, s7
+; GFX12-NEXT:    v_dual_mov_b32 v10, s8 :: v_dual_mov_b32 v11, s9
+; GFX12-NEXT:    v_dual_mov_b32 v12, s10 :: v_dual_mov_b32 v13, s11
+; GFX12-NEXT:    v_dual_mov_b32 v14, s12 :: v_dual_mov_b32 v15, s13
+; GFX12-NEXT:    v_dual_mov_b32 v16, s14 :: v_dual_mov_b32 v17, s15
 ; GFX12-NEXT:    s_clause 0x3
 ; GFX12-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX12-NEXT:    global_store_b128 v[0:1], v[6:9], off offset:16
@@ -1618,8 +1614,8 @@ define amdgpu_ps void @load_uniform_P4_v2i32(ptr addrspace(4) inreg %ptra, ptr a
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_i32 s0, s4, s0
 ; GFX7-NEXT:    s_add_i32 s1, s5, s1
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX7-NEXT:    buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64
 ; GFX7-NEXT:    s_endpgm
@@ -1676,12 +1672,12 @@ define amdgpu_ps void @load_uniform_P4_v3i32_gfx12(ptr addrspace(4) inreg %ptra,
 ; GFX7-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_readfirstlane_b32 s7, v4
-; GFX7-NEXT:    s_add_i32 s4, s0, s4
-; GFX7-NEXT:    s_add_i32 s5, s1, s5
-; GFX7-NEXT:    s_add_i32 s6, s7, s6
-; GFX7-NEXT:    v_mov_b32_e32 v2, s4
-; GFX7-NEXT:    v_mov_b32_e32 v3, s5
-; GFX7-NEXT:    v_mov_b32_e32 v4, s6
+; GFX7-NEXT:    s_add_i32 s0, s0, s4
+; GFX7-NEXT:    s_add_i32 s1, s1, s5
+; GFX7-NEXT:    s_add_i32 s4, s7, s6
+; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
+; GFX7-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX7-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX7-NEXT:    buffer_store_dwordx3 v[2:4], v[0:1], s[0:3], 0 addr64
 ; GFX7-NEXT:    s_endpgm
@@ -1698,9 +1694,8 @@ define amdgpu_ps void @load_uniform_P4_v3i32_gfx12(ptr addrspace(4) inreg %ptra,
 ; GFX11-NEXT:    s_add_i32 s0, s3, s0
 ; GFX11-NEXT:    s_add_i32 s1, s4, s1
 ; GFX11-NEXT:    s_add_i32 s2, s5, s2
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v4, s2
-; GFX11-NEXT:    v_mov_b32_e32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX11-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX11-NEXT:    global_store_b96 v[0:1], v[2:4], off
 ; GFX11-NEXT:    s_endpgm
 ;
@@ -1717,9 +1712,8 @@ define amdgpu_ps void @load_uniform_P4_v3i32_gfx12(ptr addrspace(4) inreg %ptra,
 ; GFX12-NEXT:    s_add_co_i32 s0, s3, s0
 ; GFX12-NEXT:    s_add_co_i32 s1, s4, s1
 ; GFX12-NEXT:    s_add_co_i32 s2, s5, s2
-; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX12-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v4, s2
-; GFX12-NEXT:    v_mov_b32_e32 v2, s0
+; GFX12-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX12-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX12-NEXT:    global_store_b96 v[0:1], v[2:4], off
 ; GFX12-NEXT:    s_endpgm
   %a = load <3 x i32>, ptr addrspace(4) %ptra, align 2
@@ -1744,14 +1738,14 @@ define amdgpu_ps void @load_uniform_P4_v4i32(ptr addrspace(4) inreg %ptra, ptr a
 ; GFX7-NEXT:    v_readfirstlane_b32 s8, v4
 ; GFX7-NEXT:    v_readfirstlane_b32 s9, v5
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-NEXT:    s_add_i32 s4, s0, s4
-; GFX7-NEXT:    s_add_i32 s5, s1, s5
-; GFX7-NEXT:    s_add_i32 s6, s8, s6
-; GFX7-NEXT:    s_add_i32 s7, s9, s7
-; GFX7-NEXT:    v_mov_b32_e32 v2, s4
-; GFX7-NEXT:    v_mov_b32_e32 v3, s5
-; GFX7-NEXT:    v_mov_b32_e32 v4, s6
-; GFX7-NEXT:    v_mov_b32_e32 v5, s7
+; GFX7-NEXT:    s_add_i32 s0, s0, s4
+; GFX7-NEXT:    s_add_i32 s1, s1, s5
+; GFX7-NEXT:    s_add_i32 s4, s8, s6
+; GFX7-NEXT:    s_add_i32 s5, s9, s7
+; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
+; GFX7-NEXT:    v_mov_b32_e32 v4, s4
+; GFX7-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX7-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX7-NEXT:    buffer_store_dwordx4 v[2:5], v[0:1], s[0:3], 0 addr64
 ; GFX7-NEXT:    s_endpgm
@@ -1771,9 +1765,8 @@ define amdgpu_ps void @load_uniform_P4_v4i32(ptr addrspace(4) inreg %ptra, ptr a
 ; GFX11-NEXT:    s_add_i32 s1, s5, s1
 ; GFX11-NEXT:    s_add_i32 s2, s6, s2
 ; GFX11-NEXT:    s_add_i32 s3, s7, s3
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX11-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX11-NEXT:    s_endpgm
 ;
@@ -1792,9 +1785,8 @@ define amdgpu_ps void @load_uniform_P4_v4i32(ptr addrspace(4) inreg %ptra, ptr a
 ; GFX12-NEXT:    s_add_co_i32 s1, s5, s1
 ; GFX12-NEXT:    s_add_co_i32 s2, s6, s2
 ; GFX12-NEXT:    s_add_co_i32 s3, s7, s3
-; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX12-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX12-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
-; GFX12-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
 ; GFX12-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX12-NEXT:    s_endpgm
   %a = load <4 x i32>, ptr addrspace(4) %ptra, align 2
@@ -1872,10 +1864,10 @@ define amdgpu_ps void @load_uniform_P4_v8i32(ptr addrspace(4) inreg %ptra, ptr a
 ; GFX11-NEXT:    s_add_i32 s5, s13, s5
 ; GFX11-NEXT:    s_add_i32 s6, s14, s6
 ; GFX11-NEXT:    s_add_i32 s7, s15, s7
-; GFX11-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
-; GFX11-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX11-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
+; GFX11-NEXT:    v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s5
+; GFX11-NEXT:    v_dual_mov_b32 v8, s6 :: v_dual_mov_b32 v9, s7
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[6:9], off offset:16
@@ -1907,10 +1899,10 @@ define amdgpu_ps void @load_uniform_P4_v8i32(ptr addrspace(4) inreg %ptra, ptr a
 ; GFX12-NEXT:    s_add_co_i32 s5, s13, s5
 ; GFX12-NEXT:    s_add_co_i32 s6, s14, s6
 ; GFX12-NEXT:    s_add_co_i32 s7, s15, s7
-; GFX12-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
-; GFX12-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
-; GFX12-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX12-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
+; GFX12-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX12-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
+; GFX12-NEXT:    v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s5
+; GFX12-NEXT:    v_dual_mov_b32 v8, s6 :: v_dual_mov_b32 v9, s7
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX12-NEXT:    global_store_b128 v[0:1], v[6:9], off offset:16
@@ -2036,18 +2028,18 @@ define amdgpu_ps void @load_uniform_P4_v16i32(ptr addrspace(4) inreg %ptra, ptr
 ; GFX11-NEXT:    s_add_i32 s9, s25, s9
 ; GFX11-NEXT:    s_add_i32 s10, s26, s10
 ; GFX11-NEXT:    s_add_i32 s11, s27, s11
-; GFX11-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX11-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
 ; GFX11-NEXT:    s_add_i32 s12, s28, s12
 ; GFX11-NEXT:    s_add_i32 s13, s29, s13
 ; GFX11-NEXT:    s_add_i32 s14, s30, s14
 ; GFX11-NEXT:    s_add_i32 s15, s31, s15
-; GFX11-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX11-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
-; GFX11-NEXT:    v_dual_mov_b32 v13, s11 :: v_dual_mov_b32 v12, s10
-; GFX11-NEXT:    v_dual_mov_b32 v11, s9 :: v_dual_mov_b32 v10, s8
-; GFX11-NEXT:    v_dual_mov_b32 v17, s15 :: v_dual_mov_b32 v16, s14
-; GFX11-NEXT:    v_dual_mov_b32 v15, s13 :: v_dual_mov_b32 v14, s12
+; GFX11-NEXT:    v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s5
+; GFX11-NEXT:    v_dual_mov_b32 v8, s6 :: v_dual_mov_b32 v9, s7
+; GFX11-NEXT:    v_dual_mov_b32 v10, s8 :: v_dual_mov_b32 v11, s9
+; GFX11-NEXT:    v_dual_mov_b32 v12, s10 :: v_dual_mov_b32 v13, s11
+; GFX11-NEXT:    v_dual_mov_b32 v14, s12 :: v_dual_mov_b32 v15, s13
+; GFX11-NEXT:    v_dual_mov_b32 v16, s14 :: v_dual_mov_b32 v17, s15
 ; GFX11-NEXT:    s_clause 0x3
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[6:9], off offset:16
@@ -2097,18 +2089,18 @@ define amdgpu_ps void @load_uniform_P4_v16i32(ptr addrspace(4) inreg %ptra, ptr
 ; GFX12-NEXT:    s_add_co_i32 s9, s25, s9
 ; GFX12-NEXT:    s_add_co_i32 s10, s26, s10
 ; GFX12-NEXT:    s_add_co_i32 s11, s27, s11
-; GFX12-NEXT:    v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
-; GFX12-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX12-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX12-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
 ; GFX12-NEXT:    s_add_co_i32 s12, s28, s12
 ; GFX12-NEXT:    s_add_co_i32 s13, s29, s13
 ; GFX12-NEXT:    s_add_co_i32 s14, s30, s14
 ; GFX12-NEXT:    s_add_co_i32 s15, s31, s15
-; GFX12-NEXT:    v_dual_mov_b32 v9, s7 :: v_dual_mov_b32 v8, s6
-; GFX12-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v6, s4
-; GFX12-NEXT:    v_dual_mov_b32 v13, s11 :: v_dual_mov_b32 v12, s10
-; GFX12-NEXT:    v_dual_mov_b32 v11, s9 :: v_dual_mov_b32 v10, s8
-; GFX12-NEXT:    v_dual_mov_b32 v17, s15 :: v_dual_mov_b32 v16, s14
-; GFX12-NEXT:    v_dual_mov_b32 v15, s13 :: v_dual_mov_b32 v14, s12
+; GFX12-NEXT:    v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s5
+; GFX12-NEXT:    v_dual_mov_b32 v8, s6 :: v_dual_mov_b32 v9, s7
+; GFX12-NEXT:    v_dual_mov_b32 v10, s8 :: v_dual_mov_b32 v11, s9
+; GFX12-NEXT:    v_dual_mov_b32 v12, s10 :: v_dual_mov_b32 v13, s11
+; GFX12-NEXT:    v_dual_mov_b32 v14, s12 :: v_dual_mov_b32 v15, s13
+; GFX12-NEXT:    v_dual_mov_b32 v16, s14 :: v_dual_mov_b32 v17, s15
 ; GFX12-NEXT:    s_clause 0x3
 ; GFX12-NEXT:    global_store_b128 v[0:1], v[2:5], off
 ; GFX12-NEXT:    global_store_b128 v[0:1], v[6:9], off offset:16
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/mad.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/mad.ll
index 89c023539aa74..e56c32ce704fe 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/mad.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/mad.ll
@@ -13,9 +13,9 @@ define amdgpu_ps void @mad_i64_uniform(i64 inreg %a, i64 inreg %b, ptr addrspace
 ; GFX7-NEXT:    s_mul_i32 s1, s1, s2
 ; GFX7-NEXT:    v_readfirstlane_b32 s5, v2
 ; GFX7-NEXT:    s_add_u32 s0, s0, s5
-; GFX7-NEXT:    s_add_u32 s5, s1, s0
+; GFX7-NEXT:    s_add_u32 s0, s1, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s4
-; GFX7-NEXT:    v_mov_b32_e32 v3, s5
+; GFX7-NEXT:    v_mov_b32_e32 v3, s0
 ; GFX7-NEXT:    s_mov_b32 s2, 0
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_mov_b64 s[0:1], 0
@@ -61,8 +61,8 @@ define amdgpu_ps void @mad_u64_u32_uniform_carry(i32 inreg %a, i32 inreg %b, i64
 ; GFX7-NEXT:    s_mov_b32 s2, 0
 ; GFX7-NEXT:    v_readfirstlane_b32 s1, v2
 ; GFX7-NEXT:    s_addc_u32 s1, s1, s3
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX7-NEXT:    buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64
@@ -81,11 +81,11 @@ define amdgpu_ps void @mad_u64_u32_uniform_carry(i32 inreg %a, i32 inreg %b, i64
 ; GFX11-LABEL: mad_u64_u32_uniform_carry:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_mul_i32 s4, s0, s1
-; GFX11-NEXT:    s_mul_hi_u32 s1, s0, s1
-; GFX11-NEXT:    s_add_u32 s0, s4, s2
-; GFX11-NEXT:    s_addc_u32 s1, s1, s3
+; GFX11-NEXT:    s_mul_hi_u32 s0, s0, s1
+; GFX11-NEXT:    s_add_u32 s1, s4, s2
+; GFX11-NEXT:    s_addc_u32 s0, s0, s3
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX11-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v3, s0
 ; GFX11-NEXT:    global_store_b64 v[0:1], v[2:3], off
 ; GFX11-NEXT:    s_endpgm
   %a.ext = zext i32 %a to i64
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/mul.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/mul.ll
index 51209640b185b..b593c8af83e34 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/mul.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/mul.ll
@@ -3867,11 +3867,11 @@ define amdgpu_kernel void @s_mul_u64_zext_with_sregs(ptr addrspace(1) %out, ptr
 ; GFX7-NEXT:    s_mov_b32 s2, -1
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mul_hi_u32 v0, s3, v0
-; GFX7-NEXT:    s_mul_i32 s4, s3, 0x50
+; GFX7-NEXT:    s_mulk_i32 s3, 0x50
+; GFX7-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX7-NEXT:    v_mov_b32_e32 v0, s3
+; GFX7-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7-NEXT:    v_readfirstlane_b32 s5, v0
-; GFX7-NEXT:    v_mov_b32_e32 v0, s4
-; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX7-NEXT:    s_endpgm
 ;
@@ -3897,12 +3897,12 @@ define amdgpu_kernel void @s_mul_u64_zext_with_sregs(ptr addrspace(1) %out, ptr
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s3, s[2:3], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_mul_i32 s2, s3, 0x50
-; GFX9-NEXT:    s_mul_hi_u32 s3, s3, 0x50
-; GFX9-NEXT:    v_mov_b32_e32 v0, s2
-; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    s_mul_i32 s3, s2, 0x50
+; GFX9-NEXT:    s_mul_hi_u32 s2, s2, 0x50
+; GFX9-NEXT:    v_mov_b32_e32 v0, s3
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
@@ -3911,12 +3911,12 @@ define amdgpu_kernel void @s_mul_u64_zext_with_sregs(ptr addrspace(1) %out, ptr
 ; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX10-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    s_load_dword s3, s[2:3], 0x0
+; GFX10-NEXT:    s_load_dword s2, s[2:3], 0x0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    s_mul_i32 s2, s3, 0x50
-; GFX10-NEXT:    s_mul_hi_u32 s3, s3, 0x50
-; GFX10-NEXT:    v_mov_b32_e32 v0, s2
-; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    s_mul_i32 s3, s2, 0x50
+; GFX10-NEXT:    s_mul_hi_u32 s2, s2, 0x50
+; GFX10-NEXT:    v_mov_b32_e32 v0, s3
+; GFX10-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX10-NEXT:    s_endpgm
 ;
@@ -3925,11 +3925,11 @@ define amdgpu_kernel void @s_mul_u64_zext_with_sregs(ptr addrspace(1) %out, ptr
 ; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
 ; GFX11-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    s_load_b32 s3, s[2:3], 0x0
+; GFX11-NEXT:    s_load_b32 s2, s[2:3], 0x0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    s_mul_i32 s2, s3, 0x50
-; GFX11-NEXT:    s_mul_hi_u32 s3, s3, 0x50
-; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-NEXT:    s_mul_i32 s3, s2, 0x50
+; GFX11-NEXT:    s_mul_hi_u32 s2, s2, 0x50
+; GFX11-NEXT:    v_dual_mov_b32 v0, s3 :: v_dual_mov_b32 v1, s2
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX11-NEXT:    s_endpgm
 ;
@@ -4091,13 +4091,13 @@ define amdgpu_kernel void @s_mul_u64_sext_with_sregs(ptr addrspace(1) %out, ptr
 ; GFX7-NEXT:    s_mov_b32 s2, -1
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mul_hi_u32 v0, s3, v0
-; GFX7-NEXT:    s_ashr_i32 s5, s3, 31
-; GFX7-NEXT:    s_mul_i32 s4, s3, 0x50
-; GFX7-NEXT:    s_mulk_i32 s5, 0x50
-; GFX7-NEXT:    v_readfirstlane_b32 s3, v0
-; GFX7-NEXT:    s_add_u32 s5, s5, s3
-; GFX7-NEXT:    v_mov_b32_e32 v0, s4
-; GFX7-NEXT:    v_mov_b32_e32 v1, s5
+; GFX7-NEXT:    s_ashr_i32 s4, s3, 31
+; GFX7-NEXT:    s_mulk_i32 s4, 0x50
+; GFX7-NEXT:    s_mulk_i32 s3, 0x50
+; GFX7-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX7-NEXT:    s_add_u32 s4, s4, s5
+; GFX7-NEXT:    v_mov_b32_e32 v0, s3
+; GFX7-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX7-NEXT:    s_endpgm
@@ -4127,15 +4127,15 @@ define amdgpu_kernel void @s_mul_u64_sext_with_sregs(ptr addrspace(1) %out, ptr
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s3, s[2:3], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_ashr_i32 s4, s3, 31
-; GFX9-NEXT:    s_mul_i32 s2, s3, 0x50
-; GFX9-NEXT:    s_mul_hi_u32 s3, s3, 0x50
-; GFX9-NEXT:    s_mulk_i32 s4, 0x50
-; GFX9-NEXT:    s_add_u32 s3, s4, s3
-; GFX9-NEXT:    v_mov_b32_e32 v0, s2
-; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    s_ashr_i32 s3, s2, 31
+; GFX9-NEXT:    s_mul_i32 s4, s2, 0x50
+; GFX9-NEXT:    s_mul_hi_u32 s2, s2, 0x50
+; GFX9-NEXT:    s_mulk_i32 s3, 0x50
+; GFX9-NEXT:    s_add_u32 s2, s3, s2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/or.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/or.ll
index 0a12cd16159b0..41595685896c9 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/or.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/or.ll
@@ -595,14 +595,13 @@ define amdgpu_kernel void @s_or_u64_zext_with_sregs(ptr addrspace(1) %out, ptr a
 ; GFX7-LABEL: s_or_u64_zext_with_sregs:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
-; GFX7-NEXT:    s_mov_b32 s5, 0
+; GFX7-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_load_dword s3, s[2:3], 0x0
 ; GFX7-NEXT:    s_mov_b32 s2, -1
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-NEXT:    s_or_b32 s4, s3, 0x50
-; GFX7-NEXT:    v_mov_b32_e32 v0, s4
-; GFX7-NEXT:    v_mov_b32_e32 v1, s5
+; GFX7-NEXT:    s_or_b32 s3, s3, 0x50
+; GFX7-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX7-NEXT:    s_endpgm
@@ -610,42 +609,39 @@ define amdgpu_kernel void @s_or_u64_zext_with_sregs(ptr addrspace(1) %out, ptr a
 ; GFX8-LABEL: s_or_u64_zext_with_sregs:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    s_load_dword s2, s[2:3], 0x0
-; GFX8-NEXT:    s_mov_b32 s3, 0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_or_b32 s2, s2, 0x50
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: s_or_u64_zext_with_sregs:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    s_load_dword s2, s[2:3], 0x0
-; GFX9-NEXT:    s_mov_b32 s3, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    s_or_b32 s2, s2, 0x50
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s2
-; GFX9-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: s_or_u64_zext_with_sregs:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_load_dword s2, s[2:3], 0x0
-; GFX10-NEXT:    s_mov_b32 s3, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_or_b32 s2, s2, 0x50
-; GFX10-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX10-NEXT:    s_endpgm
@@ -656,10 +652,9 @@ define amdgpu_kernel void @s_or_u64_zext_with_sregs(ptr addrspace(1) %out, ptr a
 ; GFX11-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    s_load_b32 s2, s[2:3], 0x0
-; GFX11-NEXT:    s_mov_b32 s3, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    s_or_b32 s2, s2, 0x50
-; GFX11-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v0, s2
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX11-NEXT:    s_endpgm
 ;
@@ -669,11 +664,10 @@ define amdgpu_kernel void @s_or_u64_zext_with_sregs(ptr addrspace(1) %out, ptr a
 ; GFX12-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-NEXT:    s_load_b32 s2, s[2:3], 0x0
-; GFX12-NEXT:    s_mov_b32 s3, 0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-NEXT:    s_or_b32 s2, s2, 0x50
 ; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX12-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v0, s2
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
 ; GFX12-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX12-NEXT:    s_endpgm
   %val = load i32, ptr addrspace(1) %in, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/sdivrem.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/sdivrem.ll
index 61bd7cb51f0d0..b7120b5f2e39b 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/sdivrem.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/sdivrem.ll
@@ -736,10 +736,10 @@ define amdgpu_kernel void @sdivrem_v2i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX8-NEXT:    s_xor_b32 s6, s6, s8
 ; GFX8-NEXT:    v_cvt_f32_u32_e32 v0, s6
 ; GFX8-NEXT:    s_sub_i32 s9, 0, s6
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s3
+; GFX8-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX8-NEXT:    v_readfirstlane_b32 s10, v0
@@ -775,54 +775,54 @@ define amdgpu_kernel void @sdivrem_v2i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX8-NEXT:    s_add_i32 s7, s7, s6
 ; GFX8-NEXT:    s_xor_b32 s7, s7, s6
 ; GFX8-NEXT:    v_cvt_f32_u32_e32 v0, s7
-; GFX8-NEXT:    s_ashr_i32 s3, s5, 31
-; GFX8-NEXT:    s_add_i32 s5, s5, s3
-; GFX8-NEXT:    s_xor_b32 s5, s5, s3
+; GFX8-NEXT:    s_sub_i32 s2, 0, s7
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v4, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    s_sub_i32 s1, 0, s7
 ; GFX8-NEXT:    s_xor_b32 s0, s9, s8
+; GFX8-NEXT:    s_ashr_i32 s1, s5, 31
 ; GFX8-NEXT:    v_mul_f32_e32 v4, 0x4f7ffffe, v4
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v4, v4
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v4
-; GFX8-NEXT:    s_mul_i32 s1, s1, s2
-; GFX8-NEXT:    v_mul_hi_u32 v4, v4, s1
-; GFX8-NEXT:    s_xor_b32 s1, s10, s0
-; GFX8-NEXT:    s_sub_i32 s0, s1, s0
+; GFX8-NEXT:    s_add_i32 s5, s5, s1
+; GFX8-NEXT:    s_xor_b32 s5, s5, s1
+; GFX8-NEXT:    v_readfirstlane_b32 s3, v4
+; GFX8-NEXT:    s_mul_i32 s2, s2, s3
+; GFX8-NEXT:    v_mul_hi_u32 v4, v4, s2
+; GFX8-NEXT:    s_xor_b32 s2, s10, s0
+; GFX8-NEXT:    s_sub_i32 s0, s2, s0
 ; GFX8-NEXT:    v_readfirstlane_b32 s8, v4
-; GFX8-NEXT:    s_add_i32 s2, s2, s8
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    s_add_i32 s3, s3, s8
+; GFX8-NEXT:    v_mov_b32_e32 v4, s3
 ; GFX8-NEXT:    v_mul_hi_u32 v4, s5, v4
-; GFX8-NEXT:    s_xor_b32 s2, s4, s9
-; GFX8-NEXT:    s_sub_i32 s2, s2, s9
-; GFX8-NEXT:    v_readfirstlane_b32 s1, v4
-; GFX8-NEXT:    s_mul_i32 s4, s1, s7
+; GFX8-NEXT:    s_xor_b32 s3, s4, s9
+; GFX8-NEXT:    s_sub_i32 s2, s3, s9
+; GFX8-NEXT:    v_mov_b32_e32 v6, s2
+; GFX8-NEXT:    v_readfirstlane_b32 s3, v4
+; GFX8-NEXT:    s_mul_i32 s4, s3, s7
 ; GFX8-NEXT:    s_sub_i32 s4, s5, s4
 ; GFX8-NEXT:    s_cmp_ge_u32 s4, s7
 ; GFX8-NEXT:    s_cselect_b32 s5, 1, 0
-; GFX8-NEXT:    s_add_i32 s8, s1, 1
+; GFX8-NEXT:    s_add_i32 s8, s3, 1
 ; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
-; GFX8-NEXT:    s_cselect_b32 s1, s8, s1
+; GFX8-NEXT:    s_cselect_b32 s3, s8, s3
 ; GFX8-NEXT:    s_sub_i32 s8, s4, s7
 ; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
 ; GFX8-NEXT:    s_cselect_b32 s4, s8, s4
 ; GFX8-NEXT:    s_cmp_ge_u32 s4, s7
 ; GFX8-NEXT:    s_cselect_b32 s5, 1, 0
-; GFX8-NEXT:    s_add_i32 s8, s1, 1
+; GFX8-NEXT:    s_add_i32 s8, s3, 1
 ; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
-; GFX8-NEXT:    s_cselect_b32 s1, s8, s1
+; GFX8-NEXT:    s_cselect_b32 s3, s8, s3
 ; GFX8-NEXT:    s_sub_i32 s7, s4, s7
 ; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
 ; GFX8-NEXT:    s_cselect_b32 s4, s7, s4
-; GFX8-NEXT:    s_xor_b32 s5, s3, s6
-; GFX8-NEXT:    s_xor_b32 s1, s1, s5
-; GFX8-NEXT:    s_xor_b32 s4, s4, s3
-; GFX8-NEXT:    s_sub_i32 s1, s1, s5
-; GFX8-NEXT:    s_sub_i32 s3, s4, s3
-; GFX8-NEXT:    v_mov_b32_e32 v5, s1
+; GFX8-NEXT:    s_xor_b32 s5, s1, s6
+; GFX8-NEXT:    s_xor_b32 s3, s3, s5
+; GFX8-NEXT:    s_xor_b32 s4, s4, s1
+; GFX8-NEXT:    s_sub_i32 s3, s3, s5
+; GFX8-NEXT:    s_sub_i32 s1, s4, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
-; GFX8-NEXT:    v_mov_b32_e32 v7, s3
-; GFX8-NEXT:    v_mov_b32_e32 v6, s2
+; GFX8-NEXT:    v_mov_b32_e32 v5, s3
+; GFX8-NEXT:    v_mov_b32_e32 v7, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[0:1], v[4:5]
 ; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[6:7]
 ; GFX8-NEXT:    s_endpgm
@@ -866,22 +866,22 @@ define amdgpu_kernel void @sdivrem_v2i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX9-NEXT:    s_sub_i32 s6, s4, s6
 ; GFX9-NEXT:    s_cmp_lg_u32 s11, 0
 ; GFX9-NEXT:    s_cselect_b32 s4, s6, s4
-; GFX9-NEXT:    s_ashr_i32 s11, s7, 31
-; GFX9-NEXT:    s_add_i32 s6, s7, s11
-; GFX9-NEXT:    s_xor_b32 s7, s6, s11
+; GFX9-NEXT:    s_ashr_i32 s6, s7, 31
+; GFX9-NEXT:    s_add_i32 s7, s7, s6
+; GFX9-NEXT:    s_xor_b32 s7, s7, s6
 ; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s7
-; GFX9-NEXT:    s_xor_b32 s6, s9, s8
-; GFX9-NEXT:    s_xor_b32 s10, s10, s6
-; GFX9-NEXT:    s_xor_b32 s12, s4, s9
+; GFX9-NEXT:    s_xor_b32 s8, s9, s8
+; GFX9-NEXT:    s_xor_b32 s10, s10, s8
+; GFX9-NEXT:    s_xor_b32 s4, s4, s9
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
-; GFX9-NEXT:    s_sub_i32 s4, s10, s6
-; GFX9-NEXT:    s_sub_i32 s6, s12, s9
+; GFX9-NEXT:    s_sub_i32 s8, s10, s8
+; GFX9-NEXT:    s_sub_i32 s4, s4, s9
 ; GFX9-NEXT:    s_sub_i32 s9, 0, s7
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX9-NEXT:    s_ashr_i32 s8, s5, 31
-; GFX9-NEXT:    s_add_i32 s5, s5, s8
-; GFX9-NEXT:    s_xor_b32 s5, s5, s8
+; GFX9-NEXT:    s_ashr_i32 s11, s5, 31
+; GFX9-NEXT:    s_add_i32 s5, s5, s11
+; GFX9-NEXT:    s_xor_b32 s5, s5, s11
 ; GFX9-NEXT:    v_readfirstlane_b32 s10, v0
 ; GFX9-NEXT:    s_mul_i32 s9, s9, s10
 ; GFX9-NEXT:    s_mul_hi_u32 s9, s10, s9
@@ -904,17 +904,17 @@ define amdgpu_kernel void @sdivrem_v2i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX9-NEXT:    s_cselect_b32 s9, s12, s9
 ; GFX9-NEXT:    s_sub_i32 s7, s5, s7
 ; GFX9-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX9-NEXT:    s_cselect_b32 s7, s7, s5
-; GFX9-NEXT:    s_xor_b32 s5, s8, s11
-; GFX9-NEXT:    s_xor_b32 s9, s9, s5
-; GFX9-NEXT:    s_sub_i32 s5, s9, s5
-; GFX9-NEXT:    s_xor_b32 s7, s7, s8
-; GFX9-NEXT:    s_sub_i32 s7, s7, s8
+; GFX9-NEXT:    s_cselect_b32 s5, s7, s5
+; GFX9-NEXT:    s_xor_b32 s6, s11, s6
+; GFX9-NEXT:    s_xor_b32 s7, s9, s6
+; GFX9-NEXT:    s_sub_i32 s6, s7, s6
+; GFX9-NEXT:    s_xor_b32 s5, s5, s11
+; GFX9-NEXT:    s_sub_i32 s5, s5, s11
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s6
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s5
-; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
-; GFX9-NEXT:    v_mov_b32_e32 v0, s6
-; GFX9-NEXT:    v_mov_b32_e32 v1, s7
 ; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
 ; GFX9-NEXT:    s_endpgm
 ;
@@ -957,54 +957,54 @@ define amdgpu_kernel void @sdivrem_v2i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX10-NEXT:    s_sub_i32 s6, s4, s6
 ; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
 ; GFX10-NEXT:    s_cselect_b32 s4, s6, s4
-; GFX10-NEXT:    s_ashr_i32 s10, s7, 31
+; GFX10-NEXT:    s_ashr_i32 s6, s7, 31
 ; GFX10-NEXT:    s_ashr_i32 s13, s5, 31
-; GFX10-NEXT:    s_add_i32 s6, s7, s10
+; GFX10-NEXT:    s_add_i32 s7, s7, s6
 ; GFX10-NEXT:    s_add_i32 s5, s5, s13
-; GFX10-NEXT:    s_xor_b32 s7, s6, s10
+; GFX10-NEXT:    s_xor_b32 s7, s7, s6
 ; GFX10-NEXT:    s_xor_b32 s5, s5, s13
 ; GFX10-NEXT:    v_cvt_f32_u32_e32 v0, s7
 ; GFX10-NEXT:    s_sub_i32 s12, 0, s7
 ; GFX10-NEXT:    s_xor_b32 s8, s11, s8
+; GFX10-NEXT:    s_xor_b32 s4, s4, s11
 ; GFX10-NEXT:    s_xor_b32 s9, s9, s8
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX10-NEXT:    s_sub_i32 s8, s9, s8
+; GFX10-NEXT:    s_sub_i32 s4, s4, s11
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX10-NEXT:    v_readfirstlane_b32 s6, v0
-; GFX10-NEXT:    s_mul_i32 s12, s12, s6
-; GFX10-NEXT:    s_mul_hi_u32 s12, s6, s12
-; GFX10-NEXT:    s_add_i32 s6, s6, s12
-; GFX10-NEXT:    s_mul_hi_u32 s12, s5, s6
-; GFX10-NEXT:    s_xor_b32 s6, s4, s11
-; GFX10-NEXT:    s_mul_i32 s14, s12, s7
-; GFX10-NEXT:    s_sub_i32 s4, s9, s8
-; GFX10-NEXT:    s_sub_i32 s5, s5, s14
-; GFX10-NEXT:    s_sub_i32 s6, s6, s11
+; GFX10-NEXT:    v_readfirstlane_b32 s10, v0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s8
+; GFX10-NEXT:    s_mul_i32 s12, s12, s10
+; GFX10-NEXT:    s_mul_hi_u32 s12, s10, s12
+; GFX10-NEXT:    s_add_i32 s10, s10, s12
+; GFX10-NEXT:    s_mul_hi_u32 s10, s5, s10
+; GFX10-NEXT:    s_mul_i32 s12, s10, s7
+; GFX10-NEXT:    s_sub_i32 s5, s5, s12
 ; GFX10-NEXT:    s_cmp_ge_u32 s5, s7
-; GFX10-NEXT:    s_cselect_b32 s8, 1, 0
-; GFX10-NEXT:    s_add_i32 s9, s12, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s8, 0
-; GFX10-NEXT:    s_cselect_b32 s9, s9, s12
+; GFX10-NEXT:    s_cselect_b32 s9, 1, 0
+; GFX10-NEXT:    s_add_i32 s11, s10, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
+; GFX10-NEXT:    s_cselect_b32 s10, s11, s10
 ; GFX10-NEXT:    s_sub_i32 s11, s5, s7
-; GFX10-NEXT:    s_cmp_lg_u32 s8, 0
+; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
 ; GFX10-NEXT:    s_cselect_b32 s5, s11, s5
 ; GFX10-NEXT:    s_cmp_ge_u32 s5, s7
-; GFX10-NEXT:    s_cselect_b32 s8, 1, 0
-; GFX10-NEXT:    s_add_i32 s11, s9, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s8, 0
-; GFX10-NEXT:    s_cselect_b32 s9, s11, s9
+; GFX10-NEXT:    s_cselect_b32 s9, 1, 0
+; GFX10-NEXT:    s_add_i32 s11, s10, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
+; GFX10-NEXT:    s_cselect_b32 s10, s11, s10
 ; GFX10-NEXT:    s_sub_i32 s7, s5, s7
-; GFX10-NEXT:    s_cmp_lg_u32 s8, 0
+; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
 ; GFX10-NEXT:    s_cselect_b32 s5, s7, s5
-; GFX10-NEXT:    s_xor_b32 s7, s13, s10
-; GFX10-NEXT:    s_xor_b32 s8, s9, s7
-; GFX10-NEXT:    s_xor_b32 s9, s5, s13
-; GFX10-NEXT:    s_sub_i32 s5, s8, s7
-; GFX10-NEXT:    s_sub_i32 s7, s9, s13
-; GFX10-NEXT:    v_mov_b32_e32 v0, s4
-; GFX10-NEXT:    v_mov_b32_e32 v1, s5
-; GFX10-NEXT:    v_mov_b32_e32 v2, s6
-; GFX10-NEXT:    v_mov_b32_e32 v3, s7
+; GFX10-NEXT:    s_xor_b32 s6, s13, s6
+; GFX10-NEXT:    s_xor_b32 s5, s5, s13
+; GFX10-NEXT:    s_xor_b32 s7, s10, s6
+; GFX10-NEXT:    s_sub_i32 s5, s5, s13
+; GFX10-NEXT:    s_sub_i32 s6, s7, s6
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
+; GFX10-NEXT:    v_mov_b32_e32 v1, s6
 ; GFX10-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1]
 ; GFX10-NEXT:    global_store_dwordx2 v4, v[2:3], s[2:3]
 ; GFX10-NEXT:    s_endpgm
@@ -1060,151 +1060,151 @@ define amdgpu_kernel void @sdivrem_v4i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX8-NEXT:    s_sub_i32 s4, s0, s4
 ; GFX8-NEXT:    s_cmp_lg_u32 s13, 0
 ; GFX8-NEXT:    s_cselect_b32 s0, s4, s0
-; GFX8-NEXT:    s_ashr_i32 s13, s5, 31
-; GFX8-NEXT:    s_add_i32 s4, s5, s13
-; GFX8-NEXT:    s_xor_b32 s5, s4, s13
+; GFX8-NEXT:    s_ashr_i32 s4, s5, 31
+; GFX8-NEXT:    s_add_i32 s5, s5, s4
+; GFX8-NEXT:    s_xor_b32 s5, s5, s4
 ; GFX8-NEXT:    v_cvt_f32_u32_e32 v0, s5
 ; GFX8-NEXT:    s_sub_i32 s14, 0, s5
-; GFX8-NEXT:    s_xor_b32 s4, s11, s10
-; GFX8-NEXT:    s_ashr_i32 s10, s1, 31
+; GFX8-NEXT:    s_ashr_i32 s13, s1, 31
+; GFX8-NEXT:    s_add_i32 s1, s1, s13
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v0, v0
-; GFX8-NEXT:    s_add_i32 s1, s1, s10
-; GFX8-NEXT:    s_xor_b32 s1, s1, s10
-; GFX8-NEXT:    s_xor_b32 s12, s12, s4
+; GFX8-NEXT:    s_xor_b32 s10, s11, s10
+; GFX8-NEXT:    s_xor_b32 s12, s12, s10
+; GFX8-NEXT:    s_xor_b32 s0, s0, s11
 ; GFX8-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; GFX8-NEXT:    s_sub_i32 s0, s0, s11
+; GFX8-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NEXT:    v_readfirstlane_b32 s15, v0
 ; GFX8-NEXT:    s_mul_i32 s14, s14, s15
 ; GFX8-NEXT:    v_mul_hi_u32 v0, v0, s14
-; GFX8-NEXT:    v_readfirstlane_b32 s14, v0
-; GFX8-NEXT:    s_add_i32 s14, s15, s14
-; GFX8-NEXT:    v_mov_b32_e32 v0, s14
-; GFX8-NEXT:    v_mul_hi_u32 v0, s1, v0
-; GFX8-NEXT:    s_xor_b32 s14, s0, s11
-; GFX8-NEXT:    s_sub_i32 s0, s12, s4
-; GFX8-NEXT:    s_sub_i32 s4, s14, s11
-; GFX8-NEXT:    v_readfirstlane_b32 s11, v0
-; GFX8-NEXT:    s_mul_i32 s12, s11, s5
-; GFX8-NEXT:    s_sub_i32 s1, s1, s12
-; GFX8-NEXT:    s_cmp_ge_u32 s1, s5
+; GFX8-NEXT:    s_xor_b32 s14, s1, s13
+; GFX8-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX8-NEXT:    s_add_i32 s1, s15, s1
+; GFX8-NEXT:    v_mov_b32_e32 v0, s1
+; GFX8-NEXT:    v_mul_hi_u32 v0, s14, v0
+; GFX8-NEXT:    s_sub_i32 s1, s12, s10
+; GFX8-NEXT:    v_readfirstlane_b32 s10, v0
+; GFX8-NEXT:    s_mul_i32 s11, s10, s5
+; GFX8-NEXT:    s_sub_i32 s11, s14, s11
+; GFX8-NEXT:    s_cmp_ge_u32 s11, s5
 ; GFX8-NEXT:    s_cselect_b32 s12, 1, 0
-; GFX8-NEXT:    s_add_i32 s14, s11, 1
+; GFX8-NEXT:    s_add_i32 s14, s10, 1
 ; GFX8-NEXT:    s_cmp_lg_u32 s12, 0
-; GFX8-NEXT:    s_cselect_b32 s11, s14, s11
-; GFX8-NEXT:    s_sub_i32 s14, s1, s5
+; GFX8-NEXT:    s_cselect_b32 s10, s14, s10
+; GFX8-NEXT:    s_sub_i32 s14, s11, s5
 ; GFX8-NEXT:    s_cmp_lg_u32 s12, 0
-; GFX8-NEXT:    s_cselect_b32 s1, s14, s1
-; GFX8-NEXT:    s_cmp_ge_u32 s1, s5
+; GFX8-NEXT:    s_cselect_b32 s11, s14, s11
+; GFX8-NEXT:    s_cmp_ge_u32 s11, s5
 ; GFX8-NEXT:    s_cselect_b32 s12, 1, 0
-; GFX8-NEXT:    s_add_i32 s14, s11, 1
+; GFX8-NEXT:    s_add_i32 s14, s10, 1
 ; GFX8-NEXT:    s_cmp_lg_u32 s12, 0
-; GFX8-NEXT:    s_cselect_b32 s11, s14, s11
-; GFX8-NEXT:    s_sub_i32 s5, s1, s5
+; GFX8-NEXT:    s_cselect_b32 s10, s14, s10
+; GFX8-NEXT:    s_sub_i32 s5, s11, s5
 ; GFX8-NEXT:    s_cmp_lg_u32 s12, 0
-; GFX8-NEXT:    s_cselect_b32 s1, s5, s1
-; GFX8-NEXT:    s_ashr_i32 s12, s6, 31
-; GFX8-NEXT:    s_add_i32 s5, s6, s12
-; GFX8-NEXT:    s_xor_b32 s6, s5, s12
+; GFX8-NEXT:    s_cselect_b32 s5, s5, s11
+; GFX8-NEXT:    s_ashr_i32 s11, s6, 31
+; GFX8-NEXT:    s_add_i32 s6, s6, s11
+; GFX8-NEXT:    s_xor_b32 s6, s6, s11
 ; GFX8-NEXT:    v_cvt_f32_u32_e32 v0, s6
 ; GFX8-NEXT:    s_sub_i32 s14, 0, s6
-; GFX8-NEXT:    s_xor_b32 s5, s10, s13
-; GFX8-NEXT:    s_ashr_i32 s13, s2, 31
+; GFX8-NEXT:    s_ashr_i32 s12, s2, 31
+; GFX8-NEXT:    s_add_i32 s2, s2, s12
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v0, v0
-; GFX8-NEXT:    s_add_i32 s2, s2, s13
-; GFX8-NEXT:    s_xor_b32 s2, s2, s13
-; GFX8-NEXT:    s_xor_b32 s11, s11, s5
+; GFX8-NEXT:    s_xor_b32 s2, s2, s12
+; GFX8-NEXT:    s_xor_b32 s4, s13, s4
+; GFX8-NEXT:    s_xor_b32 s10, s10, s4
 ; GFX8-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; GFX8-NEXT:    s_xor_b32 s5, s5, s13
+; GFX8-NEXT:    s_sub_i32 s10, s10, s4
+; GFX8-NEXT:    s_sub_i32 s13, s5, s13
 ; GFX8-NEXT:    v_readfirstlane_b32 s15, v0
 ; GFX8-NEXT:    s_mul_i32 s14, s14, s15
 ; GFX8-NEXT:    v_mul_hi_u32 v0, v0, s14
+; GFX8-NEXT:    v_mov_b32_e32 v1, s10
+; GFX8-NEXT:    v_mov_b32_e32 v5, s13
 ; GFX8-NEXT:    v_readfirstlane_b32 s14, v0
 ; GFX8-NEXT:    s_add_i32 s14, s15, s14
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s14
 ; GFX8-NEXT:    v_mul_hi_u32 v0, s2, v0
-; GFX8-NEXT:    s_xor_b32 s14, s1, s10
-; GFX8-NEXT:    s_sub_i32 s1, s11, s5
-; GFX8-NEXT:    s_sub_i32 s5, s14, s10
-; GFX8-NEXT:    v_readfirstlane_b32 s10, v0
-; GFX8-NEXT:    s_mul_i32 s11, s10, s6
-; GFX8-NEXT:    s_sub_i32 s2, s2, s11
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX8-NEXT:    s_mul_i32 s5, s4, s6
+; GFX8-NEXT:    s_sub_i32 s2, s2, s5
 ; GFX8-NEXT:    s_cmp_ge_u32 s2, s6
-; GFX8-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX8-NEXT:    s_add_i32 s14, s10, 1
-; GFX8-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX8-NEXT:    s_cselect_b32 s10, s14, s10
+; GFX8-NEXT:    s_cselect_b32 s5, 1, 0
+; GFX8-NEXT:    s_add_i32 s14, s4, 1
+; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
+; GFX8-NEXT:    s_cselect_b32 s4, s14, s4
 ; GFX8-NEXT:    s_sub_i32 s14, s2, s6
-; GFX8-NEXT:    s_cmp_lg_u32 s11, 0
+; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
 ; GFX8-NEXT:    s_cselect_b32 s2, s14, s2
 ; GFX8-NEXT:    s_cmp_ge_u32 s2, s6
-; GFX8-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX8-NEXT:    s_add_i32 s14, s10, 1
-; GFX8-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX8-NEXT:    s_cselect_b32 s14, s14, s10
-; GFX8-NEXT:    s_sub_i32 s6, s2, s6
-; GFX8-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX8-NEXT:    s_cselect_b32 s2, s6, s2
+; GFX8-NEXT:    s_cselect_b32 s5, 1, 0
+; GFX8-NEXT:    s_add_i32 s14, s4, 1
+; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
+; GFX8-NEXT:    s_cselect_b32 s14, s14, s4
+; GFX8-NEXT:    s_sub_i32 s4, s2, s6
+; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
+; GFX8-NEXT:    s_cselect_b32 s2, s4, s2
 ; GFX8-NEXT:    s_ashr_i32 s15, s7, 31
-; GFX8-NEXT:    s_add_i32 s6, s7, s15
-; GFX8-NEXT:    s_xor_b32 s7, s6, s15
-; GFX8-NEXT:    v_cvt_f32_u32_e32 v0, s7
-; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x0
-; GFX8-NEXT:    s_xor_b32 s6, s13, s12
+; GFX8-NEXT:    s_add_i32 s4, s7, s15
+; GFX8-NEXT:    s_xor_b32 s16, s4, s15
+; GFX8-NEXT:    v_cvt_f32_u32_e32 v0, s16
+; GFX8-NEXT:    s_load_dwordx4 s[4:7], s[8:9], 0x0
+; GFX8-NEXT:    s_xor_b32 s2, s2, s12
+; GFX8-NEXT:    s_sub_i32 s2, s2, s12
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX8-NEXT:    v_mov_b32_e32 v6, s2
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v8, s8
-; GFX8-NEXT:    v_mov_b32_e32 v9, s9
+; GFX8-NEXT:    v_mov_b32_e32 v11, s7
+; GFX8-NEXT:    v_mov_b32_e32 v10, s6
 ; GFX8-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v10, s10
-; GFX8-NEXT:    v_mov_b32_e32 v11, s11
-; GFX8-NEXT:    s_sub_i32 s9, 0, s7
-; GFX8-NEXT:    v_readfirstlane_b32 s10, v0
-; GFX8-NEXT:    s_mul_i32 s9, s9, s10
-; GFX8-NEXT:    v_mul_hi_u32 v0, v0, s9
-; GFX8-NEXT:    s_ashr_i32 s8, s3, 31
-; GFX8-NEXT:    s_add_i32 s3, s3, s8
-; GFX8-NEXT:    s_xor_b32 s3, s3, s8
-; GFX8-NEXT:    v_readfirstlane_b32 s11, v0
-; GFX8-NEXT:    s_add_i32 s10, s10, s11
-; GFX8-NEXT:    v_mov_b32_e32 v0, s10
+; GFX8-NEXT:    s_sub_i32 s6, 0, s16
+; GFX8-NEXT:    v_mov_b32_e32 v9, s5
+; GFX8-NEXT:    v_mov_b32_e32 v8, s4
+; GFX8-NEXT:    v_readfirstlane_b32 s7, v0
+; GFX8-NEXT:    s_mul_i32 s6, s6, s7
+; GFX8-NEXT:    v_mul_hi_u32 v0, v0, s6
+; GFX8-NEXT:    s_ashr_i32 s5, s3, 31
+; GFX8-NEXT:    s_add_i32 s3, s3, s5
+; GFX8-NEXT:    s_xor_b32 s3, s3, s5
+; GFX8-NEXT:    v_readfirstlane_b32 s8, v0
+; GFX8-NEXT:    s_add_i32 s7, s7, s8
+; GFX8-NEXT:    v_mov_b32_e32 v0, s7
 ; GFX8-NEXT:    v_mul_hi_u32 v0, s3, v0
-; GFX8-NEXT:    s_xor_b32 s9, s14, s6
-; GFX8-NEXT:    s_xor_b32 s10, s2, s13
-; GFX8-NEXT:    s_sub_i32 s2, s9, s6
-; GFX8-NEXT:    v_readfirstlane_b32 s9, v0
-; GFX8-NEXT:    s_sub_i32 s6, s10, s13
-; GFX8-NEXT:    s_mul_i32 s10, s9, s7
-; GFX8-NEXT:    s_sub_i32 s3, s3, s10
-; GFX8-NEXT:    s_cmp_ge_u32 s3, s7
-; GFX8-NEXT:    s_cselect_b32 s10, 1, 0
-; GFX8-NEXT:    s_add_i32 s11, s9, 1
-; GFX8-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX8-NEXT:    s_cselect_b32 s9, s11, s9
-; GFX8-NEXT:    s_sub_i32 s11, s3, s7
-; GFX8-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX8-NEXT:    s_cselect_b32 s3, s11, s3
-; GFX8-NEXT:    s_cmp_ge_u32 s3, s7
-; GFX8-NEXT:    s_cselect_b32 s10, 1, 0
-; GFX8-NEXT:    s_add_i32 s11, s9, 1
-; GFX8-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX8-NEXT:    s_cselect_b32 s9, s11, s9
-; GFX8-NEXT:    s_sub_i32 s7, s3, s7
-; GFX8-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX8-NEXT:    s_cselect_b32 s3, s7, s3
-; GFX8-NEXT:    s_xor_b32 s7, s8, s15
-; GFX8-NEXT:    s_xor_b32 s9, s9, s7
-; GFX8-NEXT:    s_xor_b32 s10, s3, s8
-; GFX8-NEXT:    s_sub_i32 s3, s9, s7
-; GFX8-NEXT:    s_sub_i32 s7, s10, s8
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    v_mov_b32_e32 v6, s6
-; GFX8-NEXT:    v_mov_b32_e32 v7, s7
+; GFX8-NEXT:    s_xor_b32 s4, s12, s11
+; GFX8-NEXT:    s_xor_b32 s6, s14, s4
+; GFX8-NEXT:    s_sub_i32 s4, s6, s4
+; GFX8-NEXT:    v_readfirstlane_b32 s6, v0
+; GFX8-NEXT:    s_mul_i32 s7, s6, s16
+; GFX8-NEXT:    s_sub_i32 s3, s3, s7
+; GFX8-NEXT:    s_cmp_ge_u32 s3, s16
+; GFX8-NEXT:    s_cselect_b32 s7, 1, 0
+; GFX8-NEXT:    s_add_i32 s8, s6, 1
+; GFX8-NEXT:    s_cmp_lg_u32 s7, 0
+; GFX8-NEXT:    s_cselect_b32 s6, s8, s6
+; GFX8-NEXT:    s_sub_i32 s8, s3, s16
+; GFX8-NEXT:    s_cmp_lg_u32 s7, 0
+; GFX8-NEXT:    s_cselect_b32 s3, s8, s3
+; GFX8-NEXT:    s_cmp_ge_u32 s3, s16
+; GFX8-NEXT:    s_cselect_b32 s7, 1, 0
+; GFX8-NEXT:    s_add_i32 s8, s6, 1
+; GFX8-NEXT:    s_cmp_lg_u32 s7, 0
+; GFX8-NEXT:    s_cselect_b32 s6, s8, s6
+; GFX8-NEXT:    s_sub_i32 s8, s3, s16
+; GFX8-NEXT:    s_cmp_lg_u32 s7, 0
+; GFX8-NEXT:    s_cselect_b32 s3, s8, s3
+; GFX8-NEXT:    s_xor_b32 s7, s5, s15
+; GFX8-NEXT:    s_xor_b32 s6, s6, s7
+; GFX8-NEXT:    s_xor_b32 s3, s3, s5
+; GFX8-NEXT:    s_sub_i32 s6, s6, s7
+; GFX8-NEXT:    s_sub_i32 s3, s3, s5
+; GFX8-NEXT:    v_mov_b32_e32 v0, s1
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s6
+; GFX8-NEXT:    v_mov_b32_e32 v7, s3
 ; GFX8-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
 ; GFX8-NEXT:    flat_store_dwordx4 v[10:11], v[4:7]
 ; GFX8-NEXT:    s_endpgm
@@ -1248,22 +1248,22 @@ define amdgpu_kernel void @sdivrem_v4i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX9-NEXT:    s_sub_i32 s4, s0, s4
 ; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
 ; GFX9-NEXT:    s_cselect_b32 s0, s4, s0
-; GFX9-NEXT:    s_ashr_i32 s13, s5, 31
-; GFX9-NEXT:    s_add_i32 s4, s5, s13
-; GFX9-NEXT:    s_xor_b32 s5, s4, s13
+; GFX9-NEXT:    s_ashr_i32 s4, s5, 31
+; GFX9-NEXT:    s_add_i32 s5, s5, s4
+; GFX9-NEXT:    s_xor_b32 s5, s5, s4
 ; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s5
-; GFX9-NEXT:    s_xor_b32 s4, s11, s10
-; GFX9-NEXT:    s_xor_b32 s12, s12, s4
-; GFX9-NEXT:    s_xor_b32 s14, s0, s11
+; GFX9-NEXT:    s_xor_b32 s10, s11, s10
+; GFX9-NEXT:    s_xor_b32 s12, s12, s10
+; GFX9-NEXT:    s_xor_b32 s0, s0, s11
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
-; GFX9-NEXT:    s_sub_i32 s0, s12, s4
-; GFX9-NEXT:    s_sub_i32 s4, s14, s11
+; GFX9-NEXT:    s_sub_i32 s10, s12, s10
+; GFX9-NEXT:    s_sub_i32 s0, s0, s11
 ; GFX9-NEXT:    s_sub_i32 s11, 0, s5
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX9-NEXT:    s_ashr_i32 s10, s1, 31
-; GFX9-NEXT:    s_add_i32 s1, s1, s10
-; GFX9-NEXT:    s_xor_b32 s1, s1, s10
+; GFX9-NEXT:    s_ashr_i32 s13, s1, 31
+; GFX9-NEXT:    s_add_i32 s1, s1, s13
+; GFX9-NEXT:    s_xor_b32 s1, s1, s13
 ; GFX9-NEXT:    v_readfirstlane_b32 s12, v0
 ; GFX9-NEXT:    s_mul_i32 s11, s11, s12
 ; GFX9-NEXT:    s_mul_hi_u32 s11, s12, s11
@@ -1289,100 +1289,100 @@ define amdgpu_kernel void @sdivrem_v4i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX9-NEXT:    s_cselect_b32 s1, s5, s1
 ; GFX9-NEXT:    s_ashr_i32 s12, s6, 31
 ; GFX9-NEXT:    s_add_i32 s5, s6, s12
-; GFX9-NEXT:    s_xor_b32 s6, s5, s12
-; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s6
-; GFX9-NEXT:    s_xor_b32 s5, s10, s13
-; GFX9-NEXT:    s_xor_b32 s11, s11, s5
-; GFX9-NEXT:    s_xor_b32 s14, s1, s10
+; GFX9-NEXT:    s_xor_b32 s5, s5, s12
+; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s5
+; GFX9-NEXT:    s_xor_b32 s4, s13, s4
+; GFX9-NEXT:    s_xor_b32 s6, s11, s4
+; GFX9-NEXT:    s_sub_i32 s11, s6, s4
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
-; GFX9-NEXT:    s_sub_i32 s1, s11, s5
-; GFX9-NEXT:    s_sub_i32 s5, s14, s10
-; GFX9-NEXT:    s_sub_i32 s11, 0, s6
+; GFX9-NEXT:    s_sub_i32 s6, 0, s5
+; GFX9-NEXT:    s_ashr_i32 s14, s2, 31
+; GFX9-NEXT:    s_add_i32 s2, s2, s14
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX9-NEXT:    s_ashr_i32 s13, s2, 31
-; GFX9-NEXT:    s_add_i32 s2, s2, s13
-; GFX9-NEXT:    s_xor_b32 s2, s2, s13
-; GFX9-NEXT:    v_readfirstlane_b32 s10, v0
-; GFX9-NEXT:    s_mul_i32 s11, s11, s10
-; GFX9-NEXT:    s_mul_hi_u32 s11, s10, s11
-; GFX9-NEXT:    s_add_i32 s10, s10, s11
-; GFX9-NEXT:    s_mul_hi_u32 s10, s2, s10
-; GFX9-NEXT:    s_mul_i32 s11, s10, s6
-; GFX9-NEXT:    s_sub_i32 s2, s2, s11
-; GFX9-NEXT:    s_cmp_ge_u32 s2, s6
-; GFX9-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX9-NEXT:    s_add_i32 s14, s10, 1
-; GFX9-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX9-NEXT:    s_cselect_b32 s10, s14, s10
-; GFX9-NEXT:    s_sub_i32 s14, s2, s6
-; GFX9-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX9-NEXT:    s_cselect_b32 s2, s14, s2
-; GFX9-NEXT:    s_cmp_ge_u32 s2, s6
-; GFX9-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX9-NEXT:    s_add_i32 s14, s10, 1
-; GFX9-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX9-NEXT:    s_cselect_b32 s14, s14, s10
-; GFX9-NEXT:    s_sub_i32 s6, s2, s6
-; GFX9-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX9-NEXT:    s_cselect_b32 s6, s6, s2
+; GFX9-NEXT:    s_xor_b32 s2, s2, s14
+; GFX9-NEXT:    s_xor_b32 s1, s1, s13
+; GFX9-NEXT:    s_sub_i32 s1, s1, s13
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX9-NEXT:    s_mul_i32 s6, s6, s4
+; GFX9-NEXT:    s_mul_hi_u32 s6, s4, s6
+; GFX9-NEXT:    s_add_i32 s4, s4, s6
+; GFX9-NEXT:    s_mul_hi_u32 s4, s2, s4
+; GFX9-NEXT:    s_mul_i32 s6, s4, s5
+; GFX9-NEXT:    s_sub_i32 s2, s2, s6
+; GFX9-NEXT:    s_cmp_ge_u32 s2, s5
+; GFX9-NEXT:    s_cselect_b32 s6, 1, 0
+; GFX9-NEXT:    s_add_i32 s13, s4, 1
+; GFX9-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX9-NEXT:    s_cselect_b32 s4, s13, s4
+; GFX9-NEXT:    s_sub_i32 s13, s2, s5
+; GFX9-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX9-NEXT:    s_cselect_b32 s2, s13, s2
+; GFX9-NEXT:    s_cmp_ge_u32 s2, s5
+; GFX9-NEXT:    s_cselect_b32 s6, 1, 0
+; GFX9-NEXT:    s_add_i32 s13, s4, 1
+; GFX9-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX9-NEXT:    s_cselect_b32 s13, s13, s4
+; GFX9-NEXT:    s_sub_i32 s4, s2, s5
+; GFX9-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX9-NEXT:    s_cselect_b32 s2, s4, s2
 ; GFX9-NEXT:    s_ashr_i32 s15, s7, 31
-; GFX9-NEXT:    s_add_i32 s2, s7, s15
-; GFX9-NEXT:    s_xor_b32 s7, s2, s15
-; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s7
-; GFX9-NEXT:    s_xor_b32 s2, s13, s12
-; GFX9-NEXT:    s_xor_b32 s6, s6, s13
-; GFX9-NEXT:    s_xor_b32 s12, s14, s2
+; GFX9-NEXT:    s_add_i32 s4, s7, s15
+; GFX9-NEXT:    s_xor_b32 s16, s4, s15
+; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s16
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[8:9], 0x0
+; GFX9-NEXT:    s_xor_b32 s8, s14, s12
+; GFX9-NEXT:    s_xor_b32 s9, s13, s8
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
-; GFX9-NEXT:    s_sub_i32 s6, s6, s13
-; GFX9-NEXT:    s_sub_i32 s14, 0, s7
-; GFX9-NEXT:    s_sub_i32 s2, s12, s2
+; GFX9-NEXT:    s_sub_i32 s13, 0, s16
+; GFX9-NEXT:    s_sub_i32 s8, s9, s8
+; GFX9-NEXT:    s_ashr_i32 s9, s3, 31
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX9-NEXT:    s_ashr_i32 s12, s3, 31
-; GFX9-NEXT:    s_add_i32 s3, s3, s12
-; GFX9-NEXT:    s_xor_b32 s3, s3, s12
-; GFX9-NEXT:    v_readfirstlane_b32 s13, v0
-; GFX9-NEXT:    s_mul_i32 s14, s14, s13
-; GFX9-NEXT:    s_mul_hi_u32 s14, s13, s14
-; GFX9-NEXT:    s_add_i32 s13, s13, s14
-; GFX9-NEXT:    s_mul_hi_u32 s13, s3, s13
-; GFX9-NEXT:    s_mul_i32 s14, s13, s7
-; GFX9-NEXT:    s_sub_i32 s3, s3, s14
-; GFX9-NEXT:    s_cmp_ge_u32 s3, s7
-; GFX9-NEXT:    s_cselect_b32 s14, 1, 0
-; GFX9-NEXT:    s_add_i32 s16, s13, 1
-; GFX9-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX9-NEXT:    s_cselect_b32 s13, s16, s13
-; GFX9-NEXT:    s_sub_i32 s16, s3, s7
-; GFX9-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX9-NEXT:    s_cselect_b32 s3, s16, s3
-; GFX9-NEXT:    s_cmp_ge_u32 s3, s7
-; GFX9-NEXT:    s_cselect_b32 s14, 1, 0
-; GFX9-NEXT:    s_add_i32 s16, s13, 1
-; GFX9-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX9-NEXT:    s_cselect_b32 s13, s16, s13
-; GFX9-NEXT:    s_sub_i32 s7, s3, s7
-; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x0
-; GFX9-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX9-NEXT:    s_cselect_b32 s7, s7, s3
-; GFX9-NEXT:    s_xor_b32 s3, s12, s15
-; GFX9-NEXT:    s_xor_b32 s13, s13, s3
-; GFX9-NEXT:    s_sub_i32 s3, s13, s3
-; GFX9-NEXT:    s_xor_b32 s7, s7, s12
-; GFX9-NEXT:    s_sub_i32 s7, s7, s12
+; GFX9-NEXT:    s_add_i32 s3, s3, s9
+; GFX9-NEXT:    s_xor_b32 s3, s3, s9
+; GFX9-NEXT:    s_xor_b32 s2, s2, s14
+; GFX9-NEXT:    v_readfirstlane_b32 s12, v0
+; GFX9-NEXT:    s_mul_i32 s13, s13, s12
+; GFX9-NEXT:    s_mul_hi_u32 s13, s12, s13
+; GFX9-NEXT:    s_add_i32 s12, s12, s13
+; GFX9-NEXT:    s_mul_hi_u32 s12, s3, s12
+; GFX9-NEXT:    s_mul_i32 s13, s12, s16
+; GFX9-NEXT:    s_sub_i32 s2, s2, s14
+; GFX9-NEXT:    s_sub_i32 s3, s3, s13
+; GFX9-NEXT:    s_cmp_ge_u32 s3, s16
+; GFX9-NEXT:    s_cselect_b32 s13, 1, 0
+; GFX9-NEXT:    s_add_i32 s14, s12, 1
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s12, s14, s12
+; GFX9-NEXT:    s_sub_i32 s14, s3, s16
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s3, s14, s3
+; GFX9-NEXT:    s_cmp_ge_u32 s3, s16
+; GFX9-NEXT:    s_cselect_b32 s13, 1, 0
+; GFX9-NEXT:    s_add_i32 s14, s12, 1
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s12, s14, s12
+; GFX9-NEXT:    s_sub_i32 s14, s3, s16
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s3, s14, s3
+; GFX9-NEXT:    s_xor_b32 s13, s9, s15
+; GFX9-NEXT:    s_xor_b32 s12, s12, s13
+; GFX9-NEXT:    s_sub_i32 s12, s12, s13
+; GFX9-NEXT:    s_xor_b32 s3, s3, s9
+; GFX9-NEXT:    s_sub_i32 s3, s3, s9
+; GFX9-NEXT:    v_mov_b32_e32 v0, s10
+; GFX9-NEXT:    v_mov_b32_e32 v1, s11
+; GFX9-NEXT:    v_mov_b32_e32 v2, s8
+; GFX9-NEXT:    v_mov_b32_e32 v3, s12
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5]
+; GFX9-NEXT:    s_nop 0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX9-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s3
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[8:9]
-; GFX9-NEXT:    s_nop 0
-; GFX9-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9-NEXT:    v_mov_b32_e32 v1, s5
-; GFX9-NEXT:    v_mov_b32_e32 v2, s6
-; GFX9-NEXT:    v_mov_b32_e32 v3, s7
-; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[10:11]
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: sdivrem_v4i32:
@@ -1424,140 +1424,140 @@ define amdgpu_kernel void @sdivrem_v4i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX10-NEXT:    s_sub_i32 s4, s0, s4
 ; GFX10-NEXT:    s_cmp_lg_u32 s12, 0
 ; GFX10-NEXT:    s_cselect_b32 s0, s4, s0
-; GFX10-NEXT:    s_ashr_i32 s12, s5, 31
+; GFX10-NEXT:    s_ashr_i32 s4, s5, 31
 ; GFX10-NEXT:    s_ashr_i32 s15, s1, 31
-; GFX10-NEXT:    s_add_i32 s4, s5, s12
+; GFX10-NEXT:    s_add_i32 s5, s5, s4
 ; GFX10-NEXT:    s_add_i32 s1, s1, s15
-; GFX10-NEXT:    s_xor_b32 s5, s4, s12
+; GFX10-NEXT:    s_xor_b32 s5, s5, s4
 ; GFX10-NEXT:    s_xor_b32 s1, s1, s15
 ; GFX10-NEXT:    v_cvt_f32_u32_e32 v0, s5
 ; GFX10-NEXT:    s_sub_i32 s14, 0, s5
 ; GFX10-NEXT:    s_xor_b32 s10, s13, s10
+; GFX10-NEXT:    s_xor_b32 s0, s0, s13
 ; GFX10-NEXT:    s_xor_b32 s11, s11, s10
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX10-NEXT:    s_sub_i32 s10, s11, s10
+; GFX10-NEXT:    s_sub_i32 s11, s0, s13
+; GFX10-NEXT:    v_mov_b32_e32 v4, s11
 ; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX10-NEXT:    v_readfirstlane_b32 s4, v0
-; GFX10-NEXT:    s_mul_i32 s14, s14, s4
-; GFX10-NEXT:    s_mul_hi_u32 s14, s4, s14
-; GFX10-NEXT:    s_add_i32 s4, s4, s14
-; GFX10-NEXT:    s_mul_hi_u32 s14, s1, s4
-; GFX10-NEXT:    s_xor_b32 s4, s0, s13
-; GFX10-NEXT:    s_mul_i32 s16, s14, s5
-; GFX10-NEXT:    s_sub_i32 s0, s11, s10
-; GFX10-NEXT:    s_sub_i32 s1, s1, s16
-; GFX10-NEXT:    s_sub_i32 s4, s4, s13
+; GFX10-NEXT:    v_readfirstlane_b32 s12, v0
+; GFX10-NEXT:    s_mul_i32 s14, s14, s12
+; GFX10-NEXT:    s_mul_hi_u32 s14, s12, s14
+; GFX10-NEXT:    s_add_i32 s12, s12, s14
+; GFX10-NEXT:    s_mul_hi_u32 s12, s1, s12
+; GFX10-NEXT:    s_mul_i32 s14, s12, s5
+; GFX10-NEXT:    s_sub_i32 s1, s1, s14
 ; GFX10-NEXT:    s_cmp_ge_u32 s1, s5
-; GFX10-NEXT:    s_cselect_b32 s10, 1, 0
-; GFX10-NEXT:    s_add_i32 s11, s14, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX10-NEXT:    s_cselect_b32 s11, s11, s14
+; GFX10-NEXT:    s_cselect_b32 s0, 1, 0
+; GFX10-NEXT:    s_add_i32 s13, s12, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX10-NEXT:    s_cselect_b32 s12, s13, s12
 ; GFX10-NEXT:    s_sub_i32 s13, s1, s5
-; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX10-NEXT:    s_cselect_b32 s1, s13, s1
-; GFX10-NEXT:    s_cmp_ge_u32 s1, s5
-; GFX10-NEXT:    s_cselect_b32 s10, 1, 0
-; GFX10-NEXT:    s_add_i32 s13, s11, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX10-NEXT:    s_cselect_b32 s11, s13, s11
-; GFX10-NEXT:    s_sub_i32 s5, s1, s5
-; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX10-NEXT:    s_cselect_b32 s1, s5, s1
-; GFX10-NEXT:    s_ashr_i32 s10, s6, 31
+; GFX10-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX10-NEXT:    s_cselect_b32 s0, s13, s1
+; GFX10-NEXT:    s_cmp_ge_u32 s0, s5
+; GFX10-NEXT:    s_cselect_b32 s1, 1, 0
+; GFX10-NEXT:    s_add_i32 s13, s12, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s1, 0
+; GFX10-NEXT:    s_cselect_b32 s12, s13, s12
+; GFX10-NEXT:    s_sub_i32 s5, s0, s5
+; GFX10-NEXT:    s_cmp_lg_u32 s1, 0
+; GFX10-NEXT:    s_cselect_b32 s0, s5, s0
+; GFX10-NEXT:    s_ashr_i32 s1, s6, 31
 ; GFX10-NEXT:    s_ashr_i32 s14, s2, 31
-; GFX10-NEXT:    s_add_i32 s5, s6, s10
+; GFX10-NEXT:    s_add_i32 s5, s6, s1
 ; GFX10-NEXT:    s_add_i32 s2, s2, s14
-; GFX10-NEXT:    s_xor_b32 s6, s5, s10
+; GFX10-NEXT:    s_xor_b32 s5, s5, s1
 ; GFX10-NEXT:    s_xor_b32 s2, s2, s14
-; GFX10-NEXT:    v_cvt_f32_u32_e32 v0, s6
-; GFX10-NEXT:    s_sub_i32 s13, 0, s6
-; GFX10-NEXT:    s_xor_b32 s12, s15, s12
-; GFX10-NEXT:    s_xor_b32 s11, s11, s12
+; GFX10-NEXT:    v_cvt_f32_u32_e32 v0, s5
+; GFX10-NEXT:    s_sub_i32 s13, 0, s5
+; GFX10-NEXT:    s_xor_b32 s4, s15, s4
+; GFX10-NEXT:    s_xor_b32 s0, s0, s15
+; GFX10-NEXT:    s_xor_b32 s12, s12, s4
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX10-NEXT:    s_sub_i32 s4, s12, s4
+; GFX10-NEXT:    s_sub_i32 s12, s0, s15
+; GFX10-NEXT:    v_mov_b32_e32 v1, s4
+; GFX10-NEXT:    v_mov_b32_e32 v5, s12
 ; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX10-NEXT:    v_readfirstlane_b32 s5, v0
-; GFX10-NEXT:    s_mul_i32 s13, s13, s5
-; GFX10-NEXT:    s_mul_hi_u32 s13, s5, s13
-; GFX10-NEXT:    s_add_i32 s5, s5, s13
-; GFX10-NEXT:    s_mul_hi_u32 s13, s2, s5
-; GFX10-NEXT:    s_xor_b32 s5, s1, s15
-; GFX10-NEXT:    s_mul_i32 s16, s13, s6
-; GFX10-NEXT:    s_sub_i32 s1, s11, s12
-; GFX10-NEXT:    s_sub_i32 s2, s2, s16
-; GFX10-NEXT:    s_sub_i32 s5, s5, s15
-; GFX10-NEXT:    s_cmp_ge_u32 s2, s6
-; GFX10-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX10-NEXT:    s_add_i32 s12, s13, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX10-NEXT:    s_cselect_b32 s12, s12, s13
-; GFX10-NEXT:    s_sub_i32 s13, s2, s6
-; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX10-NEXT:    s_cselect_b32 s2, s13, s2
-; GFX10-NEXT:    s_cmp_ge_u32 s2, s6
-; GFX10-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX10-NEXT:    s_add_i32 s13, s12, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX10-NEXT:    s_cselect_b32 s12, s13, s12
-; GFX10-NEXT:    s_sub_i32 s6, s2, s6
-; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX10-NEXT:    s_cselect_b32 s2, s6, s2
-; GFX10-NEXT:    s_ashr_i32 s13, s7, 31
+; GFX10-NEXT:    v_readfirstlane_b32 s6, v0
+; GFX10-NEXT:    s_mul_i32 s13, s13, s6
+; GFX10-NEXT:    s_mul_hi_u32 s13, s6, s13
+; GFX10-NEXT:    s_add_i32 s6, s6, s13
+; GFX10-NEXT:    s_mul_hi_u32 s6, s2, s6
+; GFX10-NEXT:    s_mul_i32 s13, s6, s5
+; GFX10-NEXT:    s_sub_i32 s2, s2, s13
+; GFX10-NEXT:    s_cmp_ge_u32 s2, s5
+; GFX10-NEXT:    s_cselect_b32 s0, 1, 0
+; GFX10-NEXT:    s_add_i32 s13, s6, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX10-NEXT:    s_cselect_b32 s6, s13, s6
+; GFX10-NEXT:    s_sub_i32 s13, s2, s5
+; GFX10-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX10-NEXT:    s_cselect_b32 s0, s13, s2
+; GFX10-NEXT:    s_cmp_ge_u32 s0, s5
+; GFX10-NEXT:    s_cselect_b32 s2, 1, 0
+; GFX10-NEXT:    s_add_i32 s13, s6, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s2, 0
+; GFX10-NEXT:    s_cselect_b32 s6, s13, s6
+; GFX10-NEXT:    s_sub_i32 s5, s0, s5
+; GFX10-NEXT:    s_cmp_lg_u32 s2, 0
+; GFX10-NEXT:    s_cselect_b32 s0, s5, s0
+; GFX10-NEXT:    s_ashr_i32 s5, s7, 31
 ; GFX10-NEXT:    s_ashr_i32 s15, s3, 31
-; GFX10-NEXT:    s_add_i32 s6, s7, s13
+; GFX10-NEXT:    s_add_i32 s2, s7, s5
 ; GFX10-NEXT:    s_add_i32 s3, s3, s15
-; GFX10-NEXT:    s_xor_b32 s7, s6, s13
-; GFX10-NEXT:    s_xor_b32 s10, s14, s10
-; GFX10-NEXT:    v_cvt_f32_u32_e32 v0, s7
-; GFX10-NEXT:    s_sub_i32 s11, 0, s7
+; GFX10-NEXT:    s_xor_b32 s7, s2, s5
 ; GFX10-NEXT:    s_xor_b32 s3, s3, s15
+; GFX10-NEXT:    v_cvt_f32_u32_e32 v0, s7
+; GFX10-NEXT:    s_sub_i32 s13, 0, s7
+; GFX10-NEXT:    s_xor_b32 s1, s14, s1
+; GFX10-NEXT:    s_xor_b32 s0, s0, s14
+; GFX10-NEXT:    s_xor_b32 s6, s6, s1
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX10-NEXT:    s_sub_i32 s6, s6, s1
+; GFX10-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX10-NEXT:    v_readfirstlane_b32 s6, v0
-; GFX10-NEXT:    s_mul_i32 s11, s11, s6
-; GFX10-NEXT:    s_mul_hi_u32 s11, s6, s11
-; GFX10-NEXT:    s_add_i32 s6, s6, s11
-; GFX10-NEXT:    s_xor_b32 s11, s12, s10
-; GFX10-NEXT:    s_mul_hi_u32 s12, s3, s6
-; GFX10-NEXT:    s_xor_b32 s6, s2, s14
-; GFX10-NEXT:    s_mul_i32 s16, s12, s7
-; GFX10-NEXT:    s_sub_i32 s2, s11, s10
-; GFX10-NEXT:    s_sub_i32 s3, s3, s16
-; GFX10-NEXT:    s_sub_i32 s6, s6, s14
-; GFX10-NEXT:    s_cmp_ge_u32 s3, s7
-; GFX10-NEXT:    s_cselect_b32 s10, 1, 0
-; GFX10-NEXT:    s_add_i32 s11, s12, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX10-NEXT:    s_cselect_b32 s12, s11, s12
-; GFX10-NEXT:    s_sub_i32 s11, s3, s7
-; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX10-NEXT:    s_cselect_b32 s3, s11, s3
-; GFX10-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x0
-; GFX10-NEXT:    s_cmp_ge_u32 s3, s7
-; GFX10-NEXT:    s_cselect_b32 s14, 1, 0
-; GFX10-NEXT:    s_add_i32 s16, s12, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX10-NEXT:    s_cselect_b32 s12, s16, s12
-; GFX10-NEXT:    s_sub_i32 s7, s3, s7
-; GFX10-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX10-NEXT:    s_cselect_b32 s3, s7, s3
-; GFX10-NEXT:    s_xor_b32 s7, s15, s13
-; GFX10-NEXT:    s_xor_b32 s13, s3, s15
-; GFX10-NEXT:    s_xor_b32 s12, s12, s7
-; GFX10-NEXT:    s_sub_i32 s3, s12, s7
-; GFX10-NEXT:    s_sub_i32 s7, s13, s15
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s5
-; GFX10-NEXT:    v_mov_b32_e32 v6, s6
+; GFX10-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s10
+; GFX10-NEXT:    s_mul_i32 s13, s13, s2
+; GFX10-NEXT:    s_mul_hi_u32 s13, s2, s13
+; GFX10-NEXT:    s_add_i32 s2, s2, s13
+; GFX10-NEXT:    s_mul_hi_u32 s2, s3, s2
+; GFX10-NEXT:    s_mul_i32 s13, s2, s7
+; GFX10-NEXT:    s_sub_i32 s1, s3, s13
+; GFX10-NEXT:    s_sub_i32 s13, s0, s14
+; GFX10-NEXT:    s_cmp_ge_u32 s1, s7
+; GFX10-NEXT:    v_mov_b32_e32 v6, s13
+; GFX10-NEXT:    s_cselect_b32 s0, 1, 0
+; GFX10-NEXT:    s_add_i32 s3, s2, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX10-NEXT:    s_cselect_b32 s14, s3, s2
+; GFX10-NEXT:    s_sub_i32 s2, s1, s7
+; GFX10-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX10-NEXT:    s_cselect_b32 s16, s2, s1
+; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX10-NEXT:    s_cmp_ge_u32 s16, s7
+; GFX10-NEXT:    s_cselect_b32 s17, 1, 0
+; GFX10-NEXT:    s_add_i32 s8, s14, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s17, 0
+; GFX10-NEXT:    s_cselect_b32 s8, s8, s14
+; GFX10-NEXT:    s_sub_i32 s7, s16, s7
+; GFX10-NEXT:    s_cmp_lg_u32 s17, 0
+; GFX10-NEXT:    s_cselect_b32 s7, s7, s16
+; GFX10-NEXT:    s_xor_b32 s5, s15, s5
+; GFX10-NEXT:    s_xor_b32 s7, s7, s15
+; GFX10-NEXT:    s_xor_b32 s8, s8, s5
+; GFX10-NEXT:    s_sub_i32 s7, s7, s15
+; GFX10-NEXT:    s_sub_i32 s5, s8, s5
 ; GFX10-NEXT:    v_mov_b32_e32 v7, s7
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX10-NEXT:    global_store_dwordx4 v8, v[4:7], s[10:11]
+; GFX10-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX10-NEXT:    global_store_dwordx4 v8, v[4:7], s[2:3]
 ; GFX10-NEXT:    s_endpgm
   %div = sdiv <4 x i32> %x, %y
   store <4 x i32> %div, ptr addrspace(1) %out0
@@ -1749,26 +1749,26 @@ define amdgpu_kernel void @sdivrem_v2i64(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX8-NEXT:    s_cmp_lg_u32 s27, 0
 ; GFX8-NEXT:    s_cselect_b64 s[4:5], s[4:5], s[8:9]
 ; GFX8-NEXT:    s_cmp_lg_u32 s24, 0
-; GFX8-NEXT:    s_cselect_b64 s[8:9], s[4:5], s[18:19]
-; GFX8-NEXT:    s_xor_b64 s[4:5], s[12:13], s[14:15]
-; GFX8-NEXT:    s_xor_b64 s[14:15], s[16:17], s[4:5]
-; GFX8-NEXT:    s_sub_u32 s4, s14, s4
-; GFX8-NEXT:    s_subb_u32 s5, s15, s5
-; GFX8-NEXT:    s_xor_b64 s[8:9], s[8:9], s[12:13]
-; GFX8-NEXT:    s_sub_u32 s8, s8, s12
-; GFX8-NEXT:    s_subb_u32 s9, s9, s12
-; GFX8-NEXT:    s_ashr_i32 s12, s7, 31
-; GFX8-NEXT:    s_ashr_i32 s14, s11, 31
-; GFX8-NEXT:    s_add_u32 s16, s6, s12
-; GFX8-NEXT:    s_addc_u32 s17, s7, s12
-; GFX8-NEXT:    s_add_u32 s6, s10, s14
-; GFX8-NEXT:    s_mov_b32 s15, s14
-; GFX8-NEXT:    s_addc_u32 s7, s11, s14
-; GFX8-NEXT:    s_xor_b64 s[6:7], s[6:7], s[14:15]
+; GFX8-NEXT:    s_cselect_b64 s[4:5], s[4:5], s[18:19]
+; GFX8-NEXT:    s_xor_b64 s[8:9], s[12:13], s[14:15]
+; GFX8-NEXT:    s_xor_b64 s[14:15], s[16:17], s[8:9]
+; GFX8-NEXT:    s_sub_u32 s14, s14, s8
+; GFX8-NEXT:    s_subb_u32 s15, s15, s9
+; GFX8-NEXT:    s_xor_b64 s[4:5], s[4:5], s[12:13]
+; GFX8-NEXT:    s_sub_u32 s13, s4, s12
+; GFX8-NEXT:    s_subb_u32 s12, s5, s12
+; GFX8-NEXT:    s_ashr_i32 s4, s7, 31
+; GFX8-NEXT:    s_ashr_i32 s8, s11, 31
+; GFX8-NEXT:    s_add_u32 s16, s6, s4
+; GFX8-NEXT:    s_addc_u32 s17, s7, s4
+; GFX8-NEXT:    s_add_u32 s6, s10, s8
+; GFX8-NEXT:    s_mov_b32 s9, s8
+; GFX8-NEXT:    s_addc_u32 s7, s11, s8
+; GFX8-NEXT:    s_xor_b64 s[6:7], s[6:7], s[8:9]
 ; GFX8-NEXT:    v_cvt_f32_u32_e32 v0, s7
 ; GFX8-NEXT:    v_cvt_f32_u32_e32 v1, s6
-; GFX8-NEXT:    s_mov_b32 s13, s12
-; GFX8-NEXT:    s_xor_b64 s[10:11], s[16:17], s[12:13]
+; GFX8-NEXT:    s_mov_b32 s5, s4
+; GFX8-NEXT:    s_xor_b64 s[10:11], s[16:17], s[4:5]
 ; GFX8-NEXT:    v_mul_f32_e32 v0, 0x4f800000, v0
 ; GFX8-NEXT:    v_add_f32_e32 v0, v0, v1
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v0, v0
@@ -1929,24 +1929,24 @@ define amdgpu_kernel void @sdivrem_v2i64(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX8-NEXT:    s_cmp_lg_u32 s27, 0
 ; GFX8-NEXT:    s_cselect_b64 s[6:7], s[6:7], s[10:11]
 ; GFX8-NEXT:    s_cmp_lg_u32 s24, 0
-; GFX8-NEXT:    s_cselect_b64 s[10:11], s[6:7], s[18:19]
-; GFX8-NEXT:    s_xor_b64 s[6:7], s[12:13], s[14:15]
-; GFX8-NEXT:    s_xor_b64 s[14:15], s[16:17], s[6:7]
-; GFX8-NEXT:    s_sub_u32 s6, s14, s6
-; GFX8-NEXT:    s_subb_u32 s7, s15, s7
-; GFX8-NEXT:    s_xor_b64 s[10:11], s[10:11], s[12:13]
-; GFX8-NEXT:    s_sub_u32 s10, s10, s12
-; GFX8-NEXT:    s_subb_u32 s11, s11, s12
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    v_mov_b32_e32 v2, s6
-; GFX8-NEXT:    v_mov_b32_e32 v3, s7
+; GFX8-NEXT:    s_cselect_b64 s[6:7], s[6:7], s[18:19]
+; GFX8-NEXT:    s_xor_b64 s[8:9], s[4:5], s[8:9]
+; GFX8-NEXT:    s_xor_b64 s[10:11], s[16:17], s[8:9]
+; GFX8-NEXT:    s_sub_u32 s8, s10, s8
+; GFX8-NEXT:    s_subb_u32 s9, s11, s9
+; GFX8-NEXT:    s_xor_b64 s[6:7], s[6:7], s[4:5]
+; GFX8-NEXT:    s_sub_u32 s5, s6, s4
+; GFX8-NEXT:    s_subb_u32 s4, s7, s4
+; GFX8-NEXT:    v_mov_b32_e32 v0, s14
+; GFX8-NEXT:    v_mov_b32_e32 v1, s15
+; GFX8-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NEXT:    v_mov_b32_e32 v3, s9
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v0, s8
-; GFX8-NEXT:    v_mov_b32_e32 v1, s9
-; GFX8-NEXT:    v_mov_b32_e32 v2, s10
-; GFX8-NEXT:    v_mov_b32_e32 v3, s11
+; GFX8-NEXT:    v_mov_b32_e32 v0, s13
+; GFX8-NEXT:    v_mov_b32_e32 v1, s12
+; GFX8-NEXT:    v_mov_b32_e32 v2, s5
+; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_endpgm
@@ -1957,18 +1957,18 @@ define amdgpu_kernel void @sdivrem_v2i64(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX9-NEXT:    s_load_dwordx4 s[12:15], s[8:9], 0x20
 ; GFX9-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_ashr_i32 s10, s5, 31
-; GFX9-NEXT:    s_ashr_i32 s8, s13, 31
-; GFX9-NEXT:    s_add_u32 s16, s4, s10
-; GFX9-NEXT:    s_addc_u32 s17, s5, s10
-; GFX9-NEXT:    s_add_u32 s4, s12, s8
-; GFX9-NEXT:    s_mov_b32 s9, s8
-; GFX9-NEXT:    s_addc_u32 s5, s13, s8
-; GFX9-NEXT:    s_xor_b64 s[4:5], s[4:5], s[8:9]
+; GFX9-NEXT:    s_ashr_i32 s8, s5, 31
+; GFX9-NEXT:    s_ashr_i32 s10, s13, 31
+; GFX9-NEXT:    s_add_u32 s16, s4, s8
+; GFX9-NEXT:    s_addc_u32 s17, s5, s8
+; GFX9-NEXT:    s_add_u32 s4, s12, s10
+; GFX9-NEXT:    s_mov_b32 s11, s10
+; GFX9-NEXT:    s_addc_u32 s5, s13, s10
+; GFX9-NEXT:    s_xor_b64 s[4:5], s[4:5], s[10:11]
 ; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s5
 ; GFX9-NEXT:    v_cvt_f32_u32_e32 v1, s4
-; GFX9-NEXT:    s_mov_b32 s11, s10
-; GFX9-NEXT:    s_xor_b64 s[12:13], s[16:17], s[10:11]
+; GFX9-NEXT:    s_mov_b32 s9, s8
+; GFX9-NEXT:    s_xor_b64 s[12:13], s[16:17], s[8:9]
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f800000, v0
 ; GFX9-NEXT:    v_add_f32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
@@ -2107,31 +2107,31 @@ define amdgpu_kernel void @sdivrem_v2i64(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX9-NEXT:    s_cmp_lg_u32 s27, 0
 ; GFX9-NEXT:    s_cselect_b64 s[4:5], s[4:5], s[12:13]
 ; GFX9-NEXT:    s_cmp_lg_u32 s24, 0
-; GFX9-NEXT:    s_cselect_b64 s[12:13], s[4:5], s[18:19]
-; GFX9-NEXT:    s_xor_b64 s[4:5], s[10:11], s[8:9]
-; GFX9-NEXT:    s_xor_b64 s[8:9], s[16:17], s[4:5]
-; GFX9-NEXT:    s_sub_u32 s4, s8, s4
-; GFX9-NEXT:    s_subb_u32 s5, s9, s5
-; GFX9-NEXT:    s_xor_b64 s[8:9], s[12:13], s[10:11]
-; GFX9-NEXT:    s_sub_u32 s8, s8, s10
-; GFX9-NEXT:    s_subb_u32 s9, s9, s10
-; GFX9-NEXT:    s_ashr_i32 s12, s7, 31
-; GFX9-NEXT:    s_ashr_i32 s10, s15, 31
-; GFX9-NEXT:    s_add_u32 s16, s6, s12
-; GFX9-NEXT:    s_addc_u32 s17, s7, s12
-; GFX9-NEXT:    s_add_u32 s6, s14, s10
-; GFX9-NEXT:    s_mov_b32 s11, s10
-; GFX9-NEXT:    s_addc_u32 s7, s15, s10
-; GFX9-NEXT:    s_xor_b64 s[6:7], s[6:7], s[10:11]
+; GFX9-NEXT:    s_cselect_b64 s[4:5], s[4:5], s[18:19]
+; GFX9-NEXT:    s_xor_b64 s[10:11], s[8:9], s[10:11]
+; GFX9-NEXT:    s_xor_b64 s[12:13], s[16:17], s[10:11]
+; GFX9-NEXT:    s_sub_u32 s22, s12, s10
+; GFX9-NEXT:    s_subb_u32 s23, s13, s11
+; GFX9-NEXT:    s_xor_b64 s[4:5], s[4:5], s[8:9]
+; GFX9-NEXT:    s_sub_u32 s10, s4, s8
+; GFX9-NEXT:    s_subb_u32 s11, s5, s8
+; GFX9-NEXT:    s_ashr_i32 s4, s7, 31
+; GFX9-NEXT:    s_ashr_i32 s8, s15, 31
+; GFX9-NEXT:    s_add_u32 s12, s6, s4
+; GFX9-NEXT:    s_addc_u32 s13, s7, s4
+; GFX9-NEXT:    s_add_u32 s6, s14, s8
+; GFX9-NEXT:    s_mov_b32 s9, s8
+; GFX9-NEXT:    s_addc_u32 s7, s15, s8
+; GFX9-NEXT:    s_xor_b64 s[6:7], s[6:7], s[8:9]
 ; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s7
 ; GFX9-NEXT:    v_cvt_f32_u32_e32 v1, s6
-; GFX9-NEXT:    s_mov_b32 s13, s12
-; GFX9-NEXT:    s_xor_b64 s[14:15], s[16:17], s[12:13]
+; GFX9-NEXT:    s_mov_b32 s5, s4
+; GFX9-NEXT:    s_xor_b64 s[12:13], s[12:13], s[4:5]
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f800000, v0
 ; GFX9-NEXT:    v_add_f32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
-; GFX9-NEXT:    s_sub_u32 s18, 0, s6
-; GFX9-NEXT:    s_subb_u32 s19, 0, s7
+; GFX9-NEXT:    s_sub_u32 s16, 0, s6
+; GFX9-NEXT:    s_subb_u32 s17, 0, s7
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x5f7ffffc, v0
 ; GFX9-NEXT:    v_mul_f32_e32 v1, 0x2f800000, v0
 ; GFX9-NEXT:    v_trunc_f32_e32 v1, v1
@@ -2139,150 +2139,150 @@ define amdgpu_kernel void @sdivrem_v2i64(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX9-NEXT:    v_add_f32_e32 v0, v2, v0
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v1, v1
-; GFX9-NEXT:    v_readfirstlane_b32 s16, v0
-; GFX9-NEXT:    v_readfirstlane_b32 s17, v1
-; GFX9-NEXT:    s_mul_hi_u32 s21, s18, s16
-; GFX9-NEXT:    s_mul_i32 s22, s18, s17
-; GFX9-NEXT:    s_add_u32 s21, s22, s21
-; GFX9-NEXT:    s_mul_i32 s22, s19, s16
-; GFX9-NEXT:    s_mul_i32 s20, s18, s16
-; GFX9-NEXT:    s_add_u32 s21, s22, s21
-; GFX9-NEXT:    s_mul_hi_u32 s22, s16, s20
-; GFX9-NEXT:    s_mul_i32 s23, s16, s21
-; GFX9-NEXT:    s_mul_i32 s24, s17, s20
-; GFX9-NEXT:    s_add_u32 s22, s22, s23
-; GFX9-NEXT:    s_cselect_b32 s23, 1, 0
-; GFX9-NEXT:    s_add_u32 s22, s22, s24
-; GFX9-NEXT:    s_cselect_b32 s22, 1, 0
-; GFX9-NEXT:    s_add_i32 s23, s23, s22
-; GFX9-NEXT:    s_mul_hi_u32 s20, s17, s20
-; GFX9-NEXT:    s_mul_hi_u32 s22, s16, s21
-; GFX9-NEXT:    s_mul_i32 s24, s17, s21
-; GFX9-NEXT:    s_add_u32 s20, s20, s22
-; GFX9-NEXT:    s_cselect_b32 s22, 1, 0
+; GFX9-NEXT:    v_readfirstlane_b32 s14, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s15, v1
+; GFX9-NEXT:    s_mul_hi_u32 s19, s16, s14
+; GFX9-NEXT:    s_mul_i32 s20, s16, s15
+; GFX9-NEXT:    s_add_u32 s19, s20, s19
+; GFX9-NEXT:    s_mul_i32 s20, s17, s14
+; GFX9-NEXT:    s_mul_i32 s18, s16, s14
+; GFX9-NEXT:    s_add_u32 s19, s20, s19
+; GFX9-NEXT:    s_mul_hi_u32 s20, s14, s18
+; GFX9-NEXT:    s_mul_i32 s21, s14, s19
+; GFX9-NEXT:    s_mul_i32 s24, s15, s18
+; GFX9-NEXT:    s_add_u32 s20, s20, s21
+; GFX9-NEXT:    s_cselect_b32 s21, 1, 0
 ; GFX9-NEXT:    s_add_u32 s20, s20, s24
+; GFX9-NEXT:    s_cselect_b32 s20, 1, 0
+; GFX9-NEXT:    s_add_i32 s21, s21, s20
+; GFX9-NEXT:    s_mul_hi_u32 s18, s15, s18
+; GFX9-NEXT:    s_mul_hi_u32 s20, s14, s19
+; GFX9-NEXT:    s_mul_i32 s24, s15, s19
+; GFX9-NEXT:    s_add_u32 s18, s18, s20
+; GFX9-NEXT:    s_cselect_b32 s20, 1, 0
+; GFX9-NEXT:    s_add_u32 s18, s18, s24
 ; GFX9-NEXT:    s_cselect_b32 s24, 1, 0
-; GFX9-NEXT:    s_add_i32 s22, s22, s24
-; GFX9-NEXT:    s_add_u32 s20, s20, s23
-; GFX9-NEXT:    s_cselect_b32 s23, 1, 0
-; GFX9-NEXT:    s_add_i32 s22, s22, s23
-; GFX9-NEXT:    s_mul_hi_u32 s21, s17, s21
-; GFX9-NEXT:    s_add_i32 s21, s21, s22
-; GFX9-NEXT:    s_add_u32 s16, s16, s20
-; GFX9-NEXT:    s_addc_u32 s17, s17, s21
-; GFX9-NEXT:    s_mul_i32 s20, s18, s16
-; GFX9-NEXT:    s_mul_hi_u32 s21, s18, s16
-; GFX9-NEXT:    s_mul_i32 s18, s18, s17
+; GFX9-NEXT:    s_add_i32 s20, s20, s24
 ; GFX9-NEXT:    s_add_u32 s18, s18, s21
-; GFX9-NEXT:    s_mul_i32 s19, s19, s16
-; GFX9-NEXT:    s_add_u32 s18, s19, s18
-; GFX9-NEXT:    s_mul_hi_u32 s19, s16, s20
-; GFX9-NEXT:    s_mul_i32 s21, s16, s18
-; GFX9-NEXT:    s_mul_i32 s22, s17, s20
-; GFX9-NEXT:    s_add_u32 s19, s19, s21
-; GFX9-NEXT:    s_cselect_b32 s21, 1, 0
-; GFX9-NEXT:    s_add_u32 s19, s19, s22
-; GFX9-NEXT:    s_cselect_b32 s19, 1, 0
-; GFX9-NEXT:    s_add_i32 s21, s21, s19
-; GFX9-NEXT:    s_mul_hi_u32 s19, s17, s20
-; GFX9-NEXT:    s_mul_hi_u32 s20, s16, s18
-; GFX9-NEXT:    s_mul_i32 s22, s17, s18
-; GFX9-NEXT:    s_add_u32 s19, s19, s20
-; GFX9-NEXT:    s_cselect_b32 s20, 1, 0
-; GFX9-NEXT:    s_add_u32 s19, s19, s22
-; GFX9-NEXT:    s_cselect_b32 s22, 1, 0
-; GFX9-NEXT:    s_add_i32 s20, s20, s22
-; GFX9-NEXT:    s_add_u32 s19, s19, s21
 ; GFX9-NEXT:    s_cselect_b32 s21, 1, 0
 ; GFX9-NEXT:    s_add_i32 s20, s20, s21
-; GFX9-NEXT:    s_mul_hi_u32 s18, s17, s18
-; GFX9-NEXT:    s_add_i32 s18, s18, s20
+; GFX9-NEXT:    s_mul_hi_u32 s19, s15, s19
+; GFX9-NEXT:    s_add_i32 s19, s19, s20
+; GFX9-NEXT:    s_add_u32 s14, s14, s18
+; GFX9-NEXT:    s_addc_u32 s15, s15, s19
+; GFX9-NEXT:    s_mul_i32 s18, s16, s14
+; GFX9-NEXT:    s_mul_hi_u32 s19, s16, s14
+; GFX9-NEXT:    s_mul_i32 s16, s16, s15
 ; GFX9-NEXT:    s_add_u32 s16, s16, s19
-; GFX9-NEXT:    s_addc_u32 s17, s17, s18
+; GFX9-NEXT:    s_mul_i32 s17, s17, s14
+; GFX9-NEXT:    s_add_u32 s16, s17, s16
+; GFX9-NEXT:    s_mul_hi_u32 s17, s14, s18
+; GFX9-NEXT:    s_mul_i32 s19, s14, s16
+; GFX9-NEXT:    s_mul_i32 s20, s15, s18
+; GFX9-NEXT:    s_add_u32 s17, s17, s19
+; GFX9-NEXT:    s_cselect_b32 s19, 1, 0
+; GFX9-NEXT:    s_add_u32 s17, s17, s20
+; GFX9-NEXT:    s_cselect_b32 s17, 1, 0
+; GFX9-NEXT:    s_add_i32 s19, s19, s17
+; GFX9-NEXT:    s_mul_hi_u32 s17, s15, s18
 ; GFX9-NEXT:    s_mul_hi_u32 s18, s14, s16
-; GFX9-NEXT:    s_mul_i32 s19, s14, s17
 ; GFX9-NEXT:    s_mul_i32 s20, s15, s16
-; GFX9-NEXT:    s_add_u32 s18, s18, s19
-; GFX9-NEXT:    s_cselect_b32 s19, 1, 0
-; GFX9-NEXT:    s_add_u32 s18, s18, s20
-; GFX9-NEXT:    s_cselect_b32 s18, 1, 0
-; GFX9-NEXT:    s_add_i32 s19, s19, s18
-; GFX9-NEXT:    s_mul_hi_u32 s16, s15, s16
-; GFX9-NEXT:    s_mul_hi_u32 s18, s14, s17
-; GFX9-NEXT:    s_mul_i32 s20, s15, s17
-; GFX9-NEXT:    s_add_u32 s16, s16, s18
+; GFX9-NEXT:    s_add_u32 s17, s17, s18
 ; GFX9-NEXT:    s_cselect_b32 s18, 1, 0
-; GFX9-NEXT:    s_add_u32 s16, s16, s20
+; GFX9-NEXT:    s_add_u32 s17, s17, s20
 ; GFX9-NEXT:    s_cselect_b32 s20, 1, 0
 ; GFX9-NEXT:    s_add_i32 s18, s18, s20
-; GFX9-NEXT:    s_add_u32 s16, s16, s19
+; GFX9-NEXT:    s_add_u32 s17, s17, s19
 ; GFX9-NEXT:    s_cselect_b32 s19, 1, 0
 ; GFX9-NEXT:    s_add_i32 s18, s18, s19
-; GFX9-NEXT:    s_mul_hi_u32 s17, s15, s17
-; GFX9-NEXT:    s_add_i32 s17, s17, s18
-; GFX9-NEXT:    s_mul_hi_u32 s19, s6, s16
-; GFX9-NEXT:    s_mul_i32 s20, s6, s17
-; GFX9-NEXT:    s_add_u32 s19, s20, s19
-; GFX9-NEXT:    s_mul_i32 s20, s7, s16
-; GFX9-NEXT:    s_mul_i32 s18, s6, s16
-; GFX9-NEXT:    s_add_u32 s20, s20, s19
-; GFX9-NEXT:    s_sub_u32 s18, s14, s18
-; GFX9-NEXT:    s_cselect_b32 s21, 1, 0
-; GFX9-NEXT:    s_subb_u32 s19, s15, s20
-; GFX9-NEXT:    s_sub_i32 s15, s15, s20
-; GFX9-NEXT:    s_cmp_ge_u32 s19, s7
-; GFX9-NEXT:    s_cselect_b32 s14, -1, 0
-; GFX9-NEXT:    s_cmp_ge_u32 s18, s6
-; GFX9-NEXT:    s_cselect_b32 s20, -1, 0
-; GFX9-NEXT:    s_cmp_eq_u32 s19, s7
-; GFX9-NEXT:    s_cselect_b32 s24, s20, s14
-; GFX9-NEXT:    s_sub_u32 s14, s18, s6
+; GFX9-NEXT:    s_mul_hi_u32 s16, s15, s16
+; GFX9-NEXT:    s_add_i32 s16, s16, s18
+; GFX9-NEXT:    s_add_u32 s14, s14, s17
+; GFX9-NEXT:    s_addc_u32 s15, s15, s16
+; GFX9-NEXT:    s_mul_hi_u32 s16, s12, s14
+; GFX9-NEXT:    s_mul_i32 s17, s12, s15
+; GFX9-NEXT:    s_mul_i32 s18, s13, s14
+; GFX9-NEXT:    s_add_u32 s16, s16, s17
+; GFX9-NEXT:    s_cselect_b32 s17, 1, 0
+; GFX9-NEXT:    s_add_u32 s16, s16, s18
+; GFX9-NEXT:    s_cselect_b32 s16, 1, 0
+; GFX9-NEXT:    s_add_i32 s17, s17, s16
+; GFX9-NEXT:    s_mul_hi_u32 s14, s13, s14
+; GFX9-NEXT:    s_mul_hi_u32 s16, s12, s15
+; GFX9-NEXT:    s_mul_i32 s18, s13, s15
+; GFX9-NEXT:    s_add_u32 s14, s14, s16
+; GFX9-NEXT:    s_cselect_b32 s16, 1, 0
+; GFX9-NEXT:    s_add_u32 s14, s14, s18
+; GFX9-NEXT:    s_cselect_b32 s18, 1, 0
+; GFX9-NEXT:    s_add_i32 s16, s16, s18
+; GFX9-NEXT:    s_add_u32 s14, s14, s17
+; GFX9-NEXT:    s_cselect_b32 s17, 1, 0
+; GFX9-NEXT:    s_add_i32 s16, s16, s17
+; GFX9-NEXT:    s_mul_hi_u32 s15, s13, s15
+; GFX9-NEXT:    s_add_i32 s15, s15, s16
+; GFX9-NEXT:    s_mul_hi_u32 s17, s6, s14
+; GFX9-NEXT:    s_mul_i32 s18, s6, s15
+; GFX9-NEXT:    s_add_u32 s17, s18, s17
+; GFX9-NEXT:    s_mul_i32 s18, s7, s14
+; GFX9-NEXT:    s_mul_i32 s16, s6, s14
+; GFX9-NEXT:    s_add_u32 s18, s18, s17
+; GFX9-NEXT:    s_sub_u32 s16, s12, s16
+; GFX9-NEXT:    s_cselect_b32 s19, 1, 0
+; GFX9-NEXT:    s_subb_u32 s17, s13, s18
+; GFX9-NEXT:    s_sub_i32 s13, s13, s18
+; GFX9-NEXT:    s_cmp_ge_u32 s17, s7
+; GFX9-NEXT:    s_cselect_b32 s12, -1, 0
+; GFX9-NEXT:    s_cmp_ge_u32 s16, s6
+; GFX9-NEXT:    s_cselect_b32 s18, -1, 0
+; GFX9-NEXT:    s_cmp_eq_u32 s17, s7
+; GFX9-NEXT:    s_cselect_b32 s24, s18, s12
+; GFX9-NEXT:    s_sub_u32 s12, s16, s6
 ; GFX9-NEXT:    s_cselect_b32 s25, 1, 0
-; GFX9-NEXT:    s_cmp_lg_u32 s21, 0
-; GFX9-NEXT:    s_subb_u32 s26, s15, s7
+; GFX9-NEXT:    s_cmp_lg_u32 s19, 0
+; GFX9-NEXT:    s_subb_u32 s26, s13, s7
 ; GFX9-NEXT:    s_cmp_lg_u32 s25, 0
-; GFX9-NEXT:    s_subb_u32 s15, s26, 0
-; GFX9-NEXT:    s_add_u32 s20, s16, 1
-; GFX9-NEXT:    s_addc_u32 s21, s17, 0
-; GFX9-NEXT:    s_cmp_ge_u32 s15, s7
-; GFX9-NEXT:    s_cselect_b32 s22, -1, 0
-; GFX9-NEXT:    s_cmp_ge_u32 s14, s6
-; GFX9-NEXT:    s_cselect_b32 s23, -1, 0
-; GFX9-NEXT:    s_cmp_eq_u32 s15, s7
-; GFX9-NEXT:    s_cselect_b32 s27, s23, s22
-; GFX9-NEXT:    s_add_u32 s22, s20, 1
-; GFX9-NEXT:    s_addc_u32 s23, s21, 0
-; GFX9-NEXT:    s_sub_u32 s6, s14, s6
+; GFX9-NEXT:    s_subb_u32 s13, s26, 0
+; GFX9-NEXT:    s_add_u32 s18, s14, 1
+; GFX9-NEXT:    s_addc_u32 s19, s15, 0
+; GFX9-NEXT:    s_cmp_ge_u32 s13, s7
+; GFX9-NEXT:    s_cselect_b32 s20, -1, 0
+; GFX9-NEXT:    s_cmp_ge_u32 s12, s6
+; GFX9-NEXT:    s_cselect_b32 s21, -1, 0
+; GFX9-NEXT:    s_cmp_eq_u32 s13, s7
+; GFX9-NEXT:    s_cselect_b32 s27, s21, s20
+; GFX9-NEXT:    s_add_u32 s20, s18, 1
+; GFX9-NEXT:    s_addc_u32 s21, s19, 0
+; GFX9-NEXT:    s_sub_u32 s6, s12, s6
 ; GFX9-NEXT:    s_cselect_b32 s28, 1, 0
 ; GFX9-NEXT:    s_cmp_lg_u32 s25, 0
 ; GFX9-NEXT:    s_subb_u32 s7, s26, s7
 ; GFX9-NEXT:    s_cmp_lg_u32 s28, 0
 ; GFX9-NEXT:    s_subb_u32 s7, s7, 0
 ; GFX9-NEXT:    s_cmp_lg_u32 s27, 0
-; GFX9-NEXT:    s_cselect_b64 s[20:21], s[22:23], s[20:21]
+; GFX9-NEXT:    s_cselect_b64 s[18:19], s[20:21], s[18:19]
 ; GFX9-NEXT:    s_cmp_lg_u32 s24, 0
-; GFX9-NEXT:    s_cselect_b64 s[16:17], s[20:21], s[16:17]
+; GFX9-NEXT:    s_cselect_b64 s[14:15], s[18:19], s[14:15]
 ; GFX9-NEXT:    s_cmp_lg_u32 s27, 0
-; GFX9-NEXT:    s_cselect_b64 s[6:7], s[6:7], s[14:15]
+; GFX9-NEXT:    s_cselect_b64 s[6:7], s[6:7], s[12:13]
 ; GFX9-NEXT:    s_cmp_lg_u32 s24, 0
-; GFX9-NEXT:    s_cselect_b64 s[14:15], s[6:7], s[18:19]
-; GFX9-NEXT:    s_xor_b64 s[6:7], s[12:13], s[10:11]
-; GFX9-NEXT:    s_xor_b64 s[10:11], s[16:17], s[6:7]
-; GFX9-NEXT:    s_sub_u32 s6, s10, s6
-; GFX9-NEXT:    s_subb_u32 s7, s11, s7
-; GFX9-NEXT:    s_xor_b64 s[10:11], s[14:15], s[12:13]
-; GFX9-NEXT:    s_sub_u32 s10, s10, s12
-; GFX9-NEXT:    s_subb_u32 s11, s11, s12
-; GFX9-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9-NEXT:    v_mov_b32_e32 v1, s5
-; GFX9-NEXT:    v_mov_b32_e32 v2, s6
-; GFX9-NEXT:    v_mov_b32_e32 v3, s7
+; GFX9-NEXT:    s_cselect_b64 s[6:7], s[6:7], s[16:17]
+; GFX9-NEXT:    s_xor_b64 s[8:9], s[4:5], s[8:9]
+; GFX9-NEXT:    s_xor_b64 s[12:13], s[14:15], s[8:9]
+; GFX9-NEXT:    s_sub_u32 s8, s12, s8
+; GFX9-NEXT:    s_subb_u32 s9, s13, s9
+; GFX9-NEXT:    s_xor_b64 s[6:7], s[6:7], s[4:5]
+; GFX9-NEXT:    s_sub_u32 s5, s6, s4
+; GFX9-NEXT:    s_subb_u32 s4, s7, s4
+; GFX9-NEXT:    v_mov_b32_e32 v0, s22
+; GFX9-NEXT:    v_mov_b32_e32 v1, s23
+; GFX9-NEXT:    v_mov_b32_e32 v2, s8
+; GFX9-NEXT:    v_mov_b32_e32 v3, s9
 ; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX9-NEXT:    s_nop 0
-; GFX9-NEXT:    v_mov_b32_e32 v0, s8
-; GFX9-NEXT:    v_mov_b32_e32 v1, s9
-; GFX9-NEXT:    v_mov_b32_e32 v2, s10
-; GFX9-NEXT:    v_mov_b32_e32 v3, s11
+; GFX9-NEXT:    v_mov_b32_e32 v0, s10
+; GFX9-NEXT:    v_mov_b32_e32 v1, s11
+; GFX9-NEXT:    v_mov_b32_e32 v2, s5
+; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[2:3]
 ; GFX9-NEXT:    s_endpgm
 ;
@@ -2293,16 +2293,16 @@ define amdgpu_kernel void @sdivrem_v2i64(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX10-NEXT:    s_load_dwordx8 s[0:7], s[8:9], 0x0
 ; GFX10-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    s_ashr_i32 s8, s13, 31
-; GFX10-NEXT:    s_ashr_i32 s10, s5, 31
-; GFX10-NEXT:    s_mov_b32 s9, s8
-; GFX10-NEXT:    s_add_u32 s16, s4, s10
-; GFX10-NEXT:    s_addc_u32 s17, s5, s10
-; GFX10-NEXT:    s_add_u32 s4, s12, s8
-; GFX10-NEXT:    s_addc_u32 s5, s13, s8
+; GFX10-NEXT:    s_ashr_i32 s10, s13, 31
+; GFX10-NEXT:    s_ashr_i32 s8, s5, 31
 ; GFX10-NEXT:    s_mov_b32 s11, s10
-; GFX10-NEXT:    s_xor_b64 s[4:5], s[4:5], s[8:9]
-; GFX10-NEXT:    s_xor_b64 s[12:13], s[16:17], s[10:11]
+; GFX10-NEXT:    s_add_u32 s16, s4, s8
+; GFX10-NEXT:    s_addc_u32 s17, s5, s8
+; GFX10-NEXT:    s_add_u32 s4, s12, s10
+; GFX10-NEXT:    s_addc_u32 s5, s13, s10
+; GFX10-NEXT:    s_mov_b32 s9, s8
+; GFX10-NEXT:    s_xor_b64 s[4:5], s[4:5], s[10:11]
+; GFX10-NEXT:    s_xor_b64 s[12:13], s[16:17], s[8:9]
 ; GFX10-NEXT:    v_cvt_f32_u32_e32 v0, s5
 ; GFX10-NEXT:    v_cvt_f32_u32_e32 v1, s4
 ; GFX10-NEXT:    s_sub_u32 s16, 0, s4
@@ -2443,29 +2443,31 @@ define amdgpu_kernel void @sdivrem_v2i64(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX10-NEXT:    s_cmp_lg_u32 s27, 0
 ; GFX10-NEXT:    s_cselect_b64 s[4:5], s[4:5], s[12:13]
 ; GFX10-NEXT:    s_cmp_lg_u32 s24, 0
-; GFX10-NEXT:    s_cselect_b64 s[12:13], s[4:5], s[18:19]
-; GFX10-NEXT:    s_xor_b64 s[4:5], s[10:11], s[8:9]
-; GFX10-NEXT:    s_xor_b64 s[8:9], s[16:17], s[4:5]
-; GFX10-NEXT:    s_sub_u32 s4, s8, s4
-; GFX10-NEXT:    s_subb_u32 s5, s9, s5
-; GFX10-NEXT:    s_xor_b64 s[8:9], s[12:13], s[10:11]
-; GFX10-NEXT:    s_sub_u32 s8, s8, s10
-; GFX10-NEXT:    s_subb_u32 s9, s9, s10
-; GFX10-NEXT:    s_ashr_i32 s12, s7, 31
-; GFX10-NEXT:    s_ashr_i32 s10, s15, 31
-; GFX10-NEXT:    s_add_u32 s16, s6, s12
-; GFX10-NEXT:    s_addc_u32 s17, s7, s12
-; GFX10-NEXT:    s_add_u32 s6, s14, s10
-; GFX10-NEXT:    s_mov_b32 s11, s10
-; GFX10-NEXT:    s_addc_u32 s7, s15, s10
-; GFX10-NEXT:    s_mov_b32 s13, s12
-; GFX10-NEXT:    s_xor_b64 s[6:7], s[6:7], s[10:11]
-; GFX10-NEXT:    s_xor_b64 s[14:15], s[16:17], s[12:13]
+; GFX10-NEXT:    s_cselect_b64 s[4:5], s[4:5], s[18:19]
+; GFX10-NEXT:    s_xor_b64 s[10:11], s[8:9], s[10:11]
+; GFX10-NEXT:    s_xor_b64 s[12:13], s[16:17], s[10:11]
+; GFX10-NEXT:    s_sub_u32 s16, s12, s10
+; GFX10-NEXT:    s_subb_u32 s17, s13, s11
+; GFX10-NEXT:    s_xor_b64 s[4:5], s[4:5], s[8:9]
+; GFX10-NEXT:    s_sub_u32 s12, s4, s8
+; GFX10-NEXT:    s_subb_u32 s13, s5, s8
+; GFX10-NEXT:    s_ashr_i32 s4, s7, 31
+; GFX10-NEXT:    s_ashr_i32 s8, s15, 31
+; GFX10-NEXT:    s_add_u32 s10, s6, s4
+; GFX10-NEXT:    s_addc_u32 s11, s7, s4
+; GFX10-NEXT:    s_add_u32 s6, s14, s8
+; GFX10-NEXT:    s_mov_b32 s9, s8
+; GFX10-NEXT:    s_addc_u32 s7, s15, s8
+; GFX10-NEXT:    s_mov_b32 s5, s4
+; GFX10-NEXT:    s_xor_b64 s[6:7], s[6:7], s[8:9]
+; GFX10-NEXT:    s_xor_b64 s[10:11], s[10:11], s[4:5]
 ; GFX10-NEXT:    v_cvt_f32_u32_e32 v0, s7
 ; GFX10-NEXT:    v_cvt_f32_u32_e32 v1, s6
-; GFX10-NEXT:    s_sub_u32 s16, 0, s6
-; GFX10-NEXT:    s_subb_u32 s17, 0, s7
+; GFX10-NEXT:    s_sub_u32 s18, 0, s6
+; GFX10-NEXT:    s_subb_u32 s19, 0, s7
+; GFX10-NEXT:    v_mov_b32_e32 v4, s12
 ; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f800000, v0
+; GFX10-NEXT:    v_mov_b32_e32 v5, s13
 ; GFX10-NEXT:    v_add_f32_e32 v0, v0, v1
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX10-NEXT:    v_mul_f32_e32 v0, 0x5f7ffffc, v0
@@ -2474,121 +2476,123 @@ define amdgpu_kernel void @sdivrem_v2i64(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX10-NEXT:    v_mul_f32_e32 v2, 0xcf800000, v1
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX10-NEXT:    v_add_f32_e32 v0, v2, v0
-; GFX10-NEXT:    v_readfirstlane_b32 s18, v1
+; GFX10-NEXT:    v_readfirstlane_b32 s14, v1
+; GFX10-NEXT:    v_mov_b32_e32 v1, s17
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX10-NEXT:    s_mul_i32 s20, s16, s18
-; GFX10-NEXT:    v_readfirstlane_b32 s19, v0
-; GFX10-NEXT:    s_mul_hi_u32 s22, s16, s19
-; GFX10-NEXT:    s_mul_i32 s23, s17, s19
+; GFX10-NEXT:    s_mul_i32 s20, s18, s14
+; GFX10-NEXT:    v_readfirstlane_b32 s15, v0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s16
+; GFX10-NEXT:    s_mul_hi_u32 s22, s18, s15
+; GFX10-NEXT:    s_mul_i32 s23, s19, s15
 ; GFX10-NEXT:    s_add_i32 s20, s22, s20
-; GFX10-NEXT:    s_mul_i32 s21, s16, s19
+; GFX10-NEXT:    s_mul_i32 s21, s18, s15
 ; GFX10-NEXT:    s_add_i32 s20, s20, s23
-; GFX10-NEXT:    s_mul_hi_u32 s22, s19, s21
-; GFX10-NEXT:    s_mul_i32 s23, s19, s20
-; GFX10-NEXT:    s_mul_i32 s24, s18, s21
+; GFX10-NEXT:    s_mul_hi_u32 s22, s15, s21
+; GFX10-NEXT:    s_mul_i32 s23, s15, s20
+; GFX10-NEXT:    s_mul_i32 s24, s14, s21
 ; GFX10-NEXT:    s_add_u32 s22, s22, s23
 ; GFX10-NEXT:    s_cselect_b32 s23, 1, 0
 ; GFX10-NEXT:    s_add_u32 s22, s22, s24
 ; GFX10-NEXT:    s_cselect_b32 s22, 1, 0
-; GFX10-NEXT:    s_mul_hi_u32 s21, s18, s21
-; GFX10-NEXT:    s_mul_hi_u32 s25, s19, s20
+; GFX10-NEXT:    s_mul_hi_u32 s21, s14, s21
+; GFX10-NEXT:    s_mul_hi_u32 s25, s15, s20
 ; GFX10-NEXT:    s_add_i32 s23, s23, s22
-; GFX10-NEXT:    s_mul_i32 s24, s18, s20
+; GFX10-NEXT:    s_mul_i32 s24, s14, s20
 ; GFX10-NEXT:    s_add_u32 s21, s21, s25
 ; GFX10-NEXT:    s_cselect_b32 s22, 1, 0
 ; GFX10-NEXT:    s_add_u32 s21, s21, s24
 ; GFX10-NEXT:    s_cselect_b32 s24, 1, 0
-; GFX10-NEXT:    s_mul_hi_u32 s20, s18, s20
+; GFX10-NEXT:    s_mul_hi_u32 s20, s14, s20
 ; GFX10-NEXT:    s_add_i32 s22, s22, s24
 ; GFX10-NEXT:    s_add_u32 s21, s21, s23
 ; GFX10-NEXT:    s_cselect_b32 s23, 1, 0
 ; GFX10-NEXT:    s_add_i32 s22, s22, s23
 ; GFX10-NEXT:    s_add_i32 s20, s20, s22
-; GFX10-NEXT:    s_add_u32 s19, s19, s21
-; GFX10-NEXT:    s_addc_u32 s18, s18, s20
-; GFX10-NEXT:    s_mul_i32 s20, s16, s19
-; GFX10-NEXT:    s_mul_hi_u32 s21, s16, s19
-; GFX10-NEXT:    s_mul_i32 s16, s16, s18
-; GFX10-NEXT:    s_mul_i32 s17, s17, s19
-; GFX10-NEXT:    s_add_i32 s16, s21, s16
-; GFX10-NEXT:    s_mul_hi_u32 s22, s19, s20
-; GFX10-NEXT:    s_add_i32 s16, s16, s17
-; GFX10-NEXT:    s_mul_i32 s21, s18, s20
-; GFX10-NEXT:    s_mul_hi_u32 s17, s18, s20
-; GFX10-NEXT:    s_mul_i32 s20, s19, s16
-; GFX10-NEXT:    s_mul_hi_u32 s23, s19, s16
+; GFX10-NEXT:    s_add_u32 s15, s15, s21
+; GFX10-NEXT:    s_addc_u32 s14, s14, s20
+; GFX10-NEXT:    s_mul_i32 s20, s18, s15
+; GFX10-NEXT:    s_mul_hi_u32 s21, s18, s15
+; GFX10-NEXT:    s_mul_i32 s18, s18, s14
+; GFX10-NEXT:    s_mul_i32 s19, s19, s15
+; GFX10-NEXT:    s_add_i32 s18, s21, s18
+; GFX10-NEXT:    s_mul_hi_u32 s22, s15, s20
+; GFX10-NEXT:    s_add_i32 s18, s18, s19
+; GFX10-NEXT:    s_mul_i32 s21, s14, s20
+; GFX10-NEXT:    s_mul_hi_u32 s19, s14, s20
+; GFX10-NEXT:    s_mul_i32 s20, s15, s18
+; GFX10-NEXT:    s_mul_hi_u32 s23, s15, s18
 ; GFX10-NEXT:    s_add_u32 s20, s22, s20
 ; GFX10-NEXT:    s_cselect_b32 s22, 1, 0
 ; GFX10-NEXT:    s_add_u32 s20, s20, s21
 ; GFX10-NEXT:    s_cselect_b32 s20, 1, 0
-; GFX10-NEXT:    s_mul_i32 s21, s18, s16
+; GFX10-NEXT:    s_mul_i32 s21, s14, s18
 ; GFX10-NEXT:    s_add_i32 s22, s22, s20
-; GFX10-NEXT:    s_add_u32 s17, s17, s23
+; GFX10-NEXT:    s_add_u32 s19, s19, s23
 ; GFX10-NEXT:    s_cselect_b32 s20, 1, 0
-; GFX10-NEXT:    s_add_u32 s17, s17, s21
+; GFX10-NEXT:    s_add_u32 s19, s19, s21
 ; GFX10-NEXT:    s_cselect_b32 s21, 1, 0
-; GFX10-NEXT:    s_mul_hi_u32 s16, s18, s16
+; GFX10-NEXT:    s_mul_hi_u32 s18, s14, s18
 ; GFX10-NEXT:    s_add_i32 s20, s20, s21
-; GFX10-NEXT:    s_add_u32 s17, s17, s22
+; GFX10-NEXT:    s_add_u32 s19, s19, s22
 ; GFX10-NEXT:    s_cselect_b32 s21, 1, 0
 ; GFX10-NEXT:    s_add_i32 s20, s20, s21
-; GFX10-NEXT:    s_add_i32 s16, s16, s20
-; GFX10-NEXT:    s_add_u32 s17, s19, s17
-; GFX10-NEXT:    s_addc_u32 s18, s18, s16
-; GFX10-NEXT:    s_mul_hi_u32 s16, s14, s17
-; GFX10-NEXT:    s_mul_i32 s20, s14, s18
-; GFX10-NEXT:    s_mul_i32 s19, s15, s17
-; GFX10-NEXT:    s_add_u32 s16, s16, s20
+; GFX10-NEXT:    s_add_i32 s18, s18, s20
+; GFX10-NEXT:    s_add_u32 s15, s15, s19
+; GFX10-NEXT:    s_addc_u32 s18, s14, s18
+; GFX10-NEXT:    s_mul_hi_u32 s14, s10, s15
+; GFX10-NEXT:    s_mul_i32 s20, s10, s18
+; GFX10-NEXT:    s_mul_i32 s19, s11, s15
+; GFX10-NEXT:    s_add_u32 s14, s14, s20
 ; GFX10-NEXT:    s_cselect_b32 s20, 1, 0
-; GFX10-NEXT:    s_add_u32 s16, s16, s19
-; GFX10-NEXT:    s_cselect_b32 s16, 1, 0
-; GFX10-NEXT:    s_mul_hi_u32 s17, s15, s17
-; GFX10-NEXT:    s_mul_hi_u32 s21, s14, s18
-; GFX10-NEXT:    s_add_i32 s20, s20, s16
-; GFX10-NEXT:    s_mul_i32 s19, s15, s18
-; GFX10-NEXT:    s_add_u32 s16, s17, s21
-; GFX10-NEXT:    s_cselect_b32 s17, 1, 0
-; GFX10-NEXT:    s_add_u32 s16, s16, s19
+; GFX10-NEXT:    s_add_u32 s14, s14, s19
+; GFX10-NEXT:    s_cselect_b32 s14, 1, 0
+; GFX10-NEXT:    s_mul_hi_u32 s15, s11, s15
+; GFX10-NEXT:    s_mul_hi_u32 s21, s10, s18
+; GFX10-NEXT:    s_add_i32 s20, s20, s14
+; GFX10-NEXT:    s_mul_i32 s19, s11, s18
+; GFX10-NEXT:    s_add_u32 s14, s15, s21
+; GFX10-NEXT:    s_cselect_b32 s15, 1, 0
+; GFX10-NEXT:    s_add_u32 s14, s14, s19
 ; GFX10-NEXT:    s_cselect_b32 s19, 1, 0
-; GFX10-NEXT:    s_mul_hi_u32 s18, s15, s18
-; GFX10-NEXT:    s_add_i32 s17, s17, s19
-; GFX10-NEXT:    s_add_u32 s16, s16, s20
+; GFX10-NEXT:    s_mul_hi_u32 s18, s11, s18
+; GFX10-NEXT:    s_add_i32 s15, s15, s19
+; GFX10-NEXT:    s_add_u32 s14, s14, s20
 ; GFX10-NEXT:    s_cselect_b32 s19, 1, 0
-; GFX10-NEXT:    s_mul_i32 s20, s7, s16
-; GFX10-NEXT:    s_add_i32 s17, s17, s19
-; GFX10-NEXT:    s_add_i32 s17, s18, s17
-; GFX10-NEXT:    s_mul_hi_u32 s18, s6, s16
-; GFX10-NEXT:    s_mul_i32 s19, s6, s17
+; GFX10-NEXT:    s_mul_i32 s20, s7, s14
+; GFX10-NEXT:    s_add_i32 s15, s15, s19
+; GFX10-NEXT:    s_add_i32 s15, s18, s15
+; GFX10-NEXT:    s_mul_hi_u32 s18, s6, s14
+; GFX10-NEXT:    s_mul_i32 s19, s6, s15
 ; GFX10-NEXT:    s_add_i32 s18, s18, s19
-; GFX10-NEXT:    s_mul_i32 s19, s6, s16
+; GFX10-NEXT:    s_mul_i32 s19, s6, s14
 ; GFX10-NEXT:    s_add_i32 s20, s18, s20
-; GFX10-NEXT:    s_sub_u32 s18, s14, s19
+; GFX10-NEXT:    s_sub_u32 s18, s10, s19
 ; GFX10-NEXT:    s_cselect_b32 s21, 1, 0
-; GFX10-NEXT:    s_subb_u32 s19, s15, s20
-; GFX10-NEXT:    s_sub_i32 s15, s15, s20
+; GFX10-NEXT:    s_subb_u32 s19, s11, s20
+; GFX10-NEXT:    s_sub_i32 s11, s11, s20
 ; GFX10-NEXT:    s_cmp_ge_u32 s19, s7
-; GFX10-NEXT:    s_cselect_b32 s14, -1, 0
+; GFX10-NEXT:    s_cselect_b32 s10, -1, 0
 ; GFX10-NEXT:    s_cmp_ge_u32 s18, s6
 ; GFX10-NEXT:    s_cselect_b32 s20, -1, 0
 ; GFX10-NEXT:    s_cmp_eq_u32 s19, s7
-; GFX10-NEXT:    s_cselect_b32 s24, s20, s14
-; GFX10-NEXT:    s_sub_u32 s14, s18, s6
+; GFX10-NEXT:    s_cselect_b32 s24, s20, s10
+; GFX10-NEXT:    s_sub_u32 s10, s18, s6
 ; GFX10-NEXT:    s_cselect_b32 s25, 1, 0
 ; GFX10-NEXT:    s_cmp_lg_u32 s21, 0
-; GFX10-NEXT:    s_subb_u32 s26, s15, s7
+; GFX10-NEXT:    s_subb_u32 s26, s11, s7
 ; GFX10-NEXT:    s_cmp_lg_u32 s25, 0
-; GFX10-NEXT:    s_subb_u32 s15, s26, 0
-; GFX10-NEXT:    s_add_u32 s20, s16, 1
-; GFX10-NEXT:    s_addc_u32 s21, s17, 0
-; GFX10-NEXT:    s_cmp_ge_u32 s15, s7
+; GFX10-NEXT:    s_subb_u32 s11, s26, 0
+; GFX10-NEXT:    s_add_u32 s20, s14, 1
+; GFX10-NEXT:    s_addc_u32 s21, s15, 0
+; GFX10-NEXT:    s_cmp_ge_u32 s11, s7
 ; GFX10-NEXT:    s_cselect_b32 s22, -1, 0
-; GFX10-NEXT:    s_cmp_ge_u32 s14, s6
+; GFX10-NEXT:    s_cmp_ge_u32 s10, s6
 ; GFX10-NEXT:    s_cselect_b32 s23, -1, 0
-; GFX10-NEXT:    s_cmp_eq_u32 s15, s7
+; GFX10-NEXT:    s_cmp_eq_u32 s11, s7
 ; GFX10-NEXT:    s_cselect_b32 s27, s23, s22
 ; GFX10-NEXT:    s_add_u32 s22, s20, 1
 ; GFX10-NEXT:    s_addc_u32 s23, s21, 0
-; GFX10-NEXT:    s_sub_u32 s6, s14, s6
+; GFX10-NEXT:    s_sub_u32 s6, s10, s6
 ; GFX10-NEXT:    s_cselect_b32 s28, 1, 0
 ; GFX10-NEXT:    s_cmp_lg_u32 s25, 0
 ; GFX10-NEXT:    s_subb_u32 s7, s26, s7
@@ -2597,26 +2601,22 @@ define amdgpu_kernel void @sdivrem_v2i64(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX10-NEXT:    s_cmp_lg_u32 s27, 0
 ; GFX10-NEXT:    s_cselect_b64 s[20:21], s[22:23], s[20:21]
 ; GFX10-NEXT:    s_cmp_lg_u32 s24, 0
-; GFX10-NEXT:    s_cselect_b64 s[16:17], s[20:21], s[16:17]
+; GFX10-NEXT:    s_cselect_b64 s[14:15], s[20:21], s[14:15]
 ; GFX10-NEXT:    s_cmp_lg_u32 s27, 0
-; GFX10-NEXT:    s_cselect_b64 s[6:7], s[6:7], s[14:15]
+; GFX10-NEXT:    s_cselect_b64 s[6:7], s[6:7], s[10:11]
 ; GFX10-NEXT:    s_cmp_lg_u32 s24, 0
-; GFX10-NEXT:    s_cselect_b64 s[14:15], s[6:7], s[18:19]
-; GFX10-NEXT:    s_xor_b64 s[6:7], s[12:13], s[10:11]
-; GFX10-NEXT:    s_xor_b64 s[10:11], s[16:17], s[6:7]
-; GFX10-NEXT:    s_sub_u32 s6, s10, s6
-; GFX10-NEXT:    s_subb_u32 s7, s11, s7
-; GFX10-NEXT:    s_xor_b64 s[10:11], s[14:15], s[12:13]
-; GFX10-NEXT:    v_mov_b32_e32 v0, s4
-; GFX10-NEXT:    s_sub_u32 s10, s10, s12
-; GFX10-NEXT:    s_subb_u32 s11, s11, s12
-; GFX10-NEXT:    v_mov_b32_e32 v1, s5
-; GFX10-NEXT:    v_mov_b32_e32 v2, s6
-; GFX10-NEXT:    v_mov_b32_e32 v3, s7
-; GFX10-NEXT:    v_mov_b32_e32 v4, s8
-; GFX10-NEXT:    v_mov_b32_e32 v5, s9
-; GFX10-NEXT:    v_mov_b32_e32 v6, s10
-; GFX10-NEXT:    v_mov_b32_e32 v7, s11
+; GFX10-NEXT:    s_cselect_b64 s[6:7], s[6:7], s[18:19]
+; GFX10-NEXT:    s_xor_b64 s[8:9], s[4:5], s[8:9]
+; GFX10-NEXT:    s_xor_b64 s[10:11], s[14:15], s[8:9]
+; GFX10-NEXT:    s_sub_u32 s8, s10, s8
+; GFX10-NEXT:    s_subb_u32 s9, s11, s9
+; GFX10-NEXT:    s_xor_b64 s[6:7], s[6:7], s[4:5]
+; GFX10-NEXT:    v_mov_b32_e32 v2, s8
+; GFX10-NEXT:    s_sub_u32 s5, s6, s4
+; GFX10-NEXT:    s_subb_u32 s4, s7, s4
+; GFX10-NEXT:    v_mov_b32_e32 v3, s9
+; GFX10-NEXT:    v_mov_b32_e32 v6, s5
+; GFX10-NEXT:    v_mov_b32_e32 v7, s4
 ; GFX10-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
 ; GFX10-NEXT:    global_store_dwordx4 v8, v[4:7], s[2:3]
 ; GFX10-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/strict_fma.f32.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/strict_fma.f32.ll
index 7ef6adb8fcc4d..9a4ccfc43987b 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/strict_fma.f32.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/strict_fma.f32.ll
@@ -278,8 +278,8 @@ define void @v_constained_fma_v3f32_fpexcept_strict_uni(<3 x float> inreg %x, <3
 ; GFX942-NEXT:    v_pk_fma_f32 v[2:3], s[2:3], v[2:3], v[4:5]
 ; GFX942-NEXT:    s_nop 0
 ; GFX942-NEXT:    v_readfirstlane_b32 s2, v2
-; GFX942-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX942-NEXT:    v_mov_b32_e32 v2, s0
+; GFX942-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX942-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX942-NEXT:    global_store_dwordx3 v[0:1], v[2:4], off
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/strict_fma.f64.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/strict_fma.f64.ll
index 5d864664e012e..ad195cd94e095 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/strict_fma.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/strict_fma.f64.ll
@@ -121,10 +121,10 @@ define void @v_constained_fma_v2f64_fpexcept_strict_uni(<2 x double> inreg %x, <
 ; GFX8-NEXT:    v_fma_f64 v[4:5], s[18:19], v[4:5], v[6:7]
 ; GFX8-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX8-NEXT:    v_readfirstlane_b32 s5, v3
-; GFX8-NEXT:    v_readfirstlane_b32 s6, v4
-; GFX8-NEXT:    v_readfirstlane_b32 s7, v5
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX8-NEXT:    v_readfirstlane_b32 s7, v5
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
@@ -146,10 +146,10 @@ define void @v_constained_fma_v2f64_fpexcept_strict_uni(<2 x double> inreg %x, <
 ; GFX900-NEXT:    v_fma_f64 v[4:5], s[18:19], v[4:5], v[6:7]
 ; GFX900-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX900-NEXT:    v_readfirstlane_b32 s5, v3
-; GFX900-NEXT:    v_readfirstlane_b32 s6, v4
-; GFX900-NEXT:    v_readfirstlane_b32 s7, v5
 ; GFX900-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX900-NEXT:    v_mov_b32_e32 v3, s5
+; GFX900-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX900-NEXT:    v_readfirstlane_b32 s7, v5
 ; GFX900-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX900-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX900-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
@@ -284,24 +284,24 @@ define void @v_constained_fma_v3f64_fpexcept_strict_uni(<3 x double> inreg %x, <
 ; GFX8-NEXT:    v_fma_f64 v[0:1], s[16:17], v[0:1], v[6:7]
 ; GFX8-NEXT:    v_mov_b32_e32 v6, s24
 ; GFX8-NEXT:    v_mov_b32_e32 v7, s25
-; GFX8-NEXT:    v_mov_b32_e32 v9, s5
 ; GFX8-NEXT:    v_mov_b32_e32 v8, s4
+; GFX8-NEXT:    v_mov_b32_e32 v9, s5
 ; GFX8-NEXT:    v_fma_f64 v[6:7], s[18:19], v[6:7], v[8:9]
-; GFX8-NEXT:    v_readfirstlane_b32 s8, v2
-; GFX8-NEXT:    v_readfirstlane_b32 s9, v3
-; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
-; GFX8-NEXT:    v_readfirstlane_b32 s5, v1
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX8-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX8-NEXT:    v_readfirstlane_b32 s6, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s7, v1
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s26
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s27
-; GFX8-NEXT:    v_mov_b32_e32 v2, s8
-; GFX8-NEXT:    v_mov_b32_e32 v3, s9
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX8-NEXT:    v_fma_f64 v[8:9], s[20:21], v[0:1], v[2:3]
-; GFX8-NEXT:    v_readfirstlane_b32 s6, v6
-; GFX8-NEXT:    v_readfirstlane_b32 s7, v7
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    v_mov_b32_e32 v2, s6
-; GFX8-NEXT:    v_mov_b32_e32 v3, s7
+; GFX8-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v7
+; GFX8-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NEXT:    v_mov_b32_e32 v1, s7
+; GFX8-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_add_u32_e32 v0, vcc, 16, v4
@@ -322,15 +322,15 @@ define void @v_constained_fma_v3f64_fpexcept_strict_uni(<3 x double> inreg %x, <
 ; GFX900-NEXT:    v_fma_f64 v[0:1], s[16:17], v[0:1], v[6:7]
 ; GFX900-NEXT:    v_mov_b32_e32 v6, s24
 ; GFX900-NEXT:    v_mov_b32_e32 v7, s25
-; GFX900-NEXT:    v_mov_b32_e32 v9, s5
 ; GFX900-NEXT:    v_mov_b32_e32 v8, s4
+; GFX900-NEXT:    v_mov_b32_e32 v9, s5
 ; GFX900-NEXT:    v_fma_f64 v[6:7], s[18:19], v[6:7], v[8:9]
 ; GFX900-NEXT:    v_readfirstlane_b32 s6, v2
-; GFX900-NEXT:    v_readfirstlane_b32 s7, v3
+; GFX900-NEXT:    v_readfirstlane_b32 s4, v3
 ; GFX900-NEXT:    v_mov_b32_e32 v2, s26
 ; GFX900-NEXT:    v_mov_b32_e32 v3, s27
-; GFX900-NEXT:    v_mov_b32_e32 v9, s7
 ; GFX900-NEXT:    v_mov_b32_e32 v8, s6
+; GFX900-NEXT:    v_mov_b32_e32 v9, s4
 ; GFX900-NEXT:    v_fma_f64 v[8:9], s[20:21], v[2:3], v[8:9]
 ; GFX900-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX900-NEXT:    v_readfirstlane_b32 s5, v1
@@ -510,8 +510,8 @@ define void @v_constained_fma_v4f64_fpexcept_strict_uni(<4 x double> inreg %x, <
 ; GFX8-NEXT:    v_mov_b32_e32 v6, s10
 ; GFX8-NEXT:    v_mov_b32_e32 v7, s11
 ; GFX8-NEXT:    v_fma_f64 v[4:5], s[20:21], v[4:5], v[6:7]
-; GFX8-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX8-NEXT:    v_mov_b32_e32 v6, s4
+; GFX8-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX8-NEXT:    v_mov_b32_e32 v8, s12
 ; GFX8-NEXT:    v_mov_b32_e32 v9, s13
 ; GFX8-NEXT:    v_fma_f64 v[6:7], s[22:23], v[6:7], v[8:9]
@@ -566,8 +566,8 @@ define void @v_constained_fma_v4f64_fpexcept_strict_uni(<4 x double> inreg %x, <
 ; GFX900-NEXT:    v_mov_b32_e32 v6, s10
 ; GFX900-NEXT:    v_mov_b32_e32 v7, s11
 ; GFX900-NEXT:    v_fma_f64 v[4:5], s[20:21], v[4:5], v[6:7]
-; GFX900-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX900-NEXT:    v_mov_b32_e32 v6, s4
+; GFX900-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX900-NEXT:    v_mov_b32_e32 v8, s12
 ; GFX900-NEXT:    v_mov_b32_e32 v9, s13
 ; GFX900-NEXT:    v_fma_f64 v[6:7], s[22:23], v[6:7], v[8:9]
@@ -663,9 +663,9 @@ define void @v_constained_fma_v4f64_fpexcept_strict_uni(<4 x double> inreg %x, <
 ; GFX11-NEXT:    v_readfirstlane_b32 s7, v9
 ; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
-; GFX11-NEXT:    v_dual_mov_b32 v11, s7 :: v_dual_mov_b32 v10, s6
-; GFX11-NEXT:    v_dual_mov_b32 v9, s5 :: v_dual_mov_b32 v8, s4
+; GFX11-NEXT:    v_dual_mov_b32 v8, s4 :: v_dual_mov_b32 v9, s5
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_4)
+; GFX11-NEXT:    v_dual_mov_b32 v10, s6 :: v_dual_mov_b32 v11, s7
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    global_store_b128 v[6:7], v[0:3], off
 ; GFX11-NEXT:    global_store_b128 v[6:7], v[8:11], off offset:16
@@ -706,9 +706,8 @@ define void @v_constained_fma_v4f64_fpexcept_strict_uni(<4 x double> inreg %x, <
 ; GFX12-NEXT:    s_wait_alu depctr_va_sdst(0)
 ; GFX12-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX12-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_3)
-; GFX12-NEXT:    v_dual_mov_b32 v11, s7 :: v_dual_mov_b32 v10, s6
-; GFX12-NEXT:    v_dual_mov_b32 v9, s5 :: v_dual_mov_b32 v8, s4
+; GFX12-NEXT:    v_dual_mov_b32 v8, s4 :: v_dual_mov_b32 v9, s5
+; GFX12-NEXT:    v_dual_mov_b32 v10, s6 :: v_dual_mov_b32 v11, s7
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    global_store_b128 v[6:7], v[0:3], off
 ; GFX12-NEXT:    global_store_b128 v[6:7], v[8:11], off offset:16
@@ -1104,9 +1103,9 @@ define void @v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_uni(<2 x double> i
 ; GFX8-NEXT:    v_fma_f64 v[4:5], -s[18:19], -v[4:5], v[6:7]
 ; GFX8-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX8-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX8-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s7
@@ -1129,9 +1128,9 @@ define void @v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_uni(<2 x double> i
 ; GFX900-NEXT:    v_fma_f64 v[4:5], -s[18:19], -v[4:5], v[6:7]
 ; GFX900-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX900-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX900-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX900-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX900-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX900-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX900-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX900-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX900-NEXT:    v_mov_b32_e32 v5, s7
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll
index a910fb96a8d9c..77d8ffd040984 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll
@@ -428,7 +428,7 @@ define void @s_usubo_usube(i64 inreg %a, i64 inreg %b, ptr addrspace(1) %res, pt
 ; GFX11-NEXT:    s_sub_u32 s0, s0, s2
 ; GFX11-NEXT:    s_subb_u32 s1, s1, s3
 ; GFX11-NEXT:    s_cselect_b32 s2, 1, 0
-; GFX11-NEXT:    v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0
+; GFX11-NEXT:    v_dual_mov_b32 v4, s0 :: v_dual_mov_b32 v5, s1
 ; GFX11-NEXT:    v_mov_b32_e32 v6, s2
 ; GFX11-NEXT:    global_store_b64 v[0:1], v[4:5], off
 ; GFX11-NEXT:    global_store_b32 v[2:3], v6, off
@@ -445,7 +445,7 @@ define void @s_usubo_usube(i64 inreg %a, i64 inreg %b, ptr addrspace(1) %res, pt
 ; GFX12-NEXT:    s_sub_co_ci_u32 s1, s1, s3
 ; GFX12-NEXT:    s_cselect_b32 s2, 1, 0
 ; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0
+; GFX12-NEXT:    v_dual_mov_b32 v4, s0 :: v_dual_mov_b32 v5, s1
 ; GFX12-NEXT:    v_mov_b32_e32 v6, s2
 ; GFX12-NEXT:    global_store_b64 v[0:1], v[4:5], off
 ; GFX12-NEXT:    global_store_b32 v[2:3], v6, off
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/udivrem.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/udivrem.ll
index 2ec45c8f5e7ff..4202200db53d9 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/udivrem.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/udivrem.ll
@@ -673,22 +673,22 @@ define amdgpu_kernel void @udivrem_v2i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX8-NEXT:    s_cselect_b32 s9, s11, s9
 ; GFX8-NEXT:    s_sub_i32 s11, s4, s6
 ; GFX8-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX8-NEXT:    s_cselect_b32 s10, s11, s4
-; GFX8-NEXT:    s_cmp_ge_u32 s10, s6
-; GFX8-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX8-NEXT:    s_add_i32 s4, s9, 1
-; GFX8-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX8-NEXT:    s_cselect_b32 s4, s4, s9
-; GFX8-NEXT:    s_sub_i32 s6, s10, s6
-; GFX8-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX8-NEXT:    s_cselect_b32 s6, s6, s10
-; GFX8-NEXT:    s_sub_i32 s9, 0, s7
-; GFX8-NEXT:    s_mul_i32 s9, s9, s8
-; GFX8-NEXT:    v_mul_hi_u32 v0, v1, s9
+; GFX8-NEXT:    s_cselect_b32 s4, s11, s4
+; GFX8-NEXT:    s_cmp_ge_u32 s4, s6
+; GFX8-NEXT:    s_cselect_b32 s10, 1, 0
+; GFX8-NEXT:    s_add_i32 s11, s9, 1
+; GFX8-NEXT:    s_cmp_lg_u32 s10, 0
+; GFX8-NEXT:    s_cselect_b32 s9, s11, s9
+; GFX8-NEXT:    s_sub_i32 s6, s4, s6
+; GFX8-NEXT:    s_cmp_lg_u32 s10, 0
+; GFX8-NEXT:    s_cselect_b32 s4, s6, s4
+; GFX8-NEXT:    s_sub_i32 s6, 0, s7
+; GFX8-NEXT:    s_mul_i32 s6, s6, s8
+; GFX8-NEXT:    v_mul_hi_u32 v0, v1, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_readfirstlane_b32 s9, v0
-; GFX8-NEXT:    s_add_i32 s8, s8, s9
-; GFX8-NEXT:    v_mov_b32_e32 v0, s8
+; GFX8-NEXT:    v_readfirstlane_b32 s6, v0
+; GFX8-NEXT:    s_add_i32 s6, s8, s6
+; GFX8-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX8-NEXT:    v_mul_hi_u32 v2, s5, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX8-NEXT:    v_readfirstlane_b32 s0, v2
@@ -696,25 +696,25 @@ define amdgpu_kernel void @udivrem_v2i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX8-NEXT:    s_sub_i32 s1, s5, s1
 ; GFX8-NEXT:    s_cmp_ge_u32 s1, s7
 ; GFX8-NEXT:    s_cselect_b32 s5, 1, 0
-; GFX8-NEXT:    s_add_i32 s8, s0, 1
+; GFX8-NEXT:    s_add_i32 s6, s0, 1
 ; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
-; GFX8-NEXT:    s_cselect_b32 s0, s8, s0
-; GFX8-NEXT:    s_sub_i32 s8, s1, s7
+; GFX8-NEXT:    s_cselect_b32 s0, s6, s0
+; GFX8-NEXT:    s_sub_i32 s6, s1, s7
 ; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
-; GFX8-NEXT:    s_cselect_b32 s1, s8, s1
+; GFX8-NEXT:    s_cselect_b32 s1, s6, s1
 ; GFX8-NEXT:    s_cmp_ge_u32 s1, s7
-; GFX8-NEXT:    s_cselect_b32 s8, 1, 0
-; GFX8-NEXT:    s_add_i32 s5, s0, 1
-; GFX8-NEXT:    s_cmp_lg_u32 s8, 0
-; GFX8-NEXT:    s_cselect_b32 s5, s5, s0
-; GFX8-NEXT:    s_sub_i32 s0, s1, s7
-; GFX8-NEXT:    s_cmp_lg_u32 s8, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s4
-; GFX8-NEXT:    v_mov_b32_e32 v3, s5
-; GFX8-NEXT:    s_cselect_b32 s7, s0, s1
+; GFX8-NEXT:    s_cselect_b32 s5, 1, 0
+; GFX8-NEXT:    s_add_i32 s6, s0, 1
+; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
+; GFX8-NEXT:    s_cselect_b32 s0, s6, s0
+; GFX8-NEXT:    s_sub_i32 s6, s1, s7
+; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s9
+; GFX8-NEXT:    v_mov_b32_e32 v3, s0
+; GFX8-NEXT:    s_cselect_b32 s0, s6, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v0, s6
-; GFX8-NEXT:    v_mov_b32_e32 v1, s7
+; GFX8-NEXT:    v_mov_b32_e32 v0, s4
+; GFX8-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
@@ -748,44 +748,44 @@ define amdgpu_kernel void @udivrem_v2i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX9-NEXT:    s_cselect_b32 s8, s11, s8
 ; GFX9-NEXT:    s_sub_i32 s11, s4, s6
 ; GFX9-NEXT:    s_cmp_lg_u32 s9, 0
-; GFX9-NEXT:    s_cselect_b32 s9, s11, s4
-; GFX9-NEXT:    s_cmp_ge_u32 s9, s6
-; GFX9-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX9-NEXT:    s_add_i32 s4, s8, 1
-; GFX9-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX9-NEXT:    s_cselect_b32 s4, s4, s8
-; GFX9-NEXT:    s_sub_i32 s6, s9, s6
-; GFX9-NEXT:    s_cmp_lg_u32 s11, 0
+; GFX9-NEXT:    s_cselect_b32 s4, s11, s4
+; GFX9-NEXT:    s_cmp_ge_u32 s4, s6
+; GFX9-NEXT:    s_cselect_b32 s9, 1, 0
+; GFX9-NEXT:    s_add_i32 s11, s8, 1
+; GFX9-NEXT:    s_cmp_lg_u32 s9, 0
+; GFX9-NEXT:    s_cselect_b32 s8, s11, s8
+; GFX9-NEXT:    s_sub_i32 s6, s4, s6
+; GFX9-NEXT:    s_cmp_lg_u32 s9, 0
 ; GFX9-NEXT:    v_readfirstlane_b32 s10, v1
-; GFX9-NEXT:    s_cselect_b32 s6, s6, s9
-; GFX9-NEXT:    s_sub_i32 s8, 0, s7
-; GFX9-NEXT:    s_mul_i32 s8, s8, s10
-; GFX9-NEXT:    s_mul_hi_u32 s8, s10, s8
-; GFX9-NEXT:    s_add_i32 s10, s10, s8
-; GFX9-NEXT:    s_mul_hi_u32 s8, s5, s10
-; GFX9-NEXT:    s_mul_i32 s9, s8, s7
+; GFX9-NEXT:    s_cselect_b32 s4, s6, s4
+; GFX9-NEXT:    s_sub_i32 s6, 0, s7
+; GFX9-NEXT:    s_mul_i32 s6, s6, s10
+; GFX9-NEXT:    s_mul_hi_u32 s6, s10, s6
+; GFX9-NEXT:    s_add_i32 s10, s10, s6
+; GFX9-NEXT:    s_mul_hi_u32 s6, s5, s10
+; GFX9-NEXT:    s_mul_i32 s9, s6, s7
 ; GFX9-NEXT:    s_sub_i32 s5, s5, s9
 ; GFX9-NEXT:    s_cmp_ge_u32 s5, s7
 ; GFX9-NEXT:    s_cselect_b32 s9, 1, 0
-; GFX9-NEXT:    s_add_i32 s10, s8, 1
+; GFX9-NEXT:    s_add_i32 s10, s6, 1
 ; GFX9-NEXT:    s_cmp_lg_u32 s9, 0
-; GFX9-NEXT:    s_cselect_b32 s8, s10, s8
+; GFX9-NEXT:    s_cselect_b32 s6, s10, s6
 ; GFX9-NEXT:    s_sub_i32 s10, s5, s7
 ; GFX9-NEXT:    s_cmp_lg_u32 s9, 0
-; GFX9-NEXT:    s_cselect_b32 s9, s10, s5
-; GFX9-NEXT:    s_cmp_ge_u32 s9, s7
-; GFX9-NEXT:    s_cselect_b32 s10, 1, 0
-; GFX9-NEXT:    s_add_i32 s5, s8, 1
-; GFX9-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX9-NEXT:    s_cselect_b32 s5, s5, s8
-; GFX9-NEXT:    s_sub_i32 s7, s9, s7
-; GFX9-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX9-NEXT:    s_cselect_b32 s7, s7, s9
+; GFX9-NEXT:    s_cselect_b32 s5, s10, s5
+; GFX9-NEXT:    s_cmp_ge_u32 s5, s7
+; GFX9-NEXT:    s_cselect_b32 s9, 1, 0
+; GFX9-NEXT:    s_add_i32 s10, s6, 1
+; GFX9-NEXT:    s_cmp_lg_u32 s9, 0
+; GFX9-NEXT:    s_cselect_b32 s6, s10, s6
+; GFX9-NEXT:    s_sub_i32 s7, s5, s7
+; GFX9-NEXT:    s_cmp_lg_u32 s9, 0
+; GFX9-NEXT:    s_cselect_b32 s5, s7, s5
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s6
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s5
-; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
-; GFX9-NEXT:    v_mov_b32_e32 v0, s6
-; GFX9-NEXT:    v_mov_b32_e32 v1, s7
 ; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
 ; GFX9-NEXT:    s_endpgm
 ;
@@ -814,46 +814,46 @@ define amdgpu_kernel void @udivrem_v2i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX10-NEXT:    s_cselect_b32 s9, 1, 0
 ; GFX10-NEXT:    s_add_i32 s10, s8, 1
 ; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
-; GFX10-NEXT:    v_readfirstlane_b32 s11, v0
 ; GFX10-NEXT:    s_cselect_b32 s8, s10, s8
 ; GFX10-NEXT:    s_sub_i32 s10, s4, s6
 ; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
-; GFX10-NEXT:    s_cselect_b32 s9, s10, s4
-; GFX10-NEXT:    s_cmp_ge_u32 s9, s6
-; GFX10-NEXT:    s_cselect_b32 s10, 1, 0
-; GFX10-NEXT:    s_add_i32 s4, s8, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX10-NEXT:    s_cselect_b32 s4, s4, s8
-; GFX10-NEXT:    s_sub_i32 s6, s9, s6
-; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX10-NEXT:    s_cselect_b32 s6, s6, s9
-; GFX10-NEXT:    s_sub_i32 s8, 0, s7
-; GFX10-NEXT:    s_mul_i32 s8, s8, s11
-; GFX10-NEXT:    s_mul_hi_u32 s8, s11, s8
-; GFX10-NEXT:    s_add_i32 s11, s11, s8
-; GFX10-NEXT:    s_mul_hi_u32 s8, s5, s11
-; GFX10-NEXT:    s_mul_i32 s9, s8, s7
+; GFX10-NEXT:    s_cselect_b32 s4, s10, s4
+; GFX10-NEXT:    v_readfirstlane_b32 s10, v0
+; GFX10-NEXT:    s_cmp_ge_u32 s4, s6
+; GFX10-NEXT:    s_cselect_b32 s9, 1, 0
+; GFX10-NEXT:    s_add_i32 s11, s8, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
+; GFX10-NEXT:    s_cselect_b32 s8, s11, s8
+; GFX10-NEXT:    s_sub_i32 s6, s4, s6
+; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s8
+; GFX10-NEXT:    s_cselect_b32 s4, s6, s4
+; GFX10-NEXT:    s_sub_i32 s6, 0, s7
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    s_mul_i32 s6, s6, s10
+; GFX10-NEXT:    s_mul_hi_u32 s6, s10, s6
+; GFX10-NEXT:    s_add_i32 s10, s10, s6
+; GFX10-NEXT:    s_mul_hi_u32 s6, s5, s10
+; GFX10-NEXT:    s_mul_i32 s9, s6, s7
 ; GFX10-NEXT:    s_sub_i32 s5, s5, s9
 ; GFX10-NEXT:    s_cmp_ge_u32 s5, s7
 ; GFX10-NEXT:    s_cselect_b32 s9, 1, 0
-; GFX10-NEXT:    s_add_i32 s10, s8, 1
+; GFX10-NEXT:    s_add_i32 s10, s6, 1
 ; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
-; GFX10-NEXT:    s_cselect_b32 s8, s10, s8
+; GFX10-NEXT:    s_cselect_b32 s6, s10, s6
 ; GFX10-NEXT:    s_sub_i32 s10, s5, s7
 ; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
-; GFX10-NEXT:    s_cselect_b32 s9, s10, s5
-; GFX10-NEXT:    s_cmp_ge_u32 s9, s7
-; GFX10-NEXT:    s_cselect_b32 s10, 1, 0
-; GFX10-NEXT:    s_add_i32 s5, s8, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX10-NEXT:    s_cselect_b32 s5, s5, s8
-; GFX10-NEXT:    s_sub_i32 s7, s9, s7
-; GFX10-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s4
-; GFX10-NEXT:    s_cselect_b32 s7, s7, s9
-; GFX10-NEXT:    v_mov_b32_e32 v1, s5
-; GFX10-NEXT:    v_mov_b32_e32 v2, s6
-; GFX10-NEXT:    v_mov_b32_e32 v3, s7
+; GFX10-NEXT:    s_cselect_b32 s5, s10, s5
+; GFX10-NEXT:    s_cmp_ge_u32 s5, s7
+; GFX10-NEXT:    s_cselect_b32 s9, 1, 0
+; GFX10-NEXT:    s_add_i32 s10, s6, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
+; GFX10-NEXT:    s_cselect_b32 s6, s10, s6
+; GFX10-NEXT:    s_sub_i32 s7, s5, s7
+; GFX10-NEXT:    s_cmp_lg_u32 s9, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s6
+; GFX10-NEXT:    s_cselect_b32 s5, s7, s5
+; GFX10-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX10-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1]
 ; GFX10-NEXT:    global_store_dwordx2 v4, v[2:3], s[2:3]
 ; GFX10-NEXT:    s_endpgm
@@ -905,12 +905,12 @@ define amdgpu_kernel void @udivrem_v4i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX8-NEXT:    s_cmp_lg_u32 s12, 0
 ; GFX8-NEXT:    s_cselect_b32 s0, s13, s0
 ; GFX8-NEXT:    s_cmp_ge_u32 s0, s4
-; GFX8-NEXT:    s_cselect_b32 s13, 1, 0
-; GFX8-NEXT:    s_add_i32 s12, s11, 1
-; GFX8-NEXT:    s_cmp_lg_u32 s13, 0
-; GFX8-NEXT:    s_cselect_b32 s12, s12, s11
+; GFX8-NEXT:    s_cselect_b32 s12, 1, 0
+; GFX8-NEXT:    s_add_i32 s13, s11, 1
+; GFX8-NEXT:    s_cmp_lg_u32 s12, 0
+; GFX8-NEXT:    s_cselect_b32 s13, s13, s11
 ; GFX8-NEXT:    s_sub_i32 s4, s0, s4
-; GFX8-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX8-NEXT:    s_cmp_lg_u32 s12, 0
 ; GFX8-NEXT:    s_cselect_b32 s0, s4, s0
 ; GFX8-NEXT:    s_sub_i32 s4, 0, s5
 ; GFX8-NEXT:    s_mul_i32 s4, s4, s10
@@ -927,17 +927,17 @@ define amdgpu_kernel void @udivrem_v4i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX8-NEXT:    s_sub_i32 s1, s1, s11
 ; GFX8-NEXT:    s_cmp_ge_u32 s1, s5
 ; GFX8-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX8-NEXT:    s_add_i32 s13, s10, 1
+; GFX8-NEXT:    s_add_i32 s12, s10, 1
 ; GFX8-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX8-NEXT:    s_cselect_b32 s10, s13, s10
-; GFX8-NEXT:    s_sub_i32 s13, s1, s5
+; GFX8-NEXT:    s_cselect_b32 s10, s12, s10
+; GFX8-NEXT:    s_sub_i32 s12, s1, s5
 ; GFX8-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX8-NEXT:    s_cselect_b32 s1, s13, s1
+; GFX8-NEXT:    s_cselect_b32 s1, s12, s1
 ; GFX8-NEXT:    s_cmp_ge_u32 s1, s5
 ; GFX8-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX8-NEXT:    s_add_i32 s13, s10, 1
+; GFX8-NEXT:    s_add_i32 s12, s10, 1
 ; GFX8-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX8-NEXT:    s_cselect_b32 s13, s13, s10
+; GFX8-NEXT:    s_cselect_b32 s12, s12, s10
 ; GFX8-NEXT:    s_sub_i32 s5, s1, s5
 ; GFX8-NEXT:    s_cmp_lg_u32 s11, 0
 ; GFX8-NEXT:    s_cselect_b32 s1, s5, s1
@@ -966,44 +966,44 @@ define amdgpu_kernel void @udivrem_v4i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX8-NEXT:    s_cselect_b32 s10, 1, 0
 ; GFX8-NEXT:    s_add_i32 s11, s5, 1
 ; GFX8-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX8-NEXT:    s_cselect_b32 s14, s11, s5
-; GFX8-NEXT:    s_sub_i32 s5, s2, s6
+; GFX8-NEXT:    s_cselect_b32 s5, s11, s5
+; GFX8-NEXT:    s_sub_i32 s6, s2, s6
 ; GFX8-NEXT:    s_cmp_lg_u32 s10, 0
-; GFX8-NEXT:    s_cselect_b32 s2, s5, s2
-; GFX8-NEXT:    s_sub_i32 s5, 0, s7
-; GFX8-NEXT:    s_mul_i32 s5, s5, s4
-; GFX8-NEXT:    v_mul_hi_u32 v0, v1, s5
+; GFX8-NEXT:    s_cselect_b32 s2, s6, s2
+; GFX8-NEXT:    s_sub_i32 s6, 0, s7
+; GFX8-NEXT:    s_mul_i32 s6, s6, s4
+; GFX8-NEXT:    v_mul_hi_u32 v0, v1, s6
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x0
-; GFX8-NEXT:    v_readfirstlane_b32 s5, v0
-; GFX8-NEXT:    s_add_i32 s4, s4, s5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s12
+; GFX8-NEXT:    v_mov_b32_e32 v2, s5
+; GFX8-NEXT:    v_readfirstlane_b32 s6, v0
+; GFX8-NEXT:    s_add_i32 s4, s4, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    v_mul_hi_u32 v0, s3, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s8
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s9
 ; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
-; GFX8-NEXT:    s_mul_i32 s5, s4, s7
-; GFX8-NEXT:    s_sub_i32 s3, s3, s5
+; GFX8-NEXT:    s_mul_i32 s6, s4, s7
+; GFX8-NEXT:    s_sub_i32 s3, s3, s6
 ; GFX8-NEXT:    s_cmp_ge_u32 s3, s7
-; GFX8-NEXT:    s_cselect_b32 s5, 1, 0
-; GFX8-NEXT:    s_add_i32 s6, s4, 1
-; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
-; GFX8-NEXT:    s_cselect_b32 s4, s6, s4
-; GFX8-NEXT:    s_sub_i32 s6, s3, s7
-; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
-; GFX8-NEXT:    s_cselect_b32 s3, s6, s3
+; GFX8-NEXT:    s_cselect_b32 s6, 1, 0
+; GFX8-NEXT:    s_add_i32 s14, s4, 1
+; GFX8-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX8-NEXT:    s_cselect_b32 s4, s14, s4
+; GFX8-NEXT:    s_sub_i32 s14, s3, s7
+; GFX8-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX8-NEXT:    s_cselect_b32 s3, s14, s3
 ; GFX8-NEXT:    s_cmp_ge_u32 s3, s7
-; GFX8-NEXT:    s_cselect_b32 s5, 1, 0
-; GFX8-NEXT:    s_add_i32 s6, s4, 1
-; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
-; GFX8-NEXT:    s_cselect_b32 s15, s6, s4
-; GFX8-NEXT:    s_sub_i32 s4, s3, s7
-; GFX8-NEXT:    s_cmp_lg_u32 s5, 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s12
-; GFX8-NEXT:    v_mov_b32_e32 v1, s13
-; GFX8-NEXT:    v_mov_b32_e32 v2, s14
-; GFX8-NEXT:    v_mov_b32_e32 v3, s15
-; GFX8-NEXT:    s_cselect_b32 s3, s4, s3
+; GFX8-NEXT:    s_cselect_b32 s6, 1, 0
+; GFX8-NEXT:    s_add_i32 s14, s4, 1
+; GFX8-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX8-NEXT:    s_cselect_b32 s4, s14, s4
+; GFX8-NEXT:    s_sub_i32 s7, s3, s7
+; GFX8-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s13
+; GFX8-NEXT:    v_mov_b32_e32 v3, s4
+; GFX8-NEXT:    s_cselect_b32 s3, s7, s3
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s10
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s0
@@ -1043,105 +1043,105 @@ define amdgpu_kernel void @udivrem_v4i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX9-NEXT:    s_cselect_b32 s12, s15, s12
 ; GFX9-NEXT:    s_sub_i32 s15, s0, s4
 ; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
-; GFX9-NEXT:    s_cselect_b32 s13, s15, s0
-; GFX9-NEXT:    s_cmp_ge_u32 s13, s4
-; GFX9-NEXT:    s_cselect_b32 s15, 1, 0
-; GFX9-NEXT:    s_add_i32 s0, s12, 1
-; GFX9-NEXT:    s_cmp_lg_u32 s15, 0
-; GFX9-NEXT:    s_cselect_b32 s0, s0, s12
-; GFX9-NEXT:    s_sub_i32 s4, s13, s4
-; GFX9-NEXT:    s_cmp_lg_u32 s15, 0
+; GFX9-NEXT:    s_cselect_b32 s0, s15, s0
+; GFX9-NEXT:    s_cmp_ge_u32 s0, s4
+; GFX9-NEXT:    s_cselect_b32 s13, 1, 0
+; GFX9-NEXT:    s_add_i32 s15, s12, 1
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s12, s15, s12
+; GFX9-NEXT:    s_sub_i32 s4, s0, s4
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
 ; GFX9-NEXT:    v_readfirstlane_b32 s14, v1
-; GFX9-NEXT:    s_cselect_b32 s4, s4, s13
-; GFX9-NEXT:    s_sub_i32 s12, 0, s5
-; GFX9-NEXT:    s_mul_i32 s12, s12, s14
-; GFX9-NEXT:    s_mul_hi_u32 s12, s14, s12
-; GFX9-NEXT:    s_add_i32 s14, s14, s12
-; GFX9-NEXT:    s_mul_hi_u32 s12, s1, s14
-; GFX9-NEXT:    s_mul_i32 s13, s12, s5
+; GFX9-NEXT:    s_cselect_b32 s0, s4, s0
+; GFX9-NEXT:    s_sub_i32 s4, 0, s5
+; GFX9-NEXT:    s_mul_i32 s4, s4, s14
+; GFX9-NEXT:    s_mul_hi_u32 s4, s14, s4
+; GFX9-NEXT:    s_add_i32 s14, s14, s4
+; GFX9-NEXT:    s_mul_hi_u32 s4, s1, s14
+; GFX9-NEXT:    s_mul_i32 s13, s4, s5
 ; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s6
 ; GFX9-NEXT:    s_sub_i32 s1, s1, s13
 ; GFX9-NEXT:    s_cmp_ge_u32 s1, s5
 ; GFX9-NEXT:    s_cselect_b32 s13, 1, 0
-; GFX9-NEXT:    s_add_i32 s14, s12, 1
+; GFX9-NEXT:    s_add_i32 s14, s4, 1
 ; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
-; GFX9-NEXT:    s_cselect_b32 s12, s14, s12
+; GFX9-NEXT:    s_cselect_b32 s4, s14, s4
 ; GFX9-NEXT:    s_sub_i32 s14, s1, s5
 ; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
-; GFX9-NEXT:    s_cselect_b32 s13, s14, s1
-; GFX9-NEXT:    s_cmp_ge_u32 s13, s5
+; GFX9-NEXT:    s_cselect_b32 s1, s14, s1
+; GFX9-NEXT:    s_cmp_ge_u32 s1, s5
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
-; GFX9-NEXT:    s_cselect_b32 s14, 1, 0
-; GFX9-NEXT:    s_add_i32 s1, s12, 1
+; GFX9-NEXT:    s_cselect_b32 s13, 1, 0
+; GFX9-NEXT:    s_add_i32 s14, s4, 1
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX9-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX9-NEXT:    s_cselect_b32 s1, s1, s12
-; GFX9-NEXT:    s_sub_i32 s5, s13, s5
-; GFX9-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX9-NEXT:    s_cselect_b32 s5, s5, s13
-; GFX9-NEXT:    v_readfirstlane_b32 s12, v0
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s4, s14, s4
+; GFX9-NEXT:    s_sub_i32 s5, s1, s5
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s1, s5, s1
+; GFX9-NEXT:    v_readfirstlane_b32 s5, v0
 ; GFX9-NEXT:    s_sub_i32 s13, 0, s6
-; GFX9-NEXT:    s_mul_i32 s13, s13, s12
-; GFX9-NEXT:    s_mul_hi_u32 s13, s12, s13
-; GFX9-NEXT:    s_add_i32 s12, s12, s13
-; GFX9-NEXT:    s_mul_hi_u32 s12, s2, s12
-; GFX9-NEXT:    s_mul_i32 s13, s12, s6
+; GFX9-NEXT:    s_mul_i32 s13, s13, s5
+; GFX9-NEXT:    s_mul_hi_u32 s13, s5, s13
+; GFX9-NEXT:    s_add_i32 s5, s5, s13
+; GFX9-NEXT:    s_mul_hi_u32 s5, s2, s5
+; GFX9-NEXT:    s_mul_i32 s13, s5, s6
 ; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s7
 ; GFX9-NEXT:    s_sub_i32 s2, s2, s13
 ; GFX9-NEXT:    s_cmp_ge_u32 s2, s6
 ; GFX9-NEXT:    s_cselect_b32 s13, 1, 0
-; GFX9-NEXT:    s_add_i32 s14, s12, 1
+; GFX9-NEXT:    s_add_i32 s14, s5, 1
 ; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
-; GFX9-NEXT:    s_cselect_b32 s12, s14, s12
+; GFX9-NEXT:    s_cselect_b32 s5, s14, s5
 ; GFX9-NEXT:    s_sub_i32 s14, s2, s6
 ; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
-; GFX9-NEXT:    s_cselect_b32 s13, s14, s2
-; GFX9-NEXT:    s_cmp_ge_u32 s13, s6
+; GFX9-NEXT:    s_cselect_b32 s2, s14, s2
+; GFX9-NEXT:    s_cmp_ge_u32 s2, s6
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
-; GFX9-NEXT:    s_cselect_b32 s14, 1, 0
-; GFX9-NEXT:    s_add_i32 s2, s12, 1
+; GFX9-NEXT:    s_cselect_b32 s13, 1, 0
+; GFX9-NEXT:    s_add_i32 s14, s5, 1
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX9-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX9-NEXT:    s_cselect_b32 s2, s2, s12
-; GFX9-NEXT:    s_sub_i32 s6, s13, s6
-; GFX9-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX9-NEXT:    s_cselect_b32 s6, s6, s13
-; GFX9-NEXT:    v_readfirstlane_b32 s12, v0
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s5, s14, s5
+; GFX9-NEXT:    s_sub_i32 s6, s2, s6
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s2, s6, s2
+; GFX9-NEXT:    v_readfirstlane_b32 s6, v0
 ; GFX9-NEXT:    s_sub_i32 s13, 0, s7
-; GFX9-NEXT:    s_mul_i32 s13, s13, s12
-; GFX9-NEXT:    s_mul_hi_u32 s13, s12, s13
-; GFX9-NEXT:    s_add_i32 s12, s12, s13
-; GFX9-NEXT:    s_mul_hi_u32 s12, s3, s12
-; GFX9-NEXT:    s_mul_i32 s13, s12, s7
+; GFX9-NEXT:    s_mul_i32 s13, s13, s6
+; GFX9-NEXT:    s_mul_hi_u32 s13, s6, s13
+; GFX9-NEXT:    s_add_i32 s6, s6, s13
+; GFX9-NEXT:    s_mul_hi_u32 s6, s3, s6
+; GFX9-NEXT:    s_mul_i32 s13, s6, s7
 ; GFX9-NEXT:    s_sub_i32 s3, s3, s13
 ; GFX9-NEXT:    s_cmp_ge_u32 s3, s7
 ; GFX9-NEXT:    s_cselect_b32 s13, 1, 0
-; GFX9-NEXT:    s_add_i32 s14, s12, 1
+; GFX9-NEXT:    s_add_i32 s14, s6, 1
 ; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
-; GFX9-NEXT:    s_cselect_b32 s12, s14, s12
+; GFX9-NEXT:    s_cselect_b32 s6, s14, s6
 ; GFX9-NEXT:    s_sub_i32 s14, s3, s7
 ; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
-; GFX9-NEXT:    s_cselect_b32 s13, s14, s3
-; GFX9-NEXT:    s_cmp_ge_u32 s13, s7
-; GFX9-NEXT:    s_cselect_b32 s14, 1, 0
-; GFX9-NEXT:    s_add_i32 s3, s12, 1
-; GFX9-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX9-NEXT:    s_cselect_b32 s3, s3, s12
-; GFX9-NEXT:    s_sub_i32 s7, s13, s7
-; GFX9-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX9-NEXT:    s_cselect_b32 s7, s7, s13
+; GFX9-NEXT:    s_cselect_b32 s3, s14, s3
+; GFX9-NEXT:    s_cmp_ge_u32 s3, s7
+; GFX9-NEXT:    s_cselect_b32 s13, 1, 0
+; GFX9-NEXT:    s_add_i32 s14, s6, 1
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s6, s14, s6
+; GFX9-NEXT:    s_sub_i32 s7, s3, s7
+; GFX9-NEXT:    s_cmp_lg_u32 s13, 0
+; GFX9-NEXT:    s_cselect_b32 s3, s7, s3
+; GFX9-NEXT:    v_mov_b32_e32 v0, s12
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
+; GFX9-NEXT:    v_mov_b32_e32 v2, s5
+; GFX9-NEXT:    v_mov_b32_e32 v3, s6
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[8:9]
+; GFX9-NEXT:    s_nop 0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX9-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s3
-; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[8:9]
-; GFX9-NEXT:    s_nop 0
-; GFX9-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9-NEXT:    v_mov_b32_e32 v1, s5
-; GFX9-NEXT:    v_mov_b32_e32 v2, s6
-; GFX9-NEXT:    v_mov_b32_e32 v3, s7
 ; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[10:11]
 ; GFX9-NEXT:    s_endpgm
 ;
@@ -1172,108 +1172,108 @@ define amdgpu_kernel void @udivrem_v4i32(ptr addrspace(1) %out0, ptr addrspace(1
 ; GFX10-NEXT:    s_cselect_b32 s11, 1, 0
 ; GFX10-NEXT:    s_add_i32 s12, s10, 1
 ; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX10-NEXT:    v_readfirstlane_b32 s13, v0
 ; GFX10-NEXT:    s_cselect_b32 s10, s12, s10
 ; GFX10-NEXT:    s_sub_i32 s12, s0, s4
 ; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
+; GFX10-NEXT:    s_cselect_b32 s0, s12, s0
+; GFX10-NEXT:    v_readfirstlane_b32 s12, v0
+; GFX10-NEXT:    s_cmp_ge_u32 s0, s4
 ; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v1
-; GFX10-NEXT:    s_cselect_b32 s11, s12, s0
+; GFX10-NEXT:    s_cselect_b32 s11, 1, 0
+; GFX10-NEXT:    s_add_i32 s13, s10, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
 ; GFX10-NEXT:    v_cvt_f32_u32_e32 v1, s7
-; GFX10-NEXT:    s_cmp_ge_u32 s11, s4
-; GFX10-NEXT:    s_cselect_b32 s12, 1, 0
-; GFX10-NEXT:    s_add_i32 s0, s10, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s12, 0
+; GFX10-NEXT:    s_cselect_b32 s10, s13, s10
+; GFX10-NEXT:    s_sub_i32 s4, s0, s4
+; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX10-NEXT:    s_cselect_b32 s0, s0, s10
-; GFX10-NEXT:    s_sub_i32 s4, s11, s4
-; GFX10-NEXT:    s_cmp_lg_u32 s12, 0
+; GFX10-NEXT:    s_cselect_b32 s4, s4, s0
+; GFX10-NEXT:    s_sub_i32 s0, 0, s5
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v1, v1
-; GFX10-NEXT:    s_cselect_b32 s4, s4, s11
-; GFX10-NEXT:    s_sub_i32 s10, 0, s5
-; GFX10-NEXT:    s_mul_i32 s10, s10, s13
-; GFX10-NEXT:    s_mul_hi_u32 s10, s13, s10
-; GFX10-NEXT:    s_add_i32 s13, s13, s10
-; GFX10-NEXT:    s_mul_hi_u32 s10, s1, s13
-; GFX10-NEXT:    v_readfirstlane_b32 s13, v0
-; GFX10-NEXT:    s_mul_i32 s11, s10, s5
-; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v1
+; GFX10-NEXT:    s_mul_i32 s0, s0, s12
+; GFX10-NEXT:    v_mov_b32_e32 v4, s4
+; GFX10-NEXT:    s_mul_hi_u32 s0, s12, s0
+; GFX10-NEXT:    s_add_i32 s12, s12, s0
+; GFX10-NEXT:    s_mul_hi_u32 s0, s1, s12
+; GFX10-NEXT:    s_mul_i32 s11, s0, s5
 ; GFX10-NEXT:    s_sub_i32 s1, s1, s11
 ; GFX10-NEXT:    s_cmp_ge_u32 s1, s5
 ; GFX10-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX10-NEXT:    s_add_i32 s12, s10, 1
+; GFX10-NEXT:    s_add_i32 s12, s0, 1
 ; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX10-NEXT:    s_cselect_b32 s10, s12, s10
+; GFX10-NEXT:    s_cselect_b32 s0, s12, s0
 ; GFX10-NEXT:    s_sub_i32 s12, s1, s5
 ; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX10-NEXT:    s_cselect_b32 s11, s12, s1
-; GFX10-NEXT:    s_cmp_ge_u32 s11, s5
-; GFX10-NEXT:    s_cselect_b32 s12, 1, 0
-; GFX10-NEXT:    s_add_i32 s1, s10, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s12, 0
-; GFX10-NEXT:    s_cselect_b32 s1, s1, s10
-; GFX10-NEXT:    s_sub_i32 s5, s11, s5
-; GFX10-NEXT:    s_cmp_lg_u32 s12, 0
-; GFX10-NEXT:    s_cselect_b32 s5, s5, s11
-; GFX10-NEXT:    s_sub_i32 s10, 0, s6
-; GFX10-NEXT:    s_mul_i32 s10, s10, s13
-; GFX10-NEXT:    s_mul_hi_u32 s10, s13, s10
-; GFX10-NEXT:    s_add_i32 s13, s13, s10
-; GFX10-NEXT:    s_mul_hi_u32 s10, s2, s13
-; GFX10-NEXT:    v_readfirstlane_b32 s13, v0
-; GFX10-NEXT:    s_mul_i32 s11, s10, s6
-; GFX10-NEXT:    s_sub_i32 s2, s2, s11
-; GFX10-NEXT:    s_cmp_ge_u32 s2, s6
+; GFX10-NEXT:    s_cselect_b32 s1, s12, s1
+; GFX10-NEXT:    v_readfirstlane_b32 s12, v0
+; GFX10-NEXT:    s_cmp_ge_u32 s1, s5
+; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v1
 ; GFX10-NEXT:    s_cselect_b32 s11, 1, 0
-; GFX10-NEXT:    s_add_i32 s12, s10, 1
+; GFX10-NEXT:    s_add_i32 s13, s0, 1
 ; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX10-NEXT:    s_cselect_b32 s10, s12, s10
-; GFX10-NEXT:    s_sub_i32 s12, s2, s6
+; GFX10-NEXT:    s_cselect_b32 s13, s13, s0
+; GFX10-NEXT:    s_sub_i32 s0, s1, s5
 ; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
-; GFX10-NEXT:    s_cselect_b32 s11, s12, s2
-; GFX10-NEXT:    s_cmp_ge_u32 s11, s6
-; GFX10-NEXT:    s_cselect_b32 s12, 1, 0
-; GFX10-NEXT:    s_add_i32 s2, s10, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s12, 0
-; GFX10-NEXT:    s_cselect_b32 s2, s2, s10
-; GFX10-NEXT:    s_sub_i32 s6, s11, s6
-; GFX10-NEXT:    s_cmp_lg_u32 s12, 0
-; GFX10-NEXT:    s_cselect_b32 s6, s6, s11
-; GFX10-NEXT:    s_sub_i32 s10, 0, s7
-; GFX10-NEXT:    s_mul_i32 s10, s10, s13
-; GFX10-NEXT:    s_mul_hi_u32 s10, s13, s10
-; GFX10-NEXT:    s_add_i32 s13, s13, s10
-; GFX10-NEXT:    s_mul_hi_u32 s12, s3, s13
-; GFX10-NEXT:    s_mul_i32 s10, s12, s7
-; GFX10-NEXT:    s_sub_i32 s3, s3, s10
-; GFX10-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x0
-; GFX10-NEXT:    s_cmp_ge_u32 s3, s7
-; GFX10-NEXT:    s_cselect_b32 s13, 1, 0
-; GFX10-NEXT:    s_add_i32 s14, s12, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s13, 0
-; GFX10-NEXT:    s_cselect_b32 s12, s14, s12
-; GFX10-NEXT:    s_sub_i32 s14, s3, s7
-; GFX10-NEXT:    s_cmp_lg_u32 s13, 0
-; GFX10-NEXT:    s_cselect_b32 s13, s14, s3
-; GFX10-NEXT:    s_cmp_ge_u32 s13, s7
-; GFX10-NEXT:    s_cselect_b32 s14, 1, 0
-; GFX10-NEXT:    s_add_i32 s3, s12, 1
-; GFX10-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX10-NEXT:    s_cselect_b32 s3, s3, s12
-; GFX10-NEXT:    s_sub_i32 s7, s13, s7
-; GFX10-NEXT:    s_cmp_lg_u32 s14, 0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-NEXT:    s_cselect_b32 s7, s7, s13
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
-; GFX10-NEXT:    v_mov_b32_e32 v4, s4
+; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; GFX10-NEXT:    s_cselect_b32 s5, s0, s1
+; GFX10-NEXT:    s_sub_i32 s0, 0, s6
+; GFX10-NEXT:    v_mov_b32_e32 v1, s13
+; GFX10-NEXT:    s_mul_i32 s0, s0, s12
 ; GFX10-NEXT:    v_mov_b32_e32 v5, s5
+; GFX10-NEXT:    s_mul_hi_u32 s0, s12, s0
+; GFX10-NEXT:    s_add_i32 s12, s12, s0
+; GFX10-NEXT:    s_mul_hi_u32 s0, s2, s12
+; GFX10-NEXT:    s_mul_i32 s1, s0, s6
+; GFX10-NEXT:    s_sub_i32 s1, s2, s1
+; GFX10-NEXT:    s_cmp_ge_u32 s1, s6
+; GFX10-NEXT:    s_cselect_b32 s2, 1, 0
+; GFX10-NEXT:    s_add_i32 s11, s0, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s2, 0
+; GFX10-NEXT:    s_cselect_b32 s0, s11, s0
+; GFX10-NEXT:    s_sub_i32 s11, s1, s6
+; GFX10-NEXT:    s_cmp_lg_u32 s2, 0
+; GFX10-NEXT:    s_cselect_b32 s1, s11, s1
+; GFX10-NEXT:    v_readfirstlane_b32 s11, v0
+; GFX10-NEXT:    s_cmp_ge_u32 s1, s6
+; GFX10-NEXT:    v_mov_b32_e32 v0, s10
+; GFX10-NEXT:    s_cselect_b32 s2, 1, 0
+; GFX10-NEXT:    s_add_i32 s12, s0, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s2, 0
+; GFX10-NEXT:    s_cselect_b32 s12, s12, s0
+; GFX10-NEXT:    s_sub_i32 s0, s1, s6
+; GFX10-NEXT:    s_cmp_lg_u32 s2, 0
+; GFX10-NEXT:    v_mov_b32_e32 v2, s12
+; GFX10-NEXT:    s_cselect_b32 s6, s0, s1
+; GFX10-NEXT:    s_sub_i32 s0, 0, s7
 ; GFX10-NEXT:    v_mov_b32_e32 v6, s6
+; GFX10-NEXT:    s_mul_i32 s0, s0, s11
+; GFX10-NEXT:    s_mul_hi_u32 s0, s11, s0
+; GFX10-NEXT:    s_add_i32 s11, s11, s0
+; GFX10-NEXT:    s_mul_hi_u32 s11, s3, s11
+; GFX10-NEXT:    s_mul_i32 s0, s11, s7
+; GFX10-NEXT:    s_sub_i32 s14, s3, s0
+; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX10-NEXT:    s_cmp_ge_u32 s14, s7
+; GFX10-NEXT:    s_cselect_b32 s15, 1, 0
+; GFX10-NEXT:    s_add_i32 s8, s11, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s15, 0
+; GFX10-NEXT:    s_cselect_b32 s8, s8, s11
+; GFX10-NEXT:    s_sub_i32 s9, s14, s7
+; GFX10-NEXT:    s_cmp_lg_u32 s15, 0
+; GFX10-NEXT:    s_cselect_b32 s9, s9, s14
+; GFX10-NEXT:    s_cmp_ge_u32 s9, s7
+; GFX10-NEXT:    s_cselect_b32 s11, 1, 0
+; GFX10-NEXT:    s_add_i32 s14, s8, 1
+; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
+; GFX10-NEXT:    s_cselect_b32 s8, s14, s8
+; GFX10-NEXT:    s_sub_i32 s7, s9, s7
+; GFX10-NEXT:    s_cmp_lg_u32 s11, 0
+; GFX10-NEXT:    v_mov_b32_e32 v3, s8
+; GFX10-NEXT:    s_cselect_b32 s7, s7, s9
 ; GFX10-NEXT:    v_mov_b32_e32 v7, s7
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX10-NEXT:    global_store_dwordx4 v8, v[4:7], s[10:11]
+; GFX10-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX10-NEXT:    global_store_dwordx4 v8, v[4:7], s[2:3]
 ; GFX10-NEXT:    s_endpgm
   %div = udiv <4 x i32> %x, %y
   store <4 x i32> %div, ptr addrspace(1) %out0
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/widen-i8-i16-scalar-loads.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/widen-i8-i16-scalar-loads.ll
index ca73d1a0eb839..d0089b5588bd7 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/widen-i8-i16-scalar-loads.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/widen-i8-i16-scalar-loads.ll
@@ -398,9 +398,9 @@ define amdgpu_kernel void @constant_zextload_i8_align2(ptr addrspace(1) %out, pt
 ; GFX8-NEXT:    s_add_u32 s2, s0, 2
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    s_addc_u32 s0, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
+; GFX8-NEXT:    v_mov_b32_e32 v3, s0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    flat_store_short v[0:1], v4
 ; GFX8-NEXT:    flat_store_short v[2:3], v5
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/wmma-gfx12-w32-imm.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/wmma-gfx12-w32-imm.ll
index 29146314e673d..3b03abdd82e6f 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/wmma-gfx12-w32-imm.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/wmma-gfx12-w32-imm.ll
@@ -19,20 +19,12 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_f16_imm_non_inlineable(<8 x half> %A, <8 x half> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_f16_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s5, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s7, s0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GCN-NEXT:    v_dual_mov_b32 v16, s6 :: v_dual_mov_b32 v17, s7
-; GCN-NEXT:    v_dual_mov_b32 v15, s5 :: v_dual_mov_b32 v14, s4
-; GCN-NEXT:    v_dual_mov_b32 v13, s3 :: v_dual_mov_b32 v12, s2
-; GCN-NEXT:    v_dual_mov_b32 v11, s1 :: v_dual_mov_b32 v10, s0
+; GCN-NEXT:    v_mov_b32_e32 v10, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v11, v10 :: v_dual_mov_b32 v12, v10
+; GCN-NEXT:    v_dual_mov_b32 v13, v10 :: v_dual_mov_b32 v14, v10
+; GCN-NEXT:    v_dual_mov_b32 v15, v10 :: v_dual_mov_b32 v16, v10
+; GCN-NEXT:    v_mov_b32_e32 v17, v10
 ; GCN-NEXT:    v_wmma_f32_16x16x16_f16 v[10:17], v[0:3], v[4:7], v[10:17]
 ; GCN-NEXT:    s_clause 0x1
 ; GCN-NEXT:    global_store_b128 v[8:9], v[10:13], off
@@ -61,20 +53,12 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_imm_non_inlineable(<8 x i16> %A, <8 x i16> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_bf16_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s5, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s7, s0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GCN-NEXT:    v_dual_mov_b32 v16, s6 :: v_dual_mov_b32 v17, s7
-; GCN-NEXT:    v_dual_mov_b32 v15, s5 :: v_dual_mov_b32 v14, s4
-; GCN-NEXT:    v_dual_mov_b32 v13, s3 :: v_dual_mov_b32 v12, s2
-; GCN-NEXT:    v_dual_mov_b32 v11, s1 :: v_dual_mov_b32 v10, s0
+; GCN-NEXT:    v_mov_b32_e32 v10, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v11, v10 :: v_dual_mov_b32 v12, v10
+; GCN-NEXT:    v_dual_mov_b32 v13, v10 :: v_dual_mov_b32 v14, v10
+; GCN-NEXT:    v_dual_mov_b32 v15, v10 :: v_dual_mov_b32 v16, v10
+; GCN-NEXT:    v_mov_b32_e32 v17, v10
 ; GCN-NEXT:    v_wmma_f32_16x16x16_bf16 v[10:17], v[0:3], v[4:7], v[10:17]
 ; GCN-NEXT:    s_clause 0x1
 ; GCN-NEXT:    global_store_b128 v[8:9], v[10:13], off
@@ -101,14 +85,10 @@ bb:
 define amdgpu_ps void @test_wmma_f16_16x16x16_f16_imm_non_inlineable(<8 x half> %A, <8 x half> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f16_16x16x16_f16_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x42004200
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v13, s3
-; GCN-NEXT:    v_dual_mov_b32 v11, s1 :: v_dual_mov_b32 v10, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v10, 0x42004200
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v11, v10 :: v_dual_mov_b32 v12, v10
+; GCN-NEXT:    v_mov_b32_e32 v13, v10
 ; GCN-NEXT:    v_wmma_f16_16x16x16_f16 v[10:13], v[0:3], v[4:7], v[10:13]
 ; GCN-NEXT:    global_store_b128 v[8:9], v[10:13], off
 ; GCN-NEXT:    s_endpgm
@@ -121,14 +101,10 @@ bb:
 define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_imm(<8 x i16> %A, <8 x i16> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_bf16_16x16x16_bf16_imm:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x3f803f80
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v13, s3
-; GCN-NEXT:    v_dual_mov_b32 v11, s1 :: v_dual_mov_b32 v10, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v10, 0x3f803f80
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v11, v10 :: v_dual_mov_b32 v12, v10
+; GCN-NEXT:    v_mov_b32_e32 v13, v10
 ; GCN-NEXT:    v_wmma_bf16_16x16x16_bf16 v[10:13], v[0:3], v[4:7], v[10:13]
 ; GCN-NEXT:    global_store_b128 v[8:9], v[10:13], off
 ; GCN-NEXT:    s_endpgm
@@ -141,14 +117,10 @@ bb:
 define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_imm_non_inlineable(<8 x i16> %A, <8 x i16> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_bf16_16x16x16_bf16_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x3fc03fc0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v13, s3
-; GCN-NEXT:    v_dual_mov_b32 v11, s1 :: v_dual_mov_b32 v10, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v10, 0x3fc03fc0
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v11, v10 :: v_dual_mov_b32 v12, v10
+; GCN-NEXT:    v_mov_b32_e32 v13, v10
 ; GCN-NEXT:    v_wmma_bf16_16x16x16_bf16 v[10:13], v[0:3], v[4:7], v[10:13]
 ; GCN-NEXT:    global_store_b128 v[8:9], v[10:13], off
 ; GCN-NEXT:    s_endpgm
@@ -175,20 +147,12 @@ bb:
 define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_imm_non_inlineable(<2 x i32> %A, <2 x i32> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_i32_16x16x16_iu8_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_movk_i32 s0, 0x80
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s5, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s7, s0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GCN-NEXT:    v_dual_mov_b32 v12, s6 :: v_dual_mov_b32 v13, s7
-; GCN-NEXT:    v_dual_mov_b32 v11, s5 :: v_dual_mov_b32 v10, s4
-; GCN-NEXT:    v_dual_mov_b32 v9, s3 :: v_dual_mov_b32 v8, s2
-; GCN-NEXT:    v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x80
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v7, v6 :: v_dual_mov_b32 v8, v6
+; GCN-NEXT:    v_dual_mov_b32 v9, v6 :: v_dual_mov_b32 v10, v6
+; GCN-NEXT:    v_dual_mov_b32 v11, v6 :: v_dual_mov_b32 v12, v6
+; GCN-NEXT:    v_mov_b32_e32 v13, v6
 ; GCN-NEXT:    v_wmma_i32_16x16x16_iu8 v[6:13], v[0:1], v[2:3], v[6:13]
 ; GCN-NEXT:    s_clause 0x1
 ; GCN-NEXT:    global_store_b128 v[4:5], v[6:9], off
@@ -217,20 +181,12 @@ bb:
 define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_imm_non_inlineable(i32 %A, i32 %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_i32_16x16x16_iu4_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_movk_i32 s0, 0x80
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s5, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s7, s0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GCN-NEXT:    v_dual_mov_b32 v10, s6 :: v_dual_mov_b32 v11, s7
-; GCN-NEXT:    v_dual_mov_b32 v9, s5 :: v_dual_mov_b32 v8, s4
-; GCN-NEXT:    v_dual_mov_b32 v7, s3 :: v_dual_mov_b32 v6, s2
-; GCN-NEXT:    v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0
+; GCN-NEXT:    v_mov_b32_e32 v4, 0x80
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v5, v4 :: v_dual_mov_b32 v6, v4
+; GCN-NEXT:    v_dual_mov_b32 v7, v4 :: v_dual_mov_b32 v8, v4
+; GCN-NEXT:    v_dual_mov_b32 v9, v4 :: v_dual_mov_b32 v10, v4
+; GCN-NEXT:    v_mov_b32_e32 v11, v4
 ; GCN-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:11], v0, v1, v[4:11]
 ; GCN-NEXT:    s_clause 0x1
 ; GCN-NEXT:    global_store_b128 v[2:3], v[4:7], off
@@ -259,20 +215,12 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_imm_non_inlineable(<2 x i32> %A, <2 x i32> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_fp8_fp8_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s5, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s7, s0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GCN-NEXT:    v_dual_mov_b32 v12, s6 :: v_dual_mov_b32 v13, s7
-; GCN-NEXT:    v_dual_mov_b32 v11, s5 :: v_dual_mov_b32 v10, s4
-; GCN-NEXT:    v_dual_mov_b32 v9, s3 :: v_dual_mov_b32 v8, s2
-; GCN-NEXT:    v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v7, v6 :: v_dual_mov_b32 v8, v6
+; GCN-NEXT:    v_dual_mov_b32 v9, v6 :: v_dual_mov_b32 v10, v6
+; GCN-NEXT:    v_dual_mov_b32 v11, v6 :: v_dual_mov_b32 v12, v6
+; GCN-NEXT:    v_mov_b32_e32 v13, v6
 ; GCN-NEXT:    v_wmma_f32_16x16x16_fp8_fp8 v[6:13], v[0:1], v[2:3], v[6:13]
 ; GCN-NEXT:    s_clause 0x1
 ; GCN-NEXT:    global_store_b128 v[4:5], v[6:9], off
@@ -301,20 +249,12 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_imm_non_inlineable(<2 x i32> %A, <2 x i32> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_bf8_fp8_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s5, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s7, s0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GCN-NEXT:    v_dual_mov_b32 v12, s6 :: v_dual_mov_b32 v13, s7
-; GCN-NEXT:    v_dual_mov_b32 v11, s5 :: v_dual_mov_b32 v10, s4
-; GCN-NEXT:    v_dual_mov_b32 v9, s3 :: v_dual_mov_b32 v8, s2
-; GCN-NEXT:    v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v7, v6 :: v_dual_mov_b32 v8, v6
+; GCN-NEXT:    v_dual_mov_b32 v9, v6 :: v_dual_mov_b32 v10, v6
+; GCN-NEXT:    v_dual_mov_b32 v11, v6 :: v_dual_mov_b32 v12, v6
+; GCN-NEXT:    v_mov_b32_e32 v13, v6
 ; GCN-NEXT:    v_wmma_f32_16x16x16_bf8_fp8 v[6:13], v[0:1], v[2:3], v[6:13]
 ; GCN-NEXT:    s_clause 0x1
 ; GCN-NEXT:    global_store_b128 v[4:5], v[6:9], off
@@ -343,20 +283,12 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_imm_non_inlineable(<2 x i32> %A, <2 x i32> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_fp8_bf8_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s5, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s7, s0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GCN-NEXT:    v_dual_mov_b32 v12, s6 :: v_dual_mov_b32 v13, s7
-; GCN-NEXT:    v_dual_mov_b32 v11, s5 :: v_dual_mov_b32 v10, s4
-; GCN-NEXT:    v_dual_mov_b32 v9, s3 :: v_dual_mov_b32 v8, s2
-; GCN-NEXT:    v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v7, v6 :: v_dual_mov_b32 v8, v6
+; GCN-NEXT:    v_dual_mov_b32 v9, v6 :: v_dual_mov_b32 v10, v6
+; GCN-NEXT:    v_dual_mov_b32 v11, v6 :: v_dual_mov_b32 v12, v6
+; GCN-NEXT:    v_mov_b32_e32 v13, v6
 ; GCN-NEXT:    v_wmma_f32_16x16x16_fp8_bf8 v[6:13], v[0:1], v[2:3], v[6:13]
 ; GCN-NEXT:    s_clause 0x1
 ; GCN-NEXT:    global_store_b128 v[4:5], v[6:9], off
@@ -385,20 +317,12 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_imm_non_inlineable(<2 x i32> %A, <2 x i32> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_bf8_bf8_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s5, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s7, s0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GCN-NEXT:    v_dual_mov_b32 v12, s6 :: v_dual_mov_b32 v13, s7
-; GCN-NEXT:    v_dual_mov_b32 v11, s5 :: v_dual_mov_b32 v10, s4
-; GCN-NEXT:    v_dual_mov_b32 v9, s3 :: v_dual_mov_b32 v8, s2
-; GCN-NEXT:    v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v7, v6 :: v_dual_mov_b32 v8, v6
+; GCN-NEXT:    v_dual_mov_b32 v9, v6 :: v_dual_mov_b32 v10, v6
+; GCN-NEXT:    v_dual_mov_b32 v11, v6 :: v_dual_mov_b32 v12, v6
+; GCN-NEXT:    v_mov_b32_e32 v13, v6
 ; GCN-NEXT:    v_wmma_f32_16x16x16_bf8_bf8 v[6:13], v[0:1], v[2:3], v[6:13]
 ; GCN-NEXT:    s_clause 0x1
 ; GCN-NEXT:    global_store_b128 v[4:5], v[6:9], off
@@ -427,20 +351,12 @@ bb:
 define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_imm_non_inlineable(<2 x i32> %A, <2 x i32> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_i32_16x16x32_iu4_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_movk_i32 s0, 0x80
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s5, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s7, s0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GCN-NEXT:    v_dual_mov_b32 v12, s6 :: v_dual_mov_b32 v13, s7
-; GCN-NEXT:    v_dual_mov_b32 v11, s5 :: v_dual_mov_b32 v10, s4
-; GCN-NEXT:    v_dual_mov_b32 v9, s3 :: v_dual_mov_b32 v8, s2
-; GCN-NEXT:    v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x80
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_dual_mov_b32 v7, v6 :: v_dual_mov_b32 v8, v6
+; GCN-NEXT:    v_dual_mov_b32 v9, v6 :: v_dual_mov_b32 v10, v6
+; GCN-NEXT:    v_dual_mov_b32 v11, v6 :: v_dual_mov_b32 v12, v6
+; GCN-NEXT:    v_mov_b32_e32 v13, v6
 ; GCN-NEXT:    v_wmma_i32_16x16x32_iu4 v[6:13], v[0:1], v[2:3], v[6:13]
 ; GCN-NEXT:    s_clause 0x1
 ; GCN-NEXT:    global_store_b128 v[4:5], v[6:9], off
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/wmma-gfx12-w64-imm.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/wmma-gfx12-w64-imm.ll
index 6deaf5e4c650b..b63233b6fb2aa 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/wmma-gfx12-w64-imm.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/wmma-gfx12-w64-imm.ll
@@ -17,16 +17,11 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_f16_imm_non_inlineable(<4 x half> %A, <4 x half> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_f16_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_mov_b32_e32 v8, s2
-; GCN-NEXT:    v_mov_b32_e32 v9, s3
-; GCN-NEXT:    v_mov_b32_e32 v7, s1
-; GCN-NEXT:    v_mov_b32_e32 v6, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v7, v6
+; GCN-NEXT:    v_mov_b32_e32 v8, v6
+; GCN-NEXT:    v_mov_b32_e32 v9, v6
 ; GCN-NEXT:    v_wmma_f32_16x16x16_f16 v[6:9], v[0:1], v[2:3], v[6:9]
 ; GCN-NEXT:    global_store_b128 v[4:5], v[6:9], off
 ; GCN-NEXT:    s_endpgm
@@ -51,16 +46,11 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_imm_non_inlineable(<4 x i16> %A, <4 x i16> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_bf16_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_mov_b32_e32 v8, s2
-; GCN-NEXT:    v_mov_b32_e32 v9, s3
-; GCN-NEXT:    v_mov_b32_e32 v7, s1
-; GCN-NEXT:    v_mov_b32_e32 v6, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v7, v6
+; GCN-NEXT:    v_mov_b32_e32 v8, v6
+; GCN-NEXT:    v_mov_b32_e32 v9, v6
 ; GCN-NEXT:    v_wmma_f32_16x16x16_bf16 v[6:9], v[0:1], v[2:3], v[6:9]
 ; GCN-NEXT:    global_store_b128 v[4:5], v[6:9], off
 ; GCN-NEXT:    s_endpgm
@@ -85,11 +75,9 @@ bb:
 define amdgpu_ps void @test_wmma_f16_16x16x16_f16_imm_non_inlineable(<4 x half> %A, <4 x half> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f16_16x16x16_f16_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x42004200
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    v_mov_b32_e32 v6, s0
-; GCN-NEXT:    v_mov_b32_e32 v7, s1
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x42004200
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v7, v6
 ; GCN-NEXT:    v_wmma_f16_16x16x16_f16 v[6:7], v[0:1], v[2:3], v[6:7]
 ; GCN-NEXT:    global_store_b64 v[4:5], v[6:7], off
 ; GCN-NEXT:    s_endpgm
@@ -102,11 +90,9 @@ bb:
 define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_imm(<4 x i16> %A, <4 x i16> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_bf16_16x16x16_bf16_imm:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x3f803f80
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    v_mov_b32_e32 v6, s0
-; GCN-NEXT:    v_mov_b32_e32 v7, s1
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x3f803f80
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v7, v6
 ; GCN-NEXT:    v_wmma_bf16_16x16x16_bf16 v[6:7], v[0:1], v[2:3], v[6:7]
 ; GCN-NEXT:    global_store_b64 v[4:5], v[6:7], off
 ; GCN-NEXT:    s_endpgm
@@ -119,11 +105,9 @@ bb:
 define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_imm_non_inlineable(<4 x i16> %A, <4 x i16> %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_bf16_16x16x16_bf16_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x3fc03fc0
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    v_mov_b32_e32 v6, s0
-; GCN-NEXT:    v_mov_b32_e32 v7, s1
+; GCN-NEXT:    v_mov_b32_e32 v6, 0x3fc03fc0
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v7, v6
 ; GCN-NEXT:    v_wmma_bf16_16x16x16_bf16 v[6:7], v[0:1], v[2:3], v[6:7]
 ; GCN-NEXT:    global_store_b64 v[4:5], v[6:7], off
 ; GCN-NEXT:    s_endpgm
@@ -148,16 +132,11 @@ bb:
 define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_imm_non_inlineable(i32 %A, i32 %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_i32_16x16x16_iu8_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_movk_i32 s0, 0x80
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_mov_b32_e32 v6, s2
-; GCN-NEXT:    v_mov_b32_e32 v7, s3
-; GCN-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v4, 0x80
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v5, v4
+; GCN-NEXT:    v_mov_b32_e32 v6, v4
+; GCN-NEXT:    v_mov_b32_e32 v7, v4
 ; GCN-NEXT:    v_wmma_i32_16x16x16_iu8 v[4:7], v0, v1, v[4:7]
 ; GCN-NEXT:    global_store_b128 v[2:3], v[4:7], off
 ; GCN-NEXT:    s_endpgm
@@ -182,16 +161,11 @@ bb:
 define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_imm_non_inlineable(i32 %A, i32 %B, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_i32_16x16x16_iu4_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_movk_i32 s0, 0x80
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_mov_b32_e32 v6, s2
-; GCN-NEXT:    v_mov_b32_e32 v7, s3
-; GCN-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v4, 0x80
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v5, v4
+; GCN-NEXT:    v_mov_b32_e32 v6, v4
+; GCN-NEXT:    v_mov_b32_e32 v7, v4
 ; GCN-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v0, v1, v[4:7]
 ; GCN-NEXT:    global_store_b128 v[2:3], v[4:7], off
 ; GCN-NEXT:    s_endpgm
@@ -216,16 +190,11 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_imm_non_inlineable(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_fp8_fp8_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_mov_b32_e32 v6, s2
-; GCN-NEXT:    v_mov_b32_e32 v7, s3
-; GCN-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v4, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v5, v4
+; GCN-NEXT:    v_mov_b32_e32 v6, v4
+; GCN-NEXT:    v_mov_b32_e32 v7, v4
 ; GCN-NEXT:    v_wmma_f32_16x16x16_fp8_fp8 v[4:7], v0, v1, v[4:7]
 ; GCN-NEXT:    global_store_b128 v[2:3], v[4:7], off
 ; GCN-NEXT:    s_endpgm
@@ -250,16 +219,11 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_imm_non_inlineable(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_bf8_fp8_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_mov_b32_e32 v6, s2
-; GCN-NEXT:    v_mov_b32_e32 v7, s3
-; GCN-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v4, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v5, v4
+; GCN-NEXT:    v_mov_b32_e32 v6, v4
+; GCN-NEXT:    v_mov_b32_e32 v7, v4
 ; GCN-NEXT:    v_wmma_f32_16x16x16_bf8_fp8 v[4:7], v0, v1, v[4:7]
 ; GCN-NEXT:    global_store_b128 v[2:3], v[4:7], off
 ; GCN-NEXT:    s_endpgm
@@ -284,16 +248,11 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_imm_non_inlineable(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_fp8_bf8_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_mov_b32_e32 v6, s2
-; GCN-NEXT:    v_mov_b32_e32 v7, s3
-; GCN-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v4, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v5, v4
+; GCN-NEXT:    v_mov_b32_e32 v6, v4
+; GCN-NEXT:    v_mov_b32_e32 v7, v4
 ; GCN-NEXT:    v_wmma_f32_16x16x16_fp8_bf8 v[4:7], v0, v1, v[4:7]
 ; GCN-NEXT:    global_store_b128 v[2:3], v[4:7], off
 ; GCN-NEXT:    s_endpgm
@@ -318,16 +277,11 @@ bb:
 define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_imm_non_inlineable(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_f32_16x16x16_bf8_bf8_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_mov_b32 s0, 0x40400000
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_mov_b32_e32 v6, s2
-; GCN-NEXT:    v_mov_b32_e32 v7, s3
-; GCN-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v4, 0x40400000
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v5, v4
+; GCN-NEXT:    v_mov_b32_e32 v6, v4
+; GCN-NEXT:    v_mov_b32_e32 v7, v4
 ; GCN-NEXT:    v_wmma_f32_16x16x16_bf8_bf8 v[4:7], v0, v1, v[4:7]
 ; GCN-NEXT:    global_store_b128 v[2:3], v[4:7], off
 ; GCN-NEXT:    s_endpgm
@@ -352,16 +306,11 @@ bb:
 define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_imm_non_inlineable(i32 %A, i32 %B, <4 x i32> %C, ptr addrspace(1) %out) {
 ; GCN-LABEL: test_wmma_i32_16x16x32_iu4_imm_non_inlineable:
 ; GCN:       ; %bb.0: ; %bb
-; GCN-NEXT:    s_movk_i32 s0, 0x80
-; GCN-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GCN-NEXT:    s_mov_b32 s1, s0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s3, s0
-; GCN-NEXT:    v_mov_b32_e32 v6, s2
-; GCN-NEXT:    v_mov_b32_e32 v7, s3
-; GCN-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v4, 0x80
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GCN-NEXT:    v_mov_b32_e32 v5, v4
+; GCN-NEXT:    v_mov_b32_e32 v6, v4
+; GCN-NEXT:    v_mov_b32_e32 v7, v4
 ; GCN-NEXT:    v_wmma_i32_16x16x32_iu4 v[4:7], v0, v1, v[4:7]
 ; GCN-NEXT:    global_store_b128 v[2:3], v[4:7], off
 ; GCN-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/add.ll b/llvm/test/CodeGen/AMDGPU/add.ll
index 3a8d9a0a30e1e..403f166809845 100644
--- a/llvm/test/CodeGen/AMDGPU/add.ll
+++ b/llvm/test/CodeGen/AMDGPU/add.ll
@@ -332,28 +332,28 @@ define amdgpu_kernel void @s_add_v8i32(ptr addrspace(1) %out, <8 x i32> %a, <8 x
 ; GFX8-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x44
 ; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_add_i32 s4, s11, s19
-; GFX8-NEXT:    s_add_i32 s5, s10, s18
-; GFX8-NEXT:    s_add_i32 s6, s9, s17
-; GFX8-NEXT:    s_add_i32 s7, s8, s16
-; GFX8-NEXT:    s_add_i32 s2, s15, s23
-; GFX8-NEXT:    s_add_i32 s3, s14, s22
+; GFX8-NEXT:    s_add_i32 s2, s11, s19
+; GFX8-NEXT:    s_add_i32 s3, s10, s18
+; GFX8-NEXT:    s_add_i32 s4, s9, s17
+; GFX8-NEXT:    s_add_i32 s5, s8, s16
+; GFX8-NEXT:    s_add_i32 s6, s15, s23
+; GFX8-NEXT:    s_add_i32 s7, s14, s22
 ; GFX8-NEXT:    s_add_i32 s8, s13, s21
 ; GFX8-NEXT:    s_add_i32 s9, s12, s20
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NEXT:    v_mov_b32_e32 v2, s3
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s6
+; GFX8-NEXT:    s_add_u32 s6, s0, 16
+; GFX8-NEXT:    v_mov_b32_e32 v2, s7
+; GFX8-NEXT:    s_addc_u32 s7, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s9
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s8
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s1
-; GFX8-NEXT:    v_mov_b32_e32 v0, s7
-; GFX8-NEXT:    v_mov_b32_e32 v1, s6
-; GFX8-NEXT:    v_mov_b32_e32 v2, s5
-; GFX8-NEXT:    v_mov_b32_e32 v3, s4
+; GFX8-NEXT:    v_mov_b32_e32 v0, s5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
+; GFX8-NEXT:    v_mov_b32_e32 v2, s3
+; GFX8-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_endpgm
@@ -521,54 +521,56 @@ define amdgpu_kernel void @s_add_v16i32(ptr addrspace(1) %out, <16 x i32> %a, <1
 ; GFX8-NEXT:    s_load_dwordx16 s[36:51], s[4:5], 0xa4
 ; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_add_i32 s4, s11, s39
-; GFX8-NEXT:    s_add_i32 s5, s10, s38
-; GFX8-NEXT:    s_add_i32 s6, s9, s37
-; GFX8-NEXT:    s_add_i32 s7, s8, s36
-; GFX8-NEXT:    s_add_i32 s8, s15, s43
-; GFX8-NEXT:    s_add_i32 s9, s14, s42
-; GFX8-NEXT:    s_add_i32 s10, s13, s41
-; GFX8-NEXT:    s_add_i32 s11, s12, s40
-; GFX8-NEXT:    s_add_i32 s12, s19, s47
-; GFX8-NEXT:    s_add_i32 s13, s18, s46
-; GFX8-NEXT:    s_add_i32 s14, s17, s45
-; GFX8-NEXT:    s_add_i32 s15, s16, s44
-; GFX8-NEXT:    s_add_i32 s2, s23, s51
-; GFX8-NEXT:    s_add_i32 s3, s22, s50
+; GFX8-NEXT:    s_add_i32 s2, s11, s39
+; GFX8-NEXT:    s_add_i32 s3, s10, s38
+; GFX8-NEXT:    s_add_i32 s4, s9, s37
+; GFX8-NEXT:    s_add_i32 s5, s8, s36
+; GFX8-NEXT:    s_add_i32 s6, s15, s43
+; GFX8-NEXT:    s_add_i32 s7, s14, s42
+; GFX8-NEXT:    s_add_i32 s8, s13, s41
+; GFX8-NEXT:    s_add_i32 s9, s12, s40
+; GFX8-NEXT:    s_add_i32 s10, s19, s47
+; GFX8-NEXT:    s_add_i32 s11, s18, s46
+; GFX8-NEXT:    s_add_i32 s12, s17, s45
+; GFX8-NEXT:    s_add_i32 s13, s16, s44
+; GFX8-NEXT:    s_add_i32 s14, s23, s51
+; GFX8-NEXT:    s_add_i32 s15, s22, s50
 ; GFX8-NEXT:    s_add_i32 s16, s21, s49
 ; GFX8-NEXT:    s_add_i32 s17, s20, s48
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 48
-; GFX8-NEXT:    v_mov_b32_e32 v2, s3
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 32
+; GFX8-NEXT:    v_mov_b32_e32 v3, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 48
+; GFX8-NEXT:    v_mov_b32_e32 v2, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s17
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s16
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NEXT:    v_mov_b32_e32 v0, s15
-; GFX8-NEXT:    v_mov_b32_e32 v1, s14
-; GFX8-NEXT:    v_mov_b32_e32 v2, s13
-; GFX8-NEXT:    v_mov_b32_e32 v3, s12
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 32
+; GFX8-NEXT:    v_mov_b32_e32 v2, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s13
+; GFX8-NEXT:    v_mov_b32_e32 v1, s12
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v0, s11
-; GFX8-NEXT:    v_mov_b32_e32 v1, s10
-; GFX8-NEXT:    v_mov_b32_e32 v2, s9
-; GFX8-NEXT:    v_mov_b32_e32 v3, s8
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s6
+; GFX8-NEXT:    s_add_u32 s6, s0, 16
+; GFX8-NEXT:    v_mov_b32_e32 v2, s7
+; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s9
+; GFX8-NEXT:    v_mov_b32_e32 v1, s8
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s1
-; GFX8-NEXT:    v_mov_b32_e32 v0, s7
-; GFX8-NEXT:    v_mov_b32_e32 v1, s6
-; GFX8-NEXT:    v_mov_b32_e32 v2, s5
-; GFX8-NEXT:    v_mov_b32_e32 v3, s4
+; GFX8-NEXT:    v_mov_b32_e32 v0, s5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
+; GFX8-NEXT:    v_mov_b32_e32 v2, s3
+; GFX8-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_endpgm
@@ -981,8 +983,8 @@ define amdgpu_kernel void @add64(ptr addrspace(1) %out, i64 %a, i64 %b) {
 ; GFX8-NEXT:    s_add_u32 s0, s2, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    s_addc_u32 s1, s3, s5
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX8-NEXT:    s_endpgm
 ;
@@ -1079,8 +1081,8 @@ define amdgpu_kernel void @add64_sgpr_vgpr(ptr addrspace(1) %out, i64 %a, ptr ad
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_add_u32 s0, s2, s4
 ; GFX8-NEXT:    s_addc_u32 s1, s3, s5
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX8-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/always-uniform.ll b/llvm/test/CodeGen/AMDGPU/always-uniform.ll
index fd67d08d05292..98f7c4fdb44da 100644
--- a/llvm/test/CodeGen/AMDGPU/always-uniform.ll
+++ b/llvm/test/CodeGen/AMDGPU/always-uniform.ll
@@ -13,14 +13,14 @@ define amdgpu_kernel void @readfirstlane_uniform(ptr addrspace(1) noalias nocapt
 ; GCN-NEXT:    s_lshl_b32 s4, s4, 2
 ; GCN-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NEXT:    s_load_dword s4, s[0:1], s4
-; GCN-NEXT:    s_add_u32 s0, s2, 40
-; GCN-NEXT:    s_addc_u32 s1, s3, 0
+; GCN-NEXT:    s_load_dword s0, s[0:1], s4
+; GCN-NEXT:    s_add_u32 s1, s2, 40
+; GCN-NEXT:    s_addc_u32 s2, s3, 0
 ; GCN-NEXT:    s_mov_b32 flat_scratch_lo, s13
-; GCN-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-NEXT:    v_mov_b32_e32 v1, s1
+; GCN-NEXT:    v_mov_b32_e32 v0, s1
+; GCN-NEXT:    v_mov_b32_e32 v1, s2
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NEXT:    v_mov_b32_e32 v2, s0
 ; GCN-NEXT:    flat_store_dword v[0:1], v2
 ; GCN-NEXT:    s_endpgm
   %tid = tail call i32 @llvm.amdgcn.workitem.id.x()
diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-cs-chain-cc.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-cs-chain-cc.ll
index 283c03d8df5bc..7efa40c2e6a89 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgpu-cs-chain-cc.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgpu-cs-chain-cc.ll
@@ -936,17 +936,13 @@ define amdgpu_cs_chain void @amdgpu_cs_chain_dont_realign_stack(i32 %idx) #0 {
 ; GISEL-GFX11-LABEL: amdgpu_cs_chain_dont_realign_stack:
 ; GISEL-GFX11:       ; %bb.0:
 ; GISEL-GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GISEL-GFX11-NEXT:    s_mov_b32 s3, 4
-; GISEL-GFX11-NEXT:    s_mov_b32 s2, 3
-; GISEL-GFX11-NEXT:    s_mov_b32 s1, 2
-; GISEL-GFX11-NEXT:    s_mov_b32 s0, 1
-; GISEL-GFX11-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_lshlrev_b32 v0, 4, v8
+; GISEL-GFX11-NEXT:    v_dual_mov_b32 v3, 4 :: v_dual_lshlrev_b32 v0, 4, v8
 ; GISEL-GFX11-NEXT:    s_add_i32 s33, s32, 31
-; GISEL-GFX11-NEXT:    v_mov_b32_e32 v2, s2
+; GISEL-GFX11-NEXT:    v_mov_b32_e32 v2, 3
 ; GISEL-GFX11-NEXT:    s_and_not1_b32 s33, s33, 31
 ; GISEL-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GISEL-GFX11-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_add_nc_u32 v4, s33, v0
-; GISEL-GFX11-NEXT:    v_mov_b32_e32 v0, s0
+; GISEL-GFX11-NEXT:    v_dual_mov_b32 v1, 2 :: v_dual_add_nc_u32 v4, s33, v0
+; GISEL-GFX11-NEXT:    v_mov_b32_e32 v0, 1
 ; GISEL-GFX11-NEXT:    s_mov_b32 s34, s32
 ; GISEL-GFX11-NEXT:    s_addk_i32 s32, 0xc0
 ; GISEL-GFX11-NEXT:    scratch_store_b128 v4, v[0:3], off dlc
@@ -1026,13 +1022,9 @@ define amdgpu_cs_chain void @amdgpu_cs_chain_realign_stack_chain_call(i32 %idx,
 ; GISEL-GFX11-NEXT:    s_add_i32 s33, s32, 31
 ; GISEL-GFX11-NEXT:    s_mov_b32 s34, s32
 ; GISEL-GFX11-NEXT:    s_and_not1_b32 s33, s33, 31
-; GISEL-GFX11-NEXT:    s_mov_b32 s7, 4
-; GISEL-GFX11-NEXT:    s_mov_b32 s6, 3
-; GISEL-GFX11-NEXT:    s_mov_b32 s5, 2
-; GISEL-GFX11-NEXT:    s_mov_b32 s4, 1
-; GISEL-GFX11-NEXT:    v_dual_mov_b32 v1, s5 :: v_dual_lshlrev_b32 v4, 4, v8
-; GISEL-GFX11-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v3, s7
-; GISEL-GFX11-NEXT:    v_mov_b32_e32 v2, s6
+; GISEL-GFX11-NEXT:    v_dual_mov_b32 v3, 4 :: v_dual_lshlrev_b32 v4, 4, v8
+; GISEL-GFX11-NEXT:    v_dual_mov_b32 v2, 3 :: v_dual_mov_b32 v1, 2
+; GISEL-GFX11-NEXT:    v_mov_b32_e32 v0, 1
 ; GISEL-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
 ; GISEL-GFX11-NEXT:    v_add_nc_u32_e32 v4, s33, v4
 ; GISEL-GFX11-NEXT:    s_addk_i32 s32, 0xc0
diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-cs-chain-preserve-cc.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-cs-chain-preserve-cc.ll
index 181ffe3ce6aed..45d78d55d8ec8 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgpu-cs-chain-preserve-cc.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgpu-cs-chain-preserve-cc.ll
@@ -594,29 +594,24 @@ define amdgpu_cs_chain_preserve void @amdgpu_cs_chain_preserve_realign_stack(i32
 ; GISEL-GFX11-NEXT:    s_and_not1_b32 s33, s33, 31
 ; GISEL-GFX11-NEXT:    s_addk_i32 s32, 0xe0
 ; GISEL-GFX11-NEXT:    scratch_store_b32 off, v16, s33 ; 4-byte Folded Spill
-; GISEL-GFX11-NEXT:    s_mov_b32 s7, 4
-; GISEL-GFX11-NEXT:    s_mov_b32 s6, 3
-; GISEL-GFX11-NEXT:    s_mov_b32 s5, 2
-; GISEL-GFX11-NEXT:    s_mov_b32 s4, 1
-; GISEL-GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GISEL-GFX11-NEXT:    v_dual_mov_b32 v1, s4 :: v_dual_lshlrev_b32 v0, 4, v8
-; GISEL-GFX11-NEXT:    v_mov_b32_e32 v4, s7
-; GISEL-GFX11-NEXT:    v_dual_mov_b32 v2, s5 :: v_dual_mov_b32 v3, s6
+; GISEL-GFX11-NEXT:    v_dual_mov_b32 v3, 3 :: v_dual_lshlrev_b32 v0, 4, v8
+; GISEL-GFX11-NEXT:    v_mov_b32_e32 v2, 2
+; GISEL-GFX11-NEXT:    v_dual_mov_b32 v4, 4 :: v_dual_mov_b32 v1, 1
+; GISEL-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
 ; GISEL-GFX11-NEXT:    v_add_nc_u32_e32 v5, s33, v0
 ; GISEL-GFX11-NEXT:    s_mov_b32 s3, s0
 ; GISEL-GFX11-NEXT:    ;;#ASMSTART
 ; GISEL-GFX11-NEXT:    s_nop
 ; GISEL-GFX11-NEXT:    ;;#ASMEND
-; GISEL-GFX11-NEXT:    v_dual_mov_b32 v8, v9 :: v_dual_mov_b32 v9, v10
-; GISEL-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GISEL-GFX11-NEXT:    v_dual_mov_b32 v10, v11 :: v_dual_add_nc_u32 v5, 32, v5
 ; GISEL-GFX11-NEXT:    s_mov_b32 s4, chain_preserve_callee at abs32@lo
 ; GISEL-GFX11-NEXT:    s_mov_b32 s5, chain_preserve_callee at abs32@hi
+; GISEL-GFX11-NEXT:    v_dual_mov_b32 v8, v9 :: v_dual_add_nc_u32 v5, 32, v5
+; GISEL-GFX11-NEXT:    v_dual_mov_b32 v9, v10 :: v_dual_mov_b32 v10, v11
 ; GISEL-GFX11-NEXT:    s_mov_b32 s0, s3
-; GISEL-GFX11-NEXT:    s_mov_b32 s32, s34
 ; GISEL-GFX11-NEXT:    scratch_store_b128 v5, v[1:4], off dlc
 ; GISEL-GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GISEL-GFX11-NEXT:    scratch_load_b32 v16, off, s33 ; 4-byte Folded Reload
+; GISEL-GFX11-NEXT:    s_mov_b32 s32, s34
 ; GISEL-GFX11-NEXT:    s_mov_b32 exec_lo, -1
 ; GISEL-GFX11-NEXT:    s_setpc_b64 s[4:5]
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/and.ll b/llvm/test/CodeGen/AMDGPU/and.ll
index 198aaf7a5cd7b..4367ab6ef74f6 100644
--- a/llvm/test/CodeGen/AMDGPU/and.ll
+++ b/llvm/test/CodeGen/AMDGPU/and.ll
@@ -735,13 +735,13 @@ define amdgpu_kernel void @s_and_multi_use_inline_imm_i64(ptr addrspace(1) %out,
 ; GFX6-NEXT:    s_lshl_b32 s6, s6, 1
 ; GFX6-NEXT:    s_lshl_b32 s7, s7, 1
 ; GFX6-NEXT:    s_and_b32 s6, s6, 62
-; GFX6-NEXT:    s_and_b32 s8, s7, 62
+; GFX6-NEXT:    s_and_b32 s7, s7, 62
 ; GFX6-NEXT:    s_add_u32 s6, s6, s4
-; GFX6-NEXT:    s_addc_u32 s7, 0, s5
-; GFX6-NEXT:    s_add_u32 s4, s8, s4
+; GFX6-NEXT:    s_addc_u32 s8, 0, s5
+; GFX6-NEXT:    s_add_u32 s4, s7, s4
 ; GFX6-NEXT:    s_addc_u32 s5, 0, s5
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s6
-; GFX6-NEXT:    v_mov_b32_e32 v1, s7
+; GFX6-NEXT:    v_mov_b32_e32 v1, s8
 ; GFX6-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
@@ -762,13 +762,13 @@ define amdgpu_kernel void @s_and_multi_use_inline_imm_i64(ptr addrspace(1) %out,
 ; GFX8-NEXT:    s_lshl_b32 s6, s6, 1
 ; GFX8-NEXT:    s_lshl_b32 s7, s7, 1
 ; GFX8-NEXT:    s_and_b32 s6, s6, 62
-; GFX8-NEXT:    s_and_b32 s8, s7, 62
+; GFX8-NEXT:    s_and_b32 s7, s7, 62
 ; GFX8-NEXT:    s_add_u32 s6, s6, s4
-; GFX8-NEXT:    s_addc_u32 s7, 0, s5
-; GFX8-NEXT:    s_add_u32 s4, s8, s4
+; GFX8-NEXT:    s_addc_u32 s8, 0, s5
+; GFX8-NEXT:    s_add_u32 s4, s7, s4
 ; GFX8-NEXT:    s_addc_u32 s5, 0, s5
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s6
-; GFX8-NEXT:    v_mov_b32_e32 v1, s7
+; GFX8-NEXT:    v_mov_b32_e32 v1, s8
 ; GFX8-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s4
diff --git a/llvm/test/CodeGen/AMDGPU/any_extend_vector_inreg.ll b/llvm/test/CodeGen/AMDGPU/any_extend_vector_inreg.ll
index 7b6197e3fcb70..bdf9804d5cad1 100644
--- a/llvm/test/CodeGen/AMDGPU/any_extend_vector_inreg.ll
+++ b/llvm/test/CodeGen/AMDGPU/any_extend_vector_inreg.ll
@@ -57,92 +57,92 @@ define amdgpu_kernel void @any_extend_vector_inreg_v16i8_to_v4i32(ptr addrspace(
 ; GFX8-NEXT:    s_load_dwordx8 s[0:7], s[8:9], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_load_dword s0, s[8:9], 0x20
-; GFX8-NEXT:    s_lshr_b32 s6, s5, 24
-; GFX8-NEXT:    s_lshr_b32 s8, s2, 24
+; GFX8-NEXT:    s_lshr_b32 s4, s5, 24
+; GFX8-NEXT:    s_lshr_b32 s5, s2, 24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_lshl_b64 s[2:3], s[0:1], 8
-; GFX8-NEXT:    s_add_u32 s4, s10, 13
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 15
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
-; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 14
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
-; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 8
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
-; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 11
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
-; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 10
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
-; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 4
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
-; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 6
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
-; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 1
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
-; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 3
+; GFX8-NEXT:    s_add_u32 s0, s10, 13
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 15
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
+; GFX8-NEXT:    flat_store_byte v[0:1], v2
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 14
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
+; GFX8-NEXT:    flat_store_byte v[0:1], v2
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 8
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
+; GFX8-NEXT:    flat_store_byte v[0:1], v2
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 11
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
+; GFX8-NEXT:    flat_store_byte v[0:1], v2
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 10
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
+; GFX8-NEXT:    flat_store_byte v[0:1], v2
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 4
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
+; GFX8-NEXT:    flat_store_byte v[0:1], v2
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 6
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
+; GFX8-NEXT:    flat_store_byte v[0:1], v2
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 1
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
+; GFX8-NEXT:    flat_store_byte v[0:1], v2
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    s_add_u32 s0, s10, 3
 ; GFX8-NEXT:    flat_store_byte v[0:1], v2
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s10
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s11
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
 ; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 9
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 9
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
 ; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    s_add_u32 s4, s10, 2
-; GFX8-NEXT:    s_addc_u32 s5, s11, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 2
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s7
-; GFX8-NEXT:    s_add_u32 s0, s10, 5
+; GFX8-NEXT:    s_addc_u32 s2, s11, 0
 ; GFX8-NEXT:    flat_store_byte v[0:1], v2
-; GFX8-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    s_add_u32 s0, s10, 5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX8-NEXT:    s_addc_u32 s1, s11, 0
 ; GFX8-NEXT:    flat_store_byte v[0:1], v2
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    s_add_u32 s0, s10, 12
-; GFX8-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
+; GFX8-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX8-NEXT:    s_addc_u32 s1, s11, 0
 ; GFX8-NEXT:    flat_store_byte v[0:1], v2
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    s_add_u32 s0, s10, 7
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX8-NEXT:    s_addc_u32 s1, s11, 0
 ; GFX8-NEXT:    flat_store_byte v[0:1], v2
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s6
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    flat_store_byte v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 bb:
diff --git a/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_cond.ll b/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_cond.ll
index bbc9ccdf23b2c..d3785d6d6d882 100644
--- a/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_cond.ll
+++ b/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_cond.ll
@@ -116,13 +116,13 @@ define amdgpu_kernel void @flat_atomic_usub_cond_rtn_u32(ptr %addr, i32 %in, ptr
 ; GFX9-SDAG-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-SDAG-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-SDAG-NEXT:    s_add_u32 s8, s0, 16
+; GFX9-SDAG-NEXT:    s_add_u32 s7, s0, 16
 ; GFX9-SDAG-NEXT:    v_subrev_u32_e32 v0, s6, v1
-; GFX9-SDAG-NEXT:    s_addc_u32 s9, s1, 0
+; GFX9-SDAG-NEXT:    s_addc_u32 s8, s1, 0
 ; GFX9-SDAG-NEXT:    v_cmp_le_u32_e32 vcc, s6, v1
 ; GFX9-SDAG-NEXT:    v_cndmask_b32_e32 v0, v1, v0, vcc
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v2, s8
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v2, s7
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v3, s8
 ; GFX9-SDAG-NEXT:    flat_atomic_cmpswap v0, v[2:3], v[0:1] glc
 ; GFX9-SDAG-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-SDAG-NEXT:    buffer_wbinvl1_vol
@@ -172,13 +172,13 @@ define amdgpu_kernel void @flat_atomic_usub_cond_rtn_u32(ptr %addr, i32 %in, ptr
 ; GFX9-GISEL-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-GISEL-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-GISEL-NEXT:    s_add_u32 s8, s0, 16
+; GFX9-GISEL-NEXT:    s_add_u32 s7, s0, 16
 ; GFX9-GISEL-NEXT:    v_subrev_u32_e32 v0, s6, v1
-; GFX9-GISEL-NEXT:    s_addc_u32 s9, s1, 0
+; GFX9-GISEL-NEXT:    s_addc_u32 s8, s1, 0
 ; GFX9-GISEL-NEXT:    v_cmp_le_u32_e32 vcc, s6, v1
 ; GFX9-GISEL-NEXT:    v_cndmask_b32_e32 v0, v1, v0, vcc
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v2, s8
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v3, s9
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v2, s7
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v3, s8
 ; GFX9-GISEL-NEXT:    flat_atomic_cmpswap v0, v[2:3], v[0:1] glc
 ; GFX9-GISEL-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-GISEL-NEXT:    buffer_wbinvl1_vol
diff --git a/llvm/test/CodeGen/AMDGPU/carryout-selection.ll b/llvm/test/CodeGen/AMDGPU/carryout-selection.ll
index 0a7aa5d32e21d..170a42d49c376 100644
--- a/llvm/test/CodeGen/AMDGPU/carryout-selection.ll
+++ b/llvm/test/CodeGen/AMDGPU/carryout-selection.ll
@@ -39,8 +39,8 @@ define amdgpu_kernel void @sadd64rr(ptr addrspace(1) %out, i64 %a, i64 %b) {
 ; VI-NEXT:    s_add_u32 s0, s2, s4
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-NEXT:    s_addc_u32 s1, s3, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; VI-NEXT:    s_endpgm
 ;
@@ -197,8 +197,8 @@ define amdgpu_kernel void @sadd64ri(ptr addrspace(1) %out, i64 %a) {
 ; VI-NEXT:    s_add_u32 s0, s2, 0x56789876
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-NEXT:    s_addc_u32 s1, s3, 0x1234
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; VI-NEXT:    s_endpgm
 ;
@@ -1362,8 +1362,8 @@ define amdgpu_kernel void @ssub64rr(ptr addrspace(1) %out, i64 %a, i64 %b) {
 ; VI-NEXT:    s_sub_u32 s0, s2, s4
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-NEXT:    s_subb_u32 s1, s3, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; VI-NEXT:    s_endpgm
 ;
@@ -1520,8 +1520,8 @@ define amdgpu_kernel void @ssub64ri(ptr addrspace(1) %out, i64 %a) {
 ; VI-NEXT:    s_sub_u32 s0, 0x56789876, s2
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-NEXT:    s_subb_u32 s1, 0x1234, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; VI-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/cluster_stores.ll b/llvm/test/CodeGen/AMDGPU/cluster_stores.ll
index 0aaa8167b5fe2..9bf5d4f2556b5 100644
--- a/llvm/test/CodeGen/AMDGPU/cluster_stores.ll
+++ b/llvm/test/CodeGen/AMDGPU/cluster_stores.ll
@@ -54,17 +54,17 @@ define amdgpu_kernel void @cluster_load_cluster_store(ptr noalias %lb, ptr noali
 ; GFX10-NEXT:    s_add_u32 s4, s0, 8
 ; GFX10-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX10-NEXT:    s_add_u32 s6, s0, 16
-; GFX10-NEXT:    s_addc_u32 s7, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v2, s4
+; GFX10-NEXT:    s_addc_u32 s4, s1, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-NEXT:    s_add_u32 s0, s0, 24
-; GFX10-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX10-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX10-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v4, s6
-; GFX10-NEXT:    v_mov_b32_e32 v5, s7
-; GFX10-NEXT:    v_mov_b32_e32 v7, s1
+; GFX10-NEXT:    v_mov_b32_e32 v5, s4
 ; GFX10-NEXT:    v_mov_b32_e32 v6, s0
+; GFX10-NEXT:    v_mov_b32_e32 v7, s1
 ; GFX10-NEXT:    s_clause 0x3
 ; GFX10-NEXT:    flat_load_dword v8, v[0:1]
 ; GFX10-NEXT:    flat_load_dword v9, v[2:3]
@@ -73,17 +73,17 @@ define amdgpu_kernel void @cluster_load_cluster_store(ptr noalias %lb, ptr noali
 ; GFX10-NEXT:    s_add_u32 s0, s2, 8
 ; GFX10-NEXT:    s_addc_u32 s1, s3, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v2, s0
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    s_add_u32 s0, s2, 16
+; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    s_addc_u32 s1, s3, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-NEXT:    s_add_u32 s2, s2, 24
 ; GFX10-NEXT:    s_addc_u32 s3, s3, 0
-; GFX10-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX10-NEXT:    v_mov_b32_e32 v4, s0
-; GFX10-NEXT:    v_mov_b32_e32 v7, s3
+; GFX10-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX10-NEXT:    v_mov_b32_e32 v6, s2
+; GFX10-NEXT:    v_mov_b32_e32 v7, s3
 ; GFX10-NEXT:    s_waitcnt vmcnt(3) lgkmcnt(3)
 ; GFX10-NEXT:    flat_store_dword v[0:1], v8
 ; GFX10-NEXT:    s_waitcnt vmcnt(2) lgkmcnt(3)
@@ -182,16 +182,16 @@ define amdgpu_kernel void @cluster_load_valu_cluster_store(ptr noalias %lb, ptr
 ; GFX10-NEXT:    s_add_u32 s6, s0, 16
 ; GFX10-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX10-NEXT:    v_mov_b32_e32 v3, s5
-; GFX10-NEXT:    s_addc_u32 s7, s1, 0
+; GFX10-NEXT:    s_addc_u32 s4, s1, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-NEXT:    s_add_u32 s0, s0, 24
 ; GFX10-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v4, s6
-; GFX10-NEXT:    v_mov_b32_e32 v5, s7
+; GFX10-NEXT:    v_mov_b32_e32 v5, s4
 ; GFX10-NEXT:    flat_load_dword v6, v[2:3]
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    v_mov_b32_e32 v2, s0
+; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    s_clause 0x2
 ; GFX10-NEXT:    flat_load_dword v8, v[0:1]
 ; GFX10-NEXT:    flat_load_dword v9, v[4:5]
@@ -199,19 +199,19 @@ define amdgpu_kernel void @cluster_load_valu_cluster_store(ptr noalias %lb, ptr
 ; GFX10-NEXT:    s_add_u32 s0, s2, 8
 ; GFX10-NEXT:    s_addc_u32 s1, s3, 0
 ; GFX10-NEXT:    s_add_u32 s4, s2, 16
-; GFX10-NEXT:    s_addc_u32 s5, s3, 0
+; GFX10-NEXT:    v_mov_b32_e32 v2, s0
+; GFX10-NEXT:    s_addc_u32 s0, s3, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-NEXT:    v_mov_b32_e32 v3, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s0
-; GFX10-NEXT:    s_add_u32 s0, s2, 24
+; GFX10-NEXT:    s_add_u32 s1, s2, 24
 ; GFX10-NEXT:    v_mov_b32_e32 v4, s4
-; GFX10-NEXT:    v_mov_b32_e32 v5, s5
-; GFX10-NEXT:    s_addc_u32 s1, s3, 0
-; GFX10-NEXT:    v_mov_b32_e32 v7, s1
+; GFX10-NEXT:    v_mov_b32_e32 v5, s0
+; GFX10-NEXT:    s_addc_u32 s0, s3, 0
+; GFX10-NEXT:    v_mov_b32_e32 v7, s0
 ; GFX10-NEXT:    s_waitcnt vmcnt(3) lgkmcnt(3)
 ; GFX10-NEXT:    v_add_nc_u32_e32 v11, 1, v6
-; GFX10-NEXT:    v_mov_b32_e32 v6, s0
+; GFX10-NEXT:    v_mov_b32_e32 v6, s1
 ; GFX10-NEXT:    s_waitcnt vmcnt(2) lgkmcnt(2)
 ; GFX10-NEXT:    flat_store_dword v[0:1], v8
 ; GFX10-NEXT:    s_waitcnt vmcnt(1) lgkmcnt(2)
diff --git a/llvm/test/CodeGen/AMDGPU/ctlz.ll b/llvm/test/CodeGen/AMDGPU/ctlz.ll
index d9b6edf00b82e..81b69c3101cab 100644
--- a/llvm/test/CodeGen/AMDGPU/ctlz.ll
+++ b/llvm/test/CodeGen/AMDGPU/ctlz.ll
@@ -658,12 +658,11 @@ define amdgpu_kernel void @s_ctlz_i64(ptr addrspace(1) noalias %out, [8 x i32],
 ; GFX10-GISEL-NEXT:    s_clause 0x1
 ; GFX10-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x4c
 ; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-GISEL-NEXT:    s_flbit_i32_b64 s0, s[0:1]
-; GFX10-GISEL-NEXT:    s_mov_b32 s1, 0
 ; GFX10-GISEL-NEXT:    s_min_u32 s0, s0, 64
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
 ; GFX10-GISEL-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/ctlz_zero_poison.ll b/llvm/test/CodeGen/AMDGPU/ctlz_zero_poison.ll
index b9a47766cb089..c53601ed7b05b 100644
--- a/llvm/test/CodeGen/AMDGPU/ctlz_zero_poison.ll
+++ b/llvm/test/CodeGen/AMDGPU/ctlz_zero_poison.ll
@@ -562,12 +562,11 @@ define amdgpu_kernel void @s_ctlz_zero_poison_i64_with_select(ptr addrspace(1) n
 ; GFX9-GISEL-LABEL: s_ctlz_zero_poison_i64_with_select:
 ; GFX9-GISEL:       ; %bb.0:
 ; GFX9-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; GFX9-GISEL-NEXT:    s_mov_b32 s5, 0
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-GISEL-NEXT:    s_flbit_i32_b64 s4, s[2:3]
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9-GISEL-NEXT:    s_flbit_i32_b64 s2, s[2:3]
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX9-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9-GISEL-NEXT:    s_endpgm
   %ctlz = tail call i64 @llvm.ctlz.i64(i64 %val, i1 true) nounwind readnone
@@ -702,11 +701,11 @@ define amdgpu_kernel void @v_ctlz_zero_poison_i16_with_select(ptr addrspace(1) n
 ; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s4, s2, 1
-; VI-NEXT:    s_addc_u32 s5, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    s_addc_u32 s2, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v2, s4
+; VI-NEXT:    v_mov_b32_e32 v3, s2
 ; VI-NEXT:    flat_load_ubyte v2, v[2:3]
 ; VI-NEXT:    flat_load_ubyte v3, v[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v0, s0
@@ -823,8 +822,8 @@ define amdgpu_kernel void @v_ctlz_zero_poison_i32_with_select(ptr addrspace(1) n
 ; VI-NEXT:    s_add_u32 s4, s2, 1
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 3
+; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
@@ -832,8 +831,8 @@ define amdgpu_kernel void @v_ctlz_zero_poison_i32_with_select(ptr addrspace(1) n
 ; VI-NEXT:    v_mov_b32_e32 v4, s4
 ; VI-NEXT:    v_mov_b32_e32 v5, s5
 ; VI-NEXT:    s_addc_u32 s3, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    v_mov_b32_e32 v6, s2
+; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    flat_load_ubyte v2, v[2:3]
 ; VI-NEXT:    flat_load_ubyte v3, v[0:1]
 ; VI-NEXT:    flat_load_ubyte v4, v[4:5]
@@ -969,24 +968,24 @@ define amdgpu_kernel void @v_ctlz_zero_poison_i64_with_select(ptr addrspace(1) n
 ; VI-NEXT:    s_add_u32 s4, s2, 5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s4
-; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 7
+; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v4, s4
-; VI-NEXT:    v_mov_b32_e32 v5, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 6
+; VI-NEXT:    v_mov_b32_e32 v5, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v7, s5
 ; VI-NEXT:    v_mov_b32_e32 v6, s4
 ; VI-NEXT:    s_add_u32 s4, s2, 1
+; VI-NEXT:    v_mov_b32_e32 v7, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v9, s5
 ; VI-NEXT:    v_mov_b32_e32 v8, s4
 ; VI-NEXT:    s_add_u32 s4, s2, 3
+; VI-NEXT:    v_mov_b32_e32 v9, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    flat_load_ubyte v10, v[0:1]
 ; VI-NEXT:    flat_load_ubyte v11, v[2:3]
@@ -999,8 +998,8 @@ define amdgpu_kernel void @v_ctlz_zero_poison_i64_with_select(ptr addrspace(1) n
 ; VI-NEXT:    v_mov_b32_e32 v2, s4
 ; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_addc_u32 s3, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    flat_load_ubyte v8, v[0:1]
 ; VI-NEXT:    flat_load_ubyte v2, v[2:3]
 ; VI-NEXT:    flat_load_ubyte v3, v[4:5]
@@ -1259,12 +1258,11 @@ define amdgpu_kernel void @s_ctlz_zero_poison_i64(ptr addrspace(1) noalias %out,
 ; GFX9-GISEL:       ; %bb.0:
 ; GFX9-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x4c
 ; GFX9-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x24
-; GFX9-GISEL-NEXT:    s_mov_b32 s5, 0
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-GISEL-NEXT:    s_flbit_i32_b64 s4, s[0:1]
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9-GISEL-NEXT:    s_flbit_i32_b64 s0, s[0:1]
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX9-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
 ; GFX9-GISEL-NEXT:    s_endpgm
   %ctlz = call i64 @llvm.ctlz.i64(i64 %val, i1 true)
diff --git a/llvm/test/CodeGen/AMDGPU/ctpop64.ll b/llvm/test/CodeGen/AMDGPU/ctpop64.ll
index edf9ef109a42d..f36abb3071df1 100644
--- a/llvm/test/CodeGen/AMDGPU/ctpop64.ll
+++ b/llvm/test/CodeGen/AMDGPU/ctpop64.ll
@@ -277,16 +277,16 @@ define amdgpu_kernel void @s_ctpop_v4i64(ptr addrspace(1) noalias %out, <4 x i64
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    s_load_b256 s[8:15], s[4:5], 0x44
-; GFX12-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24
+; GFX12-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX12-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    s_bcnt1_i32_b64 s0, s[8:9]
-; GFX12-NEXT:    s_bcnt1_i32_b64 s1, s[10:11]
-; GFX12-NEXT:    s_bcnt1_i32_b64 s2, s[12:13]
-; GFX12-NEXT:    s_bcnt1_i32_b64 s3, s[14:15]
-; GFX12-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5]
+; GFX12-NEXT:    s_bcnt1_i32_b64 s2, s[8:9]
+; GFX12-NEXT:    s_bcnt1_i32_b64 s3, s[10:11]
+; GFX12-NEXT:    s_bcnt1_i32_b64 s4, s[12:13]
+; GFX12-NEXT:    s_bcnt1_i32_b64 s5, s[14:15]
+; GFX12-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s5
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1]
 ; GFX12-NEXT:    s_endpgm
   %ctpop = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %val) nounwind readnone
   %truncctpop = trunc <4 x i64> %ctpop to <4 x i32>
diff --git a/llvm/test/CodeGen/AMDGPU/cttz.ll b/llvm/test/CodeGen/AMDGPU/cttz.ll
index 81227d358ad29..731a446ef16f2 100644
--- a/llvm/test/CodeGen/AMDGPU/cttz.ll
+++ b/llvm/test/CodeGen/AMDGPU/cttz.ll
@@ -569,12 +569,11 @@ define amdgpu_kernel void @s_cttz_i64(ptr addrspace(1) noalias %out, [8 x i32],
 ; GFX10-GISEL-NEXT:    s_clause 0x1
 ; GFX10-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x4c
 ; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-GISEL-NEXT:    s_ff1_i32_b64 s0, s[0:1]
-; GFX10-GISEL-NEXT:    s_mov_b32 s1, 0
 ; GFX10-GISEL-NEXT:    s_min_u32 s0, s0, 64
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
 ; GFX10-GISEL-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/cttz_zero_poison.ll b/llvm/test/CodeGen/AMDGPU/cttz_zero_poison.ll
index 8cce8f0c143ca..af929749b5060 100644
--- a/llvm/test/CodeGen/AMDGPU/cttz_zero_poison.ll
+++ b/llvm/test/CodeGen/AMDGPU/cttz_zero_poison.ll
@@ -539,12 +539,11 @@ define amdgpu_kernel void @s_cttz_zero_poison_i64_with_select(ptr addrspace(1) n
 ; GFX9-GISEL-LABEL: s_cttz_zero_poison_i64_with_select:
 ; GFX9-GISEL:       ; %bb.0:
 ; GFX9-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; GFX9-GISEL-NEXT:    s_mov_b32 s5, 0
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-GISEL-NEXT:    s_ff1_i32_b64 s4, s[2:3]
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9-GISEL-NEXT:    s_ff1_i32_b64 s2, s[2:3]
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX9-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9-GISEL-NEXT:    s_endpgm
   %cttz = tail call i64 @llvm.cttz.i64(i64 %val, i1 true) nounwind readnone
@@ -673,11 +672,11 @@ define amdgpu_kernel void @v_cttz_zero_poison_i16_with_select(ptr addrspace(1) n
 ; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s4, s2, 1
-; VI-NEXT:    s_addc_u32 s5, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    s_addc_u32 s2, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v2, s4
+; VI-NEXT:    v_mov_b32_e32 v3, s2
 ; VI-NEXT:    flat_load_ubyte v2, v[2:3]
 ; VI-NEXT:    flat_load_ubyte v3, v[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v0, s0
@@ -788,8 +787,8 @@ define amdgpu_kernel void @v_cttz_zero_poison_i32_with_select(ptr addrspace(1) n
 ; VI-NEXT:    s_add_u32 s4, s2, 1
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 3
+; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
@@ -797,8 +796,8 @@ define amdgpu_kernel void @v_cttz_zero_poison_i32_with_select(ptr addrspace(1) n
 ; VI-NEXT:    v_mov_b32_e32 v4, s4
 ; VI-NEXT:    v_mov_b32_e32 v5, s5
 ; VI-NEXT:    s_addc_u32 s3, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    v_mov_b32_e32 v6, s2
+; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    flat_load_ubyte v2, v[2:3]
 ; VI-NEXT:    flat_load_ubyte v3, v[0:1]
 ; VI-NEXT:    flat_load_ubyte v4, v[4:5]
@@ -933,24 +932,24 @@ define amdgpu_kernel void @v_cttz_zero_poison_i64_with_select(ptr addrspace(1) n
 ; VI-NEXT:    s_add_u32 s4, s2, 5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s4
-; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 7
+; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v4, s4
-; VI-NEXT:    v_mov_b32_e32 v5, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 6
+; VI-NEXT:    v_mov_b32_e32 v5, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v7, s5
 ; VI-NEXT:    v_mov_b32_e32 v6, s4
 ; VI-NEXT:    s_add_u32 s4, s2, 1
+; VI-NEXT:    v_mov_b32_e32 v7, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v9, s5
 ; VI-NEXT:    v_mov_b32_e32 v8, s4
 ; VI-NEXT:    s_add_u32 s4, s2, 3
+; VI-NEXT:    v_mov_b32_e32 v9, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v11, s3
 ; VI-NEXT:    v_mov_b32_e32 v10, s2
@@ -1125,8 +1124,8 @@ define amdgpu_kernel void @v_cttz_i32_sel_eq_neg1(ptr addrspace(1) noalias %out,
 ; VI-NEXT:    s_add_u32 s4, s2, 1
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 3
+; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
@@ -1134,8 +1133,8 @@ define amdgpu_kernel void @v_cttz_i32_sel_eq_neg1(ptr addrspace(1) noalias %out,
 ; VI-NEXT:    v_mov_b32_e32 v4, s4
 ; VI-NEXT:    v_mov_b32_e32 v5, s5
 ; VI-NEXT:    s_addc_u32 s3, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    v_mov_b32_e32 v6, s2
+; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    flat_load_ubyte v2, v[2:3]
 ; VI-NEXT:    flat_load_ubyte v3, v[0:1]
 ; VI-NEXT:    flat_load_ubyte v4, v[4:5]
@@ -1255,8 +1254,8 @@ define amdgpu_kernel void @v_cttz_i32_sel_ne_neg1(ptr addrspace(1) noalias %out,
 ; VI-NEXT:    s_add_u32 s4, s2, 1
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 3
+; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
@@ -1264,8 +1263,8 @@ define amdgpu_kernel void @v_cttz_i32_sel_ne_neg1(ptr addrspace(1) noalias %out,
 ; VI-NEXT:    v_mov_b32_e32 v4, s4
 ; VI-NEXT:    v_mov_b32_e32 v5, s5
 ; VI-NEXT:    s_addc_u32 s3, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    v_mov_b32_e32 v6, s2
+; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    flat_load_ubyte v2, v[2:3]
 ; VI-NEXT:    flat_load_ubyte v3, v[0:1]
 ; VI-NEXT:    flat_load_ubyte v4, v[4:5]
@@ -1388,8 +1387,8 @@ define amdgpu_kernel void @v_cttz_i32_sel_ne_bitwidth(ptr addrspace(1) noalias %
 ; VI-NEXT:    s_add_u32 s4, s2, 1
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 3
+; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
@@ -1397,8 +1396,8 @@ define amdgpu_kernel void @v_cttz_i32_sel_ne_bitwidth(ptr addrspace(1) noalias %
 ; VI-NEXT:    v_mov_b32_e32 v4, s4
 ; VI-NEXT:    v_mov_b32_e32 v5, s5
 ; VI-NEXT:    s_addc_u32 s3, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    v_mov_b32_e32 v6, s2
+; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    flat_load_ubyte v2, v[2:3]
 ; VI-NEXT:    flat_load_ubyte v3, v[0:1]
 ; VI-NEXT:    flat_load_ubyte v4, v[4:5]
diff --git a/llvm/test/CodeGen/AMDGPU/extract_vector_dynelt.ll b/llvm/test/CodeGen/AMDGPU/extract_vector_dynelt.ll
index aabed8e2814c0..ddccd6e0549ff 100644
--- a/llvm/test/CodeGen/AMDGPU/extract_vector_dynelt.ll
+++ b/llvm/test/CodeGen/AMDGPU/extract_vector_dynelt.ll
@@ -752,46 +752,30 @@ entry:
 define amdgpu_kernel void @double8_extelt(ptr addrspace(1) %out, i32 %sel) {
 ; GCN-LABEL: double8_extelt:
 ; GCN:       ; %bb.0: ; %entry
-; GCN-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
-; GCN-NEXT:    s_load_dword s18, s[4:5], 0x2c
-; GCN-NEXT:    s_mov_b32 s15, 0x40200000
-; GCN-NEXT:    s_mov_b32 s1, 0x3ff00000
-; GCN-NEXT:    s_mov_b32 s0, 0
-; GCN-NEXT:    s_mov_b32 s13, 0x401c0000
-; GCN-NEXT:    s_mov_b32 s11, 0x40180000
-; GCN-NEXT:    s_mov_b32 s9, 0x40140000
-; GCN-NEXT:    s_mov_b32 s7, 0x40100000
-; GCN-NEXT:    s_mov_b32 s5, 0x40080000
-; GCN-NEXT:    s_mov_b32 s3, 2.0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s8, s0
-; GCN-NEXT:    s_mov_b32 s10, s0
-; GCN-NEXT:    s_mov_b32 s12, s0
-; GCN-NEXT:    s_mov_b32 s14, s0
-; GCN-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-NEXT:    v_mov_b32_e32 v1, s1
-; GCN-NEXT:    v_mov_b32_e32 v15, s15
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GCN-NEXT:    s_load_dword s2, s[4:5], 0x2c
+; GCN-NEXT:    v_mov_b32_e32 v15, 0x40200000
+; GCN-NEXT:    v_mov_b32_e32 v1, 0x3ff00000
+; GCN-NEXT:    v_mov_b32_e32 v0, 0
+; GCN-NEXT:    v_mov_b32_e32 v13, 0x401c0000
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NEXT:    s_lshl_b32 m0, s18, 1
-; GCN-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-NEXT:    v_mov_b32_e32 v3, s3
-; GCN-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NEXT:    v_mov_b32_e32 v6, s6
-; GCN-NEXT:    v_mov_b32_e32 v7, s7
-; GCN-NEXT:    v_mov_b32_e32 v8, s8
-; GCN-NEXT:    v_mov_b32_e32 v9, s9
-; GCN-NEXT:    v_mov_b32_e32 v10, s10
-; GCN-NEXT:    v_mov_b32_e32 v11, s11
-; GCN-NEXT:    v_mov_b32_e32 v12, s12
-; GCN-NEXT:    v_mov_b32_e32 v13, s13
-; GCN-NEXT:    v_mov_b32_e32 v14, s14
+; GCN-NEXT:    s_lshl_b32 m0, s2, 1
+; GCN-NEXT:    v_mov_b32_e32 v11, 0x40180000
+; GCN-NEXT:    v_mov_b32_e32 v9, 0x40140000
+; GCN-NEXT:    v_mov_b32_e32 v7, 0x40100000
+; GCN-NEXT:    v_mov_b32_e32 v5, 0x40080000
+; GCN-NEXT:    v_mov_b32_e32 v3, 2.0
+; GCN-NEXT:    v_mov_b32_e32 v2, v0
+; GCN-NEXT:    v_mov_b32_e32 v4, v0
+; GCN-NEXT:    v_mov_b32_e32 v6, v0
+; GCN-NEXT:    v_mov_b32_e32 v8, v0
+; GCN-NEXT:    v_mov_b32_e32 v10, v0
+; GCN-NEXT:    v_mov_b32_e32 v12, v0
+; GCN-NEXT:    v_mov_b32_e32 v14, v0
 ; GCN-NEXT:    v_movrels_b32_e32 v16, v1
 ; GCN-NEXT:    v_movrels_b32_e32 v15, v0
-; GCN-NEXT:    v_mov_b32_e32 v0, s16
-; GCN-NEXT:    v_mov_b32_e32 v1, s17
+; GCN-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN-NEXT:    flat_store_dwordx2 v[0:1], v[15:16]
 ; GCN-NEXT:    s_endpgm
 ;
@@ -904,45 +888,29 @@ entry:
 define amdgpu_kernel void @double7_extelt(ptr addrspace(1) %out, i32 %sel) {
 ; GCN-LABEL: double7_extelt:
 ; GCN:       ; %bb.0: ; %entry
-; GCN-NEXT:    s_load_dwordx2 s[14:15], s[4:5], 0x24
-; GCN-NEXT:    s_load_dword s16, s[4:5], 0x2c
-; GCN-NEXT:    s_mov_b32 s1, 0x3ff00000
-; GCN-NEXT:    s_mov_b32 s0, 0
-; GCN-NEXT:    s_mov_b32 s13, 0x401c0000
-; GCN-NEXT:    s_mov_b32 s11, 0x40180000
-; GCN-NEXT:    s_mov_b32 s9, 0x40140000
-; GCN-NEXT:    s_mov_b32 s7, 0x40100000
-; GCN-NEXT:    s_mov_b32 s5, 0x40080000
-; GCN-NEXT:    s_mov_b32 s3, 2.0
-; GCN-NEXT:    s_mov_b32 s2, s0
-; GCN-NEXT:    s_mov_b32 s4, s0
-; GCN-NEXT:    s_mov_b32 s6, s0
-; GCN-NEXT:    s_mov_b32 s8, s0
-; GCN-NEXT:    s_mov_b32 s10, s0
-; GCN-NEXT:    s_mov_b32 s12, s0
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GCN-NEXT:    s_load_dword s2, s[4:5], 0x2c
+; GCN-NEXT:    v_mov_b32_e32 v13, 0x401c0000
+; GCN-NEXT:    v_mov_b32_e32 v1, 0x3ff00000
+; GCN-NEXT:    v_mov_b32_e32 v0, 0
+; GCN-NEXT:    v_mov_b32_e32 v11, 0x40180000
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    s_lshl_b32 m0, s2, 1
+; GCN-NEXT:    v_mov_b32_e32 v9, 0x40140000
+; GCN-NEXT:    v_mov_b32_e32 v7, 0x40100000
+; GCN-NEXT:    v_mov_b32_e32 v5, 0x40080000
+; GCN-NEXT:    v_mov_b32_e32 v3, 2.0
+; GCN-NEXT:    v_mov_b32_e32 v2, v0
+; GCN-NEXT:    v_mov_b32_e32 v4, v0
+; GCN-NEXT:    v_mov_b32_e32 v6, v0
+; GCN-NEXT:    v_mov_b32_e32 v8, v0
+; GCN-NEXT:    v_mov_b32_e32 v10, v0
+; GCN-NEXT:    v_mov_b32_e32 v12, v0
+; GCN-NEXT:    v_movrels_b32_e32 v14, v1
+; GCN-NEXT:    v_movrels_b32_e32 v13, v0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s0
 ; GCN-NEXT:    v_mov_b32_e32 v1, s1
-; GCN-NEXT:    v_mov_b32_e32 v15, s15
-; GCN-NEXT:    s_lshl_b32 m0, s16, 1
-; GCN-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-NEXT:    v_mov_b32_e32 v3, s3
-; GCN-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NEXT:    v_mov_b32_e32 v6, s6
-; GCN-NEXT:    v_mov_b32_e32 v7, s7
-; GCN-NEXT:    v_mov_b32_e32 v8, s8
-; GCN-NEXT:    v_mov_b32_e32 v9, s9
-; GCN-NEXT:    v_mov_b32_e32 v10, s10
-; GCN-NEXT:    v_mov_b32_e32 v11, s11
-; GCN-NEXT:    v_mov_b32_e32 v12, s12
-; GCN-NEXT:    v_mov_b32_e32 v13, s13
-; GCN-NEXT:    v_mov_b32_e32 v14, s14
-; GCN-NEXT:    v_movrels_b32_e32 v16, v1
-; GCN-NEXT:    v_movrels_b32_e32 v15, v0
-; GCN-NEXT:    v_mov_b32_e32 v0, s14
-; GCN-NEXT:    v_mov_b32_e32 v1, s15
-; GCN-NEXT:    flat_store_dwordx2 v[0:1], v[15:16]
+; GCN-NEXT:    flat_store_dwordx2 v[0:1], v[13:14]
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-O0-LABEL: double7_extelt:
@@ -1154,75 +1122,43 @@ define amdgpu_kernel void @double15_extelt(ptr addrspace(1) %out, i32 %sel) {
 ; GCN:       ; %bb.0: ; %entry
 ; GCN-NEXT:    s_load_dword s2, s[4:5], 0x2c
 ; GCN-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GCN-NEXT:    s_mov_b32 s37, 0x3ff00000
-; GCN-NEXT:    s_mov_b32 s36, 0
-; GCN-NEXT:    s_mov_b32 s65, 0x402e0000
-; GCN-NEXT:    s_mov_b32 s63, 0x402c0000
-; GCN-NEXT:    s_mov_b32 s61, 0x402a0000
-; GCN-NEXT:    s_mov_b32 s59, 0x40280000
-; GCN-NEXT:    s_mov_b32 s57, 0x40260000
-; GCN-NEXT:    s_mov_b32 s55, 0x40240000
-; GCN-NEXT:    s_mov_b32 s53, 0x40220000
-; GCN-NEXT:    s_mov_b32 s51, 0x40200000
-; GCN-NEXT:    s_mov_b32 s49, 0x401c0000
-; GCN-NEXT:    s_mov_b32 s47, 0x40180000
-; GCN-NEXT:    s_mov_b32 s45, 0x40140000
-; GCN-NEXT:    s_mov_b32 s43, 0x40100000
-; GCN-NEXT:    s_mov_b32 s41, 0x40080000
-; GCN-NEXT:    s_mov_b32 s39, 2.0
-; GCN-NEXT:    s_mov_b32 s38, s36
-; GCN-NEXT:    s_mov_b32 s40, s36
-; GCN-NEXT:    s_mov_b32 s42, s36
-; GCN-NEXT:    s_mov_b32 s44, s36
-; GCN-NEXT:    s_mov_b32 s46, s36
-; GCN-NEXT:    s_mov_b32 s48, s36
-; GCN-NEXT:    s_mov_b32 s50, s36
-; GCN-NEXT:    s_mov_b32 s52, s36
-; GCN-NEXT:    s_mov_b32 s54, s36
-; GCN-NEXT:    s_mov_b32 s56, s36
-; GCN-NEXT:    s_mov_b32 s58, s36
-; GCN-NEXT:    s_mov_b32 s60, s36
-; GCN-NEXT:    s_mov_b32 s62, s36
-; GCN-NEXT:    s_mov_b32 s64, s36
-; GCN-NEXT:    v_mov_b32_e32 v0, s36
-; GCN-NEXT:    v_mov_b32_e32 v1, s37
-; GCN-NEXT:    v_mov_b32_e32 v31, s67
+; GCN-NEXT:    v_mov_b32_e32 v29, 0x402e0000
+; GCN-NEXT:    v_mov_b32_e32 v1, 0x3ff00000
+; GCN-NEXT:    v_mov_b32_e32 v0, 0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_lshl_b32 m0, s2, 1
-; GCN-NEXT:    v_mov_b32_e32 v2, s38
-; GCN-NEXT:    v_mov_b32_e32 v3, s39
-; GCN-NEXT:    v_mov_b32_e32 v4, s40
-; GCN-NEXT:    v_mov_b32_e32 v5, s41
-; GCN-NEXT:    v_mov_b32_e32 v6, s42
-; GCN-NEXT:    v_mov_b32_e32 v7, s43
-; GCN-NEXT:    v_mov_b32_e32 v8, s44
-; GCN-NEXT:    v_mov_b32_e32 v9, s45
-; GCN-NEXT:    v_mov_b32_e32 v10, s46
-; GCN-NEXT:    v_mov_b32_e32 v11, s47
-; GCN-NEXT:    v_mov_b32_e32 v12, s48
-; GCN-NEXT:    v_mov_b32_e32 v13, s49
-; GCN-NEXT:    v_mov_b32_e32 v14, s50
-; GCN-NEXT:    v_mov_b32_e32 v15, s51
-; GCN-NEXT:    v_mov_b32_e32 v16, s52
-; GCN-NEXT:    v_mov_b32_e32 v17, s53
-; GCN-NEXT:    v_mov_b32_e32 v18, s54
-; GCN-NEXT:    v_mov_b32_e32 v19, s55
-; GCN-NEXT:    v_mov_b32_e32 v20, s56
-; GCN-NEXT:    v_mov_b32_e32 v21, s57
-; GCN-NEXT:    v_mov_b32_e32 v22, s58
-; GCN-NEXT:    v_mov_b32_e32 v23, s59
-; GCN-NEXT:    v_mov_b32_e32 v24, s60
-; GCN-NEXT:    v_mov_b32_e32 v25, s61
-; GCN-NEXT:    v_mov_b32_e32 v26, s62
-; GCN-NEXT:    v_mov_b32_e32 v27, s63
-; GCN-NEXT:    v_mov_b32_e32 v28, s64
-; GCN-NEXT:    v_mov_b32_e32 v29, s65
-; GCN-NEXT:    v_mov_b32_e32 v30, s66
-; GCN-NEXT:    v_movrels_b32_e32 v32, v1
-; GCN-NEXT:    v_movrels_b32_e32 v31, v0
+; GCN-NEXT:    v_mov_b32_e32 v27, 0x402c0000
+; GCN-NEXT:    v_mov_b32_e32 v25, 0x402a0000
+; GCN-NEXT:    v_mov_b32_e32 v23, 0x40280000
+; GCN-NEXT:    v_mov_b32_e32 v21, 0x40260000
+; GCN-NEXT:    v_mov_b32_e32 v19, 0x40240000
+; GCN-NEXT:    v_mov_b32_e32 v17, 0x40220000
+; GCN-NEXT:    v_mov_b32_e32 v15, 0x40200000
+; GCN-NEXT:    v_mov_b32_e32 v13, 0x401c0000
+; GCN-NEXT:    v_mov_b32_e32 v11, 0x40180000
+; GCN-NEXT:    v_mov_b32_e32 v9, 0x40140000
+; GCN-NEXT:    v_mov_b32_e32 v7, 0x40100000
+; GCN-NEXT:    v_mov_b32_e32 v5, 0x40080000
+; GCN-NEXT:    v_mov_b32_e32 v3, 2.0
+; GCN-NEXT:    v_mov_b32_e32 v2, v0
+; GCN-NEXT:    v_mov_b32_e32 v4, v0
+; GCN-NEXT:    v_mov_b32_e32 v6, v0
+; GCN-NEXT:    v_mov_b32_e32 v8, v0
+; GCN-NEXT:    v_mov_b32_e32 v10, v0
+; GCN-NEXT:    v_mov_b32_e32 v12, v0
+; GCN-NEXT:    v_mov_b32_e32 v14, v0
+; GCN-NEXT:    v_mov_b32_e32 v16, v0
+; GCN-NEXT:    v_mov_b32_e32 v18, v0
+; GCN-NEXT:    v_mov_b32_e32 v20, v0
+; GCN-NEXT:    v_mov_b32_e32 v22, v0
+; GCN-NEXT:    v_mov_b32_e32 v24, v0
+; GCN-NEXT:    v_mov_b32_e32 v26, v0
+; GCN-NEXT:    v_mov_b32_e32 v28, v0
+; GCN-NEXT:    v_movrels_b32_e32 v30, v1
+; GCN-NEXT:    v_movrels_b32_e32 v29, v0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s0
 ; GCN-NEXT:    v_mov_b32_e32 v1, s1
-; GCN-NEXT:    flat_store_dwordx2 v[0:1], v[31:32]
+; GCN-NEXT:    flat_store_dwordx2 v[0:1], v[29:30]
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-O0-LABEL: double15_extelt:
@@ -1416,72 +1352,40 @@ define amdgpu_kernel void @double16_extelt(ptr addrspace(1) %out, i32 %sel) {
 ; GCN:       ; %bb.0: ; %entry
 ; GCN-NEXT:    s_load_dword s2, s[4:5], 0x2c
 ; GCN-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GCN-NEXT:    s_mov_b32 s67, 0x40300000
-; GCN-NEXT:    s_mov_b32 s37, 0x3ff00000
-; GCN-NEXT:    s_mov_b32 s36, 0
-; GCN-NEXT:    s_mov_b32 s65, 0x402e0000
-; GCN-NEXT:    s_mov_b32 s63, 0x402c0000
-; GCN-NEXT:    s_mov_b32 s61, 0x402a0000
-; GCN-NEXT:    s_mov_b32 s59, 0x40280000
-; GCN-NEXT:    s_mov_b32 s57, 0x40260000
-; GCN-NEXT:    s_mov_b32 s55, 0x40240000
-; GCN-NEXT:    s_mov_b32 s53, 0x40220000
-; GCN-NEXT:    s_mov_b32 s51, 0x40200000
-; GCN-NEXT:    s_mov_b32 s49, 0x401c0000
-; GCN-NEXT:    s_mov_b32 s47, 0x40180000
-; GCN-NEXT:    s_mov_b32 s45, 0x40140000
-; GCN-NEXT:    s_mov_b32 s43, 0x40100000
-; GCN-NEXT:    s_mov_b32 s41, 0x40080000
-; GCN-NEXT:    s_mov_b32 s39, 2.0
-; GCN-NEXT:    s_mov_b32 s38, s36
-; GCN-NEXT:    s_mov_b32 s40, s36
-; GCN-NEXT:    s_mov_b32 s42, s36
-; GCN-NEXT:    s_mov_b32 s44, s36
-; GCN-NEXT:    s_mov_b32 s46, s36
-; GCN-NEXT:    s_mov_b32 s48, s36
-; GCN-NEXT:    s_mov_b32 s50, s36
-; GCN-NEXT:    s_mov_b32 s52, s36
-; GCN-NEXT:    s_mov_b32 s54, s36
-; GCN-NEXT:    s_mov_b32 s56, s36
-; GCN-NEXT:    s_mov_b32 s58, s36
-; GCN-NEXT:    s_mov_b32 s60, s36
-; GCN-NEXT:    s_mov_b32 s62, s36
-; GCN-NEXT:    s_mov_b32 s64, s36
-; GCN-NEXT:    s_mov_b32 s66, s36
-; GCN-NEXT:    v_mov_b32_e32 v0, s36
-; GCN-NEXT:    v_mov_b32_e32 v1, s37
-; GCN-NEXT:    v_mov_b32_e32 v31, s67
+; GCN-NEXT:    v_mov_b32_e32 v31, 0x40300000
+; GCN-NEXT:    v_mov_b32_e32 v1, 0x3ff00000
+; GCN-NEXT:    v_mov_b32_e32 v0, 0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_lshl_b32 m0, s2, 1
-; GCN-NEXT:    v_mov_b32_e32 v2, s38
-; GCN-NEXT:    v_mov_b32_e32 v3, s39
-; GCN-NEXT:    v_mov_b32_e32 v4, s40
-; GCN-NEXT:    v_mov_b32_e32 v5, s41
-; GCN-NEXT:    v_mov_b32_e32 v6, s42
-; GCN-NEXT:    v_mov_b32_e32 v7, s43
-; GCN-NEXT:    v_mov_b32_e32 v8, s44
-; GCN-NEXT:    v_mov_b32_e32 v9, s45
-; GCN-NEXT:    v_mov_b32_e32 v10, s46
-; GCN-NEXT:    v_mov_b32_e32 v11, s47
-; GCN-NEXT:    v_mov_b32_e32 v12, s48
-; GCN-NEXT:    v_mov_b32_e32 v13, s49
-; GCN-NEXT:    v_mov_b32_e32 v14, s50
-; GCN-NEXT:    v_mov_b32_e32 v15, s51
-; GCN-NEXT:    v_mov_b32_e32 v16, s52
-; GCN-NEXT:    v_mov_b32_e32 v17, s53
-; GCN-NEXT:    v_mov_b32_e32 v18, s54
-; GCN-NEXT:    v_mov_b32_e32 v19, s55
-; GCN-NEXT:    v_mov_b32_e32 v20, s56
-; GCN-NEXT:    v_mov_b32_e32 v21, s57
-; GCN-NEXT:    v_mov_b32_e32 v22, s58
-; GCN-NEXT:    v_mov_b32_e32 v23, s59
-; GCN-NEXT:    v_mov_b32_e32 v24, s60
-; GCN-NEXT:    v_mov_b32_e32 v25, s61
-; GCN-NEXT:    v_mov_b32_e32 v26, s62
-; GCN-NEXT:    v_mov_b32_e32 v27, s63
-; GCN-NEXT:    v_mov_b32_e32 v28, s64
-; GCN-NEXT:    v_mov_b32_e32 v29, s65
-; GCN-NEXT:    v_mov_b32_e32 v30, s66
+; GCN-NEXT:    v_mov_b32_e32 v29, 0x402e0000
+; GCN-NEXT:    v_mov_b32_e32 v27, 0x402c0000
+; GCN-NEXT:    v_mov_b32_e32 v25, 0x402a0000
+; GCN-NEXT:    v_mov_b32_e32 v23, 0x40280000
+; GCN-NEXT:    v_mov_b32_e32 v21, 0x40260000
+; GCN-NEXT:    v_mov_b32_e32 v19, 0x40240000
+; GCN-NEXT:    v_mov_b32_e32 v17, 0x40220000
+; GCN-NEXT:    v_mov_b32_e32 v15, 0x40200000
+; GCN-NEXT:    v_mov_b32_e32 v13, 0x401c0000
+; GCN-NEXT:    v_mov_b32_e32 v11, 0x40180000
+; GCN-NEXT:    v_mov_b32_e32 v9, 0x40140000
+; GCN-NEXT:    v_mov_b32_e32 v7, 0x40100000
+; GCN-NEXT:    v_mov_b32_e32 v5, 0x40080000
+; GCN-NEXT:    v_mov_b32_e32 v3, 2.0
+; GCN-NEXT:    v_mov_b32_e32 v2, v0
+; GCN-NEXT:    v_mov_b32_e32 v4, v0
+; GCN-NEXT:    v_mov_b32_e32 v6, v0
+; GCN-NEXT:    v_mov_b32_e32 v8, v0
+; GCN-NEXT:    v_mov_b32_e32 v10, v0
+; GCN-NEXT:    v_mov_b32_e32 v12, v0
+; GCN-NEXT:    v_mov_b32_e32 v14, v0
+; GCN-NEXT:    v_mov_b32_e32 v16, v0
+; GCN-NEXT:    v_mov_b32_e32 v18, v0
+; GCN-NEXT:    v_mov_b32_e32 v20, v0
+; GCN-NEXT:    v_mov_b32_e32 v22, v0
+; GCN-NEXT:    v_mov_b32_e32 v24, v0
+; GCN-NEXT:    v_mov_b32_e32 v26, v0
+; GCN-NEXT:    v_mov_b32_e32 v28, v0
+; GCN-NEXT:    v_mov_b32_e32 v30, v0
 ; GCN-NEXT:    v_movrels_b32_e32 v32, v1
 ; GCN-NEXT:    v_movrels_b32_e32 v31, v0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s0
diff --git a/llvm/test/CodeGen/AMDGPU/fast-unaligned-load-store.global.ll b/llvm/test/CodeGen/AMDGPU/fast-unaligned-load-store.global.ll
index b42690eb3bf2b..cb27e0cdde70b 100644
--- a/llvm/test/CodeGen/AMDGPU/fast-unaligned-load-store.global.ll
+++ b/llvm/test/CodeGen/AMDGPU/fast-unaligned-load-store.global.ll
@@ -82,11 +82,11 @@ define amdgpu_kernel void @global_store_2xi16_align2(ptr addrspace(1) %p, ptr ad
 ; GFX7-ALIGNED-NEXT:    s_add_u32 s2, s0, 2
 ; GFX7-ALIGNED-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-ALIGNED-NEXT:    v_mov_b32_e32 v1, s1
-; GFX7-ALIGNED-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-ALIGNED-NEXT:    s_addc_u32 s0, s1, 0
 ; GFX7-ALIGNED-NEXT:    flat_store_short v[0:1], v2
 ; GFX7-ALIGNED-NEXT:    v_mov_b32_e32 v2, 2
 ; GFX7-ALIGNED-NEXT:    v_mov_b32_e32 v0, s2
-; GFX7-ALIGNED-NEXT:    v_mov_b32_e32 v1, s3
+; GFX7-ALIGNED-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX7-ALIGNED-NEXT:    flat_store_short v[0:1], v2
 ; GFX7-ALIGNED-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/fcopysign.f64.ll b/llvm/test/CodeGen/AMDGPU/fcopysign.f64.ll
index 819d4e289b4a9..6c005ad2354b5 100644
--- a/llvm/test/CodeGen/AMDGPU/fcopysign.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fcopysign.f64.ll
@@ -705,8 +705,8 @@ define amdgpu_kernel void @s_test_copysign_v3f64(ptr addrspace(1) %out, <3 x dou
 ; VI-NEXT:    s_add_u32 s2, s0, 16
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v4, s12
-; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    v_mov_b32_e32 v6, s2
+; VI-NEXT:    v_mov_b32_e32 v7, s3
 ; VI-NEXT:    flat_store_dwordx2 v[6:7], v[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v0, s8
 ; VI-NEXT:    v_mov_b32_e32 v2, s10
@@ -789,8 +789,8 @@ define amdgpu_kernel void @s_test_copysign_v4f64(ptr addrspace(1) %out, <4 x dou
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v4, s12
 ; VI-NEXT:    v_mov_b32_e32 v6, s14
-; VI-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-NEXT:    v_mov_b32_e32 v8, s2
+; VI-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-NEXT:    v_mov_b32_e32 v0, s8
 ; VI-NEXT:    v_mov_b32_e32 v2, s10
diff --git a/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll b/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
index 990b13ca76fb8..dd71d19d40493 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
@@ -981,11 +981,9 @@ define amdgpu_ps float @flat_load_i8_vgpr64_sgpr32(ptr %vbase, i32 inreg %soffse
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    s_mov_b32 s3, 0
-; GFX1250-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, 0
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX1250-GISEL-NEXT:    flat_load_u8 v0, v[0:1]
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
@@ -1032,11 +1030,9 @@ define amdgpu_ps float @flat_load_i8_vgpr64_sgpr32_offset_8388607(ptr %vbase, i3
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    s_mov_b32 s3, 0
-; GFX1250-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, 0
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX1250-GISEL-NEXT:    flat_load_u8 v0, v[0:1] offset:8388607
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
diff --git a/llvm/test/CodeGen/AMDGPU/flat-scratch.ll b/llvm/test/CodeGen/AMDGPU/flat-scratch.ll
index 0bb612c75d447..7eb77092997ec 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-scratch.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-scratch.ll
@@ -17,16 +17,13 @@
 define amdgpu_kernel void @zero_init_kernel() {
 ; GFX9-LABEL: zero_init_kernel:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s0, 0
 ; GFX9-NEXT:    s_add_u32 flat_scratch_lo, s8, s13
-; GFX9-NEXT:    s_mov_b32 s1, s0
-; GFX9-NEXT:    s_mov_b32 s2, s0
-; GFX9-NEXT:    s_mov_b32 s3, s0
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX9-NEXT:    s_addc_u32 flat_scratch_hi, s9, 0
-; GFX9-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-NEXT:    v_mov_b32_e32 v3, v0
+; GFX9-NEXT:    s_mov_b32 s0, 0
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:48
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:32
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:16
@@ -39,14 +36,10 @@ define amdgpu_kernel void @zero_init_kernel() {
 ; GFX10-NEXT:    s_addc_u32 s9, s9, 0
 ; GFX10-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s8
 ; GFX10-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s9
-; GFX10-NEXT:    s_mov_b32 s0, 0
-; GFX10-NEXT:    s_mov_b32 s1, s0
-; GFX10-NEXT:    s_mov_b32 s2, s0
-; GFX10-NEXT:    s_mov_b32 s3, s0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, v0
+; GFX10-NEXT:    v_mov_b32_e32 v2, v0
+; GFX10-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:48
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:32
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:16
@@ -55,13 +48,11 @@ define amdgpu_kernel void @zero_init_kernel() {
 ;
 ; GFX11-LABEL: zero_init_kernel:
 ; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_mov_b32 s0, 0
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    s_mov_b32 s1, s0
-; GFX11-NEXT:    s_mov_b32 s2, s0
-; GFX11-NEXT:    s_mov_b32 s3, s0
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-NEXT:    s_clause 0x3
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], off offset:48
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], off offset:32
@@ -71,13 +62,10 @@ define amdgpu_kernel void @zero_init_kernel() {
 ;
 ; GFX12-LABEL: zero_init_kernel:
 ; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_mov_b32 s0, 0
-; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX12-NEXT:    s_mov_b32 s1, s0
-; GFX12-NEXT:    s_mov_b32 s2, s0
-; GFX12-NEXT:    s_mov_b32 s3, s0
-; GFX12-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-NEXT:    s_clause 0x3
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], off offset:48
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], off offset:32
@@ -90,18 +78,15 @@ define amdgpu_kernel void @zero_init_kernel() {
 ; GFX9-PAL-NEXT:    s_getpc_b64 s[12:13]
 ; GFX9-PAL-NEXT:    s_mov_b32 s12, s0
 ; GFX9-PAL-NEXT:    s_load_dwordx2 s[12:13], s[12:13], 0x0
-; GFX9-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX9-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s3, s0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-PAL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-PAL-NEXT:    s_and_b32 s13, s13, 0xffff
 ; GFX9-PAL-NEXT:    s_add_u32 flat_scratch_lo, s12, s11
 ; GFX9-PAL-NEXT:    s_addc_u32 flat_scratch_hi, s13, 0
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-PAL-NEXT:    s_mov_b32 s0, 0
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:48
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:32
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:16
@@ -133,14 +118,11 @@ define amdgpu_kernel void @zero_init_kernel() {
 ; GFX1010-PAL-NEXT:    s_addc_u32 s13, s13, 0
 ; GFX1010-PAL-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s12
 ; GFX1010-PAL-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s13
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1010-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX1010-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX1010-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX1010-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX1010-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:48
 ; GFX1010-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:32
 ; GFX1010-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:16
@@ -158,14 +140,10 @@ define amdgpu_kernel void @zero_init_kernel() {
 ; GFX1030-PAL-NEXT:    s_addc_u32 s13, s13, 0
 ; GFX1030-PAL-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s12
 ; GFX1030-PAL-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s13
-; GFX1030-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX1030-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX1030-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX1030-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:48
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:32
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:16
@@ -174,13 +152,11 @@ define amdgpu_kernel void @zero_init_kernel() {
 ;
 ; GFX11-PAL-LABEL: zero_init_kernel:
 ; GFX11-PAL:       ; %bb.0:
-; GFX11-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX11-PAL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-PAL-NEXT:    s_clause 0x3
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:48
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:32
@@ -190,13 +166,10 @@ define amdgpu_kernel void @zero_init_kernel() {
 ;
 ; GFX12-PAL-LABEL: zero_init_kernel:
 ; GFX12-PAL:       ; %bb.0:
-; GFX12-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX12-PAL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX12-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-PAL-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-PAL-NEXT:    s_clause 0x3
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:48
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:32
@@ -212,14 +185,10 @@ define void @zero_init_foo() {
 ; GFX9-LABEL: zero_init_foo:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    s_mov_b32 s0, 0
-; GFX9-NEXT:    s_mov_b32 s1, s0
-; GFX9-NEXT:    s_mov_b32 s2, s0
-; GFX9-NEXT:    s_mov_b32 s3, s0
-; GFX9-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:48
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:32
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:16
@@ -230,14 +199,10 @@ define void @zero_init_foo() {
 ; GFX10-LABEL: zero_init_foo:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX10-NEXT:    s_mov_b32 s0, 0
-; GFX10-NEXT:    s_mov_b32 s1, s0
-; GFX10-NEXT:    s_mov_b32 s2, s0
-; GFX10-NEXT:    s_mov_b32 s3, s0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, v0
+; GFX10-NEXT:    v_mov_b32_e32 v2, v0
+; GFX10-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:48
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:32
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:16
@@ -247,13 +212,11 @@ define void @zero_init_foo() {
 ; GFX11-LABEL: zero_init_foo:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT:    s_mov_b32 s0, 0
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    s_mov_b32 s1, s0
-; GFX11-NEXT:    s_mov_b32 s2, s0
-; GFX11-NEXT:    s_mov_b32 s3, s0
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-NEXT:    s_clause 0x3
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:48
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:32
@@ -268,14 +231,10 @@ define void @zero_init_foo() {
 ; GFX12-NEXT:    s_wait_samplecnt 0x0
 ; GFX12-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    s_mov_b32 s0, 0
-; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    s_mov_b32 s1, s0
-; GFX12-NEXT:    s_mov_b32 s2, s0
-; GFX12-NEXT:    s_mov_b32 s3, s0
-; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-NEXT:    s_clause 0x3
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:48
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:32
@@ -286,14 +245,10 @@ define void @zero_init_foo() {
 ; GFX9-PAL-LABEL: zero_init_foo:
 ; GFX9-PAL:       ; %bb.0:
 ; GFX9-PAL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX9-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:48
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:32
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:16
@@ -320,14 +275,10 @@ define void @zero_init_foo() {
 ; GFX10-PAL-LABEL: zero_init_foo:
 ; GFX10-PAL:       ; %bb.0:
 ; GFX10-PAL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX10-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX10-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX10-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX10-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX10-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX10-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX10-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX10-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:48
 ; GFX10-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:32
 ; GFX10-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:16
@@ -337,13 +288,11 @@ define void @zero_init_foo() {
 ; GFX11-PAL-LABEL: zero_init_foo:
 ; GFX11-PAL:       ; %bb.0:
 ; GFX11-PAL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX11-PAL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-PAL-NEXT:    s_clause 0x3
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:48
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:32
@@ -358,14 +307,10 @@ define void @zero_init_foo() {
 ; GFX12-PAL-NEXT:    s_wait_samplecnt 0x0
 ; GFX12-PAL-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-PAL-NEXT:    s_wait_kmcnt 0x0
-; GFX12-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX12-PAL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX12-PAL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-PAL-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-PAL-NEXT:    s_clause 0x3
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:48
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:32
@@ -1080,13 +1025,10 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX9-NEXT:    s_mov_b32 s0, 0
 ; GFX9-NEXT:    scratch_load_dword v0, off, s0 glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    s_mov_b32 s1, s0
-; GFX9-NEXT:    s_mov_b32 s2, s0
-; GFX9-NEXT:    s_mov_b32 s3, s0
-; GFX9-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:256
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:272
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:288
@@ -1101,14 +1043,10 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX10-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s9
 ; GFX10-NEXT:    scratch_load_dword v0, off, off glc dlc
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    s_mov_b32 s0, 0
-; GFX10-NEXT:    s_mov_b32 s1, s0
-; GFX10-NEXT:    s_mov_b32 s2, s0
-; GFX10-NEXT:    s_mov_b32 s3, s0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, v0
+; GFX10-NEXT:    v_mov_b32_e32 v2, v0
+; GFX10-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:256
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:272
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:288
@@ -1119,13 +1057,11 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    scratch_load_b32 v0, off, off glc dlc
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-NEXT:    s_mov_b32 s0, 0
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    s_mov_b32 s1, s0
-; GFX11-NEXT:    s_mov_b32 s2, s0
-; GFX11-NEXT:    s_mov_b32 s3, s0
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-NEXT:    s_clause 0x3
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], off offset:256
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], off offset:272
@@ -1137,13 +1073,10 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    scratch_load_b32 v0, off, off scope:SCOPE_SYS
 ; GFX12-NEXT:    s_wait_loadcnt 0x0
-; GFX12-NEXT:    s_mov_b32 s0, 0
-; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX12-NEXT:    s_mov_b32 s1, s0
-; GFX12-NEXT:    s_mov_b32 s2, s0
-; GFX12-NEXT:    s_mov_b32 s3, s0
-; GFX12-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-NEXT:    s_clause 0x3
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], off offset:256
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], off offset:272
@@ -1157,19 +1090,16 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX9-PAL-NEXT:    s_mov_b32 s12, s0
 ; GFX9-PAL-NEXT:    s_load_dwordx2 s[12:13], s[12:13], 0x0
 ; GFX9-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX9-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s3, s0
 ; GFX9-PAL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-PAL-NEXT:    s_and_b32 s13, s13, 0xffff
 ; GFX9-PAL-NEXT:    s_add_u32 flat_scratch_lo, s12, s11
 ; GFX9-PAL-NEXT:    s_addc_u32 flat_scratch_hi, s13, 0
 ; GFX9-PAL-NEXT:    scratch_load_dword v0, off, s0 glc
 ; GFX9-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:256
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:272
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:288
@@ -1206,13 +1136,10 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX1010-PAL-NEXT:    s_mov_b32 s0, 0
 ; GFX1010-PAL-NEXT:    scratch_load_dword v0, off, s0 glc dlc
 ; GFX1010-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX1010-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX1010-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX1010-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX1010-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:256
 ; GFX1010-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:272
 ; GFX1010-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:288
@@ -1232,14 +1159,10 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX1030-PAL-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s13
 ; GFX1030-PAL-NEXT:    scratch_load_dword v0, off, off glc dlc
 ; GFX1030-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX1030-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX1030-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX1030-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX1030-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:256
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:272
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], off offset:288
@@ -1250,13 +1173,11 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX11-PAL:       ; %bb.0:
 ; GFX11-PAL-NEXT:    scratch_load_b32 v0, off, off glc dlc
 ; GFX11-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX11-PAL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-PAL-NEXT:    s_clause 0x3
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:256
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:272
@@ -1268,13 +1189,10 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX12-PAL:       ; %bb.0:
 ; GFX12-PAL-NEXT:    scratch_load_b32 v0, off, off scope:SCOPE_SYS
 ; GFX12-PAL-NEXT:    s_wait_loadcnt 0x0
-; GFX12-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX12-PAL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX12-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-PAL-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-PAL-NEXT:    s_clause 0x3
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:256
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:272
@@ -1295,14 +1213,10 @@ define void @zero_init_small_offset_foo() {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    scratch_load_dword v0, off, s32 glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    s_mov_b32 s0, 0
-; GFX9-NEXT:    s_mov_b32 s1, s0
-; GFX9-NEXT:    s_mov_b32 s2, s0
-; GFX9-NEXT:    s_mov_b32 s3, s0
-; GFX9-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:256
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:272
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:288
@@ -1315,14 +1229,10 @@ define void @zero_init_small_offset_foo() {
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    scratch_load_dword v0, off, s32 glc dlc
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    s_mov_b32 s0, 0
-; GFX10-NEXT:    s_mov_b32 s1, s0
-; GFX10-NEXT:    s_mov_b32 s2, s0
-; GFX10-NEXT:    s_mov_b32 s3, s0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, v0
+; GFX10-NEXT:    v_mov_b32_e32 v2, v0
+; GFX10-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:256
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:272
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:288
@@ -1334,13 +1244,11 @@ define void @zero_init_small_offset_foo() {
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    scratch_load_b32 v0, off, s32 glc dlc
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-NEXT:    s_mov_b32 s0, 0
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    s_mov_b32 s1, s0
-; GFX11-NEXT:    s_mov_b32 s2, s0
-; GFX11-NEXT:    s_mov_b32 s3, s0
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-NEXT:    s_clause 0x3
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:256
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:272
@@ -1357,14 +1265,10 @@ define void @zero_init_small_offset_foo() {
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-NEXT:    scratch_load_b32 v0, off, s32 scope:SCOPE_SYS
 ; GFX12-NEXT:    s_wait_loadcnt 0x0
-; GFX12-NEXT:    s_mov_b32 s0, 0
-; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    s_mov_b32 s1, s0
-; GFX12-NEXT:    s_mov_b32 s2, s0
-; GFX12-NEXT:    s_mov_b32 s3, s0
-; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-NEXT:    s_clause 0x3
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:256
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:272
@@ -1377,14 +1281,10 @@ define void @zero_init_small_offset_foo() {
 ; GFX9-PAL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-PAL-NEXT:    scratch_load_dword v0, off, s32 glc
 ; GFX9-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX9-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:256
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:272
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:288
@@ -1415,14 +1315,10 @@ define void @zero_init_small_offset_foo() {
 ; GFX10-PAL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-PAL-NEXT:    scratch_load_dword v0, off, s32 glc dlc
 ; GFX10-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX10-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX10-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX10-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX10-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX10-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX10-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX10-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:256
 ; GFX10-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:272
 ; GFX10-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s32 offset:288
@@ -1434,13 +1330,11 @@ define void @zero_init_small_offset_foo() {
 ; GFX11-PAL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-PAL-NEXT:    scratch_load_b32 v0, off, s32 glc dlc
 ; GFX11-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX11-PAL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-PAL-NEXT:    s_clause 0x3
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:256
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:272
@@ -1457,14 +1351,10 @@ define void @zero_init_small_offset_foo() {
 ; GFX12-PAL-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-PAL-NEXT:    scratch_load_b32 v0, off, s32 scope:SCOPE_SYS
 ; GFX12-PAL-NEXT:    s_wait_loadcnt 0x0
-; GFX12-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX12-PAL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX12-PAL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-PAL-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-PAL-NEXT:    s_clause 0x3
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:256
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:272
@@ -2316,13 +2206,10 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX9-NEXT:    s_mov_b32 s0, 0
 ; GFX9-NEXT:    scratch_load_dword v0, off, s0 offset:4 glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    s_mov_b32 s1, s0
-; GFX9-NEXT:    s_mov_b32 s2, s0
-; GFX9-NEXT:    s_mov_b32 s3, s0
-; GFX9-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    s_movk_i32 s0, 0x4004
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s0
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:16
@@ -2338,15 +2225,11 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX10-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s9
 ; GFX10-NEXT:    scratch_load_dword v0, off, off offset:4 glc dlc
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    s_mov_b32 s0, 0
-; GFX10-NEXT:    s_mov_b32 s1, s0
-; GFX10-NEXT:    s_mov_b32 s2, s0
-; GFX10-NEXT:    s_mov_b32 s3, s0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-NEXT:    s_movk_i32 s0, 0x4004
+; GFX10-NEXT:    v_mov_b32_e32 v1, v0
+; GFX10-NEXT:    v_mov_b32_e32 v2, v0
+; GFX10-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s0
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:16
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:32
@@ -2357,14 +2240,12 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    scratch_load_b32 v0, off, off offset:4 glc dlc
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-NEXT:    s_mov_b32 s0, 0
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    s_mov_b32 s1, s0
-; GFX11-NEXT:    s_mov_b32 s2, s0
-; GFX11-NEXT:    s_mov_b32 s3, s0
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-NEXT:    s_movk_i32 s0, 0x4004
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-NEXT:    s_clause 0x3
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s0
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s0 offset:16
@@ -2376,13 +2257,10 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    scratch_load_b32 v0, off, off scope:SCOPE_SYS
 ; GFX12-NEXT:    s_wait_loadcnt 0x0
-; GFX12-NEXT:    s_mov_b32 s0, 0
-; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX12-NEXT:    s_mov_b32 s1, s0
-; GFX12-NEXT:    s_mov_b32 s2, s0
-; GFX12-NEXT:    s_mov_b32 s3, s0
-; GFX12-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-NEXT:    s_clause 0x3
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], off offset:16384
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], off offset:16400
@@ -2396,19 +2274,16 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX9-PAL-NEXT:    s_mov_b32 s12, s0
 ; GFX9-PAL-NEXT:    s_load_dwordx2 s[12:13], s[12:13], 0x0
 ; GFX9-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX9-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s3, s0
 ; GFX9-PAL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-PAL-NEXT:    s_and_b32 s13, s13, 0xffff
 ; GFX9-PAL-NEXT:    s_add_u32 flat_scratch_lo, s12, s11
 ; GFX9-PAL-NEXT:    s_addc_u32 flat_scratch_hi, s13, 0
 ; GFX9-PAL-NEXT:    scratch_load_dword v0, off, s0 offset:4 glc
 ; GFX9-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-PAL-NEXT:    s_movk_i32 s0, 0x4004
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:16
@@ -2447,14 +2322,11 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX1010-PAL-NEXT:    s_mov_b32 s0, 0
 ; GFX1010-PAL-NEXT:    scratch_load_dword v0, off, s0 offset:4 glc dlc
 ; GFX1010-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX1010-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX1010-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX1010-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1010-PAL-NEXT:    s_movk_i32 s0, 0x4004
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX1010-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0
 ; GFX1010-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:16
 ; GFX1010-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:32
@@ -2474,15 +2346,11 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX1030-PAL-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s13
 ; GFX1030-PAL-NEXT:    scratch_load_dword v0, off, off offset:4 glc dlc
 ; GFX1030-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX1030-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX1030-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX1030-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX1030-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1030-PAL-NEXT:    s_movk_i32 s0, 0x4004
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:16
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:32
@@ -2493,14 +2361,12 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX11-PAL:       ; %bb.0:
 ; GFX11-PAL-NEXT:    scratch_load_b32 v0, off, off offset:4 glc dlc
 ; GFX11-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX11-PAL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-PAL-NEXT:    s_movk_i32 s0, 0x4004
+; GFX11-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-PAL-NEXT:    s_clause 0x3
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], s0
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], s0 offset:16
@@ -2512,13 +2378,10 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX12-PAL:       ; %bb.0:
 ; GFX12-PAL-NEXT:    scratch_load_b32 v0, off, off scope:SCOPE_SYS
 ; GFX12-PAL-NEXT:    s_wait_loadcnt 0x0
-; GFX12-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX12-PAL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX12-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-PAL-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-PAL-NEXT:    s_clause 0x3
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:16384
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:16400
@@ -2539,14 +2402,10 @@ define void @zero_init_large_offset_foo() {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    scratch_load_dword v0, off, s32 offset:4 glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    s_mov_b32 s0, 0
-; GFX9-NEXT:    s_mov_b32 s1, s0
-; GFX9-NEXT:    s_mov_b32 s2, s0
-; GFX9-NEXT:    s_mov_b32 s3, s0
-; GFX9-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    s_add_i32 s0, s32, 0x4004
 ; GFX9-NEXT:    scratch_store_dwordx4 off, v[0:3], s0
 ; GFX9-NEXT:    s_add_i32 s0, s32, 0x4004
@@ -2563,15 +2422,11 @@ define void @zero_init_large_offset_foo() {
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    scratch_load_dword v0, off, s32 offset:4 glc dlc
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    s_mov_b32 s0, 0
-; GFX10-NEXT:    s_mov_b32 s1, s0
-; GFX10-NEXT:    s_mov_b32 s2, s0
-; GFX10-NEXT:    s_mov_b32 s3, s0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-NEXT:    s_add_i32 s0, s32, 0x4004
+; GFX10-NEXT:    v_mov_b32_e32 v1, v0
+; GFX10-NEXT:    v_mov_b32_e32 v2, v0
+; GFX10-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s0
 ; GFX10-NEXT:    s_add_i32 s0, s32, 0x4004
 ; GFX10-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:16
@@ -2586,14 +2441,12 @@ define void @zero_init_large_offset_foo() {
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    scratch_load_b32 v0, off, s32 offset:4 glc dlc
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-NEXT:    s_mov_b32 s0, 0
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    s_mov_b32 s1, s0
-; GFX11-NEXT:    s_mov_b32 s2, s0
-; GFX11-NEXT:    s_mov_b32 s3, s0
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-NEXT:    s_add_i32 s0, s32, 0x4004
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s0
 ; GFX11-NEXT:    s_add_i32 s0, s32, 0x4004
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s0 offset:16
@@ -2612,14 +2465,10 @@ define void @zero_init_large_offset_foo() {
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-NEXT:    scratch_load_b32 v0, off, s32 scope:SCOPE_SYS
 ; GFX12-NEXT:    s_wait_loadcnt 0x0
-; GFX12-NEXT:    s_mov_b32 s0, 0
-; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    s_mov_b32 s1, s0
-; GFX12-NEXT:    s_mov_b32 s2, s0
-; GFX12-NEXT:    s_mov_b32 s3, s0
-; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-NEXT:    s_clause 0x3
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:16384
 ; GFX12-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:16400
@@ -2632,14 +2481,10 @@ define void @zero_init_large_offset_foo() {
 ; GFX9-PAL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-PAL-NEXT:    scratch_load_dword v0, off, s32 offset:4 glc
 ; GFX9-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX9-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX9-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX9-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-PAL-NEXT:    s_add_i32 s0, s32, 0x4004
 ; GFX9-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0
 ; GFX9-PAL-NEXT:    s_add_i32 s0, s32, 0x4004
@@ -2678,15 +2523,11 @@ define void @zero_init_large_offset_foo() {
 ; GFX1010-PAL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX1010-PAL-NEXT:    scratch_load_dword v0, off, s32 offset:4 glc dlc
 ; GFX1010-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX1010-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX1010-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX1010-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX1010-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX1010-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1010-PAL-NEXT:    s_add_i32 s0, s32, 0x4004
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX1010-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX1010-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0
 ; GFX1010-PAL-NEXT:    s_waitcnt_depctr depctr_vm_vsrc(0)
 ; GFX1010-PAL-NEXT:    s_add_i32 s0, s32, 0x4004
@@ -2704,15 +2545,11 @@ define void @zero_init_large_offset_foo() {
 ; GFX1030-PAL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX1030-PAL-NEXT:    scratch_load_dword v0, off, s32 offset:4 glc dlc
 ; GFX1030-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX1030-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX1030-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX1030-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX1030-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX1030-PAL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1030-PAL-NEXT:    s_add_i32 s0, s32, 0x4004
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX1030-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0
 ; GFX1030-PAL-NEXT:    s_add_i32 s0, s32, 0x4004
 ; GFX1030-PAL-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:16
@@ -2727,14 +2564,12 @@ define void @zero_init_large_offset_foo() {
 ; GFX11-PAL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-PAL-NEXT:    scratch_load_b32 v0, off, s32 offset:4 glc dlc
 ; GFX11-PAL-NEXT:    s_waitcnt vmcnt(0)
-; GFX11-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX11-PAL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX11-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-PAL-NEXT:    s_add_i32 s0, s32, 0x4004
+; GFX11-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], s0
 ; GFX11-PAL-NEXT:    s_add_i32 s0, s32, 0x4004
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], s0 offset:16
@@ -2753,14 +2588,10 @@ define void @zero_init_large_offset_foo() {
 ; GFX12-PAL-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-PAL-NEXT:    scratch_load_b32 v0, off, s32 scope:SCOPE_SYS
 ; GFX12-PAL-NEXT:    s_wait_loadcnt 0x0
-; GFX12-PAL-NEXT:    s_mov_b32 s0, 0
-; GFX12-PAL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-PAL-NEXT:    s_mov_b32 s1, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s2, s0
-; GFX12-PAL-NEXT:    s_mov_b32 s3, s0
-; GFX12-PAL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-PAL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-PAL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-PAL-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v2, v0
+; GFX12-PAL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX12-PAL-NEXT:    s_clause 0x3
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:16384
 ; GFX12-PAL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:16400
diff --git a/llvm/test/CodeGen/AMDGPU/flat_atomics.ll b/llvm/test/CodeGen/AMDGPU/flat_atomics.ll
index 232a170b7b2b3..ae9e7aab08465 100644
--- a/llvm/test/CodeGen/AMDGPU/flat_atomics.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat_atomics.ll
@@ -5751,8 +5751,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64_offset(ptr %out, i32 %in, i
 ; GCN1-NEXT:    s_add_u32 s0, s0, 16
 ; GCN1-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN1-NEXT:    v_mov_b32_e32 v1, s4
-; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v2, s0
+; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    flat_atomic_cmpswap v[2:3], v[0:1]
 ; GCN1-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN1-NEXT:    buffer_wbinvl1_vol
@@ -5772,8 +5772,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64_offset(ptr %out, i32 %in, i
 ; GCN2-NEXT:    s_add_u32 s0, s0, 16
 ; GCN2-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN2-NEXT:    v_mov_b32_e32 v1, s4
-; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    flat_atomic_cmpswap v[2:3], v[0:1]
 ; GCN2-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN2-NEXT:    buffer_wbinvl1_vol
@@ -5791,8 +5791,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64_offset(ptr %out, i32 %in, i
 ; GCN3-NEXT:    s_add_u32 s0, s2, s0
 ; GCN3-NEXT:    s_addc_u32 s1, s3, s1
 ; GCN3-NEXT:    v_mov_b32_e32 v1, s7
-; GCN3-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN3-NEXT:    v_mov_b32_e32 v2, s0
+; GCN3-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN3-NEXT:    flat_atomic_cmpswap v[2:3], v[0:1] offset:16
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
@@ -5811,7 +5811,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64_offset(ptr %out, i32 %in, i
 ; GFX11-NEXT:    s_add_u32 s0, s2, s0
 ; GFX11-NEXT:    s_addc_u32 s1, s3, s1
 ; GFX11-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s4
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX11-NEXT:    flat_atomic_cmpswap_b32 v[2:3], v[0:1] offset:16
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
@@ -5840,8 +5840,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64_offset(ptr %out, ptr %o
 ; GCN1-NEXT:    s_add_u32 s0, s0, 16
 ; GCN1-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN1-NEXT:    v_mov_b32_e32 v1, s9
-; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v2, s0
+; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    flat_atomic_cmpswap v2, v[2:3], v[0:1] glc
 ; GCN1-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN1-NEXT:    buffer_wbinvl1_vol
@@ -5864,8 +5864,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64_offset(ptr %out, ptr %o
 ; GCN2-NEXT:    s_add_u32 s0, s0, 16
 ; GCN2-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN2-NEXT:    v_mov_b32_e32 v1, s9
-; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    flat_atomic_cmpswap v2, v[2:3], v[0:1] glc
 ; GCN2-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN2-NEXT:    buffer_wbinvl1_vol
@@ -5886,8 +5886,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64_offset(ptr %out, ptr %o
 ; GCN3-NEXT:    s_add_u32 s0, s0, s4
 ; GCN3-NEXT:    s_addc_u32 s1, s1, s5
 ; GCN3-NEXT:    v_mov_b32_e32 v1, s9
-; GCN3-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN3-NEXT:    v_mov_b32_e32 v2, s0
+; GCN3-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN3-NEXT:    flat_atomic_cmpswap v2, v[2:3], v[0:1] offset:16 glc
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
@@ -5909,7 +5909,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64_offset(ptr %out, ptr %o
 ; GFX11-NEXT:    s_add_u32 s0, s0, s4
 ; GFX11-NEXT:    s_addc_u32 s1, s1, s5
 ; GFX11-NEXT:    v_dual_mov_b32 v0, s8 :: v_dual_mov_b32 v1, s9
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX11-NEXT:    flat_atomic_cmpswap_b32 v2, v[2:3], v[0:1] offset:16 glc
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    buffer_gl1_inv
@@ -6070,8 +6070,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64(ptr %out, i32 %in, i64 %ind
 ; GCN1-NEXT:    s_add_u32 s0, s2, s0
 ; GCN1-NEXT:    s_addc_u32 s1, s3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v1, s4
-; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v2, s0
+; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    flat_atomic_cmpswap v[2:3], v[0:1]
 ; GCN1-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN1-NEXT:    buffer_wbinvl1_vol
@@ -6089,8 +6089,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64(ptr %out, i32 %in, i64 %ind
 ; GCN2-NEXT:    s_add_u32 s0, s2, s0
 ; GCN2-NEXT:    s_addc_u32 s1, s3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v1, s4
-; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    flat_atomic_cmpswap v[2:3], v[0:1]
 ; GCN2-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN2-NEXT:    buffer_wbinvl1_vol
@@ -6108,8 +6108,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64(ptr %out, i32 %in, i64 %ind
 ; GCN3-NEXT:    s_add_u32 s0, s2, s0
 ; GCN3-NEXT:    s_addc_u32 s1, s3, s1
 ; GCN3-NEXT:    v_mov_b32_e32 v1, s7
-; GCN3-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN3-NEXT:    v_mov_b32_e32 v2, s0
+; GCN3-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN3-NEXT:    flat_atomic_cmpswap v[2:3], v[0:1]
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
@@ -6128,7 +6128,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64(ptr %out, i32 %in, i64 %ind
 ; GFX11-NEXT:    s_add_u32 s0, s2, s0
 ; GFX11-NEXT:    s_addc_u32 s1, s3, s1
 ; GFX11-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s4
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX11-NEXT:    flat_atomic_cmpswap_b32 v[2:3], v[0:1]
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
@@ -6154,8 +6154,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64(ptr %out, ptr %out2, i3
 ; GCN1-NEXT:    s_add_u32 s0, s0, s4
 ; GCN1-NEXT:    s_addc_u32 s1, s1, s5
 ; GCN1-NEXT:    v_mov_b32_e32 v1, s9
-; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v2, s0
+; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    flat_atomic_cmpswap v2, v[2:3], v[0:1] glc
 ; GCN1-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN1-NEXT:    buffer_wbinvl1_vol
@@ -6176,8 +6176,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64(ptr %out, ptr %out2, i3
 ; GCN2-NEXT:    s_add_u32 s0, s0, s4
 ; GCN2-NEXT:    s_addc_u32 s1, s1, s5
 ; GCN2-NEXT:    v_mov_b32_e32 v1, s9
-; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    flat_atomic_cmpswap v2, v[2:3], v[0:1] glc
 ; GCN2-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN2-NEXT:    buffer_wbinvl1_vol
@@ -6198,8 +6198,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64(ptr %out, ptr %out2, i3
 ; GCN3-NEXT:    s_add_u32 s0, s0, s4
 ; GCN3-NEXT:    s_addc_u32 s1, s1, s5
 ; GCN3-NEXT:    v_mov_b32_e32 v1, s9
-; GCN3-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN3-NEXT:    v_mov_b32_e32 v2, s0
+; GCN3-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN3-NEXT:    flat_atomic_cmpswap v2, v[2:3], v[0:1] glc
 ; GCN3-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN3-NEXT:    buffer_wbinvl1_vol
@@ -6221,7 +6221,7 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64(ptr %out, ptr %out2, i3
 ; GFX11-NEXT:    s_add_u32 s0, s0, s4
 ; GFX11-NEXT:    s_addc_u32 s1, s1, s5
 ; GFX11-NEXT:    v_dual_mov_b32 v0, s8 :: v_dual_mov_b32 v1, s9
-; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
 ; GFX11-NEXT:    flat_atomic_cmpswap_b32 v2, v[2:3], v[0:1] glc
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    buffer_gl1_inv
@@ -8329,8 +8329,9 @@ define amdgpu_kernel void @atomic_store_i8_addr64_offset(i8 %in, ptr %out, i64 %
 ; GFX11-TRUE16-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-TRUE16-NEXT:    s_add_u32 s0, s0, s2
 ; GFX11-TRUE16-NEXT:    s_addc_u32 s1, s1, s3
+; GFX11-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
 ; GFX11-TRUE16-NEXT:    v_mov_b16_e32 v0.l, s4
-; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
 ; GFX11-TRUE16-NEXT:    flat_store_b8 v[1:2], v0 offset:16
 ; GFX11-TRUE16-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/flat_atomics_i64.ll b/llvm/test/CodeGen/AMDGPU/flat_atomics_i64.ll
index 5559c3e84e9b7..186b1b783a9f8 100644
--- a/llvm/test/CodeGen/AMDGPU/flat_atomics_i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat_atomics_i64.ll
@@ -12944,8 +12944,8 @@ define amdgpu_kernel void @atomic_store_i64_offset(i64 %in, ptr %out) {
 ; GCN1-NEXT:    s_add_u32 s0, s2, 32
 ; GCN1-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN1-NEXT:    s_addc_u32 s1, s3, 0
-; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v2, s0
+; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN1-NEXT:    s_endpgm
 ;
@@ -12957,8 +12957,8 @@ define amdgpu_kernel void @atomic_store_i64_offset(i64 %in, ptr %out) {
 ; GCN2-NEXT:    s_add_u32 s0, s2, 32
 ; GCN2-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN2-NEXT:    s_addc_u32 s1, s3, 0
-; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN2-NEXT:    s_endpgm
 ;
@@ -13029,8 +13029,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64_offset(i64 %in, ptr %out, i64
 ; GCN1-NEXT:    s_addc_u32 s1, s3, s1
 ; GCN1-NEXT:    s_add_u32 s0, s0, 32
 ; GCN1-NEXT:    s_addc_u32 s1, s1, 0
-; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v2, s0
+; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN1-NEXT:    s_endpgm
 ;
@@ -13046,8 +13046,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64_offset(i64 %in, ptr %out, i64
 ; GCN2-NEXT:    s_addc_u32 s1, s3, s1
 ; GCN2-NEXT:    s_add_u32 s0, s0, 32
 ; GCN2-NEXT:    s_addc_u32 s1, s1, 0
-; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN2-NEXT:    s_endpgm
 ;
@@ -13084,8 +13084,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64(i64 %in, ptr %out, i64 %index
 ; GCN1-NEXT:    s_lshl_b64 s[0:1], s[4:5], 3
 ; GCN1-NEXT:    s_add_u32 s0, s2, s0
 ; GCN1-NEXT:    s_addc_u32 s1, s3, s1
-; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v2, s0
+; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN1-NEXT:    s_endpgm
 ;
@@ -13099,8 +13099,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64(i64 %in, ptr %out, i64 %index
 ; GCN2-NEXT:    s_lshl_b64 s[0:1], s[4:5], 3
 ; GCN2-NEXT:    s_add_u32 s0, s2, s0
 ; GCN2-NEXT:    s_addc_u32 s1, s3, s1
-; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN2-NEXT:    s_endpgm
 ;
@@ -14827,8 +14827,8 @@ define amdgpu_kernel void @atomic_store_f64_offset(double %in, ptr %out) {
 ; GCN1-NEXT:    s_add_u32 s0, s2, 32
 ; GCN1-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN1-NEXT:    s_addc_u32 s1, s3, 0
-; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v2, s0
+; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN1-NEXT:    s_endpgm
 ;
@@ -14840,8 +14840,8 @@ define amdgpu_kernel void @atomic_store_f64_offset(double %in, ptr %out) {
 ; GCN2-NEXT:    s_add_u32 s0, s2, 32
 ; GCN2-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN2-NEXT:    s_addc_u32 s1, s3, 0
-; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN2-NEXT:    s_endpgm
 ;
@@ -14912,8 +14912,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64_offset(double %in, ptr %out,
 ; GCN1-NEXT:    s_addc_u32 s1, s3, s1
 ; GCN1-NEXT:    s_add_u32 s0, s0, 32
 ; GCN1-NEXT:    s_addc_u32 s1, s1, 0
-; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v2, s0
+; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN1-NEXT:    s_endpgm
 ;
@@ -14929,8 +14929,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64_offset(double %in, ptr %out,
 ; GCN2-NEXT:    s_addc_u32 s1, s3, s1
 ; GCN2-NEXT:    s_add_u32 s0, s0, 32
 ; GCN2-NEXT:    s_addc_u32 s1, s1, 0
-; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN2-NEXT:    s_endpgm
 ;
@@ -14967,8 +14967,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64(double %in, ptr %out, i64 %in
 ; GCN1-NEXT:    s_lshl_b64 s[0:1], s[4:5], 3
 ; GCN1-NEXT:    s_add_u32 s0, s2, s0
 ; GCN1-NEXT:    s_addc_u32 s1, s3, s1
-; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    v_mov_b32_e32 v2, s0
+; GCN1-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN1-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN1-NEXT:    s_endpgm
 ;
@@ -14982,8 +14982,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64(double %in, ptr %out, i64 %in
 ; GCN2-NEXT:    s_lshl_b64 s[0:1], s[4:5], 3
 ; GCN2-NEXT:    s_add_u32 s0, s2, s0
 ; GCN2-NEXT:    s_addc_u32 s1, s3, s1
-; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN2-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN2-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_noprivate.ll b/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_noprivate.ll
index 0d8f57a9e42bf..2104c15af9b1e 100644
--- a/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_noprivate.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat_atomics_i64_noprivate.ll
@@ -12,8 +12,8 @@ define amdgpu_kernel void @atomic_add_i64_offset(ptr %out, i64 %in) {
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s3
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_add_x2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -27,8 +27,8 @@ define amdgpu_kernel void @atomic_add_i64_offset(ptr %out, i64 %in) {
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_add_x2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -60,8 +60,8 @@ define amdgpu_kernel void @atomic_add_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX7-NEXT:    s_add_u32 s0, s0, 32
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s7
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_add_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -79,8 +79,8 @@ define amdgpu_kernel void @atomic_add_i64_ret_offset(ptr %out, ptr %out2, i64 %i
 ; GFX8-NEXT:    s_add_u32 s0, s0, 32
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s7
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_add_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -123,8 +123,8 @@ define amdgpu_kernel void @atomic_add_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX7-NEXT:    s_addc_u32 s1, s1, s3
 ; GFX7-NEXT:    s_add_u32 s0, s0, 32
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_add_x2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -142,8 +142,8 @@ define amdgpu_kernel void @atomic_add_i64_addr64_offset(ptr %out, i64 %in, i64 %
 ; GFX8-NEXT:    s_addc_u32 s1, s1, s3
 ; GFX8-NEXT:    s_add_u32 s0, s0, 32
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_add_x2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -183,8 +183,8 @@ define amdgpu_kernel void @atomic_add_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX7-NEXT:    s_addc_u32 s1, s1, s5
 ; GFX7-NEXT:    s_add_u32 s0, s0, 32
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_add_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -204,8 +204,8 @@ define amdgpu_kernel void @atomic_add_i64_ret_addr64_offset(ptr %out, ptr %out2,
 ; GFX8-NEXT:    s_addc_u32 s1, s1, s5
 ; GFX8-NEXT:    s_add_u32 s0, s0, 32
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_add_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -345,8 +345,8 @@ define amdgpu_kernel void @atomic_add_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX7-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; GFX7-NEXT:    s_add_u32 s0, s0, s2
 ; GFX7-NEXT:    s_addc_u32 s1, s1, s3
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_add_x2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -362,8 +362,8 @@ define amdgpu_kernel void @atomic_add_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX8-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; GFX8-NEXT:    s_add_u32 s0, s0, s2
 ; GFX8-NEXT:    s_addc_u32 s1, s1, s3
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_add_x2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -400,8 +400,8 @@ define amdgpu_kernel void @atomic_add_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX7-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; GFX7-NEXT:    s_add_u32 s0, s0, s4
 ; GFX7-NEXT:    s_addc_u32 s1, s1, s5
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_add_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -419,8 +419,8 @@ define amdgpu_kernel void @atomic_add_i64_ret_addr64(ptr %out, ptr %out2, i64 %i
 ; GFX8-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; GFX8-NEXT:    s_add_u32 s0, s0, s4
 ; GFX8-NEXT:    s_addc_u32 s1, s1, s5
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_add_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -5476,8 +5476,8 @@ define amdgpu_kernel void @atomic_xchg_i64_offset(ptr %out, i64 %in) {
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s3
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -5491,8 +5491,8 @@ define amdgpu_kernel void @atomic_xchg_i64_offset(ptr %out, i64 %in) {
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -5523,8 +5523,8 @@ define amdgpu_kernel void @atomic_xchg_f64_offset(ptr %out, double %in) {
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s3
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -5538,8 +5538,8 @@ define amdgpu_kernel void @atomic_xchg_f64_offset(ptr %out, double %in) {
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -5570,8 +5570,8 @@ define amdgpu_kernel void @atomic_xchg_pointer_offset(ptr %out, ptr %in) {
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s3
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -5585,8 +5585,8 @@ define amdgpu_kernel void @atomic_xchg_pointer_offset(ptr %out, ptr %in) {
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -5618,8 +5618,8 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_offset(ptr %out, ptr %out2, i64 %
 ; GFX7-NEXT:    s_add_u32 s0, s0, 32
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s7
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_swap_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -5637,8 +5637,8 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_offset(ptr %out, ptr %out2, i64 %
 ; GFX8-NEXT:    s_add_u32 s0, s0, 32
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s7
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_swap_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -5681,8 +5681,8 @@ define amdgpu_kernel void @atomic_xchg_i64_addr64_offset(ptr %out, i64 %in, i64
 ; GFX7-NEXT:    s_addc_u32 s1, s1, s3
 ; GFX7-NEXT:    s_add_u32 s0, s0, 32
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -5700,8 +5700,8 @@ define amdgpu_kernel void @atomic_xchg_i64_addr64_offset(ptr %out, i64 %in, i64
 ; GFX8-NEXT:    s_addc_u32 s1, s1, s3
 ; GFX8-NEXT:    s_add_u32 s0, s0, 32
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -5741,8 +5741,8 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX7-NEXT:    s_addc_u32 s1, s1, s5
 ; GFX7-NEXT:    s_add_u32 s0, s0, 32
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_swap_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -5762,8 +5762,8 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_addr64_offset(ptr %out, ptr %out2
 ; GFX8-NEXT:    s_addc_u32 s1, s1, s5
 ; GFX8-NEXT:    s_add_u32 s0, s0, 32
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_swap_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -5903,8 +5903,8 @@ define amdgpu_kernel void @atomic_xchg_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX7-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; GFX7-NEXT:    s_add_u32 s0, s0, s2
 ; GFX7-NEXT:    s_addc_u32 s1, s1, s3
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -5920,8 +5920,8 @@ define amdgpu_kernel void @atomic_xchg_i64_addr64(ptr %out, i64 %in, i64 %index)
 ; GFX8-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; GFX8-NEXT:    s_add_u32 s0, s0, s2
 ; GFX8-NEXT:    s_addc_u32 s1, s1, s3
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -5958,8 +5958,8 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_addr64(ptr %out, ptr %out2, i64 %
 ; GFX7-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; GFX7-NEXT:    s_add_u32 s0, s0, s4
 ; GFX7-NEXT:    s_addc_u32 s1, s1, s5
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_atomic_swap_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -5977,8 +5977,8 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_addr64(ptr %out, ptr %out2, i64 %
 ; GFX8-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; GFX8-NEXT:    s_add_u32 s0, s0, s4
 ; GFX8-NEXT:    s_addc_u32 s1, s1, s5
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_atomic_swap_x2 v[0:1], v[2:3], v[0:1] glc
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -6936,8 +6936,8 @@ define amdgpu_kernel void @atomic_store_i64_offset(i64 %in, ptr %out) {
 ; GFX7-NEXT:    s_add_u32 s0, s2, 32
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    s_addc_u32 s1, s3, 0
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_endpgm
 ;
@@ -6949,8 +6949,8 @@ define amdgpu_kernel void @atomic_store_i64_offset(i64 %in, ptr %out) {
 ; GFX8-NEXT:    s_add_u32 s0, s2, 32
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    s_addc_u32 s1, s3, 0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_endpgm
 ;
@@ -7021,8 +7021,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64_offset(i64 %in, ptr %out, i64
 ; GFX7-NEXT:    s_addc_u32 s1, s3, s1
 ; GFX7-NEXT:    s_add_u32 s0, s0, 32
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_endpgm
 ;
@@ -7038,8 +7038,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64_offset(i64 %in, ptr %out, i64
 ; GFX8-NEXT:    s_addc_u32 s1, s3, s1
 ; GFX8-NEXT:    s_add_u32 s0, s0, 32
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_endpgm
 ;
@@ -7076,8 +7076,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64(i64 %in, ptr %out, i64 %index
 ; GFX7-NEXT:    s_lshl_b64 s[0:1], s[4:5], 3
 ; GFX7-NEXT:    s_add_u32 s0, s2, s0
 ; GFX7-NEXT:    s_addc_u32 s1, s3, s1
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_endpgm
 ;
@@ -7091,8 +7091,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64(i64 %in, ptr %out, i64 %index
 ; GFX8-NEXT:    s_lshl_b64 s[0:1], s[4:5], 3
 ; GFX8-NEXT:    s_add_u32 s0, s2, s0
 ; GFX8-NEXT:    s_addc_u32 s1, s3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_endpgm
 ;
@@ -7129,8 +7129,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_offset(ptr %out, i64 %in, i64 %old
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s5
-; GFX7-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-NEXT:    flat_atomic_cmpswap_x2 v[4:5], v[0:3]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -7147,8 +7147,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_offset(ptr %out, i64 %in, i64 %old
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s5
-; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    flat_atomic_cmpswap_x2 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -7185,8 +7185,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_soffset(ptr %out, i64 %in, i64 %ol
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s5
-; GFX7-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-NEXT:    flat_atomic_cmpswap_x2 v[4:5], v[0:3]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -7203,8 +7203,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_soffset(ptr %out, i64 %in, i64 %ol
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s5
-; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    flat_atomic_cmpswap_x2 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -7240,8 +7240,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_offset(ptr %out, ptr %out2, i6
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s7
-; GFX7-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-NEXT:    flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] glc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -7260,8 +7260,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_offset(ptr %out, ptr %out2, i6
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s7
-; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] glc
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -7305,8 +7305,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_addr64_offset(ptr %out, i64 %in, i
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s7
-; GFX7-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-NEXT:    flat_atomic_cmpswap_x2 v[4:5], v[0:3]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -7325,8 +7325,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_addr64_offset(ptr %out, i64 %in, i
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s7
-; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    flat_atomic_cmpswap_x2 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -7361,14 +7361,14 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64_offset(ptr %out, ptr %o
 ; GFX7-NEXT:    s_lshl_b64 s[2:3], s[14:15], 3
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX7-NEXT:    s_add_u32 s0, s8, s2
-; GFX7-NEXT:    s_addc_u32 s3, s9, s3
-; GFX7-NEXT:    s_add_u32 s2, s0, 32
-; GFX7-NEXT:    s_addc_u32 s3, s3, 0
+; GFX7-NEXT:    s_addc_u32 s2, s9, s3
+; GFX7-NEXT:    s_add_u32 s0, s0, 32
+; GFX7-NEXT:    s_addc_u32 s2, s2, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s12
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s13
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s1
-; GFX7-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-NEXT:    v_mov_b32_e32 v5, s2
 ; GFX7-NEXT:    flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] glc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -7385,14 +7385,14 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64_offset(ptr %out, ptr %o
 ; GFX8-NEXT:    s_lshl_b64 s[2:3], s[14:15], 3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX8-NEXT:    s_add_u32 s0, s8, s2
-; GFX8-NEXT:    s_addc_u32 s3, s9, s3
-; GFX8-NEXT:    s_add_u32 s2, s0, 32
-; GFX8-NEXT:    s_addc_u32 s3, s3, 0
+; GFX8-NEXT:    s_addc_u32 s2, s9, s3
+; GFX8-NEXT:    s_add_u32 s0, s0, 32
+; GFX8-NEXT:    s_addc_u32 s2, s2, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s12
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s13
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s2
 ; GFX8-NEXT:    flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] glc
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -7547,8 +7547,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_addr64(ptr %out, i64 %in, i64 %ind
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s7
-; GFX7-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-NEXT:    flat_atomic_cmpswap_x2 v[4:5], v[0:3]
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -7565,8 +7565,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_addr64(ptr %out, i64 %in, i64 %ind
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s7
-; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    flat_atomic_cmpswap_x2 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -7598,14 +7598,14 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64(ptr %out, ptr %out2, i6
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x11
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_lshl_b64 s[2:3], s[14:15], 3
-; GFX7-NEXT:    s_add_u32 s2, s8, s2
-; GFX7-NEXT:    s_addc_u32 s3, s9, s3
+; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    s_add_u32 s0, s8, s2
+; GFX7-NEXT:    s_addc_u32 s2, s9, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s12
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s13
-; GFX7-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s1
-; GFX7-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-NEXT:    v_mov_b32_e32 v5, s2
 ; GFX7-NEXT:    flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] glc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
@@ -7620,14 +7620,14 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64(ptr %out, ptr %out2, i6
 ; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x44
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_lshl_b64 s[2:3], s[14:15], 3
-; GFX8-NEXT:    s_add_u32 s2, s8, s2
-; GFX8-NEXT:    s_addc_u32 s3, s9, s3
+; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    s_add_u32 s0, s8, s2
+; GFX8-NEXT:    s_addc_u32 s2, s9, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s12
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s13
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s2
 ; GFX8-NEXT:    flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] glc
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    buffer_wbinvl1_vol
@@ -7892,8 +7892,8 @@ define amdgpu_kernel void @atomic_store_f64_offset(double %in, ptr %out) {
 ; GFX7-NEXT:    s_add_u32 s0, s2, 32
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    s_addc_u32 s1, s3, 0
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_endpgm
 ;
@@ -7905,8 +7905,8 @@ define amdgpu_kernel void @atomic_store_f64_offset(double %in, ptr %out) {
 ; GFX8-NEXT:    s_add_u32 s0, s2, 32
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    s_addc_u32 s1, s3, 0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_endpgm
 ;
@@ -7977,8 +7977,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64_offset(double %in, ptr %out,
 ; GFX7-NEXT:    s_addc_u32 s1, s3, s1
 ; GFX7-NEXT:    s_add_u32 s0, s0, 32
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_endpgm
 ;
@@ -7994,8 +7994,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64_offset(double %in, ptr %out,
 ; GFX8-NEXT:    s_addc_u32 s1, s3, s1
 ; GFX8-NEXT:    s_add_u32 s0, s0, 32
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_endpgm
 ;
@@ -8032,8 +8032,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64(double %in, ptr %out, i64 %in
 ; GFX7-NEXT:    s_lshl_b64 s[0:1], s[4:5], 3
 ; GFX7-NEXT:    s_add_u32 s0, s2, s0
 ; GFX7-NEXT:    s_addc_u32 s1, s3, s1
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX7-NEXT:    s_endpgm
 ;
@@ -8047,8 +8047,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64(double %in, ptr %out, i64 %in
 ; GFX8-NEXT:    s_lshl_b64 s[0:1], s[4:5], 3
 ; GFX8-NEXT:    s_add_u32 s0, s2, s0
 ; GFX8-NEXT:    s_addc_u32 s1, s3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GFX8-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/fmaximum.ll b/llvm/test/CodeGen/AMDGPU/fmaximum.ll
index 213d6ffeb9f00..aa1e2cf4e9166 100644
--- a/llvm/test/CodeGen/AMDGPU/fmaximum.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmaximum.ll
@@ -936,8 +936,8 @@ define amdgpu_ps <2 x float> @test_fmaximum_f64_vv(double %a, double %b) {
 define amdgpu_ps <2 x float> @test_fmaximum_f64_ss(double inreg %a, double inreg %b) {
 ; GFX9-SDAG-LABEL: test_fmaximum_f64_ss:
 ; GFX9-SDAG:       ; %bb.0:
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX9-SDAG-NEXT:    v_max_f64 v[2:3], s[0:1], v[0:1]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e32 vcc, s[0:1], v[0:1]
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, 0x7ff80000
@@ -999,12 +999,12 @@ define amdgpu_ps <2 x float> @test_fmaximum_f64_ss(double inreg %a, double inreg
 define amdgpu_ps <4 x float> @test_fmaximum_v2f64_ss(<2 x double> inreg %a, <2 x double> inreg %b) {
 ; GFX9-SDAG-LABEL: test_fmaximum_v2f64_ss:
 ; GFX9-SDAG:       ; %bb.0:
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-SDAG-NEXT:    v_max_f64 v[2:3], s[0:1], v[0:1]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e32 vcc, s[0:1], v[0:1]
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX9-SDAG-NEXT:    v_max_f64 v[4:5], s[2:3], v[0:1]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e64 s[0:1], s[2:3], v[0:1]
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v6, 0x7ff80000
@@ -1153,8 +1153,8 @@ define amdgpu_ps <8 x float> @test_fmaximum_v4f64(<4 x double> %a, <4 x double>
 define amdgpu_ps <8 x float> @test_fmaximum_v4f64_ss(<4 x double> inreg %a, <4 x double> inreg %b) {
 ; GFX9-SDAG-LABEL: test_fmaximum_v4f64_ss:
 ; GFX9-SDAG:       ; %bb.0:
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; GFX9-SDAG-NEXT:    v_max_f64 v[2:3], s[0:1], v[0:1]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e32 vcc, s[0:1], v[0:1]
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s10
@@ -1163,12 +1163,12 @@ define amdgpu_ps <8 x float> @test_fmaximum_v4f64_ss(<4 x double> inreg %a, <4 x
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v2, s11
 ; GFX9-SDAG-NEXT:    v_max_f64 v[4:5], s[2:3], v[1:2]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e64 s[0:1], s[2:3], v[1:2]
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s12
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v2, s13
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s12
 ; GFX9-SDAG-NEXT:    v_max_f64 v[6:7], s[4:5], v[1:2]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e64 s[2:3], s[4:5], v[1:2]
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s14
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v2, s15
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s14
 ; GFX9-SDAG-NEXT:    v_max_f64 v[8:9], s[6:7], v[1:2]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e64 s[4:5], s[6:7], v[1:2]
 ; GFX9-SDAG-NEXT:    v_cndmask_b32_e32 v1, v3, v10, vcc
diff --git a/llvm/test/CodeGen/AMDGPU/fmaxnum.ll b/llvm/test/CodeGen/AMDGPU/fmaxnum.ll
index 25090bf693552..08f0e2c5687a3 100644
--- a/llvm/test/CodeGen/AMDGPU/fmaxnum.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmaxnum.ll
@@ -335,18 +335,18 @@ define amdgpu_kernel void @test_fmax_v3f32(ptr addrspace(1) %out, <3 x float> %a
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v5, s14, s14
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
-; GFX12-GISEL-NEXT:    s_max_num_f32 s4, s2, s3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
+; GFX12-GISEL-NEXT:    s_max_num_f32 s2, s2, s3
+; GFX12-GISEL-NEXT:    s_max_num_f32 s3, s4, s5
+; GFX12-GISEL-NEXT:    s_max_num_f32 s4, s6, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_2) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
 ; GFX12-GISEL-NEXT:    s_mov_b32 s2, -1
-; GFX12-GISEL-NEXT:    s_max_num_f32 s5, s5, s6
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0x31016000
-; GFX12-GISEL-NEXT:    s_max_num_f32 s6, s7, s8
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_2)
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX12-GISEL-NEXT:    buffer_store_b96 v[0:2], off, s[0:3], null
 ; GFX12-GISEL-NEXT:    s_endpgm
   %val = call <3 x float> @llvm.maxnum.v3f32(<3 x float> %a, <3 x float> %b)
@@ -481,20 +481,20 @@ define amdgpu_kernel void @test_fmax_v4f32(ptr addrspace(1) %out, <4 x float> %a
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v7, s15, s15
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v6
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s10, v7
-; GFX12-GISEL-NEXT:    s_max_num_f32 s4, s2, s3
-; GFX12-GISEL-NEXT:    s_max_num_f32 s5, s5, s6
-; GFX12-GISEL-NEXT:    s_max_num_f32 s6, s7, s8
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v7
+; GFX12-GISEL-NEXT:    s_max_num_f32 s2, s2, s3
+; GFX12-GISEL-NEXT:    s_max_num_f32 s3, s4, s5
+; GFX12-GISEL-NEXT:    s_max_num_f32 s4, s6, s7
+; GFX12-GISEL-NEXT:    s_max_num_f32 s5, s8, s9
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_2)
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s5
 ; GFX12-GISEL-NEXT:    s_mov_b32 s2, -1
-; GFX12-GISEL-NEXT:    s_max_num_f32 s7, s9, s10
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_2)
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s6 :: v_dual_mov_b32 v3, s7
 ; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX12-GISEL-NEXT:    buffer_store_b128 v[0:3], off, s[0:3], null
 ; GFX12-GISEL-NEXT:    s_endpgm
@@ -702,33 +702,33 @@ define amdgpu_kernel void @test_fmax_v8f32(ptr addrspace(1) %out, <8 x float> %a
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v15, s23, s23
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v6
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s10, v7
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s11, v8
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v9
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v10
-; GFX12-GISEL-NEXT:    s_max_num_f32 s4, s2, s3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v11
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v12
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s14, v13
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s15, v14
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s16, v15
-; GFX12-GISEL-NEXT:    s_max_num_f32 s5, s5, s6
-; GFX12-GISEL-NEXT:    s_max_num_f32 s6, s7, s8
-; GFX12-GISEL-NEXT:    s_max_num_f32 s7, s9, s10
-; GFX12-GISEL-NEXT:    s_max_num_f32 s8, s11, s12
-; GFX12-GISEL-NEXT:    s_max_num_f32 s9, s13, s2
-; GFX12-GISEL-NEXT:    s_max_num_f32 s10, s3, s14
-; GFX12-GISEL-NEXT:    s_max_num_f32 s11, s15, s16
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s6 :: v_dual_mov_b32 v3, s7
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v7
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s10, v8
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s11, v9
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v10
+; GFX12-GISEL-NEXT:    s_max_num_f32 s2, s2, s3
+; GFX12-GISEL-NEXT:    s_max_num_f32 s3, s4, s5
+; GFX12-GISEL-NEXT:    s_max_num_f32 s4, s6, s7
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v11
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v12
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v13
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v14
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s14, v15
+; GFX12-GISEL-NEXT:    s_max_num_f32 s8, s8, s9
+; GFX12-GISEL-NEXT:    s_max_num_f32 s9, s10, s11
+; GFX12-GISEL-NEXT:    s_max_num_f32 s5, s12, s5
+; GFX12-GISEL-NEXT:    s_max_num_f32 s6, s6, s7
+; GFX12-GISEL-NEXT:    s_max_num_f32 s7, s13, s14
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
 ; GFX12-GISEL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s8 :: v_dual_mov_b32 v5, s9
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s10 :: v_dual_mov_b32 v7, s11
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s8
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s9 :: v_dual_mov_b32 v5, s5
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
 ; GFX12-GISEL-NEXT:    s_mov_b32 s2, -1
 ; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX12-GISEL-NEXT:    s_clause 0x1
@@ -1068,6 +1068,7 @@ define amdgpu_kernel void @test_fmax_v16f32(ptr addrspace(1) %out, <16 x float>
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v7, s11, s11
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v8, s40, s40
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v9, s12, s12
+; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v10, s41, s41
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v11, s13, s13
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
@@ -1075,80 +1076,79 @@ define amdgpu_kernel void @test_fmax_v16f32(ptr addrspace(1) %out, <16 x float>
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s11, v6
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v7
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v8
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s24, v9
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v7
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s10, v8
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s11, v9
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v10
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v11
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v0, s42, s42
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v1, s14, s14
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v2, s43, s43
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v3, s15, s15
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v4, s44, s44
-; GFX12-GISEL-NEXT:    s_max_num_f32 s8, s2, s3
-; GFX12-GISEL-NEXT:    s_max_num_f32 s9, s4, s5
-; GFX12-GISEL-NEXT:    s_max_num_f32 s10, s6, s7
-; GFX12-GISEL-NEXT:    s_max_num_f32 s11, s11, s12
-; GFX12-GISEL-NEXT:    s_max_num_f32 s4, s13, s24
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v4
+; GFX12-GISEL-NEXT:    s_max_num_f32 s2, s2, s3
+; GFX12-GISEL-NEXT:    s_max_num_f32 s3, s4, s5
+; GFX12-GISEL-NEXT:    s_max_num_f32 s4, s6, s7
+; GFX12-GISEL-NEXT:    s_max_num_f32 s5, s8, s9
+; GFX12-GISEL-NEXT:    s_max_num_f32 s6, s10, s11
+; GFX12-GISEL-NEXT:    s_max_num_f32 s7, s12, s13
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v1
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s10, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s11, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v4
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v0, s16, s16
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v1, s45, s45
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v2, s17, s17
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v3, s46, s46
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v4, s18, s18
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s14, v0
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s15, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s16, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s17, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s18, v4
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s14, v1
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s15, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s16, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s17, v4
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v0, s47, s47
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v1, s19, s19
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v2, s48, s48
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v3, s20, s20
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v4, s49, s49
-; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v10, s41, s41
-; GFX12-GISEL-NEXT:    s_max_num_f32 s6, s2, s3
-; GFX12-GISEL-NEXT:    s_max_num_f32 s7, s7, s12
-; GFX12-GISEL-NEXT:    s_max_num_f32 s12, s13, s14
-; GFX12-GISEL-NEXT:    s_max_num_f32 s13, s15, s16
-; GFX12-GISEL-NEXT:    s_max_num_f32 s14, s17, s18
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s16, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s17, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s18, v4
+; GFX12-GISEL-NEXT:    s_max_num_f32 s8, s8, s9
+; GFX12-GISEL-NEXT:    s_max_num_f32 s9, s10, s11
+; GFX12-GISEL-NEXT:    s_max_num_f32 s10, s12, s13
+; GFX12-GISEL-NEXT:    s_max_num_f32 s11, s14, s15
+; GFX12-GISEL-NEXT:    s_max_num_f32 s12, s16, s17
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s14, v1
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s15, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s16, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s17, v4
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v0, s21, s21
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v1, s50, s50
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v2, s22, s22
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v3, s51, s51
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v4, s23, s23
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s25, v10
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s26, v11
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s19, v0
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s20, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s21, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s22, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s23, v4
-; GFX12-GISEL-NEXT:    s_max_num_f32 s5, s25, s26
-; GFX12-GISEL-NEXT:    s_max_num_f32 s15, s2, s3
-; GFX12-GISEL-NEXT:    s_max_num_f32 s16, s16, s17
-; GFX12-GISEL-NEXT:    s_max_num_f32 s17, s18, s19
-; GFX12-GISEL-NEXT:    s_max_num_f32 s18, s20, s21
-; GFX12-GISEL-NEXT:    s_max_num_f32 s19, s22, s23
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s18, v0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s19, v1
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s20, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s21, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s22, v4
+; GFX12-GISEL-NEXT:    s_max_num_f32 s13, s13, s14
+; GFX12-GISEL-NEXT:    s_max_num_f32 s14, s15, s16
+; GFX12-GISEL-NEXT:    s_max_num_f32 s15, s17, s18
+; GFX12-GISEL-NEXT:    s_max_num_f32 s16, s19, s20
+; GFX12-GISEL-NEXT:    s_max_num_f32 s17, s21, s22
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s5
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v5, s7
 ; GFX12-GISEL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s8 :: v_dual_mov_b32 v1, s9
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s10 :: v_dual_mov_b32 v3, s11
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v5, s5
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v8, s12 :: v_dual_mov_b32 v9, s13
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v10, s14 :: v_dual_mov_b32 v11, s15
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s8 :: v_dual_mov_b32 v7, s9
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v8, s10 :: v_dual_mov_b32 v9, s11
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v10, s12 :: v_dual_mov_b32 v11, s13
 ; GFX12-GISEL-NEXT:    s_mov_b32 s2, -1
 ; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0x31016000
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v12, s16 :: v_dual_mov_b32 v13, s17
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v14, s18 :: v_dual_mov_b32 v15, s19
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v12, s14 :: v_dual_mov_b32 v13, s15
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v14, s16 :: v_dual_mov_b32 v15, s17
 ; GFX12-GISEL-NEXT:    s_clause 0x3
 ; GFX12-GISEL-NEXT:    buffer_store_b128 v[0:3], off, s[0:3], null
 ; GFX12-GISEL-NEXT:    buffer_store_b128 v[4:7], off, s[0:3], null offset:16
diff --git a/llvm/test/CodeGen/AMDGPU/fminimum.ll b/llvm/test/CodeGen/AMDGPU/fminimum.ll
index 094dd76e0f945..3f24f37f7c8b5 100644
--- a/llvm/test/CodeGen/AMDGPU/fminimum.ll
+++ b/llvm/test/CodeGen/AMDGPU/fminimum.ll
@@ -936,8 +936,8 @@ define amdgpu_ps <2 x float> @test_fminimum_f64_vv(double %a, double %b) {
 define amdgpu_ps <2 x float> @test_fminimum_f64_ss(double inreg %a, double inreg %b) {
 ; GFX9-SDAG-LABEL: test_fminimum_f64_ss:
 ; GFX9-SDAG:       ; %bb.0:
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX9-SDAG-NEXT:    v_min_f64 v[2:3], s[0:1], v[0:1]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e32 vcc, s[0:1], v[0:1]
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, 0x7ff80000
@@ -999,12 +999,12 @@ define amdgpu_ps <2 x float> @test_fminimum_f64_ss(double inreg %a, double inreg
 define amdgpu_ps <4 x float> @test_fminimum_v2f64_ss(<2 x double> inreg %a, <2 x double> inreg %b) {
 ; GFX9-SDAG-LABEL: test_fminimum_v2f64_ss:
 ; GFX9-SDAG:       ; %bb.0:
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-SDAG-NEXT:    v_min_f64 v[2:3], s[0:1], v[0:1]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e32 vcc, s[0:1], v[0:1]
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX9-SDAG-NEXT:    v_min_f64 v[4:5], s[2:3], v[0:1]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e64 s[0:1], s[2:3], v[0:1]
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v6, 0x7ff80000
@@ -1153,8 +1153,8 @@ define amdgpu_ps <8 x float> @test_fminimum_v4f64(<4 x double> %a, <4 x double>
 define amdgpu_ps <8 x float> @test_fminimum_v4f64_ss(<4 x double> inreg %a, <4 x double> inreg %b) {
 ; GFX9-SDAG-LABEL: test_fminimum_v4f64_ss:
 ; GFX9-SDAG:       ; %bb.0:
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; GFX9-SDAG-NEXT:    v_min_f64 v[2:3], s[0:1], v[0:1]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e32 vcc, s[0:1], v[0:1]
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s10
@@ -1163,12 +1163,12 @@ define amdgpu_ps <8 x float> @test_fminimum_v4f64_ss(<4 x double> inreg %a, <4 x
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v2, s11
 ; GFX9-SDAG-NEXT:    v_min_f64 v[4:5], s[2:3], v[1:2]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e64 s[0:1], s[2:3], v[1:2]
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s12
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v2, s13
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s12
 ; GFX9-SDAG-NEXT:    v_min_f64 v[6:7], s[4:5], v[1:2]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e64 s[2:3], s[4:5], v[1:2]
-; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s14
 ; GFX9-SDAG-NEXT:    v_mov_b32_e32 v2, s15
+; GFX9-SDAG-NEXT:    v_mov_b32_e32 v1, s14
 ; GFX9-SDAG-NEXT:    v_min_f64 v[8:9], s[6:7], v[1:2]
 ; GFX9-SDAG-NEXT:    v_cmp_u_f64_e64 s[4:5], s[6:7], v[1:2]
 ; GFX9-SDAG-NEXT:    v_cndmask_b32_e32 v1, v3, v10, vcc
diff --git a/llvm/test/CodeGen/AMDGPU/fminnum.ll b/llvm/test/CodeGen/AMDGPU/fminnum.ll
index 06fb492792ae8..6d8dd7503024a 100644
--- a/llvm/test/CodeGen/AMDGPU/fminnum.ll
+++ b/llvm/test/CodeGen/AMDGPU/fminnum.ll
@@ -396,20 +396,20 @@ define amdgpu_kernel void @test_fmin_v4f32(ptr addrspace(1) %out, <4 x float> %a
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v7, s15, s15
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v6
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s10, v7
-; GFX12-GISEL-NEXT:    s_min_num_f32 s4, s2, s3
-; GFX12-GISEL-NEXT:    s_min_num_f32 s5, s5, s6
-; GFX12-GISEL-NEXT:    s_min_num_f32 s6, s7, s8
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v7
+; GFX12-GISEL-NEXT:    s_min_num_f32 s2, s2, s3
+; GFX12-GISEL-NEXT:    s_min_num_f32 s3, s4, s5
+; GFX12-GISEL-NEXT:    s_min_num_f32 s4, s6, s7
+; GFX12-GISEL-NEXT:    s_min_num_f32 s5, s8, s9
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_2)
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s5
 ; GFX12-GISEL-NEXT:    s_mov_b32 s2, -1
-; GFX12-GISEL-NEXT:    s_min_num_f32 s7, s9, s10
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_2)
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s6 :: v_dual_mov_b32 v3, s7
 ; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX12-GISEL-NEXT:    buffer_store_b128 v[0:3], off, s[0:3], null
 ; GFX12-GISEL-NEXT:    s_endpgm
@@ -617,33 +617,33 @@ define amdgpu_kernel void @test_fmin_v8f32(ptr addrspace(1) %out, <8 x float> %a
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v15, s23, s23
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v6
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s10, v7
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s11, v8
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v9
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v10
-; GFX12-GISEL-NEXT:    s_min_num_f32 s4, s2, s3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v11
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v12
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s14, v13
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s15, v14
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s16, v15
-; GFX12-GISEL-NEXT:    s_min_num_f32 s5, s5, s6
-; GFX12-GISEL-NEXT:    s_min_num_f32 s6, s7, s8
-; GFX12-GISEL-NEXT:    s_min_num_f32 s7, s9, s10
-; GFX12-GISEL-NEXT:    s_min_num_f32 s8, s11, s12
-; GFX12-GISEL-NEXT:    s_min_num_f32 s9, s13, s2
-; GFX12-GISEL-NEXT:    s_min_num_f32 s10, s3, s14
-; GFX12-GISEL-NEXT:    s_min_num_f32 s11, s15, s16
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s6 :: v_dual_mov_b32 v3, s7
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v7
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s10, v8
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s11, v9
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v10
+; GFX12-GISEL-NEXT:    s_min_num_f32 s2, s2, s3
+; GFX12-GISEL-NEXT:    s_min_num_f32 s3, s4, s5
+; GFX12-GISEL-NEXT:    s_min_num_f32 s4, s6, s7
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v11
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v12
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v13
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v14
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s14, v15
+; GFX12-GISEL-NEXT:    s_min_num_f32 s8, s8, s9
+; GFX12-GISEL-NEXT:    s_min_num_f32 s9, s10, s11
+; GFX12-GISEL-NEXT:    s_min_num_f32 s5, s12, s5
+; GFX12-GISEL-NEXT:    s_min_num_f32 s6, s6, s7
+; GFX12-GISEL-NEXT:    s_min_num_f32 s7, s13, s14
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
 ; GFX12-GISEL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s8 :: v_dual_mov_b32 v5, s9
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s10 :: v_dual_mov_b32 v7, s11
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s8
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s9 :: v_dual_mov_b32 v5, s5
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
 ; GFX12-GISEL-NEXT:    s_mov_b32 s2, -1
 ; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX12-GISEL-NEXT:    s_clause 0x1
@@ -983,6 +983,7 @@ define amdgpu_kernel void @test_fmin_v16f32(ptr addrspace(1) %out, <16 x float>
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v7, s11, s11
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v8, s40, s40
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v9, s12, s12
+; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v10, s41, s41
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v11, s13, s13
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
@@ -990,80 +991,79 @@ define amdgpu_kernel void @test_fmin_v16f32(ptr addrspace(1) %out, <16 x float>
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s11, v6
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v7
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v8
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s24, v9
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v7
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s10, v8
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s11, v9
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v10
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v11
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v0, s42, s42
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v1, s14, s14
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v2, s43, s43
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v3, s15, s15
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v4, s44, s44
-; GFX12-GISEL-NEXT:    s_min_num_f32 s8, s2, s3
-; GFX12-GISEL-NEXT:    s_min_num_f32 s9, s4, s5
-; GFX12-GISEL-NEXT:    s_min_num_f32 s10, s6, s7
-; GFX12-GISEL-NEXT:    s_min_num_f32 s11, s11, s12
-; GFX12-GISEL-NEXT:    s_min_num_f32 s4, s13, s24
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s7, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v4
+; GFX12-GISEL-NEXT:    s_min_num_f32 s2, s2, s3
+; GFX12-GISEL-NEXT:    s_min_num_f32 s3, s4, s5
+; GFX12-GISEL-NEXT:    s_min_num_f32 s4, s6, s7
+; GFX12-GISEL-NEXT:    s_min_num_f32 s5, s8, s9
+; GFX12-GISEL-NEXT:    s_min_num_f32 s6, s10, s11
+; GFX12-GISEL-NEXT:    s_min_num_f32 s7, s12, s13
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s9, v1
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s10, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s11, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s12, v4
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v0, s16, s16
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v1, s45, s45
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v2, s17, s17
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v3, s46, s46
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v4, s18, s18
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s14, v0
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s15, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s16, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s17, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s18, v4
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s14, v1
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s15, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s16, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s17, v4
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v0, s47, s47
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v1, s19, s19
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v2, s48, s48
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v3, s20, s20
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v4, s49, s49
-; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v10, s41, s41
-; GFX12-GISEL-NEXT:    s_min_num_f32 s6, s2, s3
-; GFX12-GISEL-NEXT:    s_min_num_f32 s7, s7, s12
-; GFX12-GISEL-NEXT:    s_min_num_f32 s12, s13, s14
-; GFX12-GISEL-NEXT:    s_min_num_f32 s13, s15, s16
-; GFX12-GISEL-NEXT:    s_min_num_f32 s14, s17, s18
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s16, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s17, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s18, v4
+; GFX12-GISEL-NEXT:    s_min_num_f32 s8, s8, s9
+; GFX12-GISEL-NEXT:    s_min_num_f32 s9, s10, s11
+; GFX12-GISEL-NEXT:    s_min_num_f32 s10, s12, s13
+; GFX12-GISEL-NEXT:    s_min_num_f32 s11, s14, s15
+; GFX12-GISEL-NEXT:    s_min_num_f32 s12, s16, s17
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s13, v0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s14, v1
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s15, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s16, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s17, v4
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v0, s21, s21
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v1, s50, s50
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v2, s22, s22
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v3, s51, s51
 ; GFX12-GISEL-NEXT:    v_max_num_f32_e64 v4, s23, s23
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s25, v10
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s26, v11
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s19, v0
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s20, v1
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s21, v2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s22, v3
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s23, v4
-; GFX12-GISEL-NEXT:    s_min_num_f32 s5, s25, s26
-; GFX12-GISEL-NEXT:    s_min_num_f32 s15, s2, s3
-; GFX12-GISEL-NEXT:    s_min_num_f32 s16, s16, s17
-; GFX12-GISEL-NEXT:    s_min_num_f32 s17, s18, s19
-; GFX12-GISEL-NEXT:    s_min_num_f32 s18, s20, s21
-; GFX12-GISEL-NEXT:    s_min_num_f32 s19, s22, s23
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s18, v0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s19, v1
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s20, v2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s21, v3
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s22, v4
+; GFX12-GISEL-NEXT:    s_min_num_f32 s13, s13, s14
+; GFX12-GISEL-NEXT:    s_min_num_f32 s14, s15, s16
+; GFX12-GISEL-NEXT:    s_min_num_f32 s15, s17, s18
+; GFX12-GISEL-NEXT:    s_min_num_f32 s16, s19, s20
+; GFX12-GISEL-NEXT:    s_min_num_f32 s17, s21, s22
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s5
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v5, s7
 ; GFX12-GISEL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s8 :: v_dual_mov_b32 v1, s9
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s10 :: v_dual_mov_b32 v3, s11
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v5, s5
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v8, s12 :: v_dual_mov_b32 v9, s13
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v10, s14 :: v_dual_mov_b32 v11, s15
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s8 :: v_dual_mov_b32 v7, s9
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v8, s10 :: v_dual_mov_b32 v9, s11
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v10, s12 :: v_dual_mov_b32 v11, s13
 ; GFX12-GISEL-NEXT:    s_mov_b32 s2, -1
 ; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0x31016000
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v12, s16 :: v_dual_mov_b32 v13, s17
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v14, s18 :: v_dual_mov_b32 v15, s19
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v12, s14 :: v_dual_mov_b32 v13, s15
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v14, s16 :: v_dual_mov_b32 v15, s17
 ; GFX12-GISEL-NEXT:    s_clause 0x3
 ; GFX12-GISEL-NEXT:    buffer_store_b128 v[0:3], off, s[0:3], null
 ; GFX12-GISEL-NEXT:    buffer_store_b128 v[4:7], off, s[0:3], null offset:16
diff --git a/llvm/test/CodeGen/AMDGPU/fmul-2-combine-multi-use.ll b/llvm/test/CodeGen/AMDGPU/fmul-2-combine-multi-use.ll
index 3d7a5833123d9..63f8cc49e2e3e 100644
--- a/llvm/test/CodeGen/AMDGPU/fmul-2-combine-multi-use.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmul-2-combine-multi-use.ll
@@ -83,22 +83,22 @@ define amdgpu_kernel void @multiple_use_fadd_fmac_f32(ptr addrspace(1) %out, flo
 ; VI-LABEL: multiple_use_fadd_fmac_f32:
 ; VI:       ; %bb.0:
 ; VI-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
-; VI-NEXT:    s_load_dword s4, s[8:9], 0x8
+; VI-NEXT:    s_load_dword s2, s[8:9], 0x8
 ; VI-NEXT:    s_load_dword s3, s[8:9], 0x2c
 ; VI-NEXT:    s_add_i32 s12, s12, s17
 ; VI-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-NEXT:    s_add_u32 s2, s0, 4
+; VI-NEXT:    s_add_u32 s4, s0, 4
 ; VI-NEXT:    s_mov_b32 flat_scratch_lo, s13
-; VI-NEXT:    v_add_f32_e64 v2, s4, s4
+; VI-NEXT:    v_add_f32_e64 v2, s2, s2
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
-; VI-NEXT:    v_mac_f32_e64 v3, s4, 2.0
+; VI-NEXT:    v_mac_f32_e64 v3, s2, 2.0
 ; VI-NEXT:    flat_store_dword v[0:1], v2
 ; VI-NEXT:    s_waitcnt vmcnt(0)
-; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v0, s4
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    flat_store_dword v[0:1], v3
 ; VI-NEXT:    s_waitcnt vmcnt(0)
@@ -156,12 +156,12 @@ define amdgpu_kernel void @multiple_use_fadd_fmad_f32(ptr addrspace(1) %out, flo
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
-; VI-NEXT:    s_addc_u32 s5, s1, 0
+; VI-NEXT:    s_addc_u32 s0, s1, 0
 ; VI-NEXT:    v_mad_f32 v3, |s2|, 2.0, v3
 ; VI-NEXT:    flat_store_dword v[0:1], v2
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v0, s4
-; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    v_mov_b32_e32 v1, s0
 ; VI-NEXT:    flat_store_dword v[0:1], v3
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    s_endpgm
@@ -209,18 +209,18 @@ define amdgpu_kernel void @multiple_use_fadd_multi_fmad_f32(ptr addrspace(1) %ou
 ; VI-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; VI-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-NEXT:    s_add_u32 s6, s4, 4
+; VI-NEXT:    s_add_u32 s3, s4, 4
 ; VI-NEXT:    v_mov_b32_e32 v0, s1
 ; VI-NEXT:    v_mov_b32_e32 v1, s2
-; VI-NEXT:    s_addc_u32 s7, s5, 0
+; VI-NEXT:    s_addc_u32 s1, s5, 0
 ; VI-NEXT:    v_mad_f32 v2, |s0|, 2.0, v0
 ; VI-NEXT:    v_mad_f32 v3, |s0|, 2.0, v1
 ; VI-NEXT:    v_mov_b32_e32 v0, s4
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    flat_store_dword v[0:1], v2
 ; VI-NEXT:    s_waitcnt vmcnt(0)
-; VI-NEXT:    v_mov_b32_e32 v0, s6
-; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_mov_b32_e32 v0, s3
+; VI-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-NEXT:    flat_store_dword v[0:1], v3
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    s_endpgm
@@ -549,23 +549,23 @@ define amdgpu_kernel void @multiple_fadd_use_test_f16(ptr addrspace(1) %out, i16
 define amdgpu_kernel void @multiple_use_fadd_fmac_f16(ptr addrspace(1) %out, i16 zeroext %x.arg, i16 zeroext %y.arg) #0 {
 ; VI-DENORM-LABEL: multiple_use_fadd_fmac_f16:
 ; VI-DENORM:       ; %bb.0:
-; VI-DENORM-NEXT:    s_load_dword s4, s[8:9], 0x8
+; VI-DENORM-NEXT:    s_load_dword s2, s[8:9], 0x8
 ; VI-DENORM-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
 ; VI-DENORM-NEXT:    s_add_i32 s12, s12, s17
 ; VI-DENORM-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; VI-DENORM-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; VI-DENORM-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-DENORM-NEXT:    s_lshr_b32 s3, s4, 16
-; VI-DENORM-NEXT:    s_add_u32 s2, s0, 2
+; VI-DENORM-NEXT:    s_lshr_b32 s3, s2, 16
+; VI-DENORM-NEXT:    s_add_u32 s4, s0, 2
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s3
-; VI-DENORM-NEXT:    v_add_f16_e64 v2, s4, s4
+; VI-DENORM-NEXT:    v_add_f16_e64 v2, s2, s2
 ; VI-DENORM-NEXT:    s_addc_u32 s3, s1, 0
-; VI-DENORM-NEXT:    v_fma_f16 v3, s4, 2.0, v0
+; VI-DENORM-NEXT:    v_fma_f16 v3, s2, 2.0, v0
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-DENORM-NEXT:    flat_store_short v[0:1], v2
 ; VI-DENORM-NEXT:    s_waitcnt vmcnt(0)
-; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s2
+; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s4
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-DENORM-NEXT:    flat_store_short v[0:1], v3
 ; VI-DENORM-NEXT:    s_waitcnt vmcnt(0)
@@ -573,23 +573,23 @@ define amdgpu_kernel void @multiple_use_fadd_fmac_f16(ptr addrspace(1) %out, i16
 ;
 ; VI-FLUSH-LABEL: multiple_use_fadd_fmac_f16:
 ; VI-FLUSH:       ; %bb.0:
-; VI-FLUSH-NEXT:    s_load_dword s4, s[8:9], 0x8
+; VI-FLUSH-NEXT:    s_load_dword s2, s[8:9], 0x8
 ; VI-FLUSH-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
 ; VI-FLUSH-NEXT:    s_add_i32 s12, s12, s17
 ; VI-FLUSH-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; VI-FLUSH-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; VI-FLUSH-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-FLUSH-NEXT:    s_lshr_b32 s3, s4, 16
-; VI-FLUSH-NEXT:    s_add_u32 s2, s0, 2
-; VI-FLUSH-NEXT:    v_add_f16_e64 v2, s4, s4
+; VI-FLUSH-NEXT:    s_lshr_b32 s3, s2, 16
+; VI-FLUSH-NEXT:    s_add_u32 s4, s0, 2
+; VI-FLUSH-NEXT:    v_add_f16_e64 v2, s2, s2
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-FLUSH-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v1, s1
-; VI-FLUSH-NEXT:    v_mac_f16_e64 v3, s4, 2.0
+; VI-FLUSH-NEXT:    v_mac_f16_e64 v3, s2, 2.0
 ; VI-FLUSH-NEXT:    flat_store_short v[0:1], v2
 ; VI-FLUSH-NEXT:    s_waitcnt vmcnt(0)
-; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s2
+; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s4
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-FLUSH-NEXT:    flat_store_short v[0:1], v3
 ; VI-FLUSH-NEXT:    s_waitcnt vmcnt(0)
@@ -705,23 +705,23 @@ define amdgpu_kernel void @multiple_use_fadd_fmac_f16(ptr addrspace(1) %out, i16
 define amdgpu_kernel void @multiple_use_fadd_fmad_f16(ptr addrspace(1) %out, i16 zeroext %x.arg, i16 zeroext %y.arg) #0 {
 ; VI-DENORM-LABEL: multiple_use_fadd_fmad_f16:
 ; VI-DENORM:       ; %bb.0:
-; VI-DENORM-NEXT:    s_load_dword s4, s[8:9], 0x8
+; VI-DENORM-NEXT:    s_load_dword s2, s[8:9], 0x8
 ; VI-DENORM-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
 ; VI-DENORM-NEXT:    s_add_i32 s12, s12, s17
 ; VI-DENORM-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; VI-DENORM-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; VI-DENORM-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-DENORM-NEXT:    s_lshr_b32 s3, s4, 16
-; VI-DENORM-NEXT:    s_add_u32 s2, s0, 2
+; VI-DENORM-NEXT:    s_lshr_b32 s3, s2, 16
+; VI-DENORM-NEXT:    s_add_u32 s4, s0, 2
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s3
-; VI-DENORM-NEXT:    v_add_f16_e64 v2, |s4|, |s4|
+; VI-DENORM-NEXT:    v_add_f16_e64 v2, |s2|, |s2|
 ; VI-DENORM-NEXT:    s_addc_u32 s3, s1, 0
-; VI-DENORM-NEXT:    v_fma_f16 v3, |s4|, 2.0, v0
+; VI-DENORM-NEXT:    v_fma_f16 v3, |s2|, 2.0, v0
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-DENORM-NEXT:    flat_store_short v[0:1], v2
 ; VI-DENORM-NEXT:    s_waitcnt vmcnt(0)
-; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s2
+; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s4
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-DENORM-NEXT:    flat_store_short v[0:1], v3
 ; VI-DENORM-NEXT:    s_waitcnt vmcnt(0)
@@ -729,23 +729,23 @@ define amdgpu_kernel void @multiple_use_fadd_fmad_f16(ptr addrspace(1) %out, i16
 ;
 ; VI-FLUSH-LABEL: multiple_use_fadd_fmad_f16:
 ; VI-FLUSH:       ; %bb.0:
-; VI-FLUSH-NEXT:    s_load_dword s4, s[8:9], 0x8
+; VI-FLUSH-NEXT:    s_load_dword s2, s[8:9], 0x8
 ; VI-FLUSH-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
 ; VI-FLUSH-NEXT:    s_add_i32 s12, s12, s17
 ; VI-FLUSH-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; VI-FLUSH-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; VI-FLUSH-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-FLUSH-NEXT:    s_lshr_b32 s3, s4, 16
-; VI-FLUSH-NEXT:    s_add_u32 s2, s0, 2
+; VI-FLUSH-NEXT:    s_lshr_b32 s3, s2, 16
+; VI-FLUSH-NEXT:    s_add_u32 s4, s0, 2
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s3
-; VI-FLUSH-NEXT:    v_add_f16_e64 v2, |s4|, |s4|
+; VI-FLUSH-NEXT:    v_add_f16_e64 v2, |s2|, |s2|
 ; VI-FLUSH-NEXT:    s_addc_u32 s3, s1, 0
-; VI-FLUSH-NEXT:    v_mad_f16 v3, |s4|, 2.0, v0
+; VI-FLUSH-NEXT:    v_mad_f16 v3, |s2|, 2.0, v0
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-FLUSH-NEXT:    flat_store_short v[0:1], v2
 ; VI-FLUSH-NEXT:    s_waitcnt vmcnt(0)
-; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s2
+; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s4
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-FLUSH-NEXT:    flat_store_short v[0:1], v3
 ; VI-FLUSH-NEXT:    s_waitcnt vmcnt(0)
@@ -867,19 +867,19 @@ define amdgpu_kernel void @multiple_use_fadd_multi_fmad_f16(ptr addrspace(1) %ou
 ; VI-DENORM-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; VI-DENORM-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; VI-DENORM-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-DENORM-NEXT:    s_lshr_b32 s5, s2, 16
+; VI-DENORM-NEXT:    s_lshr_b32 s4, s2, 16
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s3
-; VI-DENORM-NEXT:    s_add_u32 s4, s0, 2
-; VI-DENORM-NEXT:    v_mov_b32_e32 v1, s5
+; VI-DENORM-NEXT:    s_add_u32 s3, s0, 2
+; VI-DENORM-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-DENORM-NEXT:    v_fma_f16 v2, |s2|, 2.0, v0
-; VI-DENORM-NEXT:    s_addc_u32 s5, s1, 0
+; VI-DENORM-NEXT:    s_addc_u32 s4, s1, 0
 ; VI-DENORM-NEXT:    v_fma_f16 v3, |s2|, 2.0, v1
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-DENORM-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-DENORM-NEXT:    flat_store_short v[0:1], v3
 ; VI-DENORM-NEXT:    s_waitcnt vmcnt(0)
-; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s4
-; VI-DENORM-NEXT:    v_mov_b32_e32 v1, s5
+; VI-DENORM-NEXT:    v_mov_b32_e32 v0, s3
+; VI-DENORM-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-DENORM-NEXT:    flat_store_short v[0:1], v2
 ; VI-DENORM-NEXT:    s_waitcnt vmcnt(0)
 ; VI-DENORM-NEXT:    s_endpgm
@@ -891,19 +891,19 @@ define amdgpu_kernel void @multiple_use_fadd_multi_fmad_f16(ptr addrspace(1) %ou
 ; VI-FLUSH-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; VI-FLUSH-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; VI-FLUSH-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-FLUSH-NEXT:    s_lshr_b32 s5, s2, 16
+; VI-FLUSH-NEXT:    s_lshr_b32 s4, s2, 16
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s3
-; VI-FLUSH-NEXT:    s_add_u32 s4, s0, 2
-; VI-FLUSH-NEXT:    v_mov_b32_e32 v1, s5
+; VI-FLUSH-NEXT:    s_add_u32 s3, s0, 2
+; VI-FLUSH-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-FLUSH-NEXT:    v_mad_f16 v2, |s2|, 2.0, v0
-; VI-FLUSH-NEXT:    s_addc_u32 s5, s1, 0
+; VI-FLUSH-NEXT:    s_addc_u32 s4, s1, 0
 ; VI-FLUSH-NEXT:    v_mad_f16 v3, |s2|, 2.0, v1
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-FLUSH-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-FLUSH-NEXT:    flat_store_short v[0:1], v3
 ; VI-FLUSH-NEXT:    s_waitcnt vmcnt(0)
-; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s4
-; VI-FLUSH-NEXT:    v_mov_b32_e32 v1, s5
+; VI-FLUSH-NEXT:    v_mov_b32_e32 v0, s3
+; VI-FLUSH-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-FLUSH-NEXT:    flat_store_short v[0:1], v2
 ; VI-FLUSH-NEXT:    s_waitcnt vmcnt(0)
 ; VI-FLUSH-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/fnearbyint.ll b/llvm/test/CodeGen/AMDGPU/fnearbyint.ll
index 6a55d358734a8..6cda5ad134598 100644
--- a/llvm/test/CodeGen/AMDGPU/fnearbyint.ll
+++ b/llvm/test/CodeGen/AMDGPU/fnearbyint.ll
@@ -254,22 +254,20 @@ define amdgpu_kernel void @nearbyint_v2f64(ptr addrspace(1) %out, <2 x double> %
 ; SI:       ; %bb.0: ; %entry
 ; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0xd
 ; SI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x9
-; SI-NEXT:    s_brev_b32 s10, -2
+; SI-NEXT:    s_brev_b32 s8, -2
 ; SI-NEXT:    v_mov_b32_e32 v6, 0x43300000
 ; SI-NEXT:    v_mov_b32_e32 v0, 0
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v7, s3
-; SI-NEXT:    v_bfi_b32 v1, s10, v6, v7
-; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    v_bfi_b32 v1, s8, v6, v7
 ; SI-NEXT:    v_add_f64 v[2:3], s[2:3], v[0:1]
-; SI-NEXT:    s_mov_b32 s9, 0x432fffff
-; SI-NEXT:    s_mov_b32 s8, s6
-; SI-NEXT:    v_mov_b32_e32 v4, s8
-; SI-NEXT:    v_mov_b32_e32 v5, s9
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    v_mov_b32_e32 v5, 0x432fffff
+; SI-NEXT:    v_mov_b32_e32 v4, s6
 ; SI-NEXT:    v_add_f64 v[2:3], v[2:3], -v[0:1]
 ; SI-NEXT:    v_cmp_gt_f64_e64 vcc, |s[2:3]|, v[4:5]
 ; SI-NEXT:    v_mov_b32_e32 v9, s1
-; SI-NEXT:    v_bfi_b32 v1, s10, v6, v9
+; SI-NEXT:    v_bfi_b32 v1, s8, v6, v9
 ; SI-NEXT:    v_cndmask_b32_e32 v3, v3, v7, vcc
 ; SI-NEXT:    v_add_f64 v[6:7], s[0:1], v[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v8, s2
@@ -337,12 +335,9 @@ define amdgpu_kernel void @nearbyint_v4f64(ptr addrspace(1) %out, <4 x double> %
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, s11
 ; SI-NEXT:    v_bfi_b32 v5, s6, v10, v2
-; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
 ; SI-NEXT:    v_add_f64 v[0:1], s[10:11], v[4:5]
-; SI-NEXT:    s_mov_b32 s5, 0x432fffff
-; SI-NEXT:    s_mov_b32 s4, s2
-; SI-NEXT:    v_mov_b32_e32 v9, s5
-; SI-NEXT:    v_mov_b32_e32 v8, s4
+; SI-NEXT:    v_mov_b32_e32 v9, 0x432fffff
+; SI-NEXT:    v_mov_b32_e32 v8, s2
 ; SI-NEXT:    v_mov_b32_e32 v11, s9
 ; SI-NEXT:    v_add_f64 v[0:1], v[0:1], -v[4:5]
 ; SI-NEXT:    v_cmp_gt_f64_e64 vcc, |s[10:11]|, v[8:9]
@@ -363,6 +358,7 @@ define amdgpu_kernel void @nearbyint_v4f64(ptr addrspace(1) %out, <4 x double> %
 ; SI-NEXT:    v_add_f64 v[6:7], v[6:7], -v[4:5]
 ; SI-NEXT:    v_cmp_gt_f64_e64 vcc, |s[14:15]|, v[8:9]
 ; SI-NEXT:    v_bfi_b32 v5, s6, v10, v13
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
 ; SI-NEXT:    v_add_f64 v[10:11], s[12:13], v[4:5]
 ; SI-NEXT:    v_cndmask_b32_e32 v7, v7, v12, vcc
 ; SI-NEXT:    v_mov_b32_e32 v12, s14
@@ -403,11 +399,11 @@ define amdgpu_kernel void @nearbyint_v4f64(ptr addrspace(1) %out, <4 x double> %
 ; VI-NEXT:    v_rndne_f64_e32 v[2:3], s[10:11]
 ; VI-NEXT:    v_rndne_f64_e32 v[0:1], s[8:9]
 ; VI-NEXT:    s_add_u32 s2, s0, 16
-; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v11, s3
-; VI-NEXT:    v_mov_b32_e32 v10, s2
 ; VI-NEXT:    v_mov_b32_e32 v9, s1
 ; VI-NEXT:    v_mov_b32_e32 v8, s0
+; VI-NEXT:    s_addc_u32 s0, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v10, s2
+; VI-NEXT:    v_mov_b32_e32 v11, s0
 ; VI-NEXT:    flat_store_dwordx4 v[10:11], v[4:7]
 ; VI-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
 ; VI-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/fneg-fabs.f64.ll b/llvm/test/CodeGen/AMDGPU/fneg-fabs.f64.ll
index 140f0f6863128..a2df43ca75864 100644
--- a/llvm/test/CodeGen/AMDGPU/fneg-fabs.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fneg-fabs.f64.ll
@@ -238,24 +238,24 @@ define amdgpu_kernel void @fneg_fabs_v4f64(ptr addrspace(1) %out, <4 x double> %
 ; VI-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x44
 ; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-NEXT:    s_or_b32 s4, s11, 0x80000000
-; VI-NEXT:    s_or_b32 s5, s9, 0x80000000
-; VI-NEXT:    s_or_b32 s2, s15, 0x80000000
-; VI-NEXT:    s_or_b32 s3, s13, 0x80000000
-; VI-NEXT:    v_mov_b32_e32 v3, s2
-; VI-NEXT:    s_add_u32 s2, s0, 16
-; VI-NEXT:    v_mov_b32_e32 v1, s3
-; VI-NEXT:    s_addc_u32 s3, s1, 0
+; VI-NEXT:    s_or_b32 s2, s11, 0x80000000
+; VI-NEXT:    s_or_b32 s3, s9, 0x80000000
+; VI-NEXT:    s_or_b32 s4, s15, 0x80000000
+; VI-NEXT:    s_or_b32 s5, s13, 0x80000000
+; VI-NEXT:    v_mov_b32_e32 v3, s4
+; VI-NEXT:    s_add_u32 s4, s0, 16
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_addc_u32 s5, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s12
 ; VI-NEXT:    v_mov_b32_e32 v2, s14
-; VI-NEXT:    v_mov_b32_e32 v5, s3
-; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v5, s5
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v0, s8
-; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    v_mov_b32_e32 v2, s10
-; VI-NEXT:    v_mov_b32_e32 v3, s4
+; VI-NEXT:    v_mov_b32_e32 v3, s2
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/fp-min-max-num-flat-atomics.ll b/llvm/test/CodeGen/AMDGPU/fp-min-max-num-flat-atomics.ll
index d0815c583e5bd..0abdb5fa9500a 100644
--- a/llvm/test/CodeGen/AMDGPU/fp-min-max-num-flat-atomics.ll
+++ b/llvm/test/CodeGen/AMDGPU/fp-min-max-num-flat-atomics.ll
@@ -8,19 +8,35 @@ declare float @llvm.amdgcn.flat.atomic.fmin.num.f32.p1.f32(ptr %ptr, float %data
 declare float @llvm.amdgcn.flat.atomic.fmax.num.f32.p1.f32(ptr %ptr, float %data)
 
 define amdgpu_cs void @flat_atomic_fmin_num_f32_noret(ptr %ptr, float %data) {
-; GCN-LABEL: flat_atomic_fmin_num_f32_noret:
-; GCN:  ; %bb.0:
-; GCN:    flat_atomic_min_num_f32 v[0:1], v2
-; GCN:    s_endpgm
+; GFX12-LABEL: flat_atomic_fmin_num_f32_noret:
+; GFX12:  ; %bb.0:
+; GFX12:    flat_atomic_min_num_f32 v[0:1], v2
+; GFX12:    s_endpgm
+;
+; GFX1250-LABEL: flat_atomic_fmin_num_f32_noret:
+; GFX1250:  ; %bb.0:
+; GFX1250:    s_mov_b64 s[64:65], 0
+; GFX1250:    v_nop
+; GFX1250:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
+; GFX1250:    flat_atomic_min_num_f32 v[0:1], v2
+; GFX1250:    s_endpgm
   %ret = call float @llvm.amdgcn.flat.atomic.fmin.num.f32.p1.f32(ptr %ptr, float %data)
   ret void
 }
 
 define amdgpu_cs void @flat_atomic_fmax_num_f32_noret(ptr %ptr, float %data) {
-; GCN-LABEL: flat_atomic_fmax_num_f32_noret:
-; GCN:  ; %bb.0:
-; GCN:    flat_atomic_max_num_f32 v[0:1], v2
-; GCN:    s_endpgm
+; GFX12-LABEL: flat_atomic_fmax_num_f32_noret:
+; GFX12:  ; %bb.0:
+; GFX12:    flat_atomic_max_num_f32 v[0:1], v2
+; GFX12:    s_endpgm
+;
+; GFX1250-LABEL: flat_atomic_fmax_num_f32_noret:
+; GFX1250:  ; %bb.0:
+; GFX1250:    s_mov_b64 s[64:65], 0
+; GFX1250:    v_nop
+; GFX1250:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
+; GFX1250:    flat_atomic_max_num_f32 v[0:1], v2
+; GFX1250:    s_endpgm
   %ret = call float @llvm.amdgcn.flat.atomic.fmax.num.f32.p1.f32(ptr %ptr, float %data)
   ret void
 }
@@ -34,6 +50,9 @@ define amdgpu_cs float @flat_atomic_fmin_num_f32_rtn(float %data, ptr %ptr) {
 ;
 ; GFX1250-SDAG-LABEL: flat_atomic_fmin_num_f32_rtn:
 ; GFX1250-SDAG:  ; %bb.0:
+; GFX1250-SDAG:    s_mov_b64 s[64:65], 0
+; GFX1250-SDAG:    v_nop
+; GFX1250-SDAG:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-SDAG:    v_dual_mov_b32 v3, v2 :: v_dual_mov_b32 v2, v1
 ; GFX1250-SDAG:    flat_atomic_min_num_f32 v0, v[2:3], v0 th:TH_ATOMIC_RETURN
 ; GFX1250-SDAG:    s_wait_loadcnt_dscnt 0x0
@@ -41,6 +60,9 @@ define amdgpu_cs float @flat_atomic_fmin_num_f32_rtn(float %data, ptr %ptr) {
 ;
 ; GFX1250-GISEL-LABEL: flat_atomic_fmin_num_f32_rtn:
 ; GFX1250-GISEL:  ; %bb.0:
+; GFX1250-GISEL:    s_mov_b64 s[64:65], 0
+; GFX1250-GISEL:    v_nop
+; GFX1250-GISEL:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL:    v_dual_mov_b32 v4, v1 :: v_dual_mov_b32 v5, v2
 ; GFX1250-GISEL:    flat_atomic_min_num_f32 v0, v[4:5], v0 th:TH_ATOMIC_RETURN
 ; GFX1250-GISEL:    s_wait_loadcnt_dscnt 0x0
@@ -58,6 +80,9 @@ define amdgpu_cs float @flat_atomic_fmax_num_f32_rtn(float %data, ptr %ptr) {
 ;
 ; GFX1250-SDAG-LABEL: flat_atomic_fmax_num_f32_rtn:
 ; GFX1250-SDAG:  ; %bb.0:
+; GFX1250-SDAG:    s_mov_b64 s[64:65], 0
+; GFX1250-SDAG:    v_nop
+; GFX1250-SDAG:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-SDAG:    v_dual_mov_b32 v3, v2 :: v_dual_mov_b32 v2, v1
 ; GFX1250-SDAG:    flat_atomic_max_num_f32 v0, v[2:3], v0 th:TH_ATOMIC_RETURN
 ; GFX1250-SDAG:    s_wait_loadcnt_dscnt 0x0
@@ -65,6 +90,9 @@ define amdgpu_cs float @flat_atomic_fmax_num_f32_rtn(float %data, ptr %ptr) {
 ;
 ; GFX1250-GISEL-LABEL: flat_atomic_fmax_num_f32_rtn:
 ; GFX1250-GISEL:  ; %bb.0:
+; GFX1250-GISEL:    s_mov_b64 s[64:65], 0
+; GFX1250-GISEL:    v_nop
+; GFX1250-GISEL:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL:    v_dual_mov_b32 v4, v1 :: v_dual_mov_b32 v5, v2
 ; GFX1250-GISEL:    flat_atomic_max_num_f32 v0, v[4:5], v0 th:TH_ATOMIC_RETURN
 ; GFX1250-GISEL:    s_wait_loadcnt_dscnt 0x0
@@ -74,20 +102,17 @@ define amdgpu_cs float @flat_atomic_fmax_num_f32_rtn(float %data, ptr %ptr) {
 }
 
 define amdgpu_ps void @flat_atomic_fmin_num_f32_noret_saddr(ptr inreg %ptr, float %data) {
-; GFX12-SDAG-LABEL: flat_atomic_fmin_num_f32_noret_saddr:
-; GFX12-SDAG:  ; %bb.0:
-; GFX12-SDAG:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
-; GFX12-SDAG:    flat_atomic_min_num_f32 v[1:2], v0
-; GFX12-SDAG:    s_endpgm
-;
-; GFX12-GISEL-LABEL: flat_atomic_fmin_num_f32_noret_saddr:
-; GFX12-GISEL:  ; %bb.0:
-; GFX12-GISEL:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
-; GFX12-GISEL:    flat_atomic_min_num_f32 v[1:2], v0
-; GFX12-GISEL:    s_endpgm
+; GFX12-LABEL: flat_atomic_fmin_num_f32_noret_saddr:
+; GFX12:  ; %bb.0:
+; GFX12:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
+; GFX12:    flat_atomic_min_num_f32 v[1:2], v0
+; GFX12:    s_endpgm
 ;
 ; GFX1250-LABEL: flat_atomic_fmin_num_f32_noret_saddr:
 ; GFX1250:  ; %bb.0:
+; GFX1250:    s_mov_b64 s[64:65], 0
+; GFX1250:    v_nop
+; GFX1250:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250:    v_mov_b32_e32 v1, 0
 ; GFX1250:    flat_atomic_min_num_f32 v1, v0, s[0:1]
 ; GFX1250:    s_endpgm
@@ -96,20 +121,17 @@ define amdgpu_ps void @flat_atomic_fmin_num_f32_noret_saddr(ptr inreg %ptr, floa
 }
 
 define amdgpu_ps void @flat_atomic_fmax_num_f32_noret_saddr(ptr inreg %ptr, float %data) {
-; GFX12-SDAG-LABEL: flat_atomic_fmax_num_f32_noret_saddr:
-; GFX12-SDAG:  ; %bb.0:
-; GFX12-SDAG:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
-; GFX12-SDAG:    flat_atomic_max_num_f32 v[1:2], v0
-; GFX12-SDAG:    s_endpgm
-;
-; GFX12-GISEL-LABEL: flat_atomic_fmax_num_f32_noret_saddr:
-; GFX12-GISEL:  ; %bb.0:
-; GFX12-GISEL:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
-; GFX12-GISEL:    flat_atomic_max_num_f32 v[1:2], v0
-; GFX12-GISEL:    s_endpgm
+; GFX12-LABEL: flat_atomic_fmax_num_f32_noret_saddr:
+; GFX12:  ; %bb.0:
+; GFX12:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
+; GFX12:    flat_atomic_max_num_f32 v[1:2], v0
+; GFX12:    s_endpgm
 ;
 ; GFX1250-LABEL: flat_atomic_fmax_num_f32_noret_saddr:
 ; GFX1250:  ; %bb.0:
+; GFX1250:    s_mov_b64 s[64:65], 0
+; GFX1250:    v_nop
+; GFX1250:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250:    v_mov_b32_e32 v1, 0
 ; GFX1250:    flat_atomic_max_num_f32 v1, v0, s[0:1]
 ; GFX1250:    s_endpgm
@@ -118,22 +140,18 @@ define amdgpu_ps void @flat_atomic_fmax_num_f32_noret_saddr(ptr inreg %ptr, floa
 }
 
 define amdgpu_ps float @flat_atomic_fmin_num_f32_rtn_saddr(ptr inreg %ptr, float %data) {
-; GFX12-SDAG-LABEL: flat_atomic_fmin_num_f32_rtn_saddr:
-; GFX12-SDAG:  ; %bb.0:
-; GFX12-SDAG:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
-; GFX12-SDAG:    flat_atomic_min_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
-; GFX12-SDAG:    s_wait_loadcnt_dscnt 0x0
-; GFX12-SDAG:    ; return to shader part epilog
-;
-; GFX12-GISEL-LABEL: flat_atomic_fmin_num_f32_rtn_saddr:
-; GFX12-GISEL:  ; %bb.0:
-; GFX12-GISEL:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
-; GFX12-GISEL:    flat_atomic_min_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
-; GFX12-GISEL:    s_wait_loadcnt_dscnt 0x0
-; GFX12-GISEL:    ; return to shader part epilog
+; GFX12-LABEL: flat_atomic_fmin_num_f32_rtn_saddr:
+; GFX12:  ; %bb.0:
+; GFX12:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
+; GFX12:    flat_atomic_min_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
+; GFX12:    s_wait_loadcnt_dscnt 0x0
+; GFX12:    ; return to shader part epilog
 ;
 ; GFX1250-LABEL: flat_atomic_fmin_num_f32_rtn_saddr:
 ; GFX1250:  ; %bb.0:
+; GFX1250:    s_mov_b64 s[64:65], 0
+; GFX1250:    v_nop
+; GFX1250:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250:    v_mov_b32_e32 v1, 0
 ; GFX1250:    flat_atomic_min_num_f32 v0, v1, v0, s[0:1] th:TH_ATOMIC_RETURN
 ; GFX1250:    s_wait_loadcnt_dscnt 0x0
@@ -143,22 +161,18 @@ define amdgpu_ps float @flat_atomic_fmin_num_f32_rtn_saddr(ptr inreg %ptr, float
 }
 
 define amdgpu_ps float @flat_atomic_fmax_num_f32_rtn_saddr(ptr inreg %ptr, float %data) {
-; GFX12-SDAG-LABEL: flat_atomic_fmax_num_f32_rtn_saddr:
-; GFX12-SDAG:  ; %bb.0:
-; GFX12-SDAG:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
-; GFX12-SDAG:    flat_atomic_max_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
-; GFX12-SDAG:    s_wait_loadcnt_dscnt 0x0
-; GFX12-SDAG:    ; return to shader part epilog
-;
-; GFX12-GISEL-LABEL: flat_atomic_fmax_num_f32_rtn_saddr:
-; GFX12-GISEL:  ; %bb.0:
-; GFX12-GISEL:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
-; GFX12-GISEL:    flat_atomic_max_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
-; GFX12-GISEL:    s_wait_loadcnt_dscnt 0x0
-; GFX12-GISEL:    ; return to shader part epilog
+; GFX12-LABEL: flat_atomic_fmax_num_f32_rtn_saddr:
+; GFX12:  ; %bb.0:
+; GFX12:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
+; GFX12:    flat_atomic_max_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
+; GFX12:    s_wait_loadcnt_dscnt 0x0
+; GFX12:    ; return to shader part epilog
 ;
 ; GFX1250-LABEL: flat_atomic_fmax_num_f32_rtn_saddr:
 ; GFX1250:  ; %bb.0:
+; GFX1250:    s_mov_b64 s[64:65], 0
+; GFX1250:    v_nop
+; GFX1250:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250:    v_mov_b32_e32 v1, 0
 ; GFX1250:    flat_atomic_max_num_f32 v0, v1, v0, s[0:1] th:TH_ATOMIC_RETURN
 ; GFX1250:    s_wait_loadcnt_dscnt 0x0
@@ -166,3 +180,7 @@ define amdgpu_ps float @flat_atomic_fmax_num_f32_rtn_saddr(ptr inreg %ptr, float
   %ret = call float @llvm.amdgcn.flat.atomic.fmax.num.f32.p1.f32(ptr %ptr, float %data)
   ret float %ret
 }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GCN: {{.*}}
+; GFX12-GISEL: {{.*}}
+; GFX12-SDAG: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/fp_to_sint.ll b/llvm/test/CodeGen/AMDGPU/fp_to_sint.ll
index 2618c670b8c23..5b5e6fb585914 100644
--- a/llvm/test/CodeGen/AMDGPU/fp_to_sint.ll
+++ b/llvm/test/CodeGen/AMDGPU/fp_to_sint.ll
@@ -548,9 +548,9 @@ define amdgpu_kernel void @fp_to_sint_v2i64(ptr addrspace(1) %out, <2 x float> %
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    v_trunc_f32_e32 v0, s2
 ; GFX11-GISEL-NEXT:    v_trunc_f32_e32 v2, s3
-; GFX11-GISEL-NEXT:    s_ashr_i32 s6, s2, 31
+; GFX11-GISEL-NEXT:    s_ashr_i32 s4, s2, 31
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-GISEL-NEXT:    s_mov_b32 s7, s6
+; GFX11-GISEL-NEXT:    s_mov_b32 s5, s4
 ; GFX11-GISEL-NEXT:    v_mul_f32_e64 v1, 0x2f800000, |v0|
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
 ; GFX11-GISEL-NEXT:    v_mul_f32_e64 v3, 0x2f800000, |v2|
@@ -564,24 +564,24 @@ define amdgpu_kernel void @fp_to_sint_v2i64(ptr addrspace(1) %out, <2 x float> %
 ; GFX11-GISEL-NEXT:    v_cvt_u32_f32_e32 v3, v3
 ; GFX11-GISEL-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s5, v1
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s7, v1
 ; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s9, v3
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s6, v0
 ; GFX11-GISEL-NEXT:    v_cvt_u32_f32_e32 v0, v2
-; GFX11-GISEL-NEXT:    s_xor_b64 s[4:5], s[4:5], s[6:7]
+; GFX11-GISEL-NEXT:    s_xor_b64 s[6:7], s[6:7], s[4:5]
 ; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
-; GFX11-GISEL-NEXT:    s_sub_u32 s4, s4, s6
-; GFX11-GISEL-NEXT:    s_subb_u32 s5, s5, s6
+; GFX11-GISEL-NEXT:    s_sub_u32 s6, s6, s4
+; GFX11-GISEL-NEXT:    s_subb_u32 s7, s7, s4
 ; GFX11-GISEL-NEXT:    s_ashr_i32 s2, s3, 31
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
 ; GFX11-GISEL-NEXT:    s_mov_b32 s3, s2
-; GFX11-GISEL-NEXT:    s_xor_b64 s[6:7], s[8:9], s[2:3]
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-GISEL-NEXT:    s_xor_b64 s[4:5], s[8:9], s[2:3]
+; GFX11-GISEL-NEXT:    s_sub_u32 s3, s4, s2
+; GFX11-GISEL-NEXT:    s_subb_u32 s2, s5, s2
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-GISEL-NEXT:    s_sub_u32 s6, s6, s2
-; GFX11-GISEL-NEXT:    s_subb_u32 s7, s7, s2
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, s6 :: v_dual_mov_b32 v3, s7
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, s3 :: v_dual_mov_b32 v3, s2
 ; GFX11-GISEL-NEXT:    global_store_b128 v4, v[0:3], s[0:1]
 ; GFX11-GISEL-NEXT:    s_endpgm
 ;
@@ -885,35 +885,34 @@ define amdgpu_kernel void @fp_to_sint_v4i64(ptr addrspace(1) %out, <4 x float> %
 ; GFX11-GISEL-NEXT:    s_xor_b64 s[8:9], s[8:9], s[6:7]
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3)
 ; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s10, v0
-; GFX11-GISEL-NEXT:    s_sub_u32 s8, s8, s6
-; GFX11-GISEL-NEXT:    s_subb_u32 s9, s9, s6
+; GFX11-GISEL-NEXT:    s_sub_u32 s12, s8, s6
+; GFX11-GISEL-NEXT:    s_subb_u32 s13, s9, s6
 ; GFX11-GISEL-NEXT:    s_ashr_i32 s0, s1, 31
 ; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s7, v6
 ; GFX11-GISEL-NEXT:    s_mov_b32 s1, s0
 ; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s6, v2
-; GFX11-GISEL-NEXT:    s_xor_b64 s[10:11], s[10:11], s[0:1]
+; GFX11-GISEL-NEXT:    s_xor_b64 s[8:9], s[10:11], s[0:1]
 ; GFX11-GISEL-NEXT:    v_cvt_u32_f32_e32 v0, v4
-; GFX11-GISEL-NEXT:    s_sub_u32 s10, s10, s0
-; GFX11-GISEL-NEXT:    s_subb_u32 s11, s11, s0
-; GFX11-GISEL-NEXT:    s_ashr_i32 s12, s2, 31
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s15, v1
-; GFX11-GISEL-NEXT:    s_mov_b32 s13, s12
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s14, v0
-; GFX11-GISEL-NEXT:    s_xor_b64 s[0:1], s[6:7], s[12:13]
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s8 :: v_dual_mov_b32 v1, s9
-; GFX11-GISEL-NEXT:    s_sub_u32 s0, s0, s12
-; GFX11-GISEL-NEXT:    s_subb_u32 s1, s1, s12
-; GFX11-GISEL-NEXT:    s_ashr_i32 s6, s3, 31
+; GFX11-GISEL-NEXT:    s_sub_u32 s10, s8, s0
+; GFX11-GISEL-NEXT:    s_subb_u32 s11, s9, s0
+; GFX11-GISEL-NEXT:    s_ashr_i32 s0, s2, 31
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s9, v1
+; GFX11-GISEL-NEXT:    s_mov_b32 s1, s0
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
+; GFX11-GISEL-NEXT:    s_xor_b64 s[6:7], s[6:7], s[0:1]
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s12 :: v_dual_mov_b32 v1, s13
+; GFX11-GISEL-NEXT:    s_sub_u32 s6, s6, s0
+; GFX11-GISEL-NEXT:    s_subb_u32 s7, s7, s0
+; GFX11-GISEL-NEXT:    s_ashr_i32 s0, s3, 31
 ; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, s10 :: v_dual_mov_b32 v3, s11
-; GFX11-GISEL-NEXT:    s_mov_b32 s7, s6
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11-GISEL-NEXT:    s_xor_b64 s[2:3], s[14:15], s[6:7]
-; GFX11-GISEL-NEXT:    s_sub_u32 s2, s2, s6
-; GFX11-GISEL-NEXT:    s_subb_u32 s3, s3, s6
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v8, 0 :: v_dual_mov_b32 v7, s3
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v6, s2 :: v_dual_mov_b32 v5, s1
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v4, s0
+; GFX11-GISEL-NEXT:    s_mov_b32 s1, s0
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v8, 0 :: v_dual_mov_b32 v5, s7
+; GFX11-GISEL-NEXT:    s_xor_b64 s[2:3], s[8:9], s[0:1]
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-GISEL-NEXT:    s_sub_u32 s1, s2, s0
+; GFX11-GISEL-NEXT:    s_subb_u32 s0, s3, s0
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v7, s0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v6, s1
 ; GFX11-GISEL-NEXT:    s_clause 0x1
 ; GFX11-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[4:5]
 ; GFX11-GISEL-NEXT:    global_store_b128 v8, v[4:7], s[4:5] offset:16
diff --git a/llvm/test/CodeGen/AMDGPU/fp_to_uint.ll b/llvm/test/CodeGen/AMDGPU/fp_to_uint.ll
index 2fb9e0b764391..f1a6c4d6390e6 100644
--- a/llvm/test/CodeGen/AMDGPU/fp_to_uint.ll
+++ b/llvm/test/CodeGen/AMDGPU/fp_to_uint.ll
@@ -429,16 +429,16 @@ define amdgpu_kernel void @fp_to_uint_v2f32_to_v2i64(ptr addrspace(1) %out, <2 x
 ; GFX11-GISEL-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX11-GISEL-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s5, v2
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s7, v3
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
 ; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s6, v1
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, s5
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v3, s7 :: v_dual_mov_b32 v0, s4
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v0, s4
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3)
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, s6
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX11-GISEL-NEXT:    global_store_b128 v4, v[0:3], s[0:1]
 ; GFX11-GISEL-NEXT:    s_endpgm
 ;
@@ -643,14 +643,13 @@ define amdgpu_kernel void @fp_to_uint_v4f32_to_v4i64(ptr addrspace(1) %out, <4 x
 ;
 ; GFX11-GISEL-LABEL: fp_to_uint_v4f32_to_v4i64:
 ; GFX11-GISEL:       ; %bb.0:
-; GFX11-GISEL-NEXT:    s_clause 0x1
 ; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[4:5], 0x34
-; GFX11-GISEL-NEXT:    s_load_b64 s[8:9], s[4:5], 0x24
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    v_trunc_f32_e32 v0, s0
 ; GFX11-GISEL-NEXT:    v_trunc_f32_e32 v1, s1
 ; GFX11-GISEL-NEXT:    v_trunc_f32_e32 v2, s2
 ; GFX11-GISEL-NEXT:    v_trunc_f32_e32 v3, s3
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
 ; GFX11-GISEL-NEXT:    v_dual_mul_f32 v4, 0x2f800000, v0 :: v_dual_mul_f32 v5, 0x2f800000, v1
 ; GFX11-GISEL-NEXT:    v_dual_mul_f32 v6, 0x2f800000, v2 :: v_dual_mul_f32 v7, 0x2f800000, v3
@@ -672,22 +671,23 @@ define amdgpu_kernel void @fp_to_uint_v4f32_to_v4i64(ptr addrspace(1) %out, <4 x
 ; GFX11-GISEL-NEXT:    v_cvt_u32_f32_e32 v7, v7
 ; GFX11-GISEL-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; GFX11-GISEL-NEXT:    v_cvt_u32_f32_e32 v3, v3
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s2, v4
 ; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s3, v5
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s2, v1
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s5, v6
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s7, v7
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v8, 0 :: v_dual_mov_b32 v1, s1
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v5, s5
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v7, s7
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v6, s6
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s6, v0
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s7, v1
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v6
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s5, v7
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s8, v2
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s9, v3
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v8, 0 :: v_dual_mov_b32 v1, s2
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v0, s6
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, s7 :: v_dual_mov_b32 v5, s4
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v7, s5 :: v_dual_mov_b32 v4, s8
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v6, s9
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    s_clause 0x1
-; GFX11-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[8:9]
-; GFX11-GISEL-NEXT:    global_store_b128 v8, v[4:7], s[8:9] offset:16
+; GFX11-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1]
+; GFX11-GISEL-NEXT:    global_store_b128 v8, v[4:7], s[0:1] offset:16
 ; GFX11-GISEL-NEXT:    s_endpgm
 ;
 ; EG-LABEL: fp_to_uint_v4f32_to_v4i64:
diff --git a/llvm/test/CodeGen/AMDGPU/frem.ll b/llvm/test/CodeGen/AMDGPU/frem.ll
index 4134420e78a2e..a4c389b08422d 100644
--- a/llvm/test/CodeGen/AMDGPU/frem.ll
+++ b/llvm/test/CodeGen/AMDGPU/frem.ll
@@ -1416,8 +1416,8 @@ define amdgpu_kernel void @fast_frem_f16(ptr addrspace(1) %out, ptr addrspace(1)
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
 ; VI-NEXT:    flat_load_ushort v4, v[2:3]
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_load_ushort v2, v[2:3]
 ; VI-NEXT:    s_waitcnt vmcnt(1)
 ; VI-NEXT:    v_cvt_f32_f16_e32 v3, v4
@@ -1899,8 +1899,8 @@ define amdgpu_kernel void @unsafe_frem_f16(ptr addrspace(1) %out, ptr addrspace(
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
 ; VI-NEXT:    flat_load_ushort v4, v[2:3]
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_load_ushort v2, v[2:3]
 ; VI-NEXT:    s_waitcnt vmcnt(1)
 ; VI-NEXT:    v_cvt_f32_f16_e32 v3, v4
@@ -3133,8 +3133,8 @@ define amdgpu_kernel void @fast_frem_f32(ptr addrspace(1) %out, ptr addrspace(1)
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
 ; VI-NEXT:    flat_load_dword v4, v[2:3]
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_load_dword v2, v[2:3]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_div_scale_f32 v5, s[0:1], v2, v2, v4
@@ -3411,8 +3411,8 @@ define amdgpu_kernel void @unsafe_frem_f32(ptr addrspace(1) %out, ptr addrspace(
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
 ; VI-NEXT:    flat_load_dword v4, v[2:3]
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_load_dword v2, v[2:3]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_div_scale_f32 v5, s[0:1], v2, v2, v4
@@ -14083,8 +14083,8 @@ define amdgpu_kernel void @frem_v4f32(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; VI-NEXT:    s_add_u32 s2, s4, 64
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    s_addc_u32 s3, s5, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; VI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
@@ -16909,8 +16909,8 @@ define amdgpu_kernel void @frem_v2f64(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; VI-NEXT:    s_add_u32 s2, s4, 64
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    s_addc_u32 s3, s5, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; VI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/fsqrt.f32.ll b/llvm/test/CodeGen/AMDGPU/fsqrt.f32.ll
index ea61e1122c6a9..1790832c9198c 100644
--- a/llvm/test/CodeGen/AMDGPU/fsqrt.f32.ll
+++ b/llvm/test/CodeGen/AMDGPU/fsqrt.f32.ll
@@ -4542,13 +4542,13 @@ define amdgpu_kernel void @elim_redun_check_v2(ptr addrspace(1) %out, <2 x float
 ; GISEL-IEEE-NEXT:    v_cmp_lt_f32_e32 vcc, s3, v0
 ; GISEL-IEEE-NEXT:    s_cselect_b32 s4, 1, 0
 ; GISEL-IEEE-NEXT:    s_or_b64 s[2:3], vcc, vcc
-; GISEL-IEEE-NEXT:    s_cselect_b32 s3, 1, 0
+; GISEL-IEEE-NEXT:    s_cselect_b32 s2, 1, 0
 ; GISEL-IEEE-NEXT:    s_cmp_lg_u32 s4, 0
-; GISEL-IEEE-NEXT:    s_cselect_b32 s2, 0x7fc00000, s6
-; GISEL-IEEE-NEXT:    s_cmp_lg_u32 s3, 0
-; GISEL-IEEE-NEXT:    s_cselect_b32 s3, 0x7fc00000, s7
-; GISEL-IEEE-NEXT:    v_mov_b32_e32 v0, s2
-; GISEL-IEEE-NEXT:    v_mov_b32_e32 v1, s3
+; GISEL-IEEE-NEXT:    s_cselect_b32 s3, 0x7fc00000, s6
+; GISEL-IEEE-NEXT:    s_cmp_lg_u32 s2, 0
+; GISEL-IEEE-NEXT:    s_cselect_b32 s2, 0x7fc00000, s7
+; GISEL-IEEE-NEXT:    v_mov_b32_e32 v0, s3
+; GISEL-IEEE-NEXT:    v_mov_b32_e32 v1, s2
 ; GISEL-IEEE-NEXT:    s_mov_b32 s2, -1
 ; GISEL-IEEE-NEXT:    s_mov_b32 s3, 0xf000
 ; GISEL-IEEE-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
@@ -4653,13 +4653,13 @@ define amdgpu_kernel void @elim_redun_check_v2(ptr addrspace(1) %out, <2 x float
 ; GISEL-DAZ-NEXT:    v_cmp_lt_f32_e32 vcc, s3, v0
 ; GISEL-DAZ-NEXT:    s_cselect_b32 s4, 1, 0
 ; GISEL-DAZ-NEXT:    s_or_b64 s[2:3], vcc, vcc
-; GISEL-DAZ-NEXT:    s_cselect_b32 s3, 1, 0
+; GISEL-DAZ-NEXT:    s_cselect_b32 s2, 1, 0
 ; GISEL-DAZ-NEXT:    s_cmp_lg_u32 s4, 0
-; GISEL-DAZ-NEXT:    s_cselect_b32 s2, 0x7fc00000, s6
-; GISEL-DAZ-NEXT:    s_cmp_lg_u32 s3, 0
-; GISEL-DAZ-NEXT:    s_cselect_b32 s3, 0x7fc00000, s7
-; GISEL-DAZ-NEXT:    v_mov_b32_e32 v0, s2
-; GISEL-DAZ-NEXT:    v_mov_b32_e32 v1, s3
+; GISEL-DAZ-NEXT:    s_cselect_b32 s3, 0x7fc00000, s6
+; GISEL-DAZ-NEXT:    s_cmp_lg_u32 s2, 0
+; GISEL-DAZ-NEXT:    s_cselect_b32 s2, 0x7fc00000, s7
+; GISEL-DAZ-NEXT:    v_mov_b32_e32 v0, s3
+; GISEL-DAZ-NEXT:    v_mov_b32_e32 v1, s2
 ; GISEL-DAZ-NEXT:    s_mov_b32 s2, -1
 ; GISEL-DAZ-NEXT:    s_mov_b32 s3, 0xf000
 ; GISEL-DAZ-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
@@ -4789,13 +4789,13 @@ define amdgpu_kernel void @elim_redun_check_v2_ult(ptr addrspace(1) %out, <2 x f
 ; GISEL-IEEE-NEXT:    v_cmp_nge_f32_e32 vcc, s3, v0
 ; GISEL-IEEE-NEXT:    s_cselect_b32 s4, 1, 0
 ; GISEL-IEEE-NEXT:    s_or_b64 s[2:3], vcc, vcc
-; GISEL-IEEE-NEXT:    s_cselect_b32 s3, 1, 0
+; GISEL-IEEE-NEXT:    s_cselect_b32 s2, 1, 0
 ; GISEL-IEEE-NEXT:    s_cmp_lg_u32 s4, 0
-; GISEL-IEEE-NEXT:    s_cselect_b32 s2, 0x7fc00000, s6
-; GISEL-IEEE-NEXT:    s_cmp_lg_u32 s3, 0
-; GISEL-IEEE-NEXT:    s_cselect_b32 s3, 0x7fc00000, s7
-; GISEL-IEEE-NEXT:    v_mov_b32_e32 v0, s2
-; GISEL-IEEE-NEXT:    v_mov_b32_e32 v1, s3
+; GISEL-IEEE-NEXT:    s_cselect_b32 s3, 0x7fc00000, s6
+; GISEL-IEEE-NEXT:    s_cmp_lg_u32 s2, 0
+; GISEL-IEEE-NEXT:    s_cselect_b32 s2, 0x7fc00000, s7
+; GISEL-IEEE-NEXT:    v_mov_b32_e32 v0, s3
+; GISEL-IEEE-NEXT:    v_mov_b32_e32 v1, s2
 ; GISEL-IEEE-NEXT:    s_mov_b32 s2, -1
 ; GISEL-IEEE-NEXT:    s_mov_b32 s3, 0xf000
 ; GISEL-IEEE-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
@@ -4900,13 +4900,13 @@ define amdgpu_kernel void @elim_redun_check_v2_ult(ptr addrspace(1) %out, <2 x f
 ; GISEL-DAZ-NEXT:    v_cmp_nge_f32_e32 vcc, s3, v0
 ; GISEL-DAZ-NEXT:    s_cselect_b32 s4, 1, 0
 ; GISEL-DAZ-NEXT:    s_or_b64 s[2:3], vcc, vcc
-; GISEL-DAZ-NEXT:    s_cselect_b32 s3, 1, 0
+; GISEL-DAZ-NEXT:    s_cselect_b32 s2, 1, 0
 ; GISEL-DAZ-NEXT:    s_cmp_lg_u32 s4, 0
-; GISEL-DAZ-NEXT:    s_cselect_b32 s2, 0x7fc00000, s6
-; GISEL-DAZ-NEXT:    s_cmp_lg_u32 s3, 0
-; GISEL-DAZ-NEXT:    s_cselect_b32 s3, 0x7fc00000, s7
-; GISEL-DAZ-NEXT:    v_mov_b32_e32 v0, s2
-; GISEL-DAZ-NEXT:    v_mov_b32_e32 v1, s3
+; GISEL-DAZ-NEXT:    s_cselect_b32 s3, 0x7fc00000, s6
+; GISEL-DAZ-NEXT:    s_cmp_lg_u32 s2, 0
+; GISEL-DAZ-NEXT:    s_cselect_b32 s2, 0x7fc00000, s7
+; GISEL-DAZ-NEXT:    v_mov_b32_e32 v0, s3
+; GISEL-DAZ-NEXT:    v_mov_b32_e32 v1, s2
 ; GISEL-DAZ-NEXT:    s_mov_b32 s2, -1
 ; GISEL-DAZ-NEXT:    s_mov_b32 s3, 0xf000
 ; GISEL-DAZ-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
diff --git a/llvm/test/CodeGen/AMDGPU/gfx-callable-return-types.ll b/llvm/test/CodeGen/AMDGPU/gfx-callable-return-types.ll
index f805b11d35db5..ed48489be4cf5 100644
--- a/llvm/test/CodeGen/AMDGPU/gfx-callable-return-types.ll
+++ b/llvm/test/CodeGen/AMDGPU/gfx-callable-return-types.ll
@@ -2015,12 +2015,10 @@ define amdgpu_gfx <512 x i32> @return_512xi32() #0 {
 ; GFX11-LABEL: return_512xi32:
 ; GFX11:       ; %bb.0: ; %entry
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT:    s_mov_b32 s0, 0
-; GFX11-NEXT:    s_mov_b32 s1, s0
-; GFX11-NEXT:    s_mov_b32 s2, s0
-; GFX11-NEXT:    s_mov_b32 s3, s0
-; GFX11-NEXT:    v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
-; GFX11-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    v_mov_b32_e32 v2, v1
+; GFX11-NEXT:    v_mov_b32_e32 v3, v1
+; GFX11-NEXT:    v_mov_b32_e32 v4, v1
 ; GFX11-NEXT:    s_clause 0x1f
 ; GFX11-NEXT:    scratch_store_b128 v0, v[1:4], off offset:2032
 ; GFX11-NEXT:    scratch_store_b128 v0, v[1:4], off offset:2016
@@ -2222,14 +2220,14 @@ define amdgpu_gfx void @call_512xi32() #0 {
 ; GFX11-LABEL: call_512xi32:
 ; GFX11:       ; %bb.0: ; %entry
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT:    s_mov_b32 s35, s33
+; GFX11-NEXT:    s_mov_b32 s2, s33
 ; GFX11-NEXT:    s_add_i32 s33, s32, 0x7ff
 ; GFX11-NEXT:    s_and_b32 s33, s33, 0xfffff800
 ; GFX11-NEXT:    s_xor_saveexec_b32 s0, -1
 ; GFX11-NEXT:    scratch_store_b32 off, v5, s33 offset:2048 ; 4-byte Folded Spill
 ; GFX11-NEXT:    s_mov_b32 exec_lo, s0
 ; GFX11-NEXT:    v_writelane_b32 v5, s30, 0
-; GFX11-NEXT:    s_mov_b32 s36, s34
+; GFX11-NEXT:    s_mov_b32 s3, s34
 ; GFX11-NEXT:    s_mov_b32 s34, s32
 ; GFX11-NEXT:    s_addk_i32 s32, 0x1800
 ; GFX11-NEXT:    v_writelane_b32 v5, s31, 1
@@ -2240,11 +2238,11 @@ define amdgpu_gfx void @call_512xi32() #0 {
 ; GFX11-NEXT:    v_readlane_b32 s30, v5, 0
 ; GFX11-NEXT:    v_readlane_b32 s31, v5, 1
 ; GFX11-NEXT:    s_mov_b32 s32, s34
-; GFX11-NEXT:    s_mov_b32 s34, s36
+; GFX11-NEXT:    s_mov_b32 s34, s3
 ; GFX11-NEXT:    s_xor_saveexec_b32 s0, -1
 ; GFX11-NEXT:    scratch_load_b32 v5, off, s33 offset:2048 ; 4-byte Folded Reload
 ; GFX11-NEXT:    s_mov_b32 exec_lo, s0
-; GFX11-NEXT:    s_mov_b32 s33, s35
+; GFX11-NEXT:    s_mov_b32 s33, s2
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    s_setpc_b64 s[30:31]
 entry:
@@ -3258,15 +3256,15 @@ define amdgpu_gfx void @call_72xi32() #1 {
 ; GFX11-NEXT:    scratch_store_b32 off, v61, s33
 ; GFX11-NEXT:    v_writelane_b32 v62, s30, 0
 ; GFX11-NEXT:    v_writelane_b32 v62, s31, 1
-; GFX11-NEXT:    s_mov_b32 s0, 0
-; GFX11-NEXT:    v_mov_b32_e32 v4, 0
-; GFX11-NEXT:    s_mov_b32 s1, s0
-; GFX11-NEXT:    s_mov_b32 s2, s0
-; GFX11-NEXT:    s_mov_b32 s3, s0
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-NEXT:    s_add_i32 s0, s32, 0xa0
 ; GFX11-NEXT:    s_add_i32 s1, s32, 0x90
+; GFX11-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v9, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v4, 0
+; GFX11-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-NEXT:    v_dual_mov_b32 v3, v0 :: v_dual_mov_b32 v8, 0
+; GFX11-NEXT:    v_mov_b32_e32 v11, 0
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s32
 ; GFX11-NEXT:    scratch_store_b32 off, v4, s0
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s1
@@ -3287,12 +3285,10 @@ define amdgpu_gfx void @call_72xi32() #1 {
 ; GFX11-NEXT:    s_add_i32 s2, s33, 0x200
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s0
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], s1
-; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v3, 0
-; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
-; GFX11-NEXT:    v_dual_mov_b32 v5, 0 :: v_dual_mov_b32 v4, 0
-; GFX11-NEXT:    v_dual_mov_b32 v7, 0 :: v_dual_mov_b32 v6, 0
-; GFX11-NEXT:    v_dual_mov_b32 v9, 0 :: v_dual_mov_b32 v8, 0
-; GFX11-NEXT:    v_dual_mov_b32 v11, 0 :: v_dual_mov_b32 v10, 0
+; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, 0
+; GFX11-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v5, 0
+; GFX11-NEXT:    v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v4, 0
+; GFX11-NEXT:    v_dual_mov_b32 v7, 0 :: v_dual_mov_b32 v10, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v13, 0 :: v_dual_mov_b32 v12, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v15, 0 :: v_dual_mov_b32 v14, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v17, 0 :: v_dual_mov_b32 v16, 0
diff --git a/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll b/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
index a1130abf89a7a..bde2505b35309 100644
--- a/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
@@ -1938,9 +1938,8 @@ define amdgpu_ps float @global_load_i8_vgpr64_sgpr32(ptr addrspace(1) %vbase, i3
 ;
 ; GFX12-GISEL-LABEL: global_load_i8_vgpr64_sgpr32:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0
 ; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX12-GISEL-NEXT:    v_add_co_u32 v0, vcc, v0, v2
 ; GFX12-GISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc
@@ -1995,9 +1994,8 @@ define amdgpu_ps float @global_load_i8_vgpr64_sgpr32_offset_4095(ptr addrspace(1
 ;
 ; GFX12-GISEL-LABEL: global_load_i8_vgpr64_sgpr32_offset_4095:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0
 ; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s2
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX12-GISEL-NEXT:    v_add_co_u32 v0, vcc, v0, v2
 ; GFX12-GISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics.ll b/llvm/test/CodeGen/AMDGPU/global_atomics.ll
index fdc8eff962d24..17beec185fca0 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics.ll
@@ -5468,8 +5468,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    s_add_u32 s0, s0, 16
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_cmpswap v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -5557,8 +5557,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64_offset(ptr addrspace(1)
 ; VI-NEXT:    s_add_u32 s0, s0, 16
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, s9
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_cmpswap v0, v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -5781,8 +5781,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_addr64(ptr addrspace(1) %out, i32
 ; VI-NEXT:    s_add_u32 s0, s2, s0
 ; VI-NEXT:    s_addc_u32 s1, s3, s1
 ; VI-NEXT:    v_mov_b32_e32 v1, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_cmpswap v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -5867,8 +5867,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i32_ret_addr64(ptr addrspace(1) %out,
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    v_mov_b32_e32 v1, s9
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_cmpswap v0, v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll
index 0ad48c2d6b683..48c4e21265440 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll
@@ -162,8 +162,8 @@ define amdgpu_kernel void @atomic_add_i64_addr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_add_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -240,8 +240,8 @@ define amdgpu_kernel void @atomic_add_i64_ret_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_add_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -448,8 +448,8 @@ define amdgpu_kernel void @atomic_add_i64_addr64(ptr addrspace(1) %out, i64 %in,
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_add_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -523,8 +523,8 @@ define amdgpu_kernel void @atomic_add_i64_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_add_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -730,8 +730,8 @@ define amdgpu_kernel void @atomic_and_i64_addr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_and_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -808,8 +808,8 @@ define amdgpu_kernel void @atomic_and_i64_ret_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_and_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -1016,8 +1016,8 @@ define amdgpu_kernel void @atomic_and_i64_addr64(ptr addrspace(1) %out, i64 %in,
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_and_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -1091,8 +1091,8 @@ define amdgpu_kernel void @atomic_and_i64_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_and_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -1298,8 +1298,8 @@ define amdgpu_kernel void @atomic_sub_i64_addr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_sub_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -1376,8 +1376,8 @@ define amdgpu_kernel void @atomic_sub_i64_ret_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_sub_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -1584,8 +1584,8 @@ define amdgpu_kernel void @atomic_sub_i64_addr64(ptr addrspace(1) %out, i64 %in,
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_sub_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -1659,8 +1659,8 @@ define amdgpu_kernel void @atomic_sub_i64_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_sub_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -1855,8 +1855,8 @@ define amdgpu_kernel void @atomic_max_i64_addr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_smax_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -1928,8 +1928,8 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_smax_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_mov_b32 s7, 0xf000
 ; VI-NEXT:    s_mov_b32 s6, -1
@@ -2123,8 +2123,8 @@ define amdgpu_kernel void @atomic_max_i64_addr64(ptr addrspace(1) %out, i64 %in,
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_smax_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -2193,8 +2193,8 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_smax_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_mov_b32 s7, 0xf000
 ; VI-NEXT:    s_mov_b32 s6, -1
@@ -2387,8 +2387,8 @@ define amdgpu_kernel void @atomic_umax_i64_addr64_offset(ptr addrspace(1) %out,
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_umax_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -2460,8 +2460,8 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64_offset(ptr addrspace(1) %o
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_umax_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_mov_b32 s7, 0xf000
 ; VI-NEXT:    s_mov_b32 s6, -1
@@ -2655,8 +2655,8 @@ define amdgpu_kernel void @atomic_umax_i64_addr64(ptr addrspace(1) %out, i64 %in
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_umax_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -2725,8 +2725,8 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_umax_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_mov_b32 s7, 0xf000
 ; VI-NEXT:    s_mov_b32 s6, -1
@@ -2919,8 +2919,8 @@ define amdgpu_kernel void @atomic_min_i64_addr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_smin_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -2992,8 +2992,8 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_smin_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_mov_b32 s7, 0xf000
 ; VI-NEXT:    s_mov_b32 s6, -1
@@ -3187,8 +3187,8 @@ define amdgpu_kernel void @atomic_min_i64_addr64(ptr addrspace(1) %out, i64 %in,
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_smin_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -3257,8 +3257,8 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_smin_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_mov_b32 s7, 0xf000
 ; VI-NEXT:    s_mov_b32 s6, -1
@@ -3451,8 +3451,8 @@ define amdgpu_kernel void @atomic_umin_i64_addr64_offset(ptr addrspace(1) %out,
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_umin_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -3524,8 +3524,8 @@ define amdgpu_kernel void @atomic_umin_i64_ret_addr64_offset(ptr addrspace(1) %o
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_umin_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_mov_b32 s7, 0xf000
 ; VI-NEXT:    s_mov_b32 s6, -1
@@ -3719,8 +3719,8 @@ define amdgpu_kernel void @atomic_umin_i64_addr64(ptr addrspace(1) %out, i64 %in
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_umin_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -3789,8 +3789,8 @@ define amdgpu_kernel void @atomic_umin_i64_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_umin_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_mov_b32 s7, 0xf000
 ; VI-NEXT:    s_mov_b32 s6, -1
@@ -3994,8 +3994,8 @@ define amdgpu_kernel void @atomic_or_i64_addr64_offset(ptr addrspace(1) %out, i6
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_or_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -4072,8 +4072,8 @@ define amdgpu_kernel void @atomic_or_i64_ret_addr64_offset(ptr addrspace(1) %out
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_or_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -4280,8 +4280,8 @@ define amdgpu_kernel void @atomic_or_i64_addr64(ptr addrspace(1) %out, i64 %in,
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_or_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -4355,8 +4355,8 @@ define amdgpu_kernel void @atomic_or_i64_ret_addr64(ptr addrspace(1) %out, ptr a
 ; VI-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_or_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -4672,8 +4672,8 @@ define amdgpu_kernel void @atomic_xchg_i64_addr64_offset(ptr addrspace(1) %out,
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -4750,8 +4750,8 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_addr64_offset(ptr addrspace(1) %o
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_swap_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -4958,8 +4958,8 @@ define amdgpu_kernel void @atomic_xchg_i64_addr64(ptr addrspace(1) %out, i64 %in
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_swap_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -5033,8 +5033,8 @@ define amdgpu_kernel void @atomic_xchg_i64_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_swap_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -5240,8 +5240,8 @@ define amdgpu_kernel void @atomic_xor_i64_addr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_xor_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -5318,8 +5318,8 @@ define amdgpu_kernel void @atomic_xor_i64_ret_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_xor_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -5526,8 +5526,8 @@ define amdgpu_kernel void @atomic_xor_i64_addr64(ptr addrspace(1) %out, i64 %in,
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_xor_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -5601,8 +5601,8 @@ define amdgpu_kernel void @atomic_xor_i64_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    s_lshl_b64 s[4:5], s[6:7], 3
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_xor_x2 v[0:1], v[2:3], v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -5903,8 +5903,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    v_mov_b32_e32 v2, s6
 ; VI-NEXT:    v_mov_b32_e32 v3, s7
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    flat_atomic_cmpswap_x2 v[4:5], v[0:3]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -5981,14 +5981,14 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64_offset(ptr addrspace(1)
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[14:15], 3
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    s_add_u32 s0, s8, s2
-; VI-NEXT:    s_addc_u32 s3, s9, s3
-; VI-NEXT:    s_add_u32 s2, s0, 32
-; VI-NEXT:    s_addc_u32 s3, s3, 0
+; VI-NEXT:    s_addc_u32 s2, s9, s3
+; VI-NEXT:    s_add_u32 s0, s0, 32
+; VI-NEXT:    s_addc_u32 s2, s2, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s12
 ; VI-NEXT:    v_mov_b32_e32 v1, s13
 ; VI-NEXT:    v_mov_b32_e32 v3, s1
-; VI-NEXT:    v_mov_b32_e32 v5, s3
-; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s2
 ; VI-NEXT:    flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -6220,8 +6220,8 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_addr64(ptr addrspace(1) %out, i64
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    v_mov_b32_e32 v2, s6
 ; VI-NEXT:    v_mov_b32_e32 v3, s7
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    flat_atomic_cmpswap_x2 v[4:5], v[0:3]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -6295,14 +6295,14 @@ define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64(ptr addrspace(1) %out,
 ; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x44
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_lshl_b64 s[2:3], s[14:15], 3
-; VI-NEXT:    s_add_u32 s2, s8, s2
-; VI-NEXT:    s_addc_u32 s3, s9, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    s_add_u32 s0, s8, s2
+; VI-NEXT:    s_addc_u32 s2, s9, s3
 ; VI-NEXT:    v_mov_b32_e32 v0, s12
 ; VI-NEXT:    v_mov_b32_e32 v1, s13
-; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    v_mov_b32_e32 v3, s1
-; VI-NEXT:    v_mov_b32_e32 v5, s3
-; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s2
 ; VI-NEXT:    flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -6807,8 +6807,8 @@ define amdgpu_kernel void @atomic_store_i64_offset(i64 %in, ptr addrspace(1) %ou
 ; VI-NEXT:    s_add_u32 s0, s2, 32
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-NEXT:    s_addc_u32 s1, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -6917,8 +6917,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64_offset(i64 %in, ptr addrspace
 ; VI-NEXT:    s_addc_u32 s1, s3, s1
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -6985,8 +6985,8 @@ define amdgpu_kernel void @atomic_store_i64_addr64(i64 %in, ptr addrspace(1) %ou
 ; VI-NEXT:    s_lshl_b64 s[0:1], s[4:5], 3
 ; VI-NEXT:    s_add_u32 s0, s2, s0
 ; VI-NEXT:    s_addc_u32 s1, s3, s1
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -7054,8 +7054,8 @@ define amdgpu_kernel void @atomic_store_f64_addr64_offset(double %in, ptr addrsp
 ; VI-NEXT:    s_addc_u32 s1, s3, s1
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
 ;
@@ -7253,8 +7253,8 @@ define amdgpu_kernel void @atomic_inc_i64_incr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
@@ -7456,8 +7456,8 @@ define amdgpu_kernel void @atomic_dec_i64_decr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
diff --git a/llvm/test/CodeGen/AMDGPU/half.ll b/llvm/test/CodeGen/AMDGPU/half.ll
index 8f274a55c53b8..c7583dc68e147 100644
--- a/llvm/test/CodeGen/AMDGPU/half.ll
+++ b/llvm/test/CodeGen/AMDGPU/half.ll
@@ -99,12 +99,12 @@ define amdgpu_kernel void @load_v3f16_arg(ptr addrspace(1) %out, <3 x half> %arg
 ; CIVI-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; CIVI-NEXT:    s_waitcnt lgkmcnt(0)
 ; CIVI-NEXT:    s_add_u32 s4, s0, 4
-; CIVI-NEXT:    s_addc_u32 s5, s1, 0
-; CIVI-NEXT:    v_mov_b32_e32 v4, s3
-; CIVI-NEXT:    v_mov_b32_e32 v2, s4
-; CIVI-NEXT:    v_mov_b32_e32 v3, s5
 ; CIVI-NEXT:    v_mov_b32_e32 v0, s0
 ; CIVI-NEXT:    v_mov_b32_e32 v1, s1
+; CIVI-NEXT:    s_addc_u32 s0, s1, 0
+; CIVI-NEXT:    v_mov_b32_e32 v4, s3
+; CIVI-NEXT:    v_mov_b32_e32 v2, s4
+; CIVI-NEXT:    v_mov_b32_e32 v3, s0
 ; CIVI-NEXT:    v_mov_b32_e32 v5, s2
 ; CIVI-NEXT:    flat_store_short v[2:3], v4
 ; CIVI-NEXT:    flat_store_dword v[0:1], v5
@@ -474,8 +474,8 @@ define amdgpu_kernel void @extload_v8f16_to_v8f32_arg(ptr addrspace(1) %out, <8
 ; CI-NEXT:    v_cvt_f32_f16_e32 v0, s0
 ; CI-NEXT:    s_add_u32 s0, s4, 16
 ; CI-NEXT:    s_addc_u32 s1, s5, 0
-; CI-NEXT:    v_mov_b32_e32 v9, s1
 ; CI-NEXT:    v_mov_b32_e32 v8, s0
+; CI-NEXT:    v_mov_b32_e32 v9, s1
 ; CI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; CI-NEXT:    s_nop 0
 ; CI-NEXT:    v_mov_b32_e32 v4, s4
@@ -505,8 +505,8 @@ define amdgpu_kernel void @extload_v8f16_to_v8f32_arg(ptr addrspace(1) %out, <8
 ; VI-NEXT:    v_cvt_f32_f16_e32 v0, s0
 ; VI-NEXT:    s_add_u32 s0, s4, 16
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
-; VI-NEXT:    v_mov_b32_e32 v9, s1
 ; VI-NEXT:    v_mov_b32_e32 v8, s0
+; VI-NEXT:    v_mov_b32_e32 v9, s1
 ; VI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-NEXT:    s_nop 0
 ; VI-NEXT:    v_mov_b32_e32 v4, s4
@@ -668,8 +668,8 @@ define amdgpu_kernel void @extload_v3f16_to_v3f64_arg(ptr addrspace(1) %out, <3
 ; CIVI-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
 ; CIVI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v2
 ; CIVI-NEXT:    s_addc_u32 s3, s1, 0
-; CIVI-NEXT:    v_mov_b32_e32 v7, s3
 ; CIVI-NEXT:    v_mov_b32_e32 v6, s2
+; CIVI-NEXT:    v_mov_b32_e32 v7, s3
 ; CIVI-NEXT:    flat_store_dwordx2 v[6:7], v[4:5]
 ; CIVI-NEXT:    v_mov_b32_e32 v5, s1
 ; CIVI-NEXT:    v_mov_b32_e32 v4, s0
@@ -719,8 +719,8 @@ define amdgpu_kernel void @extload_v4f16_to_v4f64_arg(ptr addrspace(1) %out, <4
 ; CI-NEXT:    v_cvt_f64_f32_e32 v[4:5], v4
 ; CI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v6
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v9, s3
 ; CI-NEXT:    v_mov_b32_e32 v8, s2
+; CI-NEXT:    v_mov_b32_e32 v9, s3
 ; CI-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
 ; CI-NEXT:    s_nop 0
 ; CI-NEXT:    v_mov_b32_e32 v0, s0
@@ -747,8 +747,8 @@ define amdgpu_kernel void @extload_v4f16_to_v4f64_arg(ptr addrspace(1) %out, <4
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[4:5], v4
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v6
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-NEXT:    v_mov_b32_e32 v8, s2
+; VI-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
 ; VI-NEXT:    s_nop 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s0
@@ -810,23 +810,23 @@ define amdgpu_kernel void @extload_v8f16_to_v8f64_arg(ptr addrspace(1) %out, <8
 ; CI-NEXT:    s_add_u32 s0, s4, 48
 ; CI-NEXT:    s_addc_u32 s1, s5, 0
 ; CI-NEXT:    v_cvt_f64_f32_e32 v[8:9], v8
-; CI-NEXT:    v_mov_b32_e32 v17, s1
 ; CI-NEXT:    v_mov_b32_e32 v16, s0
 ; CI-NEXT:    s_add_u32 s0, s4, 32
 ; CI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v1
 ; CI-NEXT:    v_cvt_f64_f32_e32 v[4:5], v4
+; CI-NEXT:    v_mov_b32_e32 v17, s1
 ; CI-NEXT:    s_addc_u32 s1, s5, 0
 ; CI-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
 ; CI-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; CI-NEXT:    s_nop 0
-; CI-NEXT:    v_mov_b32_e32 v13, s1
 ; CI-NEXT:    v_mov_b32_e32 v12, s0
 ; CI-NEXT:    s_add_u32 s0, s4, 16
+; CI-NEXT:    v_mov_b32_e32 v13, s1
 ; CI-NEXT:    s_addc_u32 s1, s5, 0
 ; CI-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; CI-NEXT:    s_nop 0
-; CI-NEXT:    v_mov_b32_e32 v9, s1
 ; CI-NEXT:    v_mov_b32_e32 v8, s0
+; CI-NEXT:    v_mov_b32_e32 v9, s1
 ; CI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; CI-NEXT:    s_nop 0
 ; CI-NEXT:    v_mov_b32_e32 v4, s4
@@ -861,23 +861,23 @@ define amdgpu_kernel void @extload_v8f16_to_v8f64_arg(ptr addrspace(1) %out, <8
 ; VI-NEXT:    s_add_u32 s0, s4, 48
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[8:9], v8
-; VI-NEXT:    v_mov_b32_e32 v17, s1
 ; VI-NEXT:    v_mov_b32_e32 v16, s0
 ; VI-NEXT:    s_add_u32 s0, s4, 32
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v1
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[4:5], v4
+; VI-NEXT:    v_mov_b32_e32 v17, s1
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
 ; VI-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-NEXT:    s_nop 0
-; VI-NEXT:    v_mov_b32_e32 v13, s1
 ; VI-NEXT:    v_mov_b32_e32 v12, s0
 ; VI-NEXT:    s_add_u32 s0, s4, 16
+; VI-NEXT:    v_mov_b32_e32 v13, s1
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
 ; VI-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-NEXT:    s_nop 0
-; VI-NEXT:    v_mov_b32_e32 v9, s1
 ; VI-NEXT:    v_mov_b32_e32 v8, s0
+; VI-NEXT:    v_mov_b32_e32 v9, s1
 ; VI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-NEXT:    s_nop 0
 ; VI-NEXT:    v_mov_b32_e32 v4, s4
@@ -1342,9 +1342,9 @@ define amdgpu_kernel void @global_extload_v8f16_to_v8f32(ptr addrspace(1) %out,
 ; CI-NEXT:    v_mov_b32_e32 v1, s3
 ; CI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; CI-NEXT:    s_add_u32 s2, s0, 16
-; CI-NEXT:    s_addc_u32 s3, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v13, s1
 ; CI-NEXT:    v_mov_b32_e32 v12, s0
+; CI-NEXT:    s_addc_u32 s0, s1, 0
 ; CI-NEXT:    s_waitcnt vmcnt(0)
 ; CI-NEXT:    v_lshrrev_b32_e32 v5, 16, v2
 ; CI-NEXT:    v_lshrrev_b32_e32 v9, 16, v3
@@ -1359,7 +1359,7 @@ define amdgpu_kernel void @global_extload_v8f16_to_v8f32(ptr addrspace(1) %out,
 ; CI-NEXT:    v_cvt_f32_f16_e32 v7, v7
 ; CI-NEXT:    v_cvt_f32_f16_e32 v5, v14
 ; CI-NEXT:    v_mov_b32_e32 v0, s2
-; CI-NEXT:    v_mov_b32_e32 v1, s3
+; CI-NEXT:    v_mov_b32_e32 v1, s0
 ; CI-NEXT:    flat_store_dwordx4 v[0:1], v[8:11]
 ; CI-NEXT:    flat_store_dwordx4 v[12:13], v[4:7]
 ; CI-NEXT:    s_endpgm
@@ -1375,9 +1375,9 @@ define amdgpu_kernel void @global_extload_v8f16_to_v8f32(ptr addrspace(1) %out,
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; VI-NEXT:    s_add_u32 s2, s0, 16
-; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v13, s1
 ; VI-NEXT:    v_mov_b32_e32 v12, s0
+; VI-NEXT:    s_addc_u32 s0, s1, 0
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_cvt_f32_f16_e32 v10, v3
 ; VI-NEXT:    v_cvt_f32_f16_e32 v8, v2
@@ -1388,7 +1388,7 @@ define amdgpu_kernel void @global_extload_v8f16_to_v8f32(ptr addrspace(1) %out,
 ; VI-NEXT:    v_cvt_f32_f16_sdwa v7, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; VI-NEXT:    v_cvt_f32_f16_sdwa v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
-; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v1, s0
 ; VI-NEXT:    flat_store_dwordx4 v[0:1], v[8:11]
 ; VI-NEXT:    flat_store_dwordx4 v[12:13], v[4:7]
 ; VI-NEXT:    s_endpgm
@@ -1459,13 +1459,13 @@ define amdgpu_kernel void @global_extload_v16f16_to_v16f32(ptr addrspace(1) %out
 ; CI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; CI-NEXT:    s_add_u32 s2, s2, 16
 ; CI-NEXT:    s_addc_u32 s3, s3, 0
-; CI-NEXT:    v_mov_b32_e32 v5, s3
 ; CI-NEXT:    v_mov_b32_e32 v4, s2
+; CI-NEXT:    v_mov_b32_e32 v5, s3
 ; CI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; CI-NEXT:    s_add_u32 s2, s0, 16
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v16, s3
 ; CI-NEXT:    v_mov_b32_e32 v15, s2
+; CI-NEXT:    v_mov_b32_e32 v16, s3
 ; CI-NEXT:    s_add_u32 s2, s0, 48
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
 ; CI-NEXT:    s_waitcnt vmcnt(1)
@@ -1499,10 +1499,10 @@ define amdgpu_kernel void @global_extload_v16f16_to_v16f32(ptr addrspace(1) %out
 ; CI-NEXT:    v_cvt_f32_f16_e32 v3, v3
 ; CI-NEXT:    v_cvt_f32_f16_e32 v1, v16
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v11, s3
 ; CI-NEXT:    v_mov_b32_e32 v10, s2
-; CI-NEXT:    v_mov_b32_e32 v17, s1
+; CI-NEXT:    v_mov_b32_e32 v11, s3
 ; CI-NEXT:    v_mov_b32_e32 v16, s0
+; CI-NEXT:    v_mov_b32_e32 v17, s1
 ; CI-NEXT:    flat_store_dwordx4 v[4:5], v[12:15]
 ; CI-NEXT:    flat_store_dwordx4 v[10:11], v[6:9]
 ; CI-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
@@ -1520,21 +1520,21 @@ define amdgpu_kernel void @global_extload_v16f16_to_v16f32(ptr addrspace(1) %out
 ; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; VI-NEXT:    s_add_u32 s2, s2, 16
 ; VI-NEXT:    s_addc_u32 s3, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; VI-NEXT:    s_add_u32 s2, s0, 16
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v19, s3
 ; VI-NEXT:    v_mov_b32_e32 v18, s2
+; VI-NEXT:    v_mov_b32_e32 v19, s3
 ; VI-NEXT:    s_add_u32 s2, s0, 48
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v17, s1
 ; VI-NEXT:    v_mov_b32_e32 v16, s0
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v21, s3
 ; VI-NEXT:    v_mov_b32_e32 v20, s2
+; VI-NEXT:    v_mov_b32_e32 v21, s3
 ; VI-NEXT:    s_waitcnt vmcnt(1)
 ; VI-NEXT:    v_cvt_f32_f16_e32 v14, v3
 ; VI-NEXT:    v_cvt_f32_f16_e32 v12, v2
@@ -1554,8 +1554,8 @@ define amdgpu_kernel void @global_extload_v16f16_to_v16f32(ptr addrspace(1) %out
 ; VI-NEXT:    v_cvt_f32_f16_e32 v0, v4
 ; VI-NEXT:    v_cvt_f32_f16_sdwa v3, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; VI-NEXT:    v_cvt_f32_f16_sdwa v1, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    flat_store_dwordx4 v[16:17], v[8:11]
 ; VI-NEXT:    flat_store_dwordx4 v[20:21], v[12:15]
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -1792,8 +1792,8 @@ define amdgpu_kernel void @global_extload_v3f16_to_v3f64(ptr addrspace(1) %out,
 ; CI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
 ; CI-NEXT:    s_add_u32 s2, s0, 16
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v7, s3
 ; CI-NEXT:    v_mov_b32_e32 v6, s2
+; CI-NEXT:    v_mov_b32_e32 v7, s3
 ; CI-NEXT:    s_waitcnt vmcnt(0)
 ; CI-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
 ; CI-NEXT:    v_cvt_f32_f16_e32 v1, v1
@@ -1819,10 +1819,10 @@ define amdgpu_kernel void @global_extload_v3f16_to_v3f64(ptr addrspace(1) %out,
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
 ; VI-NEXT:    s_add_u32 s2, s0, 16
-; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    s_addc_u32 s0, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v9, s0
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; VI-NEXT:    v_cvt_f32_f16_e32 v2, v0
@@ -1894,11 +1894,11 @@ define amdgpu_kernel void @global_extload_v4f16_to_v4f64(ptr addrspace(1) %out,
 ; CI-NEXT:    v_mov_b32_e32 v1, s3
 ; CI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
 ; CI-NEXT:    s_add_u32 s2, s0, 16
-; CI-NEXT:    s_addc_u32 s3, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v11, s3
-; CI-NEXT:    v_mov_b32_e32 v10, s2
 ; CI-NEXT:    v_mov_b32_e32 v9, s1
 ; CI-NEXT:    v_mov_b32_e32 v8, s0
+; CI-NEXT:    s_addc_u32 s0, s1, 0
+; CI-NEXT:    v_mov_b32_e32 v10, s2
+; CI-NEXT:    v_mov_b32_e32 v11, s0
 ; CI-NEXT:    s_waitcnt vmcnt(0)
 ; CI-NEXT:    v_lshrrev_b32_e32 v2, 16, v1
 ; CI-NEXT:    v_lshrrev_b32_e32 v3, 16, v0
@@ -1925,10 +1925,10 @@ define amdgpu_kernel void @global_extload_v4f16_to_v4f64(ptr addrspace(1) %out,
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
 ; VI-NEXT:    s_add_u32 s2, s0, 16
-; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v11, s3
 ; VI-NEXT:    v_mov_b32_e32 v9, s1
 ; VI-NEXT:    v_mov_b32_e32 v8, s0
+; VI-NEXT:    s_addc_u32 s0, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v11, s0
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; VI-NEXT:    v_cvt_f32_f16_sdwa v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
@@ -2011,15 +2011,15 @@ define amdgpu_kernel void @global_extload_v8f16_to_v8f64(ptr addrspace(1) %out,
 ; CI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; CI-NEXT:    s_add_u32 s2, s0, 48
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v7, s3
 ; CI-NEXT:    v_mov_b32_e32 v6, s2
+; CI-NEXT:    v_mov_b32_e32 v7, s3
 ; CI-NEXT:    s_add_u32 s2, s0, 32
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v13, s1
 ; CI-NEXT:    v_mov_b32_e32 v12, s0
 ; CI-NEXT:    s_add_u32 s0, s0, 16
-; CI-NEXT:    v_mov_b32_e32 v15, s3
 ; CI-NEXT:    v_mov_b32_e32 v14, s2
+; CI-NEXT:    v_mov_b32_e32 v15, s3
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
 ; CI-NEXT:    s_waitcnt vmcnt(0)
 ; CI-NEXT:    v_lshrrev_b32_e32 v4, 16, v3
@@ -2043,8 +2043,8 @@ define amdgpu_kernel void @global_extload_v8f16_to_v8f64(ptr addrspace(1) %out,
 ; CI-NEXT:    v_cvt_f64_f32_e32 v[10:11], v19
 ; CI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v18
 ; CI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v17
-; CI-NEXT:    v_mov_b32_e32 v17, s1
 ; CI-NEXT:    v_mov_b32_e32 v16, s0
+; CI-NEXT:    v_mov_b32_e32 v17, s1
 ; CI-NEXT:    flat_store_dwordx4 v[14:15], v[8:11]
 ; CI-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
 ; CI-NEXT:    flat_store_dwordx4 v[12:13], v[4:7]
@@ -2062,15 +2062,15 @@ define amdgpu_kernel void @global_extload_v8f16_to_v8f64(ptr addrspace(1) %out,
 ; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; VI-NEXT:    s_add_u32 s2, s0, 48
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v8, s3
 ; VI-NEXT:    v_mov_b32_e32 v7, s2
+; VI-NEXT:    v_mov_b32_e32 v8, s3
 ; VI-NEXT:    s_add_u32 s2, s0, 32
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v13, s1
 ; VI-NEXT:    v_mov_b32_e32 v12, s0
 ; VI-NEXT:    s_add_u32 s0, s0, 16
-; VI-NEXT:    v_mov_b32_e32 v15, s3
 ; VI-NEXT:    v_mov_b32_e32 v14, s2
+; VI-NEXT:    v_mov_b32_e32 v15, s3
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_cvt_f32_f16_e32 v9, v0
@@ -2090,8 +2090,8 @@ define amdgpu_kernel void @global_extload_v8f16_to_v8f64(ptr addrspace(1) %out,
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[10:11], v2
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v17
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v16
-; VI-NEXT:    v_mov_b32_e32 v17, s1
 ; VI-NEXT:    v_mov_b32_e32 v16, s0
+; VI-NEXT:    v_mov_b32_e32 v17, s1
 ; VI-NEXT:    flat_store_dwordx4 v[14:15], v[8:11]
 ; VI-NEXT:    flat_store_dwordx4 v[16:17], v[4:7]
 ; VI-NEXT:    flat_store_dwordx4 v[12:13], v[0:3]
@@ -2180,91 +2180,91 @@ define amdgpu_kernel void @global_extload_v16f16_to_v16f64(ptr addrspace(1) %out
 ; CI-NEXT:    s_waitcnt lgkmcnt(0)
 ; CI-NEXT:    v_mov_b32_e32 v0, s2
 ; CI-NEXT:    v_mov_b32_e32 v1, s3
-; CI-NEXT:    flat_load_dwordx4 v[4:7], v[0:1]
+; CI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; CI-NEXT:    s_add_u32 s2, s2, 16
 ; CI-NEXT:    s_addc_u32 s3, s3, 0
-; CI-NEXT:    v_mov_b32_e32 v0, s2
-; CI-NEXT:    v_mov_b32_e32 v1, s3
-; CI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; CI-NEXT:    v_mov_b32_e32 v4, s2
+; CI-NEXT:    v_mov_b32_e32 v5, s3
+; CI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; CI-NEXT:    s_add_u32 s2, s0, 48
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v14, s3
-; CI-NEXT:    v_mov_b32_e32 v13, s2
+; CI-NEXT:    v_mov_b32_e32 v14, s2
+; CI-NEXT:    v_mov_b32_e32 v15, s3
 ; CI-NEXT:    s_add_u32 s2, s0, 32
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v16, s3
-; CI-NEXT:    v_mov_b32_e32 v15, s2
+; CI-NEXT:    v_mov_b32_e32 v16, s2
+; CI-NEXT:    v_mov_b32_e32 v17, s3
 ; CI-NEXT:    s_add_u32 s2, s0, 16
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v18, s3
-; CI-NEXT:    v_mov_b32_e32 v17, s2
-; CI-NEXT:    v_mov_b32_e32 v12, s1
-; CI-NEXT:    v_mov_b32_e32 v11, s0
+; CI-NEXT:    v_mov_b32_e32 v18, s2
+; CI-NEXT:    v_mov_b32_e32 v19, s3
+; CI-NEXT:    v_mov_b32_e32 v13, s1
+; CI-NEXT:    v_mov_b32_e32 v12, s0
 ; CI-NEXT:    s_add_u32 s2, s0, 0x70
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
 ; CI-NEXT:    s_waitcnt vmcnt(1)
-; CI-NEXT:    v_lshrrev_b32_e32 v8, 16, v7
-; CI-NEXT:    v_cvt_f32_f16_e32 v7, v7
-; CI-NEXT:    v_cvt_f32_f16_e32 v9, v8
-; CI-NEXT:    v_cvt_f64_f32_e32 v[7:8], v7
-; CI-NEXT:    v_cvt_f64_f32_e32 v[9:10], v9
+; CI-NEXT:    v_lshrrev_b32_e32 v8, 16, v3
+; CI-NEXT:    v_cvt_f32_f16_e32 v3, v3
+; CI-NEXT:    v_cvt_f32_f16_e32 v10, v8
+; CI-NEXT:    v_cvt_f64_f32_e32 v[8:9], v3
+; CI-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
+; CI-NEXT:    v_cvt_f64_f32_e32 v[10:11], v10
+; CI-NEXT:    v_cvt_f32_f16_e32 v2, v2
+; CI-NEXT:    v_cvt_f32_f16_e32 v3, v3
 ; CI-NEXT:    s_waitcnt vmcnt(0)
-; CI-NEXT:    v_lshrrev_b32_e32 v19, 16, v2
-; CI-NEXT:    v_cvt_f32_f16_e32 v20, v2
-; CI-NEXT:    v_cvt_f32_f16_e32 v21, v3
-; CI-NEXT:    flat_store_dwordx4 v[13:14], v[7:10]
-; CI-NEXT:    v_mov_b32_e32 v14, s3
-; CI-NEXT:    v_lshrrev_b32_e32 v7, 16, v6
+; CI-NEXT:    v_lshrrev_b32_e32 v20, 16, v5
+; CI-NEXT:    flat_store_dwordx4 v[14:15], v[8:11]
+; CI-NEXT:    v_cvt_f32_f16_e32 v22, v5
+; CI-NEXT:    v_cvt_f64_f32_e32 v[8:9], v2
+; CI-NEXT:    v_cvt_f64_f32_e32 v[10:11], v3
+; CI-NEXT:    v_lshrrev_b32_e32 v2, 16, v1
+; CI-NEXT:    v_cvt_f32_f16_e32 v1, v1
+; CI-NEXT:    v_cvt_f32_f16_e32 v2, v2
+; CI-NEXT:    flat_store_dwordx4 v[16:17], v[8:11]
+; CI-NEXT:    v_cvt_f32_f16_e32 v5, v7
+; CI-NEXT:    v_lshrrev_b32_e32 v8, 16, v0
+; CI-NEXT:    v_cvt_f32_f16_e32 v9, v0
+; CI-NEXT:    v_cvt_f64_f32_e32 v[0:1], v1
+; CI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v2
+; CI-NEXT:    v_cvt_f32_f16_e32 v8, v8
+; CI-NEXT:    v_lshrrev_b32_e32 v10, 16, v7
+; CI-NEXT:    v_cvt_f32_f16_e32 v7, v10
+; CI-NEXT:    flat_store_dwordx4 v[18:19], v[0:3]
+; CI-NEXT:    v_lshrrev_b32_e32 v11, 16, v6
+; CI-NEXT:    v_cvt_f64_f32_e32 v[0:1], v9
+; CI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v8
 ; CI-NEXT:    v_cvt_f32_f16_e32 v6, v6
-; CI-NEXT:    v_cvt_f32_f16_e32 v8, v7
-; CI-NEXT:    v_lshrrev_b32_e32 v10, 16, v3
-; CI-NEXT:    v_mov_b32_e32 v13, s2
-; CI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v6
-; CI-NEXT:    v_cvt_f64_f32_e32 v[8:9], v8
+; CI-NEXT:    v_cvt_f32_f16_e32 v10, v11
+; CI-NEXT:    v_lshrrev_b32_e32 v21, 16, v4
+; CI-NEXT:    flat_store_dwordx4 v[12:13], v[0:3]
+; CI-NEXT:    v_mov_b32_e32 v14, s2
+; CI-NEXT:    v_cvt_f64_f32_e32 v[0:1], v5
+; CI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v7
+; CI-NEXT:    v_cvt_f32_f16_e32 v7, v20
 ; CI-NEXT:    s_add_u32 s2, s0, 0x60
+; CI-NEXT:    v_cvt_f32_f16_e32 v4, v4
+; CI-NEXT:    v_cvt_f32_f16_e32 v21, v21
+; CI-NEXT:    v_mov_b32_e32 v15, s3
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
-; CI-NEXT:    flat_store_dwordx4 v[15:16], v[6:9]
-; CI-NEXT:    v_mov_b32_e32 v16, s3
-; CI-NEXT:    v_lshrrev_b32_e32 v6, 16, v5
-; CI-NEXT:    v_cvt_f32_f16_e32 v5, v5
-; CI-NEXT:    v_cvt_f32_f16_e32 v6, v6
-; CI-NEXT:    v_lshrrev_b32_e32 v8, 16, v4
-; CI-NEXT:    v_cvt_f32_f16_e32 v9, v4
-; CI-NEXT:    v_cvt_f64_f32_e32 v[4:5], v5
-; CI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v6
-; CI-NEXT:    v_cvt_f32_f16_e32 v8, v8
-; CI-NEXT:    v_mov_b32_e32 v15, s2
+; CI-NEXT:    v_mov_b32_e32 v16, s2
 ; CI-NEXT:    s_add_u32 s2, s0, 0x50
-; CI-NEXT:    flat_store_dwordx4 v[17:18], v[4:7]
-; CI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v8
-; CI-NEXT:    v_lshrrev_b32_e32 v4, 16, v1
-; CI-NEXT:    v_lshrrev_b32_e32 v5, 16, v0
-; CI-NEXT:    v_cvt_f32_f16_e32 v6, v0
-; CI-NEXT:    v_cvt_f32_f16_e32 v7, v1
-; CI-NEXT:    v_cvt_f64_f32_e32 v[0:1], v9
-; CI-NEXT:    v_cvt_f32_f16_e32 v8, v10
-; CI-NEXT:    v_cvt_f32_f16_e32 v10, v19
-; CI-NEXT:    v_cvt_f32_f16_e32 v22, v5
-; CI-NEXT:    flat_store_dwordx4 v[11:12], v[0:3]
-; CI-NEXT:    v_cvt_f32_f16_e32 v12, v4
-; CI-NEXT:    v_cvt_f64_f32_e32 v[0:1], v21
-; CI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v8
-; CI-NEXT:    v_cvt_f64_f32_e32 v[8:9], v20
+; CI-NEXT:    v_cvt_f64_f32_e32 v[8:9], v6
 ; CI-NEXT:    v_cvt_f64_f32_e32 v[10:11], v10
+; CI-NEXT:    v_mov_b32_e32 v17, s3
 ; CI-NEXT:    s_addc_u32 s3, s1, 0
-; CI-NEXT:    flat_store_dwordx4 v[13:14], v[0:3]
+; CI-NEXT:    flat_store_dwordx4 v[14:15], v[0:3]
 ; CI-NEXT:    s_add_u32 s0, s0, 64
-; CI-NEXT:    v_cvt_f64_f32_e32 v[0:1], v7
-; CI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v12
-; CI-NEXT:    v_cvt_f64_f32_e32 v[4:5], v6
-; CI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v22
+; CI-NEXT:    v_cvt_f64_f32_e32 v[0:1], v22
+; CI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v7
+; CI-NEXT:    v_cvt_f64_f32_e32 v[4:5], v4
+; CI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v21
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v18, s3
-; CI-NEXT:    v_mov_b32_e32 v17, s2
-; CI-NEXT:    v_mov_b32_e32 v13, s1
+; CI-NEXT:    v_mov_b32_e32 v18, s2
+; CI-NEXT:    v_mov_b32_e32 v19, s3
 ; CI-NEXT:    v_mov_b32_e32 v12, s0
-; CI-NEXT:    flat_store_dwordx4 v[15:16], v[8:11]
-; CI-NEXT:    flat_store_dwordx4 v[17:18], v[0:3]
+; CI-NEXT:    v_mov_b32_e32 v13, s1
+; CI-NEXT:    flat_store_dwordx4 v[16:17], v[8:11]
+; CI-NEXT:    flat_store_dwordx4 v[18:19], v[0:3]
 ; CI-NEXT:    flat_store_dwordx4 v[12:13], v[4:7]
 ; CI-NEXT:    s_endpgm
 ;
@@ -2285,16 +2285,16 @@ define amdgpu_kernel void @global_extload_v16f16_to_v16f64(ptr addrspace(1) %out
 ; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; VI-NEXT:    s_add_u32 s2, s0, 48
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v14, s3
 ; VI-NEXT:    v_mov_b32_e32 v13, s2
+; VI-NEXT:    v_mov_b32_e32 v14, s3
 ; VI-NEXT:    s_add_u32 s2, s0, 32
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v16, s3
 ; VI-NEXT:    v_mov_b32_e32 v15, s2
+; VI-NEXT:    v_mov_b32_e32 v16, s3
 ; VI-NEXT:    s_add_u32 s2, s0, 16
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v18, s3
 ; VI-NEXT:    v_mov_b32_e32 v17, s2
+; VI-NEXT:    v_mov_b32_e32 v18, s3
 ; VI-NEXT:    v_mov_b32_e32 v12, s1
 ; VI-NEXT:    v_mov_b32_e32 v11, s0
 ; VI-NEXT:    s_add_u32 s2, s0, 0x50
@@ -2305,26 +2305,26 @@ define amdgpu_kernel void @global_extload_v16f16_to_v16f64(ptr addrspace(1) %out
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[7:8], v8
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[9:10], v9
 ; VI-NEXT:    flat_store_dwordx4 v[13:14], v[7:10]
-; VI-NEXT:    v_mov_b32_e32 v14, s3
+; VI-NEXT:    v_mov_b32_e32 v13, s2
 ; VI-NEXT:    v_cvt_f32_f16_e32 v7, v6
 ; VI-NEXT:    v_cvt_f32_f16_sdwa v8, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; VI-NEXT:    s_waitcnt vmcnt(1)
 ; VI-NEXT:    v_cvt_f32_f16_e32 v10, v2
-; VI-NEXT:    v_mov_b32_e32 v13, s2
+; VI-NEXT:    s_add_u32 s2, s0, 64
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v7
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[8:9], v8
-; VI-NEXT:    s_add_u32 s2, s0, 64
+; VI-NEXT:    v_mov_b32_e32 v14, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[15:16], v[6:9]
-; VI-NEXT:    v_mov_b32_e32 v16, s3
+; VI-NEXT:    v_mov_b32_e32 v15, s2
 ; VI-NEXT:    v_cvt_f32_f16_e32 v6, v5
 ; VI-NEXT:    v_cvt_f32_f16_sdwa v7, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; VI-NEXT:    v_cvt_f32_f16_e32 v8, v4
 ; VI-NEXT:    v_cvt_f32_f16_sdwa v9, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[4:5], v6
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[6:7], v7
-; VI-NEXT:    v_mov_b32_e32 v15, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 0x70
+; VI-NEXT:    v_mov_b32_e32 v16, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[17:18], v[4:7]
 ; VI-NEXT:    v_cvt_f32_f16_sdwa v17, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
@@ -2348,10 +2348,10 @@ define amdgpu_kernel void @global_extload_v16f16_to_v16f64(ptr addrspace(1) %out
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[2:3], v17
 ; VI-NEXT:    v_cvt_f64_f32_e32 v[7:8], v8
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v20, s3
 ; VI-NEXT:    v_mov_b32_e32 v19, s2
-; VI-NEXT:    v_mov_b32_e32 v14, s1
+; VI-NEXT:    v_mov_b32_e32 v20, s3
 ; VI-NEXT:    v_mov_b32_e32 v13, s0
+; VI-NEXT:    v_mov_b32_e32 v14, s1
 ; VI-NEXT:    flat_store_dwordx4 v[15:16], v[9:12]
 ; VI-NEXT:    flat_store_dwordx4 v[19:20], v[0:3]
 ; VI-NEXT:    flat_store_dwordx4 v[13:14], v[5:8]
@@ -2784,8 +2784,8 @@ define amdgpu_kernel void @global_truncstore_v8f32_to_v8f16(ptr addrspace(1) %ou
 ; CI-NEXT:    v_mov_b32_e32 v1, s3
 ; CI-NEXT:    s_add_u32 s2, s2, 16
 ; CI-NEXT:    s_addc_u32 s3, s3, 0
-; CI-NEXT:    v_mov_b32_e32 v5, s3
 ; CI-NEXT:    v_mov_b32_e32 v4, s2
+; CI-NEXT:    v_mov_b32_e32 v5, s3
 ; CI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; CI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; CI-NEXT:    v_mov_b32_e32 v8, s0
@@ -2822,8 +2822,8 @@ define amdgpu_kernel void @global_truncstore_v8f32_to_v8f16(ptr addrspace(1) %ou
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    s_add_u32 s2, s2, 16
 ; VI-NEXT:    s_addc_u32 s3, s3, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; VI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v8, s0
@@ -2908,8 +2908,8 @@ define amdgpu_kernel void @global_truncstore_v16f32_to_v16f16(ptr addrspace(1) %
 ; CI-NEXT:    s_add_u32 s4, s2, 32
 ; CI-NEXT:    s_addc_u32 s5, s3, 0
 ; CI-NEXT:    v_mov_b32_e32 v0, s4
-; CI-NEXT:    v_mov_b32_e32 v1, s5
 ; CI-NEXT:    s_add_u32 s4, s2, 48
+; CI-NEXT:    v_mov_b32_e32 v1, s5
 ; CI-NEXT:    s_addc_u32 s5, s3, 0
 ; CI-NEXT:    v_mov_b32_e32 v9, s3
 ; CI-NEXT:    v_mov_b32_e32 v8, s2
@@ -2919,8 +2919,8 @@ define amdgpu_kernel void @global_truncstore_v16f32_to_v16f16(ptr addrspace(1) %
 ; CI-NEXT:    v_mov_b32_e32 v5, s5
 ; CI-NEXT:    s_addc_u32 s3, s3, 0
 ; CI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
-; CI-NEXT:    v_mov_b32_e32 v13, s3
 ; CI-NEXT:    v_mov_b32_e32 v12, s2
+; CI-NEXT:    v_mov_b32_e32 v13, s3
 ; CI-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
 ; CI-NEXT:    flat_load_dwordx4 v[12:15], v[12:13]
 ; CI-NEXT:    s_add_u32 s2, s0, 16
@@ -2950,8 +2950,8 @@ define amdgpu_kernel void @global_truncstore_v16f32_to_v16f16(ptr addrspace(1) %
 ; CI-NEXT:    v_or_b32_e32 v1, v2, v3
 ; CI-NEXT:    v_lshlrev_b32_e32 v2, 16, v7
 ; CI-NEXT:    v_lshlrev_b32_e32 v7, 16, v16
-; CI-NEXT:    v_mov_b32_e32 v5, s3
 ; CI-NEXT:    v_mov_b32_e32 v4, s2
+; CI-NEXT:    v_mov_b32_e32 v5, s3
 ; CI-NEXT:    v_or_b32_e32 v0, v0, v18
 ; CI-NEXT:    v_or_b32_e32 v3, v6, v2
 ; CI-NEXT:    v_or_b32_e32 v2, v17, v7
@@ -2979,8 +2979,8 @@ define amdgpu_kernel void @global_truncstore_v16f32_to_v16f16(ptr addrspace(1) %
 ; VI-NEXT:    s_add_u32 s4, s2, 32
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s4
-; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    s_add_u32 s4, s2, 48
+; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    s_addc_u32 s5, s3, 0
 ; VI-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-NEXT:    v_mov_b32_e32 v8, s2
@@ -2990,8 +2990,8 @@ define amdgpu_kernel void @global_truncstore_v16f32_to_v16f16(ptr addrspace(1) %
 ; VI-NEXT:    s_addc_u32 s3, s3, 0
 ; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; VI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
-; VI-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-NEXT:    v_mov_b32_e32 v12, s2
+; VI-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
 ; VI-NEXT:    flat_load_dwordx4 v[12:15], v[12:13]
 ; VI-NEXT:    s_add_u32 s2, s0, 16
@@ -3016,8 +3016,8 @@ define amdgpu_kernel void @global_truncstore_v16f32_to_v16f16(ptr addrspace(1) %
 ; VI-NEXT:    v_cvt_f16_f32_e32 v14, v14
 ; VI-NEXT:    v_cvt_f16_f32_sdwa v13, v13 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD
 ; VI-NEXT:    v_cvt_f16_f32_e32 v12, v12
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_or_b32_e32 v1, v2, v3
 ; VI-NEXT:    v_or_b32_e32 v0, v0, v16
 ; VI-NEXT:    v_or_b32_e32 v3, v6, v7
diff --git a/llvm/test/CodeGen/AMDGPU/identical-subrange-spill-infloop.ll b/llvm/test/CodeGen/AMDGPU/identical-subrange-spill-infloop.ll
index fb0c03543377c..d0972b9730e59 100644
--- a/llvm/test/CodeGen/AMDGPU/identical-subrange-spill-infloop.ll
+++ b/llvm/test/CodeGen/AMDGPU/identical-subrange-spill-infloop.ll
@@ -201,13 +201,11 @@ define void @main(i1 %arg) #0 {
 ; CHECK-NEXT:    s_cbranch_execz .LBB0_9
 ; CHECK-NEXT:  ; %bb.7: ; %bb33.preheader
 ; CHECK-NEXT:    s_mov_b32 s16, 0
-; CHECK-NEXT:    s_mov_b32 s20, s16
-; CHECK-NEXT:    s_mov_b32 s21, s16
 ; CHECK-NEXT:    s_mov_b32 s17, s16
 ; CHECK-NEXT:    s_mov_b32 s18, s16
 ; CHECK-NEXT:    s_mov_b32 s19, s16
-; CHECK-NEXT:    v_mov_b32_e32 v1, s20
-; CHECK-NEXT:    v_mov_b32_e32 v2, s21
+; CHECK-NEXT:    v_mov_b32_e32 v1, s16
+; CHECK-NEXT:    v_mov_b32_e32 v2, s16
 ; CHECK-NEXT:    image_sample_lz v3, v[1:2], s[8:15], s[16:19] dmask:0x1
 ; CHECK-NEXT:    image_sample_lz v4, v[1:2], s[36:43], s[16:19] dmask:0x1
 ; CHECK-NEXT:    s_and_b64 vcc, exec, 0
diff --git a/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll b/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll
index 127f4996b7676..9c786abe9196b 100644
--- a/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll
+++ b/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll
@@ -2026,20 +2026,20 @@ define amdgpu_kernel void @insert_w_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, 0x41880000
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    v_movreld_b32_e32 v0, v16
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 32
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 16
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v8, s2
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-MOVREL-NEXT:    s_nop 0
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s1
@@ -2075,20 +2075,20 @@ define amdgpu_kernel void @insert_w_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 48
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 16
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v8, s2
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-IDXMODE-NEXT:    s_nop 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s1
@@ -2351,20 +2351,20 @@ define amdgpu_kernel void @insert_unsigned_base_plus_offset(ptr addrspace(1) %ou
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, 0x41880000
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    v_movreld_b32_e32 v1, v16
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 32
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 16
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v8, s2
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-MOVREL-NEXT:    s_nop 0
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s1
@@ -2400,20 +2400,20 @@ define amdgpu_kernel void @insert_unsigned_base_plus_offset(ptr addrspace(1) %ou
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 48
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 16
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v8, s2
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-IDXMODE-NEXT:    s_nop 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s1
@@ -2681,20 +2681,20 @@ define amdgpu_kernel void @insert_signed_base_plus_offset(ptr addrspace(1) %out,
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, 0x41880000
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    v_movreld_b32_e32 v0, v16
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 32
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 16
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v8, s2
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-MOVREL-NEXT:    s_nop 0
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s1
@@ -2731,20 +2731,20 @@ define amdgpu_kernel void @insert_signed_base_plus_offset(ptr addrspace(1) %out,
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 48
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 16
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v8, s2
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-IDXMODE-NEXT:    s_nop 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s1
@@ -3002,20 +3002,20 @@ define amdgpu_kernel void @insert_wo_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, 0x41880000
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    v_movreld_b32_e32 v0, v16
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 32
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 16
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v8, s2
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-MOVREL-NEXT:    s_nop 0
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s1
@@ -3050,20 +3050,20 @@ define amdgpu_kernel void @insert_wo_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 48
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 16
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v8, s2
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-IDXMODE-NEXT:    s_nop 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s1
@@ -3310,21 +3310,21 @@ define amdgpu_kernel void @insert_neg_offset_sgpr(ptr addrspace(1) %in, ptr addr
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, 13
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v14, 14
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v15, 15
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 32
 ; VI-MOVREL-NEXT:    v_movreld_b32_e32 v0, 16
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 16
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v8, s2
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-MOVREL-NEXT:    s_nop 0
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s1
@@ -3359,20 +3359,20 @@ define amdgpu_kernel void @insert_neg_offset_sgpr(ptr addrspace(1) %in, ptr addr
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 48
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 16
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v8, s2
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-IDXMODE-NEXT:    s_nop 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s1
@@ -3611,21 +3611,21 @@ define amdgpu_kernel void @insert_neg_offset_sgpr_loadreg(ptr addrspace(1) %in,
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s21
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v14, s22
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v15, s23
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 32
 ; VI-MOVREL-NEXT:    v_movreld_b32_e32 v0, 5
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 16
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v8, s2
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-MOVREL-NEXT:    s_nop 0
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s1
@@ -3640,7 +3640,7 @@ define amdgpu_kernel void @insert_neg_offset_sgpr_loadreg(ptr addrspace(1) %in,
 ; VI-IDXMODE-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x2c
 ; VI-IDXMODE-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v0, s8
-; VI-IDXMODE-NEXT:    s_add_i32 s3, s2, 0xfffffe00
+; VI-IDXMODE-NEXT:    s_addk_i32 s2, 0xfe00
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v1, s9
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v2, s10
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v3, s11
@@ -3656,25 +3656,25 @@ define amdgpu_kernel void @insert_neg_offset_sgpr_loadreg(ptr addrspace(1) %in,
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s21
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v14, s22
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v15, s23
-; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 48
-; VI-IDXMODE-NEXT:    s_set_gpr_idx_on s3, gpr_idx(DST)
+; VI-IDXMODE-NEXT:    s_add_u32 s3, s0, 48
+; VI-IDXMODE-NEXT:    s_set_gpr_idx_on s2, gpr_idx(DST)
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v0, 5
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
-; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s2
+; VI-IDXMODE-NEXT:    s_addc_u32 s2, s1, 0
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 16
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v8, s2
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-IDXMODE-NEXT:    s_nop 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s1
@@ -4136,21 +4136,21 @@ define amdgpu_kernel void @insert_neg_offset_vgpr(ptr addrspace(1) %in, ptr addr
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    v_cndmask_b32_e64 v13, 14, 33, vcc
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, 12, v12
-; VI-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 32
 ; VI-NEXT:    v_cndmask_b32_e64 v12, 13, 33, vcc
+; VI-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-NEXT:    s_nop 0
-; VI-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 16
+; VI-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-NEXT:    s_nop 0
-; VI-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-NEXT:    v_mov_b32_e32 v8, s2
+; VI-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-NEXT:    s_nop 0
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
@@ -4628,20 +4628,20 @@ define amdgpu_kernel void @insert_neg_inline_offset_vgpr(ptr addrspace(1) %in, p
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, 12, v12
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    v_cndmask_b32_e32 v12, 13, v16, vcc
-; VI-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 32
+; VI-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-NEXT:    s_nop 0
-; VI-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 16
+; VI-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-NEXT:    s_nop 0
-; VI-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-NEXT:    v_mov_b32_e32 v8, s2
+; VI-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-NEXT:    s_nop 0
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
@@ -6224,21 +6224,21 @@ define amdgpu_kernel void @insert_vgpr_offset_multiple_in_block(ptr addrspace(1)
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s2, s0, 48
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v19, s3
 ; VI-NEXT:    v_mov_b32_e32 v18, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 32
+; VI-NEXT:    v_mov_b32_e32 v19, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[18:19], v[14:17]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; VI-NEXT:    v_mov_b32_e32 v15, s3
 ; VI-NEXT:    v_mov_b32_e32 v14, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 16
+; VI-NEXT:    v_mov_b32_e32 v15, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
-; VI-NEXT:    v_mov_b32_e32 v11, s3
 ; VI-NEXT:    v_mov_b32_e32 v10, s2
+; VI-NEXT:    v_mov_b32_e32 v11, s3
 ; VI-NEXT:    flat_store_dwordx4 v[10:11], v[6:9]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v7, s1
@@ -6732,21 +6732,21 @@ define amdgpu_kernel void @insert_w_offset_multiple_in_block(ptr addrspace(1) %o
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, v0
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    v_movreld_b32_e32 v16, v32
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v33, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v32, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 32
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v33, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[32:33], v[12:15]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 16
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v8, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 64
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    s_add_u32 s4, s0, 0x70
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
@@ -6756,8 +6756,8 @@ define amdgpu_kernel void @insert_w_offset_multiple_in_block(ptr addrspace(1) %o
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-MOVREL-NEXT:    s_nop 0
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v0, s4
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-MOVREL-NEXT:    s_add_u32 s4, s0, 0x60
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-MOVREL-NEXT:    s_addc_u32 s5, s1, 0
 ; VI-MOVREL-NEXT:    s_add_u32 s0, s0, 0x50
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[0:1], v[28:31]
@@ -6821,21 +6821,21 @@ define amdgpu_kernel void @insert_w_offset_multiple_in_block(ptr addrspace(1) %o
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 48
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v33, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v32, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v33, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[32:33], v[12:15]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 16
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v8, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 64
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    s_add_u32 s4, s0, 0x70
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
@@ -6845,8 +6845,8 @@ define amdgpu_kernel void @insert_w_offset_multiple_in_block(ptr addrspace(1) %o
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-IDXMODE-NEXT:    s_nop 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v0, s4
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-IDXMODE-NEXT:    s_add_u32 s4, s0, 0x60
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-IDXMODE-NEXT:    s_addc_u32 s5, s1, 0
 ; VI-IDXMODE-NEXT:    s_add_u32 s0, s0, 0x50
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[0:1], v[28:31]
@@ -7798,29 +7798,29 @@ define amdgpu_kernel void @extract_largest_inbounds_offset(ptr addrspace(1) %out
 ; VI-MOVREL-LABEL: extract_largest_inbounds_offset:
 ; VI-MOVREL:       ; %bb.0: ; %entry
 ; VI-MOVREL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; VI-MOVREL-NEXT:    s_load_dword s6, s[4:5], 0x34
+; VI-MOVREL-NEXT:    s_load_dword s4, s[4:5], 0x34
 ; VI-MOVREL-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[0:3], v[0:1] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
-; VI-MOVREL-NEXT:    s_add_u32 s4, s2, 48
-; VI-MOVREL-NEXT:    s_addc_u32 s5, s3, 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s4
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s5
-; VI-MOVREL-NEXT:    s_add_u32 s4, s2, 32
-; VI-MOVREL-NEXT:    s_addc_u32 s5, s3, 0
+; VI-MOVREL-NEXT:    s_add_u32 s5, s2, 48
+; VI-MOVREL-NEXT:    s_addc_u32 s6, s3, 0
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s5
+; VI-MOVREL-NEXT:    s_add_u32 s5, s2, 32
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s6
+; VI-MOVREL-NEXT:    s_addc_u32 s6, s3, 0
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s2, 16
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[12:15], v[4:5] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s4
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s5
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s5
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s6
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s3, 0
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[8:11], v[4:5] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s2
-; VI-MOVREL-NEXT:    s_add_i32 m0, s6, 15
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s3
+; VI-MOVREL-NEXT:    s_add_i32 m0, s4, 15
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[4:7], v[4:5] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, s0
@@ -7832,14 +7832,14 @@ define amdgpu_kernel void @extract_largest_inbounds_offset(ptr addrspace(1) %out
 ; VI-IDXMODE-LABEL: extract_largest_inbounds_offset:
 ; VI-IDXMODE:       ; %bb.0: ; %entry
 ; VI-IDXMODE-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; VI-IDXMODE-NEXT:    s_load_dword s6, s[4:5], 0x34
+; VI-IDXMODE-NEXT:    s_load_dword s4, s[4:5], 0x34
 ; VI-IDXMODE-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-IDXMODE-NEXT:    s_add_u32 s4, s2, 48
-; VI-IDXMODE-NEXT:    s_addc_u32 s5, s3, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s4
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s5
-; VI-IDXMODE-NEXT:    s_add_u32 s4, s2, 32
-; VI-IDXMODE-NEXT:    s_addc_u32 s5, s3, 0
+; VI-IDXMODE-NEXT:    s_add_u32 s5, s2, 48
+; VI-IDXMODE-NEXT:    s_addc_u32 s6, s3, 0
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s5
+; VI-IDXMODE-NEXT:    s_add_u32 s5, s2, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s6
+; VI-IDXMODE-NEXT:    s_addc_u32 s6, s3, 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s2, 16
@@ -7847,19 +7847,19 @@ define amdgpu_kernel void @extract_largest_inbounds_offset(ptr addrspace(1) %out
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
 ; VI-IDXMODE-NEXT:    flat_load_dwordx4 v[12:15], v[4:5] glc
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s4
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s5
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s5
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s6
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s3, 0
 ; VI-IDXMODE-NEXT:    flat_load_dwordx4 v[8:11], v[4:5] glc
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s2
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-IDXMODE-NEXT:    flat_load_dwordx4 v[4:7], v[4:5] glc
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s1
-; VI-IDXMODE-NEXT:    s_add_i32 s6, s6, 15
-; VI-IDXMODE-NEXT:    s_set_gpr_idx_on s6, gpr_idx(SRC0)
+; VI-IDXMODE-NEXT:    s_add_i32 s4, s4, 15
+; VI-IDXMODE-NEXT:    s_set_gpr_idx_on s4, gpr_idx(SRC0)
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v0, v0
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
 ; VI-IDXMODE-NEXT:    flat_store_dword v[16:17], v0
@@ -8061,29 +8061,29 @@ define amdgpu_kernel void @extract_out_of_bounds_offset(ptr addrspace(1) %out, p
 ; VI-MOVREL-LABEL: extract_out_of_bounds_offset:
 ; VI-MOVREL:       ; %bb.0: ; %entry
 ; VI-MOVREL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; VI-MOVREL-NEXT:    s_load_dword s6, s[4:5], 0x34
+; VI-MOVREL-NEXT:    s_load_dword s4, s[4:5], 0x34
 ; VI-MOVREL-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[0:3], v[0:1] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
-; VI-MOVREL-NEXT:    s_add_u32 s4, s2, 48
-; VI-MOVREL-NEXT:    s_addc_u32 s5, s3, 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s4
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s5
-; VI-MOVREL-NEXT:    s_add_u32 s4, s2, 32
-; VI-MOVREL-NEXT:    s_addc_u32 s5, s3, 0
+; VI-MOVREL-NEXT:    s_add_u32 s5, s2, 48
+; VI-MOVREL-NEXT:    s_addc_u32 s6, s3, 0
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s5
+; VI-MOVREL-NEXT:    s_add_u32 s5, s2, 32
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s6
+; VI-MOVREL-NEXT:    s_addc_u32 s6, s3, 0
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s2, 16
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[12:15], v[4:5] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s4
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s5
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s5
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s6
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s3, 0
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[8:11], v[4:5] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s2
-; VI-MOVREL-NEXT:    s_add_i32 m0, s6, 16
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s3
+; VI-MOVREL-NEXT:    s_add_i32 m0, s4, 16
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[4:7], v[4:5] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, s0
@@ -8095,14 +8095,14 @@ define amdgpu_kernel void @extract_out_of_bounds_offset(ptr addrspace(1) %out, p
 ; VI-IDXMODE-LABEL: extract_out_of_bounds_offset:
 ; VI-IDXMODE:       ; %bb.0: ; %entry
 ; VI-IDXMODE-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; VI-IDXMODE-NEXT:    s_load_dword s6, s[4:5], 0x34
+; VI-IDXMODE-NEXT:    s_load_dword s4, s[4:5], 0x34
 ; VI-IDXMODE-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-IDXMODE-NEXT:    s_add_u32 s4, s2, 48
-; VI-IDXMODE-NEXT:    s_addc_u32 s5, s3, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s4
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s5
-; VI-IDXMODE-NEXT:    s_add_u32 s4, s2, 32
-; VI-IDXMODE-NEXT:    s_addc_u32 s5, s3, 0
+; VI-IDXMODE-NEXT:    s_add_u32 s5, s2, 48
+; VI-IDXMODE-NEXT:    s_addc_u32 s6, s3, 0
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s5
+; VI-IDXMODE-NEXT:    s_add_u32 s5, s2, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s6
+; VI-IDXMODE-NEXT:    s_addc_u32 s6, s3, 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s2, 16
@@ -8110,19 +8110,19 @@ define amdgpu_kernel void @extract_out_of_bounds_offset(ptr addrspace(1) %out, p
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
 ; VI-IDXMODE-NEXT:    flat_load_dwordx4 v[12:15], v[4:5] glc
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s4
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s5
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s5
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s6
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s3, 0
 ; VI-IDXMODE-NEXT:    flat_load_dwordx4 v[8:11], v[4:5] glc
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s2
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-IDXMODE-NEXT:    flat_load_dwordx4 v[4:7], v[4:5] glc
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s1
-; VI-IDXMODE-NEXT:    s_add_i32 s6, s6, 16
-; VI-IDXMODE-NEXT:    s_set_gpr_idx_on s6, gpr_idx(SRC0)
+; VI-IDXMODE-NEXT:    s_add_i32 s4, s4, 16
+; VI-IDXMODE-NEXT:    s_set_gpr_idx_on s4, gpr_idx(SRC0)
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v0, v0
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
 ; VI-IDXMODE-NEXT:    flat_store_dword v[16:17], v0
@@ -8324,29 +8324,29 @@ define amdgpu_kernel void @extractelement_v16i32_or_index(ptr addrspace(1) %out,
 ; VI-MOVREL-LABEL: extractelement_v16i32_or_index:
 ; VI-MOVREL:       ; %bb.0: ; %entry
 ; VI-MOVREL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; VI-MOVREL-NEXT:    s_load_dword s6, s[4:5], 0x34
+; VI-MOVREL-NEXT:    s_load_dword s4, s[4:5], 0x34
 ; VI-MOVREL-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[0:3], v[0:1] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
-; VI-MOVREL-NEXT:    s_add_u32 s4, s2, 48
-; VI-MOVREL-NEXT:    s_addc_u32 s5, s3, 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s4
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s5
-; VI-MOVREL-NEXT:    s_add_u32 s4, s2, 32
-; VI-MOVREL-NEXT:    s_addc_u32 s5, s3, 0
+; VI-MOVREL-NEXT:    s_add_u32 s5, s2, 48
+; VI-MOVREL-NEXT:    s_addc_u32 s6, s3, 0
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s5
+; VI-MOVREL-NEXT:    s_add_u32 s5, s2, 32
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s6
+; VI-MOVREL-NEXT:    s_addc_u32 s6, s3, 0
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s2, 16
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[12:15], v[4:5] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s4
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s5
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s5
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s6
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s3, 0
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[8:11], v[4:5] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v4, s2
-; VI-MOVREL-NEXT:    s_lshl_b32 m0, s6, 2
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s3
+; VI-MOVREL-NEXT:    s_lshl_b32 m0, s4, 2
 ; VI-MOVREL-NEXT:    flat_load_dwordx4 v[4:7], v[4:5] glc
 ; VI-MOVREL-NEXT:    s_waitcnt vmcnt(0)
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, s0
@@ -8358,14 +8358,14 @@ define amdgpu_kernel void @extractelement_v16i32_or_index(ptr addrspace(1) %out,
 ; VI-IDXMODE-LABEL: extractelement_v16i32_or_index:
 ; VI-IDXMODE:       ; %bb.0: ; %entry
 ; VI-IDXMODE-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; VI-IDXMODE-NEXT:    s_load_dword s6, s[4:5], 0x34
+; VI-IDXMODE-NEXT:    s_load_dword s4, s[4:5], 0x34
 ; VI-IDXMODE-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-IDXMODE-NEXT:    s_add_u32 s4, s2, 48
-; VI-IDXMODE-NEXT:    s_addc_u32 s5, s3, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s4
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s5
-; VI-IDXMODE-NEXT:    s_add_u32 s4, s2, 32
-; VI-IDXMODE-NEXT:    s_addc_u32 s5, s3, 0
+; VI-IDXMODE-NEXT:    s_add_u32 s5, s2, 48
+; VI-IDXMODE-NEXT:    s_addc_u32 s6, s3, 0
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s5
+; VI-IDXMODE-NEXT:    s_add_u32 s5, s2, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s6
+; VI-IDXMODE-NEXT:    s_addc_u32 s6, s3, 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s2, 16
@@ -8373,18 +8373,18 @@ define amdgpu_kernel void @extractelement_v16i32_or_index(ptr addrspace(1) %out,
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
 ; VI-IDXMODE-NEXT:    flat_load_dwordx4 v[12:15], v[4:5] glc
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s4
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s5
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s5
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s6
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s3, 0
 ; VI-IDXMODE-NEXT:    flat_load_dwordx4 v[8:11], v[4:5] glc
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v4, s2
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-IDXMODE-NEXT:    flat_load_dwordx4 v[4:7], v[4:5] glc
 ; VI-IDXMODE-NEXT:    s_waitcnt vmcnt(0)
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s1
-; VI-IDXMODE-NEXT:    s_lshl_b32 s0, s6, 2
+; VI-IDXMODE-NEXT:    s_lshl_b32 s0, s4, 2
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_on s0, gpr_idx(SRC0)
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v0, v1
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
@@ -8602,20 +8602,20 @@ define amdgpu_kernel void @insertelement_v16f32_or_index(ptr addrspace(1) %out,
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v15, s23
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    v_movreld_b32_e32 v1, v16
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v16, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 32
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v17, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-MOVREL-NEXT:    s_add_u32 s2, s0, 16
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-MOVREL-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-MOVREL-NEXT:    s_nop 0
-; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v8, s2
+; VI-MOVREL-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-MOVREL-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-MOVREL-NEXT:    s_nop 0
 ; VI-MOVREL-NEXT:    v_mov_b32_e32 v5, s1
@@ -8630,7 +8630,7 @@ define amdgpu_kernel void @insertelement_v16f32_or_index(ptr addrspace(1) %out,
 ; VI-IDXMODE-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, 0x40a00000
 ; VI-IDXMODE-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-IDXMODE-NEXT:    s_lshl_b32 s3, s2, 2
+; VI-IDXMODE-NEXT:    s_lshl_b32 s2, s2, 2
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v0, s8
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v1, s9
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v2, s10
@@ -8647,25 +8647,25 @@ define amdgpu_kernel void @insertelement_v16f32_or_index(ptr addrspace(1) %out,
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s21
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v14, s22
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v15, s23
-; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 48
-; VI-IDXMODE-NEXT:    s_set_gpr_idx_on s3, gpr_idx(DST)
+; VI-IDXMODE-NEXT:    s_add_u32 s3, s0, 48
+; VI-IDXMODE-NEXT:    s_set_gpr_idx_on s2, gpr_idx(DST)
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v1, v16
 ; VI-IDXMODE-NEXT:    s_set_gpr_idx_off
-; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s3
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s2
+; VI-IDXMODE-NEXT:    s_addc_u32 s2, s1, 0
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v17, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 32
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v16, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v12, s2
 ; VI-IDXMODE-NEXT:    s_add_u32 s2, s0, 16
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v13, s3
 ; VI-IDXMODE-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; VI-IDXMODE-NEXT:    s_nop 0
-; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v8, s2
+; VI-IDXMODE-NEXT:    v_mov_b32_e32 v9, s3
 ; VI-IDXMODE-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-IDXMODE-NEXT:    s_nop 0
 ; VI-IDXMODE-NEXT:    v_mov_b32_e32 v5, s1
diff --git a/llvm/test/CodeGen/AMDGPU/insert_vector_dynelt.ll b/llvm/test/CodeGen/AMDGPU/insert_vector_dynelt.ll
index 4f0083bdbd915..7c37b5d3df881 100644
--- a/llvm/test/CodeGen/AMDGPU/insert_vector_dynelt.ll
+++ b/llvm/test/CodeGen/AMDGPU/insert_vector_dynelt.ll
@@ -219,8 +219,8 @@ define amdgpu_kernel void @float8_inselt(ptr addrspace(1) %out, <8 x float> %vec
 ; GCN-NEXT:    v_mov_b32_e32 v7, s15
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    v_movreld_b32_e32 v0, 1.0
-; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    v_mov_b32_e32 v8, s2
+; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v5, s1
@@ -294,11 +294,11 @@ define amdgpu_kernel void @float16_inselt(ptr addrspace(1) %out, <16 x float> %v
 ; GCN:       ; %bb.0: ; %entry
 ; GCN-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
 ; GCN-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GCN-NEXT:    s_load_dword s4, s[4:5], 0xa4
+; GCN-NEXT:    s_load_dword s2, s[4:5], 0xa4
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NEXT:    s_add_u32 s2, s0, 48
-; GCN-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NEXT:    s_add_u32 s3, s0, 48
+; GCN-NEXT:    s_addc_u32 s4, s1, 0
 ; GCN-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-NEXT:    v_mov_b32_e32 v2, s10
 ; GCN-NEXT:    v_mov_b32_e32 v3, s11
@@ -314,22 +314,22 @@ define amdgpu_kernel void @float16_inselt(ptr addrspace(1) %out, <16 x float> %v
 ; GCN-NEXT:    v_mov_b32_e32 v13, s21
 ; GCN-NEXT:    v_mov_b32_e32 v14, s22
 ; GCN-NEXT:    v_mov_b32_e32 v15, s23
-; GCN-NEXT:    s_mov_b32 m0, s4
-; GCN-NEXT:    v_mov_b32_e32 v17, s3
-; GCN-NEXT:    v_mov_b32_e32 v16, s2
+; GCN-NEXT:    s_mov_b32 m0, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-NEXT:    v_movreld_b32_e32 v0, 1.0
+; GCN-NEXT:    v_mov_b32_e32 v16, s3
+; GCN-NEXT:    v_mov_b32_e32 v17, s4
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 16
+; GCN-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    v_mov_b32_e32 v8, s2
+; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v5, s1
@@ -488,45 +488,45 @@ define amdgpu_kernel void @float32_inselt(ptr addrspace(1) %out, <32 x float> %v
 ; GCN-NEXT:    v_mov_b32_e32 v29, s21
 ; GCN-NEXT:    v_mov_b32_e32 v30, s22
 ; GCN-NEXT:    v_mov_b32_e32 v31, s23
-; GCN-NEXT:    v_mov_b32_e32 v33, s3
 ; GCN-NEXT:    v_mov_b32_e32 v32, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 0x60
 ; GCN-NEXT:    v_movreld_b32_e32 v0, 1.0
+; GCN-NEXT:    v_mov_b32_e32 v33, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[32:33], v[28:31]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v29, s3
 ; GCN-NEXT:    v_mov_b32_e32 v28, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 0x50
+; GCN-NEXT:    v_mov_b32_e32 v29, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[28:29], v[24:27]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-NEXT:    v_mov_b32_e32 v24, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 64
+; GCN-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[24:25], v[20:23]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-NEXT:    v_mov_b32_e32 v20, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 48
+; GCN-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[20:21], v[16:19]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-NEXT:    v_mov_b32_e32 v16, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 32
+; GCN-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 16
+; GCN-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    v_mov_b32_e32 v8, s2
+; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v5, s1
@@ -2133,27 +2133,27 @@ define amdgpu_kernel void @double5_inselt(ptr addrspace(1) %out, <5 x double> %v
 ; GCN-NEXT:    s_cselect_b32 s3, 0x3ff00000, s3
 ; GCN-NEXT:    s_cselect_b32 s2, 0, s2
 ; GCN-NEXT:    s_cmp_eq_u32 s12, 0
-; GCN-NEXT:    s_cselect_b32 s13, 0x3ff00000, s1
-; GCN-NEXT:    s_cselect_b32 s14, 0, s0
+; GCN-NEXT:    s_cselect_b32 s1, 0x3ff00000, s1
+; GCN-NEXT:    s_cselect_b32 s0, 0, s0
 ; GCN-NEXT:    s_cmp_eq_u32 s12, 3
-; GCN-NEXT:    s_cselect_b32 s0, 0x3ff00000, s7
-; GCN-NEXT:    s_cselect_b32 s1, 0, s6
+; GCN-NEXT:    s_cselect_b32 s7, 0x3ff00000, s7
+; GCN-NEXT:    s_cselect_b32 s6, 0, s6
 ; GCN-NEXT:    s_cmp_eq_u32 s12, 2
-; GCN-NEXT:    s_cselect_b32 s5, 0x3ff00000, s5
 ; GCN-NEXT:    s_cselect_b32 s4, 0, s4
-; GCN-NEXT:    v_mov_b32_e32 v3, s0
-; GCN-NEXT:    s_add_u32 s0, s10, 16
-; GCN-NEXT:    v_mov_b32_e32 v2, s1
-; GCN-NEXT:    s_addc_u32 s1, s11, 0
+; GCN-NEXT:    s_cselect_b32 s5, 0x3ff00000, s5
 ; GCN-NEXT:    v_mov_b32_e32 v0, s4
+; GCN-NEXT:    s_add_u32 s4, s10, 16
 ; GCN-NEXT:    v_mov_b32_e32 v1, s5
-; GCN-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NEXT:    s_add_u32 s0, s10, 32
+; GCN-NEXT:    s_addc_u32 s5, s11, 0
+; GCN-NEXT:    v_mov_b32_e32 v2, s6
+; GCN-NEXT:    v_mov_b32_e32 v3, s7
+; GCN-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NEXT:    v_mov_b32_e32 v4, s10
-; GCN-NEXT:    v_mov_b32_e32 v0, s14
-; GCN-NEXT:    v_mov_b32_e32 v1, s13
+; GCN-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NEXT:    s_add_u32 s0, s10, 32
+; GCN-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN-NEXT:    v_mov_b32_e32 v2, s2
 ; GCN-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-NEXT:    v_mov_b32_e32 v5, s11
@@ -2162,8 +2162,8 @@ define amdgpu_kernel void @double5_inselt(ptr addrspace(1) %out, <5 x double> %v
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s8
 ; GCN-NEXT:    v_mov_b32_e32 v1, s9
-; GCN-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN-NEXT:    v_mov_b32_e32 v2, s0
+; GCN-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; GCN-NEXT:    s_endpgm
 ;
@@ -2380,20 +2380,20 @@ define amdgpu_kernel void @double8_inselt(ptr addrspace(1) %out, <8 x double> %v
 ; GCN-NEXT:    v_movreld_b32_e32 v0, 0
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    v_movreld_b32_e32 v1, v16
-; GCN-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-NEXT:    v_mov_b32_e32 v16, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 32
+; GCN-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 16
+; GCN-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    v_mov_b32_e32 v8, s2
+; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v5, s1
@@ -2548,8 +2548,8 @@ define amdgpu_kernel void @double7_inselt(ptr addrspace(1) %out, <7 x double> %v
 ; GCN-NEXT:    s_add_u32 s0, s6, 16
 ; GCN-NEXT:    v_movreld_b32_e32 v1, v16
 ; GCN-NEXT:    s_addc_u32 s1, s7, 0
-; GCN-NEXT:    v_mov_b32_e32 v15, s1
 ; GCN-NEXT:    v_mov_b32_e32 v14, s0
+; GCN-NEXT:    v_mov_b32_e32 v15, s1
 ; GCN-NEXT:    s_add_u32 s0, s6, 48
 ; GCN-NEXT:    flat_store_dwordx4 v[14:15], v[4:7]
 ; GCN-NEXT:    s_addc_u32 s1, s7, 0
@@ -2558,8 +2558,8 @@ define amdgpu_kernel void @double7_inselt(ptr addrspace(1) %out, <7 x double> %v
 ; GCN-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN-NEXT:    s_add_u32 s0, s6, 32
+; GCN-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN-NEXT:    s_addc_u32 s1, s7, 0
 ; GCN-NEXT:    flat_store_dwordx2 v[0:1], v[12:13]
 ; GCN-NEXT:    v_mov_b32_e32 v0, s0
@@ -2848,44 +2848,44 @@ define amdgpu_kernel void @double16_inselt(ptr addrspace(1) %out, <16 x double>
 ; GCN-NEXT:    v_movreld_b32_e32 v0, 0
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    v_movreld_b32_e32 v1, v32
-; GCN-NEXT:    v_mov_b32_e32 v33, s3
 ; GCN-NEXT:    v_mov_b32_e32 v32, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-NEXT:    v_mov_b32_e32 v33, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[32:33], v[28:31]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v29, s3
 ; GCN-NEXT:    v_mov_b32_e32 v28, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 0x50
+; GCN-NEXT:    v_mov_b32_e32 v29, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[28:29], v[24:27]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-NEXT:    v_mov_b32_e32 v24, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 64
+; GCN-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[24:25], v[20:23]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-NEXT:    v_mov_b32_e32 v20, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 48
+; GCN-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[20:21], v[16:19]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-NEXT:    v_mov_b32_e32 v16, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 32
+; GCN-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 16
+; GCN-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    v_mov_b32_e32 v8, s2
+; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v5, s1
@@ -3229,33 +3229,33 @@ define amdgpu_kernel void @double15_inselt(ptr addrspace(1) %out, <15 x double>
 ; GCN-NEXT:    s_add_u32 s2, s0, 0x50
 ; GCN-NEXT:    v_movreld_b32_e32 v1, v32
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NEXT:    v_mov_b32_e32 v31, s3
 ; GCN-NEXT:    v_mov_b32_e32 v30, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 64
+; GCN-NEXT:    v_mov_b32_e32 v31, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[30:31], v[20:23]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-NEXT:    v_mov_b32_e32 v20, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 48
+; GCN-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[20:21], v[16:19]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-NEXT:    v_mov_b32_e32 v16, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 32
+; GCN-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 16
+; GCN-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    v_mov_b32_e32 v8, s2
 ; GCN-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; GCN-NEXT:    s_nop 0
diff --git a/llvm/test/CodeGen/AMDGPU/kernel-args.ll b/llvm/test/CodeGen/AMDGPU/kernel-args.ll
index 5b95ef59499ac..128bd47657f38 100644
--- a/llvm/test/CodeGen/AMDGPU/kernel-args.ll
+++ b/llvm/test/CodeGen/AMDGPU/kernel-args.ll
@@ -867,8 +867,8 @@ define amdgpu_kernel void @v3i8_arg(ptr addrspace(1) nocapture %out, <3 x i8> %i
 ; VI-NEXT:    s_add_u32 s0, s0, 2
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v5, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
 ; VI-NEXT:    flat_store_byte v[2:3], v5
 ; VI-NEXT:    flat_store_short v[0:1], v4
@@ -1001,12 +1001,12 @@ define amdgpu_kernel void @v3i16_arg(ptr addrspace(1) nocapture %out, <3 x i16>
 ; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s4, s0, 4
-; VI-NEXT:    s_addc_u32 s5, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v4, s3
-; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_addc_u32 s0, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s4
+; VI-NEXT:    v_mov_b32_e32 v3, s0
 ; VI-NEXT:    v_mov_b32_e32 v5, s2
 ; VI-NEXT:    flat_store_short v[2:3], v4
 ; VI-NEXT:    flat_store_dword v[0:1], v5
@@ -1540,12 +1540,12 @@ define amdgpu_kernel void @v5i8_arg(ptr addrspace(1) nocapture %out, <5 x i8> %i
 ; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s4, s0, 4
-; VI-NEXT:    s_addc_u32 s5, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v4, s3
-; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_addc_u32 s0, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s4
+; VI-NEXT:    v_mov_b32_e32 v3, s0
 ; VI-NEXT:    v_mov_b32_e32 v5, s2
 ; VI-NEXT:    flat_store_byte v[2:3], v4
 ; VI-NEXT:    flat_store_dword v[0:1], v5
@@ -1684,12 +1684,12 @@ define amdgpu_kernel void @v5i16_arg(ptr addrspace(1) nocapture %out, <5 x i16>
 ; VI-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x34
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s4, s0, 8
-; VI-NEXT:    s_addc_u32 s5, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v4, s6
-; VI-NEXT:    v_mov_b32_e32 v2, s4
-; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    v_mov_b32_e32 v0, s0
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_addc_u32 s0, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s6
+; VI-NEXT:    v_mov_b32_e32 v2, s4
+; VI-NEXT:    v_mov_b32_e32 v3, s0
 ; VI-NEXT:    flat_store_short v[2:3], v4
 ; VI-NEXT:    v_mov_b32_e32 v2, s2
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
@@ -2011,11 +2011,11 @@ define amdgpu_kernel void @v5f32_arg(ptr addrspace(1) nocapture %out, <5 x float
 ; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x44
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s4, s6, 16
-; VI-NEXT:    s_addc_u32 s5, s7, 0
 ; VI-NEXT:    v_mov_b32_e32 v3, s8
-; VI-NEXT:    v_mov_b32_e32 v1, s4
-; VI-NEXT:    v_mov_b32_e32 v2, s5
 ; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    s_addc_u32 s0, s7, 0
+; VI-NEXT:    v_mov_b32_e32 v1, s4
+; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    flat_store_dword v[1:2], v3
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s2
@@ -2113,12 +2113,12 @@ define amdgpu_kernel void @v5i64_arg(ptr addrspace(1) nocapture %out, <5 x i64>
 ; VI-NEXT:    s_load_dwordx8 s[0:7], s[4:5], 0x64
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s12, s8, 32
-; VI-NEXT:    s_addc_u32 s13, s9, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, s10
-; VI-NEXT:    v_mov_b32_e32 v2, s11
 ; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    s_addc_u32 s4, s9, 0
+; VI-NEXT:    v_mov_b32_e32 v2, s11
 ; VI-NEXT:    v_mov_b32_e32 v3, s12
-; VI-NEXT:    v_mov_b32_e32 v4, s13
+; VI-NEXT:    v_mov_b32_e32 v4, s4
 ; VI-NEXT:    s_add_u32 s4, s8, 16
 ; VI-NEXT:    flat_store_dwordx2 v[3:4], v[1:2]
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
@@ -2248,12 +2248,12 @@ define amdgpu_kernel void @v5f64_arg(ptr addrspace(1) nocapture %out, <5 x doubl
 ; VI-NEXT:    s_load_dwordx8 s[0:7], s[4:5], 0x64
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s12, s8, 32
-; VI-NEXT:    s_addc_u32 s13, s9, 0
 ; VI-NEXT:    v_mov_b32_e32 v1, s10
-; VI-NEXT:    v_mov_b32_e32 v2, s11
 ; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    s_addc_u32 s4, s9, 0
+; VI-NEXT:    v_mov_b32_e32 v2, s11
 ; VI-NEXT:    v_mov_b32_e32 v3, s12
-; VI-NEXT:    v_mov_b32_e32 v4, s13
+; VI-NEXT:    v_mov_b32_e32 v4, s4
 ; VI-NEXT:    s_add_u32 s4, s8, 16
 ; VI-NEXT:    flat_store_dwordx2 v[3:4], v[1:2]
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
@@ -2882,8 +2882,8 @@ define amdgpu_kernel void @v8i32_arg(ptr addrspace(1) nocapture %out, <8 x i32>
 ; VI-NEXT:    v_mov_b32_e32 v1, s13
 ; VI-NEXT:    v_mov_b32_e32 v2, s14
 ; VI-NEXT:    v_mov_b32_e32 v3, s15
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v0, s8
@@ -2989,8 +2989,8 @@ define amdgpu_kernel void @v8f32_arg(ptr addrspace(1) nocapture %out, <8 x float
 ; VI-NEXT:    v_mov_b32_e32 v1, s13
 ; VI-NEXT:    v_mov_b32_e32 v2, s14
 ; VI-NEXT:    v_mov_b32_e32 v3, s15
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v0, s8
@@ -3547,8 +3547,8 @@ define amdgpu_kernel void @v16i16_arg(ptr addrspace(1) %out, <16 x i16> %in) {
 ; VI-NEXT:    v_mov_b32_e32 v1, s13
 ; VI-NEXT:    v_mov_b32_e32 v2, s14
 ; VI-NEXT:    v_mov_b32_e32 v3, s15
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v0, s8
@@ -4009,29 +4009,29 @@ define amdgpu_kernel void @v16i32_arg(ptr addrspace(1) nocapture %out, <16 x i32
 ; VI-NEXT:    v_mov_b32_e32 v0, s20
 ; VI-NEXT:    s_add_u32 s2, s0, 48
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 32
 ; VI-NEXT:    v_mov_b32_e32 v1, s21
 ; VI-NEXT:    v_mov_b32_e32 v2, s22
 ; VI-NEXT:    v_mov_b32_e32 v3, s23
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 16
 ; VI-NEXT:    v_mov_b32_e32 v0, s16
 ; VI-NEXT:    v_mov_b32_e32 v1, s17
 ; VI-NEXT:    v_mov_b32_e32 v2, s18
 ; VI-NEXT:    v_mov_b32_e32 v3, s19
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; VI-NEXT:    v_mov_b32_e32 v5, s3
+; VI-NEXT:    v_mov_b32_e32 v4, s2
 ; VI-NEXT:    v_mov_b32_e32 v0, s12
 ; VI-NEXT:    v_mov_b32_e32 v1, s13
 ; VI-NEXT:    v_mov_b32_e32 v2, s14
 ; VI-NEXT:    v_mov_b32_e32 v3, s15
-; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v0, s8
@@ -4186,29 +4186,29 @@ define amdgpu_kernel void @v16f32_arg(ptr addrspace(1) nocapture %out, <16 x flo
 ; VI-NEXT:    v_mov_b32_e32 v0, s20
 ; VI-NEXT:    s_add_u32 s2, s0, 48
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 32
 ; VI-NEXT:    v_mov_b32_e32 v1, s21
 ; VI-NEXT:    v_mov_b32_e32 v2, s22
 ; VI-NEXT:    v_mov_b32_e32 v3, s23
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    v_mov_b32_e32 v4, s2
 ; VI-NEXT:    s_add_u32 s2, s0, 16
 ; VI-NEXT:    v_mov_b32_e32 v0, s16
 ; VI-NEXT:    v_mov_b32_e32 v1, s17
 ; VI-NEXT:    v_mov_b32_e32 v2, s18
 ; VI-NEXT:    v_mov_b32_e32 v3, s19
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    s_addc_u32 s3, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; VI-NEXT:    v_mov_b32_e32 v5, s3
+; VI-NEXT:    v_mov_b32_e32 v4, s2
 ; VI-NEXT:    v_mov_b32_e32 v0, s12
 ; VI-NEXT:    v_mov_b32_e32 v1, s13
 ; VI-NEXT:    v_mov_b32_e32 v2, s14
 ; VI-NEXT:    v_mov_b32_e32 v3, s15
-; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s3
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v0, s8
@@ -4486,8 +4486,8 @@ define amdgpu_kernel void @i65_arg(ptr addrspace(1) nocapture %out, i65 %in) nou
 ; VI-NEXT:    s_add_u32 s0, s0, 8
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v6, s4
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s2
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-NEXT:    flat_store_byte v[4:5], v6
@@ -5173,11 +5173,11 @@ define amdgpu_kernel void @packed_struct_argument_alignment(<{i32, i64}> %arg0,
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-NEXT:    s_add_u32 s0, s0, 3
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-NEXT:    s_add_u32 s0, s4, 51
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
 ; VI-NEXT:    v_mov_b32_e32 v6, s0
 ; VI-NEXT:    v_mov_b32_e32 v7, s1
@@ -5623,8 +5623,8 @@ define amdgpu_kernel void @array_3xi16(i8 %arg0, [3 x i16] %arg1) {
 ; VI-NEXT:    s_add_u32 s0, s4, 42
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
-; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
 ; VI-NEXT:    v_mov_b32_e32 v1, s3
 ; VI-NEXT:    flat_load_ushort v4, v[4:5]
 ; VI-NEXT:    flat_load_ushort v2, v[2:3]
@@ -5944,36 +5944,36 @@ define amdgpu_kernel void @byref_natural_align_constant_v16i32_arg(ptr addrspace
 ; VI:       ; %bb.0:
 ; VI-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
 ; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
-; VI-NEXT:    s_load_dword s4, s[4:5], 0xa4
+; VI-NEXT:    s_load_dword s2, s[4:5], 0xa4
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v0, s20
-; VI-NEXT:    s_add_u32 s2, s0, 48
-; VI-NEXT:    s_addc_u32 s3, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v5, s3
-; VI-NEXT:    v_mov_b32_e32 v4, s2
-; VI-NEXT:    s_add_u32 s2, s0, 32
+; VI-NEXT:    s_add_u32 s3, s0, 48
+; VI-NEXT:    s_addc_u32 s4, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s3
+; VI-NEXT:    s_add_u32 s3, s0, 32
 ; VI-NEXT:    v_mov_b32_e32 v1, s21
 ; VI-NEXT:    v_mov_b32_e32 v2, s22
 ; VI-NEXT:    v_mov_b32_e32 v3, s23
-; VI-NEXT:    s_addc_u32 s3, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v5, s4
+; VI-NEXT:    s_addc_u32 s4, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
-; VI-NEXT:    v_mov_b32_e32 v5, s3
-; VI-NEXT:    v_mov_b32_e32 v4, s2
-; VI-NEXT:    s_add_u32 s2, s0, 16
+; VI-NEXT:    v_mov_b32_e32 v4, s3
+; VI-NEXT:    s_add_u32 s3, s0, 16
 ; VI-NEXT:    v_mov_b32_e32 v0, s16
 ; VI-NEXT:    v_mov_b32_e32 v1, s17
 ; VI-NEXT:    v_mov_b32_e32 v2, s18
 ; VI-NEXT:    v_mov_b32_e32 v3, s19
-; VI-NEXT:    s_addc_u32 s3, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v5, s4
+; VI-NEXT:    s_addc_u32 s4, s1, 0
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
-; VI-NEXT:    v_mov_b32_e32 v5, s3
+; VI-NEXT:    v_mov_b32_e32 v4, s3
 ; VI-NEXT:    v_mov_b32_e32 v0, s12
 ; VI-NEXT:    v_mov_b32_e32 v1, s13
 ; VI-NEXT:    v_mov_b32_e32 v2, s14
 ; VI-NEXT:    v_mov_b32_e32 v3, s15
-; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v5, s4
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
@@ -5984,7 +5984,7 @@ define amdgpu_kernel void @byref_natural_align_constant_v16i32_arg(ptr addrspace
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    s_waitcnt vmcnt(0)
-; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-NEXT:    flat_store_dword v[4:5], v0
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.dec.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.dec.ll
index 022f0eaffb0e6..c358255304b4f 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.dec.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.dec.ll
@@ -953,8 +953,8 @@ define amdgpu_kernel void @flat_atomic_dec_noret_i64_offset(ptr %ptr) nounwind {
 ; CI-NEXT:    s_waitcnt lgkmcnt(0)
 ; CI-NEXT:    s_add_u32 s0, s0, 32
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; CI-NEXT:    buffer_wbinvl1_vol
@@ -968,8 +968,8 @@ define amdgpu_kernel void @flat_atomic_dec_noret_i64_offset(ptr %ptr) nounwind {
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_dec_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.inc.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.inc.ll
index 74b3b332e0254..2b4e2997f6964 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.inc.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.inc.ll
@@ -1591,8 +1591,8 @@ define amdgpu_kernel void @flat_atomic_inc_noret_i64_offset(ptr %ptr) nounwind {
 ; CI-NEXT:    s_waitcnt lgkmcnt(0)
 ; CI-NEXT:    s_add_u32 s0, s0, 32
 ; CI-NEXT:    s_addc_u32 s1, s1, 0
-; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    v_mov_b32_e32 v2, s0
+; CI-NEXT:    v_mov_b32_e32 v3, s1
 ; CI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; CI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; CI-NEXT:    buffer_wbinvl1_vol
@@ -1606,8 +1606,8 @@ define amdgpu_kernel void @flat_atomic_inc_noret_i64_offset(ptr %ptr) nounwind {
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_add_u32 s0, s0, 32
 ; VI-NEXT:    s_addc_u32 s1, s1, 0
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_atomic_inc_x2 v[2:3], v[0:1]
 ; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; VI-NEXT:    buffer_wbinvl1_vol
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.av.load.b128.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.av.load.b128.ll
index 2495fa8a43b29..96b3af6779e8c 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.av.load.b128.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.av.load.b128.ll
@@ -12568,7 +12568,7 @@ define <4 x float> @global_load_saddr_i8_zext_vgpr_offset_4096(ptr addrspace(1)
 ; GFX1100-ISEL-LABEL: global_load_saddr_i8_zext_vgpr_offset_4096:
 ; GFX1100-ISEL:       ; %bb.0:
 ; GFX1100-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
 ; GFX1100-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1100-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v1, v0
 ; GFX1100-ISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v2, vcc_lo
@@ -12830,7 +12830,7 @@ define <4 x float> @global_load_saddr_i8_zext_vgpr_offset_neg4097(ptr addrspace(
 ; GFX1100-ISEL-LABEL: global_load_saddr_i8_zext_vgpr_offset_neg4097:
 ; GFX1100-ISEL:       ; %bb.0:
 ; GFX1100-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
 ; GFX1100-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1100-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v1, v0
 ; GFX1100-ISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v2, vcc_lo
@@ -13411,7 +13411,7 @@ define <4 x float> @global_load_saddr_i8_zext_vgpr_offset_0x7FFFFF(ptr addrspace
 ; GFX1100-ISEL-LABEL: global_load_saddr_i8_zext_vgpr_offset_0x7FFFFF:
 ; GFX1100-ISEL:       ; %bb.0:
 ; GFX1100-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
 ; GFX1100-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1100-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v1, v0
 ; GFX1100-ISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v2, vcc_lo
@@ -13559,7 +13559,7 @@ define <4 x float> @global_load_saddr_i8_zext_vgpr_offset_0xFFFFFF(ptr addrspace
 ; GFX1100-ISEL-LABEL: global_load_saddr_i8_zext_vgpr_offset_0xFFFFFF:
 ; GFX1100-ISEL:       ; %bb.0:
 ; GFX1100-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
 ; GFX1100-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1100-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v1, v0
 ; GFX1100-ISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v2, vcc_lo
@@ -14844,9 +14844,8 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32(ptr addrspace(1) %vbase,
 ; GFX906-ISEL-LABEL: global_load_saddr_i8_vgpr64_sgpr32:
 ; GFX906-ISEL:       ; %bb.0:
 ; GFX906-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX906-ISEL-NEXT:    s_mov_b32 s17, 0
 ; GFX906-ISEL-NEXT:    v_mov_b32_e32 v2, s16
-; GFX906-ISEL-NEXT:    v_mov_b32_e32 v3, s17
+; GFX906-ISEL-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX906-ISEL-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v2
 ; GFX906-ISEL-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX906-ISEL-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off glc
@@ -14856,8 +14855,8 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32(ptr addrspace(1) %vbase,
 ; GFX942-ISEL-LABEL: global_load_saddr_i8_vgpr64_sgpr32:
 ; GFX942-ISEL:       ; %bb.0:
 ; GFX942-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-ISEL-NEXT:    s_mov_b32 s1, 0
-; GFX942-ISEL-NEXT:    v_mov_b64_e32 v[2:3], s[0:1]
+; GFX942-ISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX942-ISEL-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX942-ISEL-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v2
 ; GFX942-ISEL-NEXT:    s_nop 1
 ; GFX942-ISEL-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v3, vcc
@@ -14868,9 +14867,8 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32(ptr addrspace(1) %vbase,
 ; GFX1012-ISEL-LABEL: global_load_saddr_i8_vgpr64_sgpr32:
 ; GFX1012-ISEL:       ; %bb.0:
 ; GFX1012-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1012-ISEL-NEXT:    s_mov_b32 s17, 0
 ; GFX1012-ISEL-NEXT:    v_mov_b32_e32 v2, s16
-; GFX1012-ISEL-NEXT:    v_mov_b32_e32 v3, s17
+; GFX1012-ISEL-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1012-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
 ; GFX1012-ISEL-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
 ; GFX1012-ISEL-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off glc dlc
@@ -14880,11 +14878,9 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32(ptr addrspace(1) %vbase,
 ; GFX1100-ISEL-LABEL: global_load_saddr_i8_vgpr64_sgpr32:
 ; GFX1100-ISEL:       ; %bb.0:
 ; GFX1100-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-ISEL-NEXT:    s_mov_b32 s1, 0
-; GFX1100-ISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, 0
+; GFX1100-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1100-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX1100-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1100-ISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX1100-ISEL-NEXT:    global_load_b128 v[0:3], v[0:1], off glc
 ; GFX1100-ISEL-NEXT:    s_waitcnt vmcnt(0)
@@ -14894,11 +14890,9 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32(ptr addrspace(1) %vbase,
 ; GFX1250-ISEL:       ; %bb.0:
 ; GFX1250-ISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-ISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1250-ISEL-NEXT:    s_mov_b32 s1, 0
-; GFX1250-ISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-ISEL-NEXT:    v_mov_b64_e32 v[2:3], s[0:1]
+; GFX1250-ISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, 0
+; GFX1250-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX1250-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-ISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX1250-ISEL-NEXT:    global_load_b128 v[0:3], v[0:1], off scope:SCOPE_DEV
 ; GFX1250-ISEL-NEXT:    s_wait_loadcnt 0x0
@@ -14911,11 +14905,9 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32(ptr addrspace(1) %vbase,
 ; GFX1310-ISEL-NEXT:    s_wait_samplecnt 0x0
 ; GFX1310-ISEL-NEXT:    s_wait_bvhcnt 0x0
 ; GFX1310-ISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1310-ISEL-NEXT:    s_mov_b32 s1, 0
-; GFX1310-ISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1310-ISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX1310-ISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, 0
+; GFX1310-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1310-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX1310-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1310-ISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX1310-ISEL-NEXT:    global_load_b128 v[0:3], v[0:1], off scope:SCOPE_DEV
 ; GFX1310-ISEL-NEXT:    s_wait_loadcnt 0x0
@@ -14996,9 +14988,8 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32_offset_4095(ptr addrspace
 ; GFX906-ISEL-LABEL: global_load_saddr_i8_vgpr64_sgpr32_offset_4095:
 ; GFX906-ISEL:       ; %bb.0:
 ; GFX906-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX906-ISEL-NEXT:    s_mov_b32 s17, 0
 ; GFX906-ISEL-NEXT:    v_mov_b32_e32 v2, s16
-; GFX906-ISEL-NEXT:    v_mov_b32_e32 v3, s17
+; GFX906-ISEL-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX906-ISEL-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v2
 ; GFX906-ISEL-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX906-ISEL-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off offset:4095 glc
@@ -15008,8 +14999,8 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32_offset_4095(ptr addrspace
 ; GFX942-ISEL-LABEL: global_load_saddr_i8_vgpr64_sgpr32_offset_4095:
 ; GFX942-ISEL:       ; %bb.0:
 ; GFX942-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-ISEL-NEXT:    s_mov_b32 s1, 0
-; GFX942-ISEL-NEXT:    v_mov_b64_e32 v[2:3], s[0:1]
+; GFX942-ISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX942-ISEL-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX942-ISEL-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v2
 ; GFX942-ISEL-NEXT:    s_nop 1
 ; GFX942-ISEL-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v3, vcc
@@ -15020,9 +15011,8 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32_offset_4095(ptr addrspace
 ; GFX1012-ISEL-LABEL: global_load_saddr_i8_vgpr64_sgpr32_offset_4095:
 ; GFX1012-ISEL:       ; %bb.0:
 ; GFX1012-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1012-ISEL-NEXT:    s_mov_b32 s17, 0
 ; GFX1012-ISEL-NEXT:    v_mov_b32_e32 v2, s16
-; GFX1012-ISEL-NEXT:    v_mov_b32_e32 v3, s17
+; GFX1012-ISEL-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1012-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
 ; GFX1012-ISEL-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
 ; GFX1012-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, 0xfff, v0
@@ -15034,11 +15024,9 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32_offset_4095(ptr addrspace
 ; GFX1100-ISEL-LABEL: global_load_saddr_i8_vgpr64_sgpr32_offset_4095:
 ; GFX1100-ISEL:       ; %bb.0:
 ; GFX1100-ISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-ISEL-NEXT:    s_mov_b32 s1, 0
-; GFX1100-ISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX1100-ISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, 0
+; GFX1100-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1100-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX1100-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1100-ISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX1100-ISEL-NEXT:    global_load_b128 v[0:3], v[0:1], off offset:4095 glc
 ; GFX1100-ISEL-NEXT:    s_waitcnt vmcnt(0)
@@ -15048,11 +15036,9 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32_offset_4095(ptr addrspace
 ; GFX1250-ISEL:       ; %bb.0:
 ; GFX1250-ISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-ISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1250-ISEL-NEXT:    s_mov_b32 s1, 0
-; GFX1250-ISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-ISEL-NEXT:    v_mov_b64_e32 v[2:3], s[0:1]
+; GFX1250-ISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, 0
+; GFX1250-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX1250-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-ISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX1250-ISEL-NEXT:    global_load_b128 v[0:3], v[0:1], off offset:4095 scope:SCOPE_SYS
 ; GFX1250-ISEL-NEXT:    s_wait_loadcnt 0x0
@@ -15065,11 +15051,9 @@ define <4 x float> @global_load_saddr_i8_vgpr64_sgpr32_offset_4095(ptr addrspace
 ; GFX1310-ISEL-NEXT:    s_wait_samplecnt 0x0
 ; GFX1310-ISEL-NEXT:    s_wait_bvhcnt 0x0
 ; GFX1310-ISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1310-ISEL-NEXT:    s_mov_b32 s1, 0
-; GFX1310-ISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1310-ISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
+; GFX1310-ISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, 0
+; GFX1310-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1310-ISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
-; GFX1310-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1310-ISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo
 ; GFX1310-ISEL-NEXT:    global_load_b128 v[0:3], v[0:1], off offset:4095 scope:SCOPE_SYS
 ; GFX1310-ISEL-NEXT:    s_wait_loadcnt 0x0
@@ -15255,8 +15239,8 @@ define <4 x float> @global_load_saddr_f32_natural_addressing(ptr addrspace(1) in
 ; GFX1310-ISEL-NEXT:    s_wait_bvhcnt 0x0
 ; GFX1310-ISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX1310-ISEL-NEXT:    global_load_b32 v0, v[0:1], off
-; GFX1310-ISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v3, s1
-; GFX1310-ISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1310-ISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
+; GFX1310-ISEL-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX1310-ISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1310-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1310-ISEL-NEXT:    v_lshlrev_b64_e32 v[0:1], 2, v[0:1]
@@ -15854,8 +15838,8 @@ define <4 x float> @global_load_f32_saddr_zext_vgpr_range_too_large(ptr addrspac
 ; GFX1310-ISEL-NEXT:    s_wait_bvhcnt 0x0
 ; GFX1310-ISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX1310-ISEL-NEXT:    global_load_b32 v0, v[0:1], off
-; GFX1310-ISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v3, s1
-; GFX1310-ISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1310-ISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s0
+; GFX1310-ISEL-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX1310-ISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1310-ISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1310-ISEL-NEXT:    v_lshlrev_b64_e32 v[0:1], 2, v[0:1]
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.bvh8_intersect_ray.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.bvh8_intersect_ray.ll
index 7aa4816d26413..5c58858a92d48 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.bvh8_intersect_ray.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.bvh8_intersect_ray.ll
@@ -109,24 +109,17 @@ define amdgpu_ps <10 x float> @image_bvh8_intersect_ray_ssssss(i64 inreg %node_p
 ;
 ; GFX12-GISEL-LABEL: image_bvh8_intersect_ray_ssssss:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_mov_b32 s16, s3
-; GFX12-GISEL-NEXT:    s_mov_b32 s17, s4
-; GFX12-GISEL-NEXT:    s_mov_b32 s18, s5
-; GFX12-GISEL-NEXT:    s_mov_b32 s20, s6
-; GFX12-GISEL-NEXT:    s_mov_b32 s21, s7
-; GFX12-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v7, s3 :: v_dual_mov_b32 v6, s2
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v14, s16 :: v_dual_mov_b32 v15, s17
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v16, s18 :: v_dual_mov_b32 v17, s20
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v18, s21 :: v_dual_mov_b32 v19, s22
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s0 :: v_dual_mov_b32 v5, s1
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s2 :: v_dual_mov_b32 v7, 0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v14, s3 :: v_dual_mov_b32 v15, s4
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v16, s5 :: v_dual_mov_b32 v17, s6
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v18, s7 :: v_dual_mov_b32 v19, s8
 ; GFX12-GISEL-NEXT:    v_mov_b32_e32 v8, s9
-; GFX12-GISEL-NEXT:    s_mov_b32 s4, s10
-; GFX12-GISEL-NEXT:    s_mov_b32 s5, s11
-; GFX12-GISEL-NEXT:    s_mov_b32 s6, s12
-; GFX12-GISEL-NEXT:    s_mov_b32 s7, s13
-; GFX12-GISEL-NEXT:    image_bvh8_intersect_ray v[4:13], [v[4:5], v[6:7], v[14:16], v[17:19], v8], s[4:7]
+; GFX12-GISEL-NEXT:    s_mov_b32 s16, s10
+; GFX12-GISEL-NEXT:    s_mov_b32 s17, s11
+; GFX12-GISEL-NEXT:    s_mov_b32 s18, s12
+; GFX12-GISEL-NEXT:    s_mov_b32 s19, s13
+; GFX12-GISEL-NEXT:    image_bvh8_intersect_ray v[4:13], [v[4:5], v[6:7], v[14:16], v[17:19], v8], s[16:19]
 ; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-GISEL-NEXT:    global_store_b96 v[0:1], v[14:16], off
 ; GFX12-GISEL-NEXT:    global_store_b96 v[2:3], v[17:19], off
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.scale.pk.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.scale.pk.ll
index e580abd47d016..68b1d594296d9 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.scale.pk.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.scale.pk.ll
@@ -592,8 +592,8 @@ define amdgpu_ps void @test_cvt_scale_pk16_f16_fp6_ss(<3 x i32> inreg %src, i32
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v11, s1
-; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v10, s0
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v10, s0 :: v_dual_mov_b32 v11, s1
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cvt_scale_pk16_f16_fp6 v[2:9], v[10:12], s3
 ; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
@@ -642,8 +642,8 @@ define amdgpu_ps void @test_cvt_scale_pk16_f16_fp6_sl(<3 x i32> inreg %src, ptr
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v11, s1
-; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v10, s0
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v10, s0 :: v_dual_mov_b32 v11, s1
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cvt_scale_pk16_f16_fp6 v[2:9], v[10:12], 0x64 scale_sel:1
 ; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
@@ -721,8 +721,8 @@ define amdgpu_ps void @test_cvt_scale_pk16_bf16_fp6_sl(<3 x i32> inreg %src, ptr
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v11, s1
-; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v10, s0
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v10, s0 :: v_dual_mov_b32 v11, s1
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cvt_scale_pk16_bf16_fp6 v[2:9], v[10:12], 0x64 scale_sel:3
 ; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
@@ -800,8 +800,8 @@ define amdgpu_ps void @test_cvt_scale_pk16_f16_bf6_ss(<3 x i32> inreg %src, i32
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v11, s1
-; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v10, s0
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v10, s0 :: v_dual_mov_b32 v11, s1
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cvt_scale_pk16_f16_bf6 v[2:9], v[10:12], s3 scale_sel:4
 ; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
@@ -850,8 +850,8 @@ define amdgpu_ps void @test_cvt_scale_pk16_f16_bf6_sl(<3 x i32> inreg %src, ptr
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v11, s1
-; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v10, s0
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v10, s0 :: v_dual_mov_b32 v11, s1
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cvt_scale_pk16_f16_bf6 v[2:9], v[10:12], 0x64 scale_sel:5
 ; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
@@ -929,8 +929,8 @@ define amdgpu_ps void @test_cvt_scale_pk16_bf16_bf6_sl(<3 x i32> inreg %src, ptr
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v11, s1
-; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v10, s0
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v10, s0 :: v_dual_mov_b32 v11, s1
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cvt_scale_pk16_bf16_bf6 v[2:9], v[10:12], 0x64 scale_sel:8
 ; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
@@ -1014,8 +1014,8 @@ define amdgpu_ps void @test_cvt_scale_pk16_f32_fp6_ss(<3 x i32> inreg %src, i32
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v20, s2 :: v_dual_mov_b32 v19, s1
-; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v18, s0
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v18, s0 :: v_dual_mov_b32 v19, s1
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v20, s2
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cvt_scale_pk16_f32_fp6 v[2:17], v[18:20], s3 scale_sel:5
 ; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
@@ -1112,8 +1112,8 @@ define amdgpu_ps void @test_cvt_scale_pk16_f32_bf6_ss(<3 x i32> inreg %src, i32
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v20, s2 :: v_dual_mov_b32 v19, s1
-; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v18, s0
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v18, s0 :: v_dual_mov_b32 v19, s1
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v20, s2
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cvt_scale_pk16_f32_bf6 v[2:17], v[18:20], s3 scale_sel:6
 ; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.scalef32.pk.gfx13.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.scalef32.pk.gfx13.ll
index 4bbb893087686..4fa0089300c3d 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.scalef32.pk.gfx13.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.scalef32.pk.gfx13.ll
@@ -43,8 +43,8 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_bf6_f32_v(ptr addrspace(1) %ou
 ; GFX13-GISEL-NEXT:    s_clause 0x3
 ; GFX13-GISEL-NEXT:    s_load_b512 s[36:51], s[4:5], 0xa4 nv
 ; GFX13-GISEL-NEXT:    s_load_b512 s[52:67], s[4:5], 0xe4 nv
-; GFX13-GISEL-NEXT:    s_load_b32 s0, s[4:5], 0x124 nv
-; GFX13-GISEL-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24 nv
+; GFX13-GISEL-NEXT:    s_load_b32 s2, s[4:5], 0x124 nv
+; GFX13-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX13-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v0, s36 :: v_dual_mov_b32 v1, s37
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v2, s38 :: v_dual_mov_b32 v3, s39
@@ -63,23 +63,23 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_bf6_f32_v(ptr addrspace(1) %ou
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v28, s64 :: v_dual_mov_b32 v29, s65
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v30, s66 :: v_dual_mov_b32 v31, s67
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_bf6_f32 v[0:5], v[0:31], s0
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_bf6_f32 v[0:5], v[0:31], s2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s1, v1
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s0
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s6
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX13-GISEL-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX13-GISEL-NEXT:    s_clause 0x1
-; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[4:5]
-; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[4:5] offset:16
+; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[0:1]
+; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[0:1] offset:16
 ; GFX13-GISEL-NEXT:    s_endpgm
   %cvt = tail call <6 x i32> @llvm.amdgcn.cvt.scalef32.pk32.bf6.f32(<32 x float> %src, float %scale)
   store <6 x i32> %cvt, ptr addrspace(1) %out, align 8
@@ -123,7 +123,7 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_bf6_f32_s(ptr addrspace(1) %ou
 ; GFX13-GISEL-NEXT:    s_clause 0x2
 ; GFX13-GISEL-NEXT:    s_load_b512 s[36:51], s[4:5], 0xa4 nv
 ; GFX13-GISEL-NEXT:    s_load_b512 s[52:67], s[4:5], 0xe4 nv
-; GFX13-GISEL-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24 nv
+; GFX13-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX13-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v0, s36 :: v_dual_mov_b32 v1, s37
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v2, s38 :: v_dual_mov_b32 v3, s39
@@ -143,22 +143,22 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_bf6_f32_s(ptr addrspace(1) %ou
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v30, s66 :: v_dual_mov_b32 v31, s67
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_bf6_f32 v[0:5], v[0:31], 0x42c80000
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s1, v1
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s0
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s6
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX13-GISEL-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX13-GISEL-NEXT:    s_clause 0x1
-; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[4:5]
-; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[4:5] offset:16
+; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[0:1]
+; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[0:1] offset:16
 ; GFX13-GISEL-NEXT:    s_endpgm
   %cvt = tail call <6 x i32> @llvm.amdgcn.cvt.scalef32.pk32.bf6.f32(<32 x float> %src, float 100.0)
   store <6 x i32> %cvt, ptr addrspace(1) %out, align 8
@@ -203,8 +203,8 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_bf6_f32_v_inreg_src(ptr addrsp
 ; GFX13-GISEL-NEXT:    s_clause 0x3
 ; GFX13-GISEL-NEXT:    s_load_b512 s[36:51], s[4:5], 0xa4 nv
 ; GFX13-GISEL-NEXT:    s_load_b512 s[52:67], s[4:5], 0xe4 nv
-; GFX13-GISEL-NEXT:    s_load_b32 s0, s[4:5], 0x124 nv
-; GFX13-GISEL-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24 nv
+; GFX13-GISEL-NEXT:    s_load_b32 s2, s[4:5], 0x124 nv
+; GFX13-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX13-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v0, s36 :: v_dual_mov_b32 v1, s37
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v2, s38 :: v_dual_mov_b32 v3, s39
@@ -223,23 +223,23 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_bf6_f32_v_inreg_src(ptr addrsp
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v28, s64 :: v_dual_mov_b32 v29, s65
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v30, s66 :: v_dual_mov_b32 v31, s67
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_bf6_f32 v[0:5], v[0:31], s0
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_bf6_f32 v[0:5], v[0:31], s2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s1, v1
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s0
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s6
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX13-GISEL-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX13-GISEL-NEXT:    s_clause 0x1
-; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[4:5]
-; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[4:5] offset:16
+; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[0:1]
+; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[0:1] offset:16
 ; GFX13-GISEL-NEXT:    s_endpgm
   %cvt = tail call <6 x i32> @llvm.amdgcn.cvt.scalef32.pk32.bf6.f32(<32 x float> %src, float %scale)
   store <6 x i32> %cvt, ptr addrspace(1) %out, align 8
@@ -283,7 +283,7 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_bf6_f32_s_inreg_src(ptr addrsp
 ; GFX13-GISEL-NEXT:    s_clause 0x2
 ; GFX13-GISEL-NEXT:    s_load_b512 s[36:51], s[4:5], 0xa4 nv
 ; GFX13-GISEL-NEXT:    s_load_b512 s[52:67], s[4:5], 0xe4 nv
-; GFX13-GISEL-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24 nv
+; GFX13-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX13-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v0, s36 :: v_dual_mov_b32 v1, s37
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v2, s38 :: v_dual_mov_b32 v3, s39
@@ -303,22 +303,22 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_bf6_f32_s_inreg_src(ptr addrsp
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v30, s66 :: v_dual_mov_b32 v31, s67
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_bf6_f32 v[0:5], v[0:31], 0x42c80000
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s1, v1
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s0
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s6
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX13-GISEL-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX13-GISEL-NEXT:    s_clause 0x1
-; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[4:5]
-; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[4:5] offset:16
+; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[0:1]
+; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[0:1] offset:16
 ; GFX13-GISEL-NEXT:    s_endpgm
   %cvt = tail call <6 x i32> @llvm.amdgcn.cvt.scalef32.pk32.bf6.f32(<32 x float> %src, float 100.0)
   store <6 x i32> %cvt, ptr addrspace(1) %out, align 8
@@ -363,8 +363,8 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_fp6_f32_v(ptr addrspace(1) %ou
 ; GFX13-GISEL-NEXT:    s_clause 0x3
 ; GFX13-GISEL-NEXT:    s_load_b512 s[36:51], s[4:5], 0xa4 nv
 ; GFX13-GISEL-NEXT:    s_load_b512 s[52:67], s[4:5], 0xe4 nv
-; GFX13-GISEL-NEXT:    s_load_b32 s0, s[4:5], 0x124 nv
-; GFX13-GISEL-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24 nv
+; GFX13-GISEL-NEXT:    s_load_b32 s2, s[4:5], 0x124 nv
+; GFX13-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX13-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v0, s36 :: v_dual_mov_b32 v1, s37
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v2, s38 :: v_dual_mov_b32 v3, s39
@@ -383,23 +383,23 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_fp6_f32_v(ptr addrspace(1) %ou
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v28, s64 :: v_dual_mov_b32 v29, s65
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v30, s66 :: v_dual_mov_b32 v31, s67
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_fp6_f32 v[0:5], v[0:31], s0
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_fp6_f32 v[0:5], v[0:31], s2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s1, v1
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s0
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s6
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX13-GISEL-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX13-GISEL-NEXT:    s_clause 0x1
-; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[4:5]
-; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[4:5] offset:16
+; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[0:1]
+; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[0:1] offset:16
 ; GFX13-GISEL-NEXT:    s_endpgm
   %cvt = tail call <6 x i32> @llvm.amdgcn.cvt.scalef32.pk32.fp6.f32(<32 x float> %src, float %scale)
   store <6 x i32> %cvt, ptr addrspace(1) %out, align 8
@@ -443,7 +443,7 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_fp6_f32_s(ptr addrspace(1) %ou
 ; GFX13-GISEL-NEXT:    s_clause 0x2
 ; GFX13-GISEL-NEXT:    s_load_b512 s[36:51], s[4:5], 0xa4 nv
 ; GFX13-GISEL-NEXT:    s_load_b512 s[52:67], s[4:5], 0xe4 nv
-; GFX13-GISEL-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24 nv
+; GFX13-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX13-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v0, s36 :: v_dual_mov_b32 v1, s37
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v2, s38 :: v_dual_mov_b32 v3, s39
@@ -463,22 +463,22 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_fp6_f32_s(ptr addrspace(1) %ou
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v30, s66 :: v_dual_mov_b32 v31, s67
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_fp6_f32 v[0:5], v[0:31], 0x42c80000
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s1, v1
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s0
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s6
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX13-GISEL-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX13-GISEL-NEXT:    s_clause 0x1
-; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[4:5]
-; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[4:5] offset:16
+; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[0:1]
+; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[0:1] offset:16
 ; GFX13-GISEL-NEXT:    s_endpgm
   %cvt = tail call <6 x i32> @llvm.amdgcn.cvt.scalef32.pk32.fp6.f32(<32 x float> %src, float 100.0)
   store <6 x i32> %cvt, ptr addrspace(1) %out, align 8
@@ -523,8 +523,8 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_fp6_f32_v_inreg_src(ptr addrsp
 ; GFX13-GISEL-NEXT:    s_clause 0x3
 ; GFX13-GISEL-NEXT:    s_load_b512 s[36:51], s[4:5], 0xa4 nv
 ; GFX13-GISEL-NEXT:    s_load_b512 s[52:67], s[4:5], 0xe4 nv
-; GFX13-GISEL-NEXT:    s_load_b32 s0, s[4:5], 0x124 nv
-; GFX13-GISEL-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24 nv
+; GFX13-GISEL-NEXT:    s_load_b32 s2, s[4:5], 0x124 nv
+; GFX13-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX13-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v0, s36 :: v_dual_mov_b32 v1, s37
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v2, s38 :: v_dual_mov_b32 v3, s39
@@ -543,23 +543,23 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_fp6_f32_v_inreg_src(ptr addrsp
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v28, s64 :: v_dual_mov_b32 v29, s65
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v30, s66 :: v_dual_mov_b32 v31, s67
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_fp6_f32 v[0:5], v[0:31], s0
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_fp6_f32 v[0:5], v[0:31], s2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s1, v1
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s0
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s6
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX13-GISEL-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX13-GISEL-NEXT:    s_clause 0x1
-; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[4:5]
-; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[4:5] offset:16
+; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[0:1]
+; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[0:1] offset:16
 ; GFX13-GISEL-NEXT:    s_endpgm
   %cvt = tail call <6 x i32> @llvm.amdgcn.cvt.scalef32.pk32.fp6.f32(<32 x float> %src, float %scale)
   store <6 x i32> %cvt, ptr addrspace(1) %out, align 8
@@ -603,7 +603,7 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_fp6_f32_s_inreg_src(ptr addrsp
 ; GFX13-GISEL-NEXT:    s_clause 0x2
 ; GFX13-GISEL-NEXT:    s_load_b512 s[36:51], s[4:5], 0xa4 nv
 ; GFX13-GISEL-NEXT:    s_load_b512 s[52:67], s[4:5], 0xe4 nv
-; GFX13-GISEL-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24 nv
+; GFX13-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX13-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v0, s36 :: v_dual_mov_b32 v1, s37
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v2, s38 :: v_dual_mov_b32 v3, s39
@@ -623,22 +623,22 @@ define amdgpu_kernel void @test_cvt_scalef32_pk32_fp6_f32_s_inreg_src(ptr addrsp
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v30, s66 :: v_dual_mov_b32 v31, s67
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX13-GISEL-NEXT:    v_cvt_scalef32_pk32_fp6_f32 v[0:5], v[0:31], 0x42c80000
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s1, v1
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
+; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
 ; GFX13-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s0
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v4, s6
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v0, s2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v2, s4
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s5 :: v_dual_mov_b32 v4, s6
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX13-GISEL-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX13-GISEL-NEXT:    s_clause 0x1
-; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[4:5]
-; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[4:5] offset:16
+; GFX13-GISEL-NEXT:    global_store_b128 v6, v[0:3], s[0:1]
+; GFX13-GISEL-NEXT:    global_store_b64 v6, v[4:5], s[0:1] offset:16
 ; GFX13-GISEL-NEXT:    s_endpgm
   %cvt = tail call <6 x i32> @llvm.amdgcn.cvt.scalef32.pk32.fp6.f32(<32 x float> %src, float 100.0)
   store <6 x i32> %cvt, ptr addrspace(1) %out, align 8
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.dual_intersect_ray.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.dual_intersect_ray.ll
index b843d8394f228..bb7601d580ae1 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.dual_intersect_ray.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.dual_intersect_ray.ll
@@ -112,26 +112,17 @@ define amdgpu_ps <10 x float> @image_bvh_dual_intersect_ray_ssssss(i64 inreg %no
 ;
 ; GFX12-GISEL-LABEL: image_bvh_dual_intersect_ray_ssssss:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_mov_b32 s16, s3
-; GFX12-GISEL-NEXT:    s_mov_b32 s17, s4
-; GFX12-GISEL-NEXT:    s_mov_b32 s18, s5
-; GFX12-GISEL-NEXT:    s_mov_b32 s20, s6
-; GFX12-GISEL-NEXT:    s_mov_b32 s21, s7
-; GFX12-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX12-GISEL-NEXT:    s_mov_b32 s8, s9
-; GFX12-GISEL-NEXT:    s_mov_b32 s9, s10
-; GFX12-GISEL-NEXT:    s_mov_b32 s3, 0
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v7, s3 :: v_dual_mov_b32 v6, s2
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v14, s16 :: v_dual_mov_b32 v15, s17
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v16, s18 :: v_dual_mov_b32 v17, s20
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v18, s21 :: v_dual_mov_b32 v19, s22
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
-; GFX12-GISEL-NEXT:    s_mov_b32 s4, s11
-; GFX12-GISEL-NEXT:    s_mov_b32 s5, s12
-; GFX12-GISEL-NEXT:    s_mov_b32 s6, s13
-; GFX12-GISEL-NEXT:    s_mov_b32 s7, s14
-; GFX12-GISEL-NEXT:    image_bvh_dual_intersect_ray v[4:13], [v[4:5], v[6:7], v[14:16], v[17:19], v[8:9]], s[4:7]
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s0 :: v_dual_mov_b32 v5, s1
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s2 :: v_dual_mov_b32 v7, 0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v14, s3 :: v_dual_mov_b32 v15, s4
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v16, s5 :: v_dual_mov_b32 v17, s6
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v18, s7 :: v_dual_mov_b32 v19, s8
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v8, s9 :: v_dual_mov_b32 v9, s10
+; GFX12-GISEL-NEXT:    s_mov_b32 s16, s11
+; GFX12-GISEL-NEXT:    s_mov_b32 s17, s12
+; GFX12-GISEL-NEXT:    s_mov_b32 s18, s13
+; GFX12-GISEL-NEXT:    s_mov_b32 s19, s14
+; GFX12-GISEL-NEXT:    image_bvh_dual_intersect_ray v[4:13], [v[4:5], v[6:7], v[14:16], v[17:19], v[8:9]], s[16:19]
 ; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-GISEL-NEXT:    global_store_b96 v[0:1], v[14:16], off
 ; GFX12-GISEL-NEXT:    global_store_b96 v[2:3], v[17:19], off
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.global.load.async.to.lds.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.global.load.async.to.lds.ll
index 76692c55e8762..5679dc7a5f6c1 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.global.load.async.to.lds.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.global.load.async.to.lds.ll
@@ -310,8 +310,8 @@ define amdgpu_ps void @global_load_async_to_lds_b64_saddr_no_scale_offset(ptr ad
 ;
 ; GFX13-GISEL-LABEL: global_load_async_to_lds_b64_saddr_no_scale_offset:
 ; GFX13-GISEL:       ; %bb.0: ; %entry
-; GFX13-GISEL-NEXT:    v_dual_ashrrev_i32 v2, 31, v1 :: v_dual_mov_b32 v4, s1
-; GFX13-GISEL-NEXT:    v_mov_b32_e32 v3, s0
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v3, s0 :: v_dual_ashrrev_i32 v2, 31, v1
+; GFX13-GISEL-NEXT:    v_mov_b32_e32 v4, s1
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX13-GISEL-NEXT:    v_lshlrev_b64_e32 v[1:2], 2, v[1:2]
 ; GFX13-GISEL-NEXT:    v_add_co_u32 v1, vcc_lo, v3, v1
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
index 173739fa8f35f..3f931eaad33e3 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
@@ -129,23 +129,17 @@ define amdgpu_ps <4 x float> @image_bvh_intersect_ray_ssssss(i32 inreg %node_ptr
 ;
 ; GFX11-GISEL-LABEL: image_bvh_intersect_ray_ssssss:
 ; GFX11-GISEL:       ; %bb.0:
-; GFX11-GISEL-NEXT:    s_mov_b32 s16, s2
-; GFX11-GISEL-NEXT:    s_mov_b32 s17, s3
-; GFX11-GISEL-NEXT:    s_mov_b32 s18, s4
-; GFX11-GISEL-NEXT:    s_mov_b32 s20, s5
-; GFX11-GISEL-NEXT:    s_mov_b32 s21, s6
-; GFX11-GISEL-NEXT:    s_mov_b32 s22, s7
 ; GFX11-GISEL-NEXT:    v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s1
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s16 :: v_dual_mov_b32 v1, s17
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, s18 :: v_dual_mov_b32 v3, s20
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v4, s21 :: v_dual_mov_b32 v5, s22
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s5
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v5, s7
 ; GFX11-GISEL-NEXT:    v_dual_mov_b32 v6, s8 :: v_dual_mov_b32 v7, s9
 ; GFX11-GISEL-NEXT:    v_mov_b32_e32 v8, s10
-; GFX11-GISEL-NEXT:    s_mov_b32 s4, s11
-; GFX11-GISEL-NEXT:    s_mov_b32 s5, s12
-; GFX11-GISEL-NEXT:    s_mov_b32 s6, s13
-; GFX11-GISEL-NEXT:    s_mov_b32 s7, s14
-; GFX11-GISEL-NEXT:    image_bvh_intersect_ray v[0:3], [v9, v10, v[0:2], v[3:5], v[6:8]], s[4:7]
+; GFX11-GISEL-NEXT:    s_mov_b32 s16, s11
+; GFX11-GISEL-NEXT:    s_mov_b32 s17, s12
+; GFX11-GISEL-NEXT:    s_mov_b32 s18, s13
+; GFX11-GISEL-NEXT:    s_mov_b32 s19, s14
+; GFX11-GISEL-NEXT:    image_bvh_intersect_ray v[0:3], [v9, v10, v[0:2], v[3:5], v[6:8]], s[16:19]
 ; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
 ; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v1
@@ -174,23 +168,17 @@ define amdgpu_ps <4 x float> @image_bvh_intersect_ray_ssssss(i32 inreg %node_ptr
 ;
 ; GFX12-GISEL-LABEL: image_bvh_intersect_ray_ssssss:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_mov_b32 s16, s2
-; GFX12-GISEL-NEXT:    s_mov_b32 s17, s3
-; GFX12-GISEL-NEXT:    s_mov_b32 s18, s4
-; GFX12-GISEL-NEXT:    s_mov_b32 s20, s5
-; GFX12-GISEL-NEXT:    s_mov_b32 s21, s6
-; GFX12-GISEL-NEXT:    s_mov_b32 s22, s7
 ; GFX12-GISEL-NEXT:    v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s1
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s16 :: v_dual_mov_b32 v1, s17
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s18 :: v_dual_mov_b32 v3, s20
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s21 :: v_dual_mov_b32 v5, s22
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s5
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v5, s7
 ; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s8 :: v_dual_mov_b32 v7, s9
 ; GFX12-GISEL-NEXT:    v_mov_b32_e32 v8, s10
-; GFX12-GISEL-NEXT:    s_mov_b32 s4, s11
-; GFX12-GISEL-NEXT:    s_mov_b32 s5, s12
-; GFX12-GISEL-NEXT:    s_mov_b32 s6, s13
-; GFX12-GISEL-NEXT:    s_mov_b32 s7, s14
-; GFX12-GISEL-NEXT:    image_bvh_intersect_ray v[0:3], [v9, v10, v[0:2], v[3:5], v[6:8]], s[4:7]
+; GFX12-GISEL-NEXT:    s_mov_b32 s16, s11
+; GFX12-GISEL-NEXT:    s_mov_b32 s17, s12
+; GFX12-GISEL-NEXT:    s_mov_b32 s18, s13
+; GFX12-GISEL-NEXT:    s_mov_b32 s19, s14
+; GFX12-GISEL-NEXT:    image_bvh_intersect_ray v[0:3], [v9, v10, v[0:2], v[3:5], v[6:8]], s[16:19]
 ; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v1
@@ -234,22 +222,22 @@ define amdgpu_ps <4 x float> @image_bvh_intersect_ray_a16(i32 inreg %node_ptr, f
 ; GFX1013-GISEL-NEXT:    s_mov_b32 s16, s9
 ; GFX1013-GISEL-NEXT:    v_alignbit_b32 v0, s8, s7, 16
 ; GFX1013-GISEL-NEXT:    s_lshr_b32 s9, s5, 16
-; GFX1013-GISEL-NEXT:    s_and_b32 s5, s5, 0xffff
+; GFX1013-GISEL-NEXT:    s_and_b32 s7, s7, 0xffff
 ; GFX1013-GISEL-NEXT:    s_lshl_b32 s8, s9, 16
-; GFX1013-GISEL-NEXT:    s_and_b32 s9, s7, 0xffff
-; GFX1013-GISEL-NEXT:    v_readfirstlane_b32 s7, v0
+; GFX1013-GISEL-NEXT:    s_and_b32 s5, s5, 0xffff
+; GFX1013-GISEL-NEXT:    v_readfirstlane_b32 s9, v0
 ; GFX1013-GISEL-NEXT:    s_and_b32 s6, s6, 0xffff
-; GFX1013-GISEL-NEXT:    s_lshl_b32 s9, s9, 16
+; GFX1013-GISEL-NEXT:    s_lshl_b32 s7, s7, 16
 ; GFX1013-GISEL-NEXT:    s_or_b32 s5, s5, s8
-; GFX1013-GISEL-NEXT:    s_or_b32 s6, s6, s9
+; GFX1013-GISEL-NEXT:    s_or_b32 s6, s6, s7
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v4, s4
+; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v7, s9
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v6, s6
-; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v7, s7
 ; GFX1013-GISEL-NEXT:    s_mov_b32 s17, s10
 ; GFX1013-GISEL-NEXT:    s_mov_b32 s18, s11
 ; GFX1013-GISEL-NEXT:    s_mov_b32 s19, s12
@@ -436,21 +424,12 @@ define amdgpu_ps <4 x float> @image_bvh64_intersect_ray_ssssss(i64 inreg %node_p
 ;
 ; GFX11-GISEL-LABEL: image_bvh64_intersect_ray_ssssss:
 ; GFX11-GISEL:       ; %bb.0:
-; GFX11-GISEL-NEXT:    s_mov_b32 s16, s3
-; GFX11-GISEL-NEXT:    s_mov_b32 s17, s4
-; GFX11-GISEL-NEXT:    s_mov_b32 s18, s5
-; GFX11-GISEL-NEXT:    s_mov_b32 s4, s6
-; GFX11-GISEL-NEXT:    s_mov_b32 s5, s7
-; GFX11-GISEL-NEXT:    s_mov_b32 s6, s8
-; GFX11-GISEL-NEXT:    s_mov_b32 s8, s9
-; GFX11-GISEL-NEXT:    s_mov_b32 s9, s10
-; GFX11-GISEL-NEXT:    s_mov_b32 s10, s11
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v10, s1 :: v_dual_mov_b32 v9, s0
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v11, s2 :: v_dual_mov_b32 v0, s16
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, s17 :: v_dual_mov_b32 v2, s18
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v3, s4 :: v_dual_mov_b32 v4, s5
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v5, s6 :: v_dual_mov_b32 v6, s8
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v7, s9 :: v_dual_mov_b32 v8, s10
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s1
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v11, s2 :: v_dual_mov_b32 v0, s3
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, s4 :: v_dual_mov_b32 v2, s5
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v3, s6 :: v_dual_mov_b32 v4, s7
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v5, s8 :: v_dual_mov_b32 v6, s9
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v7, s10 :: v_dual_mov_b32 v8, s11
 ; GFX11-GISEL-NEXT:    image_bvh64_intersect_ray v[0:3], [v[9:10], v11, v[0:2], v[3:5], v[6:8]], s[12:15]
 ; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
@@ -476,21 +455,12 @@ define amdgpu_ps <4 x float> @image_bvh64_intersect_ray_ssssss(i64 inreg %node_p
 ;
 ; GFX12-GISEL-LABEL: image_bvh64_intersect_ray_ssssss:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_mov_b32 s16, s3
-; GFX12-GISEL-NEXT:    s_mov_b32 s17, s4
-; GFX12-GISEL-NEXT:    s_mov_b32 s18, s5
-; GFX12-GISEL-NEXT:    s_mov_b32 s4, s6
-; GFX12-GISEL-NEXT:    s_mov_b32 s5, s7
-; GFX12-GISEL-NEXT:    s_mov_b32 s6, s8
-; GFX12-GISEL-NEXT:    s_mov_b32 s8, s9
-; GFX12-GISEL-NEXT:    s_mov_b32 s9, s10
-; GFX12-GISEL-NEXT:    s_mov_b32 s10, s11
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v10, s1 :: v_dual_mov_b32 v9, s0
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v11, s2 :: v_dual_mov_b32 v0, s16
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, s17 :: v_dual_mov_b32 v2, s18
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v3, s4 :: v_dual_mov_b32 v4, s5
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v5, s6 :: v_dual_mov_b32 v6, s8
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v7, s9 :: v_dual_mov_b32 v8, s10
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s1
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v11, s2 :: v_dual_mov_b32 v0, s3
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, s4 :: v_dual_mov_b32 v2, s5
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v3, s6 :: v_dual_mov_b32 v4, s7
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v5, s8 :: v_dual_mov_b32 v6, s9
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v7, s10 :: v_dual_mov_b32 v8, s11
 ; GFX12-GISEL-NEXT:    image_bvh64_intersect_ray v[0:3], [v[9:10], v11, v[0:2], v[3:5], v[6:8]], s[12:15]
 ; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
@@ -536,23 +506,23 @@ define amdgpu_ps <4 x float> @image_bvh64_intersect_ray_a16(i64 inreg %node_ptr,
 ; GFX1013-GISEL-NEXT:    s_mov_b32 s16, s10
 ; GFX1013-GISEL-NEXT:    v_alignbit_b32 v0, s9, s8, 16
 ; GFX1013-GISEL-NEXT:    s_lshr_b32 s10, s6, 16
-; GFX1013-GISEL-NEXT:    s_and_b32 s6, s6, 0xffff
+; GFX1013-GISEL-NEXT:    s_and_b32 s8, s8, 0xffff
 ; GFX1013-GISEL-NEXT:    s_lshl_b32 s9, s10, 16
-; GFX1013-GISEL-NEXT:    s_and_b32 s10, s8, 0xffff
-; GFX1013-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
+; GFX1013-GISEL-NEXT:    s_and_b32 s6, s6, 0xffff
+; GFX1013-GISEL-NEXT:    v_readfirstlane_b32 s10, v0
 ; GFX1013-GISEL-NEXT:    s_and_b32 s7, s7, 0xffff
-; GFX1013-GISEL-NEXT:    s_lshl_b32 s10, s10, 16
+; GFX1013-GISEL-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX1013-GISEL-NEXT:    s_or_b32 s6, s6, s9
-; GFX1013-GISEL-NEXT:    s_or_b32 s7, s7, s10
+; GFX1013-GISEL-NEXT:    s_or_b32 s7, s7, s8
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v5, s5
+; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v8, s10
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v6, s6
 ; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v7, s7
-; GFX1013-GISEL-NEXT:    v_mov_b32_e32 v8, s8
 ; GFX1013-GISEL-NEXT:    s_mov_b32 s17, s11
 ; GFX1013-GISEL-NEXT:    s_mov_b32 s18, s12
 ; GFX1013-GISEL-NEXT:    s_mov_b32 s19, s13
@@ -799,33 +769,25 @@ define amdgpu_kernel void @image_bvh_intersect_ray_nsa_reassign(ptr %p_node_ptr,
 ; GFX11-GISEL:       ; %bb.0: ; %main_body
 ; GFX11-GISEL-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
 ; GFX11-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX11-GISEL-NEXT:    s_mov_b32 s10, 0x40a00000
-; GFX11-GISEL-NEXT:    s_mov_b32 s9, 4.0
-; GFX11-GISEL-NEXT:    s_mov_b32 s8, 0x40400000
-; GFX11-GISEL-NEXT:    s_mov_b32 s14, 0x41000000
-; GFX11-GISEL-NEXT:    s_mov_b32 s13, 0x40e00000
-; GFX11-GISEL-NEXT:    s_mov_b32 s12, 0x40c00000
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v8, s14
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v6, s12 :: v_dual_mov_b32 v5, s10
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v7, s13
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v8, 0x41000000
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v6, 0x40c00000
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v5, 0x40a00000
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v7, 0x40e00000
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_lshlrev_b32 v4, 2, v0
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v2, s2
-; GFX11-GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX11-GISEL-NEXT:    s_mov_b32 s1, 1.0
-; GFX11-GISEL-NEXT:    s_mov_b32 s0, 0
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v0, s0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, s2
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-GISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
 ; GFX11-GISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-GISEL-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-GISEL-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
 ; GFX11-GISEL-NEXT:    flat_load_b32 v9, v[0:1]
 ; GFX11-GISEL-NEXT:    flat_load_b32 v10, v[2:3]
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s8
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v4, s9
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, 2.0 :: v_dual_mov_b32 v3, 0x40400000
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 1.0 :: v_dual_mov_b32 v0, 0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v4, 4.0
 ; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    image_bvh_intersect_ray v[0:3], [v9, v10, v[0:2], v[3:5], v[6:8]], s[4:7]
 ; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0)
@@ -864,34 +826,26 @@ define amdgpu_kernel void @image_bvh_intersect_ray_nsa_reassign(ptr %p_node_ptr,
 ; GFX12-GISEL:       ; %bb.0: ; %main_body
 ; GFX12-GISEL-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
 ; GFX12-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX12-GISEL-NEXT:    s_mov_b32 s10, 0x40a00000
-; GFX12-GISEL-NEXT:    s_mov_b32 s9, 4.0
-; GFX12-GISEL-NEXT:    s_mov_b32 s8, 0x40400000
-; GFX12-GISEL-NEXT:    s_mov_b32 s14, 0x41000000
-; GFX12-GISEL-NEXT:    s_mov_b32 s13, 0x40e00000
-; GFX12-GISEL-NEXT:    s_mov_b32 s12, 0x40c00000
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v8, s14
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v6, s12 :: v_dual_mov_b32 v5, s10
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v7, s13
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v8, 0x41000000
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v6, 0x40c00000
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v5, 0x40a00000
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v7, 0x40e00000
 ; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_lshlrev_b32 v4, 2, v0
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v2, s2
-; GFX12-GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX12-GISEL-NEXT:    s_mov_b32 s1, 1.0
-; GFX12-GISEL-NEXT:    s_mov_b32 s0, 0
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v0, s0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s2
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX12-GISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
 ; GFX12-GISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3)
 ; GFX12-GISEL-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
 ; GFX12-GISEL-NEXT:    s_wait_alu depctr_va_vcc(0)
 ; GFX12-GISEL-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
 ; GFX12-GISEL-NEXT:    flat_load_b32 v9, v[0:1]
 ; GFX12-GISEL-NEXT:    flat_load_b32 v10, v[2:3]
-; GFX12-GISEL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s8
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v4, s9
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, 2.0 :: v_dual_mov_b32 v3, 0x40400000
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 1.0 :: v_dual_mov_b32 v0, 0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v4, 4.0
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    image_bvh_intersect_ray v[0:3], [v9, v10, v[0:2], v[3:5], v[6:8]], s[4:7]
 ; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
@@ -1231,30 +1185,22 @@ define amdgpu_kernel void @image_bvh64_intersect_ray_nsa_reassign(ptr %p_ray, <4
 ; GFX11-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24
 ; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[4:5], 0x34
 ; GFX11-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX11-GISEL-NEXT:    s_mov_b32 s5, 1.0
-; GFX11-GISEL-NEXT:    s_mov_b32 s4, 0
-; GFX11-GISEL-NEXT:    s_mov_b32 s10, 0x40a00000
-; GFX11-GISEL-NEXT:    s_mov_b32 s9, 4.0
-; GFX11-GISEL-NEXT:    s_mov_b32 s8, 0x40400000
-; GFX11-GISEL-NEXT:    s_mov_b32 s14, 0x41000000
-; GFX11-GISEL-NEXT:    s_mov_b32 s13, 0x40e00000
-; GFX11-GISEL-NEXT:    s_mov_b32 s12, 0x40c00000
 ; GFX11-GISEL-NEXT:    v_bfrev_b32_e32 v10, 4.0
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v4, s9 :: v_dual_mov_b32 v7, s13
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v6, s12
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v8, s14 :: v_dual_mov_b32 v9, 0xb36211c7
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v5, s10
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v3, s8
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v4, 4.0 :: v_dual_mov_b32 v7, 0x40e00000
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v8, 0x41000000
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v6, 0x40c00000
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v9, 0xb36211c7
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v5, 0x40a00000
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v3, 0x40400000
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_lshlrev_b32 v2, 2, v0
 ; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s6
-; GFX11-GISEL-NEXT:    s_mov_b32 s6, 2.0
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-GISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
 ; GFX11-GISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, s6
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 2.0
 ; GFX11-GISEL-NEXT:    flat_load_b32 v11, v[0:1]
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 1.0 :: v_dual_mov_b32 v0, 0
 ; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    image_bvh64_intersect_ray v[0:3], [v[9:10], v11, v[0:2], v[3:5], v[6:8]], s[0:3]
 ; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0)
@@ -1293,31 +1239,22 @@ define amdgpu_kernel void @image_bvh64_intersect_ray_nsa_reassign(ptr %p_ray, <4
 ; GFX12-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24
 ; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[4:5], 0x34
 ; GFX12-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX12-GISEL-NEXT:    s_mov_b32 s5, 1.0
-; GFX12-GISEL-NEXT:    s_mov_b32 s4, 0
-; GFX12-GISEL-NEXT:    s_mov_b32 s10, 0x40a00000
-; GFX12-GISEL-NEXT:    s_mov_b32 s9, 4.0
-; GFX12-GISEL-NEXT:    s_mov_b32 s8, 0x40400000
-; GFX12-GISEL-NEXT:    s_mov_b32 s14, 0x41000000
-; GFX12-GISEL-NEXT:    s_mov_b32 s13, 0x40e00000
-; GFX12-GISEL-NEXT:    s_mov_b32 s12, 0x40c00000
 ; GFX12-GISEL-NEXT:    v_bfrev_b32_e32 v10, 4.0
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, s9 :: v_dual_mov_b32 v7, s13
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v6, s12
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v8, s14 :: v_dual_mov_b32 v9, 0xb36211c7
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v5, s10
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v3, s8
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v4, 4.0 :: v_dual_mov_b32 v7, 0x40e00000
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v8, 0x41000000
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v6, 0x40c00000
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v9, 0xb36211c7
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v5, 0x40a00000
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v3, 0x40400000
 ; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_lshlrev_b32 v2, 2, v0
 ; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s6
-; GFX12-GISEL-NEXT:    s_mov_b32 s6, 2.0
 ; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX12-GISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v2
 ; GFX12-GISEL-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
-; GFX12-GISEL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s6
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 2.0
 ; GFX12-GISEL-NEXT:    flat_load_b32 v11, v[0:1]
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 1.0 :: v_dual_mov_b32 v0, 0
 ; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-GISEL-NEXT:    image_bvh64_intersect_ray v[0:3], [v[9:10], v11, v[0:2], v[3:5], v[6:8]], s[0:3]
 ; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll
index 4671993a13f0d..5abfd34619de6 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll
@@ -80,8 +80,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4bf16_1k(ptr addrspace(1) %arg) #
 ; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    s_mov_b64 s[36:37], 1
 ; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[32:33], s[36:37], s[36:37] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b32 s36, 2
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[34:35], s[36:37], s[36:37] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v34, 2
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v35, s37
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x0
 ; GFX90A-GISEL-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x40
@@ -179,8 +179,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4bf16_1k(ptr addrspace(1) %arg) #
 ; GFX942-GISEL-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    s_mov_b64 s[36:37], 1
 ; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[32:33], s[36:37]
-; GFX942-GISEL-NEXT:    s_mov_b32 s36, 2
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[36:37]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v34, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v35, s37
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x0
 ; GFX942-GISEL-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x40
@@ -278,8 +278,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4bf16_1k(ptr addrspace(1) %arg) #
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
 ; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[36:37], 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[32:33], s[36:37], s[36:37] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s36, 2
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[34:35], s[36:37], s[36:37] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v34, 2
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v35, s37
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x0
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x40
@@ -377,8 +377,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4bf16_1k(ptr addrspace(1) %arg) #
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[36:37], 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[32:33], s[36:37]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s36, 2
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[36:37]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v34, 2
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v35, s37
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x0
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x40
@@ -455,8 +455,8 @@ define amdgpu_kernel void @test_mfma_f32_16x16x4bf16_1k(ptr addrspace(1) %arg) #
 ; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    s_mov_b64 s[18:19], 1
 ; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[16:17], s[18:19], s[18:19] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b32 s18, 2
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[18:19], s[18:19], s[18:19] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v18, 2
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v19, s19
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -510,8 +510,8 @@ define amdgpu_kernel void @test_mfma_f32_16x16x4bf16_1k(ptr addrspace(1) %arg) #
 ; GFX942-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    s_mov_b64 s[18:19], 1
 ; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[18:19]
-; GFX942-GISEL-NEXT:    s_mov_b32 s18, 2
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[18:19]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v18, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v19, s19
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -565,8 +565,8 @@ define amdgpu_kernel void @test_mfma_f32_16x16x4bf16_1k(ptr addrspace(1) %arg) #
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
 ; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[18:19], 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[16:17], s[18:19], s[18:19] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s18, 2
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[18:19], s[18:19], s[18:19] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v18, 2
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v19, s19
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -620,8 +620,8 @@ define amdgpu_kernel void @test_mfma_f32_16x16x4bf16_1k(ptr addrspace(1) %arg) #
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[18:19], 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[18:19]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s18, 2
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[18:19]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v18, 2
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v19, s19
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -675,8 +675,8 @@ define amdgpu_kernel void @test_mfma_f32_4x4x4bf16_1k(ptr addrspace(1) %arg) #0
 ; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    s_mov_b64 s[4:5], 1
 ; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b32 s4, 2
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[4:5], s[4:5] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v6, 2
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -712,8 +712,8 @@ define amdgpu_kernel void @test_mfma_f32_4x4x4bf16_1k(ptr addrspace(1) %arg) #0
 ; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    s_mov_b64 s[4:5], 1
 ; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-GISEL-NEXT:    s_mov_b32 s4, 2
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[4:5]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v6, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -749,8 +749,8 @@ define amdgpu_kernel void @test_mfma_f32_4x4x4bf16_1k(ptr addrspace(1) %arg) #0
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s4, 2
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[4:5], s[4:5] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, 2
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -786,8 +786,8 @@ define amdgpu_kernel void @test_mfma_f32_4x4x4bf16_1k(ptr addrspace(1) %arg) #0
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s4, 2
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[4:5]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, 2
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -842,8 +842,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8bf16_1k(ptr addrspace(1) %arg) #
 ; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    s_mov_b64 s[18:19], 1
 ; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[16:17], s[18:19], s[18:19] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b32 s18, 2
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[18:19], s[18:19], s[18:19] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v18, 2
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v19, s19
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -898,8 +898,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8bf16_1k(ptr addrspace(1) %arg) #
 ; GFX942-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    s_mov_b64 s[18:19], 1
 ; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[18:19]
-; GFX942-GISEL-NEXT:    s_mov_b32 s18, 2
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[18:19]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v18, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v19, s19
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -954,8 +954,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8bf16_1k(ptr addrspace(1) %arg) #
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
 ; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[18:19], 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[16:17], s[18:19], s[18:19] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s18, 2
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[18:19], s[18:19], s[18:19] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v18, 2
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v19, s19
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -1010,8 +1010,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8bf16_1k(ptr addrspace(1) %arg) #
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[18:19], 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[18:19]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s18, 2
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[18:19]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v18, 2
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v19, s19
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -1065,8 +1065,8 @@ define amdgpu_kernel void @test_mfma_f32_16x16x16bf16_1k(ptr addrspace(1) %arg)
 ; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    s_mov_b64 s[4:5], 1
 ; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b32 s4, 2
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[4:5], s[4:5] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v6, 2
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -1102,8 +1102,8 @@ define amdgpu_kernel void @test_mfma_f32_16x16x16bf16_1k(ptr addrspace(1) %arg)
 ; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    s_mov_b64 s[4:5], 1
 ; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-GISEL-NEXT:    s_mov_b32 s4, 2
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[4:5]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v6, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -1139,8 +1139,8 @@ define amdgpu_kernel void @test_mfma_f32_16x16x16bf16_1k(ptr addrspace(1) %arg)
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s4, 2
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[4:5], s[4:5] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, 2
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -1176,8 +1176,8 @@ define amdgpu_kernel void @test_mfma_f32_16x16x16bf16_1k(ptr addrspace(1) %arg)
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s4, 2
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[4:5]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, 2
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v7, s5
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
@@ -1826,27 +1826,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_1(ptr addrspace(1)
 ;
 ; GFX90A-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
-; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[0:1], 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3ff00000
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-GISEL-NEXT:    s_nop 15
 ; GFX90A-GISEL-NEXT:    s_nop 0
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
@@ -1868,27 +1865,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_1(ptr addrspace(1)
 ;
 ; GFX942-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
-; GFX942-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-GISEL-NEXT:    s_mov_b64 s[0:1], 1.0
-; GFX942-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], 1.0
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-GISEL-NEXT:    s_nop 15
 ; GFX942-GISEL-NEXT:    s_nop 0
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
@@ -1910,27 +1903,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_1(ptr addrspace(1)
 ;
 ; GFX90A-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
 ; GFX90A-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[0:1], 1.0
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3ff00000
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 0
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-VGPR-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
@@ -1952,27 +1942,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_1(ptr addrspace(1)
 ;
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[0:1], 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], 1.0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double 1.0), i32 0, i32 0, i32 0)
@@ -2001,27 +1987,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_neg1(ptr addrspace
 ;
 ; GFX90A-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
-; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[0:1], -1.0
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, 0xbff00000
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-GISEL-NEXT:    s_nop 15
 ; GFX90A-GISEL-NEXT:    s_nop 0
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
@@ -2043,27 +2026,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_neg1(ptr addrspace
 ;
 ; GFX942-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
-; GFX942-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-GISEL-NEXT:    s_mov_b64 s[0:1], -1.0
-; GFX942-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], -1.0
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-GISEL-NEXT:    s_nop 15
 ; GFX942-GISEL-NEXT:    s_nop 0
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
@@ -2085,27 +2064,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_neg1(ptr addrspace
 ;
 ; GFX90A-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
 ; GFX90A-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[0:1], -1.0
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, 0xbff00000
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 0
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-VGPR-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
@@ -2127,27 +2103,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_neg1(ptr addrspace
 ;
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[0:1], -1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], -1.0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double -1.0), i32 0, i32 0, i32 0)
@@ -2176,27 +2148,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64(ptr addrspa
 ;
 ; GFX90A-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
-; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[0:1], 64
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 64
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-GISEL-NEXT:    s_nop 15
 ; GFX90A-GISEL-NEXT:    s_nop 0
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
@@ -2218,27 +2187,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64(ptr addrspa
 ;
 ; GFX942-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
-; GFX942-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-GISEL-NEXT:    s_mov_b64 s[0:1], 64
-; GFX942-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], 64
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-GISEL-NEXT:    s_nop 15
 ; GFX942-GISEL-NEXT:    s_nop 0
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
@@ -2260,27 +2225,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64(ptr addrspa
 ;
 ; GFX90A-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
 ; GFX90A-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[0:1], 64
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 64
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 0
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-VGPR-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
@@ -2302,27 +2264,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64(ptr addrspa
 ;
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[0:1], 64
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], 64
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double bitcast (i64 64 to double)), i32 0, i32 0, i32 0)
@@ -2362,28 +2320,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bit
 ;
 ; GFX90A-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
-; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-GISEL-NEXT:    s_mov_b32 s0, 0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, 64
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, 64
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-GISEL-NEXT:    s_nop 15
 ; GFX90A-GISEL-NEXT:    s_nop 0
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
@@ -2416,28 +2370,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bit
 ;
 ; GFX942-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
-; GFX942-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 0
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, 64
-; GFX942-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 64
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-GISEL-NEXT:    s_nop 15
 ; GFX942-GISEL-NEXT:    s_nop 0
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
@@ -2470,28 +2420,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bit
 ;
 ; GFX90A-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
 ; GFX90A-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s0, 0
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s1, 64
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, 64
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 0
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-VGPR-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
@@ -2524,28 +2470,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bit
 ;
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s0, 0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, 64
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, 64
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double bitcast (i64 274877906944 to double)), i32 0, i32 0, i32 0)
@@ -2582,28 +2524,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and
 ;
 ; GFX90A-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
-; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-GISEL-NEXT:    s_mov_b32 s0, 64
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, 64
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], 64, 64
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-GISEL-NEXT:    s_nop 15
 ; GFX90A-GISEL-NEXT:    s_nop 0
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
@@ -2633,28 +2570,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and
 ;
 ; GFX942-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
-; GFX942-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 64
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, 64
-; GFX942-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-GISEL-NEXT:    v_pk_mov_b32 v[0:1], 64, 64
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-GISEL-NEXT:    s_nop 15
 ; GFX942-GISEL-NEXT:    s_nop 0
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
@@ -2684,28 +2616,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and
 ;
 ; GFX90A-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
 ; GFX90A-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s0, 64
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s1, 64
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], 64, 64
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 0
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-VGPR-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
@@ -2735,28 +2662,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and
 ;
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s0, 64
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, 64
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], 64, 64
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double bitcast (i64 274877907008 to double)), i32 0, i32 0, i32 0)
@@ -2793,28 +2715,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_
 ;
 ; GFX90A-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
-; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-GISEL-NEXT:    s_nop 15
 ; GFX90A-GISEL-NEXT:    s_nop 0
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
@@ -2844,28 +2762,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_
 ;
 ; GFX942-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
-; GFX942-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-GISEL-NEXT:    s_nop 15
 ; GFX942-GISEL-NEXT:    s_nop 0
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
@@ -2895,28 +2809,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_
 ;
 ; GFX90A-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
 ; GFX90A-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
 ; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 0
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-VGPR-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
@@ -2946,28 +2856,24 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_
 ;
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double bitcast (<2 x float> splat (float 1.0) to double)), i32 0, i32 0, i32 0)
@@ -3007,26 +2913,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_imm(ptr addrspace(1) %arg, d
 ;
 ; GFX90A-GISEL-LABEL: test_mfma_f64_16x16x4f64_imm:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
-; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[0:1], 0
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[6:7], 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], 0, 0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v6, 0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v7, 0x3ff00000
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-GISEL-NEXT:    s_nop 15
 ; GFX90A-GISEL-NEXT:    s_nop 0
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f64_16x16x4f64_imm:
@@ -3059,26 +2962,22 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_imm(ptr addrspace(1) %arg, d
 ;
 ; GFX942-GISEL-LABEL: test_mfma_f64_16x16x4f64_imm:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
-; GFX942-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-GISEL-NEXT:    s_mov_b64 s[0:1], 0
-; GFX942-GISEL-NEXT:    s_mov_b64 s[6:7], 1.0
-; GFX942-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], 0
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 1.0
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-GISEL-NEXT:    s_nop 15
 ; GFX942-GISEL-NEXT:    s_nop 0
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_imm:
@@ -3111,26 +3010,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_imm(ptr addrspace(1) %arg, d
 ;
 ; GFX90A-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_imm:
 ; GFX90A-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[0:1], 0
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], 1.0
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], 0, 0
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, 0
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v7, 0x3ff00000
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 0
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-VGPR-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_imm:
@@ -3163,26 +3059,22 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_imm(ptr addrspace(1) %arg, d
 ;
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_imm:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[0:1], 0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], 0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 1.0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> <double 0.0, double 0.0, double 0.0, double 1.0>, i32 0, i32 0, i32 0)
@@ -3221,27 +3113,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_lit(ptr addrspace(1) %
 ;
 ; GFX90A-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
-; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-GISEL-NEXT:    s_mov_b32 s0, 0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, 0x405ec000
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX90A-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, 0x405ec000
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX90A-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-GISEL-NEXT:    s_nop 15
 ; GFX90A-GISEL-NEXT:    s_nop 0
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
@@ -3274,27 +3162,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_lit(ptr addrspace(1) %
 ;
 ; GFX942-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
-; GFX942-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 0
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, 0x405ec000
-; GFX942-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 0x405ec000
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX942-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-GISEL-NEXT:    s_nop 15
 ; GFX942-GISEL-NEXT:    s_nop 0
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
@@ -3327,27 +3211,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_lit(ptr addrspace(1) %
 ;
 ; GFX90A-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
 ; GFX90A-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s0, 0
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b32 s1, 0x405ec000
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, 0x405ec000
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX90A-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX90A-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX90A-VGPR-GISEL-NEXT:    s_nop 0
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX90A-VGPR-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
@@ -3380,27 +3260,23 @@ define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_lit(ptr addrspace(1) %
 ;
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s0, 0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, 0x405ec000
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, 0x405ec000
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[12:13]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 15
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> <double 123.0, double 123.0, double 123.0, double 123.0>, i32 0, i32 0, i32 0)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll
index 086fc6f0ed4d7..413c2973a3931 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll
@@ -4950,31 +4950,23 @@ define amdgpu_kernel void @test_mfma_i32_16x16x4i8_splatimm_src2_64(ptr addrspac
 ; GFX90A-GISEL-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v16, 1
-; GFX90A-GISEL-NEXT:    s_mov_b32 s8, 64
-; GFX90A-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s23, s8
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 64
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v17, 2
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[8:9], s[8:9] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[12:13], s[12:13] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[14:15], s[14:15] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[16:17], s[16:17] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[18:19], s[18:19] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[12:13], s[20:21], s[20:21] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[14:15], s[22:23], s[22:23] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v14, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v15, v0
 ; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    s_nop 0
 ; GFX90A-GISEL-NEXT:    v_mfma_i32_16x16x4i8 v[0:15], v16, v17, v[0:15] cbsz:1 abid:2 blgp:3
@@ -5006,31 +4998,23 @@ define amdgpu_kernel void @test_mfma_i32_16x16x4i8_splatimm_src2_64(ptr addrspac
 ; GFX942-GISEL-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v16, 1
-; GFX942-GISEL-NEXT:    s_mov_b32 s8, 64
-; GFX942-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s23, s8
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 64
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v17, 2
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[16:17]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v14, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v15, v0
 ; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    s_nop 0
 ; GFX942-GISEL-NEXT:    v_mfma_i32_16x16x4_4b_i8 v[0:15], v16, v17, v[0:15] cbsz:1 abid:2 blgp:3
@@ -5062,31 +5046,23 @@ define amdgpu_kernel void @test_mfma_i32_16x16x4i8_splatimm_src2_64(ptr addrspac
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v16, 1
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s8, 64
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s23, s8
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 64
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v17, 2
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[16:17]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v14, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v15, v0
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_i32_16x16x4_4b_i8 v[0:15], v16, v17, v[0:15] cbsz:1 abid:2 blgp:3
@@ -5395,20 +5371,18 @@ define amdgpu_kernel void @test_mfma_i32_4x4x4i8_splat_imm_src2_1(ptr addrspace(
 ; GFX90A-GISEL-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 1
-; GFX90A-GISEL-NEXT:    s_mov_b32 s0, 1
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 1
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v5, 2
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_i32_4x4x4i8 v[0:3], v4, v5, v[0:3] cbsz:1 abid:2 blgp:3
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_nop 2
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
@@ -5427,20 +5401,18 @@ define amdgpu_kernel void @test_mfma_i32_4x4x4i8_splat_imm_src2_1(ptr addrspace(
 ; GFX942-GISEL-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 1
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 1
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 1
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v5, 2
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_i32_4x4x4_16b_i8 v[0:3], v4, v5, v[0:3] cbsz:1 abid:2 blgp:3
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_nop 2
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
@@ -5459,20 +5431,18 @@ define amdgpu_kernel void @test_mfma_i32_4x4x4i8_splat_imm_src2_1(ptr addrspace(
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 1
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s0, 1
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v5, 2
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_i32_4x4x4_16b_i8 v[0:3], v4, v5, v[0:3] cbsz:1 abid:2 blgp:3
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 2
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x i32>, ptr addrspace(1) %arg
@@ -5575,20 +5545,18 @@ define amdgpu_kernel void @test_mfma_i32_4x4x4i8_splat_k_src2_1(ptr addrspace(1)
 ; GFX90A-GISEL-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
 ; GFX90A-GISEL:       ; %bb.0:
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 1
-; GFX90A-GISEL-NEXT:    s_movk_i32 s0, 0x41
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 0x41
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v5, 2
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_i32_4x4x4i8 v[0:3], v4, v5, v[0:3] cbsz:1 abid:2 blgp:3
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_nop 2
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
@@ -5611,20 +5579,18 @@ define amdgpu_kernel void @test_mfma_i32_4x4x4i8_splat_k_src2_1(ptr addrspace(1)
 ; GFX942-GISEL-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
 ; GFX942-GISEL:       ; %bb.0:
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 1
-; GFX942-GISEL-NEXT:    s_movk_i32 s0, 0x41
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0x41
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v5, 2
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_i32_4x4x4_16b_i8 v[0:3], v4, v5, v[0:3] cbsz:1 abid:2 blgp:3
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_nop 2
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
@@ -5647,20 +5613,18 @@ define amdgpu_kernel void @test_mfma_i32_4x4x4i8_splat_k_src2_1(ptr addrspace(1)
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
 ; GFX942-VGPR-GISEL:       ; %bb.0:
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 1
-; GFX942-VGPR-GISEL-NEXT:    s_movk_i32 s0, 0x41
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 0x41
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v5, 2
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_i32_4x4x4_16b_i8 v[0:3], v4, v5, v[0:3] cbsz:1 abid:2 blgp:3
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 2
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
   %in.1 = load <4 x i32>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x i32> @llvm.amdgcn.mfma.i32.4x4x4i8(i32 1, i32 2, <4 x i32> splat (i32 65), i32 1, i32 2, i32 3)
@@ -7038,20 +7002,18 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_imm_splat(ptr addrspace(1) %ar
 ; GFX90A-GISEL-LABEL: test_mfma_f32_4x4x1f32_imm_splat:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v5, 2.0
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f32_4x4x1f32 v[0:3], v4, v5, v[0:3]
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_nop 2
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f32_4x4x1f32_imm_splat:
@@ -7070,20 +7032,18 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_imm_splat(ptr addrspace(1) %ar
 ; GFX942-GISEL-LABEL: test_mfma_f32_4x4x1f32_imm_splat:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v5, 2.0
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v5, v[0:3]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_nop 1
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f32_4x4x1f32_imm_splat:
@@ -7102,20 +7062,18 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_imm_splat(ptr addrspace(1) %ar
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f32_4x4x1f32_imm_splat:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v5, 2.0
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v5, v[0:3]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.4x4x1f32(float 1.0, float 2.0, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, i32 0, i32 0, i32 0)
@@ -7309,31 +7267,23 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm_splat(ptr addrspace(1) %
 ; GFX90A-GISEL-LABEL: test_mfma_f32_16x16x1f32_imm_splat:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v16, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s8, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s23, s8
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v17, 2.0
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[8:9], s[8:9] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[12:13], s[12:13] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[14:15], s[14:15] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[16:17], s[16:17] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[18:19], s[18:19] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[12:13], s[20:21], s[20:21] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[14:15], s[22:23], s[22:23] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v14, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v15, v0
 ; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    s_nop 0
 ; GFX90A-GISEL-NEXT:    v_mfma_f32_16x16x1f32 v[0:15], v16, v17, v[0:15]
@@ -7365,31 +7315,23 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm_splat(ptr addrspace(1) %
 ; GFX942-GISEL-LABEL: test_mfma_f32_16x16x1f32_imm_splat:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v16, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s8, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s23, s8
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v17, 2.0
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[16:17]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v14, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v15, v0
 ; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    s_nop 0
 ; GFX942-GISEL-NEXT:    v_mfma_f32_16x16x1_4b_f32 v[0:15], v16, v17, v[0:15]
@@ -7421,31 +7363,23 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm_splat(ptr addrspace(1) %
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f32_16x16x1f32_imm_splat:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v16, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s8, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s23, s8
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v17, 2.0
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[16:17]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v14, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v15, v0
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f32_16x16x1_4b_f32 v[0:15], v16, v17, v[0:15]
@@ -7523,12 +7457,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16_imm_splat(ptr addrspace(1) %
 ;
 ; GFX908-GISEL-NOLIT-SRCC-LABEL: test_mfma_f32_32x32x8f16_imm_splat:
 ; GFX908-GISEL-NOLIT-SRCC:       ; %bb.0: ; %bb
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    s_mov_b32 s1, s0
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, s0
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, s1
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    s_mov_b32 s2, 0x40004000
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    s_mov_b32 s3, s2
+; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, 0x3c003c00
+; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, v0
+; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_mov_b32_e32 v2, 0x40004000
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a0, 1.0
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a1, 1.0
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a2, 1.0
@@ -7545,9 +7476,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16_imm_splat(ptr addrspace(1) %
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a13, 1.0
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a14, 1.0
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a15, 1.0
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_mov_b32_e32 v2, s2
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_mov_b32_e32 v3, s3
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_mov_b32_e32 v3, v2
+; GFX908-GISEL-NOLIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_mov_b32_e32 v16, 0
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], a[0:15]
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    s_nop 15
@@ -7569,10 +7499,10 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16_imm_splat(ptr addrspace(1) %
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v14, a14
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v15, a15
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[0:3], s[6:7]
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[4:7], s[6:7] offset:16
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[8:11], s[6:7] offset:32
-; GFX908-GISEL-NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[12:15], s[6:7] offset:48
+; GFX908-GISEL-NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; GFX908-GISEL-NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; GFX908-GISEL-NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; GFX908-GISEL-NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
 ; GFX908-GISEL-NOLIT-SRCC-NEXT:    s_endpgm
 ;
 ; LIT-SRCC-LABEL: test_mfma_f32_32x32x8f16_imm_splat:
@@ -7613,15 +7543,11 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16_imm_splat(ptr addrspace(1) %
 ;
 ; GFX908-GISEL-LIT-SRCC-LABEL: test_mfma_f32_32x32x8f16_imm_splat:
 ; GFX908-GISEL-LIT-SRCC:       ; %bb.0: ; %bb
-; GFX908-GISEL-LIT-SRCC-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GFX908-GISEL-LIT-SRCC-NEXT:    s_mov_b32 s1, s0
-; GFX908-GISEL-LIT-SRCC-NEXT:    v_mov_b32_e32 v0, s0
-; GFX908-GISEL-LIT-SRCC-NEXT:    v_mov_b32_e32 v1, s1
-; GFX908-GISEL-LIT-SRCC-NEXT:    s_mov_b32 s0, 0x40004000
-; GFX908-GISEL-LIT-SRCC-NEXT:    s_mov_b32 s1, s0
-; GFX908-GISEL-LIT-SRCC-NEXT:    v_mov_b32_e32 v3, s1
-; GFX908-GISEL-LIT-SRCC-NEXT:    v_mov_b32_e32 v2, s0
-; GFX908-GISEL-LIT-SRCC-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x24
+; GFX908-GISEL-LIT-SRCC-NEXT:    v_mov_b32_e32 v0, 0x3c003c00
+; GFX908-GISEL-LIT-SRCC-NEXT:    v_mov_b32_e32 v1, v0
+; GFX908-GISEL-LIT-SRCC-NEXT:    v_mov_b32_e32 v2, 0x40004000
+; GFX908-GISEL-LIT-SRCC-NEXT:    v_mov_b32_e32 v3, v2
+; GFX908-GISEL-LIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX908-GISEL-LIT-SRCC-NEXT:    v_mov_b32_e32 v16, 0
 ; GFX908-GISEL-LIT-SRCC-NEXT:    v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], 1.0
 ; GFX908-GISEL-LIT-SRCC-NEXT:    s_nop 15
@@ -7643,10 +7569,10 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16_imm_splat(ptr addrspace(1) %
 ; GFX908-GISEL-LIT-SRCC-NEXT:    v_accvgpr_read_b32 v14, a14
 ; GFX908-GISEL-LIT-SRCC-NEXT:    v_accvgpr_read_b32 v15, a15
 ; GFX908-GISEL-LIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX908-GISEL-LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[0:3], s[2:3]
-; GFX908-GISEL-LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[4:7], s[2:3] offset:16
-; GFX908-GISEL-LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[8:11], s[2:3] offset:32
-; GFX908-GISEL-LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[12:15], s[2:3] offset:48
+; GFX908-GISEL-LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; GFX908-GISEL-LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; GFX908-GISEL-LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; GFX908-GISEL-LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
 ; GFX908-GISEL-LIT-SRCC-NEXT:    s_endpgm
 ;
 ; GFX90A-LABEL: test_mfma_f32_32x32x8f16_imm_splat:
@@ -7670,36 +7596,26 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16_imm_splat(ptr addrspace(1) %
 ;
 ; GFX90A-GISEL-LABEL: test_mfma_f32_32x32x8f16_imm_splat:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
-; GFX90A-GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[16:17], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_mov_b32 s2, 0x40004000
-; GFX90A-GISEL-NEXT:    s_mov_b32 s8, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s3, s2
-; GFX90A-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s23, s8
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[18:19], s[2:3], s[2:3] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[8:9], s[8:9] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[12:13], s[12:13] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[14:15], s[14:15] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[16:17], s[16:17] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[18:19], s[18:19] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[12:13], s[20:21], s[20:21] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[14:15], s[22:23], s[22:23] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v16, 0x3c003c00
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v17, v16
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v18, 0x40004000
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v19, v18
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v14, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v15, v0
 ; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    s_nop 0
 ; GFX90A-GISEL-NEXT:    v_mfma_f32_32x32x8f16 v[0:15], v[16:17], v[18:19], v[0:15]
@@ -7733,36 +7649,26 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16_imm_splat(ptr addrspace(1) %
 ;
 ; GFX942-GISEL-LABEL: test_mfma_f32_32x32x8f16_imm_splat:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[0:1]
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, 0x40004000
-; GFX942-GISEL-NEXT:    s_mov_b32 s8, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, s2
-; GFX942-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s23, s8
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[16:17]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v16, 0x3c003c00
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v17, v16
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v18, 0x40004000
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v19, v18
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v14, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v15, v0
 ; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    s_nop 0
 ; GFX942-GISEL-NEXT:    v_mfma_f32_32x32x8_f16 v[0:15], v[16:17], v[18:19], v[0:15]
@@ -7795,36 +7701,26 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16_imm_splat(ptr addrspace(1) %
 ;
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f32_32x32x8f16_imm_splat:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s2, 0x40004000
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s8, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s3, s2
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s22, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s23, s8
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[2:3]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[16:17]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v16, 0x3c003c00
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v17, v16
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v18, 0x40004000
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v19, v18
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v14, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v15, v0
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f32_32x32x8_f16 v[0:15], v[16:17], v[18:19], v[0:15]
@@ -8346,20 +8242,18 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_imm(ptr addrspace(1) %arg) #0
 ; GFX90A-GISEL-LABEL: test_mfma_f32_4x4x1f32_imm:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, 2.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v5, 2.0
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f32_4x4x1f32 v[0:3], v4, v5, v[0:3]
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_nop 2
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f32_4x4x1f32_imm:
@@ -8380,20 +8274,18 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_imm(ptr addrspace(1) %arg) #0
 ; GFX942-GISEL-LABEL: test_mfma_f32_4x4x1f32_imm:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, 2.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v5, 2.0
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v5, v[0:3]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_nop 1
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f32_4x4x1f32_imm:
@@ -8414,20 +8306,18 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_imm(ptr addrspace(1) %arg) #0
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f32_4x4x1f32_imm:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, 2.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v5, 2.0
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v5, v[0:3]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.4x4x1f32(float 1.0, float 2.0, <4 x float> <float 1.0, float 2.0, float 1.0, float 1.0>, i32 0, i32 0, i32 0)
@@ -8623,31 +8513,23 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm(ptr addrspace(1) %arg) #
 ; GFX90A-GISEL-LABEL: test_mfma_f32_16x16x1f32_imm:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v16, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s8, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s23, 2.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX90A-GISEL-NEXT:    s_mov_b32 s22, s8
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v17, 2.0
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[8:9], s[8:9] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[10:11], s[10:11] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[12:13], s[12:13] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[14:15], s[14:15] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[16:17], s[16:17] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[18:19], s[18:19] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[12:13], s[20:21], s[20:21] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[14:15], s[22:23], s[22:23] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v15, 2.0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v14, v0
 ; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    s_nop 0
 ; GFX90A-GISEL-NEXT:    v_mfma_f32_16x16x1f32 v[0:15], v16, v17, v[0:15]
@@ -8692,31 +8574,23 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm(ptr addrspace(1) %arg) #
 ; GFX942-GISEL-LABEL: test_mfma_f32_16x16x1f32_imm:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v16, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s8, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s23, 2.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX942-GISEL-NEXT:    s_mov_b32 s22, s8
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v17, 2.0
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[16:17]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v15, 2.0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v14, v0
 ; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    s_nop 0
 ; GFX942-GISEL-NEXT:    v_mfma_f32_16x16x1_4b_f32 v[0:15], v16, v17, v[0:15]
@@ -8761,31 +8635,23 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm(ptr addrspace(1) %arg) #
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f32_16x16x1f32_imm:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v16, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s8, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s23, 2.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s9, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s10, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s11, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s12, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s13, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s14, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s15, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s16, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s17, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s18, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s19, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s20, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s21, s8
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s22, s8
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v17, 2.0
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[16:17]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v15, 2.0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v9, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v10, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v11, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v12, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v13, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v14, v0
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f32_16x16x1_4b_f32 v[0:15], v16, v17, v[0:15]
@@ -9156,55 +9022,39 @@ define amdgpu_kernel void @test_mfma_f32_32x32x1f32_imm(ptr addrspace(1) %arg) #
 ; GFX90A-GISEL-LABEL: test_mfma_f32_32x32x1f32_imm:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v32, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s37, 0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s36, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s38, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s39, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s40, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s41, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s42, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s43, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s44, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s45, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s46, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s47, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s48, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s49, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s50, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s51, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s52, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s53, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s54, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s55, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s56, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s57, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s58, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s59, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s60, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s61, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s62, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s63, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s64, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s65, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s66, s37
-; GFX90A-GISEL-NEXT:    s_mov_b32 s67, s37
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v33, 2.0
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[36:37], s[36:37] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[38:39], s[38:39] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[4:5], s[40:41], s[40:41] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[6:7], s[42:43], s[42:43] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[8:9], s[44:45], s[44:45] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[10:11], s[46:47], s[46:47] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[12:13], s[48:49], s[48:49] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[14:15], s[50:51], s[50:51] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[16:17], s[52:53], s[52:53] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[18:19], s[54:55], s[54:55] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[20:21], s[56:57], s[56:57] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[22:23], s[58:59], s[58:59] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[24:25], s[60:61], s[60:61] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[26:27], s[62:63], s[62:63] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[28:29], s[64:65], s[64:65] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[30:31], s[66:67], s[66:67] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v5, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v6, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v7, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v8, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v9, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v10, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v11, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v12, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v13, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v14, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v15, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v16, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v17, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v18, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v19, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v20, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v21, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v22, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v23, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v24, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v25, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v26, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v27, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v28, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v29, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v30, v1
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v31, v1
 ; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    s_nop 0
 ; GFX90A-GISEL-NEXT:    v_mfma_f32_32x32x1f32 v[0:31], v32, v33, v[0:31]
@@ -9292,55 +9142,39 @@ define amdgpu_kernel void @test_mfma_f32_32x32x1f32_imm(ptr addrspace(1) %arg) #
 ; GFX942-GISEL-LABEL: test_mfma_f32_32x32x1f32_imm:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v32, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s37, 0
-; GFX942-GISEL-NEXT:    s_mov_b32 s36, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s38, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s39, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s40, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s41, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s42, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s43, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s44, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s45, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s46, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s47, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s48, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s49, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s50, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s51, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s52, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s53, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s54, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s55, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s56, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s57, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s58, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s59, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s60, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s61, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s62, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s63, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s64, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s65, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s66, s37
-; GFX942-GISEL-NEXT:    s_mov_b32 s67, s37
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v33, 2.0
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[36:37]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[38:39]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[40:41]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[42:43]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[44:45]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[46:47]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[48:49]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[50:51]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[52:53]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[54:55]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[56:57]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[58:59]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[60:61]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[26:27], s[62:63]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[28:29], s[64:65]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[30:31], s[66:67]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v5, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v6, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v7, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v8, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v9, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v10, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v11, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v12, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v13, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v14, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v15, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v16, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v17, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v18, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v19, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v20, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v21, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v22, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v23, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v24, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v25, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v26, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v27, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v28, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v29, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v30, v1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v31, v1
 ; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    s_nop 0
 ; GFX942-GISEL-NEXT:    v_mfma_f32_32x32x1_2b_f32 v[0:31], v32, v33, v[0:31]
@@ -9427,55 +9261,39 @@ define amdgpu_kernel void @test_mfma_f32_32x32x1f32_imm(ptr addrspace(1) %arg) #
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f32_32x32x1f32_imm:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v32, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s37, 0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s36, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s38, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s39, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s40, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s41, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s42, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s43, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s44, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s45, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s46, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s47, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s48, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s49, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s50, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s51, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s52, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s53, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s54, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s55, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s56, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s57, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s58, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s59, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s60, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s61, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s62, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s63, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s64, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s65, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s66, s37
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s67, s37
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v33, 2.0
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[36:37]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[38:39]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[40:41]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[42:43]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[44:45]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[46:47]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[48:49]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[50:51]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[52:53]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[54:55]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[56:57]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[58:59]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[60:61]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[26:27], s[62:63]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[28:29], s[64:65]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[30:31], s[66:67]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v5, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v7, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v8, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v9, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v10, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v11, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v12, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v13, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v14, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v15, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v16, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v17, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v18, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v19, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v20, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v21, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v22, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v23, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v24, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v25, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v26, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v27, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v28, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v29, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v30, v1
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v31, v1
 ; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 0
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f32_32x32x1_2b_f32 v[0:31], v32, v33, v[0:31]
@@ -9592,21 +9410,19 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_lit_splat(ptr addrspace(1) %ar
 ; GFX90A-GISEL-LABEL: test_mfma_f32_4x4x1f32_lit_splat:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s0, 0x42f60000
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX90A-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; GFX90A-GISEL-NEXT:    v_lshlrev_b32_e32 v5, 4, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 0x42f60000
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v6, 2.0
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f32_4x4x1f32 v[0:3], v4, v6, v[0:3]
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_nop 3
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v5, v[0:3], s[6:7]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v5, v[0:3], s[0:1]
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f32_4x4x1f32_lit_splat:
@@ -9630,21 +9446,19 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_lit_splat(ptr addrspace(1) %ar
 ; GFX942-GISEL-LABEL: test_mfma_f32_4x4x1f32_lit_splat:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 0x42f60000
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; GFX942-GISEL-NEXT:    v_lshlrev_b32_e32 v5, 4, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0x42f60000
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v6, 2.0
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v6, v[0:3]
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_nop 2
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v5, v[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v5, v[0:3], s[0:1]
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f32_4x4x1f32_lit_splat:
@@ -9668,21 +9482,19 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_lit_splat(ptr addrspace(1) %ar
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f32_4x4x1f32_lit_splat:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s0, 0x42f60000
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; GFX942-VGPR-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; GFX942-VGPR-GISEL-NEXT:    v_lshlrev_b32_e32 v5, 4, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 0x42f60000
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v6, 2.0
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v6, v[0:3]
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 2
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v5, v[0:3], s[6:7]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v5, v[0:3], s[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -9787,20 +9599,18 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_lit_splat_bad_code(ptr addrspa
 ; GFX90A-GISEL-LABEL: test_mfma_f32_4x4x1f32_lit_splat_bad_code:
 ; GFX90A-GISEL:       ; %bb.0: ; %bb
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s0, 0x42f60000
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v0, 0x42f60000
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v5, 2.0
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX90A-GISEL-NEXT:    s_nop 1
 ; GFX90A-GISEL-NEXT:    v_mfma_f32_4x4x1f32 v[0:3], v4, v5, v[0:3]
 ; GFX90A-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-GISEL-NEXT:    s_nop 2
-; GFX90A-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX90A-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX90A-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f32_4x4x1f32_lit_splat_bad_code:
@@ -9823,20 +9633,18 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_lit_splat_bad_code(ptr addrspa
 ; GFX942-GISEL-LABEL: test_mfma_f32_4x4x1f32_lit_splat_bad_code:
 ; GFX942-GISEL:       ; %bb.0: ; %bb
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 0x42f60000
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0x42f60000
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v5, 2.0
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-GISEL-NEXT:    s_nop 1
 ; GFX942-GISEL-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v5, v[0:3]
 ; GFX942-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-GISEL-NEXT:    s_nop 1
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-GISEL-NEXT:    s_endpgm
 ;
 ; GFX942-VGPR-LABEL: test_mfma_f32_4x4x1f32_lit_splat_bad_code:
@@ -9859,20 +9667,18 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_lit_splat_bad_code(ptr addrspa
 ; GFX942-VGPR-GISEL-LABEL: test_mfma_f32_4x4x1f32_lit_splat_bad_code:
 ; GFX942-VGPR-GISEL:       ; %bb.0: ; %bb
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 1.0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s0, 0x42f60000
-; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s1, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s2, s0
-; GFX942-VGPR-GISEL-NEXT:    s_mov_b32 s3, s0
+; GFX942-VGPR-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v0, 0x42f60000
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v5, 2.0
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX942-VGPR-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
 ; GFX942-VGPR-GISEL-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v5, v[0:3]
 ; GFX942-VGPR-GISEL-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX942-VGPR-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-VGPR-GISEL-NEXT:    s_nop 1
-; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-VGPR-GISEL-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-VGPR-GISEL-NEXT:    s_endpgm
 bb:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll
index 1d19e9225e56b..7384fb8494d23 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll
@@ -43,15 +43,14 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_integer_immediate(ptr addrspace
 ;
 ; GFX600-GISEL-LABEL: v_mqsad_u32_u8_inline_integer_immediate:
 ; GFX600-GISEL:       ; %bb.0:
-; GFX600-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x9
+; GFX600-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
 ; GFX600-GISEL-NEXT:    s_load_dword s4, s[4:5], 0xd
-; GFX600-GISEL-NEXT:    s_mov_b32 s3, 40
-; GFX600-GISEL-NEXT:    s_mov_b32 s2, 30
-; GFX600-GISEL-NEXT:    s_mov_b32 s1, 20
-; GFX600-GISEL-NEXT:    s_mov_b32 s0, 10
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v10, 40
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v9, 30
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v8, 20
 ; GFX600-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v0, s10
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v1, s11
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX600-GISEL-NEXT:    ;;#ASMSTART
 ; GFX600-GISEL-NEXT:    v_lsrlrev_b64 v[2:3], v[0:1], 1
 ; GFX600-GISEL-NEXT:    ;;#ASMEND
@@ -62,17 +61,14 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_integer_immediate(ptr addrspace
 ; GFX600-GISEL-NEXT:    v_mov_b32 v4, v2
 ; GFX600-GISEL-NEXT:    ;;#ASMEND
 ; GFX600-GISEL-NEXT:    v_mov_b32_e32 v6, v4
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v10, s3
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v9, s2
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v8, s1
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v7, s0
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v7, 10
 ; GFX600-GISEL-NEXT:    v_mqsad_u32_u8 v[2:5], v[0:1], v6, v[7:10]
-; GFX600-GISEL-NEXT:    s_mov_b32 s10, -1
-; GFX600-GISEL-NEXT:    s_mov_b32 s11, 0xf000
+; GFX600-GISEL-NEXT:    s_mov_b32 s2, -1
+; GFX600-GISEL-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX600-GISEL-NEXT:    ;;#ASMSTART
 ; GFX600-GISEL-NEXT:    ;; force constraint
 ; GFX600-GISEL-NEXT:    ;;#ASMEND
-; GFX600-GISEL-NEXT:    buffer_store_dwordx4 v[0:3], off, s[8:11], 0
+; GFX600-GISEL-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; GFX600-GISEL-NEXT:    s_endpgm
 ;
 ; GFX803-SDAG-LABEL: v_mqsad_u32_u8_inline_integer_immediate:
@@ -107,15 +103,14 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_integer_immediate(ptr addrspace
 ;
 ; GFX803-GISEL-LABEL: v_mqsad_u32_u8_inline_integer_immediate:
 ; GFX803-GISEL:       ; %bb.0:
-; GFX803-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
+; GFX803-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX803-GISEL-NEXT:    s_load_dword s4, s[4:5], 0x34
-; GFX803-GISEL-NEXT:    s_mov_b32 s3, 40
-; GFX803-GISEL-NEXT:    s_mov_b32 s2, 30
-; GFX803-GISEL-NEXT:    s_mov_b32 s1, 20
-; GFX803-GISEL-NEXT:    s_mov_b32 s0, 10
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v10, 40
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v9, 30
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v8, 20
 ; GFX803-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v0, s10
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v1, s11
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX803-GISEL-NEXT:    ;;#ASMSTART
 ; GFX803-GISEL-NEXT:    v_lsrlrev_b64 v[2:3], v[0:1], 1
 ; GFX803-GISEL-NEXT:    ;;#ASMEND
@@ -126,16 +121,13 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_integer_immediate(ptr addrspace
 ; GFX803-GISEL-NEXT:    v_mov_b32 v4, v2
 ; GFX803-GISEL-NEXT:    ;;#ASMEND
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v6, v4
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v10, s3
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v9, s2
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v8, s1
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v7, s0
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v7, 10
 ; GFX803-GISEL-NEXT:    v_mqsad_u32_u8 v[2:5], v[0:1], v6, v[7:10]
 ; GFX803-GISEL-NEXT:    ;;#ASMSTART
 ; GFX803-GISEL-NEXT:    ;; force constraint
 ; GFX803-GISEL-NEXT:    ;;#ASMEND
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v4, s8
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v5, s9
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v5, s1
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX803-GISEL-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX803-GISEL-NEXT:    s_endpgm
 ;
@@ -170,24 +162,19 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_integer_immediate(ptr addrspace
 ; GFX13-GISEL-NEXT:    s_clause 0x1
 ; GFX13-GISEL-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24 nv
 ; GFX13-GISEL-NEXT:    s_load_b32 s4, s[4:5], 0x34 nv
-; GFX13-GISEL-NEXT:    s_mov_b32 s7, 40
-; GFX13-GISEL-NEXT:    s_mov_b32 s6, 30
-; GFX13-GISEL-NEXT:    s_mov_b32 s5, 20
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v9, 30 :: v_dual_mov_b32 v8, 20
 ; GFX13-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
-; GFX13-GISEL-NEXT:    v_mov_b32_e32 v4, s4
-; GFX13-GISEL-NEXT:    s_mov_b32 s4, 10
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v10, s7 :: v_dual_mov_b32 v9, s6
-; GFX13-GISEL-NEXT:    v_mov_b32_e32 v8, s5
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v10, 40 :: v_dual_mov_b32 v0, s2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v4, s4
 ; GFX13-GISEL-NEXT:    ;;#ASMSTART
 ; GFX13-GISEL-NEXT:    v_lsrlrev_b64 v[2:3], v[0:1], 1
 ; GFX13-GISEL-NEXT:    ;;#ASMEND
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v7, s4 :: v_dual_mov_b32 v0, v2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v7, 10 :: v_dual_mov_b32 v0, v2
 ; GFX13-GISEL-NEXT:    ;;#ASMSTART
 ; GFX13-GISEL-NEXT:    v_mov_b32 v4, v4
 ; GFX13-GISEL-NEXT:    ;;#ASMEND
+; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, v3 :: v_dual_mov_b32 v6, v4
-; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX13-GISEL-NEXT:    v_mqsad_u32_u8 v[2:5], v[0:1], v6, v[7:10]
 ; GFX13-GISEL-NEXT:    v_mov_b32_e32 v6, 0
 ; GFX13-GISEL-NEXT:    ;;#ASMSTART
@@ -399,15 +386,14 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_fp_immediate(ptr addrspace(1) %
 ;
 ; GFX600-GISEL-LABEL: v_mqsad_u32_u8_inline_fp_immediate:
 ; GFX600-GISEL:       ; %bb.0:
-; GFX600-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x9
+; GFX600-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
 ; GFX600-GISEL-NEXT:    s_load_dword s4, s[4:5], 0xd
-; GFX600-GISEL-NEXT:    s_mov_b32 s1, 0
-; GFX600-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX600-GISEL-NEXT:    s_mov_b32 s2, s1
-; GFX600-GISEL-NEXT:    s_mov_b32 s3, s1
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v8, 0
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v7, 1.0
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v9, v8
 ; GFX600-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v0, s10
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v1, s11
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX600-GISEL-NEXT:    ;;#ASMSTART
 ; GFX600-GISEL-NEXT:    v_lsrlrev_b64 v[2:3], v[0:1], 1
 ; GFX600-GISEL-NEXT:    ;;#ASMEND
@@ -418,17 +404,14 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_fp_immediate(ptr addrspace(1) %
 ; GFX600-GISEL-NEXT:    v_mov_b32 v4, v2
 ; GFX600-GISEL-NEXT:    ;;#ASMEND
 ; GFX600-GISEL-NEXT:    v_mov_b32_e32 v6, v4
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v10, s3
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v9, s2
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v8, s1
-; GFX600-GISEL-NEXT:    v_mov_b32_e32 v7, s0
+; GFX600-GISEL-NEXT:    v_mov_b32_e32 v10, v8
 ; GFX600-GISEL-NEXT:    v_mqsad_u32_u8 v[2:5], v[0:1], v6, v[7:10]
-; GFX600-GISEL-NEXT:    s_mov_b32 s10, -1
-; GFX600-GISEL-NEXT:    s_mov_b32 s11, 0xf000
+; GFX600-GISEL-NEXT:    s_mov_b32 s2, -1
+; GFX600-GISEL-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX600-GISEL-NEXT:    ;;#ASMSTART
 ; GFX600-GISEL-NEXT:    ;; force constraint
 ; GFX600-GISEL-NEXT:    ;;#ASMEND
-; GFX600-GISEL-NEXT:    buffer_store_dwordx4 v[0:3], off, s[8:11], 0
+; GFX600-GISEL-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; GFX600-GISEL-NEXT:    s_endpgm
 ;
 ; GFX803-SDAG-LABEL: v_mqsad_u32_u8_inline_fp_immediate:
@@ -463,15 +446,14 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_fp_immediate(ptr addrspace(1) %
 ;
 ; GFX803-GISEL-LABEL: v_mqsad_u32_u8_inline_fp_immediate:
 ; GFX803-GISEL:       ; %bb.0:
-; GFX803-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
+; GFX803-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX803-GISEL-NEXT:    s_load_dword s4, s[4:5], 0x34
-; GFX803-GISEL-NEXT:    s_mov_b32 s1, 0
-; GFX803-GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GFX803-GISEL-NEXT:    s_mov_b32 s2, s1
-; GFX803-GISEL-NEXT:    s_mov_b32 s3, s1
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v8, 0
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v7, 1.0
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v9, v8
 ; GFX803-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v0, s10
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v1, s11
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX803-GISEL-NEXT:    ;;#ASMSTART
 ; GFX803-GISEL-NEXT:    v_lsrlrev_b64 v[2:3], v[0:1], 1
 ; GFX803-GISEL-NEXT:    ;;#ASMEND
@@ -482,16 +464,13 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_fp_immediate(ptr addrspace(1) %
 ; GFX803-GISEL-NEXT:    v_mov_b32 v4, v2
 ; GFX803-GISEL-NEXT:    ;;#ASMEND
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v6, v4
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v10, s3
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v9, s2
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v8, s1
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v7, s0
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v10, v8
 ; GFX803-GISEL-NEXT:    v_mqsad_u32_u8 v[2:5], v[0:1], v6, v[7:10]
 ; GFX803-GISEL-NEXT:    ;;#ASMSTART
 ; GFX803-GISEL-NEXT:    ;; force constraint
 ; GFX803-GISEL-NEXT:    ;;#ASMEND
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v4, s8
-; GFX803-GISEL-NEXT:    v_mov_b32_e32 v5, s9
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v5, s1
+; GFX803-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX803-GISEL-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX803-GISEL-NEXT:    s_endpgm
 ;
@@ -524,20 +503,16 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_fp_immediate(ptr addrspace(1) %
 ; GFX13-GISEL:       ; %bb.0:
 ; GFX13-GISEL-NEXT:    s_clause 0x1
 ; GFX13-GISEL-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24 nv
-; GFX13-GISEL-NEXT:    s_load_b32 s7, s[4:5], 0x34 nv
-; GFX13-GISEL-NEXT:    s_mov_b32 s5, 0
-; GFX13-GISEL-NEXT:    s_mov_b32 s4, 1.0
-; GFX13-GISEL-NEXT:    s_mov_b32 s6, s5
+; GFX13-GISEL-NEXT:    s_load_b32 s4, s[4:5], 0x34 nv
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v8, 0 :: v_dual_mov_b32 v7, 1.0
 ; GFX13-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
-; GFX13-GISEL-NEXT:    v_mov_b32_e32 v4, s7
-; GFX13-GISEL-NEXT:    s_mov_b32 s7, s5
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v9, s6 :: v_dual_mov_b32 v8, s5
-; GFX13-GISEL-NEXT:    v_mov_b32_e32 v10, s7
+; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v9, v8 :: v_dual_mov_b32 v0, s2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v1, s3 :: v_dual_mov_b32 v4, s4
 ; GFX13-GISEL-NEXT:    ;;#ASMSTART
 ; GFX13-GISEL-NEXT:    v_lsrlrev_b64 v[2:3], v[0:1], 1
 ; GFX13-GISEL-NEXT:    ;;#ASMEND
-; GFX13-GISEL-NEXT:    v_dual_mov_b32 v7, s4 :: v_dual_mov_b32 v0, v2
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v10, v8 :: v_dual_mov_b32 v0, v2
 ; GFX13-GISEL-NEXT:    ;;#ASMSTART
 ; GFX13-GISEL-NEXT:    v_mov_b32 v4, v4
 ; GFX13-GISEL-NEXT:    ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ptr.s.buffer.load.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ptr.s.buffer.load.ll
index b5a30c9511606..b73ffda3b88dc 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ptr.s.buffer.load.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ptr.s.buffer.load.ll
@@ -677,11 +677,11 @@ define amdgpu_kernel void @ptr_s_buffer_load_v8i32_divergent_offset_glc(ptr addr
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s2, s0, 16
-; GFX7-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-NEXT:    v_mov_b32_e32 v12, s3
-; GFX7-NEXT:    v_mov_b32_e32 v11, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v10, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v9, s0
+; GFX7-NEXT:    s_addc_u32 s0, s1, 0
+; GFX7-NEXT:    v_mov_b32_e32 v11, s2
+; GFX7-NEXT:    v_mov_b32_e32 v12, s0
 ; GFX7-NEXT:    s_waitcnt vmcnt(1)
 ; GFX7-NEXT:    flat_store_dwordx4 v[11:12], v[1:4]
 ; GFX7-NEXT:    s_waitcnt vmcnt(1)
@@ -700,11 +700,11 @@ define amdgpu_kernel void @ptr_s_buffer_load_v8i32_divergent_offset_glc(ptr addr
 ; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v12, s3
-; GFX8-NEXT:    v_mov_b32_e32 v11, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v10, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v9, s0
+; GFX8-NEXT:    s_addc_u32 s0, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v11, s2
+; GFX8-NEXT:    v_mov_b32_e32 v12, s0
 ; GFX8-NEXT:    s_waitcnt vmcnt(1)
 ; GFX8-NEXT:    flat_store_dwordx4 v[11:12], v[1:4]
 ; GFX8-NEXT:    s_waitcnt vmcnt(1)
@@ -872,9 +872,9 @@ define amdgpu_kernel void @ptr_s_buffer_load_v16i32_divergent_offset(ptr addrspa
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s2, s0, 48
 ; GFX7-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-NEXT:    v_mov_b32_e32 v20, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v19, s2
 ; GFX7-NEXT:    s_add_u32 s2, s0, 32
+; GFX7-NEXT:    v_mov_b32_e32 v20, s3
 ; GFX7-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v18, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v17, s0
@@ -885,8 +885,8 @@ define amdgpu_kernel void @ptr_s_buffer_load_v16i32_divergent_offset(ptr addrspa
 ; GFX7-NEXT:    flat_store_dwordx4 v[19:20], v[1:4]
 ; GFX7-NEXT:    s_nop 0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s3
-; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s0
+; GFX7-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-NEXT:    s_waitcnt vmcnt(3)
 ; GFX7-NEXT:    flat_store_dwordx4 v[0:1], v[5:8]
 ; GFX7-NEXT:    s_waitcnt vmcnt(3)
@@ -910,9 +910,9 @@ define amdgpu_kernel void @ptr_s_buffer_load_v16i32_divergent_offset(ptr addrspa
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_add_u32 s2, s0, 48
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v20, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v19, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 32
+; GFX8-NEXT:    v_mov_b32_e32 v20, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v18, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v17, s0
@@ -923,8 +923,8 @@ define amdgpu_kernel void @ptr_s_buffer_load_v16i32_divergent_offset(ptr addrspa
 ; GFX8-NEXT:    flat_store_dwordx4 v[19:20], v[1:4]
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    s_waitcnt vmcnt(3)
 ; GFX8-NEXT:    flat_store_dwordx4 v[0:1], v[5:8]
 ; GFX8-NEXT:    s_waitcnt vmcnt(3)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
index 5dee433e1143f..e1fde0c5246a8 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
@@ -367,13 +367,13 @@ define amdgpu_kernel void @test_readlane_vregs_i64(ptr addrspace(1) %out, ptr ad
 ; CHECK-GISEL-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s3, v2
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
 ; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
 ; CHECK-GISEL-NEXT:    s_nop 2
-; CHECK-GISEL-NEXT:    v_readlane_b32 s2, v0, s3
-; CHECK-GISEL-NEXT:    v_readlane_b32 s3, v1, s3
-; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s3, v0, s2
+; CHECK-GISEL-NEXT:    v_readlane_b32 s2, v1, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s3
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s2
 ; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; CHECK-GISEL-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -427,13 +427,13 @@ define amdgpu_kernel void @test_readlane_vregs_f64(ptr addrspace(1) %out, ptr ad
 ; CHECK-GISEL-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s3, v2
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
 ; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
 ; CHECK-GISEL-NEXT:    s_nop 2
-; CHECK-GISEL-NEXT:    v_readlane_b32 s2, v0, s3
-; CHECK-GISEL-NEXT:    v_readlane_b32 s3, v1, s3
-; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s3, v0, s2
+; CHECK-GISEL-NEXT:    v_readlane_b32 s2, v1, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s3
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s2
 ; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; CHECK-GISEL-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.add.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.add.ll
index 4d367c1eb3c06..bd99b78e4710c 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.add.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.add.ll
@@ -3778,8 +3778,8 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX8DAGISEL-NEXT:    v_readfirstlane_b32 s1, v2
 ; GFX8DAGISEL-NEXT:    s_mul_i32 s2, s3, s4
 ; GFX8DAGISEL-NEXT:    s_add_u32 s1, s1, s2
-; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8DAGISEL-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX8DAGISEL-NEXT:    s_endpgm
 ;
@@ -3806,15 +3806,15 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX9DAGISEL:       ; %bb.0: ; %entry
 ; GFX9DAGISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX9DAGISEL-NEXT:    s_mov_b64 s[4:5], exec
-; GFX9DAGISEL-NEXT:    s_bcnt1_i32_b64 s5, s[4:5]
+; GFX9DAGISEL-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9DAGISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9DAGISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9DAGISEL-NEXT:    s_mul_i32 s4, s2, s5
-; GFX9DAGISEL-NEXT:    s_mul_hi_u32 s2, s2, s5
-; GFX9DAGISEL-NEXT:    s_mul_i32 s3, s3, s5
-; GFX9DAGISEL-NEXT:    s_add_u32 s5, s2, s3
-; GFX9DAGISEL-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9DAGISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9DAGISEL-NEXT:    s_mul_i32 s5, s2, s4
+; GFX9DAGISEL-NEXT:    s_mul_hi_u32 s2, s2, s4
+; GFX9DAGISEL-NEXT:    s_mul_i32 s3, s3, s4
+; GFX9DAGISEL-NEXT:    s_add_u32 s2, s2, s3
+; GFX9DAGISEL-NEXT:    v_mov_b32_e32 v0, s5
+; GFX9DAGISEL-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX9DAGISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9DAGISEL-NEXT:    s_endpgm
 ;
@@ -3822,15 +3822,15 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX9GISEL:       ; %bb.0: ; %entry
 ; GFX9GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX9GISEL-NEXT:    s_mov_b64 s[4:5], exec
-; GFX9GISEL-NEXT:    s_bcnt1_i32_b64 s5, s[4:5]
+; GFX9GISEL-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9GISEL-NEXT:    s_mul_i32 s4, s2, s5
-; GFX9GISEL-NEXT:    s_mul_hi_u32 s2, s2, s5
-; GFX9GISEL-NEXT:    s_mul_i32 s3, s3, s5
-; GFX9GISEL-NEXT:    s_add_u32 s5, s2, s3
-; GFX9GISEL-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9GISEL-NEXT:    s_mul_i32 s5, s2, s4
+; GFX9GISEL-NEXT:    s_mul_hi_u32 s2, s2, s4
+; GFX9GISEL-NEXT:    s_mul_i32 s3, s3, s4
+; GFX9GISEL-NEXT:    s_add_u32 s2, s2, s3
+; GFX9GISEL-NEXT:    v_mov_b32_e32 v0, s5
+; GFX9GISEL-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX9GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9GISEL-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.sub.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.sub.ll
index ae2dcfac5a4d0..9ce193d14dafd 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.sub.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.sub.ll
@@ -3860,19 +3860,19 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX6DAGISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6DAGISEL-NEXT:    s_mov_b32 s4, s0
 ; GFX6DAGISEL-NEXT:    s_bcnt1_i32_b64 s0, s[8:9]
-; GFX6DAGISEL-NEXT:    s_mov_b32 s5, s1
-; GFX6DAGISEL-NEXT:    s_sub_i32 s1, 0, s0
-; GFX6DAGISEL-NEXT:    v_mov_b32_e32 v0, s1
-; GFX6DAGISEL-NEXT:    v_mul_hi_u32 v0, s2, v0
-; GFX6DAGISEL-NEXT:    s_ashr_i32 s0, s1, 31
-; GFX6DAGISEL-NEXT:    s_mul_i32 s8, s2, s0
-; GFX6DAGISEL-NEXT:    s_mul_i32 s0, s2, s1
-; GFX6DAGISEL-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX6DAGISEL-NEXT:    s_mul_i32 s1, s3, s1
-; GFX6DAGISEL-NEXT:    s_add_u32 s1, s2, s1
-; GFX6DAGISEL-NEXT:    s_add_u32 s1, s1, s8
+; GFX6DAGISEL-NEXT:    s_sub_i32 s0, 0, s0
 ; GFX6DAGISEL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX6DAGISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX6DAGISEL-NEXT:    v_mul_hi_u32 v0, s2, v0
+; GFX6DAGISEL-NEXT:    s_mov_b32 s5, s1
+; GFX6DAGISEL-NEXT:    s_ashr_i32 s1, s0, 31
+; GFX6DAGISEL-NEXT:    s_mul_i32 s1, s2, s1
+; GFX6DAGISEL-NEXT:    s_mul_i32 s2, s2, s0
+; GFX6DAGISEL-NEXT:    v_readfirstlane_b32 s8, v0
+; GFX6DAGISEL-NEXT:    s_mul_i32 s0, s3, s0
+; GFX6DAGISEL-NEXT:    s_add_u32 s0, s8, s0
+; GFX6DAGISEL-NEXT:    s_add_u32 s0, s0, s1
+; GFX6DAGISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX6DAGISEL-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX6DAGISEL-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
 ; GFX6DAGISEL-NEXT:    s_endpgm
 ;
@@ -3885,19 +3885,19 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX7DAGISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7DAGISEL-NEXT:    s_mov_b32 s4, s0
 ; GFX7DAGISEL-NEXT:    s_bcnt1_i32_b64 s0, s[8:9]
-; GFX7DAGISEL-NEXT:    s_mov_b32 s5, s1
-; GFX7DAGISEL-NEXT:    s_sub_i32 s1, 0, s0
-; GFX7DAGISEL-NEXT:    v_mov_b32_e32 v0, s1
-; GFX7DAGISEL-NEXT:    v_mul_hi_u32 v0, s2, v0
-; GFX7DAGISEL-NEXT:    s_ashr_i32 s0, s1, 31
-; GFX7DAGISEL-NEXT:    s_mul_i32 s8, s2, s0
-; GFX7DAGISEL-NEXT:    s_mul_i32 s0, s2, s1
-; GFX7DAGISEL-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX7DAGISEL-NEXT:    s_mul_i32 s1, s3, s1
-; GFX7DAGISEL-NEXT:    s_add_u32 s1, s2, s1
-; GFX7DAGISEL-NEXT:    s_add_u32 s1, s1, s8
+; GFX7DAGISEL-NEXT:    s_sub_i32 s0, 0, s0
 ; GFX7DAGISEL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX7DAGISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX7DAGISEL-NEXT:    v_mul_hi_u32 v0, s2, v0
+; GFX7DAGISEL-NEXT:    s_mov_b32 s5, s1
+; GFX7DAGISEL-NEXT:    s_ashr_i32 s1, s0, 31
+; GFX7DAGISEL-NEXT:    s_mul_i32 s1, s2, s1
+; GFX7DAGISEL-NEXT:    s_mul_i32 s2, s2, s0
+; GFX7DAGISEL-NEXT:    v_readfirstlane_b32 s8, v0
+; GFX7DAGISEL-NEXT:    s_mul_i32 s0, s3, s0
+; GFX7DAGISEL-NEXT:    s_add_u32 s0, s8, s0
+; GFX7DAGISEL-NEXT:    s_add_u32 s0, s0, s1
+; GFX7DAGISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX7DAGISEL-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX7DAGISEL-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
 ; GFX7DAGISEL-NEXT:    s_endpgm
 ;
@@ -3913,14 +3913,14 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX8DAGISEL-NEXT:    s_ashr_i32 s0, s4, 31
 ; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v1, s1
-; GFX8DAGISEL-NEXT:    s_mul_i32 s1, s2, s0
-; GFX8DAGISEL-NEXT:    s_mul_i32 s0, s2, s4
+; GFX8DAGISEL-NEXT:    s_mul_i32 s0, s2, s0
+; GFX8DAGISEL-NEXT:    s_mul_i32 s1, s2, s4
 ; GFX8DAGISEL-NEXT:    v_readfirstlane_b32 s2, v2
 ; GFX8DAGISEL-NEXT:    s_mul_i32 s3, s3, s4
 ; GFX8DAGISEL-NEXT:    s_add_u32 s2, s2, s3
-; GFX8DAGISEL-NEXT:    s_add_u32 s1, s2, s1
-; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8DAGISEL-NEXT:    s_add_u32 s0, s2, s0
+; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v2, s1
+; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v3, s0
 ; GFX8DAGISEL-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX8DAGISEL-NEXT:    s_endpgm
 ;
@@ -3952,18 +3952,18 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX9DAGISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX9DAGISEL-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX9DAGISEL-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
-; GFX9DAGISEL-NEXT:    s_sub_i32 s5, 0, s4
-; GFX9DAGISEL-NEXT:    s_ashr_i32 s4, s5, 31
+; GFX9DAGISEL-NEXT:    s_sub_i32 s4, 0, s4
+; GFX9DAGISEL-NEXT:    s_ashr_i32 s5, s4, 31
 ; GFX9DAGISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9DAGISEL-NEXT:    s_mul_i32 s5, s2, s5
 ; GFX9DAGISEL-NEXT:    s_mul_i32 s6, s2, s4
-; GFX9DAGISEL-NEXT:    s_mul_i32 s4, s2, s5
-; GFX9DAGISEL-NEXT:    s_mul_hi_u32 s2, s2, s5
-; GFX9DAGISEL-NEXT:    s_mul_i32 s3, s3, s5
+; GFX9DAGISEL-NEXT:    s_mul_hi_u32 s2, s2, s4
+; GFX9DAGISEL-NEXT:    s_mul_i32 s3, s3, s4
 ; GFX9DAGISEL-NEXT:    s_add_u32 s2, s2, s3
-; GFX9DAGISEL-NEXT:    s_add_u32 s5, s2, s6
+; GFX9DAGISEL-NEXT:    s_add_u32 s2, s2, s5
 ; GFX9DAGISEL-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9DAGISEL-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9DAGISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9DAGISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX9DAGISEL-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX9DAGISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9DAGISEL-NEXT:    s_endpgm
 ;
@@ -3972,17 +3972,17 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX9GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX9GISEL-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX9GISEL-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
-; GFX9GISEL-NEXT:    s_sub_i32 s5, 0, s4
-; GFX9GISEL-NEXT:    s_ashr_i32 s4, s5, 31
+; GFX9GISEL-NEXT:    s_sub_i32 s4, 0, s4
+; GFX9GISEL-NEXT:    s_ashr_i32 s5, s4, 31
 ; GFX9GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9GISEL-NEXT:    s_mul_i32 s5, s2, s5
 ; GFX9GISEL-NEXT:    s_mul_i32 s6, s2, s4
-; GFX9GISEL-NEXT:    s_mul_i32 s4, s2, s5
-; GFX9GISEL-NEXT:    s_mul_hi_u32 s2, s2, s5
-; GFX9GISEL-NEXT:    s_mul_i32 s3, s3, s5
+; GFX9GISEL-NEXT:    s_mul_hi_u32 s2, s2, s4
+; GFX9GISEL-NEXT:    s_mul_i32 s3, s3, s4
 ; GFX9GISEL-NEXT:    s_add_u32 s2, s2, s3
-; GFX9GISEL-NEXT:    s_add_u32 s5, s2, s6
-; GFX9GISEL-NEXT:    v_mov_b32_e32 v0, s4
-; GFX9GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9GISEL-NEXT:    s_add_u32 s2, s2, s5
+; GFX9GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX9GISEL-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX9GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX9GISEL-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.xor.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.xor.ll
index 7d6c296515733..74e6526af3164 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.xor.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.reduce.xor.ll
@@ -3577,12 +3577,12 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX6DAGISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6DAGISEL-NEXT:    s_mov_b32 s4, s0
 ; GFX6DAGISEL-NEXT:    s_bcnt1_i32_b64 s0, s[8:9]
+; GFX6DAGISEL-NEXT:    s_and_b32 s0, s0, 1
 ; GFX6DAGISEL-NEXT:    s_mov_b32 s5, s1
-; GFX6DAGISEL-NEXT:    s_and_b32 s1, s0, 1
-; GFX6DAGISEL-NEXT:    s_mul_i32 s0, s2, s1
-; GFX6DAGISEL-NEXT:    s_mul_i32 s1, s3, s1
-; GFX6DAGISEL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX6DAGISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX6DAGISEL-NEXT:    s_mul_i32 s1, s2, s0
+; GFX6DAGISEL-NEXT:    s_mul_i32 s0, s3, s0
+; GFX6DAGISEL-NEXT:    v_mov_b32_e32 v0, s1
+; GFX6DAGISEL-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX6DAGISEL-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
 ; GFX6DAGISEL-NEXT:    s_endpgm
 ;
@@ -3595,12 +3595,12 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX7DAGISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7DAGISEL-NEXT:    s_mov_b32 s4, s0
 ; GFX7DAGISEL-NEXT:    s_bcnt1_i32_b64 s0, s[8:9]
+; GFX7DAGISEL-NEXT:    s_and_b32 s0, s0, 1
 ; GFX7DAGISEL-NEXT:    s_mov_b32 s5, s1
-; GFX7DAGISEL-NEXT:    s_and_b32 s1, s0, 1
-; GFX7DAGISEL-NEXT:    s_mul_i32 s0, s2, s1
-; GFX7DAGISEL-NEXT:    s_mul_i32 s1, s3, s1
-; GFX7DAGISEL-NEXT:    v_mov_b32_e32 v0, s0
-; GFX7DAGISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX7DAGISEL-NEXT:    s_mul_i32 s1, s2, s0
+; GFX7DAGISEL-NEXT:    s_mul_i32 s0, s3, s0
+; GFX7DAGISEL-NEXT:    v_mov_b32_e32 v0, s1
+; GFX7DAGISEL-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX7DAGISEL-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
 ; GFX7DAGISEL-NEXT:    s_endpgm
 ;
@@ -3615,8 +3615,8 @@ define amdgpu_kernel void @uniform_value_i64(ptr addrspace(1) %out, i64 %in) #0
 ; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8DAGISEL-NEXT:    s_mul_i32 s0, s2, s4
 ; GFX8DAGISEL-NEXT:    s_mul_i32 s1, s3, s4
-; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8DAGISEL-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8DAGISEL-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX8DAGISEL-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma.imm.gfx1250.w32.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma.imm.gfx1250.w32.ll
index cbbbf45ff7ed9..b9a372b8e3d57 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma.imm.gfx1250.w32.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma.imm.gfx1250.w32.ll
@@ -56,19 +56,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x4_f32_non_splat(<2 x float> %A, <2 x
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v6, 1.0 :: v_dual_mov_b32 v8, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v7, v6 :: v_dual_mov_b32 v9, v6
+; GISEL-NEXT:    v_dual_mov_b32 v10, v6 :: v_dual_mov_b32 v11, v6
+; GISEL-NEXT:    v_dual_mov_b32 v12, v6 :: v_dual_mov_b32 v13, v6
 ; GISEL-NEXT:    v_wmma_f32_16x16x4_f32 v[6:13], v[0:1], v[2:3], v[6:13]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[4:5], v[6:9], off
@@ -105,20 +97,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x4_f32_non_inlineable(<2 x float> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v6, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v7, v6 :: v_dual_mov_b32 v8, v6
+; GISEL-NEXT:    v_dual_mov_b32 v9, v6 :: v_dual_mov_b32 v10, v6
+; GISEL-NEXT:    v_dual_mov_b32 v11, v6 :: v_dual_mov_b32 v12, v6
+; GISEL-NEXT:    v_mov_b32_e32 v13, v6
 ; GISEL-NEXT:    v_wmma_f32_16x16x4_f32 v[6:13], v[0:1], v[2:3], v[6:13]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[4:5], v[6:9], off
@@ -184,19 +168,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x32_bf16_non_splat(<16 x bfloat> %A, <
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v18, 1.0 :: v_dual_mov_b32 v20, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v21, v18
+; GISEL-NEXT:    v_dual_mov_b32 v22, v18 :: v_dual_mov_b32 v23, v18
+; GISEL-NEXT:    v_dual_mov_b32 v24, v18 :: v_dual_mov_b32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x32_bf16 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -233,20 +209,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x32_bf16_non_inlineable(<16 x bfloat>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_dual_mov_b32 v21, v18 :: v_dual_mov_b32 v22, v18
+; GISEL-NEXT:    v_dual_mov_b32 v23, v18 :: v_dual_mov_b32 v24, v18
+; GISEL-NEXT:    v_mov_b32_e32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x32_bf16 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -275,13 +243,10 @@ define amdgpu_ps void @test_wmma_bf16_16x16x32_bf16(<16 x bfloat> %A, <16 x bflo
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3f803f80
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x3f803f80
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_bf16_16x16x32_bf16 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -311,13 +276,9 @@ define amdgpu_ps void @test_wmma_bf16_16x16x32_bf16_non_splat(<16 x bfloat> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3f803f80
-; GISEL-NEXT:    s_mov_b32 s1, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v18, 0x3f803f80 :: v_dual_mov_b32 v19, 1.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v20, v18 :: v_dual_mov_b32 v21, v18
 ; GISEL-NEXT:    v_wmma_bf16_16x16x32_bf16 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -348,13 +309,10 @@ define amdgpu_ps void @test_wmma_bf16_16x16x32_bf16_non_inlineable(<16 x bfloat>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3fc03fc0
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x3fc03fc0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_bf16_16x16x32_bf16 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -412,19 +370,11 @@ define amdgpu_ps void @test_wmma_bf16f32_16x16x32_bf16_non_splat(<16 x bfloat> %
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v18, 1.0 :: v_dual_mov_b32 v20, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v21, v18
+; GISEL-NEXT:    v_dual_mov_b32 v22, v18 :: v_dual_mov_b32 v23, v18
+; GISEL-NEXT:    v_dual_mov_b32 v24, v18 :: v_dual_mov_b32 v25, v18
 ; GISEL-NEXT:    v_wmma_bf16f32_16x16x32_bf16 v[26:29], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[26:29], off
 ; GISEL-NEXT:    s_endpgm
@@ -457,20 +407,12 @@ define amdgpu_ps void @test_wmma_bf16f32_16x16x32_bf16_non_inlinable(<16 x bfloa
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_dual_mov_b32 v21, v18 :: v_dual_mov_b32 v22, v18
+; GISEL-NEXT:    v_dual_mov_b32 v23, v18 :: v_dual_mov_b32 v24, v18
+; GISEL-NEXT:    v_mov_b32_e32 v25, v18
 ; GISEL-NEXT:    v_wmma_bf16f32_16x16x32_bf16 v[26:29], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[26:29], off
 ; GISEL-NEXT:    s_endpgm
@@ -534,19 +476,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x64_fp8_fp8_non_splat(<8 x i32> %A, <8
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v18, 1.0 :: v_dual_mov_b32 v20, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v21, v18
+; GISEL-NEXT:    v_dual_mov_b32 v22, v18 :: v_dual_mov_b32 v23, v18
+; GISEL-NEXT:    v_dual_mov_b32 v24, v18 :: v_dual_mov_b32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x64_fp8_fp8 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -583,20 +517,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x64_fp8_fp8_non_inlineable(<8 x i32> %
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_dual_mov_b32 v21, v18 :: v_dual_mov_b32 v22, v18
+; GISEL-NEXT:    v_dual_mov_b32 v23, v18 :: v_dual_mov_b32 v24, v18
+; GISEL-NEXT:    v_mov_b32_e32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x64_fp8_fp8 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -662,19 +588,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x64_fp8_bf8_non_splat(<8 x i32> %A, <8
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v18, 1.0 :: v_dual_mov_b32 v20, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v21, v18
+; GISEL-NEXT:    v_dual_mov_b32 v22, v18 :: v_dual_mov_b32 v23, v18
+; GISEL-NEXT:    v_dual_mov_b32 v24, v18 :: v_dual_mov_b32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x64_fp8_bf8 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -711,20 +629,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x64_fp8_bf8_non_inlineable(<8 x i32> %
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_dual_mov_b32 v21, v18 :: v_dual_mov_b32 v22, v18
+; GISEL-NEXT:    v_dual_mov_b32 v23, v18 :: v_dual_mov_b32 v24, v18
+; GISEL-NEXT:    v_mov_b32_e32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x64_fp8_bf8 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -790,19 +700,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x64_bf8_fp8_non_splat(<8 x i32> %A, <8
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v18, 1.0 :: v_dual_mov_b32 v20, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v21, v18
+; GISEL-NEXT:    v_dual_mov_b32 v22, v18 :: v_dual_mov_b32 v23, v18
+; GISEL-NEXT:    v_dual_mov_b32 v24, v18 :: v_dual_mov_b32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x64_bf8_fp8 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -839,20 +741,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x64_bf8_fp8_non_inlineable(<8 x i32> %
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_dual_mov_b32 v21, v18 :: v_dual_mov_b32 v22, v18
+; GISEL-NEXT:    v_dual_mov_b32 v23, v18 :: v_dual_mov_b32 v24, v18
+; GISEL-NEXT:    v_mov_b32_e32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x64_bf8_fp8 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -918,19 +812,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x64_bf8_bf8_non_splat(<8 x i32> %A, <8
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v18, 1.0 :: v_dual_mov_b32 v20, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v21, v18
+; GISEL-NEXT:    v_dual_mov_b32 v22, v18 :: v_dual_mov_b32 v23, v18
+; GISEL-NEXT:    v_dual_mov_b32 v24, v18 :: v_dual_mov_b32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x64_bf8_bf8 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -967,20 +853,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x64_bf8_bf8_non_inlineable(<8 x i32> %
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_dual_mov_b32 v21, v18 :: v_dual_mov_b32 v22, v18
+; GISEL-NEXT:    v_dual_mov_b32 v23, v18 :: v_dual_mov_b32 v24, v18
+; GISEL-NEXT:    v_mov_b32_e32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x64_bf8_bf8 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -1039,13 +917,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x64_fp8_fp8_non_splat(<8 x i32> %A, <8
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GISEL-NEXT:    s_mov_b32 s1, 0x3c004000
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x3c003c00
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, 0x3c004000 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_f16_16x16x64_fp8_fp8 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -1076,13 +951,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x64_fp8_fp8_non_inlineable(<8 x i32> %
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x42004200
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x42004200
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_f16_16x16x64_fp8_fp8 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -1139,13 +1011,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x64_fp8_bf8_non_splat(<8 x i32> %A, <8
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GISEL-NEXT:    s_mov_b32 s1, 0x3c004000
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x3c003c00
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, 0x3c004000 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_f16_16x16x64_fp8_bf8 v[18:21], v[0:7], v[8:15], v[18:21] neg_hi:[0,0,1]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -1176,13 +1045,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x64_fp8_bf8_non_inlineable(<8 x i32> %
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x42004200
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x42004200
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_f16_16x16x64_fp8_bf8 v[18:21], v[0:7], v[8:15], v[18:21] neg_hi:[0,0,1]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -1239,13 +1105,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x64_bf8_fp8_non_splat(<8 x i32> %A, <8
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GISEL-NEXT:    s_mov_b32 s1, 0x3c004000
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x3c003c00
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, 0x3c004000 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_f16_16x16x64_bf8_fp8 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -1276,13 +1139,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x64_bf8_fp8_non_inlineable(<8 x i32> %
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x42004200
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x42004200
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_f16_16x16x64_bf8_fp8 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -1339,13 +1199,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x64_bf8_bf8_non_splat(<8 x i32> %A, <8
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GISEL-NEXT:    s_mov_b32 s1, 0x3c004000
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x3c003c00
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, 0x3c004000 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_f16_16x16x64_bf8_bf8 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -1376,13 +1233,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x64_bf8_bf8_non_inlineable(<8 x i32> %
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x42004200
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x42004200
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_f16_16x16x64_bf8_bf8 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -1446,19 +1300,11 @@ define amdgpu_ps void @test_wmma_i32_16x16x64_iu8_non_splat(<8 x i32> %A, <8 x i
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1
-; GISEL-NEXT:    s_mov_b32 s2, 2
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v18, 1 :: v_dual_mov_b32 v20, 2
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v21, v18
+; GISEL-NEXT:    v_dual_mov_b32 v22, v18 :: v_dual_mov_b32 v23, v18
+; GISEL-NEXT:    v_dual_mov_b32 v24, v18 :: v_dual_mov_b32 v25, v18
 ; GISEL-NEXT:    v_wmma_i32_16x16x64_iu8 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -1495,20 +1341,12 @@ define amdgpu_ps void @test_wmma_i32_16x16x64_iu8_non_inlineable(<8 x i32> %A, <
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_movk_i32 s0, 0x80
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x80
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_dual_mov_b32 v21, v18 :: v_dual_mov_b32 v22, v18
+; GISEL-NEXT:    v_dual_mov_b32 v23, v18 :: v_dual_mov_b32 v24, v18
+; GISEL-NEXT:    v_mov_b32_e32 v25, v18
 ; GISEL-NEXT:    v_wmma_i32_16x16x64_iu8 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -1574,19 +1412,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x32_f16_non_splat(<16 x half> %A, <16
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v18, 1.0 :: v_dual_mov_b32 v20, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v21, v18
+; GISEL-NEXT:    v_dual_mov_b32 v22, v18 :: v_dual_mov_b32 v23, v18
+; GISEL-NEXT:    v_dual_mov_b32 v24, v18 :: v_dual_mov_b32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x32_f16 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -1623,20 +1453,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x32_f16_non_inlineable(<16 x half> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_dual_mov_b32 v21, v18 :: v_dual_mov_b32 v22, v18
+; GISEL-NEXT:    v_dual_mov_b32 v23, v18 :: v_dual_mov_b32 v24, v18
+; GISEL-NEXT:    v_mov_b32_e32 v25, v18
 ; GISEL-NEXT:    v_wmma_f32_16x16x32_f16 v[18:25], v[0:7], v[8:15], v[18:25]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
@@ -1695,13 +1517,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x32_f16_non_splat(<16 x half> %A, <16
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GISEL-NEXT:    s_mov_b32 s1, 0x3c004000
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x3c003c00
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, 0x3c004000 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_f16_16x16x32_f16 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -1732,13 +1551,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x32_f16_non_inlineable(<16 x half> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x42004200
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v18, 0x42004200
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v19, v18 :: v_dual_mov_b32 v20, v18
+; GISEL-NEXT:    v_mov_b32_e32 v21, v18
 ; GISEL-NEXT:    v_wmma_f16_16x16x32_f16 v[18:21], v[0:7], v[8:15], v[18:21]
 ; GISEL-NEXT:    global_store_b128 v[16:17], v[18:21], off
 ; GISEL-NEXT:    s_endpgm
@@ -1802,19 +1618,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4_non_splat(<16 x i32> %A, <
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s1, 2.0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v34, 1.0 :: v_dual_mov_b32 v35, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v36, v34 :: v_dual_mov_b32 v37, v34
+; GISEL-NEXT:    v_dual_mov_b32 v38, v34 :: v_dual_mov_b32 v39, v34
+; GISEL-NEXT:    v_dual_mov_b32 v40, v34 :: v_dual_mov_b32 v41, v34
 ; GISEL-NEXT:    v_wmma_f32_16x16x128_f8f6f4 v[34:41], v[0:15], v[16:31], v[34:41]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -1851,20 +1659,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4_non_inlineable(<16 x i32>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_dual_mov_b32 v37, v34 :: v_dual_mov_b32 v38, v34
+; GISEL-NEXT:    v_dual_mov_b32 v39, v34 :: v_dual_mov_b32 v40, v34
+; GISEL-NEXT:    v_mov_b32_e32 v41, v34
 ; GISEL-NEXT:    v_wmma_f32_16x16x128_f8f6f4 v[34:41], v[0:15], v[16:31], v[34:41]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -1930,19 +1730,11 @@ define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4_non_splat(<16 x i32>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s1, 2.0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v34, 1.0 :: v_dual_mov_b32 v35, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v36, v34 :: v_dual_mov_b32 v37, v34
+; GISEL-NEXT:    v_dual_mov_b32 v38, v34 :: v_dual_mov_b32 v39, v34
+; GISEL-NEXT:    v_dual_mov_b32 v40, v34 :: v_dual_mov_b32 v41, v34
 ; GISEL-NEXT:    v_wmma_scale_f32_16x16x128_f8f6f4 v[34:41], v[0:15], v[16:31], v[34:41], 1, 2 matrix_a_scale:MATRIX_SCALE_ROW1 matrix_b_scale:MATRIX_SCALE_ROW1
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -1980,21 +1772,13 @@ define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4_non_inlineable(<16 x
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    v_mov_b32_e32 v42, 0x64
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x40400000
 ; GISEL-NEXT:    v_mov_b32_e32 v43, 0x65
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v42, 0x64 :: v_dual_mov_b32 v41, v34
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_dual_mov_b32 v37, v34 :: v_dual_mov_b32 v38, v34
+; GISEL-NEXT:    v_dual_mov_b32 v39, v34 :: v_dual_mov_b32 v40, v34
 ; GISEL-NEXT:    v_wmma_scale_f32_16x16x128_f8f6f4 v[34:41], v[0:15], v[16:31], v[34:41], v42, v43 matrix_a_scale:MATRIX_SCALE_ROW1 matrix_b_scale:MATRIX_SCALE_ROW1 matrix_a_reuse
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -2060,19 +1844,11 @@ define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4_non_splat(<16 x i3
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s1, 2.0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v34, 1.0 :: v_dual_mov_b32 v35, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v36, v34 :: v_dual_mov_b32 v37, v34
+; GISEL-NEXT:    v_dual_mov_b32 v38, v34 :: v_dual_mov_b32 v39, v34
+; GISEL-NEXT:    v_dual_mov_b32 v40, v34 :: v_dual_mov_b32 v41, v34
 ; GISEL-NEXT:    v_wmma_scale16_f32_16x16x128_f8f6f4 v[34:41], v[0:15], v[16:31], v[34:41], 1, 2 matrix_a_scale:MATRIX_SCALE_ROW1 matrix_b_scale:MATRIX_SCALE_ROW1
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -2111,21 +1887,14 @@ define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4_non_inlineable(<16
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x40400000
 ; GISEL-NEXT:    v_mov_b64_e32 v[42:43], 0x64
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
 ; GISEL-NEXT:    v_mov_b64_e32 v[44:45], 0x65
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_dual_mov_b32 v37, v34 :: v_dual_mov_b32 v38, v34
+; GISEL-NEXT:    v_dual_mov_b32 v39, v34 :: v_dual_mov_b32 v40, v34
+; GISEL-NEXT:    v_mov_b32_e32 v41, v34
 ; GISEL-NEXT:    v_wmma_scale16_f32_16x16x128_f8f6f4 v[34:41], v[0:15], v[16:31], v[34:41], v[42:43], v[44:45] matrix_a_scale:MATRIX_SCALE_ROW1 matrix_b_scale:MATRIX_SCALE_ROW1 matrix_a_reuse
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -2184,13 +1953,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x128_fp8_fp8_non_splat(<16 x i32> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GISEL-NEXT:    s_mov_b32 s1, 0x3c004000
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x3c003c00
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, 0x3c004000 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_mov_b32_e32 v37, v34
 ; GISEL-NEXT:    v_wmma_f16_16x16x128_fp8_fp8 v[34:37], v[0:15], v[16:31], v[34:37]
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
 ; GISEL-NEXT:    s_endpgm
@@ -2221,13 +1987,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x128_fp8_fp8_non_inlineable(<16 x i32>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x42004200
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x42004200
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_mov_b32_e32 v37, v34
 ; GISEL-NEXT:    v_wmma_f16_16x16x128_fp8_fp8 v[34:37], v[0:15], v[16:31], v[34:37]
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
 ; GISEL-NEXT:    s_endpgm
@@ -2284,13 +2047,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x128_fp8_bf8_non_splat(<16 x i32> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GISEL-NEXT:    s_mov_b32 s1, 0x3c004000
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x3c003c00
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, 0x3c004000 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_mov_b32_e32 v37, v34
 ; GISEL-NEXT:    v_wmma_f16_16x16x128_fp8_bf8 v[34:37], v[0:15], v[16:31], v[34:37] neg_hi:[0,0,1]
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
 ; GISEL-NEXT:    s_endpgm
@@ -2321,13 +2081,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x128_fp8_bf8_non_inlineable(<16 x i32>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x42004200
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x42004200
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_mov_b32_e32 v37, v34
 ; GISEL-NEXT:    v_wmma_f16_16x16x128_fp8_bf8 v[34:37], v[0:15], v[16:31], v[34:37] neg_hi:[0,0,1]
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
 ; GISEL-NEXT:    s_endpgm
@@ -2384,13 +2141,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x128_bf8_fp8_non_splat(<16 x i32> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GISEL-NEXT:    s_mov_b32 s1, 0x3c004000
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x3c003c00
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, 0x3c004000 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_mov_b32_e32 v37, v34
 ; GISEL-NEXT:    v_wmma_f16_16x16x128_bf8_fp8 v[34:37], v[0:15], v[16:31], v[34:37]
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
 ; GISEL-NEXT:    s_endpgm
@@ -2421,13 +2175,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x128_bf8_fp8_non_inlineable(<16 x i32>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x42004200
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x42004200
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_mov_b32_e32 v37, v34
 ; GISEL-NEXT:    v_wmma_f16_16x16x128_bf8_fp8 v[34:37], v[0:15], v[16:31], v[34:37]
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
 ; GISEL-NEXT:    s_endpgm
@@ -2484,13 +2235,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x128_bf8_bf8_non_splat(<16 x i32> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x3c003c00
-; GISEL-NEXT:    s_mov_b32 s1, 0x3c004000
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x3c003c00
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, 0x3c004000 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_mov_b32_e32 v37, v34
 ; GISEL-NEXT:    v_wmma_f16_16x16x128_bf8_bf8 v[34:37], v[0:15], v[16:31], v[34:37]
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
 ; GISEL-NEXT:    s_endpgm
@@ -2521,13 +2269,10 @@ define amdgpu_ps void @test_wmma_f16_16x16x128_bf8_bf8_non_inlineable(<16 x i32>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x42004200
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x42004200
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_mov_b32_e32 v37, v34
 ; GISEL-NEXT:    v_wmma_f16_16x16x128_bf8_bf8 v[34:37], v[0:15], v[16:31], v[34:37]
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
 ; GISEL-NEXT:    s_endpgm
@@ -2591,19 +2336,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x128_fp8_fp8_non_splat(<16 x i32> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v34, 1.0 :: v_dual_mov_b32 v36, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v37, v34
+; GISEL-NEXT:    v_dual_mov_b32 v38, v34 :: v_dual_mov_b32 v39, v34
+; GISEL-NEXT:    v_dual_mov_b32 v40, v34 :: v_dual_mov_b32 v41, v34
 ; GISEL-NEXT:    v_wmma_f32_16x16x128_fp8_fp8 v[34:41], v[0:15], v[16:31], v[34:41]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -2640,20 +2377,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x128_fp8_fp8_non_inlineable(<16 x i32>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_dual_mov_b32 v37, v34 :: v_dual_mov_b32 v38, v34
+; GISEL-NEXT:    v_dual_mov_b32 v39, v34 :: v_dual_mov_b32 v40, v34
+; GISEL-NEXT:    v_mov_b32_e32 v41, v34
 ; GISEL-NEXT:    v_wmma_f32_16x16x128_fp8_fp8 v[34:41], v[0:15], v[16:31], v[34:41]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -2719,19 +2448,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x128_fp8_bf8_non_splat(<16 x i32> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v34, 1.0 :: v_dual_mov_b32 v36, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v37, v34
+; GISEL-NEXT:    v_dual_mov_b32 v38, v34 :: v_dual_mov_b32 v39, v34
+; GISEL-NEXT:    v_dual_mov_b32 v40, v34 :: v_dual_mov_b32 v41, v34
 ; GISEL-NEXT:    v_wmma_f32_16x16x128_fp8_bf8 v[34:41], v[0:15], v[16:31], v[34:41]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -2768,20 +2489,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x128_fp8_bf8_non_inlineable(<16 x i32>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_dual_mov_b32 v37, v34 :: v_dual_mov_b32 v38, v34
+; GISEL-NEXT:    v_dual_mov_b32 v39, v34 :: v_dual_mov_b32 v40, v34
+; GISEL-NEXT:    v_mov_b32_e32 v41, v34
 ; GISEL-NEXT:    v_wmma_f32_16x16x128_fp8_bf8 v[34:41], v[0:15], v[16:31], v[34:41]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -2847,19 +2560,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x128_bf8_fp8_non_splat(<16 x i32> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v34, 1.0 :: v_dual_mov_b32 v36, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v37, v34
+; GISEL-NEXT:    v_dual_mov_b32 v38, v34 :: v_dual_mov_b32 v39, v34
+; GISEL-NEXT:    v_dual_mov_b32 v40, v34 :: v_dual_mov_b32 v41, v34
 ; GISEL-NEXT:    v_wmma_f32_16x16x128_bf8_fp8 v[34:41], v[0:15], v[16:31], v[34:41]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -2896,20 +2601,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x128_bf8_fp8_non_inlineable(<16 x i32>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_dual_mov_b32 v37, v34 :: v_dual_mov_b32 v38, v34
+; GISEL-NEXT:    v_dual_mov_b32 v39, v34 :: v_dual_mov_b32 v40, v34
+; GISEL-NEXT:    v_mov_b32_e32 v41, v34
 ; GISEL-NEXT:    v_wmma_f32_16x16x128_bf8_fp8 v[34:41], v[0:15], v[16:31], v[34:41]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -2975,19 +2672,11 @@ define amdgpu_ps void @test_wmma_f32_16x16x128_bf8_bf8_non_splat(<16 x i32> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v34, 1.0 :: v_dual_mov_b32 v36, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v37, v34
+; GISEL-NEXT:    v_dual_mov_b32 v38, v34 :: v_dual_mov_b32 v39, v34
+; GISEL-NEXT:    v_dual_mov_b32 v40, v34 :: v_dual_mov_b32 v41, v34
 ; GISEL-NEXT:    v_wmma_f32_16x16x128_bf8_bf8 v[34:41], v[0:15], v[16:31], v[34:41]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -3024,20 +2713,12 @@ define amdgpu_ps void @test_wmma_f32_16x16x128_bf8_bf8_non_inlineable(<16 x i32>
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[0:1]
-; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b32_e32 v34, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v35, v34 :: v_dual_mov_b32 v36, v34
+; GISEL-NEXT:    v_dual_mov_b32 v37, v34 :: v_dual_mov_b32 v38, v34
+; GISEL-NEXT:    v_dual_mov_b32 v39, v34 :: v_dual_mov_b32 v40, v34
+; GISEL-NEXT:    v_mov_b32_e32 v41, v34
 ; GISEL-NEXT:    v_wmma_f32_16x16x128_bf8_bf8 v[34:41], v[0:15], v[16:31], v[34:41]
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    global_store_b128 v[32:33], v[34:37], off
@@ -3114,30 +2795,15 @@ define amdgpu_ps void @test_wmma_f32_32x16x128_f4_non_splat(<16 x i32> %A, <8 x
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s2, 2.0
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    s_mov_b32 s8, s0
-; GISEL-NEXT:    s_mov_b32 s9, s0
-; GISEL-NEXT:    s_mov_b32 s10, s2
-; GISEL-NEXT:    s_mov_b32 s11, s0
-; GISEL-NEXT:    s_mov_b32 s12, s0
-; GISEL-NEXT:    s_mov_b32 s13, s0
-; GISEL-NEXT:    s_mov_b32 s14, s0
-; GISEL-NEXT:    s_mov_b32 s15, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[12:13]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[14:15]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[8:9]
-; GISEL-NEXT:    v_mov_b64_e32 v[32:33], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[30:31], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[28:29], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[26:27], s[0:1]
+; GISEL-NEXT:    v_dual_mov_b32 v26, 1.0 :: v_dual_mov_b32 v28, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v27, v26 :: v_dual_mov_b32 v29, v26
+; GISEL-NEXT:    v_dual_mov_b32 v30, v26 :: v_dual_mov_b32 v31, v26
+; GISEL-NEXT:    v_dual_mov_b32 v32, v26 :: v_dual_mov_b32 v33, v26
+; GISEL-NEXT:    v_dual_mov_b32 v34, v26 :: v_dual_mov_b32 v35, v26
+; GISEL-NEXT:    v_dual_mov_b32 v36, v28 :: v_dual_mov_b32 v37, v26
+; GISEL-NEXT:    v_dual_mov_b32 v38, v26 :: v_dual_mov_b32 v39, v26
+; GISEL-NEXT:    v_dual_mov_b32 v40, v26 :: v_dual_mov_b32 v41, v26
 ; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GISEL-NEXT:    v_wmma_f32_32x16x128_f4 v[26:41], v[0:15], v[16:23], v[26:41]
 ; GISEL-NEXT:    s_clause 0x3
@@ -3184,31 +2850,16 @@ define amdgpu_ps void @test_wmma_f32_32x16x128_f4_non_inlineable(<16 x i32> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    s_mov_b32 s8, s0
-; GISEL-NEXT:    s_mov_b32 s9, s0
-; GISEL-NEXT:    s_mov_b32 s10, s0
-; GISEL-NEXT:    s_mov_b32 s11, s0
-; GISEL-NEXT:    s_mov_b32 s12, s0
-; GISEL-NEXT:    s_mov_b32 s13, s0
-; GISEL-NEXT:    s_mov_b32 s14, s0
-; GISEL-NEXT:    s_mov_b32 s15, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[12:13]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[14:15]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[8:9]
-; GISEL-NEXT:    v_mov_b64_e32 v[32:33], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[30:31], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[28:29], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[26:27], s[0:1]
+; GISEL-NEXT:    v_mov_b32_e32 v26, 0x40400000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v27, v26 :: v_dual_mov_b32 v28, v26
+; GISEL-NEXT:    v_dual_mov_b32 v29, v26 :: v_dual_mov_b32 v30, v26
+; GISEL-NEXT:    v_dual_mov_b32 v31, v26 :: v_dual_mov_b32 v32, v26
+; GISEL-NEXT:    v_dual_mov_b32 v33, v26 :: v_dual_mov_b32 v34, v26
+; GISEL-NEXT:    v_dual_mov_b32 v35, v26 :: v_dual_mov_b32 v36, v26
+; GISEL-NEXT:    v_dual_mov_b32 v37, v26 :: v_dual_mov_b32 v38, v26
+; GISEL-NEXT:    v_dual_mov_b32 v39, v26 :: v_dual_mov_b32 v40, v26
+; GISEL-NEXT:    v_mov_b32_e32 v41, v26
 ; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GISEL-NEXT:    v_wmma_f32_32x16x128_f4 v[26:41], v[0:15], v[16:23], v[26:41]
 ; GISEL-NEXT:    s_clause 0x3
@@ -3288,30 +2939,15 @@ define amdgpu_ps void @test_wmma_scale_f32_32x16x128_f4_non_splat(<16 x i32> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s1, 2.0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    s_mov_b32 s8, s0
-; GISEL-NEXT:    s_mov_b32 s9, s0
-; GISEL-NEXT:    s_mov_b32 s10, s0
-; GISEL-NEXT:    s_mov_b32 s11, s0
-; GISEL-NEXT:    s_mov_b32 s12, s0
-; GISEL-NEXT:    s_mov_b32 s13, s0
-; GISEL-NEXT:    s_mov_b32 s14, s0
-; GISEL-NEXT:    s_mov_b32 s15, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[12:13]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[14:15]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[8:9]
-; GISEL-NEXT:    v_mov_b64_e32 v[32:33], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[30:31], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[28:29], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[26:27], s[0:1]
+; GISEL-NEXT:    v_dual_mov_b32 v26, 1.0 :: v_dual_mov_b32 v27, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v28, v26 :: v_dual_mov_b32 v29, v26
+; GISEL-NEXT:    v_dual_mov_b32 v30, v26 :: v_dual_mov_b32 v31, v26
+; GISEL-NEXT:    v_dual_mov_b32 v32, v26 :: v_dual_mov_b32 v33, v26
+; GISEL-NEXT:    v_dual_mov_b32 v34, v26 :: v_dual_mov_b32 v35, v26
+; GISEL-NEXT:    v_dual_mov_b32 v36, v26 :: v_dual_mov_b32 v37, v26
+; GISEL-NEXT:    v_dual_mov_b32 v38, v26 :: v_dual_mov_b32 v39, v26
+; GISEL-NEXT:    v_dual_mov_b32 v40, v26 :: v_dual_mov_b32 v41, v26
 ; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GISEL-NEXT:    v_wmma_scale_f32_32x16x128_f4 v[26:41], v[0:15], v[16:23], v[26:41], 1, 2 matrix_a_scale:MATRIX_SCALE_ROW1 matrix_b_scale:MATRIX_SCALE_ROW1
 ; GISEL-NEXT:    s_clause 0x3
@@ -3359,32 +2995,17 @@ define amdgpu_ps void @test_wmma_scale_f32_32x16x128_f4_non_inlineable(<16 x i32
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
-; GISEL-NEXT:    v_mov_b32_e32 v42, 0x64
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    s_mov_b32 s8, s0
-; GISEL-NEXT:    s_mov_b32 s9, s0
-; GISEL-NEXT:    s_mov_b32 s10, s0
-; GISEL-NEXT:    s_mov_b32 s11, s0
-; GISEL-NEXT:    s_mov_b32 s12, s0
-; GISEL-NEXT:    s_mov_b32 s13, s0
-; GISEL-NEXT:    s_mov_b32 s14, s0
-; GISEL-NEXT:    s_mov_b32 s15, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[12:13]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[14:15]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[8:9]
-; GISEL-NEXT:    v_mov_b64_e32 v[32:33], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[30:31], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[28:29], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[26:27], s[0:1]
+; GISEL-NEXT:    v_mov_b32_e32 v26, 0x40400000
 ; GISEL-NEXT:    v_mov_b32_e32 v43, 0x65
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GISEL-NEXT:    v_dual_mov_b32 v42, 0x64 :: v_dual_mov_b32 v41, v26
+; GISEL-NEXT:    v_dual_mov_b32 v27, v26 :: v_dual_mov_b32 v28, v26
+; GISEL-NEXT:    v_dual_mov_b32 v29, v26 :: v_dual_mov_b32 v30, v26
+; GISEL-NEXT:    v_dual_mov_b32 v31, v26 :: v_dual_mov_b32 v32, v26
+; GISEL-NEXT:    v_dual_mov_b32 v33, v26 :: v_dual_mov_b32 v34, v26
+; GISEL-NEXT:    v_dual_mov_b32 v35, v26 :: v_dual_mov_b32 v36, v26
+; GISEL-NEXT:    v_dual_mov_b32 v37, v26 :: v_dual_mov_b32 v38, v26
+; GISEL-NEXT:    v_dual_mov_b32 v39, v26 :: v_dual_mov_b32 v40, v26
 ; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GISEL-NEXT:    v_wmma_scale_f32_32x16x128_f4 v[26:41], v[0:15], v[16:23], v[26:41], v42, v43 matrix_a_scale:MATRIX_SCALE_ROW1 matrix_b_scale:MATRIX_SCALE_ROW1 matrix_a_reuse
 ; GISEL-NEXT:    s_clause 0x3
@@ -3464,30 +3085,15 @@ define amdgpu_ps void @test_wmma_scale16_f32_32x16x128_f4_non_splat(<16 x i32> %
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 1.0
-; GISEL-NEXT:    s_mov_b32 s1, 2.0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    s_mov_b32 s8, s0
-; GISEL-NEXT:    s_mov_b32 s9, s0
-; GISEL-NEXT:    s_mov_b32 s10, s0
-; GISEL-NEXT:    s_mov_b32 s11, s0
-; GISEL-NEXT:    s_mov_b32 s12, s0
-; GISEL-NEXT:    s_mov_b32 s13, s0
-; GISEL-NEXT:    s_mov_b32 s14, s0
-; GISEL-NEXT:    s_mov_b32 s15, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[12:13]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[14:15]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[8:9]
-; GISEL-NEXT:    v_mov_b64_e32 v[32:33], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[30:31], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[28:29], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[26:27], s[0:1]
+; GISEL-NEXT:    v_dual_mov_b32 v26, 1.0 :: v_dual_mov_b32 v27, 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_dual_mov_b32 v28, v26 :: v_dual_mov_b32 v29, v26
+; GISEL-NEXT:    v_dual_mov_b32 v30, v26 :: v_dual_mov_b32 v31, v26
+; GISEL-NEXT:    v_dual_mov_b32 v32, v26 :: v_dual_mov_b32 v33, v26
+; GISEL-NEXT:    v_dual_mov_b32 v34, v26 :: v_dual_mov_b32 v35, v26
+; GISEL-NEXT:    v_dual_mov_b32 v36, v26 :: v_dual_mov_b32 v37, v26
+; GISEL-NEXT:    v_dual_mov_b32 v38, v26 :: v_dual_mov_b32 v39, v26
+; GISEL-NEXT:    v_dual_mov_b32 v40, v26 :: v_dual_mov_b32 v41, v26
 ; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GISEL-NEXT:    v_wmma_scale16_f32_32x16x128_f4 v[26:41], v[0:15], v[16:23], v[26:41], 1, 2 matrix_a_scale:MATRIX_SCALE_ROW1 matrix_b_scale:MATRIX_SCALE_ROW1
 ; GISEL-NEXT:    s_clause 0x3
@@ -3536,32 +3142,18 @@ define amdgpu_ps void @test_wmma_scale16_f32_32x16x128_f4_non_inlineable(<16 x i
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b32 s0, 0x40400000
+; GISEL-NEXT:    v_mov_b32_e32 v26, 0x40400000
 ; GISEL-NEXT:    v_mov_b64_e32 v[42:43], 0x64
-; GISEL-NEXT:    s_mov_b32 s1, s0
-; GISEL-NEXT:    s_mov_b32 s2, s0
-; GISEL-NEXT:    s_mov_b32 s3, s0
-; GISEL-NEXT:    s_mov_b32 s4, s0
-; GISEL-NEXT:    s_mov_b32 s5, s0
-; GISEL-NEXT:    s_mov_b32 s6, s0
-; GISEL-NEXT:    s_mov_b32 s7, s0
-; GISEL-NEXT:    s_mov_b32 s8, s0
-; GISEL-NEXT:    s_mov_b32 s9, s0
-; GISEL-NEXT:    s_mov_b32 s10, s0
-; GISEL-NEXT:    s_mov_b32 s11, s0
-; GISEL-NEXT:    s_mov_b32 s12, s0
-; GISEL-NEXT:    s_mov_b32 s13, s0
-; GISEL-NEXT:    s_mov_b32 s14, s0
-; GISEL-NEXT:    s_mov_b32 s15, s0
-; GISEL-NEXT:    v_mov_b64_e32 v[38:39], s[12:13]
-; GISEL-NEXT:    v_mov_b64_e32 v[40:41], s[14:15]
-; GISEL-NEXT:    v_mov_b64_e32 v[36:37], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[34:35], s[8:9]
-; GISEL-NEXT:    v_mov_b64_e32 v[32:33], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[30:31], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[28:29], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[26:27], s[0:1]
 ; GISEL-NEXT:    v_mov_b64_e32 v[44:45], 0x65
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3)
+; GISEL-NEXT:    v_dual_mov_b32 v27, v26 :: v_dual_mov_b32 v28, v26
+; GISEL-NEXT:    v_dual_mov_b32 v29, v26 :: v_dual_mov_b32 v30, v26
+; GISEL-NEXT:    v_dual_mov_b32 v31, v26 :: v_dual_mov_b32 v32, v26
+; GISEL-NEXT:    v_dual_mov_b32 v33, v26 :: v_dual_mov_b32 v34, v26
+; GISEL-NEXT:    v_dual_mov_b32 v35, v26 :: v_dual_mov_b32 v36, v26
+; GISEL-NEXT:    v_dual_mov_b32 v37, v26 :: v_dual_mov_b32 v38, v26
+; GISEL-NEXT:    v_dual_mov_b32 v39, v26 :: v_dual_mov_b32 v40, v26
+; GISEL-NEXT:    v_mov_b32_e32 v41, v26
 ; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GISEL-NEXT:    v_wmma_scale16_f32_32x16x128_f4 v[26:41], v[0:15], v[16:23], v[26:41], v[42:43], v[44:45] matrix_a_scale:MATRIX_SCALE_ROW1 matrix_b_scale:MATRIX_SCALE_ROW1 matrix_a_reuse
 ; GISEL-NEXT:    s_clause 0x3
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma.imm.gfx1251.w32.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma.imm.gfx1251.w32.ll
index 7c95314e42bf5..76c9f1621fe1a 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma.imm.gfx1251.w32.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma.imm.gfx1251.w32.ll
@@ -67,22 +67,15 @@ define amdgpu_ps void @test_wmma_f64_16x16x4_f64_non_splat(<2 x double> %A, <2 x
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b64 s[0:1], 1.0
-; GISEL-NEXT:    s_mov_b64 s[4:5], 2.0
-; GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[8:9], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[10:11], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[12:13], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[14:15], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[12:13]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[14:15]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[8:9]
-; GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[0:1]
+; GISEL-NEXT:    v_mov_b64_e32 v[10:11], 1.0
+; GISEL-NEXT:    v_mov_b64_e32 v[14:15], 2.0
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GISEL-NEXT:    v_mov_b64_e32 v[12:13], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[16:17], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[18:19], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[20:21], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[22:23], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[24:25], v[10:11]
 ; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GISEL-NEXT:    v_wmma_f64_16x16x4_f64 v[10:25], v[0:3], v[4:7], v[10:25]
 ; GISEL-NEXT:    s_clause 0x3
@@ -128,23 +121,15 @@ define amdgpu_ps void @test_wmma_f64_16x16x4_f64_non_inlineable(<2 x double> %A,
 ; GISEL-NEXT:    v_nop
 ; GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GISEL-NEXT:    s_mov_b64 s[0:1], 0x4008000000000000
-; GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[4:5], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[6:7], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[8:9], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[10:11], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[12:13], s[0:1]
-; GISEL-NEXT:    s_mov_b64 s[14:15], s[0:1]
-; GISEL-NEXT:    v_mov_b64_e32 v[22:23], s[12:13]
-; GISEL-NEXT:    v_mov_b64_e32 v[24:25], s[14:15]
-; GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[8:9]
-; GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[6:7]
-; GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[4:5]
-; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[0:1]
+; GISEL-NEXT:    v_mov_b64_e32 v[10:11], 0x4008000000000000
+; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GISEL-NEXT:    v_mov_b64_e32 v[12:13], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[14:15], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[16:17], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[18:19], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[20:21], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[22:23], v[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[24:25], v[10:11]
 ; GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GISEL-NEXT:    v_wmma_f64_16x16x4_f64 v[10:25], v[0:3], v[4:7], v[10:25]
 ; GISEL-NEXT:    s_clause 0x3
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.exp.ll b/llvm/test/CodeGen/AMDGPU/llvm.exp.ll
index a8559b404c646..398f042ae5445 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.exp.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.exp.ll
@@ -615,15 +615,15 @@ define amdgpu_kernel void @s_exp_v2f32(ptr addrspace(1) %out, <2 x float> %in) {
 ; SI-GISEL-NEXT:    s_or_b64 s[4:5], vcc, vcc
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v0, v0, v1
 ; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s3, v4
-; SI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s6
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s6, v0
+; SI-GISEL-NEXT:    s_cselect_b32 s6, 0x7f800000, s6
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; SI-GISEL-NEXT:    s_or_b64 s[4:5], vcc, vcc
 ; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s3, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
-; SI-GISEL-NEXT:    s_or_b64 s[4:5], vcc, vcc
-; SI-GISEL-NEXT:    s_cselect_b32 s3, 0x7f800000, s6
-; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; SI-GISEL-NEXT:    s_cselect_b32 s4, 0, s2
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s4
+; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s2
 ; SI-GISEL-NEXT:    s_mov_b32 s2, -1
 ; SI-GISEL-NEXT:    s_mov_b32 s3, 0xf000
 ; SI-GISEL-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
@@ -957,7 +957,6 @@ define amdgpu_kernel void @s_exp_v3f32(ptr addrspace(1) %out, <3 x float> %in) {
 ; VI-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x34
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v0, 0x3fb8a000
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v1, 0x39a3b295
-; VI-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x24
 ; VI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-GISEL-NEXT:    s_and_b32 s3, s0, 0xfffff000
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v2, s3
@@ -982,15 +981,15 @@ define amdgpu_kernel void @s_exp_v3f32(ptr addrspace(1) %out, <3 x float> %in) {
 ; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s0, v3
 ; VI-GISEL-NEXT:    s_cselect_b32 s3, 0, s3
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s0, 0x7f800000, s3
-; VI-GISEL-NEXT:    s_and_b32 s3, s1, 0xfffff000
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s3
+; VI-GISEL-NEXT:    s_cselect_b32 s3, 0x7f800000, s3
+; VI-GISEL-NEXT:    s_and_b32 s0, s1, 0xfffff000
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v4, s1, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v6, 0x39a3b295, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v4, 0x3fb8a000, v4
-; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s3, v0
+; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s0, v0
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v4, v6
-; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s3, v1
+; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s0, v1
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v6, v4
 ; VI-GISEL-NEXT:    v_rndne_f32_e32 v6, v5
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v5, v5, v6
@@ -1001,37 +1000,39 @@ define amdgpu_kernel void @s_exp_v3f32(ptr addrspace(1) %out, <3 x float> %in) {
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
 ; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s1, v3
 ; VI-GISEL-NEXT:    v_ldexp_f32 v4, v4, v5
-; VI-GISEL-NEXT:    v_readfirstlane_b32 s3, v4
-; VI-GISEL-NEXT:    s_cselect_b32 s3, 0, s3
+; VI-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
+; VI-GISEL-NEXT:    s_cselect_b32 s0, 0, s0
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s1, 0x7f800000, s3
-; VI-GISEL-NEXT:    s_and_b32 s3, s2, 0xfffff000
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s3
+; VI-GISEL-NEXT:    s_cselect_b32 s6, 0x7f800000, s0
+; VI-GISEL-NEXT:    s_and_b32 s0, s2, 0xfffff000
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v4, s2, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v5, 0x39a3b295, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v4, 0x3fb8a000, v4
-; VI-GISEL-NEXT:    v_mul_f32_e32 v0, s3, v0
+; VI-GISEL-NEXT:    v_mul_f32_e32 v0, s0, v0
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v4, v5
-; VI-GISEL-NEXT:    v_mul_f32_e32 v1, s3, v1
+; VI-GISEL-NEXT:    v_mul_f32_e32 v1, s0, v1
 ; VI-GISEL-NEXT:    v_add_f32_e32 v1, v1, v4
 ; VI-GISEL-NEXT:    v_rndne_f32_e32 v4, v0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v0, v0, v4
 ; VI-GISEL-NEXT:    v_add_f32_e32 v0, v0, v1
 ; VI-GISEL-NEXT:    v_cvt_i32_f32_e32 v1, v4
 ; VI-GISEL-NEXT:    v_exp_f32_e32 v0, v0
+; VI-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; VI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s2, v2
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s2, v3
 ; VI-GISEL-NEXT:    v_ldexp_f32 v0, v0, v1
-; VI-GISEL-NEXT:    v_readfirstlane_b32 s3, v0
-; VI-GISEL-NEXT:    s_cselect_b32 s3, 0, s3
+; VI-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s2, v3
+; VI-GISEL-NEXT:    s_cselect_b32 s4, 0, s4
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s3
-; VI-GISEL-NEXT:    v_mov_b32_e32 v0, s0
-; VI-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; VI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s4
+; VI-GISEL-NEXT:    v_mov_b32_e32 v0, s3
+; VI-GISEL-NEXT:    v_mov_b32_e32 v1, s6
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v2, s2
-; VI-GISEL-NEXT:    v_mov_b32_e32 v3, s4
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s5
+; VI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s1
+; VI-GISEL-NEXT:    v_mov_b32_e32 v3, s0
 ; VI-GISEL-NEXT:    flat_store_dwordx3 v[3:4], v[0:2]
 ; VI-GISEL-NEXT:    s_endpgm
 ;
@@ -1209,69 +1210,69 @@ define amdgpu_kernel void @s_exp_v3f32(ptr addrspace(1) %out, <3 x float> %in) {
 ;
 ; SI-GISEL-LABEL: s_exp_v3f32:
 ; SI-GISEL:       ; %bb.0:
-; SI-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0xd
+; SI-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0xd
+; SI-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v0, 0x3fb8aa3b
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v1, 0x32a5705f
-; SI-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x9
 ; SI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-GISEL-NEXT:    v_mul_f32_e32 v2, s0, v0
-; SI-GISEL-NEXT:    v_fma_f32 v3, s0, v0, -v2
+; SI-GISEL-NEXT:    v_mul_f32_e32 v2, s8, v0
+; SI-GISEL-NEXT:    v_fma_f32 v3, s8, v0, -v2
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v4, v2
-; SI-GISEL-NEXT:    v_fma_f32 v3, s0, v1, v3
+; SI-GISEL-NEXT:    v_fma_f32 v3, s8, v1, v3
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v2, v2, v4
 ; SI-GISEL-NEXT:    v_add_f32_e32 v2, v2, v3
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v3, v4
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v2, v2
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v4, 0xc2ce8ed0
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s0, v4
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s8, v4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v2, v2, v3
-; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s1, v0
-; SI-GISEL-NEXT:    v_fma_f32 v5, s1, v0, -v3
+; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s9, v0
+; SI-GISEL-NEXT:    v_fma_f32 v5, s9, v0, -v3
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v6, v3
-; SI-GISEL-NEXT:    v_fma_f32 v5, s1, v1, v5
+; SI-GISEL-NEXT:    v_fma_f32 v5, s9, v1, v5
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v3, v3, v6
 ; SI-GISEL-NEXT:    v_add_f32_e32 v3, v3, v5
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v5, v6
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v3, v3
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s3, v2
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v2, 0x42b17218
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s0, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s3, 0, s3
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s8, v2
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v3, v3, v5
-; SI-GISEL-NEXT:    s_cselect_b32 s0, 0x7f800000, s3
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
-; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s2, v0
-; SI-GISEL-NEXT:    v_fma_f32 v0, s2, v0, -v3
-; SI-GISEL-NEXT:    v_fma_f32 v0, s2, v1, v0
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s10, v0
+; SI-GISEL-NEXT:    v_fma_f32 v0, s10, v0, -v3
+; SI-GISEL-NEXT:    v_fma_f32 v0, s10, v1, v0
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v1, v3
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v3, v3, v1
 ; SI-GISEL-NEXT:    v_add_f32_e32 v0, v3, v0
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v1, v1
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v0, v0
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s1, v4
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s1, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s3, 0, s3
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    s_cselect_b32 s4, 0, s4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s9, v4
+; SI-GISEL-NEXT:    s_cselect_b32 s4, 0x7f800000, s4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s9, v2
+; SI-GISEL-NEXT:    s_cselect_b32 s5, 0, s5
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v0, v0, v1
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s2, v4
-; SI-GISEL-NEXT:    s_cselect_b32 s1, 0x7f800000, s3
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s3, v0
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s2, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s6, 0, s3
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s10, v4
+; SI-GISEL-NEXT:    s_cselect_b32 s5, 0x7f800000, s5
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s6, v0
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s10, v2
+; SI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
 ; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
-; SI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s6
-; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s0
-; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s1
-; SI-GISEL-NEXT:    s_mov_b32 s6, -1
-; SI-GISEL-NEXT:    s_mov_b32 s7, 0xf000
-; SI-GISEL-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-GISEL-NEXT:    s_cselect_b32 s6, 0x7f800000, s6
+; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; SI-GISEL-NEXT:    s_mov_b32 s2, -1
+; SI-GISEL-NEXT:    s_mov_b32 s3, 0xf000
+; SI-GISEL-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; SI-GISEL-NEXT:    s_waitcnt expcnt(0)
-; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; SI-GISEL-NEXT:    buffer_store_dword v0, off, s[4:7], 0 offset:8
+; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; SI-GISEL-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:8
 ; SI-GISEL-NEXT:    s_endpgm
 ;
 ; R600-LABEL: s_exp_v3f32:
@@ -1745,7 +1746,6 @@ define amdgpu_kernel void @s_exp_v4f32(ptr addrspace(1) %out, <4 x float> %in) {
 ; VI-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x34
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v0, 0x3fb8a000
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v1, 0x39a3b295
-; VI-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x24
 ; VI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-GISEL-NEXT:    s_and_b32 s6, s0, 0xfffff000
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v2, s6
@@ -1770,15 +1770,15 @@ define amdgpu_kernel void @s_exp_v4f32(ptr addrspace(1) %out, <4 x float> %in) {
 ; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s0, v3
 ; VI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s0, 0x7f800000, s6
-; VI-GISEL-NEXT:    s_and_b32 s6, s1, 0xfffff000
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s6
+; VI-GISEL-NEXT:    s_cselect_b32 s6, 0x7f800000, s6
+; VI-GISEL-NEXT:    s_and_b32 s0, s1, 0xfffff000
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v4, s1, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v6, 0x39a3b295, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v4, 0x3fb8a000, v4
-; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s6, v0
+; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s0, v0
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v4, v6
-; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s6, v1
+; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s0, v1
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v6, v4
 ; VI-GISEL-NEXT:    v_rndne_f32_e32 v6, v5
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v5, v5, v6
@@ -1789,18 +1789,18 @@ define amdgpu_kernel void @s_exp_v4f32(ptr addrspace(1) %out, <4 x float> %in) {
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
 ; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s1, v3
 ; VI-GISEL-NEXT:    v_ldexp_f32 v4, v4, v5
-; VI-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
-; VI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
+; VI-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
+; VI-GISEL-NEXT:    s_cselect_b32 s0, 0, s0
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s1, 0x7f800000, s6
-; VI-GISEL-NEXT:    s_and_b32 s6, s2, 0xfffff000
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s6
+; VI-GISEL-NEXT:    s_cselect_b32 s7, 0x7f800000, s0
+; VI-GISEL-NEXT:    s_and_b32 s0, s2, 0xfffff000
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v4, s2, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v6, 0x39a3b295, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v4, 0x3fb8a000, v4
-; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s6, v0
+; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s0, v0
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v4, v6
-; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s6, v1
+; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s0, v1
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v6, v4
 ; VI-GISEL-NEXT:    v_rndne_f32_e32 v6, v5
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v5, v5, v6
@@ -1811,38 +1811,40 @@ define amdgpu_kernel void @s_exp_v4f32(ptr addrspace(1) %out, <4 x float> %in) {
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
 ; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s2, v3
 ; VI-GISEL-NEXT:    v_ldexp_f32 v4, v4, v5
-; VI-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
-; VI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
+; VI-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
+; VI-GISEL-NEXT:    s_cselect_b32 s0, 0, s0
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s6
-; VI-GISEL-NEXT:    s_and_b32 s6, s3, 0xfffff000
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s6
+; VI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s0
+; VI-GISEL-NEXT:    s_and_b32 s0, s3, 0xfffff000
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v4, s3, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v5, 0x39a3b295, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v4, 0x3fb8a000, v4
-; VI-GISEL-NEXT:    v_mul_f32_e32 v0, s6, v0
+; VI-GISEL-NEXT:    v_mul_f32_e32 v0, s0, v0
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v4, v5
-; VI-GISEL-NEXT:    v_mul_f32_e32 v1, s6, v1
+; VI-GISEL-NEXT:    v_mul_f32_e32 v1, s0, v1
 ; VI-GISEL-NEXT:    v_add_f32_e32 v1, v1, v4
 ; VI-GISEL-NEXT:    v_rndne_f32_e32 v4, v0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v0, v0, v4
 ; VI-GISEL-NEXT:    v_add_f32_e32 v0, v0, v1
 ; VI-GISEL-NEXT:    v_cvt_i32_f32_e32 v1, v4
 ; VI-GISEL-NEXT:    v_exp_f32_e32 v0, v0
+; VI-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; VI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s3, v2
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s3, v3
 ; VI-GISEL-NEXT:    v_ldexp_f32 v0, v0, v1
-; VI-GISEL-NEXT:    v_readfirstlane_b32 s6, v0
-; VI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
+; VI-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s3, v3
+; VI-GISEL-NEXT:    s_cselect_b32 s4, 0, s4
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s3, 0x7f800000, s6
-; VI-GISEL-NEXT:    v_mov_b32_e32 v0, s0
-; VI-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; VI-GISEL-NEXT:    s_cselect_b32 s3, 0x7f800000, s4
+; VI-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; VI-GISEL-NEXT:    v_mov_b32_e32 v1, s7
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v3, s3
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s4
-; VI-GISEL-NEXT:    v_mov_b32_e32 v5, s5
+; VI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-GISEL-NEXT:    v_mov_b32_e32 v5, s1
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-GISEL-NEXT:    s_endpgm
 ;
@@ -2063,84 +2065,84 @@ define amdgpu_kernel void @s_exp_v4f32(ptr addrspace(1) %out, <4 x float> %in) {
 ;
 ; SI-GISEL-LABEL: s_exp_v4f32:
 ; SI-GISEL:       ; %bb.0:
-; SI-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0xd
+; SI-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0xd
+; SI-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v0, 0x3fb8aa3b
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v1, 0x32a5705f
-; SI-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x9
 ; SI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-GISEL-NEXT:    v_mul_f32_e32 v2, s0, v0
-; SI-GISEL-NEXT:    v_fma_f32 v3, s0, v0, -v2
+; SI-GISEL-NEXT:    v_mul_f32_e32 v2, s8, v0
+; SI-GISEL-NEXT:    v_fma_f32 v3, s8, v0, -v2
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v4, v2
-; SI-GISEL-NEXT:    v_fma_f32 v3, s0, v1, v3
+; SI-GISEL-NEXT:    v_fma_f32 v3, s8, v1, v3
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v2, v2, v4
 ; SI-GISEL-NEXT:    v_add_f32_e32 v2, v2, v3
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v3, v4
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v2, v2
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v4, 0xc2ce8ed0
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s0, v4
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s8, v4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v2, v2, v3
-; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s1, v0
-; SI-GISEL-NEXT:    v_fma_f32 v5, s1, v0, -v3
+; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s9, v0
+; SI-GISEL-NEXT:    v_fma_f32 v5, s9, v0, -v3
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v6, v3
-; SI-GISEL-NEXT:    v_fma_f32 v5, s1, v1, v5
+; SI-GISEL-NEXT:    v_fma_f32 v5, s9, v1, v5
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v3, v3, v6
 ; SI-GISEL-NEXT:    v_add_f32_e32 v3, v3, v5
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v5, v6
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v3, v3
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s8, v2
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v2, 0x42b17218
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s0, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s8, 0, s8
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s8, v2
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v3, v3, v5
-; SI-GISEL-NEXT:    s_cselect_b32 s0, 0x7f800000, s8
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s8, v3
-; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s2, v0
-; SI-GISEL-NEXT:    v_fma_f32 v5, s2, v0, -v3
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s10, v0
+; SI-GISEL-NEXT:    v_fma_f32 v5, s10, v0, -v3
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v6, v3
-; SI-GISEL-NEXT:    v_fma_f32 v5, s2, v1, v5
+; SI-GISEL-NEXT:    v_fma_f32 v5, s10, v1, v5
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v3, v3, v6
 ; SI-GISEL-NEXT:    v_add_f32_e32 v3, v3, v5
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v5, v6
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v3, v3
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s1, v4
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s1, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s8, 0, s8
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    s_cselect_b32 s4, 0, s4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s9, v4
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v3, v3, v5
-; SI-GISEL-NEXT:    s_cselect_b32 s1, 0x7f800000, s8
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s8, v3
-; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s3, v0
-; SI-GISEL-NEXT:    v_fma_f32 v0, s3, v0, -v3
-; SI-GISEL-NEXT:    v_fma_f32 v0, s3, v1, v0
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s11, v0
+; SI-GISEL-NEXT:    v_fma_f32 v0, s11, v0, -v3
+; SI-GISEL-NEXT:    v_fma_f32 v0, s11, v1, v0
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v1, v3
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v3, v3, v1
 ; SI-GISEL-NEXT:    v_add_f32_e32 v0, v3, v0
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v1, v1
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v0, v0
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s2, v4
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s2, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s8, 0, s8
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    s_cselect_b32 s4, 0x7f800000, s4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s9, v2
+; SI-GISEL-NEXT:    s_cselect_b32 s5, 0, s5
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s10, v4
+; SI-GISEL-NEXT:    s_cselect_b32 s5, 0x7f800000, s5
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s10, v2
+; SI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v0, v0, v1
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s3, v4
-; SI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s8
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s3, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s8, 0, s8
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    s_cselect_b32 s3, 0x7f800000, s8
-; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s0
-; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s1
-; SI-GISEL-NEXT:    v_mov_b32_e32 v2, s2
-; SI-GISEL-NEXT:    v_mov_b32_e32 v3, s3
-; SI-GISEL-NEXT:    s_mov_b32 s6, -1
-; SI-GISEL-NEXT:    s_mov_b32 s7, 0xf000
-; SI-GISEL-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s11, v4
+; SI-GISEL-NEXT:    s_cselect_b32 s6, 0x7f800000, s6
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s7, v0
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s11, v2
+; SI-GISEL-NEXT:    s_cselect_b32 s7, 0, s7
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s7
+; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; SI-GISEL-NEXT:    v_mov_b32_e32 v2, s6
+; SI-GISEL-NEXT:    v_mov_b32_e32 v3, s2
+; SI-GISEL-NEXT:    s_mov_b32 s2, -1
+; SI-GISEL-NEXT:    s_mov_b32 s3, 0xf000
+; SI-GISEL-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; SI-GISEL-NEXT:    s_endpgm
 ;
 ; R600-LABEL: s_exp_v4f32:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.exp10.ll b/llvm/test/CodeGen/AMDGPU/llvm.exp10.ll
index 5578dacc38dd9..6dd6ae3eca990 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.exp10.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.exp10.ll
@@ -617,15 +617,15 @@ define amdgpu_kernel void @s_exp10_v2f32(ptr addrspace(1) %out, <2 x float> %in)
 ; SI-GISEL-NEXT:    s_or_b64 s[4:5], vcc, vcc
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v0, v0, v1
 ; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s3, v4
-; SI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s6
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s6, v0
+; SI-GISEL-NEXT:    s_cselect_b32 s6, 0x7f800000, s6
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
 ; SI-GISEL-NEXT:    s_or_b64 s[4:5], vcc, vcc
 ; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s3, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
-; SI-GISEL-NEXT:    s_or_b64 s[4:5], vcc, vcc
-; SI-GISEL-NEXT:    s_cselect_b32 s3, 0x7f800000, s6
-; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; SI-GISEL-NEXT:    s_cselect_b32 s4, 0, s2
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s4
+; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s2
 ; SI-GISEL-NEXT:    s_mov_b32 s2, -1
 ; SI-GISEL-NEXT:    s_mov_b32 s3, 0xf000
 ; SI-GISEL-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
@@ -959,7 +959,6 @@ define amdgpu_kernel void @s_exp10_v3f32(ptr addrspace(1) %out, <3 x float> %in)
 ; VI-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x34
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v0, 0x40549000
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3a2784bc
-; VI-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x24
 ; VI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-GISEL-NEXT:    s_and_b32 s3, s0, 0xfffff000
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v2, s3
@@ -984,15 +983,15 @@ define amdgpu_kernel void @s_exp10_v3f32(ptr addrspace(1) %out, <3 x float> %in)
 ; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s0, v3
 ; VI-GISEL-NEXT:    s_cselect_b32 s3, 0, s3
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s0, 0x7f800000, s3
-; VI-GISEL-NEXT:    s_and_b32 s3, s1, 0xfffff000
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s3
+; VI-GISEL-NEXT:    s_cselect_b32 s3, 0x7f800000, s3
+; VI-GISEL-NEXT:    s_and_b32 s0, s1, 0xfffff000
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v4, s1, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v6, 0x3a2784bc, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v4, 0x40549000, v4
-; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s3, v0
+; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s0, v0
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v4, v6
-; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s3, v1
+; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s0, v1
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v6, v4
 ; VI-GISEL-NEXT:    v_rndne_f32_e32 v6, v5
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v5, v5, v6
@@ -1003,37 +1002,39 @@ define amdgpu_kernel void @s_exp10_v3f32(ptr addrspace(1) %out, <3 x float> %in)
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
 ; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s1, v3
 ; VI-GISEL-NEXT:    v_ldexp_f32 v4, v4, v5
-; VI-GISEL-NEXT:    v_readfirstlane_b32 s3, v4
-; VI-GISEL-NEXT:    s_cselect_b32 s3, 0, s3
+; VI-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
+; VI-GISEL-NEXT:    s_cselect_b32 s0, 0, s0
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s1, 0x7f800000, s3
-; VI-GISEL-NEXT:    s_and_b32 s3, s2, 0xfffff000
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s3
+; VI-GISEL-NEXT:    s_cselect_b32 s6, 0x7f800000, s0
+; VI-GISEL-NEXT:    s_and_b32 s0, s2, 0xfffff000
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v4, s2, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v5, 0x3a2784bc, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v4, 0x40549000, v4
-; VI-GISEL-NEXT:    v_mul_f32_e32 v0, s3, v0
+; VI-GISEL-NEXT:    v_mul_f32_e32 v0, s0, v0
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v4, v5
-; VI-GISEL-NEXT:    v_mul_f32_e32 v1, s3, v1
+; VI-GISEL-NEXT:    v_mul_f32_e32 v1, s0, v1
 ; VI-GISEL-NEXT:    v_add_f32_e32 v1, v1, v4
 ; VI-GISEL-NEXT:    v_rndne_f32_e32 v4, v0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v0, v0, v4
 ; VI-GISEL-NEXT:    v_add_f32_e32 v0, v0, v1
 ; VI-GISEL-NEXT:    v_cvt_i32_f32_e32 v1, v4
 ; VI-GISEL-NEXT:    v_exp_f32_e32 v0, v0
+; VI-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; VI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s2, v2
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s2, v3
 ; VI-GISEL-NEXT:    v_ldexp_f32 v0, v0, v1
-; VI-GISEL-NEXT:    v_readfirstlane_b32 s3, v0
-; VI-GISEL-NEXT:    s_cselect_b32 s3, 0, s3
+; VI-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s2, v3
+; VI-GISEL-NEXT:    s_cselect_b32 s4, 0, s4
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s3
-; VI-GISEL-NEXT:    v_mov_b32_e32 v0, s0
-; VI-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; VI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s4
+; VI-GISEL-NEXT:    v_mov_b32_e32 v0, s3
+; VI-GISEL-NEXT:    v_mov_b32_e32 v1, s6
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v2, s2
-; VI-GISEL-NEXT:    v_mov_b32_e32 v3, s4
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s5
+; VI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s1
+; VI-GISEL-NEXT:    v_mov_b32_e32 v3, s0
 ; VI-GISEL-NEXT:    flat_store_dwordx3 v[3:4], v[0:2]
 ; VI-GISEL-NEXT:    s_endpgm
 ;
@@ -1211,69 +1212,69 @@ define amdgpu_kernel void @s_exp10_v3f32(ptr addrspace(1) %out, <3 x float> %in)
 ;
 ; SI-GISEL-LABEL: s_exp10_v3f32:
 ; SI-GISEL:       ; %bb.0:
-; SI-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0xd
+; SI-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0xd
+; SI-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v0, 0x40549a78
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v1, 0x33979a37
-; SI-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x9
 ; SI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-GISEL-NEXT:    v_mul_f32_e32 v2, s0, v0
-; SI-GISEL-NEXT:    v_fma_f32 v3, s0, v0, -v2
+; SI-GISEL-NEXT:    v_mul_f32_e32 v2, s8, v0
+; SI-GISEL-NEXT:    v_fma_f32 v3, s8, v0, -v2
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v4, v2
-; SI-GISEL-NEXT:    v_fma_f32 v3, s0, v1, v3
+; SI-GISEL-NEXT:    v_fma_f32 v3, s8, v1, v3
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v2, v2, v4
 ; SI-GISEL-NEXT:    v_add_f32_e32 v2, v2, v3
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v3, v4
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v2, v2
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v4, 0xc23369f4
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s0, v4
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s8, v4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v2, v2, v3
-; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s1, v0
-; SI-GISEL-NEXT:    v_fma_f32 v5, s1, v0, -v3
+; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s9, v0
+; SI-GISEL-NEXT:    v_fma_f32 v5, s9, v0, -v3
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v6, v3
-; SI-GISEL-NEXT:    v_fma_f32 v5, s1, v1, v5
+; SI-GISEL-NEXT:    v_fma_f32 v5, s9, v1, v5
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v3, v3, v6
 ; SI-GISEL-NEXT:    v_add_f32_e32 v3, v3, v5
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v5, v6
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v3, v3
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s3, v2
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v2, 0x421a209b
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s0, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s3, 0, s3
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s8, v2
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v3, v3, v5
-; SI-GISEL-NEXT:    s_cselect_b32 s0, 0x7f800000, s3
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s3, v3
-; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s2, v0
-; SI-GISEL-NEXT:    v_fma_f32 v0, s2, v0, -v3
-; SI-GISEL-NEXT:    v_fma_f32 v0, s2, v1, v0
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s10, v0
+; SI-GISEL-NEXT:    v_fma_f32 v0, s10, v0, -v3
+; SI-GISEL-NEXT:    v_fma_f32 v0, s10, v1, v0
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v1, v3
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v3, v3, v1
 ; SI-GISEL-NEXT:    v_add_f32_e32 v0, v3, v0
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v1, v1
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v0, v0
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s1, v4
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s1, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s3, 0, s3
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    s_cselect_b32 s4, 0, s4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s9, v4
+; SI-GISEL-NEXT:    s_cselect_b32 s4, 0x7f800000, s4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s9, v2
+; SI-GISEL-NEXT:    s_cselect_b32 s5, 0, s5
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v0, v0, v1
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s2, v4
-; SI-GISEL-NEXT:    s_cselect_b32 s1, 0x7f800000, s3
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s3, v0
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s2, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s6, 0, s3
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s10, v4
+; SI-GISEL-NEXT:    s_cselect_b32 s5, 0x7f800000, s5
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s6, v0
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s10, v2
+; SI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
 ; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
-; SI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s6
-; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s0
-; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s1
-; SI-GISEL-NEXT:    s_mov_b32 s6, -1
-; SI-GISEL-NEXT:    s_mov_b32 s7, 0xf000
-; SI-GISEL-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-GISEL-NEXT:    s_cselect_b32 s6, 0x7f800000, s6
+; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; SI-GISEL-NEXT:    s_mov_b32 s2, -1
+; SI-GISEL-NEXT:    s_mov_b32 s3, 0xf000
+; SI-GISEL-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; SI-GISEL-NEXT:    s_waitcnt expcnt(0)
-; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; SI-GISEL-NEXT:    buffer_store_dword v0, off, s[4:7], 0 offset:8
+; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; SI-GISEL-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:8
 ; SI-GISEL-NEXT:    s_endpgm
 ;
 ; R600-LABEL: s_exp10_v3f32:
@@ -1747,7 +1748,6 @@ define amdgpu_kernel void @s_exp10_v4f32(ptr addrspace(1) %out, <4 x float> %in)
 ; VI-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x34
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v0, 0x40549000
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3a2784bc
-; VI-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x24
 ; VI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-GISEL-NEXT:    s_and_b32 s6, s0, 0xfffff000
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v2, s6
@@ -1772,15 +1772,15 @@ define amdgpu_kernel void @s_exp10_v4f32(ptr addrspace(1) %out, <4 x float> %in)
 ; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s0, v3
 ; VI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s0, 0x7f800000, s6
-; VI-GISEL-NEXT:    s_and_b32 s6, s1, 0xfffff000
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s6
+; VI-GISEL-NEXT:    s_cselect_b32 s6, 0x7f800000, s6
+; VI-GISEL-NEXT:    s_and_b32 s0, s1, 0xfffff000
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v4, s1, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v6, 0x3a2784bc, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v4, 0x40549000, v4
-; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s6, v0
+; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s0, v0
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v4, v6
-; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s6, v1
+; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s0, v1
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v6, v4
 ; VI-GISEL-NEXT:    v_rndne_f32_e32 v6, v5
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v5, v5, v6
@@ -1791,18 +1791,18 @@ define amdgpu_kernel void @s_exp10_v4f32(ptr addrspace(1) %out, <4 x float> %in)
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
 ; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s1, v3
 ; VI-GISEL-NEXT:    v_ldexp_f32 v4, v4, v5
-; VI-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
-; VI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
+; VI-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
+; VI-GISEL-NEXT:    s_cselect_b32 s0, 0, s0
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s1, 0x7f800000, s6
-; VI-GISEL-NEXT:    s_and_b32 s6, s2, 0xfffff000
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s6
+; VI-GISEL-NEXT:    s_cselect_b32 s7, 0x7f800000, s0
+; VI-GISEL-NEXT:    s_and_b32 s0, s2, 0xfffff000
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v4, s2, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v6, 0x3a2784bc, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v4, 0x40549000, v4
-; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s6, v0
+; VI-GISEL-NEXT:    v_mul_f32_e32 v5, s0, v0
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v4, v6
-; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s6, v1
+; VI-GISEL-NEXT:    v_mul_f32_e32 v6, s0, v1
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v6, v4
 ; VI-GISEL-NEXT:    v_rndne_f32_e32 v6, v5
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v5, v5, v6
@@ -1813,38 +1813,40 @@ define amdgpu_kernel void @s_exp10_v4f32(ptr addrspace(1) %out, <4 x float> %in)
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
 ; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s2, v3
 ; VI-GISEL-NEXT:    v_ldexp_f32 v4, v4, v5
-; VI-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
-; VI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
+; VI-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
+; VI-GISEL-NEXT:    s_cselect_b32 s0, 0, s0
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s6
-; VI-GISEL-NEXT:    s_and_b32 s6, s3, 0xfffff000
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s6
+; VI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s0
+; VI-GISEL-NEXT:    s_and_b32 s0, s3, 0xfffff000
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v4, s3, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v5, 0x3a2784bc, v4
 ; VI-GISEL-NEXT:    v_mul_f32_e32 v4, 0x40549000, v4
-; VI-GISEL-NEXT:    v_mul_f32_e32 v0, s6, v0
+; VI-GISEL-NEXT:    v_mul_f32_e32 v0, s0, v0
 ; VI-GISEL-NEXT:    v_add_f32_e32 v4, v4, v5
-; VI-GISEL-NEXT:    v_mul_f32_e32 v1, s6, v1
+; VI-GISEL-NEXT:    v_mul_f32_e32 v1, s0, v1
 ; VI-GISEL-NEXT:    v_add_f32_e32 v1, v1, v4
 ; VI-GISEL-NEXT:    v_rndne_f32_e32 v4, v0
 ; VI-GISEL-NEXT:    v_sub_f32_e32 v0, v0, v4
 ; VI-GISEL-NEXT:    v_add_f32_e32 v0, v0, v1
 ; VI-GISEL-NEXT:    v_cvt_i32_f32_e32 v1, v4
 ; VI-GISEL-NEXT:    v_exp_f32_e32 v0, v0
+; VI-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
 ; VI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s3, v2
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s3, v3
 ; VI-GISEL-NEXT:    v_ldexp_f32 v0, v0, v1
-; VI-GISEL-NEXT:    v_readfirstlane_b32 s6, v0
-; VI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
+; VI-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; VI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s3, v3
+; VI-GISEL-NEXT:    s_cselect_b32 s4, 0, s4
 ; VI-GISEL-NEXT:    s_cmp_lg_u64 vcc, 0
-; VI-GISEL-NEXT:    s_cselect_b32 s3, 0x7f800000, s6
-; VI-GISEL-NEXT:    v_mov_b32_e32 v0, s0
-; VI-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; VI-GISEL-NEXT:    s_cselect_b32 s3, 0x7f800000, s4
+; VI-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; VI-GISEL-NEXT:    v_mov_b32_e32 v1, s7
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v3, s3
-; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s4
-; VI-GISEL-NEXT:    v_mov_b32_e32 v5, s5
+; VI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-GISEL-NEXT:    v_mov_b32_e32 v5, s1
+; VI-GISEL-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-GISEL-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-GISEL-NEXT:    s_endpgm
 ;
@@ -2065,84 +2067,84 @@ define amdgpu_kernel void @s_exp10_v4f32(ptr addrspace(1) %out, <4 x float> %in)
 ;
 ; SI-GISEL-LABEL: s_exp10_v4f32:
 ; SI-GISEL:       ; %bb.0:
-; SI-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0xd
+; SI-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0xd
+; SI-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v0, 0x40549a78
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v1, 0x33979a37
-; SI-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x9
 ; SI-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-GISEL-NEXT:    v_mul_f32_e32 v2, s0, v0
-; SI-GISEL-NEXT:    v_fma_f32 v3, s0, v0, -v2
+; SI-GISEL-NEXT:    v_mul_f32_e32 v2, s8, v0
+; SI-GISEL-NEXT:    v_fma_f32 v3, s8, v0, -v2
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v4, v2
-; SI-GISEL-NEXT:    v_fma_f32 v3, s0, v1, v3
+; SI-GISEL-NEXT:    v_fma_f32 v3, s8, v1, v3
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v2, v2, v4
 ; SI-GISEL-NEXT:    v_add_f32_e32 v2, v2, v3
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v3, v4
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v2, v2
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v4, 0xc23369f4
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s0, v4
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s8, v4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v2, v2, v3
-; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s1, v0
-; SI-GISEL-NEXT:    v_fma_f32 v5, s1, v0, -v3
+; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s9, v0
+; SI-GISEL-NEXT:    v_fma_f32 v5, s9, v0, -v3
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v6, v3
-; SI-GISEL-NEXT:    v_fma_f32 v5, s1, v1, v5
+; SI-GISEL-NEXT:    v_fma_f32 v5, s9, v1, v5
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v3, v3, v6
 ; SI-GISEL-NEXT:    v_add_f32_e32 v3, v3, v5
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v5, v6
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v3, v3
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s8, v2
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v2, 0x421a209b
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s0, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s8, 0, s8
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s8, v2
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v3, v3, v5
-; SI-GISEL-NEXT:    s_cselect_b32 s0, 0x7f800000, s8
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s8, v3
-; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s2, v0
-; SI-GISEL-NEXT:    v_fma_f32 v5, s2, v0, -v3
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s10, v0
+; SI-GISEL-NEXT:    v_fma_f32 v5, s10, v0, -v3
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v6, v3
-; SI-GISEL-NEXT:    v_fma_f32 v5, s2, v1, v5
+; SI-GISEL-NEXT:    v_fma_f32 v5, s10, v1, v5
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v3, v3, v6
 ; SI-GISEL-NEXT:    v_add_f32_e32 v3, v3, v5
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v5, v6
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v3, v3
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s1, v4
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s1, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s8, 0, s8
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    s_cselect_b32 s4, 0, s4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s9, v4
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v3, v3, v5
-; SI-GISEL-NEXT:    s_cselect_b32 s1, 0x7f800000, s8
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s8, v3
-; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s3, v0
-; SI-GISEL-NEXT:    v_fma_f32 v0, s3, v0, -v3
-; SI-GISEL-NEXT:    v_fma_f32 v0, s3, v1, v0
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; SI-GISEL-NEXT:    v_mul_f32_e32 v3, s11, v0
+; SI-GISEL-NEXT:    v_fma_f32 v0, s11, v0, -v3
+; SI-GISEL-NEXT:    v_fma_f32 v0, s11, v1, v0
 ; SI-GISEL-NEXT:    v_rndne_f32_e32 v1, v3
 ; SI-GISEL-NEXT:    v_sub_f32_e32 v3, v3, v1
 ; SI-GISEL-NEXT:    v_add_f32_e32 v0, v3, v0
 ; SI-GISEL-NEXT:    v_cvt_i32_f32_e32 v1, v1
 ; SI-GISEL-NEXT:    v_exp_f32_e32 v0, v0
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s2, v4
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s2, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s8, 0, s8
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
+; SI-GISEL-NEXT:    s_cselect_b32 s4, 0x7f800000, s4
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s9, v2
+; SI-GISEL-NEXT:    s_cselect_b32 s5, 0, s5
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s10, v4
+; SI-GISEL-NEXT:    s_cselect_b32 s5, 0x7f800000, s5
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s10, v2
+; SI-GISEL-NEXT:    s_cselect_b32 s6, 0, s6
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
 ; SI-GISEL-NEXT:    v_ldexp_f32_e32 v0, v0, v1
-; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s3, v4
-; SI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s8
-; SI-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s3, v2
-; SI-GISEL-NEXT:    s_cselect_b32 s8, 0, s8
-; SI-GISEL-NEXT:    s_or_b64 s[6:7], vcc, vcc
-; SI-GISEL-NEXT:    s_cselect_b32 s3, 0x7f800000, s8
-; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s0
-; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s1
-; SI-GISEL-NEXT:    v_mov_b32_e32 v2, s2
-; SI-GISEL-NEXT:    v_mov_b32_e32 v3, s3
-; SI-GISEL-NEXT:    s_mov_b32 s6, -1
-; SI-GISEL-NEXT:    s_mov_b32 s7, 0xf000
-; SI-GISEL-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; SI-GISEL-NEXT:    v_cmp_lt_f32_e32 vcc, s11, v4
+; SI-GISEL-NEXT:    s_cselect_b32 s6, 0x7f800000, s6
+; SI-GISEL-NEXT:    v_readfirstlane_b32 s7, v0
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    v_cmp_gt_f32_e32 vcc, s11, v2
+; SI-GISEL-NEXT:    s_cselect_b32 s7, 0, s7
+; SI-GISEL-NEXT:    s_or_b64 s[2:3], vcc, vcc
+; SI-GISEL-NEXT:    s_cselect_b32 s2, 0x7f800000, s7
+; SI-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; SI-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; SI-GISEL-NEXT:    v_mov_b32_e32 v2, s6
+; SI-GISEL-NEXT:    v_mov_b32_e32 v3, s2
+; SI-GISEL-NEXT:    s_mov_b32 s2, -1
+; SI-GISEL-NEXT:    s_mov_b32 s3, 0xf000
+; SI-GISEL-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; SI-GISEL-NEXT:    s_endpgm
 ;
 ; R600-LABEL: s_exp10_v4f32:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.maximum.f64.ll b/llvm/test/CodeGen/AMDGPU/llvm.maximum.f64.ll
index 3ed09b1589fb0..73730b1dda4f6 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.maximum.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.maximum.f64.ll
@@ -456,8 +456,8 @@ define void @s_maximum_f64(double inreg %src0, double inreg %src1) #0 {
 ; GFX7-LABEL: s_maximum_f64:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX7-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s19
+; GFX7-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX7-NEXT:    v_max_f64 v[2:3], s[16:17], v[0:1]
 ; GFX7-NEXT:    v_cmp_u_f64_e32 vcc, s[16:17], v[0:1]
 ; GFX7-NEXT:    v_mov_b32_e32 v4, 0x7ff80000
@@ -471,8 +471,8 @@ define void @s_maximum_f64(double inreg %src0, double inreg %src1) #0 {
 ; GFX8-LABEL: s_maximum_f64:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s19
+; GFX8-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX8-NEXT:    v_max_f64 v[2:3], s[16:17], v[0:1]
 ; GFX8-NEXT:    v_cmp_u_f64_e32 vcc, s[16:17], v[0:1]
 ; GFX8-NEXT:    v_mov_b32_e32 v4, 0x7ff80000
@@ -486,8 +486,8 @@ define void @s_maximum_f64(double inreg %src0, double inreg %src1) #0 {
 ; GFX900-LABEL: s_maximum_f64:
 ; GFX900:       ; %bb.0:
 ; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX900-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX900-NEXT:    v_mov_b32_e32 v1, s19
+; GFX900-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX900-NEXT:    v_max_f64 v[2:3], s[16:17], v[0:1]
 ; GFX900-NEXT:    v_cmp_u_f64_e32 vcc, s[16:17], v[0:1]
 ; GFX900-NEXT:    v_mov_b32_e32 v4, 0x7ff80000
@@ -896,12 +896,12 @@ define void @s_maximum_v2f64(<2 x double> inreg %src0, <2 x double> inreg %src1)
 ; GFX7-LABEL: s_maximum_v2f64:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX7-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s23
+; GFX7-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-NEXT:    v_max_f64 v[2:3], s[18:19], v[0:1]
 ; GFX7-NEXT:    v_cmp_u_f64_e32 vcc, s[18:19], v[0:1]
-; GFX7-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s21
+; GFX7-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX7-NEXT:    v_max_f64 v[4:5], s[16:17], v[0:1]
 ; GFX7-NEXT:    v_cmp_u_f64_e64 s[4:5], s[16:17], v[0:1]
 ; GFX7-NEXT:    v_mov_b32_e32 v6, 0x7ff80000
@@ -917,12 +917,12 @@ define void @s_maximum_v2f64(<2 x double> inreg %src0, <2 x double> inreg %src1)
 ; GFX8-LABEL: s_maximum_v2f64:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s23
+; GFX8-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX8-NEXT:    v_max_f64 v[2:3], s[18:19], v[0:1]
 ; GFX8-NEXT:    v_cmp_u_f64_e32 vcc, s[18:19], v[0:1]
-; GFX8-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s21
+; GFX8-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX8-NEXT:    v_max_f64 v[4:5], s[16:17], v[0:1]
 ; GFX8-NEXT:    v_cmp_u_f64_e64 s[4:5], s[16:17], v[0:1]
 ; GFX8-NEXT:    v_mov_b32_e32 v6, 0x7ff80000
@@ -938,12 +938,12 @@ define void @s_maximum_v2f64(<2 x double> inreg %src0, <2 x double> inreg %src1)
 ; GFX900-LABEL: s_maximum_v2f64:
 ; GFX900:       ; %bb.0:
 ; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX900-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX900-NEXT:    v_mov_b32_e32 v1, s23
+; GFX900-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX900-NEXT:    v_max_f64 v[2:3], s[18:19], v[0:1]
 ; GFX900-NEXT:    v_cmp_u_f64_e32 vcc, s[18:19], v[0:1]
-; GFX900-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX900-NEXT:    v_mov_b32_e32 v1, s21
+; GFX900-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX900-NEXT:    v_max_f64 v[4:5], s[16:17], v[0:1]
 ; GFX900-NEXT:    v_cmp_u_f64_e64 s[4:5], s[16:17], v[0:1]
 ; GFX900-NEXT:    v_mov_b32_e32 v6, 0x7ff80000
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.minimum.f64.ll b/llvm/test/CodeGen/AMDGPU/llvm.minimum.f64.ll
index 2f18ad05ad3e2..7e517d9caac90 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.minimum.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.minimum.f64.ll
@@ -456,8 +456,8 @@ define void @s_minimum_f64(double inreg %src0, double inreg %src1) #0 {
 ; GFX7-LABEL: s_minimum_f64:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX7-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s19
+; GFX7-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX7-NEXT:    v_min_f64 v[2:3], s[16:17], v[0:1]
 ; GFX7-NEXT:    v_cmp_u_f64_e32 vcc, s[16:17], v[0:1]
 ; GFX7-NEXT:    v_mov_b32_e32 v4, 0x7ff80000
@@ -471,8 +471,8 @@ define void @s_minimum_f64(double inreg %src0, double inreg %src1) #0 {
 ; GFX8-LABEL: s_minimum_f64:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s19
+; GFX8-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX8-NEXT:    v_min_f64 v[2:3], s[16:17], v[0:1]
 ; GFX8-NEXT:    v_cmp_u_f64_e32 vcc, s[16:17], v[0:1]
 ; GFX8-NEXT:    v_mov_b32_e32 v4, 0x7ff80000
@@ -486,8 +486,8 @@ define void @s_minimum_f64(double inreg %src0, double inreg %src1) #0 {
 ; GFX900-LABEL: s_minimum_f64:
 ; GFX900:       ; %bb.0:
 ; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX900-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX900-NEXT:    v_mov_b32_e32 v1, s19
+; GFX900-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX900-NEXT:    v_min_f64 v[2:3], s[16:17], v[0:1]
 ; GFX900-NEXT:    v_cmp_u_f64_e32 vcc, s[16:17], v[0:1]
 ; GFX900-NEXT:    v_mov_b32_e32 v4, 0x7ff80000
@@ -896,12 +896,12 @@ define void @s_minimum_v2f64(<2 x double> inreg %src0, <2 x double> inreg %src1)
 ; GFX7-LABEL: s_minimum_v2f64:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX7-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s23
+; GFX7-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-NEXT:    v_min_f64 v[2:3], s[18:19], v[0:1]
 ; GFX7-NEXT:    v_cmp_u_f64_e32 vcc, s[18:19], v[0:1]
-; GFX7-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s21
+; GFX7-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX7-NEXT:    v_min_f64 v[4:5], s[16:17], v[0:1]
 ; GFX7-NEXT:    v_cmp_u_f64_e64 s[4:5], s[16:17], v[0:1]
 ; GFX7-NEXT:    v_mov_b32_e32 v6, 0x7ff80000
@@ -917,12 +917,12 @@ define void @s_minimum_v2f64(<2 x double> inreg %src0, <2 x double> inreg %src1)
 ; GFX8-LABEL: s_minimum_v2f64:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s23
+; GFX8-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX8-NEXT:    v_min_f64 v[2:3], s[18:19], v[0:1]
 ; GFX8-NEXT:    v_cmp_u_f64_e32 vcc, s[18:19], v[0:1]
-; GFX8-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s21
+; GFX8-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX8-NEXT:    v_min_f64 v[4:5], s[16:17], v[0:1]
 ; GFX8-NEXT:    v_cmp_u_f64_e64 s[4:5], s[16:17], v[0:1]
 ; GFX8-NEXT:    v_mov_b32_e32 v6, 0x7ff80000
@@ -938,12 +938,12 @@ define void @s_minimum_v2f64(<2 x double> inreg %src0, <2 x double> inreg %src1)
 ; GFX900-LABEL: s_minimum_v2f64:
 ; GFX900:       ; %bb.0:
 ; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX900-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX900-NEXT:    v_mov_b32_e32 v1, s23
+; GFX900-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX900-NEXT:    v_min_f64 v[2:3], s[18:19], v[0:1]
 ; GFX900-NEXT:    v_cmp_u_f64_e32 vcc, s[18:19], v[0:1]
-; GFX900-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX900-NEXT:    v_mov_b32_e32 v1, s21
+; GFX900-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX900-NEXT:    v_min_f64 v[4:5], s[16:17], v[0:1]
 ; GFX900-NEXT:    v_cmp_u_f64_e64 s[4:5], s[16:17], v[0:1]
 ; GFX900-NEXT:    v_mov_b32_e32 v6, 0x7ff80000
diff --git a/llvm/test/CodeGen/AMDGPU/load-constant-i1.ll b/llvm/test/CodeGen/AMDGPU/load-constant-i1.ll
index 4a03b6522bb71..b737e7ca78f0f 100644
--- a/llvm/test/CodeGen/AMDGPU/load-constant-i1.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-constant-i1.ll
@@ -1762,8 +1762,8 @@ define amdgpu_kernel void @constant_zextload_v8i1_to_v8i32(ptr addrspace(1) %out
 ; GFX8-NEXT:    v_bfe_u32 v2, v0, 6, 1
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NEXT:    v_mov_b32_e32 v11, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v10, s0
+; GFX8-NEXT:    v_mov_b32_e32 v11, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v6, s7
@@ -1955,11 +1955,11 @@ define amdgpu_kernel void @constant_sextload_v8i1_to_v8i32(ptr addrspace(1) %out
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    flat_load_ubyte v4, v[0:1]
 ; GFX8-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v11, s3
-; GFX8-NEXT:    v_mov_b32_e32 v10, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v9, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v8, s0
+; GFX8-NEXT:    s_addc_u32 s0, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v10, s2
+; GFX8-NEXT:    v_mov_b32_e32 v11, s0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    v_bfe_i32 v3, v4, 3, 1
 ; GFX8-NEXT:    v_bfe_i32 v2, v4, 2, 1
@@ -2138,9 +2138,9 @@ define amdgpu_kernel void @constant_zextload_v16i1_to_v16i32(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_add_u32 s2, s0, 48
 ; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v19, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v18, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 32
+; GFX8-NEXT:    v_mov_b32_e32 v19, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s8
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s6
@@ -2392,18 +2392,18 @@ define amdgpu_kernel void @constant_sextload_v16i1_to_v16i32(ptr addrspace(1) %o
 ; GFX8-NEXT:    flat_load_ushort v18, v[0:1]
 ; GFX8-NEXT:    s_add_u32 s2, s0, 48
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v9, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v8, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 32
+; GFX8-NEXT:    v_mov_b32_e32 v9, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v13, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v12, s0
 ; GFX8-NEXT:    s_add_u32 s0, s0, 16
-; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v14, s2
+; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v17, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v16, s0
+; GFX8-NEXT:    v_mov_b32_e32 v17, s1
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    v_bfe_i32 v7, v18, 15, 1
 ; GFX8-NEXT:    v_bfe_i32 v6, v18, 14, 1
@@ -2673,20 +2673,20 @@ define amdgpu_kernel void @constant_zextload_v32i1_to_v32i32(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_load_dword s2, s[2:3], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_bfe_u32 s4, s2, 0x10003
-; GFX8-NEXT:    s_bfe_u32 s5, s2, 0x10001
-; GFX8-NEXT:    s_bfe_u32 s6, s2, 0x10007
-; GFX8-NEXT:    s_bfe_u32 s7, s2, 0x10005
-; GFX8-NEXT:    s_bfe_u32 s8, s2, 0x1000b
-; GFX8-NEXT:    s_bfe_u32 s9, s2, 0x10009
-; GFX8-NEXT:    s_bfe_u32 s10, s2, 0x1000f
-; GFX8-NEXT:    s_bfe_u32 s11, s2, 0x1000d
-; GFX8-NEXT:    s_bfe_u32 s12, s2, 0x10013
-; GFX8-NEXT:    s_bfe_u32 s13, s2, 0x10011
-; GFX8-NEXT:    s_bfe_u32 s14, s2, 0x10017
-; GFX8-NEXT:    s_bfe_u32 s15, s2, 0x1001b
-; GFX8-NEXT:    s_bfe_u32 s16, s2, 0x10019
-; GFX8-NEXT:    s_lshr_b32 s3, s2, 31
+; GFX8-NEXT:    s_bfe_u32 s3, s2, 0x10003
+; GFX8-NEXT:    s_bfe_u32 s4, s2, 0x10001
+; GFX8-NEXT:    s_bfe_u32 s5, s2, 0x10007
+; GFX8-NEXT:    s_bfe_u32 s6, s2, 0x10005
+; GFX8-NEXT:    s_bfe_u32 s7, s2, 0x1000b
+; GFX8-NEXT:    s_bfe_u32 s8, s2, 0x10009
+; GFX8-NEXT:    s_bfe_u32 s9, s2, 0x1000f
+; GFX8-NEXT:    s_bfe_u32 s10, s2, 0x1000d
+; GFX8-NEXT:    s_bfe_u32 s11, s2, 0x10013
+; GFX8-NEXT:    s_bfe_u32 s12, s2, 0x10011
+; GFX8-NEXT:    s_bfe_u32 s13, s2, 0x10017
+; GFX8-NEXT:    s_bfe_u32 s14, s2, 0x1001b
+; GFX8-NEXT:    s_bfe_u32 s15, s2, 0x10019
+; GFX8-NEXT:    s_lshr_b32 s16, s2, 31
 ; GFX8-NEXT:    s_bfe_u32 s17, s2, 0x1001d
 ; GFX8-NEXT:    s_and_b32 s18, s2, 1
 ; GFX8-NEXT:    s_bfe_u32 s19, s2, 0x10002
@@ -2707,72 +2707,72 @@ define amdgpu_kernel void @constant_zextload_v32i1_to_v32i32(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x1001c
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 0x70
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 0x60
+; GFX8-NEXT:    v_mov_b32_e32 v3, s16
+; GFX8-NEXT:    s_addc_u32 s16, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s17
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s34
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v5, s16
+; GFX8-NEXT:    s_add_u32 s2, s0, 0x60
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 0x50
+; GFX8-NEXT:    v_mov_b32_e32 v3, s14
+; GFX8-NEXT:    s_addc_u32 s14, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s33
-; GFX8-NEXT:    v_mov_b32_e32 v1, s16
+; GFX8-NEXT:    v_mov_b32_e32 v1, s15
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s31
-; GFX8-NEXT:    v_mov_b32_e32 v3, s15
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s14
+; GFX8-NEXT:    s_add_u32 s2, s0, 0x50
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 64
+; GFX8-NEXT:    v_mov_b32_e32 v3, s13
+; GFX8-NEXT:    s_addc_u32 s13, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s30
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s29
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s28
-; GFX8-NEXT:    v_mov_b32_e32 v3, s14
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s13
+; GFX8-NEXT:    s_add_u32 s2, s0, 64
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 48
+; GFX8-NEXT:    v_mov_b32_e32 v3, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s27
-; GFX8-NEXT:    v_mov_b32_e32 v1, s13
+; GFX8-NEXT:    v_mov_b32_e32 v1, s12
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s26
-; GFX8-NEXT:    v_mov_b32_e32 v3, s12
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_add_u32 s2, s0, 48
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 32
+; GFX8-NEXT:    v_mov_b32_e32 v3, s9
+; GFX8-NEXT:    s_addc_u32 s9, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s25
-; GFX8-NEXT:    v_mov_b32_e32 v1, s11
+; GFX8-NEXT:    v_mov_b32_e32 v1, s10
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s24
-; GFX8-NEXT:    v_mov_b32_e32 v3, s10
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s9
+; GFX8-NEXT:    s_add_u32 s2, s0, 32
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 16
+; GFX8-NEXT:    v_mov_b32_e32 v3, s7
+; GFX8-NEXT:    s_addc_u32 s7, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s23
-; GFX8-NEXT:    v_mov_b32_e32 v1, s9
+; GFX8-NEXT:    v_mov_b32_e32 v1, s8
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s22
-; GFX8-NEXT:    v_mov_b32_e32 v3, s8
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
+; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s21
-; GFX8-NEXT:    v_mov_b32_e32 v1, s7
+; GFX8-NEXT:    v_mov_b32_e32 v1, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s20
-; GFX8-NEXT:    v_mov_b32_e32 v3, s6
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s18
-; GFX8-NEXT:    v_mov_b32_e32 v1, s5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s19
-; GFX8-NEXT:    v_mov_b32_e32 v3, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_endpgm
@@ -3098,106 +3098,106 @@ define amdgpu_kernel void @constant_sextload_v32i1_to_v32i32(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_load_dword s2, s[2:3], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_bfe_i32 s4, s2, 0x10003
-; GFX8-NEXT:    s_bfe_i32 s5, s2, 0x10002
-; GFX8-NEXT:    s_bfe_i32 s6, s2, 0x10001
-; GFX8-NEXT:    s_bfe_i32 s7, s2, 0x10000
-; GFX8-NEXT:    s_bfe_i32 s8, s2, 0x10007
-; GFX8-NEXT:    s_bfe_i32 s9, s2, 0x10006
-; GFX8-NEXT:    s_bfe_i32 s10, s2, 0x10005
-; GFX8-NEXT:    s_bfe_i32 s11, s2, 0x10004
-; GFX8-NEXT:    s_bfe_i32 s12, s2, 0x1000b
-; GFX8-NEXT:    s_bfe_i32 s13, s2, 0x1000a
-; GFX8-NEXT:    s_bfe_i32 s14, s2, 0x10009
-; GFX8-NEXT:    s_bfe_i32 s15, s2, 0x10008
-; GFX8-NEXT:    s_bfe_i32 s16, s2, 0x1000f
-; GFX8-NEXT:    s_bfe_i32 s17, s2, 0x1000e
-; GFX8-NEXT:    s_bfe_i32 s18, s2, 0x1000d
-; GFX8-NEXT:    s_bfe_i32 s19, s2, 0x1000c
-; GFX8-NEXT:    s_bfe_i32 s20, s2, 0x10013
-; GFX8-NEXT:    s_bfe_i32 s21, s2, 0x10012
-; GFX8-NEXT:    s_bfe_i32 s22, s2, 0x10011
-; GFX8-NEXT:    s_bfe_i32 s23, s2, 0x10010
-; GFX8-NEXT:    s_bfe_i32 s24, s2, 0x10017
-; GFX8-NEXT:    s_bfe_i32 s25, s2, 0x10016
-; GFX8-NEXT:    s_bfe_i32 s26, s2, 0x10015
-; GFX8-NEXT:    s_bfe_i32 s27, s2, 0x10014
-; GFX8-NEXT:    s_bfe_i32 s28, s2, 0x1001b
-; GFX8-NEXT:    s_bfe_i32 s29, s2, 0x1001a
-; GFX8-NEXT:    s_bfe_i32 s30, s2, 0x10019
-; GFX8-NEXT:    s_bfe_i32 s31, s2, 0x10018
-; GFX8-NEXT:    s_ashr_i32 s3, s2, 31
+; GFX8-NEXT:    s_bfe_i32 s3, s2, 0x10003
+; GFX8-NEXT:    s_bfe_i32 s4, s2, 0x10002
+; GFX8-NEXT:    s_bfe_i32 s5, s2, 0x10001
+; GFX8-NEXT:    s_bfe_i32 s6, s2, 0x10000
+; GFX8-NEXT:    s_bfe_i32 s7, s2, 0x10007
+; GFX8-NEXT:    s_bfe_i32 s8, s2, 0x10006
+; GFX8-NEXT:    s_bfe_i32 s9, s2, 0x10005
+; GFX8-NEXT:    s_bfe_i32 s10, s2, 0x10004
+; GFX8-NEXT:    s_bfe_i32 s11, s2, 0x1000b
+; GFX8-NEXT:    s_bfe_i32 s12, s2, 0x1000a
+; GFX8-NEXT:    s_bfe_i32 s13, s2, 0x10009
+; GFX8-NEXT:    s_bfe_i32 s14, s2, 0x10008
+; GFX8-NEXT:    s_bfe_i32 s15, s2, 0x1000f
+; GFX8-NEXT:    s_bfe_i32 s16, s2, 0x1000e
+; GFX8-NEXT:    s_bfe_i32 s17, s2, 0x1000d
+; GFX8-NEXT:    s_bfe_i32 s18, s2, 0x1000c
+; GFX8-NEXT:    s_bfe_i32 s19, s2, 0x10013
+; GFX8-NEXT:    s_bfe_i32 s20, s2, 0x10012
+; GFX8-NEXT:    s_bfe_i32 s21, s2, 0x10011
+; GFX8-NEXT:    s_bfe_i32 s22, s2, 0x10010
+; GFX8-NEXT:    s_bfe_i32 s23, s2, 0x10017
+; GFX8-NEXT:    s_bfe_i32 s24, s2, 0x10016
+; GFX8-NEXT:    s_bfe_i32 s25, s2, 0x10015
+; GFX8-NEXT:    s_bfe_i32 s26, s2, 0x10014
+; GFX8-NEXT:    s_bfe_i32 s27, s2, 0x1001b
+; GFX8-NEXT:    s_bfe_i32 s28, s2, 0x1001a
+; GFX8-NEXT:    s_bfe_i32 s29, s2, 0x10019
+; GFX8-NEXT:    s_bfe_i32 s30, s2, 0x10018
+; GFX8-NEXT:    s_ashr_i32 s31, s2, 31
 ; GFX8-NEXT:    s_bfe_i32 s33, s2, 0x1001e
 ; GFX8-NEXT:    s_bfe_i32 s34, s2, 0x1001d
 ; GFX8-NEXT:    s_bfe_i32 s2, s2, 0x1001c
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 0x70
-; GFX8-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NEXT:    s_add_u32 s2, s0, 0x60
+; GFX8-NEXT:    v_mov_b32_e32 v3, s31
+; GFX8-NEXT:    s_addc_u32 s31, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s34
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s33
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v5, s31
+; GFX8-NEXT:    s_add_u32 s2, s0, 0x60
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s27
+; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s30
+; GFX8-NEXT:    v_mov_b32_e32 v1, s29
+; GFX8-NEXT:    v_mov_b32_e32 v2, s28
+; GFX8-NEXT:    v_mov_b32_e32 v5, s27
 ; GFX8-NEXT:    s_add_u32 s2, s0, 0x50
-; GFX8-NEXT:    v_mov_b32_e32 v0, s31
-; GFX8-NEXT:    v_mov_b32_e32 v1, s30
-; GFX8-NEXT:    v_mov_b32_e32 v2, s29
-; GFX8-NEXT:    v_mov_b32_e32 v3, s28
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s23
+; GFX8-NEXT:    s_addc_u32 s23, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s26
+; GFX8-NEXT:    v_mov_b32_e32 v1, s25
+; GFX8-NEXT:    v_mov_b32_e32 v2, s24
+; GFX8-NEXT:    v_mov_b32_e32 v5, s23
 ; GFX8-NEXT:    s_add_u32 s2, s0, 64
-; GFX8-NEXT:    v_mov_b32_e32 v0, s27
-; GFX8-NEXT:    v_mov_b32_e32 v1, s26
-; GFX8-NEXT:    v_mov_b32_e32 v2, s25
-; GFX8-NEXT:    v_mov_b32_e32 v3, s24
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s19
+; GFX8-NEXT:    s_addc_u32 s19, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s22
+; GFX8-NEXT:    v_mov_b32_e32 v1, s21
+; GFX8-NEXT:    v_mov_b32_e32 v2, s20
+; GFX8-NEXT:    v_mov_b32_e32 v5, s19
 ; GFX8-NEXT:    s_add_u32 s2, s0, 48
-; GFX8-NEXT:    v_mov_b32_e32 v0, s23
-; GFX8-NEXT:    v_mov_b32_e32 v1, s22
-; GFX8-NEXT:    v_mov_b32_e32 v2, s21
-; GFX8-NEXT:    v_mov_b32_e32 v3, s20
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s18
+; GFX8-NEXT:    v_mov_b32_e32 v1, s17
+; GFX8-NEXT:    v_mov_b32_e32 v2, s16
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX8-NEXT:    s_add_u32 s2, s0, 32
-; GFX8-NEXT:    v_mov_b32_e32 v0, s19
-; GFX8-NEXT:    v_mov_b32_e32 v1, s18
-; GFX8-NEXT:    v_mov_b32_e32 v2, s17
-; GFX8-NEXT:    v_mov_b32_e32 v3, s16
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s14
+; GFX8-NEXT:    v_mov_b32_e32 v1, s13
+; GFX8-NEXT:    v_mov_b32_e32 v2, s12
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
 ; GFX8-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NEXT:    v_mov_b32_e32 v0, s15
-; GFX8-NEXT:    v_mov_b32_e32 v1, s14
-; GFX8-NEXT:    v_mov_b32_e32 v2, s13
-; GFX8-NEXT:    v_mov_b32_e32 v3, s12
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NEXT:    v_mov_b32_e32 v0, s11
-; GFX8-NEXT:    v_mov_b32_e32 v1, s10
-; GFX8-NEXT:    v_mov_b32_e32 v2, s9
-; GFX8-NEXT:    v_mov_b32_e32 v3, s8
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v3, s7
+; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s10
+; GFX8-NEXT:    v_mov_b32_e32 v1, s9
+; GFX8-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s1
-; GFX8-NEXT:    v_mov_b32_e32 v0, s7
-; GFX8-NEXT:    v_mov_b32_e32 v1, s6
-; GFX8-NEXT:    v_mov_b32_e32 v2, s5
-; GFX8-NEXT:    v_mov_b32_e32 v3, s4
+; GFX8-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NEXT:    v_mov_b32_e32 v1, s5
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_endpgm
@@ -3633,215 +3633,218 @@ define amdgpu_kernel void @constant_zextload_v64i1_to_v64i32(ptr addrspace(1) %o
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dwordx2 s[26:27], s[2:3], 0x0
+; GFX8-NEXT:    s_load_dwordx2 s[30:31], s[2:3], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_bfe_u32 s2, s26, 0x10003
-; GFX8-NEXT:    s_bfe_u32 s3, s26, 0x10001
-; GFX8-NEXT:    s_bfe_u32 s4, s26, 0x10007
-; GFX8-NEXT:    s_bfe_u32 s5, s26, 0x10005
-; GFX8-NEXT:    s_bfe_u32 s6, s26, 0x1000b
-; GFX8-NEXT:    s_bfe_u32 s9, s26, 0x10009
-; GFX8-NEXT:    s_bfe_u32 s11, s26, 0x1000f
-; GFX8-NEXT:    s_bfe_u32 s13, s26, 0x1000d
-; GFX8-NEXT:    s_bfe_u32 s15, s26, 0x10013
-; GFX8-NEXT:    s_bfe_u32 s17, s26, 0x10011
-; GFX8-NEXT:    s_bfe_u32 s19, s26, 0x10017
-; GFX8-NEXT:    s_bfe_u32 s21, s26, 0x1001b
-; GFX8-NEXT:    s_bfe_u32 s23, s26, 0x10019
-; GFX8-NEXT:    s_lshr_b32 s25, s26, 31
-; GFX8-NEXT:    s_bfe_u32 s28, s26, 0x1001d
-; GFX8-NEXT:    s_bfe_u32 s29, s27, 0x10003
-; GFX8-NEXT:    s_bfe_u32 s30, s27, 0x10001
-; GFX8-NEXT:    s_bfe_u32 s31, s27, 0x10007
-; GFX8-NEXT:    s_bfe_u32 s33, s27, 0x10005
-; GFX8-NEXT:    s_bfe_u32 s34, s27, 0x1000b
-; GFX8-NEXT:    s_bfe_u32 s35, s27, 0x10009
-; GFX8-NEXT:    s_bfe_u32 s36, s27, 0x1000f
-; GFX8-NEXT:    s_bfe_u32 s37, s27, 0x1000d
-; GFX8-NEXT:    s_bfe_u32 s38, s27, 0x10013
-; GFX8-NEXT:    s_bfe_u32 s39, s27, 0x10011
-; GFX8-NEXT:    s_bfe_u32 s40, s27, 0x10017
-; GFX8-NEXT:    s_bfe_u32 s41, s27, 0x1001b
-; GFX8-NEXT:    s_bfe_u32 s42, s27, 0x10019
-; GFX8-NEXT:    s_lshr_b32 s43, s27, 31
-; GFX8-NEXT:    s_bfe_u32 s44, s27, 0x1001d
-; GFX8-NEXT:    s_and_b32 s8, s26, 1
-; GFX8-NEXT:    s_bfe_u32 s7, s26, 0x10002
-; GFX8-NEXT:    s_bfe_u32 s10, s26, 0x10006
-; GFX8-NEXT:    s_bfe_u32 s12, s26, 0x10004
-; GFX8-NEXT:    s_bfe_u32 s14, s26, 0x1000a
-; GFX8-NEXT:    s_bfe_u32 s16, s26, 0x10008
-; GFX8-NEXT:    s_bfe_u32 s18, s26, 0x1000e
-; GFX8-NEXT:    s_bfe_u32 s20, s26, 0x1000c
-; GFX8-NEXT:    s_bfe_u32 s22, s26, 0x10012
-; GFX8-NEXT:    s_bfe_u32 s24, s26, 0x10010
-; GFX8-NEXT:    s_bfe_u32 s45, s26, 0x10016
-; GFX8-NEXT:    s_bfe_u32 s46, s26, 0x10015
-; GFX8-NEXT:    s_bfe_u32 s47, s26, 0x10014
-; GFX8-NEXT:    s_bfe_u32 s48, s26, 0x1001a
-; GFX8-NEXT:    s_bfe_u32 s49, s26, 0x10018
-; GFX8-NEXT:    s_bfe_u32 s50, s26, 0x1001e
-; GFX8-NEXT:    s_bfe_u32 s51, s26, 0x1001c
-; GFX8-NEXT:    s_and_b32 s52, s27, 1
-; GFX8-NEXT:    s_bfe_u32 s53, s27, 0x10002
-; GFX8-NEXT:    s_bfe_u32 s54, s27, 0x10006
-; GFX8-NEXT:    s_bfe_u32 s55, s27, 0x10004
-; GFX8-NEXT:    s_bfe_u32 s56, s27, 0x1000a
-; GFX8-NEXT:    s_bfe_u32 s57, s27, 0x10008
-; GFX8-NEXT:    s_bfe_u32 s58, s27, 0x1000e
-; GFX8-NEXT:    s_bfe_u32 s59, s27, 0x1000c
-; GFX8-NEXT:    s_bfe_u32 s60, s27, 0x10012
-; GFX8-NEXT:    s_bfe_u32 s61, s27, 0x10010
-; GFX8-NEXT:    s_bfe_u32 s62, s27, 0x10016
-; GFX8-NEXT:    s_bfe_u32 s63, s27, 0x10015
-; GFX8-NEXT:    s_bfe_u32 s64, s27, 0x10014
-; GFX8-NEXT:    s_bfe_u32 s65, s27, 0x1001a
-; GFX8-NEXT:    s_bfe_u32 s66, s27, 0x10018
-; GFX8-NEXT:    s_bfe_u32 s26, s27, 0x1001e
-; GFX8-NEXT:    s_bfe_u32 s27, s27, 0x1001c
-; GFX8-NEXT:    v_mov_b32_e32 v2, s26
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xf0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s27
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xe0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s44
-; GFX8-NEXT:    v_mov_b32_e32 v3, s43
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
-; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xd0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s66
-; GFX8-NEXT:    v_mov_b32_e32 v1, s42
-; GFX8-NEXT:    v_mov_b32_e32 v2, s65
-; GFX8-NEXT:    v_mov_b32_e32 v3, s41
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    s_bfe_u32 s2, s30, 0x10003
+; GFX8-NEXT:    s_bfe_u32 s3, s30, 0x10001
+; GFX8-NEXT:    s_bfe_u32 s4, s30, 0x10007
+; GFX8-NEXT:    s_bfe_u32 s5, s30, 0x10005
+; GFX8-NEXT:    s_bfe_u32 s8, s30, 0x1000b
+; GFX8-NEXT:    s_bfe_u32 s9, s30, 0x10009
+; GFX8-NEXT:    s_bfe_u32 s12, s30, 0x1000f
+; GFX8-NEXT:    s_bfe_u32 s13, s30, 0x1000d
+; GFX8-NEXT:    s_bfe_u32 s16, s30, 0x10013
+; GFX8-NEXT:    s_bfe_u32 s17, s30, 0x10011
+; GFX8-NEXT:    s_bfe_u32 s20, s30, 0x10017
+; GFX8-NEXT:    s_bfe_u32 s23, s30, 0x1001b
+; GFX8-NEXT:    s_bfe_u32 s24, s30, 0x10019
+; GFX8-NEXT:    s_lshr_b32 s27, s30, 31
+; GFX8-NEXT:    s_bfe_u32 s29, s30, 0x1001d
+; GFX8-NEXT:    s_bfe_u32 s33, s31, 0x10003
+; GFX8-NEXT:    s_bfe_u32 s34, s31, 0x10001
+; GFX8-NEXT:    s_bfe_u32 s35, s31, 0x10007
+; GFX8-NEXT:    s_bfe_u32 s36, s31, 0x10005
+; GFX8-NEXT:    s_bfe_u32 s37, s31, 0x1000b
+; GFX8-NEXT:    s_bfe_u32 s38, s31, 0x10009
+; GFX8-NEXT:    s_bfe_u32 s39, s31, 0x1000f
+; GFX8-NEXT:    s_bfe_u32 s40, s31, 0x1000d
+; GFX8-NEXT:    s_bfe_u32 s41, s31, 0x10013
+; GFX8-NEXT:    s_bfe_u32 s42, s31, 0x10011
+; GFX8-NEXT:    s_bfe_u32 s43, s31, 0x10017
+; GFX8-NEXT:    s_bfe_u32 s44, s31, 0x1001b
+; GFX8-NEXT:    s_bfe_u32 s45, s31, 0x10019
+; GFX8-NEXT:    s_lshr_b32 s46, s31, 31
+; GFX8-NEXT:    s_bfe_u32 s47, s31, 0x1001d
+; GFX8-NEXT:    s_and_b32 s7, s30, 1
+; GFX8-NEXT:    s_bfe_u32 s6, s30, 0x10002
+; GFX8-NEXT:    s_bfe_u32 s10, s30, 0x10006
+; GFX8-NEXT:    s_bfe_u32 s11, s30, 0x10004
+; GFX8-NEXT:    s_bfe_u32 s14, s30, 0x1000a
+; GFX8-NEXT:    s_bfe_u32 s15, s30, 0x10008
+; GFX8-NEXT:    s_bfe_u32 s18, s30, 0x1000e
+; GFX8-NEXT:    s_bfe_u32 s19, s30, 0x1000c
+; GFX8-NEXT:    s_bfe_u32 s21, s30, 0x10012
+; GFX8-NEXT:    s_bfe_u32 s22, s30, 0x10010
+; GFX8-NEXT:    s_bfe_u32 s25, s30, 0x10016
+; GFX8-NEXT:    s_bfe_u32 s26, s30, 0x10015
+; GFX8-NEXT:    s_bfe_u32 s28, s30, 0x10014
+; GFX8-NEXT:    s_bfe_u32 s48, s30, 0x1001a
+; GFX8-NEXT:    s_bfe_u32 s49, s30, 0x10018
+; GFX8-NEXT:    s_bfe_u32 s50, s30, 0x1001e
+; GFX8-NEXT:    s_bfe_u32 s30, s30, 0x1001c
+; GFX8-NEXT:    s_and_b32 s51, s31, 1
+; GFX8-NEXT:    s_bfe_u32 s52, s31, 0x10002
+; GFX8-NEXT:    s_bfe_u32 s53, s31, 0x10006
+; GFX8-NEXT:    s_bfe_u32 s54, s31, 0x10004
+; GFX8-NEXT:    s_bfe_u32 s55, s31, 0x1000a
+; GFX8-NEXT:    s_bfe_u32 s56, s31, 0x10008
+; GFX8-NEXT:    s_bfe_u32 s57, s31, 0x1000e
+; GFX8-NEXT:    s_bfe_u32 s58, s31, 0x1000c
+; GFX8-NEXT:    s_bfe_u32 s59, s31, 0x10012
+; GFX8-NEXT:    s_bfe_u32 s60, s31, 0x10010
+; GFX8-NEXT:    s_bfe_u32 s61, s31, 0x10016
+; GFX8-NEXT:    s_bfe_u32 s62, s31, 0x10015
+; GFX8-NEXT:    s_bfe_u32 s63, s31, 0x10014
+; GFX8-NEXT:    s_bfe_u32 s64, s31, 0x1001a
+; GFX8-NEXT:    s_bfe_u32 s65, s31, 0x10018
+; GFX8-NEXT:    s_bfe_u32 s66, s31, 0x1001e
+; GFX8-NEXT:    s_bfe_u32 s31, s31, 0x1001c
+; GFX8-NEXT:    v_mov_b32_e32 v0, s31
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xf0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s46
+; GFX8-NEXT:    s_addc_u32 s46, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s47
+; GFX8-NEXT:    v_mov_b32_e32 v2, s66
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v5, s46
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xe0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xc0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s64
-; GFX8-NEXT:    v_mov_b32_e32 v1, s63
-; GFX8-NEXT:    v_mov_b32_e32 v2, s62
-; GFX8-NEXT:    v_mov_b32_e32 v3, s40
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s44
+; GFX8-NEXT:    s_addc_u32 s44, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s65
+; GFX8-NEXT:    v_mov_b32_e32 v1, s45
+; GFX8-NEXT:    v_mov_b32_e32 v2, s64
+; GFX8-NEXT:    v_mov_b32_e32 v5, s44
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xd0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xb0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s61
-; GFX8-NEXT:    v_mov_b32_e32 v1, s39
-; GFX8-NEXT:    v_mov_b32_e32 v2, s60
-; GFX8-NEXT:    v_mov_b32_e32 v3, s38
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s43
+; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s63
+; GFX8-NEXT:    v_mov_b32_e32 v1, s62
+; GFX8-NEXT:    v_mov_b32_e32 v2, s61
+; GFX8-NEXT:    v_mov_b32_e32 v5, s43
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xc0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xa0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s59
-; GFX8-NEXT:    v_mov_b32_e32 v1, s37
-; GFX8-NEXT:    v_mov_b32_e32 v2, s58
-; GFX8-NEXT:    v_mov_b32_e32 v3, s36
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s41
+; GFX8-NEXT:    s_addc_u32 s41, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s60
+; GFX8-NEXT:    v_mov_b32_e32 v1, s42
+; GFX8-NEXT:    v_mov_b32_e32 v2, s59
+; GFX8-NEXT:    v_mov_b32_e32 v5, s41
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xb0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0x90
-; GFX8-NEXT:    v_mov_b32_e32 v0, s57
-; GFX8-NEXT:    v_mov_b32_e32 v1, s35
-; GFX8-NEXT:    v_mov_b32_e32 v2, s56
-; GFX8-NEXT:    v_mov_b32_e32 v3, s34
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s39
+; GFX8-NEXT:    s_addc_u32 s39, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s58
+; GFX8-NEXT:    v_mov_b32_e32 v1, s40
+; GFX8-NEXT:    v_mov_b32_e32 v2, s57
+; GFX8-NEXT:    v_mov_b32_e32 v5, s39
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xa0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0x80
-; GFX8-NEXT:    v_mov_b32_e32 v0, s55
-; GFX8-NEXT:    v_mov_b32_e32 v1, s33
-; GFX8-NEXT:    v_mov_b32_e32 v2, s54
-; GFX8-NEXT:    v_mov_b32_e32 v3, s31
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s37
+; GFX8-NEXT:    s_addc_u32 s37, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s56
+; GFX8-NEXT:    v_mov_b32_e32 v1, s38
+; GFX8-NEXT:    v_mov_b32_e32 v2, s55
+; GFX8-NEXT:    v_mov_b32_e32 v5, s37
+; GFX8-NEXT:    s_add_u32 s31, s0, 0x90
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0x70
-; GFX8-NEXT:    v_mov_b32_e32 v0, s52
-; GFX8-NEXT:    v_mov_b32_e32 v1, s30
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s35
+; GFX8-NEXT:    s_addc_u32 s35, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s54
+; GFX8-NEXT:    v_mov_b32_e32 v1, s36
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s53
-; GFX8-NEXT:    v_mov_b32_e32 v3, s29
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s35
+; GFX8-NEXT:    s_add_u32 s31, s0, 0x80
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0x60
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s33
+; GFX8-NEXT:    s_addc_u32 s33, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s51
-; GFX8-NEXT:    v_mov_b32_e32 v1, s28
+; GFX8-NEXT:    v_mov_b32_e32 v1, s34
+; GFX8-NEXT:    v_mov_b32_e32 v2, s52
+; GFX8-NEXT:    v_mov_b32_e32 v5, s33
+; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s27
+; GFX8-NEXT:    s_add_u32 s27, s0, 0x70
+; GFX8-NEXT:    v_mov_b32_e32 v1, s29
+; GFX8-NEXT:    s_addc_u32 s29, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s30
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s50
-; GFX8-NEXT:    v_mov_b32_e32 v3, s25
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s27
+; GFX8-NEXT:    v_mov_b32_e32 v5, s29
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0x50
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s23
+; GFX8-NEXT:    s_add_u32 s23, s0, 0x60
+; GFX8-NEXT:    v_mov_b32_e32 v1, s24
+; GFX8-NEXT:    s_addc_u32 s24, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s49
-; GFX8-NEXT:    v_mov_b32_e32 v1, s23
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s48
-; GFX8-NEXT:    v_mov_b32_e32 v3, s21
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
-; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v0, s47
-; GFX8-NEXT:    v_mov_b32_e32 v1, s46
-; GFX8-NEXT:    v_mov_b32_e32 v2, s45
-; GFX8-NEXT:    v_mov_b32_e32 v3, s19
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
+; GFX8-NEXT:    v_mov_b32_e32 v4, s23
+; GFX8-NEXT:    v_mov_b32_e32 v5, s24
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s22
-; GFX8-NEXT:    s_add_u32 s22, s0, 64
+; GFX8-NEXT:    v_mov_b32_e32 v3, s20
+; GFX8-NEXT:    s_add_u32 s20, s0, 0x50
 ; GFX8-NEXT:    s_addc_u32 s23, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s24
-; GFX8-NEXT:    v_mov_b32_e32 v1, s17
-; GFX8-NEXT:    v_mov_b32_e32 v3, s15
-; GFX8-NEXT:    v_mov_b32_e32 v4, s22
+; GFX8-NEXT:    v_mov_b32_e32 v0, s28
+; GFX8-NEXT:    v_mov_b32_e32 v1, s26
+; GFX8-NEXT:    v_mov_b32_e32 v2, s25
+; GFX8-NEXT:    v_mov_b32_e32 v4, s20
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s23
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s18
-; GFX8-NEXT:    s_add_u32 s18, s0, 48
-; GFX8-NEXT:    s_addc_u32 s19, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s20
+; GFX8-NEXT:    v_mov_b32_e32 v3, s16
+; GFX8-NEXT:    s_add_u32 s16, s0, 64
+; GFX8-NEXT:    v_mov_b32_e32 v1, s17
+; GFX8-NEXT:    s_addc_u32 s17, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s22
+; GFX8-NEXT:    v_mov_b32_e32 v2, s21
+; GFX8-NEXT:    v_mov_b32_e32 v4, s16
+; GFX8-NEXT:    v_mov_b32_e32 v5, s17
+; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s12
+; GFX8-NEXT:    s_add_u32 s12, s0, 48
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s13
-; GFX8-NEXT:    v_mov_b32_e32 v3, s11
-; GFX8-NEXT:    v_mov_b32_e32 v4, s18
-; GFX8-NEXT:    v_mov_b32_e32 v5, s19
+; GFX8-NEXT:    s_addc_u32 s13, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s19
+; GFX8-NEXT:    v_mov_b32_e32 v2, s18
+; GFX8-NEXT:    v_mov_b32_e32 v4, s12
+; GFX8-NEXT:    v_mov_b32_e32 v5, s13
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s14
-; GFX8-NEXT:    s_add_u32 s14, s0, 32
-; GFX8-NEXT:    s_addc_u32 s15, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s16
+; GFX8-NEXT:    v_mov_b32_e32 v3, s8
+; GFX8-NEXT:    s_add_u32 s8, s0, 32
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s9
-; GFX8-NEXT:    v_mov_b32_e32 v3, s6
-; GFX8-NEXT:    v_mov_b32_e32 v4, s14
-; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s9, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s15
+; GFX8-NEXT:    v_mov_b32_e32 v2, s14
+; GFX8-NEXT:    v_mov_b32_e32 v4, s8
+; GFX8-NEXT:    v_mov_b32_e32 v5, s9
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_add_u32 s4, s0, 16
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX8-NEXT:    s_addc_u32 s5, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s12
+; GFX8-NEXT:    v_mov_b32_e32 v0, s11
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s10
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s1
-; GFX8-NEXT:    v_mov_b32_e32 v0, s8
+; GFX8-NEXT:    v_mov_b32_e32 v0, s7
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
-; GFX8-NEXT:    v_mov_b32_e32 v2, s7
+; GFX8-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -4441,160 +4444,162 @@ define amdgpu_kernel void @constant_sextload_v64i1_to_v64i32(ptr addrspace(1) %o
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dwordx2 s[26:27], s[2:3], 0x0
+; GFX8-NEXT:    s_load_dwordx2 s[30:31], s[2:3], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_bfe_i32 s2, s26, 0x10003
-; GFX8-NEXT:    s_bfe_i32 s3, s26, 0x10002
-; GFX8-NEXT:    s_bfe_i32 s4, s26, 0x10001
-; GFX8-NEXT:    s_bfe_i32 s5, s26, 0x10000
-; GFX8-NEXT:    s_bfe_i32 s6, s26, 0x10007
-; GFX8-NEXT:    s_bfe_i32 s7, s26, 0x10006
-; GFX8-NEXT:    s_bfe_i32 s8, s26, 0x10005
-; GFX8-NEXT:    s_bfe_i32 s9, s26, 0x10004
-; GFX8-NEXT:    s_bfe_i32 s10, s26, 0x1000b
-; GFX8-NEXT:    s_bfe_i32 s11, s26, 0x1000a
-; GFX8-NEXT:    s_bfe_i32 s12, s26, 0x10009
-; GFX8-NEXT:    s_bfe_i32 s13, s26, 0x10008
-; GFX8-NEXT:    s_bfe_i32 s14, s26, 0x1000f
-; GFX8-NEXT:    s_bfe_i32 s15, s26, 0x1000e
-; GFX8-NEXT:    s_bfe_i32 s16, s26, 0x1000d
-; GFX8-NEXT:    s_bfe_i32 s17, s26, 0x1000c
-; GFX8-NEXT:    s_bfe_i32 s18, s26, 0x10013
-; GFX8-NEXT:    s_bfe_i32 s19, s26, 0x10012
-; GFX8-NEXT:    s_bfe_i32 s20, s26, 0x10011
-; GFX8-NEXT:    s_bfe_i32 s21, s26, 0x10010
-; GFX8-NEXT:    s_bfe_i32 s22, s26, 0x10017
-; GFX8-NEXT:    s_bfe_i32 s23, s26, 0x10016
-; GFX8-NEXT:    s_bfe_i32 s24, s26, 0x10015
-; GFX8-NEXT:    s_bfe_i32 s25, s26, 0x10014
-; GFX8-NEXT:    s_bfe_i32 s28, s26, 0x1001b
-; GFX8-NEXT:    s_bfe_i32 s29, s26, 0x1001a
-; GFX8-NEXT:    s_bfe_i32 s30, s26, 0x10019
-; GFX8-NEXT:    s_bfe_i32 s31, s26, 0x10018
-; GFX8-NEXT:    s_ashr_i32 s33, s26, 31
-; GFX8-NEXT:    s_bfe_i32 s34, s26, 0x1001e
-; GFX8-NEXT:    s_bfe_i32 s35, s26, 0x1001d
-; GFX8-NEXT:    s_bfe_i32 s36, s26, 0x1001c
-; GFX8-NEXT:    s_bfe_i32 s37, s27, 0x10003
-; GFX8-NEXT:    s_bfe_i32 s38, s27, 0x10002
-; GFX8-NEXT:    s_bfe_i32 s39, s27, 0x10001
-; GFX8-NEXT:    s_bfe_i32 s40, s27, 0x10000
-; GFX8-NEXT:    s_bfe_i32 s41, s27, 0x10007
-; GFX8-NEXT:    s_bfe_i32 s42, s27, 0x10006
-; GFX8-NEXT:    s_bfe_i32 s43, s27, 0x10005
-; GFX8-NEXT:    s_bfe_i32 s44, s27, 0x10004
-; GFX8-NEXT:    s_bfe_i32 s45, s27, 0x1000b
-; GFX8-NEXT:    s_bfe_i32 s46, s27, 0x1000a
-; GFX8-NEXT:    s_bfe_i32 s47, s27, 0x10009
-; GFX8-NEXT:    s_bfe_i32 s48, s27, 0x10008
-; GFX8-NEXT:    s_bfe_i32 s49, s27, 0x1000f
-; GFX8-NEXT:    s_bfe_i32 s50, s27, 0x1000e
-; GFX8-NEXT:    s_bfe_i32 s51, s27, 0x1000d
-; GFX8-NEXT:    s_bfe_i32 s52, s27, 0x1000c
-; GFX8-NEXT:    s_bfe_i32 s53, s27, 0x10013
-; GFX8-NEXT:    s_bfe_i32 s54, s27, 0x10012
-; GFX8-NEXT:    s_bfe_i32 s55, s27, 0x10011
-; GFX8-NEXT:    s_bfe_i32 s56, s27, 0x10010
-; GFX8-NEXT:    s_bfe_i32 s57, s27, 0x10017
-; GFX8-NEXT:    s_bfe_i32 s58, s27, 0x10016
-; GFX8-NEXT:    s_bfe_i32 s59, s27, 0x10015
-; GFX8-NEXT:    s_bfe_i32 s60, s27, 0x10014
-; GFX8-NEXT:    s_bfe_i32 s61, s27, 0x1001b
-; GFX8-NEXT:    s_bfe_i32 s62, s27, 0x1001a
-; GFX8-NEXT:    s_bfe_i32 s63, s27, 0x10019
-; GFX8-NEXT:    s_bfe_i32 s64, s27, 0x10018
-; GFX8-NEXT:    s_ashr_i32 s26, s27, 31
-; GFX8-NEXT:    s_bfe_i32 s65, s27, 0x1001e
-; GFX8-NEXT:    s_bfe_i32 s66, s27, 0x1001d
-; GFX8-NEXT:    s_bfe_i32 s27, s27, 0x1001c
-; GFX8-NEXT:    v_mov_b32_e32 v3, s26
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xf0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s27
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xe0
+; GFX8-NEXT:    s_bfe_i32 s2, s30, 0x10003
+; GFX8-NEXT:    s_bfe_i32 s3, s30, 0x10002
+; GFX8-NEXT:    s_bfe_i32 s4, s30, 0x10001
+; GFX8-NEXT:    s_bfe_i32 s5, s30, 0x10000
+; GFX8-NEXT:    s_bfe_i32 s6, s30, 0x10007
+; GFX8-NEXT:    s_bfe_i32 s7, s30, 0x10006
+; GFX8-NEXT:    s_bfe_i32 s8, s30, 0x10005
+; GFX8-NEXT:    s_bfe_i32 s9, s30, 0x10004
+; GFX8-NEXT:    s_bfe_i32 s10, s30, 0x1000b
+; GFX8-NEXT:    s_bfe_i32 s11, s30, 0x1000a
+; GFX8-NEXT:    s_bfe_i32 s12, s30, 0x10009
+; GFX8-NEXT:    s_bfe_i32 s13, s30, 0x10008
+; GFX8-NEXT:    s_bfe_i32 s14, s30, 0x1000f
+; GFX8-NEXT:    s_bfe_i32 s15, s30, 0x1000e
+; GFX8-NEXT:    s_bfe_i32 s16, s30, 0x1000d
+; GFX8-NEXT:    s_bfe_i32 s17, s30, 0x1000c
+; GFX8-NEXT:    s_bfe_i32 s18, s30, 0x10013
+; GFX8-NEXT:    s_bfe_i32 s19, s30, 0x10012
+; GFX8-NEXT:    s_bfe_i32 s20, s30, 0x10011
+; GFX8-NEXT:    s_bfe_i32 s21, s30, 0x10010
+; GFX8-NEXT:    s_bfe_i32 s22, s30, 0x10017
+; GFX8-NEXT:    s_bfe_i32 s23, s30, 0x10016
+; GFX8-NEXT:    s_bfe_i32 s24, s30, 0x10015
+; GFX8-NEXT:    s_bfe_i32 s25, s30, 0x10014
+; GFX8-NEXT:    s_bfe_i32 s26, s30, 0x1001b
+; GFX8-NEXT:    s_bfe_i32 s27, s30, 0x1001a
+; GFX8-NEXT:    s_bfe_i32 s28, s30, 0x10019
+; GFX8-NEXT:    s_bfe_i32 s29, s30, 0x10018
+; GFX8-NEXT:    s_ashr_i32 s33, s30, 31
+; GFX8-NEXT:    s_bfe_i32 s34, s30, 0x1001e
+; GFX8-NEXT:    s_bfe_i32 s35, s30, 0x1001d
+; GFX8-NEXT:    s_bfe_i32 s30, s30, 0x1001c
+; GFX8-NEXT:    s_bfe_i32 s36, s31, 0x10003
+; GFX8-NEXT:    s_bfe_i32 s37, s31, 0x10002
+; GFX8-NEXT:    s_bfe_i32 s38, s31, 0x10001
+; GFX8-NEXT:    s_bfe_i32 s39, s31, 0x10000
+; GFX8-NEXT:    s_bfe_i32 s40, s31, 0x10007
+; GFX8-NEXT:    s_bfe_i32 s41, s31, 0x10006
+; GFX8-NEXT:    s_bfe_i32 s42, s31, 0x10005
+; GFX8-NEXT:    s_bfe_i32 s43, s31, 0x10004
+; GFX8-NEXT:    s_bfe_i32 s44, s31, 0x1000b
+; GFX8-NEXT:    s_bfe_i32 s45, s31, 0x1000a
+; GFX8-NEXT:    s_bfe_i32 s46, s31, 0x10009
+; GFX8-NEXT:    s_bfe_i32 s47, s31, 0x10008
+; GFX8-NEXT:    s_bfe_i32 s48, s31, 0x1000f
+; GFX8-NEXT:    s_bfe_i32 s49, s31, 0x1000e
+; GFX8-NEXT:    s_bfe_i32 s50, s31, 0x1000d
+; GFX8-NEXT:    s_bfe_i32 s51, s31, 0x1000c
+; GFX8-NEXT:    s_bfe_i32 s52, s31, 0x10013
+; GFX8-NEXT:    s_bfe_i32 s53, s31, 0x10012
+; GFX8-NEXT:    s_bfe_i32 s54, s31, 0x10011
+; GFX8-NEXT:    s_bfe_i32 s55, s31, 0x10010
+; GFX8-NEXT:    s_bfe_i32 s56, s31, 0x10017
+; GFX8-NEXT:    s_bfe_i32 s57, s31, 0x10016
+; GFX8-NEXT:    s_bfe_i32 s58, s31, 0x10015
+; GFX8-NEXT:    s_bfe_i32 s59, s31, 0x10014
+; GFX8-NEXT:    s_bfe_i32 s60, s31, 0x1001b
+; GFX8-NEXT:    s_bfe_i32 s61, s31, 0x1001a
+; GFX8-NEXT:    s_bfe_i32 s62, s31, 0x10019
+; GFX8-NEXT:    s_bfe_i32 s63, s31, 0x10018
+; GFX8-NEXT:    s_ashr_i32 s64, s31, 31
+; GFX8-NEXT:    s_bfe_i32 s65, s31, 0x1001e
+; GFX8-NEXT:    s_bfe_i32 s66, s31, 0x1001d
+; GFX8-NEXT:    s_bfe_i32 s31, s31, 0x1001c
+; GFX8-NEXT:    v_mov_b32_e32 v0, s31
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xf0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s64
+; GFX8-NEXT:    s_addc_u32 s64, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s66
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s65
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
-; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xd0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s64
-; GFX8-NEXT:    v_mov_b32_e32 v1, s63
-; GFX8-NEXT:    v_mov_b32_e32 v2, s62
-; GFX8-NEXT:    v_mov_b32_e32 v3, s61
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
-; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xc0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s60
-; GFX8-NEXT:    v_mov_b32_e32 v1, s59
-; GFX8-NEXT:    v_mov_b32_e32 v2, s58
-; GFX8-NEXT:    v_mov_b32_e32 v3, s57
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v5, s64
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xe0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xb0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s56
-; GFX8-NEXT:    v_mov_b32_e32 v1, s55
-; GFX8-NEXT:    v_mov_b32_e32 v2, s54
-; GFX8-NEXT:    v_mov_b32_e32 v3, s53
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s60
+; GFX8-NEXT:    s_addc_u32 s60, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s63
+; GFX8-NEXT:    v_mov_b32_e32 v1, s62
+; GFX8-NEXT:    v_mov_b32_e32 v2, s61
+; GFX8-NEXT:    v_mov_b32_e32 v5, s60
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xd0
+; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s56
+; GFX8-NEXT:    s_addc_u32 s56, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s59
+; GFX8-NEXT:    v_mov_b32_e32 v1, s58
+; GFX8-NEXT:    v_mov_b32_e32 v2, s57
+; GFX8-NEXT:    v_mov_b32_e32 v5, s56
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xc0
+; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s52
+; GFX8-NEXT:    s_addc_u32 s52, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s55
+; GFX8-NEXT:    v_mov_b32_e32 v1, s54
+; GFX8-NEXT:    v_mov_b32_e32 v2, s53
+; GFX8-NEXT:    v_mov_b32_e32 v5, s52
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xb0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xa0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s52
-; GFX8-NEXT:    v_mov_b32_e32 v1, s51
-; GFX8-NEXT:    v_mov_b32_e32 v2, s50
-; GFX8-NEXT:    v_mov_b32_e32 v3, s49
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s48
+; GFX8-NEXT:    s_addc_u32 s48, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s51
+; GFX8-NEXT:    v_mov_b32_e32 v1, s50
+; GFX8-NEXT:    v_mov_b32_e32 v2, s49
+; GFX8-NEXT:    v_mov_b32_e32 v5, s48
+; GFX8-NEXT:    s_add_u32 s31, s0, 0xa0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0x90
-; GFX8-NEXT:    v_mov_b32_e32 v0, s48
-; GFX8-NEXT:    v_mov_b32_e32 v1, s47
-; GFX8-NEXT:    v_mov_b32_e32 v2, s46
-; GFX8-NEXT:    v_mov_b32_e32 v3, s45
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s44
+; GFX8-NEXT:    s_addc_u32 s44, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s47
+; GFX8-NEXT:    v_mov_b32_e32 v1, s46
+; GFX8-NEXT:    v_mov_b32_e32 v2, s45
+; GFX8-NEXT:    v_mov_b32_e32 v5, s44
+; GFX8-NEXT:    s_add_u32 s31, s0, 0x90
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0x80
-; GFX8-NEXT:    v_mov_b32_e32 v0, s44
-; GFX8-NEXT:    v_mov_b32_e32 v1, s43
-; GFX8-NEXT:    v_mov_b32_e32 v2, s42
-; GFX8-NEXT:    v_mov_b32_e32 v3, s41
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s40
+; GFX8-NEXT:    s_addc_u32 s40, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s43
+; GFX8-NEXT:    v_mov_b32_e32 v1, s42
+; GFX8-NEXT:    v_mov_b32_e32 v2, s41
+; GFX8-NEXT:    v_mov_b32_e32 v5, s40
+; GFX8-NEXT:    s_add_u32 s31, s0, 0x80
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0x70
-; GFX8-NEXT:    v_mov_b32_e32 v0, s40
-; GFX8-NEXT:    v_mov_b32_e32 v1, s39
-; GFX8-NEXT:    v_mov_b32_e32 v2, s38
-; GFX8-NEXT:    v_mov_b32_e32 v3, s37
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s31
+; GFX8-NEXT:    v_mov_b32_e32 v3, s36
+; GFX8-NEXT:    s_addc_u32 s36, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s39
+; GFX8-NEXT:    v_mov_b32_e32 v1, s38
+; GFX8-NEXT:    v_mov_b32_e32 v2, s37
+; GFX8-NEXT:    v_mov_b32_e32 v5, s36
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s26, s0, 0x60
-; GFX8-NEXT:    v_mov_b32_e32 v0, s36
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s30
+; GFX8-NEXT:    s_add_u32 s30, s0, 0x70
+; GFX8-NEXT:    s_addc_u32 s31, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s35
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s34
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s33
-; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s30
+; GFX8-NEXT:    v_mov_b32_e32 v5, s31
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s26
+; GFX8-NEXT:    s_add_u32 s26, s0, 0x60
+; GFX8-NEXT:    v_mov_b32_e32 v2, s27
+; GFX8-NEXT:    s_addc_u32 s27, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s29
+; GFX8-NEXT:    v_mov_b32_e32 v1, s28
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s26
-; GFX8-NEXT:    v_mov_b32_e32 v0, s31
-; GFX8-NEXT:    v_mov_b32_e32 v1, s30
-; GFX8-NEXT:    v_mov_b32_e32 v2, s29
-; GFX8-NEXT:    v_mov_b32_e32 v3, s28
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s27
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
@@ -5818,12 +5823,12 @@ define amdgpu_kernel void @constant_zextload_v3i1_to_v3i64(ptr addrspace(1) %out
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    flat_load_ubyte v4, v[0:1]
 ; GFX8-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v9, s3
-; GFX8-NEXT:    v_mov_b32_e32 v8, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX8-NEXT:    v_mov_b32_e32 v7, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v6, s0
+; GFX8-NEXT:    s_addc_u32 s0, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v8, s2
+; GFX8-NEXT:    v_mov_b32_e32 v9, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    v_and_b32_e32 v0, 1, v4
 ; GFX8-NEXT:    v_bfe_u32 v2, v4, 1, 1
@@ -5996,11 +6001,11 @@ define amdgpu_kernel void @constant_sextload_v3i1_to_v3i64(ptr addrspace(1) %out
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    flat_load_ubyte v0, v[0:1]
 ; GFX8-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v7, s3
-; GFX8-NEXT:    v_mov_b32_e32 v6, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NEXT:    s_addc_u32 s0, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v6, s2
+; GFX8-NEXT:    v_mov_b32_e32 v7, s0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v3, 2, v0
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v2, 1, v0
@@ -6179,14 +6184,14 @@ define amdgpu_kernel void @constant_zextload_v4i1_to_v4i64(ptr addrspace(1) %out
 ; GFX8-NEXT:    flat_load_ubyte v0, v[0:1]
 ; GFX8-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v9, s1
+; GFX8-NEXT:    v_mov_b32_e32 v8, s0
+; GFX8-NEXT:    s_addc_u32 s0, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, v1
-; GFX8-NEXT:    v_mov_b32_e32 v11, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v10, s2
+; GFX8-NEXT:    v_mov_b32_e32 v11, s0
 ; GFX8-NEXT:    v_mov_b32_e32 v5, v1
 ; GFX8-NEXT:    v_mov_b32_e32 v7, v1
-; GFX8-NEXT:    v_mov_b32_e32 v9, s1
-; GFX8-NEXT:    v_mov_b32_e32 v8, s0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    v_and_b32_e32 v4, 1, v0
 ; GFX8-NEXT:    v_bfe_u32 v6, v0, 1, 1
@@ -6388,11 +6393,11 @@ define amdgpu_kernel void @constant_sextload_v4i1_to_v4i64(ptr addrspace(1) %out
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    flat_load_ubyte v0, v[0:1]
 ; GFX8-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v11, s3
-; GFX8-NEXT:    v_mov_b32_e32 v10, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v9, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v8, s0
+; GFX8-NEXT:    s_addc_u32 s0, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v10, s2
+; GFX8-NEXT:    v_mov_b32_e32 v11, s0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v3, 2, v0
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v4, 3, v0
@@ -6608,8 +6613,8 @@ define amdgpu_kernel void @constant_zextload_v8i1_to_v8i64(ptr addrspace(1) %out
 ; GFX8-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v5, v1
 ; GFX8-NEXT:    v_mov_b32_e32 v7, v1
-; GFX8-NEXT:    v_mov_b32_e32 v16, s5
 ; GFX8-NEXT:    v_mov_b32_e32 v15, s4
+; GFX8-NEXT:    v_mov_b32_e32 v16, s5
 ; GFX8-NEXT:    v_mov_b32_e32 v8, v1
 ; GFX8-NEXT:    v_mov_b32_e32 v10, v1
 ; GFX8-NEXT:    v_mov_b32_e32 v3, v1
@@ -6624,12 +6629,12 @@ define amdgpu_kernel void @constant_zextload_v8i1_to_v8i64(ptr addrspace(1) %out
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NEXT:    s_add_u32 s0, s0, 16
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v18, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v17, s0
+; GFX8-NEXT:    v_mov_b32_e32 v18, s1
 ; GFX8-NEXT:    v_and_b32_e32 v6, 0xffff, v0
 ; GFX8-NEXT:    v_bfe_u32 v7, v0, 2, 1
-; GFX8-NEXT:    v_mov_b32_e32 v16, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v15, s2
+; GFX8-NEXT:    v_mov_b32_e32 v16, s3
 ; GFX8-NEXT:    v_bfe_u32 v13, v0, 1, 1
 ; GFX8-NEXT:    v_and_b32_e32 v11, 1, v0
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v2, 7, v6
@@ -6886,9 +6891,9 @@ define amdgpu_kernel void @constant_sextload_v8i1_to_v8i64(ptr addrspace(1) %out
 ; GFX8-NEXT:    s_add_u32 s2, s0, 48
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v19, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v18, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 32
+; GFX8-NEXT:    v_mov_b32_e32 v19, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v6, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v7, s5
@@ -6899,16 +6904,16 @@ define amdgpu_kernel void @constant_sextload_v8i1_to_v8i64(ptr addrspace(1) %out
 ; GFX8-NEXT:    v_mov_b32_e32 v11, s9
 ; GFX8-NEXT:    flat_store_dwordx4 v[18:19], v[4:7]
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_bfe_i32 v0, v0, 0, 1
 ; GFX8-NEXT:    v_mov_b32_e32 v12, s10
 ; GFX8-NEXT:    v_mov_b32_e32 v13, s11
 ; GFX8-NEXT:    v_mov_b32_e32 v14, s12
 ; GFX8-NEXT:    v_mov_b32_e32 v15, s13
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[8:11]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s14
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s15
@@ -7178,16 +7183,16 @@ define amdgpu_kernel void @constant_zextload_v16i1_to_v16i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX8-NEXT:    v_and_b32_e32 v2, 0xffff, v0
-; GFX8-NEXT:    s_bfe_u32 s4, s2, 0x10009
-; GFX8-NEXT:    s_bfe_u32 s5, s2, 0x1000d
-; GFX8-NEXT:    s_bfe_u32 s6, s2, 0x10007
-; GFX8-NEXT:    s_bfe_u32 s7, s2, 0x10003
-; GFX8-NEXT:    s_bfe_u32 s8, s2, 0x10001
-; GFX8-NEXT:    s_and_b32 s9, s2, 1
-; GFX8-NEXT:    s_bfe_u32 s10, s2, 0x10002
-; GFX8-NEXT:    s_bfe_u32 s11, s2, 0x10004
-; GFX8-NEXT:    s_bfe_u32 s12, s2, 0x10006
-; GFX8-NEXT:    s_bfe_u32 s13, s2, 0x1000c
+; GFX8-NEXT:    s_bfe_u32 s3, s2, 0x10009
+; GFX8-NEXT:    s_bfe_u32 s4, s2, 0x1000d
+; GFX8-NEXT:    s_bfe_u32 s5, s2, 0x10007
+; GFX8-NEXT:    s_bfe_u32 s6, s2, 0x10003
+; GFX8-NEXT:    s_bfe_u32 s7, s2, 0x10001
+; GFX8-NEXT:    s_and_b32 s8, s2, 1
+; GFX8-NEXT:    s_bfe_u32 s9, s2, 0x10002
+; GFX8-NEXT:    s_bfe_u32 s10, s2, 0x10004
+; GFX8-NEXT:    s_bfe_u32 s11, s2, 0x10006
+; GFX8-NEXT:    s_bfe_u32 s12, s2, 0x1000c
 ; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x1000a
 ; GFX8-NEXT:    v_bfe_u32 v4, v2, 11, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v8, 15, v2
@@ -7196,52 +7201,52 @@ define amdgpu_kernel void @constant_zextload_v16i1_to_v16i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    v_bfe_u32 v0, v2, 8, 1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 0x50
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v11, s3
+; GFX8-NEXT:    s_addc_u32 s13, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v10, s2
+; GFX8-NEXT:    v_mov_b32_e32 v11, s13
 ; GFX8-NEXT:    s_add_u32 s2, s0, 64
-; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[10:11], v[2:5]
-; GFX8-NEXT:    v_mov_b32_e32 v11, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v10, s2
+; GFX8-NEXT:    v_mov_b32_e32 v2, s3
+; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    s_add_u32 s2, s0, 0x70
-; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v11, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    s_add_u32 s2, s0, 0x60
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[0:1], v[6:9]
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    s_add_u32 s2, s0, 48
-; GFX8-NEXT:    v_mov_b32_e32 v2, s13
+; GFX8-NEXT:    v_mov_b32_e32 v2, s12
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    s_add_u32 s2, s0, 32
-; GFX8-NEXT:    v_mov_b32_e32 v2, s12
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v2, s11
+; GFX8-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NEXT:    v_mov_b32_e32 v2, s11
+; GFX8-NEXT:    v_mov_b32_e32 v2, s10
 ; GFX8-NEXT:    v_mov_b32_e32 v4, v12
+; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v2, s10
-; GFX8-NEXT:    v_mov_b32_e32 v4, s7
+; GFX8-NEXT:    v_mov_b32_e32 v2, s9
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s9
-; GFX8-NEXT:    v_mov_b32_e32 v4, s8
+; GFX8-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
 ; GFX8-NEXT:    s_endpgm
@@ -7673,44 +7678,44 @@ define amdgpu_kernel void @constant_sextload_v16i1_to_v16i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_add_u32 s2, s0, 0x70
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v14, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 0x60
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s5
+; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[14:15], v[2:5]
-; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v14, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 0x50
 ; GFX8-NEXT:    v_mov_b32_e32 v6, s6
 ; GFX8-NEXT:    v_mov_b32_e32 v7, s7
 ; GFX8-NEXT:    v_mov_b32_e32 v8, s8
 ; GFX8-NEXT:    v_mov_b32_e32 v9, s9
+; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[14:15], v[6:9]
-; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v14, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 64
 ; GFX8-NEXT:    v_mov_b32_e32 v10, s10
 ; GFX8-NEXT:    v_mov_b32_e32 v11, s11
 ; GFX8-NEXT:    v_mov_b32_e32 v12, s12
 ; GFX8-NEXT:    v_mov_b32_e32 v13, s13
+; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
-; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v14, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 48
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s14
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s15
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s16
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s17
+; GFX8-NEXT:    v_mov_b32_e32 v15, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[14:15], v[2:5]
 ; GFX8-NEXT:    v_mov_b32_e32 v6, s18
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NEXT:    s_add_u32 s2, s0, 32
+; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v7, s19
 ; GFX8-NEXT:    v_mov_b32_e32 v8, s20
@@ -7721,8 +7726,8 @@ define amdgpu_kernel void @constant_sextload_v16i1_to_v16i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    v_mov_b32_e32 v12, s24
 ; GFX8-NEXT:    v_mov_b32_e32 v13, s25
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[6:9]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX8-NEXT:    v_bfe_i32 v0, v0, 0, 1
 ; GFX8-NEXT:    v_mov_b32_e32 v14, s26
@@ -7730,8 +7735,8 @@ define amdgpu_kernel void @constant_sextload_v16i1_to_v16i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    v_mov_b32_e32 v16, s28
 ; GFX8-NEXT:    v_mov_b32_e32 v17, s29
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[10:13]
-; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s30
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s31
@@ -8163,136 +8168,138 @@ define amdgpu_kernel void @constant_zextload_v32i1_to_v32i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s6, s[2:3], 0x0
+; GFX8-NEXT:    s_load_dword s10, s[2:3], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_lshr_b32 s7, s6, 31
-; GFX8-NEXT:    s_bfe_u32 s8, s6, 0x1001d
-; GFX8-NEXT:    s_bfe_u32 s9, s6, 0x1001b
-; GFX8-NEXT:    s_bfe_u32 s10, s6, 0x10019
-; GFX8-NEXT:    s_bfe_u32 s11, s6, 0x10017
-; GFX8-NEXT:    s_bfe_u32 s12, s6, 0x10013
-; GFX8-NEXT:    s_bfe_u32 s13, s6, 0x10011
-; GFX8-NEXT:    s_bfe_u32 s14, s6, 0x1000f
-; GFX8-NEXT:    s_bfe_u32 s15, s6, 0x1000d
-; GFX8-NEXT:    s_bfe_u32 s16, s6, 0x1000b
-; GFX8-NEXT:    s_bfe_u32 s17, s6, 0x10009
-; GFX8-NEXT:    s_bfe_u32 s18, s6, 0x10007
-; GFX8-NEXT:    s_bfe_u32 s19, s6, 0x10005
-; GFX8-NEXT:    s_bfe_u32 s4, s6, 0x10003
-; GFX8-NEXT:    s_bfe_u32 s2, s6, 0x10001
-; GFX8-NEXT:    s_and_b32 s3, s6, 1
-; GFX8-NEXT:    s_bfe_u32 s5, s6, 0x10002
-; GFX8-NEXT:    s_bfe_u32 s20, s6, 0x10004
-; GFX8-NEXT:    s_bfe_u32 s21, s6, 0x10006
-; GFX8-NEXT:    s_bfe_u32 s22, s6, 0x10008
-; GFX8-NEXT:    s_bfe_u32 s23, s6, 0x1000a
-; GFX8-NEXT:    s_bfe_u32 s24, s6, 0x1000c
-; GFX8-NEXT:    s_bfe_u32 s25, s6, 0x1000e
-; GFX8-NEXT:    s_bfe_u32 s26, s6, 0x10010
-; GFX8-NEXT:    s_bfe_u32 s27, s6, 0x10012
-; GFX8-NEXT:    s_bfe_u32 s28, s6, 0x10014
-; GFX8-NEXT:    s_bfe_u32 s29, s6, 0x10015
-; GFX8-NEXT:    s_bfe_u32 s30, s6, 0x10016
-; GFX8-NEXT:    s_bfe_u32 s31, s6, 0x10018
-; GFX8-NEXT:    s_bfe_u32 s33, s6, 0x1001a
-; GFX8-NEXT:    s_bfe_u32 s34, s6, 0x1001c
-; GFX8-NEXT:    s_bfe_u32 s6, s6, 0x1001e
-; GFX8-NEXT:    v_mov_b32_e32 v0, s6
-; GFX8-NEXT:    s_add_u32 s6, s0, 0xf0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s7
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 0xe0
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    s_lshr_b32 s11, s10, 31
+; GFX8-NEXT:    s_bfe_u32 s12, s10, 0x1001d
+; GFX8-NEXT:    s_bfe_u32 s13, s10, 0x1001b
+; GFX8-NEXT:    s_bfe_u32 s14, s10, 0x10019
+; GFX8-NEXT:    s_bfe_u32 s15, s10, 0x10017
+; GFX8-NEXT:    s_bfe_u32 s16, s10, 0x10013
+; GFX8-NEXT:    s_bfe_u32 s17, s10, 0x10011
+; GFX8-NEXT:    s_bfe_u32 s18, s10, 0x1000f
+; GFX8-NEXT:    s_bfe_u32 s19, s10, 0x1000d
+; GFX8-NEXT:    s_bfe_u32 s20, s10, 0x1000b
+; GFX8-NEXT:    s_bfe_u32 s21, s10, 0x10009
+; GFX8-NEXT:    s_bfe_u32 s8, s10, 0x10007
+; GFX8-NEXT:    s_bfe_u32 s6, s10, 0x10005
+; GFX8-NEXT:    s_bfe_u32 s4, s10, 0x10003
+; GFX8-NEXT:    s_bfe_u32 s2, s10, 0x10001
+; GFX8-NEXT:    s_and_b32 s3, s10, 1
+; GFX8-NEXT:    s_bfe_u32 s5, s10, 0x10002
+; GFX8-NEXT:    s_bfe_u32 s7, s10, 0x10004
+; GFX8-NEXT:    s_bfe_u32 s9, s10, 0x10006
+; GFX8-NEXT:    s_bfe_u32 s22, s10, 0x10008
+; GFX8-NEXT:    s_bfe_u32 s23, s10, 0x1000a
+; GFX8-NEXT:    s_bfe_u32 s24, s10, 0x1000c
+; GFX8-NEXT:    s_bfe_u32 s25, s10, 0x1000e
+; GFX8-NEXT:    s_bfe_u32 s26, s10, 0x10010
+; GFX8-NEXT:    s_bfe_u32 s27, s10, 0x10012
+; GFX8-NEXT:    s_bfe_u32 s28, s10, 0x10014
+; GFX8-NEXT:    s_bfe_u32 s29, s10, 0x10015
+; GFX8-NEXT:    s_bfe_u32 s30, s10, 0x10016
+; GFX8-NEXT:    s_bfe_u32 s31, s10, 0x10018
+; GFX8-NEXT:    s_bfe_u32 s33, s10, 0x1001a
+; GFX8-NEXT:    s_bfe_u32 s34, s10, 0x1001c
+; GFX8-NEXT:    s_bfe_u32 s10, s10, 0x1001e
+; GFX8-NEXT:    v_mov_b32_e32 v0, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0xf0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0xe0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 0xd0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0xd0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s34
-; GFX8-NEXT:    v_mov_b32_e32 v2, s8
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s12
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 0xc0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0xc0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s33
-; GFX8-NEXT:    v_mov_b32_e32 v2, s9
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s13
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 0xb0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0xb0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s31
-; GFX8-NEXT:    v_mov_b32_e32 v2, s10
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s14
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 0xa0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0xa0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s30
-; GFX8-NEXT:    v_mov_b32_e32 v2, s11
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s15
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 0x90
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0x90
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s28
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s29
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 0x80
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0x80
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s27
-; GFX8-NEXT:    v_mov_b32_e32 v2, s12
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s16
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 0x70
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0x70
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s26
-; GFX8-NEXT:    v_mov_b32_e32 v2, s13
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s17
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 0x60
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0x60
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s25
-; GFX8-NEXT:    v_mov_b32_e32 v2, s14
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s18
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 0x50
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 0x50
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s24
-; GFX8-NEXT:    v_mov_b32_e32 v2, s15
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s19
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 64
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 64
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s23
-; GFX8-NEXT:    v_mov_b32_e32 v2, s16
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s20
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NEXT:    s_add_u32 s6, s0, 48
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s22
-; GFX8-NEXT:    v_mov_b32_e32 v2, s17
-; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s21
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NEXT:    s_add_u32 s8, s0, 48
+; GFX8-NEXT:    v_mov_b32_e32 v0, s9
+; GFX8-NEXT:    s_addc_u32 s9, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s8
+; GFX8-NEXT:    v_mov_b32_e32 v5, s9
+; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX8-NEXT:    s_add_u32 s6, s0, 32
-; GFX8-NEXT:    v_mov_b32_e32 v0, s21
-; GFX8-NEXT:    v_mov_b32_e32 v2, s18
+; GFX8-NEXT:    v_mov_b32_e32 v0, s7
 ; GFX8-NEXT:    s_addc_u32 s7, s1, 0
-; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NEXT:    v_mov_b32_e32 v0, s20
-; GFX8-NEXT:    v_mov_b32_e32 v2, s19
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
@@ -8819,18 +8826,18 @@ define amdgpu_kernel void @constant_sextload_v32i1_to_v32i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_load_dword s2, s[2:3], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_lshr_b32 s44, s2, 30
-; GFX8-NEXT:    s_lshr_b32 s46, s2, 31
-; GFX8-NEXT:    s_lshr_b32 s48, s2, 28
-; GFX8-NEXT:    s_lshr_b32 s50, s2, 29
-; GFX8-NEXT:    s_lshr_b32 s52, s2, 26
-; GFX8-NEXT:    s_lshr_b32 s54, s2, 27
-; GFX8-NEXT:    s_lshr_b32 s56, s2, 24
-; GFX8-NEXT:    s_lshr_b32 s58, s2, 25
-; GFX8-NEXT:    s_lshr_b32 s60, s2, 22
-; GFX8-NEXT:    s_lshr_b32 s62, s2, 23
-; GFX8-NEXT:    s_lshr_b32 s64, s2, 20
-; GFX8-NEXT:    s_lshr_b32 s66, s2, 21
+; GFX8-NEXT:    s_lshr_b32 s46, s2, 30
+; GFX8-NEXT:    s_lshr_b32 s50, s2, 31
+; GFX8-NEXT:    s_lshr_b32 s52, s2, 28
+; GFX8-NEXT:    s_lshr_b32 s54, s2, 29
+; GFX8-NEXT:    s_lshr_b32 s56, s2, 26
+; GFX8-NEXT:    s_lshr_b32 s58, s2, 27
+; GFX8-NEXT:    s_lshr_b32 s60, s2, 24
+; GFX8-NEXT:    s_lshr_b32 s62, s2, 25
+; GFX8-NEXT:    s_lshr_b32 s64, s2, 22
+; GFX8-NEXT:    s_lshr_b32 s66, s2, 23
+; GFX8-NEXT:    s_lshr_b32 s48, s2, 20
+; GFX8-NEXT:    s_lshr_b32 s44, s2, 21
 ; GFX8-NEXT:    s_lshr_b32 s42, s2, 18
 ; GFX8-NEXT:    s_lshr_b32 s40, s2, 19
 ; GFX8-NEXT:    s_lshr_b32 s38, s2, 16
@@ -8870,6 +8877,8 @@ define amdgpu_kernel void @constant_sextload_v32i1_to_v32i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[40:41], s[40:41], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[42:43], s[42:43], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[44:45], s[44:45], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[48:49], s[48:49], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[66:67], s[66:67], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[64:65], s[64:65], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[62:63], s[62:63], 0x10000
@@ -8879,82 +8888,78 @@ define amdgpu_kernel void @constant_sextload_v32i1_to_v32i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_bfe_i64 s[54:55], s[54:55], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[52:53], s[52:53], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[50:51], s[50:51], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[48:49], s[48:49], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[46:47], s[46:47], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[44:45], s[44:45], 0x10000
-; GFX8-NEXT:    v_mov_b32_e32 v0, s44
-; GFX8-NEXT:    s_add_u32 s44, s0, 0xf0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s45
-; GFX8-NEXT:    s_addc_u32 s45, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v4, s44
-; GFX8-NEXT:    v_mov_b32_e32 v5, s45
-; GFX8-NEXT:    s_add_u32 s44, s0, 0xe0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s46
-; GFX8-NEXT:    v_mov_b32_e32 v3, s47
-; GFX8-NEXT:    s_addc_u32 s45, s1, 0
-; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s44
-; GFX8-NEXT:    v_mov_b32_e32 v5, s45
-; GFX8-NEXT:    s_add_u32 s44, s0, 0xd0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s48
-; GFX8-NEXT:    v_mov_b32_e32 v1, s49
+; GFX8-NEXT:    s_add_u32 s33, s0, 0xf0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s46
+; GFX8-NEXT:    s_addc_u32 s46, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0xe0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s47
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s50
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s51
-; GFX8-NEXT:    s_addc_u32 s45, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s46
+; GFX8-NEXT:    s_addc_u32 s46, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s44
-; GFX8-NEXT:    v_mov_b32_e32 v5, s45
-; GFX8-NEXT:    s_add_u32 s44, s0, 0xc0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0xd0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s52
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s53
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s54
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s55
-; GFX8-NEXT:    s_addc_u32 s45, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s46
+; GFX8-NEXT:    s_addc_u32 s46, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s44
-; GFX8-NEXT:    v_mov_b32_e32 v5, s45
-; GFX8-NEXT:    s_add_u32 s44, s0, 0xb0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0xc0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s56
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s57
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s58
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s59
-; GFX8-NEXT:    s_addc_u32 s45, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s46
+; GFX8-NEXT:    s_addc_u32 s46, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s44
-; GFX8-NEXT:    v_mov_b32_e32 v5, s45
-; GFX8-NEXT:    s_add_u32 s44, s0, 0xa0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0xb0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s60
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s61
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s62
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s63
-; GFX8-NEXT:    s_addc_u32 s45, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s46
+; GFX8-NEXT:    s_addc_u32 s46, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s44
+; GFX8-NEXT:    v_mov_b32_e32 v4, s33
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s64
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s65
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s66
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s67
-; GFX8-NEXT:    v_mov_b32_e32 v5, s45
+; GFX8-NEXT:    v_mov_b32_e32 v5, s46
+; GFX8-NEXT:    s_add_u32 s33, s0, 0xa0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s33
+; GFX8-NEXT:    v_mov_b32_e32 v2, s44
+; GFX8-NEXT:    s_addc_u32 s44, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s48
+; GFX8-NEXT:    v_mov_b32_e32 v1, s49
+; GFX8-NEXT:    v_mov_b32_e32 v3, s45
+; GFX8-NEXT:    v_mov_b32_e32 v5, s44
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x90
+; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NEXT:    v_mov_b32_e32 v4, s33
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s40
-; GFX8-NEXT:    s_add_u32 s40, s0, 0x90
-; GFX8-NEXT:    v_mov_b32_e32 v3, s41
-; GFX8-NEXT:    s_addc_u32 s41, s1, 0
+; GFX8-NEXT:    s_addc_u32 s40, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s42
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s43
-; GFX8-NEXT:    v_mov_b32_e32 v4, s40
-; GFX8-NEXT:    v_mov_b32_e32 v5, s41
+; GFX8-NEXT:    v_mov_b32_e32 v3, s41
+; GFX8-NEXT:    v_mov_b32_e32 v5, s40
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x80
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s33
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s36
-; GFX8-NEXT:    s_add_u32 s36, s0, 0x80
-; GFX8-NEXT:    v_mov_b32_e32 v3, s37
-; GFX8-NEXT:    s_addc_u32 s37, s1, 0
+; GFX8-NEXT:    s_addc_u32 s36, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s38
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s39
-; GFX8-NEXT:    v_mov_b32_e32 v4, s36
-; GFX8-NEXT:    v_mov_b32_e32 v5, s37
+; GFX8-NEXT:    v_mov_b32_e32 v3, s37
+; GFX8-NEXT:    v_mov_b32_e32 v5, s36
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s30
@@ -9683,270 +9688,275 @@ define amdgpu_kernel void @constant_zextload_v64i1_to_v64i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dwordx2 s[42:43], s[2:3], 0x0
+; GFX8-NEXT:    s_load_dwordx2 s[46:47], s[2:3], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_lshr_b32 s44, s43, 31
-; GFX8-NEXT:    s_bfe_u32 s45, s43, 0x1001d
-; GFX8-NEXT:    s_bfe_u32 s46, s43, 0x1001b
-; GFX8-NEXT:    s_bfe_u32 s47, s43, 0x10019
-; GFX8-NEXT:    s_bfe_u32 s48, s43, 0x10017
-; GFX8-NEXT:    s_bfe_u32 s49, s43, 0x10013
-; GFX8-NEXT:    s_bfe_u32 s50, s43, 0x10011
-; GFX8-NEXT:    s_bfe_u32 s51, s43, 0x1000f
-; GFX8-NEXT:    s_bfe_u32 s52, s43, 0x1000d
-; GFX8-NEXT:    s_bfe_u32 s53, s43, 0x1000b
-; GFX8-NEXT:    s_bfe_u32 s40, s43, 0x10009
-; GFX8-NEXT:    s_bfe_u32 s38, s43, 0x10007
-; GFX8-NEXT:    s_bfe_u32 s37, s43, 0x10005
-; GFX8-NEXT:    s_bfe_u32 s35, s43, 0x10003
-; GFX8-NEXT:    s_bfe_u32 s33, s43, 0x10001
-; GFX8-NEXT:    s_lshr_b32 s30, s42, 31
-; GFX8-NEXT:    s_bfe_u32 s28, s42, 0x1001d
-; GFX8-NEXT:    s_bfe_u32 s26, s42, 0x1001b
-; GFX8-NEXT:    s_bfe_u32 s25, s42, 0x10019
-; GFX8-NEXT:    s_bfe_u32 s22, s42, 0x10017
-; GFX8-NEXT:    s_bfe_u32 s19, s42, 0x10013
-; GFX8-NEXT:    s_bfe_u32 s17, s42, 0x10011
-; GFX8-NEXT:    s_bfe_u32 s15, s42, 0x1000f
-; GFX8-NEXT:    s_bfe_u32 s13, s42, 0x1000d
-; GFX8-NEXT:    s_bfe_u32 s12, s42, 0x1000b
-; GFX8-NEXT:    s_bfe_u32 s10, s42, 0x10009
-; GFX8-NEXT:    s_bfe_u32 s8, s42, 0x10007
-; GFX8-NEXT:    s_bfe_u32 s6, s42, 0x10005
-; GFX8-NEXT:    s_bfe_u32 s4, s42, 0x10003
-; GFX8-NEXT:    s_bfe_u32 s2, s42, 0x10001
-; GFX8-NEXT:    s_and_b32 s3, s42, 1
-; GFX8-NEXT:    s_bfe_u32 s5, s42, 0x10002
-; GFX8-NEXT:    s_bfe_u32 s7, s42, 0x10004
-; GFX8-NEXT:    s_bfe_u32 s9, s42, 0x10006
-; GFX8-NEXT:    s_bfe_u32 s11, s42, 0x10008
-; GFX8-NEXT:    s_bfe_u32 s14, s42, 0x1000a
-; GFX8-NEXT:    s_bfe_u32 s16, s42, 0x1000c
-; GFX8-NEXT:    s_bfe_u32 s18, s42, 0x1000e
-; GFX8-NEXT:    s_bfe_u32 s20, s42, 0x10010
-; GFX8-NEXT:    s_bfe_u32 s21, s42, 0x10012
-; GFX8-NEXT:    s_bfe_u32 s23, s42, 0x10014
-; GFX8-NEXT:    s_bfe_u32 s24, s42, 0x10015
-; GFX8-NEXT:    s_bfe_u32 s27, s42, 0x10016
-; GFX8-NEXT:    s_bfe_u32 s29, s42, 0x10018
-; GFX8-NEXT:    s_bfe_u32 s31, s42, 0x1001a
-; GFX8-NEXT:    s_bfe_u32 s34, s42, 0x1001c
-; GFX8-NEXT:    s_bfe_u32 s36, s42, 0x1001e
-; GFX8-NEXT:    s_and_b32 s39, s43, 1
-; GFX8-NEXT:    s_bfe_u32 s41, s43, 0x10002
-; GFX8-NEXT:    s_bfe_u32 s54, s43, 0x10004
-; GFX8-NEXT:    s_bfe_u32 s55, s43, 0x10006
-; GFX8-NEXT:    s_bfe_u32 s56, s43, 0x10008
-; GFX8-NEXT:    s_bfe_u32 s57, s43, 0x1000a
-; GFX8-NEXT:    s_bfe_u32 s58, s43, 0x1000c
-; GFX8-NEXT:    s_bfe_u32 s59, s43, 0x1000e
-; GFX8-NEXT:    s_bfe_u32 s60, s43, 0x10010
-; GFX8-NEXT:    s_bfe_u32 s61, s43, 0x10012
-; GFX8-NEXT:    s_bfe_u32 s62, s43, 0x10016
-; GFX8-NEXT:    s_bfe_u32 s63, s43, 0x10018
-; GFX8-NEXT:    s_bfe_u32 s64, s43, 0x1001a
-; GFX8-NEXT:    s_bfe_u32 s65, s43, 0x1001c
-; GFX8-NEXT:    s_bfe_u32 s66, s43, 0x1001e
-; GFX8-NEXT:    s_bfe_u32 s42, s43, 0x10015
-; GFX8-NEXT:    s_bfe_u32 s43, s43, 0x10014
-; GFX8-NEXT:    v_mov_b32_e32 v2, s42
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x1a0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s43
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x1f0
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
-; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x1e0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s66
-; GFX8-NEXT:    v_mov_b32_e32 v2, s44
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    s_lshr_b32 s48, s47, 31
+; GFX8-NEXT:    s_bfe_u32 s49, s47, 0x1001d
+; GFX8-NEXT:    s_bfe_u32 s50, s47, 0x1001b
+; GFX8-NEXT:    s_bfe_u32 s51, s47, 0x10019
+; GFX8-NEXT:    s_bfe_u32 s52, s47, 0x10017
+; GFX8-NEXT:    s_bfe_u32 s53, s47, 0x10013
+; GFX8-NEXT:    s_bfe_u32 s54, s47, 0x10011
+; GFX8-NEXT:    s_bfe_u32 s55, s47, 0x1000f
+; GFX8-NEXT:    s_bfe_u32 s44, s47, 0x1000d
+; GFX8-NEXT:    s_bfe_u32 s43, s47, 0x1000b
+; GFX8-NEXT:    s_bfe_u32 s41, s47, 0x10009
+; GFX8-NEXT:    s_bfe_u32 s39, s47, 0x10007
+; GFX8-NEXT:    s_bfe_u32 s37, s47, 0x10005
+; GFX8-NEXT:    s_bfe_u32 s35, s47, 0x10003
+; GFX8-NEXT:    s_bfe_u32 s33, s47, 0x10001
+; GFX8-NEXT:    s_lshr_b32 s30, s46, 31
+; GFX8-NEXT:    s_bfe_u32 s28, s46, 0x1001d
+; GFX8-NEXT:    s_bfe_u32 s27, s46, 0x1001b
+; GFX8-NEXT:    s_bfe_u32 s25, s46, 0x10019
+; GFX8-NEXT:    s_bfe_u32 s22, s46, 0x10017
+; GFX8-NEXT:    s_bfe_u32 s19, s46, 0x10013
+; GFX8-NEXT:    s_bfe_u32 s17, s46, 0x10011
+; GFX8-NEXT:    s_bfe_u32 s15, s46, 0x1000f
+; GFX8-NEXT:    s_bfe_u32 s14, s46, 0x1000d
+; GFX8-NEXT:    s_bfe_u32 s12, s46, 0x1000b
+; GFX8-NEXT:    s_bfe_u32 s10, s46, 0x10009
+; GFX8-NEXT:    s_bfe_u32 s8, s46, 0x10007
+; GFX8-NEXT:    s_bfe_u32 s6, s46, 0x10005
+; GFX8-NEXT:    s_bfe_u32 s4, s46, 0x10003
+; GFX8-NEXT:    s_bfe_u32 s2, s46, 0x10001
+; GFX8-NEXT:    s_and_b32 s3, s46, 1
+; GFX8-NEXT:    s_bfe_u32 s5, s46, 0x10002
+; GFX8-NEXT:    s_bfe_u32 s7, s46, 0x10004
+; GFX8-NEXT:    s_bfe_u32 s9, s46, 0x10006
+; GFX8-NEXT:    s_bfe_u32 s11, s46, 0x10008
+; GFX8-NEXT:    s_bfe_u32 s13, s46, 0x1000a
+; GFX8-NEXT:    s_bfe_u32 s16, s46, 0x1000c
+; GFX8-NEXT:    s_bfe_u32 s18, s46, 0x1000e
+; GFX8-NEXT:    s_bfe_u32 s20, s46, 0x10010
+; GFX8-NEXT:    s_bfe_u32 s21, s46, 0x10012
+; GFX8-NEXT:    s_bfe_u32 s23, s46, 0x10014
+; GFX8-NEXT:    s_bfe_u32 s24, s46, 0x10015
+; GFX8-NEXT:    s_bfe_u32 s26, s46, 0x10016
+; GFX8-NEXT:    s_bfe_u32 s29, s46, 0x10018
+; GFX8-NEXT:    s_bfe_u32 s31, s46, 0x1001a
+; GFX8-NEXT:    s_bfe_u32 s34, s46, 0x1001c
+; GFX8-NEXT:    s_bfe_u32 s36, s46, 0x1001e
+; GFX8-NEXT:    s_and_b32 s38, s47, 1
+; GFX8-NEXT:    s_bfe_u32 s40, s47, 0x10002
+; GFX8-NEXT:    s_bfe_u32 s42, s47, 0x10004
+; GFX8-NEXT:    s_bfe_u32 s45, s47, 0x10006
+; GFX8-NEXT:    s_bfe_u32 s46, s47, 0x10008
+; GFX8-NEXT:    s_bfe_u32 s56, s47, 0x1000a
+; GFX8-NEXT:    s_bfe_u32 s57, s47, 0x1000c
+; GFX8-NEXT:    s_bfe_u32 s58, s47, 0x1000e
+; GFX8-NEXT:    s_bfe_u32 s59, s47, 0x10010
+; GFX8-NEXT:    s_bfe_u32 s60, s47, 0x10012
+; GFX8-NEXT:    s_bfe_u32 s61, s47, 0x10016
+; GFX8-NEXT:    s_bfe_u32 s62, s47, 0x10018
+; GFX8-NEXT:    s_bfe_u32 s63, s47, 0x1001a
+; GFX8-NEXT:    s_bfe_u32 s64, s47, 0x1001c
+; GFX8-NEXT:    s_bfe_u32 s65, s47, 0x1001e
+; GFX8-NEXT:    s_bfe_u32 s66, s47, 0x10015
+; GFX8-NEXT:    s_bfe_u32 s47, s47, 0x10014
+; GFX8-NEXT:    v_mov_b32_e32 v0, s47
+; GFX8-NEXT:    s_add_u32 s47, s0, 0x1a0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s66
+; GFX8-NEXT:    s_addc_u32 s66, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s47
+; GFX8-NEXT:    v_mov_b32_e32 v5, s66
+; GFX8-NEXT:    s_add_u32 s47, s0, 0x1f0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x1d0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s47
+; GFX8-NEXT:    v_mov_b32_e32 v2, s48
+; GFX8-NEXT:    s_addc_u32 s48, s1, 0
+; GFX8-NEXT:    s_add_u32 s47, s0, 0x1e0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s65
-; GFX8-NEXT:    v_mov_b32_e32 v2, s45
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s48
+; GFX8-NEXT:    s_addc_u32 s48, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x1c0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s47
+; GFX8-NEXT:    s_add_u32 s47, s0, 0x1d0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s64
-; GFX8-NEXT:    v_mov_b32_e32 v2, s46
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s49
+; GFX8-NEXT:    v_mov_b32_e32 v5, s48
+; GFX8-NEXT:    s_addc_u32 s48, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x1b0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s47
+; GFX8-NEXT:    s_add_u32 s47, s0, 0x1c0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s63
-; GFX8-NEXT:    v_mov_b32_e32 v2, s47
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s50
+; GFX8-NEXT:    v_mov_b32_e32 v5, s48
+; GFX8-NEXT:    s_addc_u32 s48, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x190
+; GFX8-NEXT:    v_mov_b32_e32 v4, s47
+; GFX8-NEXT:    s_add_u32 s47, s0, 0x1b0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s62
-; GFX8-NEXT:    v_mov_b32_e32 v2, s48
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s51
+; GFX8-NEXT:    v_mov_b32_e32 v5, s48
+; GFX8-NEXT:    s_addc_u32 s48, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x180
+; GFX8-NEXT:    v_mov_b32_e32 v4, s47
+; GFX8-NEXT:    s_add_u32 s47, s0, 0x190
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s61
-; GFX8-NEXT:    v_mov_b32_e32 v2, s49
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s52
+; GFX8-NEXT:    v_mov_b32_e32 v5, s48
+; GFX8-NEXT:    s_addc_u32 s48, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x170
+; GFX8-NEXT:    v_mov_b32_e32 v4, s47
+; GFX8-NEXT:    s_add_u32 s47, s0, 0x180
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s60
-; GFX8-NEXT:    v_mov_b32_e32 v2, s50
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s53
+; GFX8-NEXT:    v_mov_b32_e32 v5, s48
+; GFX8-NEXT:    s_addc_u32 s48, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x160
+; GFX8-NEXT:    v_mov_b32_e32 v4, s47
+; GFX8-NEXT:    s_add_u32 s47, s0, 0x170
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s59
-; GFX8-NEXT:    v_mov_b32_e32 v2, s51
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s54
+; GFX8-NEXT:    v_mov_b32_e32 v5, s48
+; GFX8-NEXT:    s_addc_u32 s48, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x150
+; GFX8-NEXT:    v_mov_b32_e32 v4, s47
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s58
-; GFX8-NEXT:    v_mov_b32_e32 v2, s52
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s55
+; GFX8-NEXT:    v_mov_b32_e32 v5, s48
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x140
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s44
+; GFX8-NEXT:    s_add_u32 s44, s0, 0x160
+; GFX8-NEXT:    s_addc_u32 s47, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s57
-; GFX8-NEXT:    v_mov_b32_e32 v2, s53
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s44
+; GFX8-NEXT:    v_mov_b32_e32 v5, s47
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x130
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s43
+; GFX8-NEXT:    s_add_u32 s43, s0, 0x150
+; GFX8-NEXT:    s_addc_u32 s44, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s56
-; GFX8-NEXT:    v_mov_b32_e32 v2, s40
-; GFX8-NEXT:    s_addc_u32 s43, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s43
+; GFX8-NEXT:    v_mov_b32_e32 v5, s44
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s42, s0, 0x120
-; GFX8-NEXT:    v_mov_b32_e32 v0, s55
-; GFX8-NEXT:    v_mov_b32_e32 v2, s38
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s41
+; GFX8-NEXT:    s_add_u32 s41, s0, 0x140
 ; GFX8-NEXT:    s_addc_u32 s43, s1, 0
-; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s42
-; GFX8-NEXT:    v_mov_b32_e32 v0, s54
-; GFX8-NEXT:    v_mov_b32_e32 v2, s37
+; GFX8-NEXT:    v_mov_b32_e32 v0, s46
+; GFX8-NEXT:    v_mov_b32_e32 v4, s41
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s43
-; GFX8-NEXT:    s_add_u32 s40, s0, 0x110
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s41
+; GFX8-NEXT:    v_mov_b32_e32 v2, s39
+; GFX8-NEXT:    s_add_u32 s39, s0, 0x130
 ; GFX8-NEXT:    s_addc_u32 s41, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s35
-; GFX8-NEXT:    v_mov_b32_e32 v4, s40
+; GFX8-NEXT:    v_mov_b32_e32 v0, s45
+; GFX8-NEXT:    v_mov_b32_e32 v4, s39
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s41
-; GFX8-NEXT:    s_add_u32 s38, s0, 0x100
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s39
+; GFX8-NEXT:    v_mov_b32_e32 v2, s37
+; GFX8-NEXT:    s_add_u32 s37, s0, 0x120
 ; GFX8-NEXT:    s_addc_u32 s39, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s33
-; GFX8-NEXT:    v_mov_b32_e32 v4, s38
+; GFX8-NEXT:    v_mov_b32_e32 v0, s42
+; GFX8-NEXT:    v_mov_b32_e32 v4, s37
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s39
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s36
-; GFX8-NEXT:    s_add_u32 s36, s0, 0xf0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s35
+; GFX8-NEXT:    s_add_u32 s35, s0, 0x110
 ; GFX8-NEXT:    s_addc_u32 s37, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s30
-; GFX8-NEXT:    v_mov_b32_e32 v4, s36
+; GFX8-NEXT:    v_mov_b32_e32 v0, s40
+; GFX8-NEXT:    v_mov_b32_e32 v4, s35
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s37
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s34
-; GFX8-NEXT:    s_add_u32 s34, s0, 0xe0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x100
 ; GFX8-NEXT:    s_addc_u32 s35, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s28
-; GFX8-NEXT:    v_mov_b32_e32 v4, s34
+; GFX8-NEXT:    v_mov_b32_e32 v0, s38
+; GFX8-NEXT:    v_mov_b32_e32 v4, s33
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s35
-; GFX8-NEXT:    s_add_u32 s30, s0, 0xd0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s31
-; GFX8-NEXT:    s_addc_u32 s31, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s26
+; GFX8-NEXT:    v_mov_b32_e32 v2, s30
+; GFX8-NEXT:    s_add_u32 s30, s0, 0xf0
+; GFX8-NEXT:    s_addc_u32 s33, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s36
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s30
-; GFX8-NEXT:    v_mov_b32_e32 v5, s31
-; GFX8-NEXT:    s_add_u32 s28, s0, 0xc0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s33
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s29
-; GFX8-NEXT:    s_addc_u32 s29, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s25
+; GFX8-NEXT:    v_mov_b32_e32 v2, s28
+; GFX8-NEXT:    s_add_u32 s28, s0, 0xe0
+; GFX8-NEXT:    s_addc_u32 s30, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s34
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s28
-; GFX8-NEXT:    v_mov_b32_e32 v5, s29
-; GFX8-NEXT:    s_add_u32 s26, s0, 0xb0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s30
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s27
+; GFX8-NEXT:    v_mov_b32_e32 v2, s27
+; GFX8-NEXT:    s_add_u32 s27, s0, 0xd0
+; GFX8-NEXT:    s_addc_u32 s28, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s31
+; GFX8-NEXT:    v_mov_b32_e32 v4, s27
+; GFX8-NEXT:    v_mov_b32_e32 v5, s28
+; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s25
+; GFX8-NEXT:    s_add_u32 s25, s0, 0xc0
 ; GFX8-NEXT:    s_addc_u32 s27, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s22
-; GFX8-NEXT:    v_mov_b32_e32 v4, s26
+; GFX8-NEXT:    v_mov_b32_e32 v0, s29
+; GFX8-NEXT:    v_mov_b32_e32 v4, s25
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s27
-; GFX8-NEXT:    s_add_u32 s22, s0, 0xa0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s23
-; GFX8-NEXT:    s_addc_u32 s23, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s22
+; GFX8-NEXT:    s_add_u32 s22, s0, 0xb0
+; GFX8-NEXT:    s_addc_u32 s25, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s26
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s22
-; GFX8-NEXT:    v_mov_b32_e32 v5, s23
-; GFX8-NEXT:    s_add_u32 s22, s0, 0x90
-; GFX8-NEXT:    v_mov_b32_e32 v2, s24
-; GFX8-NEXT:    s_addc_u32 s23, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s25
+; GFX8-NEXT:    s_add_u32 s22, s0, 0xa0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s22
-; GFX8-NEXT:    v_mov_b32_e32 v0, s21
-; GFX8-NEXT:    v_mov_b32_e32 v2, s19
+; GFX8-NEXT:    v_mov_b32_e32 v0, s23
+; GFX8-NEXT:    s_addc_u32 s23, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s24
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s23
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s20
-; GFX8-NEXT:    s_add_u32 s20, s0, 0x80
+; GFX8-NEXT:    v_mov_b32_e32 v2, s19
+; GFX8-NEXT:    s_add_u32 s19, s0, 0x90
+; GFX8-NEXT:    v_mov_b32_e32 v0, s21
 ; GFX8-NEXT:    s_addc_u32 s21, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s17
-; GFX8-NEXT:    v_mov_b32_e32 v4, s20
+; GFX8-NEXT:    v_mov_b32_e32 v4, s19
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s21
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s18
-; GFX8-NEXT:    s_add_u32 s18, s0, 0x70
+; GFX8-NEXT:    v_mov_b32_e32 v2, s17
+; GFX8-NEXT:    s_add_u32 s17, s0, 0x80
 ; GFX8-NEXT:    s_addc_u32 s19, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s15
-; GFX8-NEXT:    v_mov_b32_e32 v4, s18
+; GFX8-NEXT:    v_mov_b32_e32 v0, s20
+; GFX8-NEXT:    v_mov_b32_e32 v4, s17
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s19
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s16
-; GFX8-NEXT:    s_add_u32 s16, s0, 0x60
+; GFX8-NEXT:    v_mov_b32_e32 v2, s15
+; GFX8-NEXT:    s_add_u32 s15, s0, 0x70
 ; GFX8-NEXT:    s_addc_u32 s17, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s13
-; GFX8-NEXT:    v_mov_b32_e32 v4, s16
+; GFX8-NEXT:    v_mov_b32_e32 v0, s18
+; GFX8-NEXT:    v_mov_b32_e32 v4, s15
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s17
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0x60
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s16
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s12
 ; GFX8-NEXT:    s_add_u32 s12, s0, 0x50
+; GFX8-NEXT:    v_mov_b32_e32 v0, s13
 ; GFX8-NEXT:    s_addc_u32 s13, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s14
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s12
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s13
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -11004,24 +11014,21 @@ define amdgpu_kernel void @constant_sextload_v64i1_to_v64i64(ptr addrspace(1) %o
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX8-NEXT:    s_mov_b32 s69, 0
-; GFX8-NEXT:    s_mov_b32 s39, s69
-; GFX8-NEXT:    ; implicit-def: $vgpr62 : SGPR spill to VGPR lane
+; GFX8-NEXT:    ; implicit-def: $vgpr60 : SGPR spill to VGPR lane
+; GFX8-NEXT:    s_mov_b32 s13, s69
 ; GFX8-NEXT:    s_mov_b32 s73, s69
 ; GFX8-NEXT:    s_mov_b32 s49, s69
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
 ; GFX8-NEXT:    s_mov_b32 s75, s69
-; GFX8-NEXT:    s_mov_b32 s43, s69
+; GFX8-NEXT:    s_mov_b32 s25, s69
 ; GFX8-NEXT:    s_mov_b32 s63, s69
-; GFX8-NEXT:    s_mov_b32 s29, s69
+; GFX8-NEXT:    s_mov_b32 s7, s69
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_lshr_b32 s38, s3, 19
 ; GFX8-NEXT:    s_lshr_b32 s4, s3, 9
-; GFX8-NEXT:    s_lshr_b32 s40, s3, 15
-; GFX8-NEXT:    s_lshr_b32 s44, s3, 12
-; GFX8-NEXT:    s_lshr_b32 s54, s3, 11
-; GFX8-NEXT:    v_writelane_b32 v62, s4, 0
-; GFX8-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x10000
+; GFX8-NEXT:    s_lshr_b32 s12, s3, 19
+; GFX8-NEXT:    s_lshr_b32 s38, s3, 12
+; GFX8-NEXT:    v_writelane_b32 v60, s4, 0
 ; GFX8-NEXT:    s_lshr_b32 s64, s3, 30
 ; GFX8-NEXT:    s_lshr_b32 s60, s3, 31
 ; GFX8-NEXT:    s_lshr_b32 s66, s3, 28
@@ -11033,21 +11040,21 @@ define amdgpu_kernel void @constant_sextload_v64i1_to_v64i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_lshr_b32 s72, s3, 22
 ; GFX8-NEXT:    s_lshr_b32 s48, s3, 23
 ; GFX8-NEXT:    s_lshr_b32 s74, s3, 20
-; GFX8-NEXT:    s_lshr_b32 s42, s3, 21
+; GFX8-NEXT:    s_lshr_b32 s24, s3, 21
 ; GFX8-NEXT:    s_lshr_b32 s62, s3, 18
-; GFX8-NEXT:    s_lshr_b32 s28, s3, 16
-; GFX8-NEXT:    s_lshr_b32 s34, s3, 17
-; GFX8-NEXT:    s_mov_b32 s35, s69
-; GFX8-NEXT:    s_lshr_b32 s36, s3, 14
-; GFX8-NEXT:    s_lshr_b32 s18, s3, 8
-; GFX8-NEXT:    v_writelane_b32 v62, s5, 1
-; GFX8-NEXT:    v_mov_b32_e32 v26, s38
-; GFX8-NEXT:    v_mov_b32_e32 v27, s39
-; GFX8-NEXT:    s_bfe_i64 s[38:39], s[54:55], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[4:5], s[44:45], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[40:41], s[40:41], 0x10000
-; GFX8-NEXT:    s_lshr_b32 s46, s3, 13
-; GFX8-NEXT:    s_mov_b32 s23, s69
+; GFX8-NEXT:    s_lshr_b32 s6, s3, 16
+; GFX8-NEXT:    s_lshr_b32 s8, s3, 17
+; GFX8-NEXT:    s_mov_b32 s9, s69
+; GFX8-NEXT:    s_lshr_b32 s10, s3, 14
+; GFX8-NEXT:    s_lshr_b32 s20, s3, 15
+; GFX8-NEXT:    s_lshr_b32 s42, s3, 13
+; GFX8-NEXT:    s_lshr_b32 s54, s3, 11
+; GFX8-NEXT:    s_lshr_b32 s14, s3, 8
+; GFX8-NEXT:    v_writelane_b32 v60, s5, 1
+; GFX8-NEXT:    s_bfe_i64 s[12:13], s[12:13], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[4:5], s[38:39], 0x10000
+; GFX8-NEXT:    s_lshr_b32 s50, s3, 10
+; GFX8-NEXT:    s_mov_b32 s37, s69
 ; GFX8-NEXT:    s_bfe_i64 s[74:75], s[74:75], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[72:73], s[72:73], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[68:69], s[68:69], 0x10000
@@ -11060,109 +11067,112 @@ define amdgpu_kernel void @constant_sextload_v64i1_to_v64i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_bfe_i64 s[56:57], s[56:57], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[52:53], s[52:53], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[48:49], s[48:49], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[42:43], s[42:43], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[36:37], s[36:37], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[34:35], s[34:35], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[28:29], s[28:29], 0x10000
-; GFX8-NEXT:    v_mov_b32_e32 v42, s40
+; GFX8-NEXT:    s_bfe_i64 s[24:25], s[24:25], 0x10000
+; GFX8-NEXT:    v_mov_b32_e32 v26, s12
+; GFX8-NEXT:    v_mov_b32_e32 v27, s13
+; GFX8-NEXT:    s_bfe_i64 s[12:13], s[54:55], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[54:55], s[42:43], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[20:21], s[20:21], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[10:11], s[10:11], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[8:9], s[8:9], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[6:7], s[6:7], 0x10000
 ; GFX8-NEXT:    v_mov_b32_e32 v28, s4
-; GFX8-NEXT:    s_lshr_b32 s4, s2, 7
-; GFX8-NEXT:    s_lshr_b32 s40, s2, 4
-; GFX8-NEXT:    v_mov_b32_e32 v34, s38
-; GFX8-NEXT:    s_bfe_i64 s[18:19], s[18:19], 0x10000
-; GFX8-NEXT:    s_lshr_b32 s38, s2, 1
-; GFX8-NEXT:    s_lshr_b32 s50, s3, 10
-; GFX8-NEXT:    s_lshr_b32 s6, s3, 6
-; GFX8-NEXT:    s_lshr_b32 s8, s3, 7
-; GFX8-NEXT:    s_lshr_b32 s10, s3, 4
-; GFX8-NEXT:    s_lshr_b32 s12, s3, 5
-; GFX8-NEXT:    s_lshr_b32 s14, s3, 2
-; GFX8-NEXT:    s_lshr_b32 s16, s3, 3
-; GFX8-NEXT:    s_lshr_b32 s20, s3, 1
-; GFX8-NEXT:    s_mov_b32 s22, s3
-; GFX8-NEXT:    s_lshr_b32 s24, s2, 30
-; GFX8-NEXT:    s_lshr_b32 s26, s2, 31
-; GFX8-NEXT:    s_lshr_b32 s30, s2, 28
-; GFX8-NEXT:    v_mov_b32_e32 v12, s64
-; GFX8-NEXT:    v_mov_b32_e32 v4, s66
+; GFX8-NEXT:    v_mov_b32_e32 v29, s5
+; GFX8-NEXT:    s_bfe_i64 s[4:5], s[14:15], 0x10000
+; GFX8-NEXT:    v_readlane_b32 s14, v60, 0
+; GFX8-NEXT:    s_lshr_b32 s16, s3, 6
+; GFX8-NEXT:    s_lshr_b32 s18, s3, 7
+; GFX8-NEXT:    s_lshr_b32 s22, s3, 4
+; GFX8-NEXT:    s_lshr_b32 s26, s3, 5
+; GFX8-NEXT:    s_lshr_b32 s28, s3, 2
+; GFX8-NEXT:    s_lshr_b32 s30, s3, 3
+; GFX8-NEXT:    s_lshr_b32 s34, s3, 1
+; GFX8-NEXT:    s_mov_b32 s36, s3
+; GFX8-NEXT:    s_lshr_b32 s40, s2, 30
+; GFX8-NEXT:    s_lshr_b32 s44, s2, 31
+; GFX8-NEXT:    s_lshr_b32 s46, s2, 28
+; GFX8-NEXT:    v_mov_b32_e32 v16, s64
+; GFX8-NEXT:    v_mov_b32_e32 v8, s66
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s70
-; GFX8-NEXT:    v_mov_b32_e32 v8, s68
-; GFX8-NEXT:    v_mov_b32_e32 v16, s72
+; GFX8-NEXT:    v_mov_b32_e32 v4, s68
+; GFX8-NEXT:    v_mov_b32_e32 v12, s72
 ; GFX8-NEXT:    v_mov_b32_e32 v20, s74
-; GFX8-NEXT:    s_lshr_b32 s82, s2, 29
-; GFX8-NEXT:    s_lshr_b32 s84, s2, 26
+; GFX8-NEXT:    s_lshr_b32 s84, s2, 29
+; GFX8-NEXT:    s_lshr_b32 s82, s2, 26
 ; GFX8-NEXT:    v_mov_b32_e32 v24, s62
 ; GFX8-NEXT:    s_lshr_b32 s80, s2, 27
 ; GFX8-NEXT:    s_lshr_b32 s78, s2, 24
 ; GFX8-NEXT:    s_lshr_b32 s76, s2, 25
-; GFX8-NEXT:    v_mov_b32_e32 v14, s60
+; GFX8-NEXT:    v_mov_b32_e32 v18, s60
 ; GFX8-NEXT:    s_lshr_b32 s74, s2, 22
 ; GFX8-NEXT:    s_lshr_b32 s72, s2, 23
-; GFX8-NEXT:    v_mov_b32_e32 v6, s58
+; GFX8-NEXT:    v_mov_b32_e32 v10, s58
 ; GFX8-NEXT:    s_lshr_b32 s70, s2, 20
 ; GFX8-NEXT:    s_lshr_b32 s68, s2, 21
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s56
 ; GFX8-NEXT:    s_lshr_b32 s66, s2, 18
 ; GFX8-NEXT:    s_lshr_b32 s64, s2, 19
-; GFX8-NEXT:    v_mov_b32_e32 v10, s52
+; GFX8-NEXT:    v_mov_b32_e32 v6, s52
 ; GFX8-NEXT:    s_lshr_b32 s62, s2, 16
 ; GFX8-NEXT:    s_lshr_b32 s60, s2, 17
-; GFX8-NEXT:    v_mov_b32_e32 v18, s48
+; GFX8-NEXT:    v_mov_b32_e32 v14, s48
 ; GFX8-NEXT:    s_lshr_b32 s58, s2, 14
 ; GFX8-NEXT:    s_lshr_b32 s56, s2, 15
-; GFX8-NEXT:    v_mov_b32_e32 v22, s42
+; GFX8-NEXT:    v_mov_b32_e32 v22, s24
 ; GFX8-NEXT:    s_lshr_b32 s86, s2, 12
 ; GFX8-NEXT:    s_lshr_b32 s52, s2, 13
 ; GFX8-NEXT:    s_lshr_b32 vcc_lo, s2, 10
-; GFX8-NEXT:    s_bfe_i64 s[54:55], s[46:47], 0x10000
-; GFX8-NEXT:    v_mov_b32_e32 v36, s28
-; GFX8-NEXT:    s_lshr_b32 s48, s2, 11
-; GFX8-NEXT:    v_mov_b32_e32 v38, s34
-; GFX8-NEXT:    s_lshr_b32 s46, s2, 8
-; GFX8-NEXT:    v_mov_b32_e32 v40, s36
-; GFX8-NEXT:    s_lshr_b32 s44, s2, 9
-; GFX8-NEXT:    v_mov_b32_e32 v43, s41
-; GFX8-NEXT:    s_lshr_b32 s42, s2, 6
-; GFX8-NEXT:    v_mov_b32_e32 v29, s5
-; GFX8-NEXT:    s_lshr_b32 s36, s2, 5
-; GFX8-NEXT:    v_mov_b32_e32 v35, s39
-; GFX8-NEXT:    s_lshr_b32 s34, s2, 2
-; GFX8-NEXT:    v_mov_b32_e32 v44, s18
-; GFX8-NEXT:    v_mov_b32_e32 v45, s19
-; GFX8-NEXT:    s_lshr_b32 s28, s2, 3
-; GFX8-NEXT:    s_bfe_i64 s[18:19], s[2:3], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[2:3], s[38:39], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[38:39], s[40:41], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[40:41], s[4:5], 0x10000
-; GFX8-NEXT:    v_readlane_b32 s4, v62, 0
 ; GFX8-NEXT:    s_bfe_i64 s[50:51], s[50:51], 0x10000
-; GFX8-NEXT:    v_readlane_b32 s5, v62, 1
-; GFX8-NEXT:    v_mov_b32_e32 v13, s65
-; GFX8-NEXT:    v_mov_b32_e32 v15, s61
-; GFX8-NEXT:    v_mov_b32_e32 v5, s67
-; GFX8-NEXT:    v_mov_b32_e32 v7, s59
+; GFX8-NEXT:    v_mov_b32_e32 v36, s6
+; GFX8-NEXT:    s_lshr_b32 s48, s2, 11
+; GFX8-NEXT:    v_mov_b32_e32 v38, s8
+; GFX8-NEXT:    s_lshr_b32 s42, s2, 8
+; GFX8-NEXT:    v_mov_b32_e32 v40, s10
+; GFX8-NEXT:    s_lshr_b32 s38, s2, 9
+; GFX8-NEXT:    v_mov_b32_e32 v42, s20
+; GFX8-NEXT:    s_lshr_b32 s24, s2, 6
+; GFX8-NEXT:    s_lshr_b32 s20, s2, 7
+; GFX8-NEXT:    v_mov_b32_e32 v30, s54
+; GFX8-NEXT:    s_lshr_b32 s54, s2, 4
+; GFX8-NEXT:    s_lshr_b32 s10, s2, 5
+; GFX8-NEXT:    v_mov_b32_e32 v34, s12
+; GFX8-NEXT:    s_lshr_b32 s8, s2, 2
+; GFX8-NEXT:    s_lshr_b32 s6, s2, 3
+; GFX8-NEXT:    s_lshr_b32 s12, s2, 1
+; GFX8-NEXT:    v_readlane_b32 s15, v60, 1
+; GFX8-NEXT:    v_mov_b32_e32 v17, s65
+; GFX8-NEXT:    v_mov_b32_e32 v19, s61
+; GFX8-NEXT:    v_mov_b32_e32 v9, s67
+; GFX8-NEXT:    v_mov_b32_e32 v11, s59
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s71
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s57
-; GFX8-NEXT:    v_mov_b32_e32 v9, s69
-; GFX8-NEXT:    v_mov_b32_e32 v11, s53
-; GFX8-NEXT:    v_mov_b32_e32 v17, s73
-; GFX8-NEXT:    v_mov_b32_e32 v19, s49
+; GFX8-NEXT:    v_mov_b32_e32 v5, s69
+; GFX8-NEXT:    v_mov_b32_e32 v7, s53
+; GFX8-NEXT:    v_mov_b32_e32 v13, s73
+; GFX8-NEXT:    v_mov_b32_e32 v15, s49
 ; GFX8-NEXT:    v_mov_b32_e32 v21, s75
-; GFX8-NEXT:    v_mov_b32_e32 v23, s43
+; GFX8-NEXT:    v_mov_b32_e32 v23, s25
 ; GFX8-NEXT:    v_mov_b32_e32 v25, s63
-; GFX8-NEXT:    v_mov_b32_e32 v37, s29
-; GFX8-NEXT:    v_mov_b32_e32 v39, s35
-; GFX8-NEXT:    v_mov_b32_e32 v41, s37
-; GFX8-NEXT:    v_mov_b32_e32 v30, s54
+; GFX8-NEXT:    v_mov_b32_e32 v37, s7
+; GFX8-NEXT:    v_mov_b32_e32 v39, s9
+; GFX8-NEXT:    v_mov_b32_e32 v41, s11
+; GFX8-NEXT:    v_mov_b32_e32 v43, s21
 ; GFX8-NEXT:    v_mov_b32_e32 v31, s55
 ; GFX8-NEXT:    v_mov_b32_e32 v32, s50
 ; GFX8-NEXT:    v_mov_b32_e32 v33, s51
-; GFX8-NEXT:    s_bfe_i64 s[28:29], s[28:29], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[34:35], s[34:35], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[36:37], s[36:37], 0x10000
+; GFX8-NEXT:    v_mov_b32_e32 v35, s13
+; GFX8-NEXT:    v_mov_b32_e32 v44, s4
+; GFX8-NEXT:    v_mov_b32_e32 v45, s5
+; GFX8-NEXT:    s_bfe_i64 s[4:5], s[2:3], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[2:3], s[12:13], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[6:7], s[6:7], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[8:9], s[8:9], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[10:11], s[10:11], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[12:13], s[54:55], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[20:21], s[20:21], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[24:25], s[24:25], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[42:43], s[42:43], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[44:45], s[44:45], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[46:47], s[46:47], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[48:49], s[48:49], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[50:51], vcc, 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[52:53], s[52:53], 0x10000
@@ -11180,259 +11190,259 @@ define amdgpu_kernel void @constant_sextload_v64i1_to_v64i64(ptr addrspace(1) %o
 ; GFX8-NEXT:    s_bfe_i64 s[76:77], s[76:77], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[78:79], s[78:79], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[80:81], s[80:81], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[84:85], s[84:85], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[82:83], s[82:83], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[84:85], s[84:85], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[46:47], s[46:47], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[44:45], s[44:45], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[40:41], s[40:41], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[36:37], s[36:37], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[34:35], s[34:35], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[30:31], s[30:31], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[28:29], s[28:29], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[26:27], s[26:27], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[24:25], s[24:25], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[22:23], s[22:23], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[20:21], s[20:21], 0x10000
+; GFX8-NEXT:    s_bfe_i64 s[18:19], s[18:19], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[16:17], s[16:17], 0x10000
 ; GFX8-NEXT:    s_bfe_i64 s[14:15], s[14:15], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[12:13], s[12:13], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[10:11], s[10:11], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[8:9], s[8:9], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[6:7], s[6:7], 0x10000
-; GFX8-NEXT:    s_bfe_i64 s[4:5], s[4:5], 0x10000
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x1f0
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v46, s86
-; GFX8-NEXT:    v_mov_b32_e32 v47, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x1e0
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v48, s86
-; GFX8-NEXT:    v_mov_b32_e32 v49, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x1d0
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v50, s86
-; GFX8-NEXT:    v_mov_b32_e32 v51, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x1c0
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v52, s86
-; GFX8-NEXT:    v_mov_b32_e32 v53, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x1b0
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v54, s86
-; GFX8-NEXT:    v_mov_b32_e32 v55, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x1a0
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v56, s86
-; GFX8-NEXT:    v_mov_b32_e32 v57, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x190
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    flat_store_dwordx4 v[46:47], v[12:15]
-; GFX8-NEXT:    flat_store_dwordx4 v[48:49], v[4:7]
-; GFX8-NEXT:    v_mov_b32_e32 v12, s86
-; GFX8-NEXT:    v_mov_b32_e32 v13, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x180
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v14, s86
-; GFX8-NEXT:    v_mov_b32_e32 v15, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x170
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v58, s86
-; GFX8-NEXT:    v_mov_b32_e32 v59, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x160
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v6, s86
-; GFX8-NEXT:    v_mov_b32_e32 v7, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x150
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v48, s86
-; GFX8-NEXT:    v_mov_b32_e32 v49, s87
-; GFX8-NEXT:    s_add_u32 s86, s0, 0x140
-; GFX8-NEXT:    s_addc_u32 s87, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v46, s4
-; GFX8-NEXT:    s_add_u32 s4, s0, 0x130
-; GFX8-NEXT:    v_mov_b32_e32 v47, s5
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x1f0
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v46, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x1e0
+; GFX8-NEXT:    v_mov_b32_e32 v47, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v48, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x1d0
+; GFX8-NEXT:    v_mov_b32_e32 v49, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v50, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x1c0
+; GFX8-NEXT:    v_mov_b32_e32 v51, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v52, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x1b0
+; GFX8-NEXT:    v_mov_b32_e32 v53, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v54, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x1a0
+; GFX8-NEXT:    v_mov_b32_e32 v55, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v56, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x190
+; GFX8-NEXT:    v_mov_b32_e32 v57, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    flat_store_dwordx4 v[46:47], v[16:19]
+; GFX8-NEXT:    flat_store_dwordx4 v[48:49], v[8:11]
+; GFX8-NEXT:    v_mov_b32_e32 v16, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x180
+; GFX8-NEXT:    v_mov_b32_e32 v17, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v18, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x170
+; GFX8-NEXT:    v_mov_b32_e32 v19, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v58, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x160
+; GFX8-NEXT:    v_mov_b32_e32 v59, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v8, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x150
+; GFX8-NEXT:    v_mov_b32_e32 v9, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v10, s33
+; GFX8-NEXT:    s_add_u32 s33, s0, 0x140
+; GFX8-NEXT:    v_mov_b32_e32 v11, s86
+; GFX8-NEXT:    s_addc_u32 s86, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v46, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0x130
+; GFX8-NEXT:    v_mov_b32_e32 v47, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[50:51], v[0:3]
-; GFX8-NEXT:    flat_store_dwordx4 v[52:53], v[8:11]
-; GFX8-NEXT:    flat_store_dwordx4 v[54:55], v[16:19]
+; GFX8-NEXT:    flat_store_dwordx4 v[52:53], v[4:7]
+; GFX8-NEXT:    flat_store_dwordx4 v[54:55], v[12:15]
 ; GFX8-NEXT:    flat_store_dwordx4 v[56:57], v[20:23]
-; GFX8-NEXT:    flat_store_dwordx4 v[12:13], v[24:27]
-; GFX8-NEXT:    v_mov_b32_e32 v9, s5
-; GFX8-NEXT:    v_mov_b32_e32 v8, s4
-; GFX8-NEXT:    s_add_u32 s4, s0, 0x120
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v11, s5
-; GFX8-NEXT:    v_mov_b32_e32 v10, s4
-; GFX8-NEXT:    s_add_u32 s4, s0, 0x110
-; GFX8-NEXT:    v_mov_b32_e32 v4, s10
-; GFX8-NEXT:    v_mov_b32_e32 v5, s11
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v60, s86
-; GFX8-NEXT:    v_mov_b32_e32 v61, s87
-; GFX8-NEXT:    v_mov_b32_e32 v0, s6
-; GFX8-NEXT:    v_mov_b32_e32 v1, s7
-; GFX8-NEXT:    v_mov_b32_e32 v2, s8
-; GFX8-NEXT:    v_mov_b32_e32 v3, s9
-; GFX8-NEXT:    flat_store_dwordx4 v[6:7], v[28:31]
-; GFX8-NEXT:    v_mov_b32_e32 v6, s12
-; GFX8-NEXT:    v_mov_b32_e32 v7, s13
-; GFX8-NEXT:    flat_store_dwordx4 v[48:49], v[32:35]
-; GFX8-NEXT:    flat_store_dwordx4 v[60:61], v[44:47]
-; GFX8-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
-; GFX8-NEXT:    flat_store_dwordx4 v[10:11], v[4:7]
-; GFX8-NEXT:    v_mov_b32_e32 v0, s14
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0x100
-; GFX8-NEXT:    v_mov_b32_e32 v1, s15
-; GFX8-NEXT:    v_mov_b32_e32 v2, s16
-; GFX8-NEXT:    v_mov_b32_e32 v3, s17
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    flat_store_dwordx4 v[16:17], v[24:27]
+; GFX8-NEXT:    v_mov_b32_e32 v12, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0x120
+; GFX8-NEXT:    v_mov_b32_e32 v13, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
+; GFX8-NEXT:    flat_store_dwordx4 v[8:9], v[28:31]
+; GFX8-NEXT:    v_mov_b32_e32 v8, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0x110
+; GFX8-NEXT:    v_mov_b32_e32 v4, s22
+; GFX8-NEXT:    v_mov_b32_e32 v9, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v48, s33
+; GFX8-NEXT:    v_mov_b32_e32 v49, s86
+; GFX8-NEXT:    v_mov_b32_e32 v0, s16
+; GFX8-NEXT:    v_mov_b32_e32 v1, s17
+; GFX8-NEXT:    v_mov_b32_e32 v2, s18
+; GFX8-NEXT:    v_mov_b32_e32 v3, s19
+; GFX8-NEXT:    v_mov_b32_e32 v5, s23
+; GFX8-NEXT:    v_mov_b32_e32 v6, s26
+; GFX8-NEXT:    v_mov_b32_e32 v7, s27
+; GFX8-NEXT:    flat_store_dwordx4 v[10:11], v[32:35]
+; GFX8-NEXT:    flat_store_dwordx4 v[48:49], v[44:47]
+; GFX8-NEXT:    flat_store_dwordx4 v[12:13], v[0:3]
+; GFX8-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
+; GFX8-NEXT:    v_mov_b32_e32 v0, s28
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0x100
+; GFX8-NEXT:    v_mov_b32_e32 v1, s29
+; GFX8-NEXT:    v_mov_b32_e32 v2, s30
+; GFX8-NEXT:    v_mov_b32_e32 v3, s31
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0xf0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s22
-; GFX8-NEXT:    v_mov_b32_e32 v1, s23
-; GFX8-NEXT:    v_mov_b32_e32 v2, s20
-; GFX8-NEXT:    v_mov_b32_e32 v3, s21
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0xf0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s36
+; GFX8-NEXT:    v_mov_b32_e32 v1, s37
+; GFX8-NEXT:    v_mov_b32_e32 v2, s34
+; GFX8-NEXT:    v_mov_b32_e32 v3, s35
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0xe0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s24
-; GFX8-NEXT:    v_mov_b32_e32 v1, s25
-; GFX8-NEXT:    v_mov_b32_e32 v2, s26
-; GFX8-NEXT:    v_mov_b32_e32 v3, s27
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0xe0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s40
+; GFX8-NEXT:    v_mov_b32_e32 v1, s41
+; GFX8-NEXT:    v_mov_b32_e32 v2, s44
+; GFX8-NEXT:    v_mov_b32_e32 v3, s45
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0xd0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s30
-; GFX8-NEXT:    v_mov_b32_e32 v1, s31
-; GFX8-NEXT:    v_mov_b32_e32 v2, s82
-; GFX8-NEXT:    v_mov_b32_e32 v3, s83
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0xd0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s46
+; GFX8-NEXT:    v_mov_b32_e32 v1, s47
+; GFX8-NEXT:    v_mov_b32_e32 v2, s84
+; GFX8-NEXT:    v_mov_b32_e32 v3, s85
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0xc0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s84
-; GFX8-NEXT:    v_mov_b32_e32 v1, s85
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0xc0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s82
+; GFX8-NEXT:    v_mov_b32_e32 v1, s83
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s80
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s81
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0xb0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0xb0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s78
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s79
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s76
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s77
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0xa0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0xa0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s74
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s75
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s72
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s73
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0x90
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0x90
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s70
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s71
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s68
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s69
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0x80
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0x80
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s66
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s67
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s64
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s65
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0x70
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0x70
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s62
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s63
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s60
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s61
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0x60
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0x60
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s58
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s59
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s56
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s57
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 0x50
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 0x50
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s54
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s55
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s52
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s53
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 64
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 64
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s50
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s51
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s48
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s49
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
-; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 48
-; GFX8-NEXT:    v_mov_b32_e32 v0, s46
-; GFX8-NEXT:    v_mov_b32_e32 v1, s47
-; GFX8-NEXT:    v_mov_b32_e32 v2, s44
-; GFX8-NEXT:    v_mov_b32_e32 v3, s45
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 32
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    s_add_u32 s14, s0, 48
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s42
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s43
-; GFX8-NEXT:    v_mov_b32_e32 v2, s40
-; GFX8-NEXT:    v_mov_b32_e32 v3, s41
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v2, s38
+; GFX8-NEXT:    v_mov_b32_e32 v3, s39
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
+; GFX8-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NEXT:    s_add_u32 s4, s0, 16
-; GFX8-NEXT:    v_mov_b32_e32 v0, s38
-; GFX8-NEXT:    v_mov_b32_e32 v1, s39
-; GFX8-NEXT:    v_mov_b32_e32 v2, s36
-; GFX8-NEXT:    v_mov_b32_e32 v3, s37
-; GFX8-NEXT:    s_addc_u32 s5, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NEXT:    v_mov_b32_e32 v0, s24
+; GFX8-NEXT:    v_mov_b32_e32 v1, s25
+; GFX8-NEXT:    v_mov_b32_e32 v2, s20
+; GFX8-NEXT:    v_mov_b32_e32 v3, s21
+; GFX8-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NEXT:    v_mov_b32_e32 v0, s34
-; GFX8-NEXT:    v_mov_b32_e32 v1, s35
-; GFX8-NEXT:    v_mov_b32_e32 v2, s28
-; GFX8-NEXT:    v_mov_b32_e32 v3, s29
-; GFX8-NEXT:    v_mov_b32_e32 v5, s5
+; GFX8-NEXT:    flat_store_dwordx4 v[18:19], v[36:39]
+; GFX8-NEXT:    v_mov_b32_e32 v2, s10
+; GFX8-NEXT:    s_add_u32 s10, s0, 32
+; GFX8-NEXT:    v_mov_b32_e32 v3, s11
+; GFX8-NEXT:    s_addc_u32 s11, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s12
+; GFX8-NEXT:    v_mov_b32_e32 v1, s13
+; GFX8-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NEXT:    flat_store_dwordx4 v[58:59], v[40:43]
+; GFX8-NEXT:    v_mov_b32_e32 v2, s6
+; GFX8-NEXT:    s_add_u32 s6, s0, 16
+; GFX8-NEXT:    v_mov_b32_e32 v3, s7
+; GFX8-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s8
+; GFX8-NEXT:    v_mov_b32_e32 v1, s9
+; GFX8-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s1
-; GFX8-NEXT:    v_mov_b32_e32 v0, s18
-; GFX8-NEXT:    v_mov_b32_e32 v1, s19
+; GFX8-NEXT:    v_mov_b32_e32 v0, s4
+; GFX8-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s0
-; GFX8-NEXT:    flat_store_dwordx4 v[14:15], v[36:39]
-; GFX8-NEXT:    flat_store_dwordx4 v[58:59], v[40:43]
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/load-constant-i16.ll b/llvm/test/CodeGen/AMDGPU/load-constant-i16.ll
index 1536ac5112e7b..0c61566435c98 100644
--- a/llvm/test/CodeGen/AMDGPU/load-constant-i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-constant-i16.ll
@@ -199,13 +199,13 @@ define amdgpu_kernel void @constant_load_v3i16(ptr addrspace(1) %out, ptr addrsp
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 4
-; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s1
+; GCN-HSA-NEXT:    s_addc_u32 s0, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s2
 ; GCN-HSA-NEXT:    flat_store_short v[2:3], v4
 ; GCN-HSA-NEXT:    flat_store_dword v[0:1], v5
@@ -217,13 +217,13 @@ define amdgpu_kernel void @constant_load_v3i16(ptr addrspace(1) %out, ptr addrsp
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s0, 4
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s1
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s0, s1, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s2
 ; GCN-NOHSA-VI-NEXT:    flat_store_short v[2:3], v4
 ; GCN-NOHSA-VI-NEXT:    flat_store_dword v[0:1], v5
@@ -612,11 +612,11 @@ define amdgpu_kernel void @constant_load_v16i16_align2(ptr addrspace(4) %ptr0) #
 ; GCN-HSA-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    s_addc_u32 s0, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s0
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
@@ -632,36 +632,36 @@ define amdgpu_kernel void @constant_load_v16i16_align2(ptr addrspace(4) %ptr0) #
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 14
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v6, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v8, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v10, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 30
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 28
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v14, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 26
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NOHSA-VI-NEXT:    flat_load_ushort v16, v[0:1]
 ; GCN-NOHSA-VI-NEXT:    flat_load_ushort v17, v[2:3]
@@ -672,35 +672,35 @@ define amdgpu_kernel void @constant_load_v16i16_align2(ptr addrspace(4) %ptr0) #
 ; GCN-NOHSA-VI-NEXT:    flat_load_ushort v10, v[12:13]
 ; GCN-NOHSA-VI-NEXT:    flat_load_ushort v11, v[14:15]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 24
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 22
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v6, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 18
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NOHSA-VI-NEXT:    flat_load_ushort v12, v[0:1]
 ; GCN-NOHSA-VI-NEXT:    flat_load_ushort v13, v[2:3]
 ; GCN-NOHSA-VI-NEXT:    flat_load_ushort v14, v[4:5]
 ; GCN-NOHSA-VI-NEXT:    flat_load_ushort v15, v[6:7]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 16
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 2
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v6, s0
 ; GCN-NOHSA-VI-NEXT:    flat_load_ushort v0, v[0:1]
@@ -1784,28 +1784,28 @@ define amdgpu_kernel void @constant_zextload_v8i16_to_v8i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_lshr_b32 s8, s5, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s9, s4, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s2, s7, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s3, s6, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s2, s5, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s3, s4, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s8, s7, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s9, s6, 16
 ; GCN-HSA-NEXT:    s_and_b32 s5, s5, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s4, s4, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s7, s7, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s6, s6, 0xffff
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-HSA-NEXT:    s_add_u32 s6, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
@@ -1816,28 +1816,28 @@ define amdgpu_kernel void @constant_zextload_v8i16_to_v8i32(ptr addrspace(1) %ou
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s8, s5, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s9, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s2, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s5, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s5, s4, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s2, s7, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s7, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s7, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s8, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s7, s7, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s9, s6, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s6, s6, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s3
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s7
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s0, 16
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s9
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s9
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s3
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
@@ -1943,28 +1943,28 @@ define amdgpu_kernel void @constant_sextload_v8i16_to_v8i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_ashr_i32 s8, s5, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s9, s4, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s2, s7, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s3, s6, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    s_sext_i32_i16 s7, s7
+; GCN-HSA-NEXT:    s_ashr_i32 s2, s5, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s3, s4, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s8, s7, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s9, s6, 16
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s6, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    s_sext_i32_i16 s5, s5
-; GCN-HSA-NEXT:    s_sext_i32_i16 s4, s4
+; GCN-HSA-NEXT:    s_sext_i32_i16 s7, s7
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-HSA-NEXT:    s_add_u32 s6, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-HSA-NEXT:    s_sext_i32_i16 s5, s5
+; GCN-HSA-NEXT:    s_sext_i32_i16 s4, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
@@ -1975,28 +1975,28 @@ define amdgpu_kernel void @constant_sextload_v8i16_to_v8i32(ptr addrspace(1) %ou
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s8, s5, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s9, s4, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s2, s7, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s3, s6, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 16
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s7, s7
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s2, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s3, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s8, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s9, s6, 16
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s6, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s5, s5
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s4, s4
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s7, s7
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s0, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s5, s5
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s4, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s9
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s9
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
@@ -2122,14 +2122,14 @@ define amdgpu_kernel void @constant_zextload_v16i16_to_v16i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_lshr_b32 s12, s5, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s13, s4, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s14, s7, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s15, s6, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s16, s9, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s17, s8, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s2, s11, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s3, s10, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s2, s5, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s3, s4, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s12, s7, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s13, s6, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s14, s9, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s15, s8, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s16, s11, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s17, s10, 16
 ; GCN-HSA-NEXT:    s_and_b32 s5, s5, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s4, s4, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s7, s7, 0xffff
@@ -2138,38 +2138,40 @@ define amdgpu_kernel void @constant_zextload_v16i16_to_v16i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_and_b32 s8, s8, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s11, s11, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s10, s10, 0xffff
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
+; GCN-HSA-NEXT:    s_add_u32 s10, s0, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s11
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    s_addc_u32 s11, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s17
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s10
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    s_nop 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s17
+; GCN-HSA-NEXT:    s_add_u32 s8, s0, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s9
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    s_addc_u32 s9, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s15
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s9
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    s_nop 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s15
+; GCN-HSA-NEXT:    s_add_u32 s6, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s14
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s13
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s13
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
@@ -2180,54 +2182,56 @@ define amdgpu_kernel void @constant_zextload_v16i16_to_v16i32(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s12, s5, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s13, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s2, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s5, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s5, s4, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s14, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s12, s7, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s7, s7, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s15, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s13, s6, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s6, s6, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s16, s9, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s14, s9, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s9, s9, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s17, s8, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s15, s8, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s8, s8, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s2, s11, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s11, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s11, s10, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s16, s11, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s11, s11, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s17, s10, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s10, s10, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 48
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s3
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s11
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s10, s0, 48
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s11
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s11, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s17
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s16
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 16
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s17
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s8, s0, 32
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s16
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s15
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s14
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s15
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s0, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s14
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s13
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s13
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s3
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
@@ -2383,54 +2387,54 @@ define amdgpu_kernel void @constant_sextload_v16i16_to_v16i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_ashr_i32 s12, s5, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s13, s4, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s14, s7, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s15, s6, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s16, s9, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s17, s8, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s2, s11, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s3, s10, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    s_sext_i32_i16 s11, s11
+; GCN-HSA-NEXT:    s_ashr_i32 s2, s5, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s3, s4, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s12, s7, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s13, s6, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s14, s9, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s15, s8, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s16, s11, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s17, s10, 16
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s10, s10
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    s_sext_i32_i16 s11, s11
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
+; GCN-HSA-NEXT:    s_add_u32 s10, s0, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s11
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    s_sext_i32_i16 s9, s9
+; GCN-HSA-NEXT:    s_addc_u32 s11, s1, 0
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s8, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s17
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s10
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s11
+; GCN-HSA-NEXT:    s_sext_i32_i16 s9, s9
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    s_sext_i32_i16 s7, s7
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s6, s6
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s17
+; GCN-HSA-NEXT:    s_add_u32 s8, s0, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s9
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    s_sext_i32_i16 s5, s5
-; GCN-HSA-NEXT:    s_sext_i32_i16 s4, s4
+; GCN-HSA-NEXT:    s_addc_u32 s9, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s15
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s9
+; GCN-HSA-NEXT:    s_sext_i32_i16 s7, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    s_sext_i32_i16 s5, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s15
+; GCN-HSA-NEXT:    s_add_u32 s6, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s14
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-HSA-NEXT:    s_sext_i32_i16 s4, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s13
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s13
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
@@ -2441,54 +2445,54 @@ define amdgpu_kernel void @constant_sextload_v16i16_to_v16i32(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s12, s5, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s13, s4, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s14, s7, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s15, s6, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s16, s9, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s17, s8, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s2, s11, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s3, s10, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 48
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s11, s11
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s2, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s3, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s12, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s13, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s14, s9, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s15, s8, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s16, s11, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s17, s10, 16
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s10, s10
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 32
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s11, s11
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s10, s0, 48
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s11
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s9, s9
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s11, s1, 0
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s8, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s17
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s16
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s11
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s9, s9
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 16
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s7, s7
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s6, s6
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s17
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s8, s0, 32
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s16
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s5, s5
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s4, s4
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s15
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s14
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s7, s7
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s5, s5
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s15
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s0, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s14
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s4, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s13
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s13
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
@@ -2706,8 +2710,8 @@ define amdgpu_kernel void @constant_zextload_v32i16_to_v32i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_lshr_b32 s31, s12, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s33, s15, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s34, s14, 16
-; GCN-HSA-NEXT:    s_and_b32 s35, s1, 0xffff
-; GCN-HSA-NEXT:    s_and_b32 s36, s0, 0xffff
+; GCN-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
+; GCN-HSA-NEXT:    s_and_b32 s0, s0, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s3, s3, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s2, s2, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s5, s5, 0xffff
@@ -2722,73 +2726,75 @@ define amdgpu_kernel void @constant_zextload_v32i16_to_v32i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_and_b32 s12, s12, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s15, s15, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s14, s14, 0xffff
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 0x70
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 0x60
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 0x50
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 64
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0x70
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s35
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s36
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s35
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s36
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s14
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s34
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s15
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s33
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[0:3]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s8
+; GCN-HSA-NEXT:    s_add_u32 s8, s16, 64
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s1
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s31
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s13
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s30
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[0:3]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s9
+; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s35
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s36
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s10
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s29
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s11
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s28
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s27
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[4:7]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[8:11]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s26
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 32
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[12:15]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s25
+; GCN-HSA-NEXT:    s_add_u32 s6, s16, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
+; GCN-HSA-NEXT:    s_addc_u32 s7, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s25
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s24
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 16
+; GCN-HSA-NEXT:    s_nop 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s23
+; GCN-HSA-NEXT:    s_add_u32 s4, s16, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-HSA-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s23
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s22
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
+; GCN-HSA-NEXT:    s_nop 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s21
+; GCN-HSA-NEXT:    s_add_u32 s2, s16, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s3
+; GCN-HSA-NEXT:    s_addc_u32 s3, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s21
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s36
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s19
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s35
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s18
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s17
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -2801,104 +2807,110 @@ define amdgpu_kernel void @constant_zextload_v32i16_to_v32i32(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx16 s[0:15], s[18:19], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s18, s1, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s19, s1, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s20, s0, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s21, s0, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s22, s3, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s1, s1, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s19, s0, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s0, s0, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s20, s3, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s3, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s23, s2, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s21, s2, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s2, s2, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s24, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s22, s5, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s25, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s23, s4, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s26, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s24, s7, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s7, s7, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s27, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s25, s6, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s6, s6, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s28, s9, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s26, s9, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s9, s9, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s29, s8, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s27, s8, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s8, s8, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s30, s11, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s28, s11, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s11, s11, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s31, s10, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s29, s10, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s10, s10, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s33, s13, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s30, s13, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s13, s13, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s34, s12, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s31, s12, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s12, s12, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s0, s15, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s1, s15, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s15, s14, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s33, s15, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s15, s15, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s34, s14, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s14, s14, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 0x70
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s1
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 0x60
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s14
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s15
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s14, s16, 0x70
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s15, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s33
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s14
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s15
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 0x50
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s12
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s34
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s12, s16, 0x60
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s13
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s33
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s13, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s31
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s30
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s13
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 64
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s31
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s10, s16, 0x50
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s11
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s30
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s11, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s29
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s28
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 48
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s29
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s8, s16, 64
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s28
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s27
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s26
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 32
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s27
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s16, 48
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s26
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s25
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s24
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 16
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s25
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 32
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s24
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s23
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s22
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s23
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s16, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s22
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s21
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s21
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s20
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s19
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s19
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s18
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s17
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -3167,103 +3179,103 @@ define amdgpu_kernel void @constant_sextload_v32i16_to_v32i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_ashr_i32 s18, s1, 16
 ; GCN-HSA-NEXT:    s_ashr_i32 s19, s0, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s22, s3, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s23, s2, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s24, s5, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s25, s4, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s26, s7, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s27, s6, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s28, s9, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s29, s8, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s30, s11, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s31, s10, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s33, s13, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s34, s12, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s35, s15, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s36, s14, 16
-; GCN-HSA-NEXT:    s_sext_i32_i16 s21, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 0x70
-; GCN-HSA-NEXT:    s_sext_i32_i16 s20, s1
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 0x60
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 0x50
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-HSA-NEXT:    s_ashr_i32 s20, s3, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s21, s2, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s22, s5, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s23, s4, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s24, s7, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s25, s6, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s26, s9, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s27, s8, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s28, s11, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s29, s10, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s30, s13, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s31, s12, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s33, s15, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s34, s14, 16
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0x70
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s35
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s36
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s35
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0x50
+; GCN-HSA-NEXT:    s_sext_i32_i16 s8, s8
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s15, s15
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s14, s14
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 64
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s14
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s36
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s15
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s35
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s36
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GCN-HSA-NEXT:    s_sext_i32_i16 s9, s9
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s13, s13
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s12, s12
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[0:3]
-; GCN-HSA-NEXT:    s_sext_i32_i16 s9, s9
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s1
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 48
-; GCN-HSA-NEXT:    s_sext_i32_i16 s8, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s34
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s33
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s8
+; GCN-HSA-NEXT:    s_add_u32 s8, s16, 64
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s11, s11
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s10, s10
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s34
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s31
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s13
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s33
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-HSA-NEXT:    s_sext_i32_i16 s7, s7
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s30
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[0:3]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s9
+; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s6, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s35
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s36
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s10
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s31
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s29
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s11
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s30
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s29
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s28
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s27
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[4:7]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[8:11]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s28
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s26
+; GCN-HSA-NEXT:    s_sext_i32_i16 s7, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[12:15]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s27
+; GCN-HSA-NEXT:    s_add_u32 s6, s16, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s26
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-HSA-NEXT:    s_sext_i32_i16 s5, s5
+; GCN-HSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s4, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s25
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s24
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
+; GCN-HSA-NEXT:    s_sext_i32_i16 s5, s5
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s16, 16
-; GCN-HSA-NEXT:    s_sext_i32_i16 s3, s3
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s2, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s25
+; GCN-HSA-NEXT:    s_add_u32 s4, s16, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s24
-; GCN-HSA-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-HSA-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s23
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s22
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
+; GCN-HSA-NEXT:    s_sext_i32_i16 s3, s3
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
+; GCN-HSA-NEXT:    s_sext_i32_i16 s1, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s23
+; GCN-HSA-NEXT:    s_add_u32 s2, s16, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s22
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-HSA-NEXT:    s_addc_u32 s3, s17, 0
+; GCN-HSA-NEXT:    s_sext_i32_i16 s0, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s21
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s20
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s21
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s19
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s20
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s18
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s17
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -3277,103 +3289,103 @@ define amdgpu_kernel void @constant_sextload_v32i16_to_v32i32(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s18, s1, 16
 ; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s19, s0, 16
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s20, s1
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s21, s0
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s22, s3, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s23, s2, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s24, s5, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s25, s4, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s26, s7, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s27, s6, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s28, s9, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s29, s8, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s30, s11, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s31, s10, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s33, s13, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s34, s12, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s0, s15, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s1, s14, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 0x70
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s1
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s15, s15
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s20, s3, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s21, s2, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s22, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s23, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s24, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s25, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s26, s9, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s27, s8, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s28, s11, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s29, s10, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s30, s13, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s31, s12, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s33, s15, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s34, s14, 16
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s14, s14
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 0x60
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s15, s15
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s14
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s14, s16, 0x70
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s15
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s13, s13
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s15, s17, 0
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s12, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s33
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s14
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s15
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s13, s13
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 0x50
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s10, s10
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s12
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s34
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s12, s16, 0x60
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s13
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s33
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s13, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s31
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s30
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s13
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s11, s11
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s10, s10
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 64
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s8, s8
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s31
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s10, s16, 0x50
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s11
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s30
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s11, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s29
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s28
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s9, s9
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s8, s8
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 48
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s6, s6
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s29
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s8, s16, 64
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s28
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s27
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s26
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s7, s7
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s6, s6
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 32
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s4, s4
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s27
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s16, 48
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s26
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s25
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s24
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s5, s5
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s4, s4
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 16
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s3, s3
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s2, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s25
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 32
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s24
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s23
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s22
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s3, s3
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s1, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s23
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s16, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s22
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s0, s0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s21
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s21
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s19
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s18
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s17
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -3773,8 +3785,8 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_lshr_b32 s64, s28, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s65, s31, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s66, s30, 16
-; GCN-HSA-NEXT:    s_and_b32 s67, s17, 0xffff
-; GCN-HSA-NEXT:    s_and_b32 s68, s16, 0xffff
+; GCN-HSA-NEXT:    s_and_b32 s17, s17, 0xffff
+; GCN-HSA-NEXT:    s_and_b32 s16, s16, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s19, s19, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s18, s18, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s21, s21, 0xffff
@@ -3789,97 +3801,91 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_and_b32 s28, s28, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s31, s31, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s30, s30, 0xffff
-; GCN-HSA-NEXT:    s_add_u32 s16, s36, 0xf0
-; GCN-HSA-NEXT:    s_addc_u32 s17, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s17
-; GCN-HSA-NEXT:    s_add_u32 s16, s36, 0xe0
-; GCN-HSA-NEXT:    s_addc_u32 s17, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s16
-; GCN-HSA-NEXT:    s_add_u32 s16, s36, 0xd0
-; GCN-HSA-NEXT:    s_addc_u32 s17, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s16
-; GCN-HSA-NEXT:    s_add_u32 s16, s36, 0xc0
-; GCN-HSA-NEXT:    s_addc_u32 s17, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s16
-; GCN-HSA-NEXT:    s_add_u32 s16, s36, 0xb0
-; GCN-HSA-NEXT:    s_addc_u32 s17, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s16
-; GCN-HSA-NEXT:    s_add_u32 s16, s36, 0xa0
-; GCN-HSA-NEXT:    s_addc_u32 s17, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s16
+; GCN-HSA-NEXT:    s_add_u32 s67, s36, 0xf0
+; GCN-HSA-NEXT:    s_addc_u32 s68, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s67
+; GCN-HSA-NEXT:    s_add_u32 s67, s36, 0xe0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s68
+; GCN-HSA-NEXT:    s_addc_u32 s68, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s67
+; GCN-HSA-NEXT:    s_add_u32 s67, s36, 0xd0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s68
+; GCN-HSA-NEXT:    s_addc_u32 s68, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s67
+; GCN-HSA-NEXT:    s_add_u32 s67, s36, 0xc0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s68
+; GCN-HSA-NEXT:    s_addc_u32 s68, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s67
+; GCN-HSA-NEXT:    s_add_u32 s67, s36, 0xb0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s68
+; GCN-HSA-NEXT:    s_addc_u32 s68, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s28
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s64
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s29
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s63
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s30
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s66
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s31
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s65
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[8:11]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s26
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s18
+; GCN-HSA-NEXT:    s_add_u32 s18, s36, 0xa0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s62
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s27
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s61
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[4:7]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s20
+; GCN-HSA-NEXT:    s_addc_u32 s20, s37, 0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[12:15]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s18
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s16
 ; GCN-HSA-NEXT:    s_add_u32 s16, s36, 0x90
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s26
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s62
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s27
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s61
-; GCN-HSA-NEXT:    s_addc_u32 s17, s37, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[8:11]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s16
+; GCN-HSA-NEXT:    s_addc_u32 s18, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v28, s16
 ; GCN-HSA-NEXT:    s_add_u32 s16, s36, 0x80
-; GCN-HSA-NEXT:    s_addc_u32 s17, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s24
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s60
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s25
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s59
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s14
-; GCN-HSA-NEXT:    s_add_u32 s14, s36, 0x70
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s30
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s66
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s31
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s65
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[12:15]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s28
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s15
-; GCN-HSA-NEXT:    s_addc_u32 s15, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s64
-; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s29
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s63
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s22
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s58
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s23
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s57
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s67
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s68
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s22
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s58
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s23
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s57
+; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s18
+; GCN-HSA-NEXT:    s_addc_u32 s18, s37, 0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s12
+; GCN-HSA-NEXT:    s_add_u32 s12, s36, 0x70
+; GCN-HSA-NEXT:    s_addc_u32 s14, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s24
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s25
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s59
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s12
 ; GCN-HSA-NEXT:    s_add_u32 s12, s36, 0x60
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s20
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[4:7]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s56
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s18
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s21
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s55
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s54
-; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s19
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s53
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s68
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s38
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s13
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s54
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s19
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s53
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[16:19]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s39
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s13
 ; GCN-HSA-NEXT:    s_addc_u32 s13, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s39
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s67
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s52
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s51
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s50
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s15
-; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s14
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[16:19]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s49
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s13
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s12
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[4:7]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[8:11]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s18
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s17
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s38
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s52
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s51
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s49
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s13
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[8:11]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[12:15]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[16:19]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
 ; GCN-HSA-NEXT:    s_add_u32 s10, s36, 0x50
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s11
@@ -3889,7 +3895,7 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s10
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s56
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
 ; GCN-HSA-NEXT:    s_add_u32 s8, s36, 64
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s9
@@ -3899,7 +3905,7 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s9
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s21
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
 ; GCN-HSA-NEXT:    s_add_u32 s6, s36, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
@@ -3909,7 +3915,7 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s55
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s36, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
@@ -3919,15 +3925,15 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s20
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s36, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s37, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s40
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s35
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s36
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -3935,6 +3941,7 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s33
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s37
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[20:23]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
@@ -3993,8 +4000,8 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s64, s12, 16
 ; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s65, s15, 16
 ; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s66, s14, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s67, s1, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s68, s0, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s1, s1, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s0, s0, 0xffff
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s3, 0xffff
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s2, s2, 0xffff
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, 0xffff
@@ -4007,142 +4014,149 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s10, s10, 0xffff
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s13, s13, 0xffff
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s12, s12, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s0, s15, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s1, s14, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xf0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s1
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xe0
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s15, s15, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s14, s14, 0xffff
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s14
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s14, s36, 0xf0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s15, s37, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s66
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s65
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s14
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s15
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xd0
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s12
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s64
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s12, s36, 0xe0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s13
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s13, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s64
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s63
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s13
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xc0
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s62
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s10, s36, 0xd0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s11
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s11, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s62
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s61
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xb0
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s60
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s8, s36, 0xc0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s60
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s59
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xa0
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s58
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s36, 0xb0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s58
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s57
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x90
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s56
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s36, 0xa0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s56
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s55
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x80
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s54
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s36, 0x90
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s3
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s54
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s53
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
+; GCN-NOHSA-VI-NEXT:    s_nop 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x80
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s1
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x70
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s68
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s52
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s67
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s51
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x60
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s30
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s50
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s31
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s49
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x50
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s28
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s39
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s29
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s38
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 64
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s26
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s48
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s27
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s47
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 48
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s24
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s46
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s25
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s45
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 32
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s22
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s44
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s23
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s43
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s20
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s42
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s21
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s41
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s18
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s40
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s19
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s35
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s36
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s16
@@ -4641,135 +4655,129 @@ define amdgpu_kernel void @constant_sextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_ashr_i32 s52, s16, 16
 ; GCN-HSA-NEXT:    s_ashr_i32 s53, s19, 16
 ; GCN-HSA-NEXT:    s_ashr_i32 s54, s18, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s57, s21, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s58, s20, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s59, s23, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s60, s22, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s61, s25, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s62, s24, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s63, s27, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s64, s26, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s65, s29, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s66, s28, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s67, s31, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s68, s30, 16
-; GCN-HSA-NEXT:    s_sext_i32_i16 s56, s18
-; GCN-HSA-NEXT:    s_add_u32 s18, s36, 0xf0
-; GCN-HSA-NEXT:    s_sext_i32_i16 s55, s19
-; GCN-HSA-NEXT:    s_addc_u32 s19, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s18
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s19
-; GCN-HSA-NEXT:    s_add_u32 s18, s36, 0xe0
-; GCN-HSA-NEXT:    s_addc_u32 s19, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s18
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s19
-; GCN-HSA-NEXT:    s_add_u32 s18, s36, 0xd0
-; GCN-HSA-NEXT:    s_addc_u32 s19, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s19
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s18
-; GCN-HSA-NEXT:    s_add_u32 s18, s36, 0xc0
-; GCN-HSA-NEXT:    s_addc_u32 s19, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s19
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s18
-; GCN-HSA-NEXT:    s_add_u32 s18, s36, 0xb0
-; GCN-HSA-NEXT:    s_addc_u32 s19, s37, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s19
-; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s18
-; GCN-HSA-NEXT:    s_add_u32 s18, s36, 0xa0
-; GCN-HSA-NEXT:    s_addc_u32 s19, s37, 0
+; GCN-HSA-NEXT:    s_ashr_i32 s55, s21, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s56, s20, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s57, s23, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s58, s22, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s59, s25, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s60, s24, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s61, s27, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s62, s26, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s63, s29, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s64, s28, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s65, s31, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s66, s30, 16
+; GCN-HSA-NEXT:    s_add_u32 s67, s36, 0xf0
+; GCN-HSA-NEXT:    s_addc_u32 s68, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s67
+; GCN-HSA-NEXT:    s_add_u32 s67, s36, 0xe0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s68
+; GCN-HSA-NEXT:    s_addc_u32 s68, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s67
+; GCN-HSA-NEXT:    s_add_u32 s67, s36, 0xd0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s68
+; GCN-HSA-NEXT:    s_addc_u32 s68, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s67
+; GCN-HSA-NEXT:    s_add_u32 s67, s36, 0xc0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s68
+; GCN-HSA-NEXT:    s_addc_u32 s68, s37, 0
+; GCN-HSA-NEXT:    s_sext_i32_i16 s29, s29
+; GCN-HSA-NEXT:    s_sext_i32_i16 s28, s28
+; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s67
+; GCN-HSA-NEXT:    s_add_u32 s67, s36, 0xb0
+; GCN-HSA-NEXT:    s_sext_i32_i16 s18, s18
+; GCN-HSA-NEXT:    s_sext_i32_i16 s31, s31
+; GCN-HSA-NEXT:    s_sext_i32_i16 s30, s30
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s68
+; GCN-HSA-NEXT:    s_addc_u32 s68, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s28
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s64
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s29
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s63
+; GCN-HSA-NEXT:    s_sext_i32_i16 s20, s20
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s27, s27
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s26, s26
-; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s19
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s30
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s66
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s31
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s65
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[8:11]
+; GCN-HSA-NEXT:    s_sext_i32_i16 s16, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s18
+; GCN-HSA-NEXT:    s_add_u32 s18, s36, 0xa0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s26
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s62
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s27
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s61
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[4:7]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s20
+; GCN-HSA-NEXT:    s_addc_u32 s20, s37, 0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[12:15]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s18
-; GCN-HSA-NEXT:    s_add_u32 s18, s36, 0x90
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s26
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s64
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s27
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s63
-; GCN-HSA-NEXT:    s_addc_u32 s19, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s16
+; GCN-HSA-NEXT:    s_add_u32 s16, s36, 0x90
+; GCN-HSA-NEXT:    s_addc_u32 s18, s37, 0
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s25, s25
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s24, s24
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[8:11]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s19
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s18
-; GCN-HSA-NEXT:    s_add_u32 s18, s36, 0x80
+; GCN-HSA-NEXT:    v_mov_b32_e32 v28, s16
+; GCN-HSA-NEXT:    s_add_u32 s16, s36, 0x80
+; GCN-HSA-NEXT:    s_sext_i32_i16 s12, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s24
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s25
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s59
+; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s18
+; GCN-HSA-NEXT:    s_addc_u32 s18, s37, 0
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s14, s14
-; GCN-HSA-NEXT:    s_sext_i32_i16 s31, s31
-; GCN-HSA-NEXT:    s_sext_i32_i16 s30, s30
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s24
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s62
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s25
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s61
-; GCN-HSA-NEXT:    s_addc_u32 s19, s37, 0
-; GCN-HSA-NEXT:    s_sext_i32_i16 s15, s15
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s23, s23
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s22, s22
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s30
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[16:19]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s67
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s12
+; GCN-HSA-NEXT:    s_add_u32 s12, s36, 0x70
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s68
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s31
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s67
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[12:15]
-; GCN-HSA-NEXT:    s_sext_i32_i16 s12, s12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s14
-; GCN-HSA-NEXT:    s_add_u32 s14, s36, 0x70
-; GCN-HSA-NEXT:    s_sext_i32_i16 s29, s29
-; GCN-HSA-NEXT:    s_sext_i32_i16 s28, s28
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s15
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s22
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s60
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s23
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s59
-; GCN-HSA-NEXT:    s_addc_u32 s15, s37, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s22
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s58
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s23
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s57
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s14
+; GCN-HSA-NEXT:    s_addc_u32 s14, s37, 0
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s13, s13
-; GCN-HSA-NEXT:    s_sext_i32_i16 s21, s21
-; GCN-HSA-NEXT:    s_sext_i32_i16 s20, s20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s28
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s66
-; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s29
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s65
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[0:3]
-; GCN-HSA-NEXT:    s_sext_i32_i16 s10, s10
+; GCN-HSA-NEXT:    s_sext_i32_i16 s19, s19
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s12
 ; GCN-HSA-NEXT:    s_add_u32 s12, s36, 0x60
+; GCN-HSA-NEXT:    s_sext_i32_i16 s10, s10
+; GCN-HSA-NEXT:    s_sext_i32_i16 s15, s15
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s17, s17
-; GCN-HSA-NEXT:    s_sext_i32_i16 s16, s16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s20
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[4:7]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s58
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s56
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s21
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s57
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s54
-; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s55
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s53
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s13
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s54
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s19
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s53
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s13
 ; GCN-HSA-NEXT:    s_addc_u32 s13, s37, 0
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s11, s11
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s52
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s19
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s18
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s51
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s39
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s38
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s50
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s15
-; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s14
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[16:19]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s49
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s13
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s12
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[4:7]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[8:11]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s52
+; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s18
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s17
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s51
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s39
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s15
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s38
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s49
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s13
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[8:11]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[12:15]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GCN-HSA-NEXT:    s_sext_i32_i16 s8, s8
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[16:19]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
 ; GCN-HSA-NEXT:    s_add_u32 s10, s36, 0x50
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s11
 ; GCN-HSA-NEXT:    s_addc_u32 s11, s37, 0
+; GCN-HSA-NEXT:    s_sext_i32_i16 s8, s8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s47
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s10
@@ -4815,10 +4823,15 @@ define amdgpu_kernel void @constant_sextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s37, 0
 ; GCN-HSA-NEXT:    s_sext_i32_i16 s0, s0
+; GCN-HSA-NEXT:    s_sext_i32_i16 s21, s21
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s40
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s35
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s56
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s21
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s55
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s20
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s36
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -4826,6 +4839,7 @@ define amdgpu_kernel void @constant_sextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s33
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s37
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[20:23]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
@@ -4854,168 +4868,167 @@ define amdgpu_kernel void @constant_sextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s50, s30, 16
 ; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s51, s1, 16
 ; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s52, s0, 16
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s53, s1
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s54, s0
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s55, s3, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s56, s2, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s57, s5, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s58, s4, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s59, s7, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s60, s6, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s61, s9, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s62, s8, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s63, s11, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s64, s10, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s65, s13, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s66, s12, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s0, s15, 16
-; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s1, s14, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xf0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s1
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s15, s15
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s53, s3, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s54, s2, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s55, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s56, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s57, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s58, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s59, s9, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s60, s8, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s61, s11, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s62, s10, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s63, s13, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s64, s12, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s65, s15, 16
+; GCN-NOHSA-VI-NEXT:    s_ashr_i32 s66, s14, 16
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s14, s14
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xe0
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s15, s15
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s14
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s14, s36, 0xf0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s15
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s13, s13
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s15, s37, 0
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s12, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s66
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s65
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s14
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s15
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s13, s13
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xd0
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s10, s10
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s12
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s66
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s12, s36, 0xe0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s13
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s65
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s13, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s64
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s63
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s13
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s11, s11
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s10, s10
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xc0
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s8, s8
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s64
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s10, s36, 0xd0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s11
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s63
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s11, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s62
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s61
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s9, s9
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s8, s8
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xb0
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s6, s6
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s62
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s8, s36, 0xc0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s61
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s60
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s59
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s7, s7
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s6, s6
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0xa0
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s4, s4
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s60
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s36, 0xb0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s59
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s58
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s57
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s5, s5
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s4, s4
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x90
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s2, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s58
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s36, 0xa0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s57
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s56
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s55
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s3, s3
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s2, s2
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x80
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s0, s0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s56
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s36, 0x90
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s55
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s37, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s54
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s53
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s1, s1
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
+; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s31, s31
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x80
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s1
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x70
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s54
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s52
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s53
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s51
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
-; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s31, s31
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s30, s30
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x60
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s30
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s50
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s31
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s49
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s29, s29
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s28, s28
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 0x50
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s28
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s48
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s29
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s47
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s27, s27
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s26, s26
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 64
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s26
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s46
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s27
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s45
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s25, s25
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s24, s24
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 48
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s24
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s44
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s25
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s43
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s23, s23
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s22, s22
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 32
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s22
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s42
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s23
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s41
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s21, s21
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s20, s20
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s36, 16
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s19, s19
@@ -5024,16 +5037,17 @@ define amdgpu_kernel void @constant_sextload_v64i16_to_v64i32(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s40
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s21
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s39
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s37, 0
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s17, s17
 ; GCN-NOHSA-VI-NEXT:    s_sext_i32_i16 s16, s16
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s18
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s38
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s19
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s35
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s36
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s16
@@ -5990,17 +6004,17 @@ define amdgpu_kernel void @constant_zextload_v4i16_to_v4i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_lshr_b32 s4, s3, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s5, s2, 16
-; GCN-HSA-NEXT:    s_and_b32 s6, s2, 0xffff
-; GCN-HSA-NEXT:    s_and_b32 s2, s3, 0xffff
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    s_and_b32 s2, s2, 0xffff
+; GCN-HSA-NEXT:    s_and_b32 s3, s3, 0xffff
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s3
+; GCN-HSA-NEXT:    s_add_u32 s3, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    s_addc_u32 s4, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s4
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -6015,18 +6029,18 @@ define amdgpu_kernel void @constant_zextload_v4i16_to_v4i64(ptr addrspace(1) %ou
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s4, s2, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s2, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s2, s3, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s2, s2, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s5, s3, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s3, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s3
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s3, s0, 16
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s3
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -6289,38 +6303,39 @@ define amdgpu_kernel void @constant_zextload_v8i16_to_v8i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_lshr_b32 s8, s5, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s2, s7, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s9, s6, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s10, s4, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s2, s5, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s3, s7, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s8, s6, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s9, s4, 16
 ; GCN-HSA-NEXT:    s_and_b32 s4, s4, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s6, s6, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s5, s5, 0xffff
-; GCN-HSA-NEXT:    s_and_b32 s3, s7, 0xffff
+; GCN-HSA-NEXT:    s_and_b32 s7, s7, 0xffff
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s3
+; GCN-HSA-NEXT:    s_add_u32 s3, s0, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s7
+; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCN-HSA-NEXT:    s_nop 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s3
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s10
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
@@ -6333,38 +6348,40 @@ define amdgpu_kernel void @constant_zextload_v8i16_to_v8i64(ptr addrspace(1) %ou
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s8, s4, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s9, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s2, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s4, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s4, s5, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s10, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s8, s6, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s6, s6, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s2, s7, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s7, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 48
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s3
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 32
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s9, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s7, s7, 0xffff
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s7
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s7, s0, 48
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s7
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 16
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s10
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s0, 32
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-NOHSA-VI-NEXT:    s_nop 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s0, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s3
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
@@ -6525,28 +6542,28 @@ define amdgpu_kernel void @constant_sextload_v8i16_to_v8i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s13
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s17
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s10
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s11
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s14
@@ -6582,19 +6599,19 @@ define amdgpu_kernel void @constant_sextload_v8i16_to_v8i64(ptr addrspace(1) %ou
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s7
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s1, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s0, 32
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s17
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s1, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s0, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s12
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s13
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s15
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s1, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
@@ -6771,14 +6788,14 @@ define amdgpu_kernel void @constant_zextload_v16i16_to_v16i64(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_lshr_b32 s12, s5, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s13, s7, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s14, s11, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s2, s9, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s15, s8, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s16, s10, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s17, s6, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s18, s4, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s2, s5, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s3, s7, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s12, s11, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s13, s9, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s14, s8, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s15, s10, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s16, s6, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s17, s4, 16
 ; GCN-HSA-NEXT:    s_and_b32 s4, s4, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s6, s6, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s10, s10, 0xffff
@@ -6786,59 +6803,61 @@ define amdgpu_kernel void @constant_zextload_v16i16_to_v16i64(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_and_b32 s5, s5, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s7, s7, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s11, s11, 0xffff
-; GCN-HSA-NEXT:    s_and_b32 s3, s9, 0xffff
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x50
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s3
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    s_and_b32 s9, s9, 0xffff
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s9
+; GCN-HSA-NEXT:    s_add_u32 s9, s0, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s13
+; GCN-HSA-NEXT:    s_addc_u32 s13, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s13
+; GCN-HSA-NEXT:    s_add_u32 s9, s0, 0x70
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s11
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s14
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    s_addc_u32 s11, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s3
+; GCN-HSA-NEXT:    s_add_u32 s3, s0, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s13
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s17
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s18
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s17
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
@@ -6851,74 +6870,78 @@ define amdgpu_kernel void @constant_zextload_v16i16_to_v16i64(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s12, s4, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s13, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s2, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s4, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s4, s5, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s14, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s12, s6, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s6, s6, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s15, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s13, s7, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s7, s7, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s16, s10, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s14, s10, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s10, s10, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s17, s11, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s15, s11, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s11, s11, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s18, s8, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s16, s8, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s8, s8, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s2, s9, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s9, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 0x50
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s3
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 64
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s17, s9, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s9, s9, 0xffff
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s9
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s9, s0, 0x50
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s17
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s17, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s9
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s17
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s18
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s8, s0, 64
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s8, s0, 0x70
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s16
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s1, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s8, s0, 0x60
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s11
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s17
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s1, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 48
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s16
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s14
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 32
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s7
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s15
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s7, s0, 48
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s8, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s13
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s7
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s8
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s0, 16
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s14
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s0, 32
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-NOHSA-VI-NEXT:    s_nop 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s0, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s13
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s1, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s3
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
@@ -7149,91 +7172,92 @@ define amdgpu_kernel void @constant_sextload_v16i16_to_v16i64(ptr addrspace(1) %
 ; GCN-HSA:       ; %bb.0:
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x0
 ; GCN-HSA-NEXT:    s_add_i32 s12, s12, s17
-; GCN-HSA-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; GCN-HSA-NEXT:    s_mov_b32 flat_scratch_lo, s13
+; GCN-HSA-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx8 s[0:7], s[10:11], 0x0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_mov_b32 s12, s7
-; GCN-HSA-NEXT:    s_ashr_i32 s25, s1, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s27, s1, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s30, s3, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s31, s3, 16
-; GCN-HSA-NEXT:    s_mov_b32 s14, s5
-; GCN-HSA-NEXT:    s_mov_b32 s16, s3
-; GCN-HSA-NEXT:    s_mov_b32 s18, s1
-; GCN-HSA-NEXT:    s_lshr_b32 s20, s6, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s22, s4, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s24, s2, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s26, s0, 16
+; GCN-HSA-NEXT:    s_mov_b32 s14, s7
+; GCN-HSA-NEXT:    s_ashr_i32 s23, s1, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s25, s1, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s27, s3, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s29, s3, 16
+; GCN-HSA-NEXT:    s_mov_b32 s16, s5
+; GCN-HSA-NEXT:    s_mov_b32 s18, s3
+; GCN-HSA-NEXT:    s_mov_b32 s20, s1
+; GCN-HSA-NEXT:    s_lshr_b32 s22, s6, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s24, s4, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s26, s2, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s28, s0, 16
 ; GCN-HSA-NEXT:    s_bfe_i64 s[10:11], s[2:3], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[2:3], s[12:13], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[2:3], s[14:15], 0x100000
 ; GCN-HSA-NEXT:    s_ashr_i32 s33, s5, 31
 ; GCN-HSA-NEXT:    s_ashr_i32 s34, s5, 16
 ; GCN-HSA-NEXT:    s_ashr_i32 s35, s7, 31
 ; GCN-HSA-NEXT:    s_ashr_i32 s36, s7, 16
 ; GCN-HSA-NEXT:    s_bfe_i64 s[0:1], s[0:1], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[28:29], s[4:5], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[6:7], s[6:7], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[12:13], s[4:5], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[30:31], s[6:7], 0x100000
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-HSA-NEXT:    s_bfe_i64 s[2:3], s[26:27], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[4:5], s[24:25], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[12:13], s[22:23], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[2:3], s[28:29], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[4:5], s[26:27], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[6:7], s[24:25], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[14:15], s[22:23], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[20:21], s[20:21], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[18:19], s[18:19], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[16:17], s[16:17], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[14:15], s[14:15], 0x100000
 ; GCN-HSA-NEXT:    s_add_u32 s22, s8, 0x70
-; GCN-HSA-NEXT:    s_addc_u32 s23, s9, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s14
-; GCN-HSA-NEXT:    s_add_u32 s14, s8, 0x50
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s15
-; GCN-HSA-NEXT:    s_addc_u32 s15, s9, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s15
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s14
-; GCN-HSA-NEXT:    s_add_u32 s14, s8, 48
-; GCN-HSA-NEXT:    s_addc_u32 s15, s9, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s15
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s14
-; GCN-HSA-NEXT:    s_add_u32 s14, s8, 16
+; GCN-HSA-NEXT:    s_addc_u32 s24, s9, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s16
+; GCN-HSA-NEXT:    s_add_u32 s16, s8, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s17
+; GCN-HSA-NEXT:    s_addc_u32 s17, s9, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s16
+; GCN-HSA-NEXT:    s_add_u32 s16, s8, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s17
+; GCN-HSA-NEXT:    s_addc_u32 s17, s9, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s16
+; GCN-HSA-NEXT:    s_add_u32 s16, s8, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s22
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s23
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s24
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s17
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s36
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s35
+; GCN-HSA-NEXT:    s_addc_u32 s17, s9, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s34
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s33
-; GCN-HSA-NEXT:    s_addc_u32 s15, s9, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s31
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s30
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s18
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s19
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s27
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s14
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s15
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s18
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s14
+; GCN-HSA-NEXT:    s_add_u32 s14, s8, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s19
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s29
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s27
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s20
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s21
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s25
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s17
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[4:7]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[8:11]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s25
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s23
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s15
+; GCN-HSA-NEXT:    s_addc_u32 s15, s9, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[12:15]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-HSA-NEXT:    s_add_u32 s6, s8, 0x60
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s7
-; GCN-HSA-NEXT:    s_addc_u32 s7, s9, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s30
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s31
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s15
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s6
 ; GCN-HSA-NEXT:    s_add_u32 s6, s8, 64
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s21
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s7
 ; GCN-HSA-NEXT:    s_addc_u32 s7, s9, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s13
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s28
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s29
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s13
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_nop 0
@@ -7294,28 +7318,28 @@ define amdgpu_kernel void @constant_sextload_v16i16_to_v16i64(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s5
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s9, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s8, 64
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s34
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s35
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s9, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s8, 0x70
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s28
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s29
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s30
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s31
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s9, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s8, 0x60
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s26
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s27
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s6
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s7
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s9, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
@@ -7330,28 +7354,28 @@ define amdgpu_kernel void @constant_sextload_v16i16_to_v16i64(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s8, 48
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s9, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s8, 32
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s20
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s21
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s9, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s8, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s17
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s18
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s19
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s9, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s14
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s15
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s12
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s13
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
@@ -7628,17 +7652,17 @@ define amdgpu_kernel void @constant_zextload_v32i16_to_v32i64(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx16 s[0:15], s[18:19], 0x0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_lshr_b32 s23, s1, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s24, s3, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s25, s5, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s26, s7, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s27, s9, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s28, s11, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s29, s13, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s30, s15, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s31, s14, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s33, s12, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s34, s10, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s25, s1, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s26, s3, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s27, s5, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s28, s7, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s29, s9, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s30, s11, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s31, s13, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s33, s15, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s34, s14, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s24, s12, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s23, s10, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s22, s8, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s21, s6, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s20, s4, 16
@@ -7648,7 +7672,7 @@ define amdgpu_kernel void @constant_zextload_v32i16_to_v32i64(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_and_b32 s2, s2, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s4, s4, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s6, s6, 0xffff
-; GCN-HSA-NEXT:    s_and_b32 s35, s8, 0xffff
+; GCN-HSA-NEXT:    s_and_b32 s8, s8, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s10, s10, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s12, s12, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s14, s14, 0xffff
@@ -7656,117 +7680,114 @@ define amdgpu_kernel void @constant_zextload_v32i16_to_v32i64(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_and_b32 s3, s3, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s5, s5, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s7, s7, 0xffff
-; GCN-HSA-NEXT:    s_and_b32 s36, s9, 0xffff
+; GCN-HSA-NEXT:    s_and_b32 s9, s9, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s11, s11, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s13, s13, 0xffff
 ; GCN-HSA-NEXT:    s_and_b32 s15, s15, 0xffff
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 0xf0
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s9
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 0xd0
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s9
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 0xb0
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s9
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 0x90
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s9
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s8
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 0x70
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s9
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s8
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 0x50
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0xf0
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s35
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0xd0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s36
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s35
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0xb0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s36
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s35
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0x90
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s36
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s35
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s36
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s15
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s30
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s9
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s8
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s33
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s35
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 0x50
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s36
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s13
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s29
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s31
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s35
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s11
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s28
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s8
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s30
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 48
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s36
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s27
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s36
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s29
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s35
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s26
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s28
+; GCN-HSA-NEXT:    s_add_u32 s35, s16, 16
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[0:3]
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 0xe0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s36
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s25
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s27
+; GCN-HSA-NEXT:    s_addc_u32 s36, s17, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[0:3]
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s35
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s24
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s26
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s36
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s23
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s9
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 0xc0
+; GCN-HSA-NEXT:    s_add_u32 s1, s16, 0xe0
+; GCN-HSA-NEXT:    s_addc_u32 s3, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s25
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s1
+; GCN-HSA-NEXT:    s_add_u32 s1, s16, 0xc0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[0:3]
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s14
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s31
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s34
+; GCN-HSA-NEXT:    s_addc_u32 s3, s17, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s9
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 0xa0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s1
+; GCN-HSA-NEXT:    s_add_u32 s1, s16, 0xa0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s33
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s24
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    s_addc_u32 s3, s17, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s9
-; GCN-HSA-NEXT:    s_add_u32 s8, s16, 0x80
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s1
+; GCN-HSA-NEXT:    s_add_u32 s1, s16, 0x80
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s34
-; GCN-HSA-NEXT:    s_addc_u32 s9, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s23
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    s_addc_u32 s3, s17, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s35
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s1
+; GCN-HSA-NEXT:    s_add_u32 s1, s16, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s22
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    s_addc_u32 s3, s17, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s1
+; GCN-HSA-NEXT:    s_add_u32 s1, s16, 64
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-HSA-NEXT:    s_add_u32 s6, s16, 0x60
-; GCN-HSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s21
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    s_addc_u32 s3, s17, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    s_add_u32 s4, s16, 64
-; GCN-HSA-NEXT:    s_addc_u32 s5, s17, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    s_add_u32 s1, s16, 32
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s16, 32
-; GCN-HSA-NEXT:    s_addc_u32 s3, s17, 0
+; GCN-HSA-NEXT:    s_addc_u32 s2, s17, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s19
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s2
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -7792,134 +7813,147 @@ define amdgpu_kernel void @constant_zextload_v32i16_to_v32i64(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s21, s3, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s3, s3, 0xffff
 ; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s22, s4, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s23, s4, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s24, s5, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s25, s5, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s26, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s23, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, 0xffff
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s24, s6, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s6, s6, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s27, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s25, s7, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s7, s7, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s28, s8, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s26, s8, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s8, s8, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s29, s9, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s27, s9, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s9, s9, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s30, s10, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s28, s10, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s10, s10, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s31, s11, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s29, s11, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s11, s11, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s33, s12, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s30, s12, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s12, s12, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s34, s13, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s31, s13, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s13, s13, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s35, s14, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s33, s14, 16
 ; GCN-NOHSA-VI-NEXT:    s_and_b32 s14, s14, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s4, s15, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s15, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0xf0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0xe0
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s34, s15, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s15, s15, 0xffff
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s15
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s15, s16, 0xf0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s34
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s34, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s15
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s34
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0xd0
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s14
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s35
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s14, s16, 0xe0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s15, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s33
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s14
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s15
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0xc0
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s13
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s34
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s13, s16, 0xd0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s14, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s31
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s13
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s14
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0xb0
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s12
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s33
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s12, s16, 0xc0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s13, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s30
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s13
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0xa0
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s11
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s31
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s11, s16, 0xb0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s12, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s29
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s11
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s12
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0x90
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s30
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s10, s16, 0xa0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s11, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s28
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0x80
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s9
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s29
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s9, s16, 0x90
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s10, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s27
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s9
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s10
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0x70
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s28
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s8, s16, 0x80
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s9, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s26
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s9
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0x60
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s7
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s27
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s7, s16, 0x70
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s8, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s25
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s7
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s8
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 0x50
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s26
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s6, s16, 0x60
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s7, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s24
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
+; GCN-NOHSA-VI-NEXT:    s_nop 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s5, s16, 0x50
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s6, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s23
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s6
+; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCN-NOHSA-VI-NEXT:    s_nop 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 64
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s25
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s24
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
-; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s22
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s4, s16, 48
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s23
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s22
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s5, s17, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s3
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s3, s16, 48
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s4, s17, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s21
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s3
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s4
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s16, 32
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s16, 16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s20
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s17, 0
-; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s1
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s1, s16, 16
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s2, s17, 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s19
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s1
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s2
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s0
@@ -8348,38 +8382,38 @@ define amdgpu_kernel void @constant_sextload_v32i16_to_v32i64(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_mov_b32 s30, s15
 ; GCN-HSA-NEXT:    s_ashr_i32 s41, s3, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s55, s3, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s57, s5, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s59, s5, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s61, s7, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s63, s7, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s65, s9, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s67, s9, 16
-; GCN-HSA-NEXT:    s_mov_b32 s42, s13
-; GCN-HSA-NEXT:    s_mov_b32 s44, s11
-; GCN-HSA-NEXT:    s_mov_b32 s46, s9
-; GCN-HSA-NEXT:    s_mov_b32 s48, s7
-; GCN-HSA-NEXT:    s_mov_b32 s50, s5
+; GCN-HSA-NEXT:    s_ashr_i32 s42, s3, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s55, s5, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s57, s7, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s59, s7, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s61, s9, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s63, s9, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s65, s11, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s67, s11, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s69, s13, 31
+; GCN-HSA-NEXT:    s_mov_b32 s44, s13
+; GCN-HSA-NEXT:    s_mov_b32 s46, s11
+; GCN-HSA-NEXT:    s_mov_b32 s48, s9
+; GCN-HSA-NEXT:    s_mov_b32 s50, s7
+; GCN-HSA-NEXT:    s_mov_b32 s52, s5
 ; GCN-HSA-NEXT:    s_mov_b32 s38, s3
 ; GCN-HSA-NEXT:    s_mov_b32 s36, s1
-; GCN-HSA-NEXT:    s_lshr_b32 s52, s14, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s54, s12, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s56, s10, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s58, s8, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s60, s6, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s62, s4, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s64, s2, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s66, s0, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s54, s14, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s56, s12, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s58, s10, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s60, s8, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s62, s6, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s64, s4, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s66, s2, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s68, s0, 16
 ; GCN-HSA-NEXT:    s_bfe_i64 s[18:19], s[2:3], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[2:3], s[30:31], 0x100000
 ; GCN-HSA-NEXT:    s_ashr_i32 s33, s1, 31
 ; GCN-HSA-NEXT:    s_ashr_i32 s40, s1, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s68, s11, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s69, s11, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s70, s13, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s71, s13, 16
-; GCN-HSA-NEXT:    s_ashr_i32 s72, s15, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s73, s15, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s43, s5, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s70, s13, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s71, s15, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s72, s15, 16
 ; GCN-HSA-NEXT:    s_bfe_i64 s[0:1], s[0:1], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[20:21], s[4:5], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[22:23], s[6:7], 0x100000
@@ -8387,112 +8421,111 @@ define amdgpu_kernel void @constant_sextload_v32i16_to_v32i64(ptr addrspace(1) %
 ; GCN-HSA-NEXT:    s_bfe_i64 s[26:27], s[10:11], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[28:29], s[12:13], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[34:35], s[14:15], 0x100000
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
-; GCN-HSA-NEXT:    s_bfe_i64 s[2:3], s[66:67], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[4:5], s[64:65], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[6:7], s[62:63], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[8:9], s[60:61], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[10:11], s[58:59], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[12:13], s[56:57], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[14:15], s[54:55], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[30:31], s[52:53], 0x100000
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s3
+; GCN-HSA-NEXT:    s_bfe_i64 s[2:3], s[68:69], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[4:5], s[66:67], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[6:7], s[64:65], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[8:9], s[62:63], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[10:11], s[60:61], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[12:13], s[58:59], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[14:15], s[56:57], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[30:31], s[54:55], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[36:37], s[36:37], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[52:53], s[52:53], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[50:51], s[50:51], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[48:49], s[48:49], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[46:47], s[46:47], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[44:45], s[44:45], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[42:43], s[42:43], 0x100000
-; GCN-HSA-NEXT:    s_add_u32 s52, s16, 0xf0
-; GCN-HSA-NEXT:    s_addc_u32 s53, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s42
-; GCN-HSA-NEXT:    s_add_u32 s42, s16, 0xd0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s43
-; GCN-HSA-NEXT:    s_addc_u32 s43, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s42
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s43
-; GCN-HSA-NEXT:    s_add_u32 s42, s16, 0xb0
-; GCN-HSA-NEXT:    s_addc_u32 s43, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s42
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s43
-; GCN-HSA-NEXT:    s_add_u32 s42, s16, 0x90
-; GCN-HSA-NEXT:    s_addc_u32 s43, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s42
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s43
-; GCN-HSA-NEXT:    s_add_u32 s42, s16, 0x70
-; GCN-HSA-NEXT:    s_addc_u32 s43, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s42
-; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s43
-; GCN-HSA-NEXT:    s_add_u32 s42, s16, 0x50
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s71
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s70
+; GCN-HSA-NEXT:    s_add_u32 s54, s16, 0xf0
+; GCN-HSA-NEXT:    s_addc_u32 s56, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s44
+; GCN-HSA-NEXT:    s_add_u32 s44, s16, 0xd0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s45
+; GCN-HSA-NEXT:    s_addc_u32 s45, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s44
+; GCN-HSA-NEXT:    s_add_u32 s44, s16, 0xb0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s45
+; GCN-HSA-NEXT:    s_addc_u32 s45, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s44
+; GCN-HSA-NEXT:    s_add_u32 s44, s16, 0x90
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s45
+; GCN-HSA-NEXT:    s_addc_u32 s45, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s44
+; GCN-HSA-NEXT:    s_add_u32 s44, s16, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s45
+; GCN-HSA-NEXT:    s_addc_u32 s45, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s69
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s54
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s56
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s72
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s71
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[18:21]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s46
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s38
+; GCN-HSA-NEXT:    s_add_u32 s38, s16, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s47
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[6:9]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s43
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s67
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s65
 ; GCN-HSA-NEXT:    s_addc_u32 s43, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s52
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s53
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s73
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s72
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[6:9]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s44
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s36
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[4:7]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s36
 ; GCN-HSA-NEXT:    s_add_u32 s36, s16, 48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s45
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[2:5]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s69
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s68
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s37
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s37
 ; GCN-HSA-NEXT:    s_addc_u32 s37, s17, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[10:13]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s36
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s37
 ; GCN-HSA-NEXT:    s_add_u32 s36, s16, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s46
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s47
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s67
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s65
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s51
+; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s44
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s45
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s59
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s57
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s37
 ; GCN-HSA-NEXT:    s_addc_u32 s37, s17, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[14:17]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s30
-; GCN-HSA-NEXT:    s_add_u32 s30, s16, 0xe0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s49
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s63
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s61
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s31
-; GCN-HSA-NEXT:    s_addc_u32 s31, s17, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s49
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s63
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s61
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s50
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[14:17]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s28
+; GCN-HSA-NEXT:    s_add_u32 s28, s16, 0xe0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s30
+; GCN-HSA-NEXT:    s_addc_u32 s30, s17, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-HSA-NEXT:    s_add_u32 s14, s16, 0xc0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s51
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s38
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s59
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s57
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s39
-; GCN-HSA-NEXT:    v_mov_b32_e32 v28, s42
-; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s43
-; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s55
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s41
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s52
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s53
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s55
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s39
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s38
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s43
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s42
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s41
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s15
 ; GCN-HSA-NEXT:    s_addc_u32 s15, s17, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s36
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s37
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s40
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s33
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s34
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s35
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s28
+; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s36
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s37
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s40
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s33
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s34
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s35
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s31
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s29
-; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s30
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s31
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[18:21]
-; GCN-HSA-NEXT:    s_nop 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s15
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s14
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[4:7]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[8:11]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[12:15]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[0:3]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s28
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s30
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[10:13]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s15
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[18:21]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[4:7]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[14:17]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
 ; GCN-HSA-NEXT:    s_nop 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s12
 ; GCN-HSA-NEXT:    s_add_u32 s12, s16, 0xa0
@@ -8558,211 +8591,207 @@ define amdgpu_kernel void @constant_sextload_v32i16_to_v32i64(ptr addrspace(1) %
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx16 s[0:15], s[18:19], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s22, s0, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s24, s0, 16
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s26, s1
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s28, s1, 16
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s30, s2, 16
-; GCN-NOHSA-VI-NEXT:    s_mov_b32 s34, s3
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s36, s3, 16
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s40, s4, 16
-; GCN-NOHSA-VI-NEXT:    s_mov_b32 s42, s5
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s44, s5, 16
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s46, s6, 16
-; GCN-NOHSA-VI-NEXT:    s_mov_b32 s48, s7
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s50, s7, 16
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s52, s8, 16
-; GCN-NOHSA-VI-NEXT:    s_mov_b32 s54, s9
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s56, s9, 16
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s58, s10, 16
-; GCN-NOHSA-VI-NEXT:    s_mov_b32 s60, s11
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s62, s11, 16
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s66, s12, 16
-; GCN-NOHSA-VI-NEXT:    s_mov_b32 s68, s13
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s70, s13, 16
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s74, s14, 16
-; GCN-NOHSA-VI-NEXT:    s_mov_b32 s76, s15
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s78, s15, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s30, s1, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s34, s2, 16
+; GCN-NOHSA-VI-NEXT:    s_mov_b32 s36, s3
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s40, s3, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s42, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_mov_b32 s44, s5
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s46, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s50, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_mov_b32 s52, s7
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s54, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s56, s8, 16
+; GCN-NOHSA-VI-NEXT:    s_mov_b32 s58, s9
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s60, s9, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s62, s10, 16
+; GCN-NOHSA-VI-NEXT:    s_mov_b32 s64, s11
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s66, s11, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s68, s12, 16
+; GCN-NOHSA-VI-NEXT:    s_mov_b32 s70, s13
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s72, s13, 16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s76, s14, 16
+; GCN-NOHSA-VI-NEXT:    s_mov_b32 s78, s15
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s80, s15, 16
 ; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[18:19], s[0:1], 0x100000
 ; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[0:1], s[2:3], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[2:3], s[4:5], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[20:21], s[6:7], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[24:25], s[8:9], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[20:21], s[4:5], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[22:23], s[6:7], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[28:29], s[8:9], 0x100000
 ; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[38:39], s[10:11], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[64:65], s[12:13], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[72:73], s[14:15], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[4:5], s[22:23], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[8:9], s[26:27], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[6:7], s[28:29], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[10:11], s[30:31], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[14:15], s[34:35], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[48:49], s[12:13], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[74:75], s[14:15], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[2:3], s[24:25], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[6:7], s[26:27], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[4:5], s[30:31], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[8:9], s[34:35], 0x100000
 ; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[12:13], s[36:37], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[22:23], s[40:41], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[28:29], s[42:43], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[10:11], s[40:41], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[14:15], s[42:43], 0x100000
 ; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[26:27], s[44:45], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[30:31], s[46:47], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[36:37], s[48:49], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[34:35], s[50:51], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[40:41], s[52:53], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[44:45], s[54:55], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[42:43], s[56:57], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[46:47], s[58:59], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[48:49], s[60:61], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[50:51], s[62:63], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[24:25], s[46:47], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[30:31], s[50:51], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[36:37], s[52:53], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[34:35], s[54:55], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[40:41], s[56:57], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[44:45], s[58:59], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[42:43], s[60:61], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[46:47], s[62:63], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[50:51], s[64:65], 0x100000
 ; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[52:53], s[66:67], 0x100000
 ; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[54:55], s[68:69], 0x100000
 ; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[56:57], s[70:71], 0x100000
-; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[58:59], s[74:75], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[58:59], s[72:73], 0x100000
 ; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[60:61], s[76:77], 0x100000
 ; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[62:63], s[78:79], 0x100000
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s60
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s60, s16, 0xf0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s61
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s61, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s62
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s63
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s60
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s61
+; GCN-NOHSA-VI-NEXT:    s_bfe_i64 s[64:65], s[80:81], 0x100000
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s33, s16, 0xf0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s62
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s62, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s63
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s64
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s65
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s33
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s62
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s33, s16, 0xe0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    s_nop 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s33
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s60
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s60, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s74
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s75
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s61
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s60
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s33, s16, 0xd0
+; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s33
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s56
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s56, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s57
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s58
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s58, s16, 0xe0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s59
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s59, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s72
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s73
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s58
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s59
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s56
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s33, s16, 0xc0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    s_nop 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s54
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s54, s16, 0xd0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s55
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s55, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s56
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s57
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s54
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s55
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s33
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s48
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s48, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s33, s16, 0xb0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s49
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s54
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s55
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s48
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s48, s17, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    s_nop 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s33
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s50
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s51
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s52
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s52, s16, 0xc0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s53
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s53, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s64
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s65
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s52
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s53
-; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    s_nop 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s48
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s48, s16, 0xb0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s49
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s49, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s50
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s51
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s48
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s49
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s48
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s33, s16, 0xa0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    s_nop 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s33
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s38
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s38, s16, 0xa0
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s38, s17, 0
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s33, s16, 0x90
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s39
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s39, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s38
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s39
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s38, s16, 0x90
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s46
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s47
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s39, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s38
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s38, s17, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s38
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s33
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s44
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s45
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s42
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s43
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s39
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s38
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_nop 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s24
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s24, s16, 0x80
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s25
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s25, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s24
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s25
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s24, s16, 0x70
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s28
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s28, s16, 0x80
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s29
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s29, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s28
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s28, s16, 0x70
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s40
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s41
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s25, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s29
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s29, s17, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s24
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s28
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s36
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s37
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s34
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s35
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s25
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s29
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_nop 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s20
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s20, s16, 0x60
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s21
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s21, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s20
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s21
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s20, s16, 0x50
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s22
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s22, s16, 0x60
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s23
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s23, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s22
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s22, s16, 0x50
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s30
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s31
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s21, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s23
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s23, s17, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s20
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s28
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s29
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s26
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s27
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s21
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s22
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s26
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s27
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s24
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s25
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s23
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_nop 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s16, 64
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-NOHSA-VI-NEXT:    s_add_u32 s2, s16, 48
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s22
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s23
-; GCN-NOHSA-VI-NEXT:    s_addc_u32 s3, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s14
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s14, s16, 64
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s15
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s15, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s21
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s14
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s15
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s14
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s15
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s12
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s13
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-NOHSA-VI-NEXT:    s_nop 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s10
+; GCN-NOHSA-VI-NEXT:    s_add_u32 s10, s16, 48
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s11
+; GCN-NOHSA-VI-NEXT:    s_addc_u32 s11, s17, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s13
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_nop 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 32
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NOHSA-VI-NEXT:    s_add_u32 s0, s16, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s10
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s11
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s9
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    s_addc_u32 s1, s17, 0
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s9
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s7
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s7
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, s16
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s18
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, s19
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s2
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, s17
 ; GCN-NOHSA-VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/load-constant-i32.ll b/llvm/test/CodeGen/AMDGPU/load-constant-i32.ll
index 3c13ead2038e1..4b642a6714960 100644
--- a/llvm/test/CodeGen/AMDGPU/load-constant-i32.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-constant-i32.ll
@@ -1017,8 +1017,8 @@ define amdgpu_kernel void @constant_load_v11i32(ptr addrspace(1) %out, ptr addrs
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s12
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s13
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s1
 ; GFX7-HSA-NEXT:    flat_store_dwordx3 v[3:4], v[0:2]
 ; GFX7-HSA-NEXT:    s_endpgm
 ;
@@ -1030,8 +1030,8 @@ define amdgpu_kernel void @constant_load_v11i32(ptr addrspace(1) %out, ptr addrs
 ; GFX8-NOHSA-NEXT:    s_load_dwordx4 s[12:15], s[2:3], 0x20
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v6, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v6, s3
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s8
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s9
@@ -1050,8 +1050,8 @@ define amdgpu_kernel void @constant_load_v11i32(ptr addrspace(1) %out, ptr addrs
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s12
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s13
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s14
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s1
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx3 v[3:4], v[0:2]
 ; GFX8-NOHSA-NEXT:    s_endpgm
 ;
@@ -2291,8 +2291,8 @@ define amdgpu_kernel void @constant_zextload_v4i32_to_v4i64(ptr addrspace(1) %ou
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s7
@@ -2313,8 +2313,8 @@ define amdgpu_kernel void @constant_zextload_v4i32_to_v4i64(ptr addrspace(1) %ou
 ; GFX8-NOHSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s7
@@ -2443,23 +2443,23 @@ define amdgpu_kernel void @constant_sextload_v4i32_to_v4i64(ptr addrspace(1) %ou
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_ashr_i32 s8, s5, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s9, s4, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s2, s7, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s3, s6, 31
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    s_ashr_i32 s2, s5, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s3, s4, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s8, s7, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s9, s6, 31
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s4
+; GFX7-HSA-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s6
+; GFX7-HSA-NEXT:    s_addc_u32 s6, s1, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s2
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s4
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s4
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s6
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[5:6], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
@@ -2471,23 +2471,23 @@ define amdgpu_kernel void @constant_sextload_v4i32_to_v4i64(ptr addrspace(1) %ou
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s8, s5, 31
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s9, s4, 31
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s2, s7, 31
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s3, s6, 31
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s3
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s2, s5, 31
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s3, s4, 31
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s8, s7, 31
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s9, s6, 31
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s4
+; GFX8-NOHSA-NEXT:    s_add_u32 s4, s0, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NOHSA-NEXT:    s_addc_u32 s6, s1, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s7
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v6, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s2
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s4
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s4
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v6, s6
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[5:6], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v6, s5
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v7, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v7, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
@@ -2626,40 +2626,42 @@ define amdgpu_kernel void @constant_zextload_v8i32_to_v8i64(ptr addrspace(1) %ou
 ;
 ; GFX7-HSA-LABEL: constant_zextload_v8i32_to_v8i64:
 ; GFX7-HSA:       ; %bb.0:
-; GFX7-HSA-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX7-HSA-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x0
 ; GFX7-HSA-NEXT:    s_add_i32 s12, s12, s17
 ; GFX7-HSA-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7-HSA-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 48
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GFX7-HSA-NEXT:    s_load_dwordx8 s[0:7], s[10:11], 0x0
+; GFX7-HSA-NEXT:    s_add_u32 s10, s8, 48
+; GFX7-HSA-NEXT:    s_addc_u32 s11, s9, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, v1
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s11
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s8
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s9
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
+; GFX7-HSA-NEXT:    s_nop 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s4
+; GFX7-HSA-NEXT:    s_add_u32 s4, s8, 32
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-HSA-NEXT:    s_addc_u32 s5, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s4
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s5
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s2
+; GFX7-HSA-NEXT:    s_add_u32 s2, s8, 16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
+; GFX7-HSA-NEXT:    s_addc_u32 s3, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s1
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s9
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
 ;
@@ -2691,8 +2693,8 @@ define amdgpu_kernel void @constant_zextload_v8i32_to_v8i64(ptr addrspace(1) %ou
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s8, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s9, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s8
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -2902,8 +2904,8 @@ define amdgpu_kernel void @constant_sextload_v8i32_to_v8i64(ptr addrspace(1) %ou
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s9, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s13
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s8
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -2954,8 +2956,8 @@ define amdgpu_kernel void @constant_sextload_v8i32_to_v8i64(ptr addrspace(1) %ou
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s9, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s13
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s12
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s8
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -3202,89 +3204,89 @@ define amdgpu_kernel void @constant_sextload_v16i32_to_v16i64(ptr addrspace(1) %
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_load_dwordx16 s[0:15], s[18:19], 0x0
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_ashr_i32 s20, s1, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s21, s0, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s22, s3, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s23, s2, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s24, s5, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s25, s4, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s26, s7, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s27, s6, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s28, s9, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s29, s8, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s30, s11, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s31, s10, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s33, s13, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s34, s12, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s35, s15, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s36, s14, 31
-; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 0x70
-; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s19
-; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 0x60
-; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s19
-; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 0x50
-; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s19
-; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 64
-; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s19
-; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 48
-; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s19
-; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 32
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s35
+; GFX7-HSA-NEXT:    s_ashr_i32 s18, s1, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s19, s0, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s20, s3, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s21, s2, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s22, s5, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s23, s4, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s24, s7, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s25, s6, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s26, s9, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s27, s8, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s28, s11, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s29, s10, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s30, s13, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s31, s12, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s33, s15, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s34, s14, 31
+; GFX7-HSA-NEXT:    s_add_u32 s35, s16, 0x70
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s16, 0x60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s16, 0x50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s16, 64
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s16, 48
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s16, 32
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s34
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s33
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s14
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s15
-; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s34
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s31
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[5:6], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s33
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s30
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s12
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s19
-; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s31
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s29
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s16, 16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s29
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s27
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[3:6]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s30
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s28
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s28
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s26
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s8
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s9
-; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s27
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s25
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[6:9]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s6
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s26
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s19
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s24
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s35
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s36
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s23
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s21
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[3:6]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s25
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s23
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s24
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s22
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s22
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s20
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s21
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s19
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[9:12]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s1
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s20
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s18
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s17
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[6:9]
@@ -3378,8 +3380,8 @@ define amdgpu_kernel void @constant_sextload_v16i32_to_v16i64(ptr addrspace(1) %
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s17, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s21
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s20
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -3755,8 +3757,8 @@ define amdgpu_kernel void @constant_zextload_v16i32_to_v16i64(ptr addrspace(1) %
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s16, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -3825,8 +3827,8 @@ define amdgpu_kernel void @constant_zextload_v16i32_to_v16i64(ptr addrspace(1) %
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s16, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s17, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -4174,28 +4176,29 @@ define amdgpu_kernel void @constant_sextload_v32i32_to_v32i64(ptr addrspace(1) %
 ; GFX7-HSA-NEXT:    s_ashr_i32 s43, s23, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s44, s22, 31
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s30
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s28
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s26
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s24
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s22
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s28
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s26
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s24
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-HSA-NEXT:    s_ashr_i32 s45, s25, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s46, s24, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s47, s27, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s26, s26, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s48, s29, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s48, s26, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s49, s29, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s28, s28, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s24, s31, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s50, s31, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s30, s30, 31
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s31
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s29
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s27
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s23
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s20
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s21
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s19
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s17
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s29
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s27
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s25
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s23
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s20
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s21
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s18
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s19
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s17
 ; GFX7-HSA-NEXT:    s_ashr_i32 s16, s1, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s17, s0, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s18, s3, 31
@@ -4204,101 +4207,103 @@ define amdgpu_kernel void @constant_sextload_v32i32_to_v32i64(ptr addrspace(1) %
 ; GFX7-HSA-NEXT:    s_ashr_i32 s21, s4, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s22, s7, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s23, s6, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s27, s9, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s29, s8, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s31, s11, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s38, s10, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s39, s13, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s49, s12, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s50, s15, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s51, s14, 31
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v28, s24
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0xf0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s25
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v30, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v29, s24
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0xe0
+; GFX7-HSA-NEXT:    s_ashr_i32 s24, s9, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s25, s8, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s26, s11, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s27, s10, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s29, s13, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s31, s12, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s38, s15, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s39, s14, 31
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s30
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
+; GFX7-HSA-NEXT:    s_add_u32 s30, s36, 0xf0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v28, s50
+; GFX7-HSA-NEXT:    s_addc_u32 s50, s37, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v29, s30
+; GFX7-HSA-NEXT:    s_add_u32 s30, s36, 0xe0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v30, s50
+; GFX7-HSA-NEXT:    s_addc_u32 s50, s37, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[29:30], v[25:28]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s28
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s24
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0xd0
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v28, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s24
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0xc0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s48
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[25:26], v[18:21]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s26
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s24
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s25
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0xb0
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s24
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s25
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0xa0
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v30, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v29, s24
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0x90
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s47
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[27:28], v[7:10]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s46
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s24
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s25
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0x80
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s24
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0x70
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v32, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v31, s24
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0x60
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s45
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v34, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v33, s24
-; GFX7-HSA-NEXT:    s_add_u32 s24, s36, 0x50
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s44
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s43
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[0:3]
-; GFX7-HSA-NEXT:    s_addc_u32 s25, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s51
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s28
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s30
+; GFX7-HSA-NEXT:    s_add_u32 s30, s36, 0xd0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s50
+; GFX7-HSA-NEXT:    s_addc_u32 s50, s37, 0
+; GFX7-HSA-NEXT:    s_add_u32 s28, s36, 0xc0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s49
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v28, s30
+; GFX7-HSA-NEXT:    s_addc_u32 s30, s37, 0
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[22:25]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v29, s50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s28
+; GFX7-HSA-NEXT:    s_add_u32 s28, s36, 0xb0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s30
+; GFX7-HSA-NEXT:    s_addc_u32 s30, s37, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s28
+; GFX7-HSA-NEXT:    s_add_u32 s28, s36, 0xa0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s30
+; GFX7-HSA-NEXT:    s_addc_u32 s30, s37, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s28
+; GFX7-HSA-NEXT:    s_add_u32 s28, s36, 0x90
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s48
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s47
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s30
+; GFX7-HSA-NEXT:    s_addc_u32 s30, s37, 0
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[8:11]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s44
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s28
+; GFX7-HSA-NEXT:    s_add_u32 s28, s36, 0x80
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s30
+; GFX7-HSA-NEXT:    s_addc_u32 s30, s37, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s28
+; GFX7-HSA-NEXT:    s_add_u32 s28, s36, 0x70
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s30
+; GFX7-HSA-NEXT:    s_addc_u32 s30, s37, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v28, s28
+; GFX7-HSA-NEXT:    s_add_u32 s28, s36, 0x60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v29, s30
+; GFX7-HSA-NEXT:    s_addc_u32 s30, s37, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s43
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v30, s28
+; GFX7-HSA-NEXT:    s_add_u32 s28, s36, 0x50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s46
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s45
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v31, s30
+; GFX7-HSA-NEXT:    s_addc_u32 s30, s37, 0
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[0:3]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[4:7]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s39
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s38
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s14
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s15
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s42
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s41
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[4:7]
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[31:32], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s49
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s39
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s42
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s41
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s31
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s29
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[0:3]
 ; GFX7-HSA-NEXT:    s_add_u32 s14, s36, 64
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s12
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v36, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v35, s24
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s40
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s35
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[29:30], v[11:14]
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[22:25]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s38
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s31
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v32, s28
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v33, s30
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s40
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s35
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s34
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[12:15]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[19:22]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s27
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s33
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s26
 ; GFX7-HSA-NEXT:    s_addc_u32 s15, s37, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[33:34], v[3:6]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[30:31], v[3:6]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s11
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s29
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s27
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[16:19]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s25
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s24
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s14
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s15
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[35:36], v[6:9]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[6:9]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s9
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s8
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[9:12]
@@ -4311,7 +4316,7 @@ define amdgpu_kernel void @constant_sextload_v32i32_to_v32i64(ptr addrspace(1) %
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s34
+; GFX7-HSA-NEXT:    s_nop 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-HSA-NEXT:    s_add_u32 s4, s36, 32
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
@@ -4321,15 +4326,15 @@ define amdgpu_kernel void @constant_sextload_v32i32_to_v32i64(ptr addrspace(1) %
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s33
+; GFX7-HSA-NEXT:    s_nop 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s36, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s37, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s36
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -4337,7 +4342,6 @@ define amdgpu_kernel void @constant_sextload_v32i32_to_v32i64(ptr addrspace(1) %
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s37
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[15:18]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
 ;
@@ -4526,8 +4530,8 @@ define amdgpu_kernel void @constant_sextload_v32i32_to_v32i64(ptr addrspace(1) %
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s37, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s38
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s35
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s36
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -5103,37 +5107,37 @@ define amdgpu_kernel void @constant_zextload_v32i32_to_v32i64(ptr addrspace(1) %
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s36, 0xf0
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s36, 0xe0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s0
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s36, 0xd0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s1
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s0
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s36, 0xc0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s1
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s0
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s36, 0xb0
 ; GFX7-HSA-NEXT:    s_load_dwordx16 s[16:31], s[38:39], 0x10
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s0
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s36, 0xa0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s1
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s0
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s36, 0x90
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s30
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s31
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s1
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s37, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s1
 ; GFX7-HSA-NEXT:    s_load_dwordx16 s[0:15], s[38:39], 0x0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_nop 0
@@ -5166,29 +5170,29 @@ define amdgpu_kernel void @constant_zextload_v32i32_to_v32i64(ptr addrspace(1) %
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s17
 ; GFX7-HSA-NEXT:    s_addc_u32 s17, s37, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s17
 ; GFX7-HSA-NEXT:    s_add_u32 s16, s36, 0x70
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s17
 ; GFX7-HSA-NEXT:    s_addc_u32 s17, s37, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s17
 ; GFX7-HSA-NEXT:    s_add_u32 s16, s36, 0x60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s17
 ; GFX7-HSA-NEXT:    s_addc_u32 s17, s37, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s17
 ; GFX7-HSA-NEXT:    s_add_u32 s16, s36, 0x50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s17
 ; GFX7-HSA-NEXT:    s_addc_u32 s17, s37, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s17
 ; GFX7-HSA-NEXT:    s_add_u32 s16, s36, 64
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s17
 ; GFX7-HSA-NEXT:    s_addc_u32 s17, s37, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s17
 ; GFX7-HSA-NEXT:    s_add_u32 s16, s36, 48
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    s_addc_u32 s17, s37, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s17
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s14
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s15
+; GFX7-HSA-NEXT:    s_addc_u32 s17, s37, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[0:3]
 ; GFX7-HSA-NEXT:    s_add_u32 s14, s36, 32
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s12
@@ -5233,10 +5237,10 @@ define amdgpu_kernel void @constant_zextload_v32i32_to_v32i64(ptr addrspace(1) %
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    s_load_dwordx16 s[16:31], s[38:39], 0x40
 ; GFX8-NOHSA-NEXT:    s_load_dwordx16 s[0:15], s[38:39], 0x0
-; GFX8-NOHSA-NEXT:    s_add_u32 s34, s36, 0xf0
-; GFX8-NOHSA-NEXT:    s_addc_u32 s35, s37, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s34
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s35
+; GFX8-NOHSA-NEXT:    s_add_u32 s33, s36, 0xf0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s34, s37, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s33
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s34
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s30
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s31
@@ -5350,8 +5354,8 @@ define amdgpu_kernel void @constant_zextload_v32i32_to_v32i64(ptr addrspace(1) %
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s36, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s37, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s36
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -5775,19 +5779,19 @@ define amdgpu_kernel void @constant_load_v32i32(ptr addrspace(1) %out, ptr addrs
 ; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 0x70
 ; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s19
 ; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 0x60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s19
 ; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
 ; GFX7-HSA-NEXT:    s_nop 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 0x50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX7-HSA-NEXT:    s_add_u32 s18, s16, 64
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX7-HSA-NEXT:    s_addc_u32 s19, s17, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[8:11]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s18
@@ -5838,10 +5842,10 @@ define amdgpu_kernel void @constant_load_v32i32(ptr addrspace(1) %out, ptr addrs
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    s_load_dwordx16 s[16:31], s[38:39], 0x40
 ; GFX8-NOHSA-NEXT:    s_load_dwordx16 s[0:15], s[38:39], 0x0
-; GFX8-NOHSA-NEXT:    s_add_u32 s34, s36, 0x70
-; GFX8-NOHSA-NEXT:    s_addc_u32 s35, s37, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s34
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v6, s35
+; GFX8-NOHSA-NEXT:    s_add_u32 s33, s36, 0x70
+; GFX8-NOHSA-NEXT:    s_addc_u32 s34, s37, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s33
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v6, s34
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s28
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s29
diff --git a/llvm/test/CodeGen/AMDGPU/load-constant-i64.ll b/llvm/test/CodeGen/AMDGPU/load-constant-i64.ll
index 9b74de1c3eb60..c17c782ec1537 100644
--- a/llvm/test/CodeGen/AMDGPU/load-constant-i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-constant-i64.ll
@@ -193,8 +193,8 @@ define amdgpu_kernel void @constant_load_v3i64(ptr addrspace(1) %out, ptr addrsp
 ; GFX7-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX7-NEXT:    s_add_u32 s2, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-NEXT:    v_mov_b32_e32 v4, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s2
+; GFX7-NEXT:    v_mov_b32_e32 v4, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v5, s8
 ; GFX7-NEXT:    v_mov_b32_e32 v6, s9
@@ -216,8 +216,8 @@ define amdgpu_kernel void @constant_load_v3i64(ptr addrspace(1) %out, ptr addrsp
 ; GFX8-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX8-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v4, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s3
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s8
 ; GFX8-NEXT:    v_mov_b32_e32 v6, s9
@@ -647,19 +647,19 @@ define amdgpu_kernel void @constant_load_v16i64(ptr addrspace(1) %out, ptr addrs
 ; GFX7-NEXT:    s_add_u32 s18, s16, 0x70
 ; GFX7-NEXT:    s_addc_u32 s19, s17, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v16, s18
-; GFX7-NEXT:    v_mov_b32_e32 v17, s19
 ; GFX7-NEXT:    s_add_u32 s18, s16, 0x60
+; GFX7-NEXT:    v_mov_b32_e32 v17, s19
 ; GFX7-NEXT:    s_addc_u32 s19, s17, 0
 ; GFX7-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
 ; GFX7-NEXT:    s_nop 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s18
-; GFX7-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX7-NEXT:    s_add_u32 s18, s16, 0x50
+; GFX7-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX7-NEXT:    s_addc_u32 s19, s17, 0
 ; GFX7-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s18
-; GFX7-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX7-NEXT:    s_add_u32 s18, s16, 64
+; GFX7-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX7-NEXT:    s_addc_u32 s19, s17, 0
 ; GFX7-NEXT:    flat_store_dwordx4 v[0:1], v[8:11]
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s18
@@ -710,10 +710,10 @@ define amdgpu_kernel void @constant_load_v16i64(ptr addrspace(1) %out, ptr addrs
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    s_load_dwordx16 s[16:31], s[38:39], 0x40
 ; GFX8-NEXT:    s_load_dwordx16 s[0:15], s[38:39], 0x0
-; GFX8-NEXT:    s_add_u32 s34, s36, 0x70
-; GFX8-NEXT:    s_addc_u32 s35, s37, 0
-; GFX8-NEXT:    v_mov_b32_e32 v5, s34
-; GFX8-NEXT:    v_mov_b32_e32 v6, s35
+; GFX8-NEXT:    s_add_u32 s33, s36, 0x70
+; GFX8-NEXT:    s_addc_u32 s34, s37, 0
+; GFX8-NEXT:    v_mov_b32_e32 v5, s33
+; GFX8-NEXT:    v_mov_b32_e32 v6, s34
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s28
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s29
diff --git a/llvm/test/CodeGen/AMDGPU/load-constant-i8.ll b/llvm/test/CodeGen/AMDGPU/load-constant-i8.ll
index b72378e801b01..f6c5d72583710 100644
--- a/llvm/test/CodeGen/AMDGPU/load-constant-i8.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-constant-i8.ll
@@ -231,11 +231,11 @@ define amdgpu_kernel void @constant_load_v3i8(ptr addrspace(1) %out, ptr addrspa
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s0, 2
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    s_lshr_b32 s0, s2, 16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX7-HSA-NEXT:    flat_store_short v[0:1], v4
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-HSA-NEXT:    flat_store_byte v[2:3], v0
@@ -250,11 +250,11 @@ define amdgpu_kernel void @constant_load_v3i8(ptr addrspace(1) %out, ptr addrspa
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NOHSA-NEXT:    s_add_u32 s0, s0, 2
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s1, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s0, s2, 16
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NOHSA-NEXT:    flat_store_short v[0:1], v4
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX8-NOHSA-NEXT:    flat_store_byte v[2:3], v0
@@ -1544,22 +1544,22 @@ define amdgpu_kernel void @constant_zextload_v8i8_to_v8i32(ptr addrspace(1) %out
 ; GFX7-HSA-NEXT:    s_lshr_b32 s6, s3, 24
 ; GFX7-HSA-NEXT:    s_bfe_u32 s7, s3, 0x80008
 ; GFX7-HSA-NEXT:    s_and_b32 s8, s2, 0xff
-; GFX7-HSA-NEXT:    s_bfe_u32 s9, s2, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s2, s3, 0xff
+; GFX7-HSA-NEXT:    s_bfe_u32 s2, s2, 0x80010
+; GFX7-HSA-NEXT:    s_and_b32 s9, s3, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s3, s3, 0x80010
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s7
+; GFX7-HSA-NEXT:    s_add_u32 s3, s0, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s6
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    s_addc_u32 s6, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s8
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -1576,22 +1576,22 @@ define amdgpu_kernel void @constant_zextload_v8i8_to_v8i32(ptr addrspace(1) %out
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s6, s3, 24
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s7, s3, 0x80008
 ; GFX8-NOHSA-NEXT:    s_and_b32 s8, s2, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s9, s2, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s2, s3, 0xff
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s2, s2, 0x80010
+; GFX8-NOHSA-NEXT:    s_and_b32 s9, s3, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s3, s3, 0x80010
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s7
+; GFX8-NOHSA-NEXT:    s_add_u32 s3, s0, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    s_addc_u32 s6, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s8
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s5
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -1701,22 +1701,22 @@ define amdgpu_kernel void @constant_sextload_v8i8_to_v8i32(ptr addrspace(1) %out
 ; GFX7-HSA-NEXT:    s_ashr_i32 s4, s2, 24
 ; GFX7-HSA-NEXT:    s_bfe_i32 s5, s2, 0x80010
 ; GFX7-HSA-NEXT:    s_bfe_i32 s6, s2, 0x80008
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s7, s2
-; GFX7-HSA-NEXT:    s_ashr_i32 s2, s3, 24
+; GFX7-HSA-NEXT:    s_ashr_i32 s7, s3, 24
 ; GFX7-HSA-NEXT:    s_bfe_i32 s8, s3, 0x80010
 ; GFX7-HSA-NEXT:    s_bfe_i32 s9, s3, 0x80008
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s3, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s3
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    s_add_u32 s3, s0, 16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s7
+; GFX7-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s2, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s8
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s6
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s4
@@ -1733,22 +1733,22 @@ define amdgpu_kernel void @constant_sextload_v8i8_to_v8i32(ptr addrspace(1) %out
 ; GFX8-NOHSA-NEXT:    s_ashr_i32 s4, s2, 24
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s5, s2, 0x80010
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s6, s2, 0x80008
-; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s7, s2
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s2, s3, 24
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s7, s3, 24
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s8, s3, 0x80010
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s9, s3, 0x80008
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s3, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s3
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    s_add_u32 s3, s0, 16
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s7
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s2, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s8
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s6
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s4
@@ -1881,14 +1881,14 @@ define amdgpu_kernel void @constant_zextload_v16i8_to_v16i32(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_lshr_b32 s8, s4, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s9, s4, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s10, s5, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s11, s5, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s12, s6, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s13, s6, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s2, s7, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s3, s7, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s2, s4, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s3, s4, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s8, s5, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s9, s5, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s10, s6, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s11, s6, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s12, s7, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s13, s7, 0x80008
 ; GFX7-HSA-NEXT:    s_and_b32 s14, s4, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s4, s4, 0x80010
 ; GFX7-HSA-NEXT:    s_and_b32 s15, s5, 0xff
@@ -1897,38 +1897,40 @@ define amdgpu_kernel void @constant_zextload_v16i8_to_v16i32(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_bfe_u32 s6, s6, 0x80010
 ; GFX7-HSA-NEXT:    s_and_b32 s17, s7, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s7, s7, 0x80010
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 48
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 32
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s17
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s7
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s6
+; GFX7-HSA-NEXT:    s_add_u32 s7, s0, 48
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s12
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    s_addc_u32 s12, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s17
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s13
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s12
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s15
+; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s6
+; GFX7-HSA-NEXT:    s_add_u32 s6, s0, 32
+; GFX7-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
+; GFX7-HSA-NEXT:    s_add_u32 s5, s0, 16
+; GFX7-HSA-NEXT:    s_addc_u32 s6, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s15
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s4
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
@@ -1939,14 +1941,14 @@ define amdgpu_kernel void @constant_zextload_v16i8_to_v16i32(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s8, s4, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s9, s4, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s10, s5, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s11, s5, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s12, s6, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s13, s6, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s2, s7, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s3, s7, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s2, s4, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s3, s4, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s8, s5, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s9, s5, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s10, s6, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s11, s6, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s12, s7, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s13, s7, 0x80008
 ; GFX8-NOHSA-NEXT:    s_and_b32 s14, s4, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s4, s4, 0x80010
 ; GFX8-NOHSA-NEXT:    s_and_b32 s15, s5, 0xff
@@ -1955,38 +1957,40 @@ define amdgpu_kernel void @constant_zextload_v16i8_to_v16i32(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s6, s6, 0x80010
 ; GFX8-NOHSA-NEXT:    s_and_b32 s17, s7, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s7, s7, 0x80010
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 48
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s3
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 32
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s17
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s7
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s13
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s6
+; GFX8-NOHSA-NEXT:    s_add_u32 s7, s0, 48
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s12
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s12, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s17
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s13
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s12
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s15
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s6
+; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 32
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s11
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s10
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s5
+; GFX8-NOHSA-NEXT:    s_add_u32 s5, s0, 16
+; GFX8-NOHSA-NEXT:    s_addc_u32 s6, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s15
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s14
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s4
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_endpgm
@@ -2144,54 +2148,54 @@ define amdgpu_kernel void @constant_sextload_v16i8_to_v16i32(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_ashr_i32 s8, s4, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s9, s4, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s10, s4, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s11, s5, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s12, s5, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s13, s5, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s14, s6, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s15, s6, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s16, s6, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s2, s7, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s3, s7, 0x80010
+; GFX7-HSA-NEXT:    s_ashr_i32 s2, s4, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s3, s4, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s8, s4, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s9, s5, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s10, s5, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s11, s5, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s12, s6, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s13, s6, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s14, s6, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s15, s7, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s16, s7, 0x80010
 ; GFX7-HSA-NEXT:    s_bfe_i32 s17, s7, 0x80008
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 48
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s7, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s17
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    s_add_u32 s7, s0, 48
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s15
+; GFX7-HSA-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s6, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s17
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s5, s5
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s15
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s14
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s4, s4
+; GFX7-HSA-NEXT:    s_add_u32 s6, s0, 32
+; GFX7-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s14
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s13
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s12
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s4, s4
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    s_add_u32 s5, s0, 16
+; GFX7-HSA-NEXT:    s_addc_u32 s6, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s11
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s10
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
@@ -2202,54 +2206,54 @@ define amdgpu_kernel void @constant_sextload_v16i8_to_v16i32(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s8, s4, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s9, s4, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s10, s4, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s11, s5, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s12, s5, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s13, s5, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s14, s6, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s15, s6, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s16, s6, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s2, s7, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s3, s7, 0x80010
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s2, s4, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s3, s4, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s8, s4, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s9, s5, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s10, s5, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s11, s5, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s12, s6, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s13, s6, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s14, s6, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s15, s7, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s16, s7, 0x80010
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s17, s7, 0x80008
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 48
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s7, s7
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 32
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s7
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s17
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    s_add_u32 s7, s0, 48
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s15
+; GFX8-NOHSA-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s6, s6
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s17
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s16
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s5, s5
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s15
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s14
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s4, s4
+; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 32
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s14
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s13
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s4, s4
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s5
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s13
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s12
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s11
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    s_add_u32 s5, s0, 16
+; GFX8-NOHSA-NEXT:    s_addc_u32 s6, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s11
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s10
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s9
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_endpgm
@@ -2455,106 +2459,108 @@ define amdgpu_kernel void @constant_zextload_v32i8_to_v32i32(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_lshr_b32 s12, s4, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s13, s4, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s14, s5, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s15, s5, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s16, s6, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s17, s6, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s18, s7, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s19, s7, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s20, s8, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s21, s8, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s22, s9, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s23, s9, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s24, s10, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s25, s10, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s26, s11, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s27, s11, 0x80008
-; GFX7-HSA-NEXT:    s_and_b32 s28, s4, 0xff
+; GFX7-HSA-NEXT:    s_lshr_b32 s2, s4, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s3, s4, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s12, s5, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s13, s5, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s14, s6, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s15, s6, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s16, s7, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s17, s7, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s18, s8, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s19, s8, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s20, s9, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s21, s9, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s22, s10, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s23, s10, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s24, s11, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s25, s11, 0x80008
+; GFX7-HSA-NEXT:    s_and_b32 s26, s4, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s4, s4, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s29, s5, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s27, s5, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s5, s5, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s30, s6, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s28, s6, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s6, s6, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s31, s7, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s29, s7, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s7, s7, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s33, s8, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s30, s8, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s8, s8, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s34, s9, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s31, s9, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s9, s9, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s35, s10, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s33, s10, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s10, s10, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s36, s11, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s34, s11, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s11, s11, 0x80010
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 0x70
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 0x60
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 0x50
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 64
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s27
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0x70
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0x60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0x50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s34
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s25
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s26
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s24
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s35
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 48
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s24
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s34
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s23
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s33
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s22
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s21
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s8
+; GFX7-HSA-NEXT:    s_add_u32 s8, s0, 64
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s9
+; GFX7-HSA-NEXT:    s_addc_u32 s9, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s33
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s23
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s10
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s22
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s7
+; GFX7-HSA-NEXT:    s_add_u32 s7, s0, 48
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s35
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s36
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s31
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s21
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s30
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s20
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s19
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[4:7]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[8:11]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s20
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s18
+; GFX7-HSA-NEXT:    s_addc_u32 s8, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[12:15]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s31
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s19
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s18
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s30
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s29
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s17
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s16
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s8
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s29
+; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s6
+; GFX7-HSA-NEXT:    s_add_u32 s6, s0, 32
+; GFX7-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s28
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s15
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s28
+; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
+; GFX7-HSA-NEXT:    s_add_u32 s5, s0, 16
+; GFX7-HSA-NEXT:    s_addc_u32 s6, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s27
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s12
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s6
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s26
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s4
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
@@ -2565,22 +2571,22 @@ define amdgpu_kernel void @constant_zextload_v32i8_to_v32i32(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s12, s4, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s13, s4, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s14, s5, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s15, s5, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s16, s6, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s17, s6, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s18, s7, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s19, s7, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s20, s8, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s21, s8, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s22, s9, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s23, s9, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s24, s10, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s25, s10, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s2, s11, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s3, s11, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s2, s4, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s3, s4, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s12, s5, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s13, s5, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s14, s6, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s15, s6, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s16, s7, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s17, s7, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s18, s8, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s19, s8, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s20, s9, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s21, s9, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s22, s10, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s23, s10, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s24, s11, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s25, s11, 0x80008
 ; GFX8-NOHSA-NEXT:    s_and_b32 s26, s4, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s4, s4, 0x80010
 ; GFX8-NOHSA-NEXT:    s_and_b32 s27, s5, 0xff
@@ -2597,74 +2603,80 @@ define amdgpu_kernel void @constant_zextload_v32i8_to_v32i32(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s10, s10, 0x80010
 ; GFX8-NOHSA-NEXT:    s_and_b32 s34, s11, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s11, s11, 0x80010
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 0x70
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s3
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 0x60
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s34
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s11
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 0x50
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s33
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s25
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s11, s0, 0x70
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s24
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s24, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s34
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s25
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s11
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s24
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 64
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s31
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 0x60
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s33
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s23
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s9
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s22
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 48
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s30
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s9
+; GFX8-NOHSA-NEXT:    s_add_u32 s9, s0, 0x50
+; GFX8-NOHSA-NEXT:    s_addc_u32 s10, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s31
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s21
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s8
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s20
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s10
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 32
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s29
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NOHSA-NEXT:    s_add_u32 s8, s0, 64
+; GFX8-NOHSA-NEXT:    s_addc_u32 s9, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s30
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s19
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s7
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s18
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s9
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s28
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s7
+; GFX8-NOHSA-NEXT:    s_add_u32 s7, s0, 48
+; GFX8-NOHSA-NEXT:    s_addc_u32 s8, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s29
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s17
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s16
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s8
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s27
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s6
+; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 32
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s28
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s15
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s14
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s5
+; GFX8-NOHSA-NEXT:    s_add_u32 s5, s0, 16
+; GFX8-NOHSA-NEXT:    s_addc_u32 s6, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s27
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s13
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s26
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s13
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s4
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_endpgm
@@ -2929,106 +2941,106 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i32(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_ashr_i32 s12, s4, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s13, s4, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s14, s4, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s15, s5, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s16, s5, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s17, s5, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s18, s6, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s19, s6, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s20, s6, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s21, s7, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s22, s7, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s23, s7, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s24, s8, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s25, s8, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s26, s8, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s27, s9, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s28, s9, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s29, s9, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s30, s10, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s31, s10, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s33, s10, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s34, s11, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s35, s11, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s36, s11, 0x80008
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 0x70
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 0x60
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 0x50
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s11, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 64
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s35
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s34
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s10, s10
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[0:3]
+; GFX7-HSA-NEXT:    s_ashr_i32 s2, s4, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s3, s4, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s12, s4, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s13, s5, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s14, s5, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s15, s5, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s16, s6, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s17, s6, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s18, s6, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s19, s7, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s20, s7, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s21, s7, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s22, s8, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s23, s8, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s24, s8, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s25, s9, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s26, s9, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s27, s9, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s28, s10, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s29, s10, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s30, s10, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s31, s11, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s33, s11, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s34, s11, 0x80008
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0x70
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0x60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0x50
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s8, s8
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s9, s9
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s10, s10
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s11, s11
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s8
+; GFX7-HSA-NEXT:    s_add_u32 s8, s0, 64
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s11
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s34
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s33
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s31
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s33
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s31
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s30
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s7, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s30
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s29
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s28
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s29
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s28
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s8
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s27
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s26
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s25
+; GFX7-HSA-NEXT:    s_addc_u32 s9, s1, 0
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s7, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s35
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s36
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s27
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s26
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[0:3]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s25
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s24
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s23
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[4:7]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[8:11]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s24
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s22
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[12:15]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s23
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s22
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s21
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    s_add_u32 s7, s0, 48
+; GFX7-HSA-NEXT:    s_addc_u32 s8, s1, 0
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s6, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s21
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s20
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s19
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s8
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s5, s5
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s20
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s19
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s18
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s4, s4
+; GFX7-HSA-NEXT:    s_add_u32 s6, s0, 32
+; GFX7-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s18
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s17
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s4, s4
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s17
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s15
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    s_add_u32 s5, s0, 16
+; GFX7-HSA-NEXT:    s_addc_u32 s6, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s15
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s14
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s13
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s12
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s12
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
@@ -3039,106 +3051,106 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i32(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s12, s4, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s13, s4, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s14, s4, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s15, s5, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s16, s5, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s17, s5, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s18, s6, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s19, s6, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s20, s6, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s21, s7, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s22, s7, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s23, s7, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s24, s8, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s25, s8, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s26, s8, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s27, s9, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s28, s9, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s29, s9, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s30, s10, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s31, s10, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s33, s10, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s2, s11, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s3, s11, 0x80010
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s2, s4, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s3, s4, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s12, s4, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s13, s5, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s14, s5, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s15, s5, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s16, s6, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s17, s6, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s18, s6, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s19, s7, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s20, s7, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s21, s7, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s22, s8, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s23, s8, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s24, s8, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s25, s9, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s26, s9, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s27, s9, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s28, s10, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s29, s10, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s30, s10, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s31, s11, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s33, s11, 0x80010
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s34, s11, 0x80008
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 0x70
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s11, s11
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 0x60
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s11
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s34
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    s_add_u32 s11, s0, 0x70
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s31
+; GFX8-NOHSA-NEXT:    s_addc_u32 s31, s1, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s10, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s34
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s33
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s11
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s31
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 0x50
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s10
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s33
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s31
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s30
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s9, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 0x60
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s30
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s29
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s28
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 64
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s9
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s29
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s28
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s27
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s8, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s9
+; GFX8-NOHSA-NEXT:    s_add_u32 s9, s0, 0x50
+; GFX8-NOHSA-NEXT:    s_addc_u32 s10, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s27
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s26
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s25
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s10
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 48
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s8
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s26
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s25
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s24
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s7, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s8
+; GFX8-NOHSA-NEXT:    s_add_u32 s8, s0, 64
+; GFX8-NOHSA-NEXT:    s_addc_u32 s9, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s24
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s23
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s22
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s9
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 32
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s7
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s23
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s22
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s21
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s6, s6
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s7
+; GFX8-NOHSA-NEXT:    s_add_u32 s7, s0, 48
+; GFX8-NOHSA-NEXT:    s_addc_u32 s8, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s21
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s20
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s19
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s8
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s5, s5
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s20
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s19
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s18
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s4, s4
+; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 32
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s18
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s17
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s16
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s4, s4
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s5
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s17
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s15
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    s_add_u32 s5, s0, 16
+; GFX8-NOHSA-NEXT:    s_addc_u32 s6, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s15
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s14
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s13
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s14
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s13
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_endpgm
@@ -3507,209 +3519,207 @@ define amdgpu_kernel void @constant_zextload_v64i8_to_v64i32(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_bfe_u32 s19, s0, 0x80008
 ; GFX7-HSA-NEXT:    s_lshr_b32 s20, s1, 24
 ; GFX7-HSA-NEXT:    s_bfe_u32 s21, s1, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s22, s2, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s23, s2, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s25, s3, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s26, s3, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s28, s4, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s29, s4, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s31, s5, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s33, s5, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s35, s6, 24
+; GFX7-HSA-NEXT:    s_lshr_b32 s23, s2, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s24, s2, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s26, s3, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s27, s3, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s29, s4, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s30, s4, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s33, s5, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s34, s5, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s36, s6, 24
 ; GFX7-HSA-NEXT:    s_bfe_u32 s37, s6, 0x80008
 ; GFX7-HSA-NEXT:    s_lshr_b32 s38, s7, 24
 ; GFX7-HSA-NEXT:    s_bfe_u32 s39, s7, 0x80008
 ; GFX7-HSA-NEXT:    s_lshr_b32 s41, s8, 24
 ; GFX7-HSA-NEXT:    s_bfe_u32 s43, s8, 0x80008
 ; GFX7-HSA-NEXT:    s_lshr_b32 s44, s9, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s45, s9, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s46, s10, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s47, s10, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s48, s11, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s49, s11, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s50, s12, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s51, s12, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s52, s13, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s53, s13, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s54, s14, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s55, s14, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s56, s15, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s57, s15, 0x80008
-; GFX7-HSA-NEXT:    s_and_b32 s24, s0, 0xff
+; GFX7-HSA-NEXT:    s_bfe_u32 s46, s9, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s47, s10, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s48, s10, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s49, s11, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s50, s11, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s51, s12, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s52, s12, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s53, s13, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s54, s13, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s55, s14, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s56, s14, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s57, s15, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s58, s15, 0x80008
+; GFX7-HSA-NEXT:    s_and_b32 s22, s0, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s0, s0, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s27, s1, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s25, s1, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s1, s1, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s30, s2, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s28, s2, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s2, s2, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s34, s3, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s31, s3, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s3, s3, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s36, s4, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s35, s4, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s4, s4, 0x80010
 ; GFX7-HSA-NEXT:    s_and_b32 s40, s5, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s5, s5, 0x80010
 ; GFX7-HSA-NEXT:    s_and_b32 s42, s6, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s6, s6, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s58, s7, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s45, s7, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s7, s7, 0x80010
 ; GFX7-HSA-NEXT:    s_and_b32 s59, s8, 0xff
-; GFX7-HSA-NEXT:    s_bfe_u32 s60, s8, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s61, s9, 0xff
-; GFX7-HSA-NEXT:    s_bfe_u32 s62, s9, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s63, s10, 0xff
+; GFX7-HSA-NEXT:    s_bfe_u32 s8, s8, 0x80010
+; GFX7-HSA-NEXT:    s_and_b32 s60, s9, 0xff
+; GFX7-HSA-NEXT:    s_bfe_u32 s9, s9, 0x80010
+; GFX7-HSA-NEXT:    s_and_b32 s61, s10, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s10, s10, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s64, s11, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s62, s11, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s11, s11, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s65, s12, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s63, s12, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s12, s12, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s66, s13, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s64, s13, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s13, s13, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s67, s14, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s65, s14, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s14, s14, 0x80010
-; GFX7-HSA-NEXT:    s_and_b32 s68, s15, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s66, s15, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s15, s15, 0x80010
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xf0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xe0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xd0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xc0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xb0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xa0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0x90
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s66
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s53
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s52
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[8:11]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0x80
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s65
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s51
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s50
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s68
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s57
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s15
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s56
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[12:15]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0x70
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s64
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s49
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s48
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s67
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s55
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s54
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[0:3]
+; GFX7-HSA-NEXT:    s_add_u32 s67, s16, 0xf0
+; GFX7-HSA-NEXT:    s_addc_u32 s68, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s67
+; GFX7-HSA-NEXT:    s_add_u32 s67, s16, 0xe0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s68
+; GFX7-HSA-NEXT:    s_addc_u32 s68, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s67
+; GFX7-HSA-NEXT:    s_add_u32 s67, s16, 0xd0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s68
+; GFX7-HSA-NEXT:    s_addc_u32 s68, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s67
+; GFX7-HSA-NEXT:    s_add_u32 s67, s16, 0xc0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s68
+; GFX7-HSA-NEXT:    s_addc_u32 s68, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s67
+; GFX7-HSA-NEXT:    s_add_u32 s67, s16, 0xb0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s68
+; GFX7-HSA-NEXT:    s_addc_u32 s68, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s65
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s56
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s14
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s55
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s66
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s58
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s15
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s57
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[8:11]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s10
+; GFX7-HSA-NEXT:    s_add_u32 s10, s16, 0xa0
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[4:7]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s11
+; GFX7-HSA-NEXT:    s_addc_u32 s11, s17, 0
+; GFX7-HSA-NEXT:    s_add_u32 s9, s16, 0x90
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s10
+; GFX7-HSA-NEXT:    s_addc_u32 s10, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v28, s9
+; GFX7-HSA-NEXT:    s_add_u32 s9, s16, 0x80
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s67
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s68
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s62
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s49
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v29, s10
+; GFX7-HSA-NEXT:    s_addc_u32 s10, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s64
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s54
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s13
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s53
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s63
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX7-HSA-NEXT:    s_add_u32 s7, s16, 0x70
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s52
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s12
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s51
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[12:15]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[16:19]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s8
+; GFX7-HSA-NEXT:    s_addc_u32 s8, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s6
 ; GFX7-HSA-NEXT:    s_add_u32 s6, s16, 0x60
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[4:7]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s47
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s46
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s7
-; GFX7-HSA-NEXT:    s_addc_u32 s7, s17, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[16:19]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s61
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s6
-; GFX7-HSA-NEXT:    s_add_u32 s6, s16, 0x50
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s45
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s62
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s44
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s42
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s37
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s35
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s46
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s44
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s45
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s7
 ; GFX7-HSA-NEXT:    s_addc_u32 s7, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s59
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s43
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s60
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s58
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s41
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s39
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s38
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s8
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[4:7]
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[8:11]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s59
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s43
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s10
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s41
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s42
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s39
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s38
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s37
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s36
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s7
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[8:11]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[12:15]
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s40
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s33
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[16:19]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s31
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX7-HSA-NEXT:    s_add_u32 s5, s16, 0x50
+; GFX7-HSA-NEXT:    s_addc_u32 s6, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s40
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s34
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s33
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s61
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX7-HSA-NEXT:    s_add_u32 s4, s16, 64
 ; GFX7-HSA-NEXT:    s_addc_u32 s5, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s35
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s30
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s29
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s5
-; GFX7-HSA-NEXT:    s_add_u32 s4, s16, 48
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s29
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s28
-; GFX7-HSA-NEXT:    s_addc_u32 s5, s17, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s4
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s34
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s26
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s48
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s5
+; GFX7-HSA-NEXT:    s_add_u32 s3, s16, 48
+; GFX7-HSA-NEXT:    s_addc_u32 s4, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s31
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s27
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s26
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s4
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s47
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s16, 32
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s28
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s24
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s23
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s16, 16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s30
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s23
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s22
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s17, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s27
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s21
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s11
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s1
+; GFX7-HSA-NEXT:    s_add_u32 s1, s16, 16
+; GFX7-HSA-NEXT:    s_addc_u32 s2, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s25
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s21
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s20
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s1
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s2
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s24
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s18
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s17
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[20:23]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
 ;
@@ -3723,205 +3733,217 @@ define amdgpu_kernel void @constant_zextload_v64i8_to_v64i32(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s19, s0, 0x80008
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s20, s1, 24
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s21, s1, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s22, s2, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s23, s2, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s25, s3, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s26, s3, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s28, s4, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s29, s4, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s31, s5, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s33, s5, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s35, s6, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s36, s6, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s37, s7, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s38, s7, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s39, s8, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s40, s8, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s41, s9, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s42, s9, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s43, s10, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s44, s10, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s45, s11, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s46, s11, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s47, s12, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s48, s12, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s49, s13, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s50, s13, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s51, s14, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s52, s14, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s53, s15, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s54, s15, 0x80008
-; GFX8-NOHSA-NEXT:    s_and_b32 s24, s0, 0xff
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s23, s2, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s24, s2, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s26, s3, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s27, s3, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s29, s4, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s30, s4, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s33, s5, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s34, s5, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s36, s6, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s37, s6, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s39, s7, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s40, s7, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s41, s8, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s42, s8, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s43, s9, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s44, s9, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s45, s10, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s46, s10, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s47, s11, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s48, s11, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s49, s12, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s50, s12, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s51, s13, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s52, s13, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s53, s14, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s54, s14, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s55, s15, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s56, s15, 0x80008
+; GFX8-NOHSA-NEXT:    s_and_b32 s22, s0, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s0, s0, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s27, s1, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s25, s1, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s1, s1, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s30, s2, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s28, s2, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s2, s2, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s34, s3, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s31, s3, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s3, s3, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s55, s4, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s35, s4, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s4, s4, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s56, s5, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s38, s5, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s5, s5, 0x80010
 ; GFX8-NOHSA-NEXT:    s_and_b32 s57, s6, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s58, s6, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s59, s7, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s60, s7, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s61, s8, 0xff
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s6, s6, 0x80010
+; GFX8-NOHSA-NEXT:    s_and_b32 s58, s7, 0xff
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s7, s7, 0x80010
+; GFX8-NOHSA-NEXT:    s_and_b32 s59, s8, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s8, s8, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s62, s9, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s60, s9, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s9, s9, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s63, s10, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s61, s10, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s10, s10, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s64, s11, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s62, s11, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s11, s11, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s65, s12, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s63, s12, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s12, s12, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s66, s13, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s64, s13, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s13, s13, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s67, s14, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s65, s14, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s14, s14, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s6, s15, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s7, s15, 0x80010
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s6
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xf0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s7
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xe0
+; GFX8-NOHSA-NEXT:    s_and_b32 s66, s15, 0xff
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s15, s15, 0x80010
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s15
+; GFX8-NOHSA-NEXT:    s_add_u32 s15, s16, 0xf0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s55
+; GFX8-NOHSA-NEXT:    s_addc_u32 s55, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s66
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s56
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s15
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s55
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s14
+; GFX8-NOHSA-NEXT:    s_add_u32 s14, s16, 0xe0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s15, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s65
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s54
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s53
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xd0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s67
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s13
+; GFX8-NOHSA-NEXT:    s_add_u32 s13, s16, 0xd0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s14, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s64
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s52
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s14
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s51
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s13
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s14
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xc0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s66
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s12
+; GFX8-NOHSA-NEXT:    s_add_u32 s12, s16, 0xc0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s13, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s63
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s50
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s13
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s49
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s13
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xb0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s65
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s11
+; GFX8-NOHSA-NEXT:    s_add_u32 s11, s16, 0xb0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s12, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s62
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s48
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s12
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s47
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s11
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s12
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xa0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s64
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s16, 0xa0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s61
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s46
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s11
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s45
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x90
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s63
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s9
+; GFX8-NOHSA-NEXT:    s_add_u32 s9, s16, 0x90
+; GFX8-NOHSA-NEXT:    s_addc_u32 s10, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s60
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s44
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s10
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s43
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s10
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x80
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s62
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NOHSA-NEXT:    s_add_u32 s8, s16, 0x80
+; GFX8-NOHSA-NEXT:    s_addc_u32 s9, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s59
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s42
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s9
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s41
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s9
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x70
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s61
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s7
+; GFX8-NOHSA-NEXT:    s_add_u32 s7, s16, 0x70
+; GFX8-NOHSA-NEXT:    s_addc_u32 s8, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s58
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s40
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s8
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s39
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s8
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x60
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s59
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s38
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s60
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s37
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s57
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s37
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s36
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x50
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s57
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s36
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s58
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s35
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s56
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s33
+; GFX8-NOHSA-NEXT:    s_nop 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s5
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s31
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NOHSA-NEXT:    s_add_u32 s5, s16, 0x50
+; GFX8-NOHSA-NEXT:    s_addc_u32 s6, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s38
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s34
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s33
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_nop 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NOHSA-NEXT:    s_add_u32 s4, s16, 64
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s5, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s35
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s30
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s29
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NOHSA-NEXT:    s_add_u32 s4, s16, 48
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s55
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s29
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s28
-; GFX8-NOHSA-NEXT:    s_addc_u32 s5, s17, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s34
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s26
+; GFX8-NOHSA-NEXT:    s_nop 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s25
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s5
+; GFX8-NOHSA-NEXT:    s_add_u32 s3, s16, 48
+; GFX8-NOHSA-NEXT:    s_addc_u32 s4, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s31
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s27
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s26
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s4
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_nop 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s16, 32
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s17, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s28
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s24
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s23
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s16, 16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s30
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s23
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s22
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s17, 0
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s27
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s21
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    s_nop 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s1
+; GFX8-NOHSA-NEXT:    s_add_u32 s1, s16, 16
+; GFX8-NOHSA-NEXT:    s_addc_u32 s2, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s25
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s21
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s20
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s1
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s2
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s24
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s18
@@ -4431,176 +4453,175 @@ define amdgpu_kernel void @constant_sextload_v64i8_to_v64i32(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_ashr_i32 s43, s8, 24
 ; GFX7-HSA-NEXT:    s_bfe_i32 s44, s8, 0x80010
 ; GFX7-HSA-NEXT:    s_bfe_i32 s45, s8, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s47, s9, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s48, s9, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s49, s9, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s51, s10, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s52, s10, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s53, s10, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s54, s11, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s55, s11, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s56, s11, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s57, s12, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s58, s12, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s59, s12, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s60, s13, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s61, s13, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s62, s13, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s63, s14, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s64, s14, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s65, s14, 0x80008
-; GFX7-HSA-NEXT:    s_ashr_i32 s66, s15, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s67, s15, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_i32 s68, s15, 0x80008
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s46, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xf0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s50, s9
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xe0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xd0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xc0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xb0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s8
-; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0xa0
-; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
+; GFX7-HSA-NEXT:    s_ashr_i32 s46, s9, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s47, s9, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s48, s9, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s49, s10, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s50, s10, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s51, s10, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s52, s11, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s53, s11, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s54, s11, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s55, s12, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s56, s12, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s57, s12, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s58, s13, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s59, s13, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s60, s13, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s61, s14, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s62, s14, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s63, s14, 0x80008
+; GFX7-HSA-NEXT:    s_ashr_i32 s64, s15, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s65, s15, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_i32 s66, s15, 0x80008
+; GFX7-HSA-NEXT:    s_add_u32 s67, s16, 0xf0
+; GFX7-HSA-NEXT:    s_addc_u32 s68, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s67
+; GFX7-HSA-NEXT:    s_add_u32 s67, s16, 0xe0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s68
+; GFX7-HSA-NEXT:    s_addc_u32 s68, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s67
+; GFX7-HSA-NEXT:    s_add_u32 s67, s16, 0xd0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s68
+; GFX7-HSA-NEXT:    s_addc_u32 s68, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s67
+; GFX7-HSA-NEXT:    s_add_u32 s67, s16, 0xc0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s68
+; GFX7-HSA-NEXT:    s_addc_u32 s68, s17, 0
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s14, s14
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s67
+; GFX7-HSA-NEXT:    s_add_u32 s67, s16, 0xb0
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s9, s9
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s15, s15
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s68
+; GFX7-HSA-NEXT:    s_addc_u32 s68, s17, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s14
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s63
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s62
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s61
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s10, s10
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s13, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s15
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s66
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s65
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s64
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[8:11]
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s8, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s9
+; GFX7-HSA-NEXT:    s_add_u32 s9, s16, 0xa0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s13
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s59
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s58
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[4:7]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s10
+; GFX7-HSA-NEXT:    s_addc_u32 s10, s17, 0
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[12:15]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s8
 ; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0x90
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s62
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s61
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s60
 ; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s12, s12
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[8:11]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s8
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s11, s11
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v28, s8
 ; GFX7-HSA-NEXT:    s_add_u32 s8, s16, 0x80
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s15, s15
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s59
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s58
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s57
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s67
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s68
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s11
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s54
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s53
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s52
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v29, s9
 ; GFX7-HSA-NEXT:    s_addc_u32 s9, s17, 0
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s7, s7
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s11, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s15
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s68
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s67
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s66
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[12:15]
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s14, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s6
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX7-HSA-NEXT:    s_add_u32 s6, s16, 0x70
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s56
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s55
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s54
 ; GFX7-HSA-NEXT:    s_addc_u32 s7, s17, 0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s10, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s65
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s64
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s63
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s6
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s12, s12
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX7-HSA-NEXT:    s_add_u32 s6, s16, 0x60
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s10
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[4:7]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s53
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s52
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s51
-; GFX7-HSA-NEXT:    s_addc_u32 s7, s17, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[16:19]
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s5, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s6
-; GFX7-HSA-NEXT:    s_add_u32 s6, s16, 0x50
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s50
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s49
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s48
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s47
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s40
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s39
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s38
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s37
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s12
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s57
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s56
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s55
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s48
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s47
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s46
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s7
 ; GFX7-HSA-NEXT:    s_addc_u32 s7, s17, 0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s4, s4
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s46
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s45
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s8
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s44
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s43
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s42
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s41
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[4:7]
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[8:11]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s45
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[16:19]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s44
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s43
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s40
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s42
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s41
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s39
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s38
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s37
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s7
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[8:11]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[12:15]
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[16:19]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s5
+; GFX7-HSA-NEXT:    s_add_u32 s5, s16, 0x50
+; GFX7-HSA-NEXT:    s_addc_u32 s6, s17, 0
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s4, s4
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s36
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s35
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s34
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s3, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-HSA-NEXT:    s_add_u32 s4, s16, 64
 ; GFX7-HSA-NEXT:    s_addc_u32 s5, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s4
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s5
-; GFX7-HSA-NEXT:    s_add_u32 s4, s16, 48
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s33
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s31
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s30
-; GFX7-HSA-NEXT:    s_addc_u32 s5, s17, 0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s2, s2
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s4
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s5
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s2, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s3
+; GFX7-HSA-NEXT:    s_add_u32 s3, s16, 48
+; GFX7-HSA-NEXT:    s_addc_u32 s4, s17, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s29
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s28
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s27
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s5
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s4
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s1, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s16, 32
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s17, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s16, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s26
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s25
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s24
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s17, 0
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s0, s0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s0, s0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s1
+; GFX7-HSA-NEXT:    s_add_u32 s1, s16, 16
+; GFX7-HSA-NEXT:    s_addc_u32 s2, s17, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s23
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s22
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s21
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s1
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s51
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s49
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s10
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -4608,6 +4629,7 @@ define amdgpu_kernel void @constant_sextload_v64i8_to_v64i32(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s19
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s18
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s17
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[20:23]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
 ;
@@ -4638,183 +4660,183 @@ define amdgpu_kernel void @constant_sextload_v64i8_to_v64i32(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_ashr_i32 s37, s6, 24
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s38, s6, 0x80010
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s39, s6, 0x80008
-; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s40, s6
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s41, s7, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s42, s7, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s43, s7, 0x80008
-; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s44, s7
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s45, s8, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s46, s8, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s47, s8, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s48, s9, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s49, s9, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s50, s9, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s51, s10, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s52, s10, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s53, s10, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s54, s11, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s55, s11, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s56, s11, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s57, s12, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s58, s12, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s59, s12, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s60, s13, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s61, s13, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s62, s13, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s63, s14, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s64, s14, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s65, s14, 0x80008
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s6, s15, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s7, s15, 0x80010
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s40, s7, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s41, s7, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s42, s7, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s43, s8, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s44, s8, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s45, s8, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s46, s9, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s47, s9, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s48, s9, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s49, s10, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s50, s10, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s51, s10, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s52, s11, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s53, s11, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s54, s11, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s55, s12, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s56, s12, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s57, s12, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s58, s13, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s59, s13, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s60, s13, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s61, s14, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s62, s14, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s63, s14, 0x80008
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s64, s15, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s65, s15, 0x80010
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s66, s15, 0x80008
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s6
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xf0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s7
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s15, s15
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xe0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s15
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s66
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    s_add_u32 s15, s16, 0xf0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s64
+; GFX8-NOHSA-NEXT:    s_addc_u32 s64, s17, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s14, s14
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s66
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s65
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s15
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s64
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xd0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s14
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s65
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s64
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s63
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s13, s13
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s14
+; GFX8-NOHSA-NEXT:    s_add_u32 s14, s16, 0xe0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s15, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s63
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s62
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s61
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s14
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xc0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s13
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s62
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s61
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s60
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s12, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s13
+; GFX8-NOHSA-NEXT:    s_add_u32 s13, s16, 0xd0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s14, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s60
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s59
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s58
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s13
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s14
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xb0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s12
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s59
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s58
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s57
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s11, s11
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s12
+; GFX8-NOHSA-NEXT:    s_add_u32 s12, s16, 0xc0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s13, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s57
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s56
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s55
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s13
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0xa0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s11
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s56
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s55
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s54
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s10, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s11
+; GFX8-NOHSA-NEXT:    s_add_u32 s11, s16, 0xb0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s12, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s54
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s53
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s52
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s11
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s12
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x90
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s10
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s53
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s52
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s51
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s9, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s16, 0xa0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s51
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s50
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s49
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x80
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s9
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s50
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s49
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s48
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s8, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s9
+; GFX8-NOHSA-NEXT:    s_add_u32 s9, s16, 0x90
+; GFX8-NOHSA-NEXT:    s_addc_u32 s10, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s48
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s47
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s46
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s10
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x70
+; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s7, s7
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s8
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s47
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s46
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s45
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    s_add_u32 s8, s16, 0x80
+; GFX8-NOHSA-NEXT:    s_addc_u32 s9, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s45
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s44
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s43
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s9
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x60
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s44
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s43
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s42
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s41
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
+; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s6, s6
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s7
+; GFX8-NOHSA-NEXT:    s_add_u32 s7, s16, 0x70
+; GFX8-NOHSA-NEXT:    s_addc_u32 s8, s17, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s42
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s41
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s40
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s8
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x50
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s5, s5
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s40
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NOHSA-NEXT:    s_add_u32 s6, s16, 0x60
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s39
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s38
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s37
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s17, 0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s4, s4
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s4, s4
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s5
+; GFX8-NOHSA-NEXT:    s_add_u32 s5, s16, 0x50
+; GFX8-NOHSA-NEXT:    s_addc_u32 s6, s17, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s36
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s35
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s34
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s3, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NOHSA-NEXT:    s_add_u32 s4, s16, 64
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s5, s17, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s5
-; GFX8-NOHSA-NEXT:    s_add_u32 s4, s16, 48
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s33
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s31
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s30
-; GFX8-NOHSA-NEXT:    s_addc_u32 s5, s17, 0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s2, s2
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s4
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s5
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s2, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s3
+; GFX8-NOHSA-NEXT:    s_add_u32 s3, s16, 48
+; GFX8-NOHSA-NEXT:    s_addc_u32 s4, s17, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s29
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s28
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s27
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s5
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s4
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s1, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s16, 32
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s17, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s16, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s26
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s25
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s24
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s17, 0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s0, s0
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    s_sext_i32_i8 s0, s0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s1
+; GFX8-NOHSA-NEXT:    s_add_u32 s1, s16, 16
+; GFX8-NOHSA-NEXT:    s_addc_u32 s2, s17, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s23
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s22
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s21
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s1
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s2
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -5838,20 +5860,20 @@ define amdgpu_kernel void @constant_zextload_v4i8_to_v4i64(ptr addrspace(1) %out
 ; GFX7-HSA-NEXT:    s_load_dword s2, s[2:3], 0x0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_bfe_u32 s4, s2, 0x80008
-; GFX7-HSA-NEXT:    s_lshr_b32 s3, s2, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s3, s2, 0x80008
+; GFX7-HSA-NEXT:    s_lshr_b32 s4, s2, 24
 ; GFX7-HSA-NEXT:    s_and_b32 s5, s2, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s2, s2, 0x80010
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s4
+; GFX7-HSA-NEXT:    s_addc_u32 s4, s1, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s4
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s4
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
@@ -5872,8 +5894,8 @@ define amdgpu_kernel void @constant_zextload_v4i8_to_v4i64(ptr addrspace(1) %out
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s5
@@ -6146,28 +6168,29 @@ define amdgpu_kernel void @constant_zextload_v8i8_to_v8i64(ptr addrspace(1) %out
 ; GFX7-HSA-NEXT:    s_bfe_u32 s7, s2, 0x80008
 ; GFX7-HSA-NEXT:    s_and_b32 s8, s2, 0xff
 ; GFX7-HSA-NEXT:    s_and_b32 s9, s3, 0xff
-; GFX7-HSA-NEXT:    s_bfe_u32 s10, s2, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_u32 s2, s3, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_u32 s2, s2, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_u32 s3, s3, 0x80010
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s3
+; GFX7-HSA-NEXT:    s_add_u32 s3, s0, 48
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
+; GFX7-HSA-NEXT:    s_addc_u32 s5, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s5
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    s_nop 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 48
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 32
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s10
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s4
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s9
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s6
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s8
@@ -6189,29 +6212,30 @@ define amdgpu_kernel void @constant_zextload_v8i8_to_v8i64(ptr addrspace(1) %out
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s6, s2, 24
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s7, s2, 0x80008
 ; GFX8-NOHSA-NEXT:    s_and_b32 s8, s2, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s9, s2, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s10, s3, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s2, s3, 0x80010
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 48
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 32
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s2, s2, 0x80010
+; GFX8-NOHSA-NEXT:    s_and_b32 s9, s3, 0xff
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s3, s3, 0x80010
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s3
+; GFX8-NOHSA-NEXT:    s_add_u32 s3, s0, 48
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s4
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s4, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX8-NOHSA-NEXT:    s_add_u32 s3, s0, 32
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s4
+; GFX8-NOHSA-NEXT:    s_addc_u32 s4, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s9
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s5
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s4
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s9
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s2
+; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
+; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s8
@@ -6389,28 +6413,28 @@ define amdgpu_kernel void @constant_sextload_v8i8_to_v8i64(ptr addrspace(1) %out
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s19
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s18
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s8
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s7
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s10
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s11
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s12
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s16
@@ -6448,28 +6472,28 @@ define amdgpu_kernel void @constant_sextload_v8i8_to_v8i64(ptr addrspace(1) %out
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 48
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 32
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s19
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s18
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s7
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s8
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s10
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s11
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s12
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s13
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s16
@@ -6647,75 +6671,77 @@ define amdgpu_kernel void @constant_zextload_v16i8_to_v16i64(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_lshr_b32 s8, s5, 24
-; GFX7-HSA-NEXT:    s_lshr_b32 s9, s4, 24
-; GFX7-HSA-NEXT:    s_lshr_b32 s10, s7, 24
-; GFX7-HSA-NEXT:    s_lshr_b32 s2, s6, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s11, s6, 0x80008
-; GFX7-HSA-NEXT:    s_bfe_u32 s12, s7, 0x80008
-; GFX7-HSA-NEXT:    s_bfe_u32 s13, s4, 0x80008
-; GFX7-HSA-NEXT:    s_bfe_u32 s14, s5, 0x80008
-; GFX7-HSA-NEXT:    s_and_b32 s15, s5, 0xff
-; GFX7-HSA-NEXT:    s_and_b32 s16, s4, 0xff
-; GFX7-HSA-NEXT:    s_and_b32 s17, s7, 0xff
-; GFX7-HSA-NEXT:    s_and_b32 s18, s6, 0xff
+; GFX7-HSA-NEXT:    s_lshr_b32 s2, s5, 24
+; GFX7-HSA-NEXT:    s_lshr_b32 s3, s4, 24
+; GFX7-HSA-NEXT:    s_lshr_b32 s8, s7, 24
+; GFX7-HSA-NEXT:    s_lshr_b32 s9, s6, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s10, s6, 0x80008
+; GFX7-HSA-NEXT:    s_bfe_u32 s11, s7, 0x80008
+; GFX7-HSA-NEXT:    s_bfe_u32 s12, s4, 0x80008
+; GFX7-HSA-NEXT:    s_bfe_u32 s13, s5, 0x80008
+; GFX7-HSA-NEXT:    s_and_b32 s14, s5, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s15, s4, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s16, s7, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s17, s6, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s5, s5, 0x80010
 ; GFX7-HSA-NEXT:    s_bfe_u32 s4, s4, 0x80010
 ; GFX7-HSA-NEXT:    s_bfe_u32 s7, s7, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_u32 s3, s6, 0x80010
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 0x50
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s3
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 0x70
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    s_bfe_u32 s6, s6, 0x80010
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s6
+; GFX7-HSA-NEXT:    s_add_u32 s6, s0, 0x50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s9
+; GFX7-HSA-NEXT:    s_addc_u32 s9, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s9
+; GFX7-HSA-NEXT:    s_add_u32 s6, s0, 0x70
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s10
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
+; GFX7-HSA-NEXT:    s_add_u32 s3, s0, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s9
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    s_addc_u32 s4, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s4
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s2
+; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 0x60
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s11
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s17
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s10
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s17
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s12
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s11
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s0, 32
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s13
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s15
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s12
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s15
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s14
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s14
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s13
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
 ;
@@ -6727,74 +6753,75 @@ define amdgpu_kernel void @constant_zextload_v16i8_to_v16i64(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s8, s5, 24
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s2, s7, 24
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s9, s6, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s10, s6, 0x80008
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s11, s7, 0x80008
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s12, s5, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s13, s4, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s14, s4, 0x80008
-; GFX8-NOHSA-NEXT:    s_and_b32 s15, s4, 0xff
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s2, s5, 24
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s3, s7, 24
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s8, s6, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s9, s6, 0x80008
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s10, s7, 0x80008
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s11, s5, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s12, s4, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s13, s4, 0x80008
+; GFX8-NOHSA-NEXT:    s_and_b32 s14, s4, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s4, s4, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s16, s5, 0xff
-; GFX8-NOHSA-NEXT:    s_and_b32 s17, s7, 0xff
-; GFX8-NOHSA-NEXT:    s_and_b32 s18, s6, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s15, s5, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s16, s7, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s17, s6, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s6, s6, 0x80010
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s5, s5, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s3, s7, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s7, s7, 0x80010
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s3
+; GFX8-NOHSA-NEXT:    s_add_u32 s3, s0, 0x70
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s7
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    s_nop 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 0x70
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s3
-; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 48
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 0x50
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s5
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 64
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 0x60
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s18
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s17
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s9
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 32
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s17
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s11
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s16
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s15
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s11
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s4
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s13
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s4
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s15
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s14
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s14
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s13
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_endpgm
@@ -7040,95 +7067,96 @@ define amdgpu_kernel void @constant_sextload_v16i8_to_v16i64(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_add_i32 s12, s12, s17
 ; GFX7-HSA-NEXT:    s_mov_b32 s9, 0
 ; GFX7-HSA-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
-; GFX7-HSA-NEXT:    s_mov_b32 s11, s9
+; GFX7-HSA-NEXT:    s_mov_b32 s15, s9
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
-; GFX7-HSA-NEXT:    s_mov_b32 s21, s9
-; GFX7-HSA-NEXT:    s_mov_b32 s23, s9
+; GFX7-HSA-NEXT:    s_mov_b32 s25, s9
+; GFX7-HSA-NEXT:    s_mov_b32 s11, s9
 ; GFX7-HSA-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_lshr_b32 s8, s7, 16
-; GFX7-HSA-NEXT:    s_lshr_b32 s10, s7, 8
-; GFX7-HSA-NEXT:    s_mov_b32 s12, s7
-; GFX7-HSA-NEXT:    s_lshr_b32 s14, s6, 16
-; GFX7-HSA-NEXT:    s_lshr_b32 s16, s6, 24
-; GFX7-HSA-NEXT:    s_ashr_i32 s29, s5, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s31, s5, 24
-; GFX7-HSA-NEXT:    s_lshr_b32 s18, s6, 8
-; GFX7-HSA-NEXT:    s_lshr_b32 s20, s5, 16
-; GFX7-HSA-NEXT:    s_lshr_b32 s22, s5, 8
-; GFX7-HSA-NEXT:    s_mov_b32 s24, s5
+; GFX7-HSA-NEXT:    s_lshr_b32 s14, s7, 8
+; GFX7-HSA-NEXT:    s_mov_b32 s16, s7
+; GFX7-HSA-NEXT:    s_lshr_b32 s18, s6, 16
+; GFX7-HSA-NEXT:    s_lshr_b32 s20, s6, 24
+; GFX7-HSA-NEXT:    s_ashr_i32 s27, s5, 31
+; GFX7-HSA-NEXT:    s_ashr_i32 s29, s5, 24
+; GFX7-HSA-NEXT:    s_ashr_i32 s31, s7, 31
+; GFX7-HSA-NEXT:    s_lshr_b32 s22, s6, 8
+; GFX7-HSA-NEXT:    s_lshr_b32 s24, s5, 16
+; GFX7-HSA-NEXT:    s_lshr_b32 s10, s5, 8
+; GFX7-HSA-NEXT:    s_mov_b32 s12, s5
 ; GFX7-HSA-NEXT:    s_lshr_b32 s26, s4, 16
 ; GFX7-HSA-NEXT:    s_lshr_b32 s28, s4, 24
 ; GFX7-HSA-NEXT:    s_lshr_b32 s30, s4, 8
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[2:3], s[4:5], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[4:5], s[8:9], 0x80000
-; GFX7-HSA-NEXT:    s_ashr_i32 s33, s7, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s36, s7, 24
+; GFX7-HSA-NEXT:    s_ashr_i32 s33, s7, 24
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[34:35], s[6:7], 0x80000
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[4:5], s[30:31], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[6:7], s[28:29], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[8:9], s[26:27], 0x80000
+; GFX7-HSA-NEXT:    s_bfe_i64 s[12:13], s[12:13], 0x80000
+; GFX7-HSA-NEXT:    s_bfe_i64 s[10:11], s[10:11], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[24:25], s[24:25], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[22:23], s[22:23], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[20:21], s[20:21], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[18:19], s[18:19], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[16:17], s[16:17], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[14:15], s[14:15], 0x80000
-; GFX7-HSA-NEXT:    s_bfe_i64 s[12:13], s[12:13], 0x80000
-; GFX7-HSA-NEXT:    s_bfe_i64 s[10:11], s[10:11], 0x80000
 ; GFX7-HSA-NEXT:    s_add_u32 s26, s0, 0x70
-; GFX7-HSA-NEXT:    s_addc_u32 s27, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s10
-; GFX7-HSA-NEXT:    s_add_u32 s10, s0, 0x60
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s11
-; GFX7-HSA-NEXT:    s_addc_u32 s11, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s10
-; GFX7-HSA-NEXT:    s_add_u32 s10, s0, 0x50
+; GFX7-HSA-NEXT:    s_addc_u32 s28, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s14
+; GFX7-HSA-NEXT:    s_add_u32 s14, s0, 0x60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s15
+; GFX7-HSA-NEXT:    s_addc_u32 s15, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s14
+; GFX7-HSA-NEXT:    s_add_u32 s14, s0, 0x50
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s26
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s27
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s33
-; GFX7-HSA-NEXT:    s_addc_u32 s11, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s28
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s33
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s31
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s15
+; GFX7-HSA-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s11
-; GFX7-HSA-NEXT:    s_add_u32 s10, s0, 64
-; GFX7-HSA-NEXT:    s_addc_u32 s11, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s11
-; GFX7-HSA-NEXT:    s_add_u32 s10, s0, 48
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s13
-; GFX7-HSA-NEXT:    s_addc_u32 s11, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s15
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s14
+; GFX7-HSA-NEXT:    s_add_u32 s14, s0, 64
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s15
+; GFX7-HSA-NEXT:    s_addc_u32 s15, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s14
+; GFX7-HSA-NEXT:    s_add_u32 s14, s0, 48
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s17
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s15
+; GFX7-HSA-NEXT:    s_addc_u32 s15, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s18
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s19
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s34
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s17
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s20
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s21
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s35
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s19
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s22
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s23
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[4:7]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[8:11]
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[12:15]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s11
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s24
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s25
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s29
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s27
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s14
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s15
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s10
 ; GFX7-HSA-NEXT:    s_add_u32 s10, s0, 32
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s20
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s21
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s31
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s29
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s11
 ; GFX7-HSA-NEXT:    s_addc_u32 s11, s1, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s12
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s13
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s10
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s24
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s25
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s22
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s23
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_nop 0
@@ -7193,28 +7221,28 @@ define amdgpu_kernel void @constant_sextload_v16i8_to_v16i64(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s7
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0x60
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s36
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s30
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0x50
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s24
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s25
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s22
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s23
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 64
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s26
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s27
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s28
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s29
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
@@ -7230,19 +7258,19 @@ define amdgpu_kernel void @constant_sextload_v16i8_to_v16i64(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX8-NOHSA-NEXT:    s_add_u32 s4, s0, 32
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s33
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s31
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s4
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX8-NOHSA-NEXT:    s_add_u32 s4, s0, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s14
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s15
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s17
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s4
@@ -7539,123 +7567,125 @@ define amdgpu_kernel void @constant_zextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_load_dwordx8 s[12:19], s[2:3], 0x0
+; GFX7-HSA-NEXT:    s_load_dwordx8 s[16:23], s[2:3], 0x0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-HSA-NEXT:    s_lshr_b32 s11, s12, 24
-; GFX7-HSA-NEXT:    s_lshr_b32 s20, s13, 24
-; GFX7-HSA-NEXT:    s_lshr_b32 s21, s14, 24
-; GFX7-HSA-NEXT:    s_lshr_b32 s22, s15, 24
-; GFX7-HSA-NEXT:    s_lshr_b32 s23, s16, 24
+; GFX7-HSA-NEXT:    s_lshr_b32 s15, s16, 24
 ; GFX7-HSA-NEXT:    s_lshr_b32 s24, s17, 24
 ; GFX7-HSA-NEXT:    s_lshr_b32 s25, s18, 24
 ; GFX7-HSA-NEXT:    s_lshr_b32 s26, s19, 24
-; GFX7-HSA-NEXT:    s_bfe_u32 s27, s19, 0x80008
-; GFX7-HSA-NEXT:    s_bfe_u32 s28, s18, 0x80008
-; GFX7-HSA-NEXT:    s_bfe_u32 s29, s17, 0x80008
-; GFX7-HSA-NEXT:    s_bfe_u32 s10, s16, 0x80008
-; GFX7-HSA-NEXT:    s_bfe_u32 s8, s15, 0x80008
-; GFX7-HSA-NEXT:    s_bfe_u32 s6, s14, 0x80008
-; GFX7-HSA-NEXT:    s_bfe_u32 s4, s13, 0x80008
-; GFX7-HSA-NEXT:    s_bfe_u32 s2, s12, 0x80008
-; GFX7-HSA-NEXT:    s_and_b32 s3, s12, 0xff
-; GFX7-HSA-NEXT:    s_and_b32 s5, s13, 0xff
-; GFX7-HSA-NEXT:    s_and_b32 s7, s14, 0xff
-; GFX7-HSA-NEXT:    s_and_b32 s9, s15, 0xff
-; GFX7-HSA-NEXT:    s_and_b32 s30, s16, 0xff
-; GFX7-HSA-NEXT:    s_and_b32 s31, s17, 0xff
-; GFX7-HSA-NEXT:    s_and_b32 s33, s18, 0xff
-; GFX7-HSA-NEXT:    s_and_b32 s34, s19, 0xff
-; GFX7-HSA-NEXT:    s_bfe_u32 s35, s12, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_u32 s36, s13, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_u32 s14, s14, 0x80010
-; GFX7-HSA-NEXT:    s_bfe_u32 s15, s15, 0x80010
+; GFX7-HSA-NEXT:    s_lshr_b32 s27, s20, 24
+; GFX7-HSA-NEXT:    s_lshr_b32 s28, s21, 24
+; GFX7-HSA-NEXT:    s_lshr_b32 s29, s22, 24
+; GFX7-HSA-NEXT:    s_lshr_b32 s30, s23, 24
+; GFX7-HSA-NEXT:    s_bfe_u32 s31, s23, 0x80008
+; GFX7-HSA-NEXT:    s_bfe_u32 s14, s22, 0x80008
+; GFX7-HSA-NEXT:    s_bfe_u32 s12, s21, 0x80008
+; GFX7-HSA-NEXT:    s_bfe_u32 s10, s20, 0x80008
+; GFX7-HSA-NEXT:    s_bfe_u32 s8, s19, 0x80008
+; GFX7-HSA-NEXT:    s_bfe_u32 s6, s18, 0x80008
+; GFX7-HSA-NEXT:    s_bfe_u32 s4, s17, 0x80008
+; GFX7-HSA-NEXT:    s_bfe_u32 s2, s16, 0x80008
+; GFX7-HSA-NEXT:    s_and_b32 s3, s16, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s5, s17, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s7, s18, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s9, s19, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s11, s20, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s13, s21, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s33, s22, 0xff
+; GFX7-HSA-NEXT:    s_and_b32 s34, s23, 0xff
 ; GFX7-HSA-NEXT:    s_bfe_u32 s16, s16, 0x80010
 ; GFX7-HSA-NEXT:    s_bfe_u32 s17, s17, 0x80010
 ; GFX7-HSA-NEXT:    s_bfe_u32 s18, s18, 0x80010
 ; GFX7-HSA-NEXT:    s_bfe_u32 s19, s19, 0x80010
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 0xf0
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s13
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 0xd0
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s13
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 0xb0
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s13
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 0x90
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s13
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 0x70
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s13
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 0x50
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s19
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s26
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s12
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 48
+; GFX7-HSA-NEXT:    s_bfe_u32 s20, s20, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_u32 s21, s21, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_u32 s22, s22, 0x80010
+; GFX7-HSA-NEXT:    s_bfe_u32 s23, s23, 0x80010
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0xf0
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0xd0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0xb0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0x90
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0x70
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s36
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s23
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s30
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s35
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 0x50
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s25
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s36
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s22
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s29
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s17
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s24
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s12
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s35
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s21
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s28
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 48
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s23
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s36
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s20
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s27
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s15
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s22
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s12
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s35
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s19
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s26
+; GFX7-HSA-NEXT:    s_add_u32 s35, s0, 16
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[0:3]
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 0xe0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s21
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s36
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s18
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s25
+; GFX7-HSA-NEXT:    s_addc_u32 s36, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[0:3]
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s20
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s35
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s17
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s24
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s35
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s13
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 0xc0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s36
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s15
+; GFX7-HSA-NEXT:    s_add_u32 s15, s0, 0xe0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s16
+; GFX7-HSA-NEXT:    s_addc_u32 s16, s1, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[0:3]
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s15
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s34
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s27
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s31
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s16
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s13
-; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 0xa0
+; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s14
+; GFX7-HSA-NEXT:    s_add_u32 s14, s0, 0xc0
+; GFX7-HSA-NEXT:    s_addc_u32 s15, s1, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s33
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s28
-; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s14
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX7-HSA-NEXT:    s_nop 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s12
+; GFX7-HSA-NEXT:    s_add_u32 s12, s0, 0xa0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s13
+; GFX7-HSA-NEXT:    s_addc_u32 s13, s1, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s31
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s29
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s13
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_nop 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s10
 ; GFX7-HSA-NEXT:    s_add_u32 s10, s0, 0x80
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s11
 ; GFX7-HSA-NEXT:    s_addc_u32 s11, s1, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s30
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s10
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -7696,136 +7726,141 @@ define amdgpu_kernel void @constant_zextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NOHSA-NEXT:    s_load_dwordx8 s[8:15], s[2:3], 0x0
+; GFX8-NOHSA-NEXT:    s_load_dwordx8 s[12:19], s[2:3], 0x0
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s16, s9, 24
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s17, s11, 24
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s18, s13, 24
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s6, s15, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s19, s15, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s20, s14, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s21, s14, 0x80008
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s22, s13, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s23, s12, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s24, s12, 0x80008
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s25, s11, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s26, s10, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s27, s10, 0x80008
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s28, s9, 0x80008
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s4, s8, 24
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s2, s8, 0x80008
-; GFX8-NOHSA-NEXT:    s_and_b32 s3, s8, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s5, s8, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s8, s9, 0xff
-; GFX8-NOHSA-NEXT:    s_and_b32 s29, s10, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s10, s10, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s30, s11, 0xff
-; GFX8-NOHSA-NEXT:    s_and_b32 s31, s12, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s12, s12, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s33, s13, 0xff
-; GFX8-NOHSA-NEXT:    s_and_b32 s34, s14, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s14, s14, 0x80010
-; GFX8-NOHSA-NEXT:    s_and_b32 s35, s15, 0xff
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s9, s9, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s11, s11, 0x80010
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s10, s13, 24
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s11, s15, 24
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s20, s17, 24
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s21, s19, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s22, s19, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s23, s18, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s24, s18, 0x80008
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s25, s17, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s26, s16, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s27, s16, 0x80008
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s28, s15, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s29, s14, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s8, s14, 0x80008
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s6, s13, 0x80008
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s4, s12, 24
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s2, s12, 0x80008
+; GFX8-NOHSA-NEXT:    s_and_b32 s3, s12, 0xff
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s5, s12, 0x80010
+; GFX8-NOHSA-NEXT:    s_and_b32 s7, s13, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s9, s14, 0xff
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s12, s14, 0x80010
+; GFX8-NOHSA-NEXT:    s_and_b32 s14, s15, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s30, s16, 0xff
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s16, s16, 0x80010
+; GFX8-NOHSA-NEXT:    s_and_b32 s31, s17, 0xff
+; GFX8-NOHSA-NEXT:    s_and_b32 s33, s18, 0xff
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s18, s18, 0x80010
+; GFX8-NOHSA-NEXT:    s_and_b32 s34, s19, 0xff
 ; GFX8-NOHSA-NEXT:    s_bfe_u32 s13, s13, 0x80010
-; GFX8-NOHSA-NEXT:    s_bfe_u32 s7, s15, 0x80010
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s6
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0xf0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s7
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0xb0
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0x70
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s13
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s18
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 48
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s11
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s17
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s15, s15, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s17, s17, 0x80010
+; GFX8-NOHSA-NEXT:    s_bfe_u32 s19, s19, 0x80010
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s19
+; GFX8-NOHSA-NEXT:    s_add_u32 s19, s0, 0xf0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s21
+; GFX8-NOHSA-NEXT:    s_addc_u32 s21, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s19
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s21
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0xe0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s9
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s16
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s17
+; GFX8-NOHSA-NEXT:    s_add_u32 s17, s0, 0xb0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s19, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s20
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s17
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s19
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0xd0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s35
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s19
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s11
+; GFX8-NOHSA-NEXT:    s_add_u32 s11, s0, 0x70
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s15
+; GFX8-NOHSA-NEXT:    s_addc_u32 s15, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s11
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0xc0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s14
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s20
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 48
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 0xe0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s13
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0xa0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 0xd0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s34
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s21
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
-; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0x90
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s33
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s22
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0x80
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 0xc0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s23
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0x60
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s31
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 0xa0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s33
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s24
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 0x50
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s30
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 0x90
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s31
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s25
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 64
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 0x80
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s26
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
-; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 32
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s29
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 0x60
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s30
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s27
-; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    s_add_u32 s10, s0, 0x50
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s14
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s28
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NOHSA-NEXT:    s_addc_u32 s11, s1, 0
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s10
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s29
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s8
+; GFX8-NOHSA-NEXT:    s_add_u32 s8, s0, 64
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s9
+; GFX8-NOHSA-NEXT:    s_addc_u32 s9, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s8
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s9
+; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s6
+; GFX8-NOHSA-NEXT:    s_add_u32 s6, s0, 32
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s7
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_nop 0
@@ -8282,25 +8317,26 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_load_dwordx8 s[0:7], s[10:11], 0x0
 ; GFX7-HSA-NEXT:    s_mov_b32 s11, 0
-; GFX7-HSA-NEXT:    s_mov_b32 s37, s11
+; GFX7-HSA-NEXT:    s_mov_b32 s39, s11
 ; GFX7-HSA-NEXT:    s_mov_b32 s57, s11
 ; GFX7-HSA-NEXT:    s_mov_b32 s47, s11
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_lshr_b32 s10, s7, 16
-; GFX7-HSA-NEXT:    s_lshr_b32 s36, s7, 8
-; GFX7-HSA-NEXT:    s_mov_b32 s38, s7
-; GFX7-HSA-NEXT:    s_lshr_b32 s40, s6, 16
-; GFX7-HSA-NEXT:    s_ashr_i32 s39, s1, 24
-; GFX7-HSA-NEXT:    s_ashr_i32 s41, s3, 31
+; GFX7-HSA-NEXT:    s_lshr_b32 s38, s7, 8
+; GFX7-HSA-NEXT:    s_mov_b32 s42, s7
+; GFX7-HSA-NEXT:    s_lshr_b32 s44, s6, 16
+; GFX7-HSA-NEXT:    s_ashr_i32 s43, s1, 24
+; GFX7-HSA-NEXT:    s_ashr_i32 s45, s3, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s50, s3, 24
-; GFX7-HSA-NEXT:    s_ashr_i32 s65, s5, 24
+; GFX7-HSA-NEXT:    s_ashr_i32 s63, s5, 24
+; GFX7-HSA-NEXT:    s_ashr_i32 s65, s7, 31
 ; GFX7-HSA-NEXT:    s_lshr_b32 s52, s6, 24
 ; GFX7-HSA-NEXT:    s_lshr_b32 s54, s6, 8
 ; GFX7-HSA-NEXT:    s_lshr_b32 s56, s5, 16
 ; GFX7-HSA-NEXT:    s_lshr_b32 s46, s5, 8
 ; GFX7-HSA-NEXT:    s_mov_b32 s48, s5
-; GFX7-HSA-NEXT:    s_lshr_b32 s44, s4, 16
-; GFX7-HSA-NEXT:    s_lshr_b32 s42, s4, 24
+; GFX7-HSA-NEXT:    s_lshr_b32 s40, s4, 16
+; GFX7-HSA-NEXT:    s_lshr_b32 s36, s4, 24
 ; GFX7-HSA-NEXT:    s_lshr_b32 s34, s4, 8
 ; GFX7-HSA-NEXT:    s_lshr_b32 s30, s3, 16
 ; GFX7-HSA-NEXT:    s_lshr_b32 s24, s3, 8
@@ -8322,8 +8358,7 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[2:3], s[10:11], 0x80000
 ; GFX7-HSA-NEXT:    s_ashr_i32 s33, s1, 31
 ; GFX7-HSA-NEXT:    s_ashr_i32 s51, s5, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s68, s7, 31
-; GFX7-HSA-NEXT:    s_ashr_i32 s69, s7, 24
+; GFX7-HSA-NEXT:    s_ashr_i32 s68, s7, 24
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[0:1], s[0:1], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[28:29], s[4:5], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[66:67], s[6:7], 0x80000
@@ -8342,108 +8377,108 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[24:25], s[24:25], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[30:31], s[30:31], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[34:35], s[34:35], 0x80000
-; GFX7-HSA-NEXT:    s_bfe_i64 s[42:43], s[42:43], 0x80000
-; GFX7-HSA-NEXT:    s_bfe_i64 s[44:45], s[44:45], 0x80000
+; GFX7-HSA-NEXT:    s_bfe_i64 s[36:37], s[36:37], 0x80000
+; GFX7-HSA-NEXT:    s_bfe_i64 s[40:41], s[40:41], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[48:49], s[48:49], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[46:47], s[46:47], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[56:57], s[56:57], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[54:55], s[54:55], 0x80000
 ; GFX7-HSA-NEXT:    s_bfe_i64 s[52:53], s[52:53], 0x80000
-; GFX7-HSA-NEXT:    s_bfe_i64 s[58:59], s[40:41], 0x80000
-; GFX7-HSA-NEXT:    s_bfe_i64 s[60:61], s[38:39], 0x80000
-; GFX7-HSA-NEXT:    s_bfe_i64 s[36:37], s[36:37], 0x80000
-; GFX7-HSA-NEXT:    s_add_u32 s62, s8, 0xf0
-; GFX7-HSA-NEXT:    s_addc_u32 s63, s9, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s36
-; GFX7-HSA-NEXT:    s_add_u32 s36, s8, 0xe0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s37
-; GFX7-HSA-NEXT:    s_addc_u32 s37, s9, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s37
-; GFX7-HSA-NEXT:    s_add_u32 s36, s8, 0xd0
-; GFX7-HSA-NEXT:    s_addc_u32 s37, s9, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s37
-; GFX7-HSA-NEXT:    s_add_u32 s36, s8, 0xc0
-; GFX7-HSA-NEXT:    s_addc_u32 s37, s9, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s37
-; GFX7-HSA-NEXT:    s_add_u32 s36, s8, 0xb0
-; GFX7-HSA-NEXT:    s_addc_u32 s37, s9, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s37
-; GFX7-HSA-NEXT:    s_add_u32 s36, s8, 0xa0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s58
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s59
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s52
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s53
-; GFX7-HSA-NEXT:    s_addc_u32 s37, s9, 0
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[8:11]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s66
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s28
+; GFX7-HSA-NEXT:    s_bfe_i64 s[58:59], s[44:45], 0x80000
+; GFX7-HSA-NEXT:    s_bfe_i64 s[60:61], s[42:43], 0x80000
+; GFX7-HSA-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x80000
+; GFX7-HSA-NEXT:    s_add_u32 s42, s8, 0xf0
+; GFX7-HSA-NEXT:    s_addc_u32 s44, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s38
+; GFX7-HSA-NEXT:    s_add_u32 s38, s8, 0xe0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s39
+; GFX7-HSA-NEXT:    s_addc_u32 s39, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s38
+; GFX7-HSA-NEXT:    s_add_u32 s38, s8, 0xd0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s39
+; GFX7-HSA-NEXT:    s_addc_u32 s39, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s38
+; GFX7-HSA-NEXT:    s_add_u32 s38, s8, 0xc0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s39
+; GFX7-HSA-NEXT:    s_addc_u32 s39, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s38
+; GFX7-HSA-NEXT:    s_add_u32 s38, s8, 0xb0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s39
+; GFX7-HSA-NEXT:    s_addc_u32 s39, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s38
+; GFX7-HSA-NEXT:    s_add_u32 s38, s8, 0xa0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s58
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s59
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s52
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s53
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s39
+; GFX7-HSA-NEXT:    s_addc_u32 s39, s9, 0
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[12:15]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s42
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s28
 ; GFX7-HSA-NEXT:    s_add_u32 s28, s8, 0x90
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s29
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s29
 ; GFX7-HSA-NEXT:    s_addc_u32 s29, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s44
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s68
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s65
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v22, s28
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s29
 ; GFX7-HSA-NEXT:    s_add_u32 s28, s8, 0x80
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s67
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s54
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s55
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s56
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s57
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[0:3]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s63
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s51
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v23, s29
 ; GFX7-HSA-NEXT:    s_addc_u32 s29, s9, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s62
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s63
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s69
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s68
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[12:15]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s66
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s67
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s54
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s55
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[4:7]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[16:19]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s26
+; GFX7-HSA-NEXT:    s_add_u32 s26, s8, 0x70
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v24, s28
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s29
-; GFX7-HSA-NEXT:    s_add_u32 s28, s8, 0x70
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s56
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s57
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s65
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s51
-; GFX7-HSA-NEXT:    s_addc_u32 s29, s9, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s60
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s61
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[16:19]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s48
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s24
+; GFX7-HSA-NEXT:    s_addc_u32 s28, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s61
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s24
 ; GFX7-HSA-NEXT:    s_add_u32 s24, s8, 0x60
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s48
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s49
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[4:7]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[8:11]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s46
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s47
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s36
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s37
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s25
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v20, s38
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v21, s39
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s25
 ; GFX7-HSA-NEXT:    s_addc_u32 s25, s9, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s44
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v8, s40
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s20
 ; GFX7-HSA-NEXT:    s_add_u32 s20, s8, 0x50
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s45
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v6, s42
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v7, s43
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v9, s41
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s36
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s37
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s27
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s24
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s25
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s21
 ; GFX7-HSA-NEXT:    s_addc_u32 s21, s9, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v10, s34
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v11, s35
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v12, s30
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v13, s31
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s26
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s50
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s41
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s27
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s28
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s29
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[4:7]
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[8:11]
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[12:15]
-; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[16:19]
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v25, s29
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v14, s34
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v15, s35
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v16, s30
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v17, s31
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v18, s50
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v19, s45
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v26, s26
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v27, s28
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[8:11]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[12:15]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[16:19]
+; GFX7-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s23
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s20
@@ -8455,16 +8490,16 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s15
 ; GFX7-HSA-NEXT:    s_addc_u32 s15, s9, 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX7-HSA-NEXT:    s_add_u32 s14, s8, 48
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s18
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s19
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX7-HSA-NEXT:    s_addc_u32 s15, s9, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s14
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s17
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s39
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s43
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s33
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s15
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -8500,22 +8535,22 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX8-NOHSA-LABEL: constant_sextload_v32i8_to_v32i64:
 ; GFX8-NOHSA:       ; %bb.0:
 ; GFX8-NOHSA-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX8-NOHSA-NEXT:    s_mov_b32 s47, 0
-; GFX8-NOHSA-NEXT:    s_mov_b32 s53, s47
-; GFX8-NOHSA-NEXT:    s_mov_b32 s65, s47
-; GFX8-NOHSA-NEXT:    s_mov_b32 s67, s47
+; GFX8-NOHSA-NEXT:    s_mov_b32 s51, 0
+; GFX8-NOHSA-NEXT:    s_mov_b32 s53, s51
+; GFX8-NOHSA-NEXT:    s_mov_b32 s65, s51
+; GFX8-NOHSA-NEXT:    s_mov_b32 s67, s51
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    s_load_dwordx8 s[0:7], s[10:11], 0x0
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s46, s7, 16
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s50, s7, 16
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s54, s7, 8
 ; GFX8-NOHSA-NEXT:    s_mov_b32 s56, s7
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s58, s6, 16
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s60, s6, 24
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s62, s6, 8
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s52, s5, 16
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s50, s5, 8
-; GFX8-NOHSA-NEXT:    s_mov_b32 s48, s5
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s46, s5, 8
+; GFX8-NOHSA-NEXT:    s_mov_b32 s44, s5
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s42, s4, 16
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s40, s4, 24
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s38, s4, 8
@@ -8534,7 +8569,7 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[10:11], s[0:1], 0x80000
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[22:23], s[2:3], 0x80000
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[36:37], s[4:5], 0x80000
-; GFX8-NOHSA-NEXT:    s_bfe_i64 s[44:45], s[6:7], 0x80000
+; GFX8-NOHSA-NEXT:    s_bfe_i64 s[48:49], s[6:7], 0x80000
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[12:13], s[12:13], 0x80000
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[14:15], s[14:15], 0x80000
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[16:17], s[16:17], 0x80000
@@ -8554,8 +8589,8 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x80000
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[40:41], s[40:41], 0x80000
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[42:43], s[42:43], 0x80000
-; GFX8-NOHSA-NEXT:    s_bfe_i64 s[48:49], s[48:49], 0x80000
-; GFX8-NOHSA-NEXT:    s_bfe_i64 s[50:51], s[50:51], 0x80000
+; GFX8-NOHSA-NEXT:    s_bfe_i64 s[44:45], s[44:45], 0x80000
+; GFX8-NOHSA-NEXT:    s_bfe_i64 s[46:47], s[46:47], 0x80000
 ; GFX8-NOHSA-NEXT:    s_ashr_i32 s64, s5, 31
 ; GFX8-NOHSA-NEXT:    s_ashr_i32 s5, s5, 24
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[52:53], s[52:53], 0x80000
@@ -8566,146 +8601,144 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_bfe_i64 s[54:55], s[54:55], 0x80000
 ; GFX8-NOHSA-NEXT:    s_ashr_i32 s65, s7, 31
 ; GFX8-NOHSA-NEXT:    s_ashr_i32 s7, s7, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i64 s[46:47], s[46:47], 0x80000
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s46
-; GFX8-NOHSA-NEXT:    s_add_u32 s46, s8, 0xf0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s47
-; GFX8-NOHSA-NEXT:    s_addc_u32 s47, s9, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s46
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s47
-; GFX8-NOHSA-NEXT:    s_add_u32 s46, s8, 0xe0
+; GFX8-NOHSA-NEXT:    s_bfe_i64 s[50:51], s[50:51], 0x80000
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s7
+; GFX8-NOHSA-NEXT:    s_add_u32 s7, s8, 0xf0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s50
+; GFX8-NOHSA-NEXT:    s_addc_u32 s50, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX8-NOHSA-NEXT:    s_add_u32 s7, s8, 0xe0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s51
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s65
-; GFX8-NOHSA-NEXT:    s_addc_u32 s47, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s50
+; GFX8-NOHSA-NEXT:    s_addc_u32 s50, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s46
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s47
-; GFX8-NOHSA-NEXT:    s_add_u32 s46, s8, 0xd0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX8-NOHSA-NEXT:    s_add_u32 s7, s8, 0xd0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s56
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s57
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s54
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s55
-; GFX8-NOHSA-NEXT:    s_addc_u32 s47, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s50
+; GFX8-NOHSA-NEXT:    s_addc_u32 s50, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s46
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s58
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s59
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s60
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s61
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s47
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s50
+; GFX8-NOHSA-NEXT:    s_add_u32 s7, s8, 0xc0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    s_nop 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s44
-; GFX8-NOHSA-NEXT:    s_add_u32 s44, s8, 0xc0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s45
-; GFX8-NOHSA-NEXT:    s_addc_u32 s45, s9, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s44
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s45
-; GFX8-NOHSA-NEXT:    s_add_u32 s44, s8, 0xb0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s7
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s48
+; GFX8-NOHSA-NEXT:    s_addc_u32 s48, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s49
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s62
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s63
-; GFX8-NOHSA-NEXT:    s_addc_u32 s45, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s48
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s44
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s45
-; GFX8-NOHSA-NEXT:    s_add_u32 s44, s8, 0xa0
+; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s5
+; GFX8-NOHSA-NEXT:    s_add_u32 s5, s8, 0xb0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NOHSA-NEXT:    s_add_u32 s5, s8, 0xa0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s52
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s53
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s64
-; GFX8-NOHSA-NEXT:    s_addc_u32 s45, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s44
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s48
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s49
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s50
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s51
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s45
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NOHSA-NEXT:    s_add_u32 s5, s8, 0x90
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s44
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s45
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s46
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s47
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    s_nop 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s40
-; GFX8-NOHSA-NEXT:    s_add_u32 s40, s8, 0x90
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s41
-; GFX8-NOHSA-NEXT:    s_addc_u32 s41, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s5
+; GFX8-NOHSA-NEXT:    s_add_u32 s5, s8, 0x80
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s42
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s43
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s40
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s41
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s40
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s41
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    s_nop 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s5
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s36
-; GFX8-NOHSA-NEXT:    s_add_u32 s36, s8, 0x80
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s37
-; GFX8-NOHSA-NEXT:    s_addc_u32 s37, s9, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s38
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s39
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s36
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s37
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_nop 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s8, 0x70
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s9, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s8, 0x60
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s66
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s33
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s8, 0x50
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s30
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s31
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s34
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s35
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    s_add_u32 s2, s8, 64
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s28
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s29
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s26
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s27
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s23
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s24
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s25
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_nop 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX8-NOHSA-NEXT:    s_add_u32 s0, s8, 48
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s1, s9, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    s_add_u32 s0, s8, 32
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s6
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s4
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s1, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    s_add_u32 s0, s8, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s19
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s20
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s21
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s1, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s17
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s14
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s15
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s8
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s10
@@ -10500,8 +10533,8 @@ define amdgpu_kernel void @constant_zextload_v16i8_to_v16i16(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s9
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s11
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s14
@@ -10556,8 +10589,8 @@ define amdgpu_kernel void @constant_zextload_v16i8_to_v16i16(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s13
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s12
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s5
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s11
@@ -10828,58 +10861,58 @@ define amdgpu_kernel void @constant_sextload_v16i8_to_v16i16(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-HSA-NEXT:    s_ashr_i32 s2, s5, 24
 ; GFX7-HSA-NEXT:    s_bfe_i32 s3, s5, 0x80010
-; GFX7-HSA-NEXT:    s_lshl_b32 s2, s2, 16
-; GFX7-HSA-NEXT:    s_and_b32 s3, s3, 0xffff
 ; GFX7-HSA-NEXT:    s_bfe_i32 s8, s5, 0x80008
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s5, s5
-; GFX7-HSA-NEXT:    s_ashr_i32 s9, s4, 24
-; GFX7-HSA-NEXT:    s_or_b32 s10, s3, s2
-; GFX7-HSA-NEXT:    s_bfe_i32 s3, s4, 0x80010
-; GFX7-HSA-NEXT:    s_lshl_b32 s8, s8, 16
-; GFX7-HSA-NEXT:    s_and_b32 s5, s5, 0xffff
-; GFX7-HSA-NEXT:    s_lshl_b32 s2, s9, 16
-; GFX7-HSA-NEXT:    s_and_b32 s3, s3, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s5, s5, s8
-; GFX7-HSA-NEXT:    s_or_b32 s8, s3, s2
-; GFX7-HSA-NEXT:    s_bfe_i32 s2, s4, 0x80008
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s3, s4
-; GFX7-HSA-NEXT:    s_lshl_b32 s2, s2, 16
-; GFX7-HSA-NEXT:    s_and_b32 s3, s3, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s4, s3, s2
-; GFX7-HSA-NEXT:    s_ashr_i32 s2, s7, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s3, s7, 0x80010
 ; GFX7-HSA-NEXT:    s_lshl_b32 s2, s2, 16
 ; GFX7-HSA-NEXT:    s_and_b32 s3, s3, 0xffff
+; GFX7-HSA-NEXT:    s_lshl_b32 s8, s8, 16
+; GFX7-HSA-NEXT:    s_and_b32 s5, s5, 0xffff
+; GFX7-HSA-NEXT:    s_ashr_i32 s9, s4, 24
 ; GFX7-HSA-NEXT:    s_or_b32 s2, s3, s2
-; GFX7-HSA-NEXT:    s_bfe_i32 s3, s7, 0x80008
+; GFX7-HSA-NEXT:    s_or_b32 s3, s5, s8
+; GFX7-HSA-NEXT:    s_bfe_i32 s8, s4, 0x80010
+; GFX7-HSA-NEXT:    s_lshl_b32 s5, s9, 16
+; GFX7-HSA-NEXT:    s_and_b32 s8, s8, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s5, s8, s5
+; GFX7-HSA-NEXT:    s_bfe_i32 s8, s4, 0x80008
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s4, s4
+; GFX7-HSA-NEXT:    s_lshl_b32 s8, s8, 16
+; GFX7-HSA-NEXT:    s_and_b32 s4, s4, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s4, s4, s8
+; GFX7-HSA-NEXT:    s_ashr_i32 s8, s7, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s9, s7, 0x80010
+; GFX7-HSA-NEXT:    s_lshl_b32 s8, s8, 16
+; GFX7-HSA-NEXT:    s_and_b32 s9, s9, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s8, s9, s8
+; GFX7-HSA-NEXT:    s_bfe_i32 s9, s7, 0x80008
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s7, s7
-; GFX7-HSA-NEXT:    s_lshl_b32 s3, s3, 16
+; GFX7-HSA-NEXT:    s_lshl_b32 s9, s9, 16
 ; GFX7-HSA-NEXT:    s_and_b32 s7, s7, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s3, s7, s3
-; GFX7-HSA-NEXT:    s_ashr_i32 s7, s6, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s9, s6, 0x80010
-; GFX7-HSA-NEXT:    s_lshl_b32 s7, s7, 16
-; GFX7-HSA-NEXT:    s_and_b32 s9, s9, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s7, s9, s7
-; GFX7-HSA-NEXT:    s_bfe_i32 s9, s6, 0x80008
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s6, s6
+; GFX7-HSA-NEXT:    s_or_b32 s7, s7, s9
+; GFX7-HSA-NEXT:    s_ashr_i32 s9, s6, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s10, s6, 0x80010
 ; GFX7-HSA-NEXT:    s_lshl_b32 s9, s9, 16
+; GFX7-HSA-NEXT:    s_and_b32 s10, s10, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s9, s10, s9
+; GFX7-HSA-NEXT:    s_bfe_i32 s10, s6, 0x80008
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s6, s6
+; GFX7-HSA-NEXT:    s_lshl_b32 s10, s10, 16
 ; GFX7-HSA-NEXT:    s_and_b32 s6, s6, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s6, s6, s9
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s2
-; GFX7-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
-; GFX7-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GFX7-HSA-NEXT:    s_or_b32 s6, s6, s10
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s7
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    s_add_u32 s6, s0, 16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s7
+; GFX7-HSA-NEXT:    s_addc_u32 s7, s1, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s9
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s8
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s8
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s10
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s5
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    s_endpgm
@@ -10945,8 +10978,8 @@ define amdgpu_kernel void @constant_sextload_v16i8_to_v16i16(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s9
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s8
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s11
@@ -11353,29 +11386,29 @@ define amdgpu_kernel void @constant_zextload_v32i8_to_v32i16(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s8, 48
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s9, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s8, 32
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s11
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s14
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s9, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    s_add_u32 s0, s8, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s17
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s10
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    s_addc_u32 s1, s9, 0
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s19
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s20
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s21
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s8
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s23
@@ -11459,29 +11492,29 @@ define amdgpu_kernel void @constant_zextload_v32i8_to_v32i16(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX8-NOHSA-NEXT:    s_add_u32 s0, s8, 48
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s1, s9, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    s_add_u32 s0, s8, 32
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s21
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s11
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s5
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s1, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    s_add_u32 s0, s8, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s20
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s12
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s3
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    s_addc_u32 s1, s9, 0
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s19
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s18
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s17
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s8
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s16
@@ -11963,113 +11996,115 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i16(ptr addrspace(1) %o
 ; GFX7-HSA-NEXT:    s_bfe_i32 s11, s1, 0x80010
 ; GFX7-HSA-NEXT:    s_bfe_i32 s12, s1, 0x80008
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s1, s1
-; GFX7-HSA-NEXT:    s_lshl_b32 s10, s10, 16
-; GFX7-HSA-NEXT:    s_and_b32 s11, s11, 0xffff
 ; GFX7-HSA-NEXT:    s_lshl_b32 s12, s12, 16
 ; GFX7-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
 ; GFX7-HSA-NEXT:    s_ashr_i32 s13, s0, 24
-; GFX7-HSA-NEXT:    s_or_b32 s10, s11, s10
-; GFX7-HSA-NEXT:    s_or_b32 s11, s1, s12
+; GFX7-HSA-NEXT:    s_lshl_b32 s10, s10, 16
+; GFX7-HSA-NEXT:    s_and_b32 s11, s11, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s1, s1, s12
 ; GFX7-HSA-NEXT:    s_bfe_i32 s12, s0, 0x80010
-; GFX7-HSA-NEXT:    s_lshl_b32 s1, s13, 16
+; GFX7-HSA-NEXT:    s_or_b32 s10, s11, s10
+; GFX7-HSA-NEXT:    s_lshl_b32 s11, s13, 16
 ; GFX7-HSA-NEXT:    s_and_b32 s12, s12, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s12, s12, s1
-; GFX7-HSA-NEXT:    s_bfe_i32 s1, s0, 0x80008
+; GFX7-HSA-NEXT:    s_or_b32 s11, s12, s11
+; GFX7-HSA-NEXT:    s_bfe_i32 s12, s0, 0x80008
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s0, s0
-; GFX7-HSA-NEXT:    s_lshl_b32 s1, s1, 16
+; GFX7-HSA-NEXT:    s_lshl_b32 s12, s12, 16
 ; GFX7-HSA-NEXT:    s_and_b32 s0, s0, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s13, s0, s1
-; GFX7-HSA-NEXT:    s_ashr_i32 s0, s3, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s1, s3, 0x80010
-; GFX7-HSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX7-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s14, s1, s0
-; GFX7-HSA-NEXT:    s_bfe_i32 s0, s3, 0x80008
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s1, s3
-; GFX7-HSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX7-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s3, s1, s0
-; GFX7-HSA-NEXT:    s_ashr_i32 s0, s2, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s1, s2, 0x80010
-; GFX7-HSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX7-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s15, s1, s0
-; GFX7-HSA-NEXT:    s_bfe_i32 s0, s2, 0x80008
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s1, s2
-; GFX7-HSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX7-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s2, s1, s0
-; GFX7-HSA-NEXT:    s_ashr_i32 s0, s5, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s1, s5, 0x80010
-; GFX7-HSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX7-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s16, s1, s0
-; GFX7-HSA-NEXT:    s_bfe_i32 s0, s5, 0x80008
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s1, s5
-; GFX7-HSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX7-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s5, s1, s0
-; GFX7-HSA-NEXT:    s_ashr_i32 s0, s4, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s1, s4, 0x80010
-; GFX7-HSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX7-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s17, s1, s0
-; GFX7-HSA-NEXT:    s_bfe_i32 s0, s4, 0x80008
-; GFX7-HSA-NEXT:    s_sext_i32_i8 s1, s4
-; GFX7-HSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX7-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s4, s1, s0
-; GFX7-HSA-NEXT:    s_ashr_i32 s0, s7, 24
-; GFX7-HSA-NEXT:    s_bfe_i32 s1, s7, 0x80010
-; GFX7-HSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX7-HSA-NEXT:    s_and_b32 s1, s1, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s0, s1, s0
-; GFX7-HSA-NEXT:    s_bfe_i32 s1, s7, 0x80008
+; GFX7-HSA-NEXT:    s_or_b32 s0, s0, s12
+; GFX7-HSA-NEXT:    s_ashr_i32 s12, s3, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s13, s3, 0x80010
+; GFX7-HSA-NEXT:    s_lshl_b32 s12, s12, 16
+; GFX7-HSA-NEXT:    s_and_b32 s13, s13, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s12, s13, s12
+; GFX7-HSA-NEXT:    s_bfe_i32 s13, s3, 0x80008
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s3, s3
+; GFX7-HSA-NEXT:    s_lshl_b32 s13, s13, 16
+; GFX7-HSA-NEXT:    s_and_b32 s3, s3, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s3, s3, s13
+; GFX7-HSA-NEXT:    s_ashr_i32 s13, s2, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s14, s2, 0x80010
+; GFX7-HSA-NEXT:    s_lshl_b32 s13, s13, 16
+; GFX7-HSA-NEXT:    s_and_b32 s14, s14, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s13, s14, s13
+; GFX7-HSA-NEXT:    s_bfe_i32 s14, s2, 0x80008
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s2, s2
+; GFX7-HSA-NEXT:    s_lshl_b32 s14, s14, 16
+; GFX7-HSA-NEXT:    s_and_b32 s2, s2, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s2, s2, s14
+; GFX7-HSA-NEXT:    s_ashr_i32 s14, s5, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s15, s5, 0x80010
+; GFX7-HSA-NEXT:    s_lshl_b32 s14, s14, 16
+; GFX7-HSA-NEXT:    s_and_b32 s15, s15, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s14, s15, s14
+; GFX7-HSA-NEXT:    s_bfe_i32 s15, s5, 0x80008
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s5, s5
+; GFX7-HSA-NEXT:    s_lshl_b32 s15, s15, 16
+; GFX7-HSA-NEXT:    s_and_b32 s5, s5, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s5, s5, s15
+; GFX7-HSA-NEXT:    s_ashr_i32 s15, s4, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s16, s4, 0x80010
+; GFX7-HSA-NEXT:    s_lshl_b32 s15, s15, 16
+; GFX7-HSA-NEXT:    s_and_b32 s16, s16, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s15, s16, s15
+; GFX7-HSA-NEXT:    s_bfe_i32 s16, s4, 0x80008
+; GFX7-HSA-NEXT:    s_sext_i32_i8 s4, s4
+; GFX7-HSA-NEXT:    s_lshl_b32 s16, s16, 16
+; GFX7-HSA-NEXT:    s_and_b32 s4, s4, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s4, s4, s16
+; GFX7-HSA-NEXT:    s_ashr_i32 s16, s7, 24
+; GFX7-HSA-NEXT:    s_bfe_i32 s17, s7, 0x80010
+; GFX7-HSA-NEXT:    s_lshl_b32 s16, s16, 16
+; GFX7-HSA-NEXT:    s_and_b32 s17, s17, 0xffff
+; GFX7-HSA-NEXT:    s_or_b32 s16, s17, s16
+; GFX7-HSA-NEXT:    s_bfe_i32 s17, s7, 0x80008
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s7, s7
-; GFX7-HSA-NEXT:    s_lshl_b32 s1, s1, 16
+; GFX7-HSA-NEXT:    s_lshl_b32 s17, s17, 16
 ; GFX7-HSA-NEXT:    s_and_b32 s7, s7, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s1, s7, s1
-; GFX7-HSA-NEXT:    s_ashr_i32 s7, s6, 24
+; GFX7-HSA-NEXT:    s_or_b32 s7, s7, s17
+; GFX7-HSA-NEXT:    s_ashr_i32 s17, s6, 24
 ; GFX7-HSA-NEXT:    s_bfe_i32 s18, s6, 0x80010
-; GFX7-HSA-NEXT:    s_lshl_b32 s7, s7, 16
+; GFX7-HSA-NEXT:    s_lshl_b32 s17, s17, 16
 ; GFX7-HSA-NEXT:    s_and_b32 s18, s18, 0xffff
-; GFX7-HSA-NEXT:    s_or_b32 s7, s18, s7
+; GFX7-HSA-NEXT:    s_or_b32 s17, s18, s17
 ; GFX7-HSA-NEXT:    s_bfe_i32 s18, s6, 0x80008
 ; GFX7-HSA-NEXT:    s_sext_i32_i8 s6, s6
 ; GFX7-HSA-NEXT:    s_lshl_b32 s18, s18, 16
 ; GFX7-HSA-NEXT:    s_and_b32 s6, s6, 0xffff
 ; GFX7-HSA-NEXT:    s_or_b32 s6, s6, s18
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s0
-; GFX7-HSA-NEXT:    s_add_u32 s0, s8, 48
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s1
-; GFX7-HSA-NEXT:    s_addc_u32 s1, s9, 0
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
-; GFX7-HSA-NEXT:    s_add_u32 s0, s8, 32
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s7
-; GFX7-HSA-NEXT:    s_addc_u32 s1, s9, 0
+; GFX7-HSA-NEXT:    s_add_u32 s6, s8, 48
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s7
+; GFX7-HSA-NEXT:    s_addc_u32 s7, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s17
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s16
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
-; GFX7-HSA-NEXT:    s_add_u32 s0, s8, 16
+; GFX7-HSA-NEXT:    s_nop 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s17
+; GFX7-HSA-NEXT:    s_add_u32 s4, s8, 32
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s5
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s16
-; GFX7-HSA-NEXT:    s_addc_u32 s1, s9, 0
+; GFX7-HSA-NEXT:    s_addc_u32 s5, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s15
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s14
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s4
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s1
+; GFX7-HSA-NEXT:    s_nop 0
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s15
+; GFX7-HSA-NEXT:    s_add_u32 s2, s8, 16
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s3
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s14
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GFX7-HSA-NEXT:    s_addc_u32 s3, s9, 0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s13
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s12
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v4, s8
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s13
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s12
-; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s11
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v0, s0
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v1, s11
+; GFX7-HSA-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v3, s10
 ; GFX7-HSA-NEXT:    v_mov_b32_e32 v5, s9
 ; GFX7-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -12081,137 +12116,139 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i16(ptr addrspace(1) %o
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NOHSA-NEXT:    s_load_dwordx8 s[0:7], s[10:11], 0x0
 ; GFX8-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s17, s1, 16
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s18, s1
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s19, s1, 0x80000
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s1, s1, 16
 ; GFX8-NOHSA-NEXT:    s_lshl_b32 s18, s18, 8
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s1, s1, 8
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s17, s17, 0x80000
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s19, s1, 0x80000
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s16, s0, 16
 ; GFX8-NOHSA-NEXT:    s_and_b32 s18, s18, 0xffff0000
 ; GFX8-NOHSA-NEXT:    s_and_b32 s19, 0xffff, s19
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, s1, 0xffff0000
-; GFX8-NOHSA-NEXT:    s_and_b32 s17, 0xffff, s17
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s17, s1, 16
 ; GFX8-NOHSA-NEXT:    s_or_b32 s18, s19, s18
 ; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s19, s0
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s20, s0, 0x80000
-; GFX8-NOHSA-NEXT:    s_or_b32 s17, s17, s1
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s1, s1, 16
 ; GFX8-NOHSA-NEXT:    s_ashr_i32 s0, s0, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s1, s16, 0x80000
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s16, s16, 0x80000
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s1, s1, 8
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s17, s17, 0x80000
 ; GFX8-NOHSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, 0xffff, s1
-; GFX8-NOHSA-NEXT:    s_or_b32 s16, s1, s0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s0, s3
+; GFX8-NOHSA-NEXT:    s_and_b32 s16, 0xffff, s16
+; GFX8-NOHSA-NEXT:    s_and_b32 s1, s1, 0xffff0000
+; GFX8-NOHSA-NEXT:    s_and_b32 s17, 0xffff, s17
+; GFX8-NOHSA-NEXT:    s_or_b32 s0, s16, s0
+; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s16, s3
 ; GFX8-NOHSA-NEXT:    s_lshl_b32 s19, s19, 8
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s0, s0, 8
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s1, s3, 0x80000
+; GFX8-NOHSA-NEXT:    s_or_b32 s1, s17, s1
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s16, s16, 8
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s17, s3, 0x80000
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s14, s2, 16
 ; GFX8-NOHSA-NEXT:    s_and_b32 s19, s19, 0xffff0000
 ; GFX8-NOHSA-NEXT:    s_and_b32 s20, 0xffff, s20
-; GFX8-NOHSA-NEXT:    s_and_b32 s0, s0, 0xffff0000
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, 0xffff, s1
-; GFX8-NOHSA-NEXT:    s_or_b32 s19, s20, s19
-; GFX8-NOHSA-NEXT:    s_or_b32 s20, s1, s0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s0, s2
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s0, s0, 8
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s1, s2, 0x80000
-; GFX8-NOHSA-NEXT:    s_and_b32 s0, s0, 0xffff0000
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NOHSA-NEXT:    s_and_b32 s16, s16, 0xffff0000
+; GFX8-NOHSA-NEXT:    s_and_b32 s17, 0xffff, s17
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s15, s3, 16
-; GFX8-NOHSA-NEXT:    s_or_b32 s21, s1, s0
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s0, s3, 16
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s0, s0, 8
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s1, s15, 0x80000
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s14, s2, 16
-; GFX8-NOHSA-NEXT:    s_and_b32 s0, s0, 0xffff0000
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, 0xffff, s1
-; GFX8-NOHSA-NEXT:    s_or_b32 s3, s1, s0
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s0, s2, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s1, s14, 0x80000
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, 0xffff, s1
-; GFX8-NOHSA-NEXT:    s_or_b32 s2, s1, s0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s0, s5
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s0, s0, 8
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s1, s5, 0x80000
-; GFX8-NOHSA-NEXT:    s_and_b32 s0, s0, 0xffff0000
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, 0xffff, s1
-; GFX8-NOHSA-NEXT:    s_or_b32 s14, s1, s0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s0, s4
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s0, s0, 8
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s1, s4, 0x80000
-; GFX8-NOHSA-NEXT:    s_lshr_b32 s13, s5, 16
-; GFX8-NOHSA-NEXT:    s_and_b32 s0, s0, 0xffff0000
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, 0xffff, s1
-; GFX8-NOHSA-NEXT:    s_or_b32 s15, s1, s0
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s0, s5, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s1, s13, 0x80000
+; GFX8-NOHSA-NEXT:    s_or_b32 s19, s20, s19
+; GFX8-NOHSA-NEXT:    s_or_b32 s16, s17, s16
+; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s17, s2
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s20, s2, 0x80000
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s3, s3, 16
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s2, s2, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s14, s14, 0x80000
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s3, s3, 8
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s15, s15, 0x80000
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s2, s2, 16
+; GFX8-NOHSA-NEXT:    s_and_b32 s14, 0xffff, s14
+; GFX8-NOHSA-NEXT:    s_and_b32 s3, s3, 0xffff0000
+; GFX8-NOHSA-NEXT:    s_and_b32 s15, 0xffff, s15
+; GFX8-NOHSA-NEXT:    s_or_b32 s2, s14, s2
+; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s14, s5
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s17, s17, 8
+; GFX8-NOHSA-NEXT:    s_or_b32 s3, s15, s3
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s14, s14, 8
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s15, s5, 0x80000
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s12, s4, 16
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, 0xffff, s1
-; GFX8-NOHSA-NEXT:    s_or_b32 s5, s1, s0
-; GFX8-NOHSA-NEXT:    s_ashr_i32 s0, s4, 24
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s1, s12, 0x80000
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, 0xffff, s1
-; GFX8-NOHSA-NEXT:    s_or_b32 s4, s1, s0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s0, s7
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s0, s0, 8
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s1, s7, 0x80000
-; GFX8-NOHSA-NEXT:    s_and_b32 s0, s0, 0xffff0000
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NOHSA-NEXT:    s_and_b32 s17, s17, 0xffff0000
+; GFX8-NOHSA-NEXT:    s_and_b32 s20, 0xffff, s20
+; GFX8-NOHSA-NEXT:    s_and_b32 s14, s14, 0xffff0000
+; GFX8-NOHSA-NEXT:    s_and_b32 s15, 0xffff, s15
+; GFX8-NOHSA-NEXT:    s_lshr_b32 s13, s5, 16
+; GFX8-NOHSA-NEXT:    s_or_b32 s17, s20, s17
+; GFX8-NOHSA-NEXT:    s_or_b32 s14, s15, s14
+; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s15, s4
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s20, s4, 0x80000
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s4, s4, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s12, s12, 0x80000
+; GFX8-NOHSA-NEXT:    s_ashr_i32 s5, s5, 24
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s13, s13, 0x80000
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s4, s4, 16
+; GFX8-NOHSA-NEXT:    s_and_b32 s12, 0xffff, s12
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s5, s5, 16
+; GFX8-NOHSA-NEXT:    s_and_b32 s13, 0xffff, s13
+; GFX8-NOHSA-NEXT:    s_or_b32 s4, s12, s4
+; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s12, s7
+; GFX8-NOHSA-NEXT:    s_or_b32 s5, s13, s5
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s12, s12, 8
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s13, s7, 0x80000
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s15, s15, 8
+; GFX8-NOHSA-NEXT:    s_and_b32 s12, s12, 0xffff0000
+; GFX8-NOHSA-NEXT:    s_and_b32 s13, 0xffff, s13
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s10, s6, 16
 ; GFX8-NOHSA-NEXT:    s_lshr_b32 s11, s7, 16
-; GFX8-NOHSA-NEXT:    s_or_b32 s0, s1, s0
-; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s1, s6
+; GFX8-NOHSA-NEXT:    s_and_b32 s15, s15, 0xffff0000
+; GFX8-NOHSA-NEXT:    s_and_b32 s20, 0xffff, s20
+; GFX8-NOHSA-NEXT:    s_or_b32 s12, s13, s12
+; GFX8-NOHSA-NEXT:    s_sext_i32_i16 s13, s6
 ; GFX8-NOHSA-NEXT:    s_ashr_i32 s7, s7, 16
-; GFX8-NOHSA-NEXT:    s_lshl_b32 s1, s1, 8
-; GFX8-NOHSA-NEXT:    s_bfe_i32 s12, s6, 0x80000
+; GFX8-NOHSA-NEXT:    s_or_b32 s15, s20, s15
+; GFX8-NOHSA-NEXT:    s_lshl_b32 s13, s13, 8
+; GFX8-NOHSA-NEXT:    s_bfe_i32 s20, s6, 0x80000
 ; GFX8-NOHSA-NEXT:    s_lshl_b32 s7, s7, 8
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s11, s11, 0x80000
 ; GFX8-NOHSA-NEXT:    s_ashr_i32 s6, s6, 24
 ; GFX8-NOHSA-NEXT:    s_bfe_i32 s10, s10, 0x80000
-; GFX8-NOHSA-NEXT:    s_and_b32 s1, s1, 0xffff0000
-; GFX8-NOHSA-NEXT:    s_and_b32 s12, 0xffff, s12
+; GFX8-NOHSA-NEXT:    s_and_b32 s13, s13, 0xffff0000
+; GFX8-NOHSA-NEXT:    s_and_b32 s20, 0xffff, s20
 ; GFX8-NOHSA-NEXT:    s_and_b32 s7, s7, 0xffff0000
 ; GFX8-NOHSA-NEXT:    s_and_b32 s11, 0xffff, s11
 ; GFX8-NOHSA-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX8-NOHSA-NEXT:    s_and_b32 s10, 0xffff, s10
-; GFX8-NOHSA-NEXT:    s_or_b32 s1, s12, s1
+; GFX8-NOHSA-NEXT:    s_or_b32 s13, s20, s13
 ; GFX8-NOHSA-NEXT:    s_or_b32 s7, s11, s7
 ; GFX8-NOHSA-NEXT:    s_or_b32 s6, s10, s6
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NOHSA-NEXT:    s_add_u32 s0, s8, 48
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s1
-; GFX8-NOHSA-NEXT:    s_addc_u32 s1, s9, 0
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
-; GFX8-NOHSA-NEXT:    s_add_u32 s0, s8, 32
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s6
+; GFX8-NOHSA-NEXT:    s_add_u32 s6, s8, 48
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s7
-; GFX8-NOHSA-NEXT:    s_addc_u32 s1, s9, 0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s7, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s13
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s12
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s6
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
-; GFX8-NOHSA-NEXT:    s_add_u32 s0, s8, 16
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s15
+; GFX8-NOHSA-NEXT:    s_nop 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s14
+; GFX8-NOHSA-NEXT:    s_add_u32 s4, s8, 32
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s5
-; GFX8-NOHSA-NEXT:    s_addc_u32 s1, s9, 0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s5, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s15
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s14
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s4
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s1
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s21
+; GFX8-NOHSA-NEXT:    s_nop 0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s2
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s20
+; GFX8-NOHSA-NEXT:    s_add_u32 s2, s8, 16
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s3
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s0
+; GFX8-NOHSA-NEXT:    s_addc_u32 s3, s9, 0
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s17
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s16
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v4, s8
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v0, s19
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s16
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v2, s18
-; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s17
+; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NOHSA-NEXT:    v_mov_b32_e32 v5, s9
 ; GFX8-NOHSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NOHSA-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/load-global-f32.ll b/llvm/test/CodeGen/AMDGPU/load-global-f32.ll
index 8d7eb6dcb843c..265997163c674 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-f32.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-f32.ll
@@ -260,19 +260,19 @@ define amdgpu_kernel void @global_load_v8f32(ptr addrspace(1) %out, ptr addrspac
 ; GCN-HSA-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 16
-; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
+; GCN-HSA-NEXT:    s_addc_u32 s2, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s2
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
@@ -343,8 +343,8 @@ define amdgpu_kernel void @global_load_v9f32(ptr addrspace(1) %out, ptr addrspac
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    flat_load_dword v14, v[6:7]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
@@ -354,10 +354,10 @@ define amdgpu_kernel void @global_load_v9f32(ptr addrspace(1) %out, ptr addrspac
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
@@ -431,10 +431,10 @@ define amdgpu_kernel void @global_load_v10f32(ptr addrspace(1) %out, ptr addrspa
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    flat_load_dwordx2 v[8:9], v[8:9]
@@ -444,10 +444,10 @@ define amdgpu_kernel void @global_load_v10f32(ptr addrspace(1) %out, ptr addrspa
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
@@ -522,10 +522,10 @@ define amdgpu_kernel void @global_load_v11f32(ptr addrspace(1) %out, ptr addrspa
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    flat_load_dwordx3 v[8:10], v[8:9]
@@ -535,10 +535,10 @@ define amdgpu_kernel void @global_load_v11f32(ptr addrspace(1) %out, ptr addrspa
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[11:12], v[0:3]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
@@ -612,10 +612,10 @@ define amdgpu_kernel void @global_load_v12f32(ptr addrspace(1) %out, ptr addrspa
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
@@ -625,10 +625,10 @@ define amdgpu_kernel void @global_load_v12f32(ptr addrspace(1) %out, ptr addrspa
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[0:3]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
@@ -709,13 +709,13 @@ define amdgpu_kernel void @global_load_v16f32(ptr addrspace(1) %out, ptr addrspa
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[12:13]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
@@ -724,16 +724,16 @@ define amdgpu_kernel void @global_load_v16f32(ptr addrspace(1) %out, ptr addrspa
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[0:3]
 ; GCN-HSA-NEXT:    s_nop 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[8:11]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
diff --git a/llvm/test/CodeGen/AMDGPU/load-global-i16.ll b/llvm/test/CodeGen/AMDGPU/load-global-i16.ll
index b150958ddd51d..9eb21b5b11036 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-i16.ll
@@ -233,11 +233,11 @@ define amdgpu_kernel void @global_load_v3i16(ptr addrspace(1) %out, ptr addrspac
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 4
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s0
+; GCN-HSA-NEXT:    s_addc_u32 s0, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    flat_store_short v[4:5], v1
 ; GCN-HSA-NEXT:    flat_store_dword v[2:3], v0
@@ -543,8 +543,8 @@ define amdgpu_kernel void @global_load_v16i16(ptr addrspace(1) %out, ptr addrspa
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
@@ -693,8 +693,8 @@ define amdgpu_kernel void @global_load_v16i16_align2(ptr addrspace(1) %in, ptr a
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    s_add_u32 s0, s2, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s3, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
@@ -1870,11 +1870,11 @@ define amdgpu_kernel void @global_zextload_v8i16_to_v8i32(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s0
+; GCN-HSA-NEXT:    s_addc_u32 s0, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 16, v3
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v9, 16, v2
@@ -2019,11 +2019,11 @@ define amdgpu_kernel void @global_sextload_v8i16_to_v8i32(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s0
+; GCN-HSA-NEXT:    s_addc_u32 s0, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v11, 16, v3
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v9, 16, v2
@@ -2190,18 +2190,18 @@ define amdgpu_kernel void @global_zextload_v16i16_to_v16i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 16, v1
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v9, 16, v0
@@ -2417,23 +2417,23 @@ define amdgpu_kernel void @global_sextload_v16i16_to_v16i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v9, 16, v0
 ; GCN-HSA-NEXT:    v_bfe_i32 v10, v1, 0, 16
@@ -2678,8 +2678,8 @@ define amdgpu_kernel void @global_zextload_v32i16_to_v32i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[0:1]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
@@ -2690,18 +2690,18 @@ define amdgpu_kernel void @global_zextload_v32i16_to_v32i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 48
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 64
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
@@ -2709,16 +2709,16 @@ define amdgpu_kernel void @global_zextload_v32i16_to_v32i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s8, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s9, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v28, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v31, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v30, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v31, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v19, 16, v5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v17, 16, v4
@@ -2741,8 +2741,8 @@ define amdgpu_kernel void @global_zextload_v32i16_to_v32i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v7, 16, v15
 ; GCN-HSA-NEXT:    v_and_b32_e32 v6, 0xffff, v15
 ; GCN-HSA-NEXT:    v_and_b32_e32 v4, 0xffff, v14
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s7
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[4:7]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(4)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v23, 16, v3
@@ -3074,8 +3074,8 @@ define amdgpu_kernel void @global_sextload_v32i16_to_v32i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[0:1]
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[0:1]
@@ -3088,19 +3088,19 @@ define amdgpu_kernel void @global_sextload_v32i16_to_v32i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v19, 16, v13
@@ -3116,9 +3116,9 @@ define amdgpu_kernel void @global_sextload_v32i16_to_v32i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v16, 16, v14
 ; GCN-HSA-NEXT:    v_bfe_i32 v17, v15, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v15, v14, 0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v28, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[15:18]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(4)
@@ -3142,18 +3142,18 @@ define amdgpu_kernel void @global_sextload_v32i16_to_v32i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v12, 16, v4
 ; GCN-HSA-NEXT:    v_bfe_i32 v13, v5, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v11, v4, 0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[19:20], v[11:14]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[7:10]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v31, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v30, s0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(6)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v10, 16, v1
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v8, 16, v0
 ; GCN-HSA-NEXT:    v_bfe_i32 v9, v1, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v7, v0, 0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v30, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v31, s1
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v6, 16, v3
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v4, 16, v2
 ; GCN-HSA-NEXT:    v_bfe_i32 v5, v3, 0, 16
@@ -3567,53 +3567,58 @@ define amdgpu_kernel void @global_zextload_v64i16_to_v64i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_addc_u32 s9, s3, 0
 ; GCN-HSA-NEXT:    s_add_u32 s10, s2, 64
 ; GCN-HSA-NEXT:    s_addc_u32 s11, s3, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s9
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[20:23], v[16:17]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s11
 ; GCN-HSA-NEXT:    s_add_u32 s10, s2, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s11
 ; GCN-HSA-NEXT:    s_addc_u32 s11, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s10
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-HSA-NEXT:    s_add_u32 s10, s2, 0x60
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
-; GCN-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[4:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-HSA-NEXT:    s_addc_u32 s11, s3, 0
+; GCN-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[4:5]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 0x70
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s10
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s11
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[28:31], v[12:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s7
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s7
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[16:19], v[12:13]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[14:15]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v37, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v36, s0
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(7)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v35, 16, v25
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v33, 16, v24
 ; GCN-HSA-NEXT:    v_and_b32_e32 v34, 0xffff, v25
 ; GCN-HSA-NEXT:    v_and_b32_e32 v32, 0xffff, v24
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s2
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[36:37], v[32:35]
+; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v33, 16, v26
+; GCN-HSA-NEXT:    v_and_b32_e32 v32, 0xffff, v26
+; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xe0
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v35, 16, v27
+; GCN-HSA-NEXT:    v_and_b32_e32 v34, 0xffff, v27
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0xf0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s2
+; GCN-HSA-NEXT:    s_addc_u32 s2, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s3
+; GCN-HSA-NEXT:    s_add_u32 s3, s0, 0xc0
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[36:37], v[32:35]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v37, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v36, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s6, s0, 0xd0
 ; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s8, s0, 0xa0
@@ -3621,51 +3626,47 @@ define amdgpu_kernel void @global_zextload_v64i16_to_v64i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_add_u32 s10, s0, 0xb0
 ; GCN-HSA-NEXT:    s_addc_u32 s11, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s12, s0, 0x80
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v35, 16, v27
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v33, 16, v26
-; GCN-HSA-NEXT:    v_and_b32_e32 v34, 0xffff, v27
-; GCN-HSA-NEXT:    v_and_b32_e32 v32, 0xffff, v26
 ; GCN-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[32:35]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[32:35]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s13
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(7)
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v27, 16, v1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v33, s13
-; GCN-HSA-NEXT:    v_mov_b32_e32 v32, s12
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v25, 16, v0
-; GCN-HSA-NEXT:    v_and_b32_e32 v26, 0xffff, v1
-; GCN-HSA-NEXT:    v_and_b32_e32 v24, 0xffff, v0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[24:27]
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v35, 16, v1
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v33, 16, v0
+; GCN-HSA-NEXT:    v_and_b32_e32 v34, 0xffff, v1
+; GCN-HSA-NEXT:    v_and_b32_e32 v32, 0xffff, v0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[32:35]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(7)
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v27, 16, v9
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v25, 16, v8
-; GCN-HSA-NEXT:    v_and_b32_e32 v26, 0xffff, v9
-; GCN-HSA-NEXT:    v_and_b32_e32 v24, 0xffff, v8
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[24:27]
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v35, 16, v9
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v33, 16, v8
+; GCN-HSA-NEXT:    v_and_b32_e32 v34, 0xffff, v9
+; GCN-HSA-NEXT:    v_and_b32_e32 v32, 0xffff, v8
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[32:35]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s10
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v27, 16, v11
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v25, 16, v10
-; GCN-HSA-NEXT:    v_and_b32_e32 v26, 0xffff, v11
-; GCN-HSA-NEXT:    v_and_b32_e32 v24, 0xffff, v10
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v35, 16, v11
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v33, 16, v10
+; GCN-HSA-NEXT:    v_and_b32_e32 v34, 0xffff, v11
+; GCN-HSA-NEXT:    v_and_b32_e32 v32, 0xffff, v10
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s11
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[24:27]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s2
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[32:35]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(8)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 16, v5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v9, 16, v4
 ; GCN-HSA-NEXT:    v_and_b32_e32 v10, 0xffff, v5
 ; GCN-HSA-NEXT:    v_and_b32_e32 v8, 0xffff, v4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x90
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v27, 16, v7
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v25, 16, v6
-; GCN-HSA-NEXT:    v_and_b32_e32 v26, 0xffff, v7
-; GCN-HSA-NEXT:    v_and_b32_e32 v24, 0xffff, v6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v33, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v32, s6
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v35, 16, v7
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v33, 16, v6
+; GCN-HSA-NEXT:    v_and_b32_e32 v34, 0xffff, v7
+; GCN-HSA-NEXT:    v_and_b32_e32 v32, 0xffff, v6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[8:11]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(8)
 ; GCN-HSA-NEXT:    v_and_b32_e32 v7, 0xffff, v28
@@ -3673,31 +3674,30 @@ define amdgpu_kernel void @global_zextload_v64i16_to_v64i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v8, 16, v28
 ; GCN-HSA-NEXT:    v_and_b32_e32 v9, 0xffff, v29
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[24:27]
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v6, 16, v3
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v27, 16, v31
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v25, 16, v30
-; GCN-HSA-NEXT:    v_and_b32_e32 v26, 0xffff, v31
-; GCN-HSA-NEXT:    v_and_b32_e32 v24, 0xffff, v30
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[36:37], v[7:10]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[24:27]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[32:35]
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v29, 16, v31
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v27, 16, v30
+; GCN-HSA-NEXT:    v_and_b32_e32 v28, 0xffff, v31
+; GCN-HSA-NEXT:    v_and_b32_e32 v26, 0xffff, v30
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[7:10]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[26:29]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v6, 16, v3
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v2
 ; GCN-HSA-NEXT:    v_and_b32_e32 v5, 0xffff, v3
 ; GCN-HSA-NEXT:    v_and_b32_e32 v3, 0xffff, v2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[3:6]
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v1, 16, v20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v3, 16, v21
 ; GCN-HSA-NEXT:    v_and_b32_e32 v2, 0xffff, v21
 ; GCN-HSA-NEXT:    v_and_b32_e32 v0, 0xffff, v20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -3705,44 +3705,44 @@ define amdgpu_kernel void @global_zextload_v64i16_to_v64i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v5, 16, v22
 ; GCN-HSA-NEXT:    v_and_b32_e32 v6, 0xffff, v23
 ; GCN-HSA-NEXT:    v_and_b32_e32 v4, 0xffff, v22
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    s_waitcnt vmcnt(12)
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v1, 16, v18
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(12)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 16, v15
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v7, 16, v17
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v5, 16, v16
-; GCN-HSA-NEXT:    v_and_b32_e32 v0, 0xffff, v18
-; GCN-HSA-NEXT:    v_and_b32_e32 v6, 0xffff, v17
 ; GCN-HSA-NEXT:    v_and_b32_e32 v4, 0xffff, v16
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v18, 16, v13
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v16, 16, v12
 ; GCN-HSA-NEXT:    v_and_b32_e32 v10, 0xffff, v15
-; GCN-HSA-NEXT:    v_and_b32_e32 v17, 0xffff, v13
 ; GCN-HSA-NEXT:    v_and_b32_e32 v15, 0xffff, v12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x50
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v1, 16, v18
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v7, 16, v17
+; GCN-HSA-NEXT:    v_and_b32_e32 v0, 0xffff, v18
+; GCN-HSA-NEXT:    v_and_b32_e32 v6, 0xffff, v17
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v18, 16, v13
+; GCN-HSA-NEXT:    v_and_b32_e32 v17, 0xffff, v13
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[15:18]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v9, 16, v14
 ; GCN-HSA-NEXT:    v_and_b32_e32 v8, 0xffff, v14
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 48
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[8:11]
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v3, 16, v19
 ; GCN-HSA-NEXT:    v_and_b32_e32 v2, 0xffff, v19
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; GCN-HSA-NEXT:    s_nop 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
@@ -4342,50 +4342,50 @@ define amdgpu_kernel void @global_sextload_v64i16_to_v64i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 0x70
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 0x60
-; GCN-HSA-NEXT:    flat_load_dwordx4 v[20:23], v[0:1]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
+; GCN-HSA-NEXT:    flat_load_dwordx4 v[20:23], v[0:1]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[0:1]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 64
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[0:1]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    s_add_u32 s6, s2, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s7, s3, 0
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[24:27], v[8:9]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s7
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s7
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[16:19], v[16:17]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v37, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v36, s0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(7)
-; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v35, 16, v29
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v33, 16, v28
-; GCN-HSA-NEXT:    v_bfe_i32 v34, v29, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v32, v28, 0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v28, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xe0
+; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v35, 16, v29
+; GCN-HSA-NEXT:    v_bfe_i32 v34, v29, 0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[36:37], v[32:35]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v36, s3
@@ -4399,87 +4399,86 @@ define amdgpu_kernel void @global_sextload_v64i16_to_v64i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[31:34]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(8)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v29, 16, v20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v32, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v31, 16, v21
 ; GCN-HSA-NEXT:    v_bfe_i32 v30, v21, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v28, v20, 0, 16
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[35:36], v[28:31]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v35, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v34, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xd0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[35:36], v[28:31]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v35, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v37, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v36, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xa0
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v31, 16, v23
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v29, 16, v22
 ; GCN-HSA-NEXT:    v_bfe_i32 v30, v23, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v28, v22, 0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v37, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[28:31]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v32, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xb0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v39, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v38, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x80
+; GCN-HSA-NEXT:    v_mov_b32_e32 v39, s3
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(9)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v23, 16, v13
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v21, 16, v12
 ; GCN-HSA-NEXT:    v_bfe_i32 v22, v13, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v20, v12, 0, 16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v31, 16, v15
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v29, 16, v14
 ; GCN-HSA-NEXT:    v_bfe_i32 v30, v15, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v28, v14, 0, 16
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[34:35], v[20:23]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[36:37], v[28:31]
-; GCN-HSA-NEXT:    s_waitcnt vmcnt(10)
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(8)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v15, 16, v5
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v13, 16, v4
 ; GCN-HSA-NEXT:    v_bfe_i32 v14, v5, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v12, v4, 0, 16
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(7)
+; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v5, 16, v0
+; GCN-HSA-NEXT:    v_bfe_i32 v4, v0, 0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x90
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[34:35], v[20:23]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[36:37], v[28:31]
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v23, 16, v7
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v21, 16, v6
 ; GCN-HSA-NEXT:    v_bfe_i32 v22, v7, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v20, v6, 0, 16
-; GCN-HSA-NEXT:    s_waitcnt vmcnt(9)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v7, 16, v1
-; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v5, 16, v0
 ; GCN-HSA-NEXT:    v_bfe_i32 v6, v1, 0, 16
-; GCN-HSA-NEXT:    v_bfe_i32 v4, v0, 0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x90
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[12:15]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[38:39], v[20:23]
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v23, 16, v3
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v21, 16, v2
 ; GCN-HSA-NEXT:    v_bfe_i32 v22, v3, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v20, v2, 0, 16
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[20:23]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(11)
-; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v3, 16, v9
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v1, 16, v8
-; GCN-HSA-NEXT:    v_bfe_i32 v2, v9, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v0, v8, 0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v3, 16, v9
+; GCN-HSA-NEXT:    v_bfe_i32 v2, v9, 0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v13, 16, v24
@@ -4488,9 +4487,9 @@ define amdgpu_kernel void @global_sextload_v64i16_to_v64i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v22, 16, v10
 ; GCN-HSA-NEXT:    v_bfe_i32 v23, v11, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v21, v10, 0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[21:24]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x50
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(12)
@@ -4502,11 +4501,12 @@ define amdgpu_kernel void @global_sextload_v64i16_to_v64i32(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v18, 16, v16
 ; GCN-HSA-NEXT:    v_bfe_i32 v19, v17, 0, 16
 ; GCN-HSA-NEXT:    v_bfe_i32 v17, v16, 0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[17:20]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 48
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v15, 16, v25
@@ -5678,14 +5678,14 @@ define amdgpu_kernel void @global_zextload_v4i16_to_v4i64(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    flat_load_dwordx2 v[8:9], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, 0
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s0
+; GCN-HSA-NEXT:    s_addc_u32 s0, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v7, v1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v2, 16, v9
 ; GCN-HSA-NEXT:    v_and_b32_e32 v0, 0xffff, v9
@@ -5823,8 +5823,8 @@ define amdgpu_kernel void @global_sextload_v4i16_to_v4i64(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
 ; GCN-HSA-NEXT:    v_bfe_i32 v4, v1, 0, 16
@@ -5985,9 +5985,9 @@ define amdgpu_kernel void @global_zextload_v8i16_to_v8i64(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s0
@@ -6000,10 +6000,10 @@ define amdgpu_kernel void @global_zextload_v8i16_to_v8i64(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, v4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, v4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, v4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v5, 16, v3
 ; GCN-HSA-NEXT:    v_and_b32_e32 v3, 0xffff, v3
@@ -6236,15 +6236,15 @@ define amdgpu_kernel void @global_sextload_v8i16_to_v8i64(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    s_bfe_i64 s[6:7], s[6:7], 0x100000
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
 ; GCN-HSA-NEXT:    s_add_u32 s6, s0, 48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s7
-; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
+; GCN-HSA-NEXT:    s_addc_u32 s2, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s19
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s18
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s7
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s7
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s6
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s13
@@ -6476,18 +6476,18 @@ define amdgpu_kernel void @global_zextload_v16i16_to_v16i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x50
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, v8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, v8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, v8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, v8
@@ -6499,28 +6499,28 @@ define amdgpu_kernel void @global_zextload_v16i16_to_v16i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 16, v1
 ; GCN-HSA-NEXT:    v_and_b32_e32 v11, 0xffff, v1
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[21:22], v[11:14]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s5
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v16, 16, v5
 ; GCN-HSA-NEXT:    v_and_b32_e32 v14, 0xffff, v5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[21:22], v[14:17]
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v19, 16, v7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_and_b32_e32 v17, 0xffff, v7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[17:20]
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v9, 16, v3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCN-HSA-NEXT:    v_and_b32_e32 v7, 0xffff, v3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[7:10]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s1
@@ -6540,10 +6540,10 @@ define amdgpu_kernel void @global_zextload_v16i16_to_v16i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v8, 16, v6
 ; GCN-HSA-NEXT:    v_and_b32_e32 v6, 0xffff, v6
 ; GCN-HSA-NEXT:    v_and_b32_e32 v14, 0xffff, v4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s1
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v22, 16, v0
 ; GCN-HSA-NEXT:    v_and_b32_e32 v20, 0xffff, v0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[10:13]
@@ -6875,8 +6875,8 @@ define amdgpu_kernel void @global_sextload_v16i16_to_v16i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s0
@@ -6938,38 +6938,38 @@ define amdgpu_kernel void @global_sextload_v16i16_to_v16i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_addc_u32 s11, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[12:15]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s21
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s11
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s10
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s11
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s19
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[4:7]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x50
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[14:17]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s17
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s15
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s6
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s7
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[18:21]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 64
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s24
@@ -6977,14 +6977,14 @@ define amdgpu_kernel void @global_sextload_v16i16_to_v16i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s5
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[8:11]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s22
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s23
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[22:25]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
@@ -7363,35 +7363,35 @@ define amdgpu_kernel void @global_zextload_v32i16_to_v32i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_add_u32 s14, s0, 0x70
 ; GCN-HSA-NEXT:    s_addc_u32 s15, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s15
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s14
 ; GCN-HSA-NEXT:    s_add_u32 s14, s0, 0x50
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, v1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s15
 ; GCN-HSA-NEXT:    s_addc_u32 s15, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v24, v1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v20, 16, v5
 ; GCN-HSA-NEXT:    v_and_b32_e32 v18, 0xffff, v5
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[18:21]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s15
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s15
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v20, 16, v3
 ; GCN-HSA-NEXT:    v_and_b32_e32 v18, 0xffff, v3
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[18:21]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s11
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s10
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s11
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(4)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v20, 16, v9
 ; GCN-HSA-NEXT:    v_and_b32_e32 v18, 0xffff, v9
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[18:21]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s13
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s13
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v20, 16, v7
 ; GCN-HSA-NEXT:    v_and_b32_e32 v18, 0xffff, v7
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[18:21]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s5
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(5)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v20, 16, v11
 ; GCN-HSA-NEXT:    v_and_b32_e32 v18, 0xffff, v11
@@ -7405,14 +7405,14 @@ define amdgpu_kernel void @global_zextload_v32i16_to_v32i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_and_b32_e32 v17, 0xffff, v17
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 32
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[21:22], v[17:20]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s9
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v19, 16, v15
 ; GCN-HSA-NEXT:    v_and_b32_e32 v17, 0xffff, v15
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[21:22], v[17:20]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v19, 16, v12
 ; GCN-HSA-NEXT:    v_and_b32_e32 v17, 0xffff, v12
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0xe0
@@ -7427,39 +7427,39 @@ define amdgpu_kernel void @global_zextload_v32i16_to_v32i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, v1
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 16, v16
 ; GCN-HSA-NEXT:    v_and_b32_e32 v9, 0xffff, v16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s5
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[9:12]
 ; GCN-HSA-NEXT:    v_and_b32_e32 v0, 0xffff, v13
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 16, v2
 ; GCN-HSA-NEXT:    v_and_b32_e32 v9, 0xffff, v2
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v2, 16, v13
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[0:3]
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v23, 16, v14
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xa0
 ; GCN-HSA-NEXT:    v_and_b32_e32 v21, 0xffff, v14
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, v1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[21:24]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x80
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v20, 16, v8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v18, 0xffff, v8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, v1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[18:21]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 64
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v17, 16, v6
@@ -8073,171 +8073,170 @@ define amdgpu_kernel void @global_sextload_v32i16_to_v32i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 48
+; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
-; GCN-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[0:1]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 32
+; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
-; GCN-HSA-NEXT:    flat_load_dwordx4 v[16:19], v[0:1]
+; GCN-HSA-NEXT:    flat_load_dwordx4 v[14:17], v[4:5]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCN-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[0:1]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-HSA-NEXT:    flat_load_dwordx4 v[2:5], v[0:1]
+; GCN-HSA-NEXT:    flat_load_dwordx4 v[10:13], v[4:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    flat_load_dwordx4 v[6:9], v[4:5]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(3)
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s7, v13
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s9, v15
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s8, v14
-; GCN-HSA-NEXT:    s_mov_b32 s10, s9
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s7, v1
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s8, v2
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s9, v3
 ; GCN-HSA-NEXT:    s_mov_b32 s12, s7
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s19, v19
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s6, v12
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s19, v17
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s6, v0
+; GCN-HSA-NEXT:    s_mov_b32 s10, s9
 ; GCN-HSA-NEXT:    s_lshr_b32 s16, s8, 16
 ; GCN-HSA-NEXT:    s_bfe_i64 s[4:5], s[8:9], 0x100000
-; GCN-HSA-NEXT:    s_ashr_i32 s49, s9, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s50, s9, 16
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s8, v16
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s9, v17
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s18, v18
+; GCN-HSA-NEXT:    s_ashr_i32 s48, s9, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s49, s9, 16
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s8, v14
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s9, v15
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s18, v16
 ; GCN-HSA-NEXT:    s_bfe_i64 s[12:13], s[12:13], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[10:11], s[10:11], 0x100000
 ; GCN-HSA-NEXT:    s_mov_b32 s20, s19
 ; GCN-HSA-NEXT:    s_lshr_b32 s14, s6, 16
 ; GCN-HSA-NEXT:    s_bfe_i64 s[2:3], s[6:7], 0x100000
 ; GCN-HSA-NEXT:    s_ashr_i32 s33, s7, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s48, s7, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s47, s7, 16
+; GCN-HSA-NEXT:    s_bfe_i64 s[26:27], s[16:17], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[10:11], s[10:11], 0x100000
 ; GCN-HSA-NEXT:    s_mov_b32 s22, s9
-; GCN-HSA-NEXT:    s_lshr_b32 s24, s18, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s28, s8, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s28, s18, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s24, s8, 16
 ; GCN-HSA-NEXT:    s_bfe_i64 s[6:7], s[8:9], 0x100000
-; GCN-HSA-NEXT:    s_ashr_i32 s51, s9, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s52, s9, 16
+; GCN-HSA-NEXT:    s_bfe_i64 s[16:17], s[18:19], 0x100000
+; GCN-HSA-NEXT:    s_ashr_i32 s50, s9, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s51, s9, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s52, s19, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s53, s19, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s13
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s8, v10
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s9, v11
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s12, v12
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s13, v13
+; GCN-HSA-NEXT:    s_bfe_i64 s[18:19], s[20:21], 0x100000
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s11
-; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s13
-; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s9, v9
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s10, v10
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s11, v11
-; GCN-HSA-NEXT:    s_bfe_i64 s[12:13], s[20:21], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[26:27], s[16:17], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[16:17], s[18:19], 0x100000
-; GCN-HSA-NEXT:    s_ashr_i32 s53, s19, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s54, s19, 16
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s8, v8
-; GCN-HSA-NEXT:    s_bfe_i64 s[18:19], s[28:29], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[30:31], s[24:25], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[40:41], s[22:23], 0x100000
-; GCN-HSA-NEXT:    s_mov_b32 s22, s11
-; GCN-HSA-NEXT:    s_mov_b32 s24, s9
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s12
-; GCN-HSA-NEXT:    s_ashr_i32 s57, s11, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s58, s11, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s12, s10, 16
-; GCN-HSA-NEXT:    s_bfe_i64 s[28:29], s[10:11], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[34:35], s[28:29], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[42:43], s[22:23], 0x100000
+; GCN-HSA-NEXT:    s_mov_b32 s20, s13
+; GCN-HSA-NEXT:    s_mov_b32 s36, s9
+; GCN-HSA-NEXT:    s_lshr_b32 s22, s8, 16
+; GCN-HSA-NEXT:    s_bfe_i64 s[10:11], s[8:9], 0x100000
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s18
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s19
+; GCN-HSA-NEXT:    s_ashr_i32 s56, s13, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s57, s13, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s8, s12, 16
+; GCN-HSA-NEXT:    s_bfe_i64 s[28:29], s[12:13], 0x100000
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s10, v2
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s11, v3
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s36, v4
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s37, v5
-; GCN-HSA-NEXT:    s_lshr_b32 s20, s8, 16
-; GCN-HSA-NEXT:    s_bfe_i64 s[34:35], s[12:13], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[38:39], s[24:25], 0x100000
-; GCN-HSA-NEXT:    s_mov_b32 s44, s11
-; GCN-HSA-NEXT:    s_lshr_b32 s12, s10, 16
-; GCN-HSA-NEXT:    s_mov_b32 s46, s37
-; GCN-HSA-NEXT:    s_lshr_b32 s24, s36, 16
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s12, v6
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s13, v7
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s18, v8
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s19, v9
+; GCN-HSA-NEXT:    s_bfe_i64 s[40:41], s[36:37], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[44:45], s[20:21], 0x100000
+; GCN-HSA-NEXT:    s_mov_b32 s36, s13
+; GCN-HSA-NEXT:    s_lshr_b32 s20, s12, 16
+; GCN-HSA-NEXT:    s_mov_b32 s38, s19
+; GCN-HSA-NEXT:    s_lshr_b32 s46, s18, 16
 ; GCN-HSA-NEXT:    s_bfe_i64 s[14:15], s[14:15], 0x100000
-; GCN-HSA-NEXT:    s_ashr_i32 s55, s9, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s56, s9, 16
-; GCN-HSA-NEXT:    s_bfe_i64 s[8:9], s[8:9], 0x100000
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s13
-; GCN-HSA-NEXT:    s_bfe_i64 s[20:21], s[20:21], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[42:43], s[22:23], 0x100000
-; GCN-HSA-NEXT:    s_ashr_i32 s59, s11, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s60, s11, 16
-; GCN-HSA-NEXT:    s_bfe_i64 s[10:11], s[10:11], 0x100000
-; GCN-HSA-NEXT:    s_ashr_i32 s61, s37, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s62, s37, 16
-; GCN-HSA-NEXT:    s_bfe_i64 s[22:23], s[36:37], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[12:13], s[12:13], 0x100000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[24:25], s[24:25], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[36:37], s[44:45], 0x100000
-; GCN-HSA-NEXT:    s_bfe_i64 s[44:45], s[46:47], 0x100000
+; GCN-HSA-NEXT:    s_ashr_i32 s54, s9, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s55, s9, 16
+; GCN-HSA-NEXT:    s_bfe_i64 s[22:23], s[22:23], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[30:31], s[8:9], 0x100000
+; GCN-HSA-NEXT:    s_ashr_i32 s58, s13, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s59, s13, 16
+; GCN-HSA-NEXT:    s_bfe_i64 s[8:9], s[12:13], 0x100000
+; GCN-HSA-NEXT:    s_ashr_i32 s60, s19, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s61, s19, 16
+; GCN-HSA-NEXT:    s_bfe_i64 s[18:19], s[18:19], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[12:13], s[20:21], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[20:21], s[46:47], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[36:37], s[36:37], 0x100000
+; GCN-HSA-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x100000
 ; GCN-HSA-NEXT:    s_add_u32 s46, s0, 48
-; GCN-HSA-NEXT:    s_addc_u32 s47, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s46
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s47
+; GCN-HSA-NEXT:    s_addc_u32 s62, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s46
 ; GCN-HSA-NEXT:    s_add_u32 s46, s0, 16
-; GCN-HSA-NEXT:    s_addc_u32 s47, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s62
+; GCN-HSA-NEXT:    s_addc_u32 s62, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s46
-; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s47
 ; GCN-HSA-NEXT:    s_add_u32 s46, s0, 0xf0
-; GCN-HSA-NEXT:    s_addc_u32 s47, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s40
-; GCN-HSA-NEXT:    s_add_u32 s40, s0, 0xd0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s41
-; GCN-HSA-NEXT:    s_addc_u32 s41, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s40
-; GCN-HSA-NEXT:    v_mov_b32_e32 v30, s41
-; GCN-HSA-NEXT:    s_add_u32 s40, s0, 0xb0
-; GCN-HSA-NEXT:    s_addc_u32 s41, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s38
-; GCN-HSA-NEXT:    s_add_u32 s38, s0, 0x90
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s50
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s49
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s39
-; GCN-HSA-NEXT:    s_addc_u32 s39, s1, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s62
+; GCN-HSA-NEXT:    s_addc_u32 s62, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s42
+; GCN-HSA-NEXT:    s_add_u32 s42, s0, 0xd0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s43
+; GCN-HSA-NEXT:    s_addc_u32 s43, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s42
+; GCN-HSA-NEXT:    s_add_u32 s42, s0, 0xb0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v30, s43
+; GCN-HSA-NEXT:    s_addc_u32 s43, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s33
+; GCN-HSA-NEXT:    s_add_u32 s33, s0, 0x90
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s49
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s40
+; GCN-HSA-NEXT:    s_addc_u32 s40, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s46
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s38
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s39
-; GCN-HSA-NEXT:    s_add_u32 s38, s0, 0x70
-; GCN-HSA-NEXT:    v_mov_b32_e32 v28, s47
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s33
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s54
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s53
-; GCN-HSA-NEXT:    s_addc_u32 s39, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s42
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s43
-; GCN-HSA-NEXT:    v_mov_b32_e32 v31, s40
-; GCN-HSA-NEXT:    v_mov_b32_e32 v32, s41
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s52
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s51
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s58
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s57
-; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s56
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[25:26], v[6:9]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s55
+; GCN-HSA-NEXT:    v_mov_b32_e32 v28, s62
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s47
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s53
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s52
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s44
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s33
+; GCN-HSA-NEXT:    s_add_u32 s33, s0, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s45
+; GCN-HSA-NEXT:    v_mov_b32_e32 v31, s42
+; GCN-HSA-NEXT:    v_mov_b32_e32 v32, s43
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s51
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s57
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s56
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s41
+; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s55
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s40
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[25:26], v[4:7]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s54
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[27:28], v[10:13]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[29:30], v[14:17]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[31:32], v[18:21]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[22:25]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s44
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s45
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s62
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s61
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s38
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s39
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s38
+; GCN-HSA-NEXT:    s_addc_u32 s38, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s39
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s61
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s33
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s38
+; GCN-HSA-NEXT:    s_add_u32 s33, s0, 0x50
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    s_nop 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s33
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s36
-; GCN-HSA-NEXT:    s_add_u32 s36, s0, 0x50
+; GCN-HSA-NEXT:    s_addc_u32 s36, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s37
-; GCN-HSA-NEXT:    s_addc_u32 s37, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s60
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s59
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s36
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s37
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s59
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s58
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s36
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_nop 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
@@ -8258,57 +8257,57 @@ define amdgpu_kernel void @global_sextload_v32i16_to_v32i64(ptr addrspace(1) %ou
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s30
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s31
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s34
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s35
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xa0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s18
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s19
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s24
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s25
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x80
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s28
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s29
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s34
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s35
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s30
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s31
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s21
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s11
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s22
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s23
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 64
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s22
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s23
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s24
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s25
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s18
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s19
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s20
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s21
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s11
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s12
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s13
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/load-global-i32.ll b/llvm/test/CodeGen/AMDGPU/load-global-i32.ll
index 930fe409c2cea..ec4da41174c1b 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-i32.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-i32.ll
@@ -378,8 +378,8 @@ define amdgpu_kernel void @global_load_v8i32(ptr addrspace(1) %out, ptr addrspac
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s0
 ; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v10, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(1)
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(1)
@@ -480,11 +480,11 @@ define amdgpu_kernel void @global_load_v9i32(ptr addrspace(1) %out, ptr addrspac
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s4
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCNX3-HSA-NEXT:    flat_load_dword v14, v[8:9]
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 16
@@ -493,10 +493,10 @@ define amdgpu_kernel void @global_load_v9i32(ptr addrspace(1) %out, ptr addrspac
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s0
 ; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v10, s2
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v12, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s1
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(2)
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(2)
@@ -611,10 +611,10 @@ define amdgpu_kernel void @global_load_v10i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s4
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx2 v[8:9], v[8:9]
@@ -624,10 +624,10 @@ define amdgpu_kernel void @global_load_v10i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v10, s0
 ; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v12, s2
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v14, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, s1
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(2)
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(2)
@@ -741,10 +741,10 @@ define amdgpu_kernel void @global_load_v11i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s4
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx3 v[8:10], v[8:9]
@@ -754,10 +754,10 @@ define amdgpu_kernel void @global_load_v11i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v11, s0
 ; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v14, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s2
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v14, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s1
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(2)
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[11:12], v[0:3]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(2)
@@ -874,10 +874,10 @@ define amdgpu_kernel void @global_load_v12i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s4
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
@@ -887,10 +887,10 @@ define amdgpu_kernel void @global_load_v12i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v12, s0
 ; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v14, s2
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v17, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v17, s1
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(2)
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[0:3]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(2)
@@ -1003,9 +1003,9 @@ define amdgpu_kernel void @global_load_v16i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v12, s4
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 48
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
@@ -1013,26 +1013,26 @@ define amdgpu_kernel void @global_load_v16i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s5
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[12:13]
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v18, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v17, s1
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s0
 ; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v20, s2
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v22, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, s1
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(3)
@@ -1710,8 +1710,8 @@ define amdgpu_kernel void @global_zextload_v4i32_to_v4i64(ptr addrspace(1) %out,
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v7, v5
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, v2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v6, v3
@@ -1835,11 +1835,11 @@ define amdgpu_kernel void @global_sextload_v4i32_to_v4i64(ptr addrspace(1) %out,
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v14, s3
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v12, s1
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v11, s0
+; GCNX3-HSA-NEXT:    s_addc_u32 s0, s1, 0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v14, s0
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v10, 31, v3
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v8, 31, v2
@@ -1980,21 +1980,21 @@ define amdgpu_kernel void @global_zextload_v8i32_to_v8i64(ptr addrspace(1) %out,
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v14, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v11, v9
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s1
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v12, s0
 ; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 32
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v17, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v17, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(1)
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, v2
@@ -2171,23 +2171,23 @@ define amdgpu_kernel void @global_sextload_v8i32_to_v8i64(ptr addrspace(1) %out,
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v18, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v17, s1
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s0
 ; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 32
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v20, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, s1
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v22, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, s1
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(1)
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v11, 31, v1
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v9, 31, v0
@@ -2407,8 +2407,8 @@ define amdgpu_kernel void @global_sextload_v16i32_to_v16i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[0:1]
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 32
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[0:1]
@@ -2421,19 +2421,19 @@ define amdgpu_kernel void @global_sextload_v16i32_to_v16i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v22, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v24, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v26, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s1
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v20, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v19, 31, v9
@@ -2441,17 +2441,17 @@ define amdgpu_kernel void @global_sextload_v16i32_to_v16i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, v8
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v18, v9
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[16:19]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v20, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x50
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v19, 31, v11
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v17, 31, v10
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, v10
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v18, v11
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[16:19]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(4)
@@ -2475,18 +2475,18 @@ define amdgpu_kernel void @global_sextload_v16i32_to_v16i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v14, v5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, v6
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v10, v7
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v22, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[12:15]
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[8:11]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v31, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v30, s0
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(6)
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v11, 31, v1
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v9, 31, v0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, v0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v10, v1
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v30, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v31, s1
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v7, 31, v3
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v5, 31, v2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, v2
@@ -2764,63 +2764,63 @@ define amdgpu_kernel void @global_zextload_v16i32_to_v16i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 48
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s7
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s6
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s7
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v12, s4
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[12:13]
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v22, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v24, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v26, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v19, v17
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s1
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v20, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, v0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v18, v1
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[16:19]
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x50
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, v2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v18, v3
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[16:19]
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, s2
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(4)
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, v4
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v18, v5
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[16:19]
-; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, v6
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v18, v7
+; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[16:19]
 ; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 48
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(5)
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, v8
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v18, v9
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[16:19]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s3
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v20, s2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, v10
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v18, v11
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v20, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[16:19]
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s0
@@ -3135,13 +3135,13 @@ define amdgpu_kernel void @global_sextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 0x60
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 0x50
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[20:23], v[0:1]
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 64
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[0:1]
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s4
@@ -3150,16 +3150,16 @@ define amdgpu_kernel void @global_sextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 48
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCNX3-HSA-NEXT:    s_add_u32 s6, s2, 32
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s4
 ; GCNX3-HSA-NEXT:    s_addc_u32 s7, s3, 0
-; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[16:19], v[8:9]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s4
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 16
+; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[16:19], v[8:9]
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s7
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s6
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s7
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 16
@@ -3167,52 +3167,51 @@ define amdgpu_kernel void @global_sextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v37, s1
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v36, s0
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(7)
-; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v35, 31, v29
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v33, 31, v28
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v32, v28
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v34, v29
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xe0
+; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v35, 31, v29
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v34, v29
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[36:37], v[32:35]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v37, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v36, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xf0
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v35, 31, v31
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v33, 31, v30
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v32, v30
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v34, v31
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v37, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[32:35]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(8)
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v31, 31, v25
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v32, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v35, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v34, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xd0
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v29, 31, v24
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, v24
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v30, v25
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v35, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[36:37], v[28:31]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v37, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v36, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xa0
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v31, 31, v27
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v29, 31, v26
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, v26
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v30, v27
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v37, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[28:31]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v32, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xb0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v39, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v38, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x80
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(9)
@@ -3220,6 +3219,7 @@ define amdgpu_kernel void @global_sextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v25, 31, v20
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v24, v20
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v26, v21
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v39, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v31, 31, v23
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v29, 31, v22
@@ -3228,67 +3228,67 @@ define amdgpu_kernel void @global_sextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[34:35], v[24:27]
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[36:37], v[28:31]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(10)
-; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v23, 31, v15
-; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v21, 31, v14
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v27, 31, v13
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v25, 31, v12
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v24, v12
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v26, v13
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v20, v14
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v22, v15
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(9)
-; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v15, 31, v5
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v13, 31, v4
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v12, v4
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v14, v5
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x90
+; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v23, 31, v15
+; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v21, 31, v14
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v20, v14
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v22, v15
+; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v15, 31, v5
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v14, v5
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[12:15]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[24:27]
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[38:39], v[20:23]
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v26, 31, v7
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v24, 31, v6
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, v6
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v25, v7
-; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, s2
+; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x70
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[23:26]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(12)
+; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v7, 31, v17
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v5, 31, v16
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, v16
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s3
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, s2
-; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x70
-; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v7, 31, v17
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v6, v17
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[4:7]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v26, 31, v19
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v24, 31, v18
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v23, v18
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v25, v19
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[23:26]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(12)
-; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v18, 31, v9
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v16, 31, v8
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v15, v8
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v17, v9
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x50
+; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v18, 31, v9
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v17, v9
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[15:18]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 48
 ; GCNX3-HSA-NEXT:    v_ashrrev_i32_e32 v14, 31, v1
@@ -3900,11 +3900,11 @@ define amdgpu_kernel void @global_zextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    s_add_u32 s8, s2, 48
 ; GCNX3-HSA-NEXT:    s_addc_u32 s9, s3, 0
 ; GCNX3-HSA-NEXT:    s_add_u32 s10, s2, 64
-; GCNX3-HSA-NEXT:    s_addc_u32 s11, s3, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
+; GCNX3-HSA-NEXT:    s_addc_u32 s11, s3, 0
+; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[32:35], v[0:1]
 ; GCNX3-HSA-NEXT:    s_add_u32 s12, s2, 0x50
-; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[28:31], v[0:1]
 ; GCNX3-HSA-NEXT:    s_addc_u32 s13, s3, 0
 ; GCNX3-HSA-NEXT:    s_add_u32 s14, s2, 0x60
 ; GCNX3-HSA-NEXT:    s_addc_u32 s15, s3, 0
@@ -3912,7 +3912,7 @@ define amdgpu_kernel void @global_zextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[32:35], v[0:1]
+; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[28:31], v[0:1]
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s14
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s15
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[24:27], v[0:1]
@@ -3927,8 +3927,8 @@ define amdgpu_kernel void @global_zextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[0:1]
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s5
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s7
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s6
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s7
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, 0
@@ -3936,78 +3936,79 @@ define amdgpu_kernel void @global_zextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(7)
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v28
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v29
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s1
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, s0
-; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[0:3]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s3
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v32
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v33
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v33, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v32, s0
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[0:3]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v32, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xe0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v30
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v31
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v31, s3
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v30, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v34
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v34, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xf0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v35
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v35, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[0:3]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s3
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, s2
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[0:3]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v32, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(8)
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v32
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v33
-; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[30:31], v[0:3]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v31, s3
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v30, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v28
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xd0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v29
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v33, s3
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v32, s2
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[34:35], v[0:3]
+; GCNX3-HSA-NEXT:    s_nop 0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v30
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v30, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xa0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v31
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v31, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v34
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v35
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v35, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v34, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0xb0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v35, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[0:3]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s3
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, s2
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[0:3]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v32, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x80
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v33, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(9)
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v24
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v25
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v24, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x90
-; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[30:31], v[0:3]
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[0:3]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v26
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v27
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v26, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x60
-; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[0:3]
-; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[30:31], v[0:3]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v27, s3
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(10)
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v20
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v21
+; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[34:35], v[0:3]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v31, s3
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, s2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v22
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v23
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v30, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x70
-; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[0:3]
-; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[0:3]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s3
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(11)
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v16
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v17
+; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[0:3]
 ; GCNX3-HSA-NEXT:    s_nop 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v18
@@ -4016,25 +4017,25 @@ define amdgpu_kernel void @global_zextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(12)
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v12
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v13
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v12, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 64
-; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[30:31], v[0:3]
-; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[0:3]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v14
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v15
+; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[0:3]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(12)
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v8
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v9
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x50
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v9
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s2
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v10
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v11
@@ -4043,13 +4044,13 @@ define amdgpu_kernel void @global_zextload_v32i32_to_v32i64(ptr addrspace(1) %ou
 ; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v4
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v5
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, v6
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, v7
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCNX3-HSA-NEXT:    s_endpgm
 ;
@@ -4396,37 +4397,37 @@ define amdgpu_kernel void @global_load_v32i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 16
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
-; GCNX3-HSA-NEXT:    s_add_u32 s6, s2, 48
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCNX3-HSA-NEXT:    s_addc_u32 s7, s3, 0
+; GCNX3-HSA-NEXT:    s_add_u32 s6, s2, 48
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v28, s4
+; GCNX3-HSA-NEXT:    s_addc_u32 s4, s3, 0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s4
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 32
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v29, s5
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s4
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 0x50
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v12, s4
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 64
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v17, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v16, s4
 ; GCNX3-HSA-NEXT:    s_add_u32 s4, s2, 0x70
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v17, s5
 ; GCNX3-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 0x60
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s6
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v20, s4
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v25, s3
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v21, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v24, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v25, s3
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[12:13]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[16:19], v[16:17]
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[20:23], v[20:21]
@@ -4434,17 +4435,17 @@ define amdgpu_kernel void @global_load_v32i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[28:31], v[28:29]
 ; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 0x60
 ; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCNX3-HSA-NEXT:    s_add_u32 s4, s0, 0x70
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v33, s1
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v32, s0
-; GCNX3-HSA-NEXT:    s_addc_u32 s5, s1, 0
+; GCNX3-HSA-NEXT:    s_add_u32 s4, s0, 0x70
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(7)
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[32:33], v[0:3]
 ; GCNX3-HSA-NEXT:    s_nop 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
+; GCNX3-HSA-NEXT:    s_addc_u32 s2, s1, 0
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCNX3-HSA-NEXT:    s_add_u32 s2, s0, 64
-; GCNX3-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCNX3-HSA-NEXT:    s_add_u32 s3, s0, 64
+; GCNX3-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCNX3-HSA-NEXT:    s_add_u32 s6, s0, 0x50
 ; GCNX3-HSA-NEXT:    s_addc_u32 s7, s1, 0
 ; GCNX3-HSA-NEXT:    s_add_u32 s8, s0, 32
@@ -4464,11 +4465,11 @@ define amdgpu_kernel void @global_load_v32i32(ptr addrspace(1) %out, ptr addrspa
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[8:11]
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v6, s6
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v7, s7
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v3, s5
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v3, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s3
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s1
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(7)
 ; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[12:15]
 ; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(7)
diff --git a/llvm/test/CodeGen/AMDGPU/load-global-i8.ll b/llvm/test/CodeGen/AMDGPU/load-global-i8.ll
index 443c062943648..e58edacb2d628 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-i8.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-i8.ll
@@ -1749,11 +1749,11 @@ define amdgpu_kernel void @global_zextload_v8i8_to_v8i32(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx2 v[8:9], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s0
+; GCN-HSA-NEXT:    s_addc_u32 s0, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v7, 24, v9
 ; GCN-HSA-NEXT:    v_bfe_u32 v5, v9, 8, 8
@@ -1901,11 +1901,11 @@ define amdgpu_kernel void @global_sextload_v8i8_to_v8i32(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx2 v[7:8], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s0
+; GCN-HSA-NEXT:    s_addc_u32 s0, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v3, 24, v7
 ; GCN-HSA-NEXT:    v_bfe_i32 v2, v7, 16, 8
@@ -2072,15 +2072,15 @@ define amdgpu_kernel void @global_zextload_v16i8_to_v16i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v10, 24, v3
@@ -2298,15 +2298,15 @@ define amdgpu_kernel void @global_sextload_v16i8_to_v16i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v11, 24, v3
@@ -2569,17 +2569,17 @@ define amdgpu_kernel void @global_zextload_v32i8_to_v32i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s0
@@ -2589,22 +2589,22 @@ define amdgpu_kernel void @global_zextload_v32i8_to_v32i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_and_b32_e32 v8, 0xff, v7
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v7, 16, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[8:11]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v10, 24, v6
 ; GCN-HSA-NEXT:    v_bfe_u32 v8, v6, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v7, 0xff, v6
 ; GCN-HSA-NEXT:    v_bfe_u32 v9, v6, 16, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[7:10]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x50
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v9, 24, v5
 ; GCN-HSA-NEXT:    v_bfe_u32 v7, v5, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v6, 0xff, v5
 ; GCN-HSA-NEXT:    v_bfe_u32 v8, v5, 16, 8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[6:9]
 ; GCN-HSA-NEXT:    v_and_b32_e32 v5, 0xff, v4
@@ -2624,8 +2624,8 @@ define amdgpu_kernel void @global_zextload_v32i8_to_v32i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_bfe_u32 v13, v2, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v12, 0xff, v2
 ; GCN-HSA-NEXT:    v_bfe_u32 v14, v2, 16, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 24, v0
 ; GCN-HSA-NEXT:    v_bfe_u32 v9, v0, 8, 8
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v6, 24, v1
@@ -2952,17 +2952,17 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s0
@@ -2972,22 +2972,22 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_bfe_i32 v8, v7, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v7, v7, 0, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[7:10]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v9, 24, v6
 ; GCN-HSA-NEXT:    v_bfe_i32 v8, v6, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v7, v6, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v6, v6, 0, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[6:9]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x50
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v8, 24, v5
 ; GCN-HSA-NEXT:    v_bfe_i32 v7, v5, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v6, v5, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v5, v5, 0, 8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[5:8]
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 64
@@ -3007,8 +3007,8 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_bfe_i32 v14, v2, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v13, v2, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v12, v2, 0, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s3
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v7, 24, v0
 ; GCN-HSA-NEXT:    v_bfe_i32 v6, v0, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v5, v0, 8, 8
@@ -3444,8 +3444,8 @@ define amdgpu_kernel void @global_zextload_v64i8_to_v64i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[12:15], v[4:5]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[8:11], v[4:5]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
@@ -3466,21 +3466,21 @@ define amdgpu_kernel void @global_zextload_v64i8_to_v64i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    s_addc_u32 s15, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s16, s0, 0xa0
 ; GCN-HSA-NEXT:    s_addc_u32 s17, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s17
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s17
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v19, 24, v2
 ; GCN-HSA-NEXT:    v_bfe_u32 v17, v2, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v16, 0xff, v2
 ; GCN-HSA-NEXT:    v_bfe_u32 v18, v2, 16, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[16:19]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s2
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v19, 24, v14
 ; GCN-HSA-NEXT:    v_bfe_u32 v17, v14, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v16, 0xff, v14
 ; GCN-HSA-NEXT:    v_bfe_u32 v18, v14, 16, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[16:19]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v19, 24, v15
@@ -3502,49 +3502,49 @@ define amdgpu_kernel void @global_zextload_v64i8_to_v64i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_and_b32_e32 v14, 0xff, v13
 ; GCN-HSA-NEXT:    v_bfe_u32 v16, v13, 16, 8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s9
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[14:17]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(6)
 ; GCN-HSA-NEXT:    v_bfe_u32 v13, v10, 8, 8
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v15, 24, v10
 ; GCN-HSA-NEXT:    v_and_b32_e32 v12, 0xff, v10
 ; GCN-HSA-NEXT:    v_bfe_u32 v14, v10, 16, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s11
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xb0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s10
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s11
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[12:15]
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xb0
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v15, 24, v11
 ; GCN-HSA-NEXT:    v_bfe_u32 v13, v11, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v12, 0xff, v11
 ; GCN-HSA-NEXT:    v_bfe_u32 v14, v11, 16, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s13
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s13
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[12:15]
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 24, v8
 ; GCN-HSA-NEXT:    v_bfe_u32 v11, v8, 8, 8
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 24, v8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v10, 0xff, v8
 ; GCN-HSA-NEXT:    v_bfe_u32 v12, v8, 16, 8
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[10:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x80
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s15
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[10:13]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s15
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 24, v9
 ; GCN-HSA-NEXT:    v_bfe_u32 v11, v9, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v10, 0xff, v9
 ; GCN-HSA-NEXT:    v_bfe_u32 v12, v9, 16, 8
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x90
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[10:13]
 ; GCN-HSA-NEXT:    v_bfe_u32 v9, v3, 8, 8
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 24, v3
 ; GCN-HSA-NEXT:    v_and_b32_e32 v8, 0xff, v3
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v3, 16, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x90
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[8:11]
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v20, 24, v1
@@ -3552,36 +3552,36 @@ define amdgpu_kernel void @global_zextload_v64i8_to_v64i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_bfe_u32 v9, v0, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v8, 0xff, v0
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v0, 16, 8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
 ; GCN-HSA-NEXT:    v_bfe_u32 v18, v1, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v17, 0xff, v1
 ; GCN-HSA-NEXT:    v_bfe_u32 v19, v1, 16, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    s_waitcnt vmcnt(10)
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v16, 24, v5
-; GCN-HSA-NEXT:    v_bfe_u32 v14, v5, 8, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[8:11]
-; GCN-HSA-NEXT:    v_and_b32_e32 v13, 0xff, v5
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(11)
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v16, 24, v5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 24, v4
 ; GCN-HSA-NEXT:    v_bfe_u32 v9, v4, 8, 8
-; GCN-HSA-NEXT:    v_bfe_u32 v15, v5, 16, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v8, 0xff, v4
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v4, 16, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    v_bfe_u32 v14, v5, 8, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[17:20]
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    v_and_b32_e32 v13, 0xff, v5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v20, 24, v6
 ; GCN-HSA-NEXT:    v_bfe_u32 v18, v6, 8, 8
+; GCN-HSA-NEXT:    v_bfe_u32 v15, v5, 16, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v17, 0xff, v6
 ; GCN-HSA-NEXT:    v_bfe_u32 v19, v6, 16, 8
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[17:20]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[17:20]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v3, 24, v7
 ; GCN-HSA-NEXT:    v_bfe_u32 v1, v7, 8, 8
@@ -4191,9 +4191,9 @@ define amdgpu_kernel void @global_sextload_v64i8_to_v64i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v19, 24, v14
@@ -4221,13 +4221,13 @@ define amdgpu_kernel void @global_sextload_v64i8_to_v64i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xe0
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xf0
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v16, 24, v13
 ; GCN-HSA-NEXT:    v_bfe_i32 v15, v13, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v14, v13, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v13, v13, 0, 8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[13:16]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s3
@@ -4239,10 +4239,10 @@ define amdgpu_kernel void @global_sextload_v64i8_to_v64i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_bfe_i32 v14, v10, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v13, v10, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v12, v10, 0, 8
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[19:20], v[12:15]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xd0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[19:20], v[12:15]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v14, 24, v11
 ; GCN-HSA-NEXT:    v_bfe_i32 v13, v11, 16, 8
@@ -4257,60 +4257,59 @@ define amdgpu_kernel void @global_sextload_v64i8_to_v64i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_bfe_i32 v12, v8, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v11, v8, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v10, v8, 0, 8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s2
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xb0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[10:13]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s2
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v12, 24, v9
 ; GCN-HSA-NEXT:    v_bfe_i32 v11, v9, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v10, v9, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v9, v9, 0, 8
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xb0
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[9:12]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(9)
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v14, 24, v6
 ; GCN-HSA-NEXT:    v_bfe_i32 v13, v6, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v12, v6, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v11, v6, 0, 8
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[11:14]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x80
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v10, 24, v7
 ; GCN-HSA-NEXT:    v_bfe_i32 v9, v7, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v8, v7, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v7, v7, 0, 8
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[11:14]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[7:10]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x90
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v9, 24, v4
 ; GCN-HSA-NEXT:    v_bfe_i32 v8, v4, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v7, v4, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v6, v4, 0, 8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    s_waitcnt vmcnt(10)
-; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v14, 24, v1
-; GCN-HSA-NEXT:    v_bfe_i32 v13, v1, 16, 8
-; GCN-HSA-NEXT:    v_bfe_i32 v12, v1, 8, 8
-; GCN-HSA-NEXT:    v_bfe_i32 v11, v1, 0, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[6:9]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(11)
+; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v14, 24, v1
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v9, 24, v0
 ; GCN-HSA-NEXT:    v_bfe_i32 v8, v0, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v7, v0, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v6, v0, 0, 8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_bfe_i32 v13, v1, 16, 8
+; GCN-HSA-NEXT:    v_bfe_i32 v12, v1, 8, 8
+; GCN-HSA-NEXT:    v_bfe_i32 v11, v1, 0, 8
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v18, 24, v5
 ; GCN-HSA-NEXT:    v_bfe_i32 v17, v5, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v16, v5, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v15, v5, 0, 8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[15:18]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
 ; GCN-HSA-NEXT:    v_ashrrev_i32_e32 v18, 24, v3
 ; GCN-HSA-NEXT:    v_bfe_i32 v17, v3, 16, 8
@@ -4320,11 +4319,12 @@ define amdgpu_kernel void @global_sextload_v64i8_to_v64i32(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_bfe_i32 v4, v2, 16, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v3, v2, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_i32 v2, v2, 0, 8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 0x50
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[15:18]
@@ -5562,14 +5562,14 @@ define amdgpu_kernel void @global_zextload_v4i8_to_v4i64(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    flat_load_dword v0, v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, 0
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s0
+; GCN-HSA-NEXT:    s_addc_u32 s0, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v7, v1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_bfe_u32 v6, v0, 8, 8
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v2, 24, v0
@@ -5713,11 +5713,11 @@ define amdgpu_kernel void @global_sextload_v4i8_to_v4i64(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    flat_load_dword v0, v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s0
+; GCN-HSA-NEXT:    s_addc_u32 s0, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v3, 16, v0
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 24, v0
@@ -5883,10 +5883,10 @@ define amdgpu_kernel void @global_zextload_v8i8_to_v8i64(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, v1
@@ -5898,16 +5898,16 @@ define amdgpu_kernel void @global_zextload_v8i8_to_v8i64(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v2, 24, v16
 ; GCN-HSA-NEXT:    v_bfe_u32 v0, v16, 16, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v5, 24, v15
 ; GCN-HSA-NEXT:    v_bfe_u32 v3, v15, 16, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s1
 ; GCN-HSA-NEXT:    v_bfe_u32 v9, v16, 8, 8
 ; GCN-HSA-NEXT:    v_bfe_u32 v13, v15, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v11, 0xff, v15
@@ -6147,9 +6147,9 @@ define amdgpu_kernel void @global_sextload_v8i8_to_v8i64(ptr addrspace(1) %out,
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s18
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s15
@@ -6411,11 +6411,11 @@ define amdgpu_kernel void @global_zextload_v16i8_to_v16i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x70
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v9, v5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, v5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, v5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, v5
@@ -6427,16 +6427,16 @@ define amdgpu_kernel void @global_zextload_v16i8_to_v16i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_bfe_u32 v8, v3, 16, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[8:11]
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 24, v0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s5
 ; GCN-HSA-NEXT:    v_bfe_u32 v11, v0, 16, 8
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[11:14]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, v5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 64
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v16, 24, v1
 ; GCN-HSA-NEXT:    v_bfe_u32 v14, v1, 16, 8
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
@@ -6444,13 +6444,13 @@ define amdgpu_kernel void @global_zextload_v16i8_to_v16i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, v5
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[14:17]
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v2, 8, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s5
 ; GCN-HSA-NEXT:    v_and_b32_e32 v8, 0xff, v2
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[8:11]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s3
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v6, 24, v2
 ; GCN-HSA-NEXT:    v_bfe_u32 v4, v2, 16, 8
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
@@ -6463,8 +6463,8 @@ define amdgpu_kernel void @global_zextload_v16i8_to_v16i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    v_bfe_u32 v13, v3, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v11, 0xff, v3
@@ -6862,20 +6862,19 @@ define amdgpu_kernel void @global_sextload_v16i8_to_v16i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[2:5]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s15
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[2:5]
@@ -6883,16 +6882,17 @@ define amdgpu_kernel void @global_sextload_v16i8_to_v16i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s17
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s18
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s19
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 16
@@ -6904,16 +6904,16 @@ define amdgpu_kernel void @global_sextload_v16i8_to_v16i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s9
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[4:7]
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[12:15]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s12
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s13
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s24
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s25
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s10
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s11
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -7320,76 +7320,76 @@ define amdgpu_kernel void @global_zextload_v32i8_to_v32i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x50
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, v1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v17, v1
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v12, 24, v4
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v4, 16, 8
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v12, 24, v3
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v3, 16, 8
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0xf0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v12, 24, v2
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v2, 16, 8
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0xd0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(3)
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v12, 24, v9
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v9, 16, 8
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0xb0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v12, 24, v8
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v8, 16, 8
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x90
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v12, 24, v7
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v7, 16, 8
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v12, 24, v6
 ; GCN-HSA-NEXT:    v_bfe_u32 v10, v6, 16, 8
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 64
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_bfe_u32 v12, v5, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v10, 0xff, v5
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_bfe_u32 v12, v4, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v10, 0xff, v4
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[10:13]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s5
 ; GCN-HSA-NEXT:    v_bfe_u32 v12, v3, 8, 8
 ; GCN-HSA-NEXT:    v_and_b32_e32 v10, 0xff, v3
 ; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0xe0
@@ -7406,19 +7406,19 @@ define amdgpu_kernel void @global_zextload_v32i8_to_v32i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[9:12]
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v2, 24, v5
 ; GCN-HSA-NEXT:    v_bfe_u32 v0, v5, 16, 8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_bfe_u32 v16, v8, 8, 8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xa0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 0x80
 ; GCN-HSA-NEXT:    v_and_b32_e32 v14, 0xff, v8
@@ -8051,7 +8051,7 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    s_mov_b32 s9, s5
 ; GCN-HSA-NEXT:    s_mov_b32 s3, s5
 ; GCN-HSA-NEXT:    s_mov_b32 s7, s5
-; GCN-HSA-NEXT:    s_mov_b32 s31, s5
+; GCN-HSA-NEXT:    s_mov_b32 s39, s5
 ; GCN-HSA-NEXT:    s_mov_b32 s27, s5
 ; GCN-HSA-NEXT:    s_mov_b32 s15, s5
 ; GCN-HSA-NEXT:    s_mov_b32 s11, s5
@@ -8063,62 +8063,62 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    s_lshr_b32 s4, s17, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s8, s17, 8
 ; GCN-HSA-NEXT:    s_mov_b32 s14, s17
-; GCN-HSA-NEXT:    s_lshr_b32 s2, s13, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s6, s13, 8
 ; GCN-HSA-NEXT:    s_mov_b32 s10, s13
 ; GCN-HSA-NEXT:    s_lshr_b32 s18, s12, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s22, s12, 24
+; GCN-HSA-NEXT:    s_lshr_b32 s20, s12, 24
 ; GCN-HSA-NEXT:    s_lshr_b32 s24, s12, 8
-; GCN-HSA-NEXT:    s_bfe_i64 s[20:21], s[12:13], 0x80000
+; GCN-HSA-NEXT:    s_bfe_i64 s[22:23], s[12:13], 0x80000
 ; GCN-HSA-NEXT:    s_lshr_b32 s12, s16, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s26, s16, 24
-; GCN-HSA-NEXT:    s_lshr_b32 s30, s16, 8
+; GCN-HSA-NEXT:    s_lshr_b32 s34, s16, 8
 ; GCN-HSA-NEXT:    s_bfe_i64 s[4:5], s[4:5], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[8:9], s[8:9], 0x80000
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s38, v0
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s39, v1
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s48, v2
-; GCN-HSA-NEXT:    v_readfirstlane_b32 s49, v3
-; GCN-HSA-NEXT:    s_ashr_i32 s59, s17, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s60, s17, 24
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s48, v0
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s49, v1
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s50, v2
+; GCN-HSA-NEXT:    v_readfirstlane_b32 s51, v3
+; GCN-HSA-NEXT:    s_lshr_b32 s2, s13, 16
+; GCN-HSA-NEXT:    s_ashr_i32 s61, s17, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s62, s17, 24
 ; GCN-HSA-NEXT:    s_bfe_i64 s[46:47], s[16:17], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[16:17], s[14:15], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[28:29], s[24:25], 0x80000
-; GCN-HSA-NEXT:    s_bfe_i64 s[34:35], s[22:23], 0x80000
+; GCN-HSA-NEXT:    s_bfe_i64 s[30:31], s[20:21], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[36:37], s[18:19], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[42:43], s[10:11], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[40:41], s[6:7], 0x80000
-; GCN-HSA-NEXT:    s_bfe_i64 s[50:51], s[2:3], 0x80000
-; GCN-HSA-NEXT:    s_bfe_i64 s[52:53], s[30:31], 0x80000
-; GCN-HSA-NEXT:    s_bfe_i64 s[54:55], s[26:27], 0x80000
-; GCN-HSA-NEXT:    s_bfe_i64 s[56:57], s[12:13], 0x80000
+; GCN-HSA-NEXT:    s_bfe_i64 s[54:55], s[34:35], 0x80000
+; GCN-HSA-NEXT:    s_bfe_i64 s[56:57], s[26:27], 0x80000
+; GCN-HSA-NEXT:    s_bfe_i64 s[58:59], s[12:13], 0x80000
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s8
-; GCN-HSA-NEXT:    s_lshr_b32 s30, s49, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s26, s49, 8
-; GCN-HSA-NEXT:    s_lshr_b32 s14, s39, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s10, s39, 8
-; GCN-HSA-NEXT:    s_mov_b32 s12, s39
-; GCN-HSA-NEXT:    s_lshr_b32 s8, s38, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s6, s38, 24
-; GCN-HSA-NEXT:    s_lshr_b32 s4, s38, 8
-; GCN-HSA-NEXT:    s_bfe_i64 s[2:3], s[38:39], 0x80000
-; GCN-HSA-NEXT:    s_mov_b32 s38, s49
-; GCN-HSA-NEXT:    s_lshr_b32 s24, s48, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s22, s48, 24
-; GCN-HSA-NEXT:    s_lshr_b32 s18, s48, 8
+; GCN-HSA-NEXT:    s_lshr_b32 s38, s51, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s26, s51, 8
+; GCN-HSA-NEXT:    s_lshr_b32 s14, s49, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s10, s49, 8
+; GCN-HSA-NEXT:    s_mov_b32 s12, s49
+; GCN-HSA-NEXT:    s_lshr_b32 s8, s48, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s6, s48, 24
+; GCN-HSA-NEXT:    s_lshr_b32 s4, s48, 8
+; GCN-HSA-NEXT:    s_mov_b32 s34, s51
+; GCN-HSA-NEXT:    s_lshr_b32 s24, s50, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s20, s50, 24
+; GCN-HSA-NEXT:    s_lshr_b32 s18, s50, 8
 ; GCN-HSA-NEXT:    s_ashr_i32 s45, s13, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s58, s13, 24
+; GCN-HSA-NEXT:    s_ashr_i32 s60, s13, 24
+; GCN-HSA-NEXT:    s_bfe_i64 s[52:53], s[2:3], 0x80000
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s17
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s9
-; GCN-HSA-NEXT:    s_ashr_i32 s33, s39, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s44, s39, 24
-; GCN-HSA-NEXT:    s_ashr_i32 s61, s49, 31
-; GCN-HSA-NEXT:    s_ashr_i32 s62, s49, 24
-; GCN-HSA-NEXT:    s_bfe_i64 s[16:17], s[48:49], 0x80000
+; GCN-HSA-NEXT:    s_ashr_i32 s33, s49, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s44, s49, 24
+; GCN-HSA-NEXT:    s_bfe_i64 s[2:3], s[48:49], 0x80000
+; GCN-HSA-NEXT:    s_ashr_i32 s48, s51, 31
+; GCN-HSA-NEXT:    s_ashr_i32 s49, s51, 24
+; GCN-HSA-NEXT:    s_bfe_i64 s[16:17], s[50:51], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[4:5], s[4:5], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[6:7], s[6:7], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[8:9], s[8:9], 0x80000
@@ -8126,21 +8126,21 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    s_bfe_i64 s[10:11], s[10:11], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[14:15], s[14:15], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[18:19], s[18:19], 0x80000
-; GCN-HSA-NEXT:    s_bfe_i64 s[22:23], s[22:23], 0x80000
+; GCN-HSA-NEXT:    s_bfe_i64 s[20:21], s[20:21], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[24:25], s[24:25], 0x80000
-; GCN-HSA-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x80000
+; GCN-HSA-NEXT:    s_bfe_i64 s[34:35], s[34:35], 0x80000
 ; GCN-HSA-NEXT:    s_bfe_i64 s[26:27], s[26:27], 0x80000
-; GCN-HSA-NEXT:    s_bfe_i64 s[30:31], s[30:31], 0x80000
-; GCN-HSA-NEXT:    s_add_u32 s48, s0, 0x70
-; GCN-HSA-NEXT:    s_addc_u32 s49, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s49
-; GCN-HSA-NEXT:    s_add_u32 s48, s0, 0x60
-; GCN-HSA-NEXT:    s_addc_u32 s49, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s49
-; GCN-HSA-NEXT:    s_add_u32 s48, s0, 0x50
-; GCN-HSA-NEXT:    s_addc_u32 s49, s1, 0
+; GCN-HSA-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x80000
+; GCN-HSA-NEXT:    s_add_u32 s50, s0, 0x70
+; GCN-HSA-NEXT:    s_addc_u32 s51, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s50
+; GCN-HSA-NEXT:    s_add_u32 s50, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s51
+; GCN-HSA-NEXT:    s_addc_u32 s51, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s50
+; GCN-HSA-NEXT:    s_add_u32 s50, s0, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s51
+; GCN-HSA-NEXT:    s_addc_u32 s51, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s46
 ; GCN-HSA-NEXT:    s_add_u32 s46, s0, 64
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s47
@@ -8150,75 +8150,75 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s43
 ; GCN-HSA-NEXT:    s_addc_u32 s43, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s62
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s36
 ; GCN-HSA-NEXT:    s_add_u32 s36, s0, 32
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s59
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s61
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s37
 ; GCN-HSA-NEXT:    s_addc_u32 s37, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[4:7]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s50
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s20
-; GCN-HSA-NEXT:    s_add_u32 s20, s0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s51
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s52
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s22
+; GCN-HSA-NEXT:    s_add_u32 s22, s0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s53
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v30, s42
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v31, s43
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s58
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s60
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s45
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s21
-; GCN-HSA-NEXT:    s_addc_u32 s21, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s23
+; GCN-HSA-NEXT:    s_addc_u32 s23, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[30:31], v[16:19]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s30
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s22
+; GCN-HSA-NEXT:    s_add_u32 s22, s0, 0xf0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s23
+; GCN-HSA-NEXT:    s_addc_u32 s23, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s22
+; GCN-HSA-NEXT:    s_add_u32 s22, s0, 0xe0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s31
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s23
+; GCN-HSA-NEXT:    s_addc_u32 s23, s1, 0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s58
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s20
+; GCN-HSA-NEXT:    s_add_u32 s20, s0, 0xd0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s59
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s56
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s57
+; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s51
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s54
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s55
+; GCN-HSA-NEXT:    v_mov_b32_e32 v28, s46
+; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s47
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s40
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s21
-; GCN-HSA-NEXT:    s_add_u32 s20, s0, 0xf0
-; GCN-HSA-NEXT:    s_addc_u32 s21, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s21
-; GCN-HSA-NEXT:    s_add_u32 s20, s0, 0xe0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s41
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v24, s36
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v25, s37
-; GCN-HSA-NEXT:    s_addc_u32 s21, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s34
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s35
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[20:23]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s56
-; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s21
-; GCN-HSA-NEXT:    s_add_u32 s20, s0, 0xd0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s57
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s54
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s55
-; GCN-HSA-NEXT:    v_mov_b32_e32 v26, s48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v27, s49
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s52
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s53
-; GCN-HSA-NEXT:    v_mov_b32_e32 v28, s46
-; GCN-HSA-NEXT:    v_mov_b32_e32 v29, s47
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s28
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s29
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
-; GCN-HSA-NEXT:    s_addc_u32 s21, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s21
+; GCN-HSA-NEXT:    s_addc_u32 s21, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[26:27], v[8:11]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[28:29], v[12:15]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s30
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s31
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s62
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s61
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s38
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s39
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s38
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s39
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s49
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s34
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s35
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[24:25], v[20:23]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s26
+; GCN-HSA-NEXT:    v_mov_b32_e32 v20, s22
+; GCN-HSA-NEXT:    v_mov_b32_e32 v21, s23
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s27
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[8:11]
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[20:21], v[12:15]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s24
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s25
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s22
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s23
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s20
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s21
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
@@ -8264,12 +8264,12 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i64(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
@@ -10541,8 +10541,8 @@ define amdgpu_kernel void @global_zextload_v16i8_to_v16i16(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s6
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s3
@@ -11032,8 +11032,8 @@ define amdgpu_kernel void @global_sextload_v16i8_to_v16i16(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s7
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s5
@@ -11658,23 +11658,23 @@ define amdgpu_kernel void @global_zextload_v32i8_to_v32i16(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    s_or_b32 s7, s16, s7
 ; GCN-HSA-NEXT:    s_or_b32 s9, s17, s9
 ; GCN-HSA-NEXT:    s_and_b32 s4, s4, 0xff00ff
-; GCN-HSA-NEXT:    s_and_b32 s10, s2, 0xff00ff
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s3
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s2
+; GCN-HSA-NEXT:    s_and_b32 s2, s2, 0xff00ff
+; GCN-HSA-NEXT:    s_add_u32 s10, s0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s2
+; GCN-HSA-NEXT:    s_addc_u32 s2, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s10
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s9
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s10
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s4
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[4:7]
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s6
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s8
@@ -12469,8 +12469,8 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i16(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(1)
@@ -12567,11 +12567,11 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i16(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s15
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
@@ -12582,20 +12582,20 @@ define amdgpu_kernel void @global_sextload_v32i8_to_v32i16(ptr addrspace(1) %out
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s14
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s13
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s11
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s10
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s7
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s21
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s22
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/local-stack-alloc-block-sp-reference.ll b/llvm/test/CodeGen/AMDGPU/local-stack-alloc-block-sp-reference.ll
index 1b624decceed6..f8bca2491a809 100644
--- a/llvm/test/CodeGen/AMDGPU/local-stack-alloc-block-sp-reference.ll
+++ b/llvm/test/CodeGen/AMDGPU/local-stack-alloc-block-sp-reference.ll
@@ -261,22 +261,18 @@ define amdgpu_kernel void @local_stack_offset_uses_sp(ptr addrspace(1) %out) #1
 ; FLATSCR-NEXT:    s_cbranch_scc1 .LBB0_1
 ; FLATSCR-NEXT:  ; %bb.2: ; %static-memset-post-expansion
 ; FLATSCR-NEXT:    s_movk_i32 s0, 0x2000
-; FLATSCR-NEXT:    s_add_i32 s6, s0, 0x3000
-; FLATSCR-NEXT:    s_mov_b32 s0, 0
-; FLATSCR-NEXT:    s_mov_b32 s1, s0
-; FLATSCR-NEXT:    s_mov_b32 s2, s0
-; FLATSCR-NEXT:    s_mov_b32 s3, s0
-; FLATSCR-NEXT:    v_mov_b32_e32 v0, s0
-; FLATSCR-NEXT:    v_mov_b32_e32 v1, s1
-; FLATSCR-NEXT:    v_mov_b32_e32 v2, s2
-; FLATSCR-NEXT:    v_mov_b32_e32 v3, s3
-; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[0:3], s6 offset:256
+; FLATSCR-NEXT:    v_mov_b32_e32 v0, 0
+; FLATSCR-NEXT:    s_addk_i32 s0, 0x3000
+; FLATSCR-NEXT:    v_mov_b32_e32 v1, v0
+; FLATSCR-NEXT:    v_mov_b32_e32 v2, v0
+; FLATSCR-NEXT:    v_mov_b32_e32 v3, v0
+; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:256
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
-; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[0:3], s6 offset:272
+; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:272
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
-; FLATSCR-NEXT:    s_movk_i32 s0, 0x3000
-; FLATSCR-NEXT:    scratch_load_dwordx2 v[0:1], off, s6 offset:208 glc
+; FLATSCR-NEXT:    scratch_load_dwordx2 v[0:1], off, s0 offset:208 glc
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; FLATSCR-NEXT:    s_movk_i32 s0, 0x3000
 ; FLATSCR-NEXT:    scratch_load_dwordx2 v[2:3], off, s0 offset:64 glc
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
 ; FLATSCR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
@@ -566,23 +562,19 @@ define void @func_local_stack_offset_uses_sp(ptr addrspace(1) %out) #1 {
 ; FLATSCR-NEXT:    s_cbranch_scc1 .LBB1_1
 ; FLATSCR-NEXT:  ; %bb.2: ; %static-memset-post-expansion
 ; FLATSCR-NEXT:    s_movk_i32 s0, 0x2000
-; FLATSCR-NEXT:    s_add_i32 s4, s33, s0
-; FLATSCR-NEXT:    s_mov_b32 s0, 0
-; FLATSCR-NEXT:    s_mov_b32 s1, s0
-; FLATSCR-NEXT:    s_mov_b32 s2, s0
-; FLATSCR-NEXT:    s_mov_b32 s3, s0
-; FLATSCR-NEXT:    s_addk_i32 s4, 0x3000
-; FLATSCR-NEXT:    v_mov_b32_e32 v5, s3
-; FLATSCR-NEXT:    v_mov_b32_e32 v4, s2
-; FLATSCR-NEXT:    v_mov_b32_e32 v3, s1
-; FLATSCR-NEXT:    v_mov_b32_e32 v2, s0
-; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[2:5], s4 offset:256
+; FLATSCR-NEXT:    s_add_i32 s1, s33, s0
+; FLATSCR-NEXT:    v_mov_b32_e32 v2, 0
+; FLATSCR-NEXT:    s_add_i32 s0, s1, 0x3000
+; FLATSCR-NEXT:    v_mov_b32_e32 v3, v2
+; FLATSCR-NEXT:    v_mov_b32_e32 v4, v2
+; FLATSCR-NEXT:    v_mov_b32_e32 v5, v2
+; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[2:5], s0 offset:256
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
-; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[2:5], s4 offset:272
+; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[2:5], s0 offset:272
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
-; FLATSCR-NEXT:    s_add_i32 s0, s33, 0x3000
-; FLATSCR-NEXT:    scratch_load_dwordx2 v[2:3], off, s4 offset:208 glc
+; FLATSCR-NEXT:    scratch_load_dwordx2 v[2:3], off, s0 offset:208 glc
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; FLATSCR-NEXT:    s_add_i32 s0, s33, 0x3000
 ; FLATSCR-NEXT:    scratch_load_dwordx2 v[4:5], off, s0 offset:64 glc
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
 ; FLATSCR-NEXT:    s_mov_b32 s32, s34
@@ -893,20 +885,16 @@ define amdgpu_kernel void @local_stack_offset_uses_sp_flat(ptr addrspace(1) %out
 ; FLATSCR-NEXT:    s_cbranch_scc1 .LBB2_1
 ; FLATSCR-NEXT:  ; %bb.2: ; %static-memset-post-expansion
 ; FLATSCR-NEXT:    s_movk_i32 s0, 0x2000
-; FLATSCR-NEXT:    s_add_i32 s6, s0, 0x4000
-; FLATSCR-NEXT:    s_mov_b32 s0, 0
-; FLATSCR-NEXT:    s_mov_b32 s1, s0
-; FLATSCR-NEXT:    s_mov_b32 s2, s0
-; FLATSCR-NEXT:    s_mov_b32 s3, s0
-; FLATSCR-NEXT:    v_mov_b32_e32 v0, s0
-; FLATSCR-NEXT:    v_mov_b32_e32 v1, s1
-; FLATSCR-NEXT:    v_mov_b32_e32 v2, s2
-; FLATSCR-NEXT:    v_mov_b32_e32 v3, s3
-; FLATSCR-NEXT:    s_movk_i32 s0, 0x1000
-; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[0:3], s6 offset:256
+; FLATSCR-NEXT:    v_mov_b32_e32 v0, 0
+; FLATSCR-NEXT:    s_addk_i32 s0, 0x4000
+; FLATSCR-NEXT:    v_mov_b32_e32 v1, v0
+; FLATSCR-NEXT:    v_mov_b32_e32 v2, v0
+; FLATSCR-NEXT:    v_mov_b32_e32 v3, v0
+; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:256
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
-; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[0:3], s6 offset:272
+; FLATSCR-NEXT:    scratch_store_dwordx4 off, v[0:3], s0 offset:272
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; FLATSCR-NEXT:    s_movk_i32 s0, 0x1000
 ; FLATSCR-NEXT:    s_addk_i32 s0, 0x4000
 ; FLATSCR-NEXT:    scratch_load_dwordx2 v[8:9], off, s0 offset:720 glc
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics-hsa.ll b/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics-hsa.ll
index 44ff13dd81303..34faa39511435 100644
--- a/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics-hsa.ll
+++ b/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics-hsa.ll
@@ -27,10 +27,9 @@ define amdgpu_kernel void @workgroup_ids_kernel() {
 ;
 ; GFX9ARCH-GISEL-LABEL: workgroup_ids_kernel:
 ; GFX9ARCH-GISEL:       ; %bb.0: ; %.entry
-; GFX9ARCH-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX9ARCH-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX9ARCH-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v0, ttmp9
 ; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX9ARCH-GISEL-NEXT:    buffer_store_dwordx3 v[0:2], off, s[0:3], 0
@@ -47,10 +46,9 @@ define amdgpu_kernel void @workgroup_ids_kernel() {
 ;
 ; GFX12-GISEL-LABEL: workgroup_ids_kernel:
 ; GFX12-GISEL:       ; %bb.0: ; %.entry
-; GFX12-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX12-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX12-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, ttmp9 :: v_dual_mov_b32 v1, s1
 ; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX12-GISEL-NEXT:    buffer_store_b96 v[0:2], off, s[0:3], null
 ; GFX12-GISEL-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics-pal.ll b/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics-pal.ll
index b2ee9119fbcad..d017e49d36484 100644
--- a/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics-pal.ll
+++ b/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics-pal.ll
@@ -34,10 +34,9 @@ define amdgpu_cs void @_amdgpu_cs_main() {
 ;
 ; GFX9ARCH-GISEL-LABEL: _amdgpu_cs_main:
 ; GFX9ARCH-GISEL:       ; %bb.0: ; %.entry
-; GFX9ARCH-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX9ARCH-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX9ARCH-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v0, ttmp9
 ; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX9ARCH-GISEL-NEXT:    buffer_store_dwordx3 v[0:2], off, s[0:3], 0
@@ -54,10 +53,9 @@ define amdgpu_cs void @_amdgpu_cs_main() {
 ;
 ; GFX12-GISEL-LABEL: _amdgpu_cs_main:
 ; GFX12-GISEL:       ; %bb.0: ; %.entry
-; GFX12-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX12-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX12-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, ttmp9 :: v_dual_mov_b32 v1, s1
 ; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX12-GISEL-NEXT:    buffer_store_b96 v[0:2], off, s[0:3], null
 ; GFX12-GISEL-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics.ll b/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics.ll
index 79ba64c22abeb..a2683b94b5a5b 100644
--- a/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics.ll
+++ b/llvm/test/CodeGen/AMDGPU/lower-work-group-id-intrinsics.ll
@@ -19,10 +19,9 @@ define amdgpu_cs void @_amdgpu_cs_main() {
 ;
 ; GFX9-GISEL-LABEL: _amdgpu_cs_main:
 ; GFX9-GISEL:       ; %bb.0: ; %.entry
-; GFX9-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX9-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX9-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, ttmp9
 ; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX9-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX9-GISEL-NEXT:    buffer_store_dwordx3 v[0:2], off, s[0:3], 0
@@ -39,10 +38,9 @@ define amdgpu_cs void @_amdgpu_cs_main() {
 ;
 ; GFX12-GISEL-LABEL: _amdgpu_cs_main:
 ; GFX12-GISEL:       ; %bb.0: ; %.entry
-; GFX12-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX12-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX12-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, ttmp9 :: v_dual_mov_b32 v1, s1
 ; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX12-GISEL-NEXT:    buffer_store_b96 v[0:2], off, s[0:3], null
 ; GFX12-GISEL-NEXT:    s_endpgm
@@ -110,11 +108,11 @@ define amdgpu_cs void @_amdgpu_cs_main() {
 ; GFX1250-GISEL-NEXT:    s_add_co_i32 s3, s3, 1
 ; GFX1250-GISEL-NEXT:    s_bfe_u32 s5, ttmp6, 0x40008
 ; GFX1250-GISEL-NEXT:    s_mul_i32 s3, s4, s3
-; GFX1250-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX1250-GISEL-NEXT:    s_add_co_i32 s5, s5, s3
 ; GFX1250-GISEL-NEXT:    s_cmp_eq_u32 s2, 0
 ; GFX1250-GISEL-NEXT:    s_cselect_b32 s2, s4, s5
-; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX1250-GISEL-NEXT:    buffer_store_b96 v[0:2], off, s[0:3], null
 ; GFX1250-GISEL-NEXT:    s_endpgm
@@ -142,10 +140,9 @@ define amdgpu_cs void @workgroup_id_no_clusters() "amdgpu-cluster-dims"="0,0,0"
 ;
 ; GFX9-GISEL-LABEL: workgroup_id_no_clusters:
 ; GFX9-GISEL:       ; %bb.0: ; %.entry
-; GFX9-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX9-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX9-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, ttmp9
 ; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX9-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX9-GISEL-NEXT:    buffer_store_dwordx3 v[0:2], off, s[0:3], 0
@@ -162,10 +159,9 @@ define amdgpu_cs void @workgroup_id_no_clusters() "amdgpu-cluster-dims"="0,0,0"
 ;
 ; GFX12-GISEL-LABEL: workgroup_id_no_clusters:
 ; GFX12-GISEL:       ; %bb.0: ; %.entry
-; GFX12-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX12-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX12-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, ttmp9 :: v_dual_mov_b32 v1, s1
 ; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX12-GISEL-NEXT:    buffer_store_b96 v[0:2], off, s[0:3], null
 ; GFX12-GISEL-NEXT:    s_endpgm
@@ -189,10 +185,9 @@ define amdgpu_cs void @workgroup_id_no_clusters() "amdgpu-cluster-dims"="0,0,0"
 ; GFX1250-GISEL-NEXT:    v_nop
 ; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX1250-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX1250-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v0, ttmp9 :: v_dual_mov_b32 v1, s1
 ; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX1250-GISEL-NEXT:    buffer_store_b96 v[0:2], off, s[0:3], null
 ; GFX1250-GISEL-NEXT:    s_endpgm
@@ -220,10 +215,9 @@ define amdgpu_cs void @workgroup_id_optimized() "amdgpu-cluster-dims"="2,3,4" {
 ;
 ; GFX9-GISEL-LABEL: workgroup_id_optimized:
 ; GFX9-GISEL:       ; %bb.0: ; %.entry
-; GFX9-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX9-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX9-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-GISEL-NEXT:    v_mov_b32_e32 v0, ttmp9
 ; GFX9-GISEL-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX9-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX9-GISEL-NEXT:    buffer_store_dwordx3 v[0:2], off, s[0:3], 0
@@ -240,10 +234,9 @@ define amdgpu_cs void @workgroup_id_optimized() "amdgpu-cluster-dims"="2,3,4" {
 ;
 ; GFX12-GISEL-LABEL: workgroup_id_optimized:
 ; GFX12-GISEL:       ; %bb.0: ; %.entry
-; GFX12-GISEL-NEXT:    s_mov_b32 s0, ttmp9
 ; GFX12-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
 ; GFX12-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, ttmp9 :: v_dual_mov_b32 v1, s1
 ; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX12-GISEL-NEXT:    buffer_store_b96 v[0:2], off, s[0:3], null
 ; GFX12-GISEL-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/mad_64_32.ll b/llvm/test/CodeGen/AMDGPU/mad_64_32.ll
index 7a52fda728fd5..025f41ea54578 100644
--- a/llvm/test/CodeGen/AMDGPU/mad_64_32.ll
+++ b/llvm/test/CodeGen/AMDGPU/mad_64_32.ll
@@ -1089,12 +1089,12 @@ define amdgpu_kernel void @mad_i64_i32_uniform(ptr addrspace(1) %out, i32 %arg0,
 ; GFX11-NEXT:    s_load_b64 s[4:5], s[4:5], 0x34
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    s_mul_i32 s6, s2, s3
-; GFX11-NEXT:    s_mul_hi_u32 s3, s2, s3
-; GFX11-NEXT:    s_add_u32 s2, s6, s4
-; GFX11-NEXT:    s_addc_u32 s3, s3, s5
+; GFX11-NEXT:    s_mul_hi_u32 s2, s2, s3
+; GFX11-NEXT:    s_add_u32 s3, s6, s4
+; GFX11-NEXT:    s_addc_u32 s2, s2, s5
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
-; GFX11-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s2
+; GFX11-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX11-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/max-hard-clause-length.ll b/llvm/test/CodeGen/AMDGPU/max-hard-clause-length.ll
index 5c3048f95f081..a89aeb73933e0 100644
--- a/llvm/test/CodeGen/AMDGPU/max-hard-clause-length.ll
+++ b/llvm/test/CodeGen/AMDGPU/max-hard-clause-length.ll
@@ -10,249 +10,238 @@
 define amdgpu_kernel void @long_store_chain(ptr addrspace(1) %p) {
 ; GFX10-LABEL: long_store_chain:
 ; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x24
-; GFX10-NEXT:    s_mov_b32 s0, 0
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10-NEXT:    s_mov_b32 s1, s0
-; GFX10-NEXT:    s_mov_b32 s2, s0
-; GFX10-NEXT:    s_mov_b32 s3, s0
-; GFX10-NEXT:    v_mov_b32_e32 v0, s0
-; GFX10-NEXT:    v_mov_b32_e32 v1, s1
-; GFX10-NEXT:    v_mov_b32_e32 v2, s2
-; GFX10-NEXT:    v_mov_b32_e32 v3, s3
+; GFX10-NEXT:    v_mov_b32_e32 v1, v0
+; GFX10-NEXT:    v_mov_b32_e32 v2, v0
+; GFX10-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5]
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:16
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:32
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:48
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:64
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:80
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:96
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:112
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:128
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:144
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:160
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:176
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:192
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:208
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:224
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:240
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:256
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:272
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:288
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:304
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:320
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:336
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:352
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:368
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:384
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:400
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:416
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:432
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:448
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:464
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:480
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:496
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:512
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:528
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:544
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:560
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:576
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:592
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:608
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:624
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:640
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:656
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:672
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:688
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:704
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:720
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:736
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:752
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:768
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:784
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:800
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:816
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:832
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:848
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:864
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:880
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:896
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:912
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:928
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:944
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:960
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:976
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:992
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:1008
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:1024
-; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[4:5] offset:1040
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:32
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:48
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:64
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:80
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:96
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:112
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:128
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:144
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:160
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:176
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:192
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:208
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:224
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:240
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:256
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:272
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:288
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:304
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:320
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:336
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:352
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:368
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:384
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:400
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:416
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:432
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:448
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:464
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:480
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:496
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:512
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:528
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:544
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:560
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:576
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:592
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:608
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:624
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:640
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:656
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:672
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:688
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:704
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:720
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:736
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:752
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:768
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:784
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:800
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:816
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:832
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:848
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:864
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:880
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:896
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:912
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:928
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:944
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:960
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:976
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:992
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:1008
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:1024
+; GFX10-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:1040
 ; GFX10-NEXT:    s_endpgm
 ;
 ; GFX11-LABEL: long_store_chain:
 ; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24
-; GFX11-NEXT:    s_mov_b32 s0, 0
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-NEXT:    s_mov_b32 s1, s0
-; GFX11-NEXT:    s_mov_b32 s2, s0
-; GFX11-NEXT:    s_mov_b32 s3, s0
-; GFX11-NEXT:    v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
-; GFX11-NEXT:    v_mov_b32_e32 v2, s2
+; GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v1, v0
+; GFX11-NEXT:    v_mov_b32_e32 v2, v0
+; GFX11-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    s_clause 0x1f
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5]
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:16
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:32
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:48
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:64
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:80
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:96
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:112
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:128
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:144
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:160
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:176
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:192
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:208
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:224
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:240
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:256
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:272
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:288
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:304
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:320
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:336
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:352
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:368
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:384
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:400
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:416
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:432
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:448
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:464
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:480
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:496
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1]
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:16
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:32
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:48
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:64
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:80
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:96
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:112
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:128
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:144
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:160
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:176
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:192
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:208
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:224
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:240
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:256
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:272
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:288
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:304
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:320
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:336
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:352
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:368
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:384
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:400
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:416
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:432
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:448
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:464
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:480
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:496
 ; GFX11-NEXT:    s_clause 0x1f
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:512
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:528
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:544
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:560
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:576
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:592
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:608
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:624
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:640
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:656
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:672
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:688
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:704
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:720
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:736
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:752
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:768
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:784
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:800
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:816
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:832
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:848
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:864
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:880
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:896
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:912
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:928
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:944
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:960
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:976
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:992
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:1008
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:512
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:528
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:544
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:560
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:576
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:592
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:608
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:624
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:640
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:656
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:672
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:688
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:704
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:720
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:736
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:752
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:768
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:784
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:800
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:816
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:832
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:848
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:864
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:880
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:896
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:912
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:928
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:944
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:960
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:976
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:992
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:1008
 ; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:1024
-; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:1040
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:1024
+; GFX11-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:1040
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX12-LABEL: long_store_chain:
 ; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_load_b64 s[4:5], s[4:5], 0x24
-; GFX12-NEXT:    s_mov_b32 s0, 0
-; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX12-NEXT:    s_mov_b32 s1, s0
-; GFX12-NEXT:    s_mov_b32 s2, s0
-; GFX12-NEXT:    s_mov_b32 s3, s0
-; GFX12-NEXT:    v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v1, s1
-; GFX12-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
-; GFX12-NEXT:    v_mov_b32_e32 v2, s2
+; GFX12-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v1, v0
+; GFX12-NEXT:    v_dual_mov_b32 v2, v0 :: v_dual_mov_b32 v3, v0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-NEXT:    s_clause 0x1f
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5]
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:16
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:32
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:48
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:64
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:80
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:96
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:112
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:128
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:144
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:160
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:176
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:192
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:208
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:224
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:240
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:256
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:272
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:288
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:304
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:320
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:336
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:352
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:368
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:384
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:400
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:416
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:432
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:448
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:464
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:480
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:496
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1]
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:16
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:32
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:48
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:64
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:80
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:96
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:112
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:128
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:144
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:160
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:176
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:192
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:208
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:224
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:240
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:256
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:272
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:288
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:304
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:320
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:336
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:352
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:368
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:384
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:400
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:416
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:432
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:448
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:464
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:480
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:496
 ; GFX12-NEXT:    s_clause 0x1f
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:512
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:528
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:544
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:560
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:576
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:592
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:608
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:624
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:640
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:656
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:672
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:688
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:704
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:720
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:736
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:752
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:768
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:784
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:800
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:816
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:832
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:848
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:864
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:880
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:896
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:912
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:928
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:944
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:960
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:976
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:992
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:1008
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:512
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:528
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:544
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:560
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:576
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:592
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:608
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:624
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:640
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:656
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:672
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:688
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:704
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:720
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:736
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:752
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:768
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:784
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:800
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:816
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:832
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:848
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:864
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:880
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:896
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:912
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:928
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:944
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:960
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:976
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:992
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:1008
 ; GFX12-NEXT:    s_clause 0x1
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:1024
-; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[4:5] offset:1040
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:1024
+; GFX12-NEXT:    global_store_b128 v4, v[0:3], s[0:1] offset:1040
 ; GFX12-NEXT:    s_endpgm
   store <4 x i32> zeroinitializer, ptr addrspace(1) %p
   %ptr1 = getelementptr <4 x i32>, ptr addrspace(1) %p, i64 1
diff --git a/llvm/test/CodeGen/AMDGPU/memintrinsic-unroll.ll b/llvm/test/CodeGen/AMDGPU/memintrinsic-unroll.ll
index 61f79ccd55f1c..2527623c860ec 100644
--- a/llvm/test/CodeGen/AMDGPU/memintrinsic-unroll.ll
+++ b/llvm/test/CodeGen/AMDGPU/memintrinsic-unroll.ll
@@ -16370,14 +16370,10 @@ define void @memset_p0_sz2048(ptr addrspace(0) %dst) #1 {
 ; UNROLL3-NEXT:    flat_store_dwordx4 v[6:7], v[2:5] offset:32
 ; UNROLL3-NEXT:    s_cbranch_vccnz .LBB10_1
 ; UNROLL3-NEXT:  ; %bb.2: ; %static-memset-post-expansion
-; UNROLL3-NEXT:    s_mov_b32 s4, 0x41414141
-; UNROLL3-NEXT:    s_mov_b32 s5, s4
-; UNROLL3-NEXT:    s_mov_b32 s6, s4
-; UNROLL3-NEXT:    s_mov_b32 s7, s4
-; UNROLL3-NEXT:    v_mov_b32_e32 v2, s4
-; UNROLL3-NEXT:    v_mov_b32_e32 v3, s5
-; UNROLL3-NEXT:    v_mov_b32_e32 v4, s6
-; UNROLL3-NEXT:    v_mov_b32_e32 v5, s7
+; UNROLL3-NEXT:    v_mov_b32_e32 v2, 0x41414141
+; UNROLL3-NEXT:    v_mov_b32_e32 v3, v2
+; UNROLL3-NEXT:    v_mov_b32_e32 v4, v2
+; UNROLL3-NEXT:    v_mov_b32_e32 v5, v2
 ; UNROLL3-NEXT:    flat_store_dwordx4 v[0:1], v[2:5] offset:2016
 ; UNROLL3-NEXT:    flat_store_dwordx4 v[0:1], v[2:5] offset:2032
 ; UNROLL3-NEXT:    s_waitcnt lgkmcnt(0)
@@ -16795,14 +16791,10 @@ define void @memset_p1_sz2048(ptr addrspace(1) %dst) #1 {
 ; UNROLL3-NEXT:    global_store_dwordx4 v[6:7], v[2:5], off offset:32
 ; UNROLL3-NEXT:    s_cbranch_vccnz .LBB11_1
 ; UNROLL3-NEXT:  ; %bb.2: ; %static-memset-post-expansion
-; UNROLL3-NEXT:    s_mov_b32 s4, 0x41414141
-; UNROLL3-NEXT:    s_mov_b32 s5, s4
-; UNROLL3-NEXT:    s_mov_b32 s6, s4
-; UNROLL3-NEXT:    s_mov_b32 s7, s4
-; UNROLL3-NEXT:    v_mov_b32_e32 v2, s4
-; UNROLL3-NEXT:    v_mov_b32_e32 v3, s5
-; UNROLL3-NEXT:    v_mov_b32_e32 v4, s6
-; UNROLL3-NEXT:    v_mov_b32_e32 v5, s7
+; UNROLL3-NEXT:    v_mov_b32_e32 v2, 0x41414141
+; UNROLL3-NEXT:    v_mov_b32_e32 v3, v2
+; UNROLL3-NEXT:    v_mov_b32_e32 v4, v2
+; UNROLL3-NEXT:    v_mov_b32_e32 v5, v2
 ; UNROLL3-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off offset:2016
 ; UNROLL3-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off offset:2032
 ; UNROLL3-NEXT:    s_setpc_b64 s[30:31]
@@ -17154,14 +17146,10 @@ define void @memset_p3_sz2048(ptr addrspace(3) %dst) #1 {
 ; UNROLL3-NEXT:    s_and_b32 vcc_lo, exec_lo, s10
 ; UNROLL3-NEXT:    s_cbranch_vccnz .LBB12_1
 ; UNROLL3-NEXT:  ; %bb.2: ; %static-memset-post-expansion
-; UNROLL3-NEXT:    s_mov_b32 s4, 0x41414141
-; UNROLL3-NEXT:    s_mov_b32 s5, s4
-; UNROLL3-NEXT:    s_mov_b32 s6, s4
-; UNROLL3-NEXT:    s_mov_b32 s7, s4
-; UNROLL3-NEXT:    v_mov_b32_e32 v1, s4
-; UNROLL3-NEXT:    v_mov_b32_e32 v2, s5
-; UNROLL3-NEXT:    v_mov_b32_e32 v3, s6
-; UNROLL3-NEXT:    v_mov_b32_e32 v4, s7
+; UNROLL3-NEXT:    v_mov_b32_e32 v1, 0x41414141
+; UNROLL3-NEXT:    v_mov_b32_e32 v2, v1
+; UNROLL3-NEXT:    v_mov_b32_e32 v3, v1
+; UNROLL3-NEXT:    v_mov_b32_e32 v4, v1
 ; UNROLL3-NEXT:    ds_write_b128 v0, v[1:4] offset:2016
 ; UNROLL3-NEXT:    ds_write_b128 v0, v[1:4] offset:2032
 ; UNROLL3-NEXT:    s_waitcnt lgkmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/memset-pattern.ll b/llvm/test/CodeGen/AMDGPU/memset-pattern.ll
index 8395587d94185..5a5e861e63d2b 100644
--- a/llvm/test/CodeGen/AMDGPU/memset-pattern.ll
+++ b/llvm/test/CodeGen/AMDGPU/memset-pattern.ll
@@ -23,29 +23,16 @@ define void @memset_pattern_i128_len1_dynvalue(ptr addrspace(1) align 16 %a, i12
 }
 
 define void @memset_pattern_i128_len1(ptr addrspace(1) align 16 %a) {
-; GFX942-SDAG-LABEL: memset_pattern_i128_len1:
-; GFX942-SDAG:       ; %bb.0: ; %memset.pattern-expansion-residual-body
-; GFX942-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v2, 0xdddddddd
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v3, 0xcccccccc
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v4, 0xbbbbbbbb
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v5, 0xaaaaaaaa
-; GFX942-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
-; GFX942-SDAG-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX942-GISEL-LABEL: memset_pattern_i128_len1:
-; GFX942-GISEL:       ; %bb.0: ; %memset.pattern-expansion-residual-body
-; GFX942-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 0xdddddddd
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, 0xcccccccc
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, 0xbbbbbbbb
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, 0xaaaaaaaa
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[0:1]
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
-; GFX942-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-GISEL-NEXT:    s_setpc_b64 s[30:31]
+; GFX942-LABEL: memset_pattern_i128_len1:
+; GFX942:       ; %bb.0: ; %memset.pattern-expansion-residual-body
+; GFX942-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b32_e32 v2, 0xdddddddd
+; GFX942-NEXT:    v_mov_b32_e32 v3, 0xcccccccc
+; GFX942-NEXT:    v_mov_b32_e32 v4, 0xbbbbbbbb
+; GFX942-NEXT:    v_mov_b32_e32 v5, 0xaaaaaaaa
+; GFX942-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX942-NEXT:    s_waitcnt vmcnt(0)
+; GFX942-NEXT:    s_setpc_b64 s[30:31]
   call void @llvm.experimental.memset.pattern(ptr addrspace(1) align 16 %a, i128 u0xaaaaaaaabbbbbbbbccccccccdddddddd, i64 1, i1 false)
   ret void
 }
@@ -176,57 +163,31 @@ define void @memset_pattern_i128_constlen_mainloop_and_residual_taken(ptr addrsp
 }
 
 define void @memset_pattern_i128_len1_lds(ptr addrspace(3) align 16 %a) {
-; GFX942-SDAG-LABEL: memset_pattern_i128_len1_lds:
-; GFX942-SDAG:       ; %bb.0: ; %memset.pattern-expansion-residual-body
-; GFX942-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v2, 0xdddddddd
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v3, 0xcccccccc
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v4, 0xbbbbbbbb
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v5, 0xaaaaaaaa
-; GFX942-SDAG-NEXT:    ds_write_b128 v0, v[2:5]
-; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX942-GISEL-LABEL: memset_pattern_i128_len1_lds:
-; GFX942-GISEL:       ; %bb.0: ; %memset.pattern-expansion-residual-body
-; GFX942-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 0xdddddddd
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, 0xcccccccc
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, 0xbbbbbbbb
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, 0xaaaaaaaa
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[0:1]
-; GFX942-GISEL-NEXT:    ds_write_b128 v0, v[2:5]
-; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    s_setpc_b64 s[30:31]
+; GFX942-LABEL: memset_pattern_i128_len1_lds:
+; GFX942:       ; %bb.0: ; %memset.pattern-expansion-residual-body
+; GFX942-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b32_e32 v2, 0xdddddddd
+; GFX942-NEXT:    v_mov_b32_e32 v3, 0xcccccccc
+; GFX942-NEXT:    v_mov_b32_e32 v4, 0xbbbbbbbb
+; GFX942-NEXT:    v_mov_b32_e32 v5, 0xaaaaaaaa
+; GFX942-NEXT:    ds_write_b128 v0, v[2:5]
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_setpc_b64 s[30:31]
   call void @llvm.experimental.memset.pattern(ptr addrspace(3) align 16 %a, i128 u0xaaaaaaaabbbbbbbbccccccccdddddddd, i64 1, i1 false)
   ret void
 }
 
 define void @memset_pattern_i128_len1_no_align(ptr addrspace(1) %a) {
-; GFX942-SDAG-LABEL: memset_pattern_i128_len1_no_align:
-; GFX942-SDAG:       ; %bb.0: ; %memset.pattern-expansion-residual-body
-; GFX942-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v2, 0xdddddddd
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v3, 0xcccccccc
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v4, 0xbbbbbbbb
-; GFX942-SDAG-NEXT:    v_mov_b32_e32 v5, 0xaaaaaaaa
-; GFX942-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
-; GFX942-SDAG-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX942-GISEL-LABEL: memset_pattern_i128_len1_no_align:
-; GFX942-GISEL:       ; %bb.0: ; %memset.pattern-expansion-residual-body
-; GFX942-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-GISEL-NEXT:    s_mov_b32 s0, 0xdddddddd
-; GFX942-GISEL-NEXT:    s_mov_b32 s1, 0xcccccccc
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, 0xbbbbbbbb
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, 0xaaaaaaaa
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[2:3]
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[0:1]
-; GFX942-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
-; GFX942-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-GISEL-NEXT:    s_setpc_b64 s[30:31]
+; GFX942-LABEL: memset_pattern_i128_len1_no_align:
+; GFX942:       ; %bb.0: ; %memset.pattern-expansion-residual-body
+; GFX942-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b32_e32 v2, 0xdddddddd
+; GFX942-NEXT:    v_mov_b32_e32 v3, 0xcccccccc
+; GFX942-NEXT:    v_mov_b32_e32 v4, 0xbbbbbbbb
+; GFX942-NEXT:    v_mov_b32_e32 v5, 0xaaaaaaaa
+; GFX942-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX942-NEXT:    s_waitcnt vmcnt(0)
+; GFX942-NEXT:    s_setpc_b64 s[30:31]
   call void @llvm.experimental.memset.pattern(ptr addrspace(1) %a, i128 u0xaaaaaaaabbbbbbbbccccccccdddddddd, i64 1, i1 false)
   ret void
 }
diff --git a/llvm/test/CodeGen/AMDGPU/min.ll b/llvm/test/CodeGen/AMDGPU/min.ll
index 65936651b52bd..943692635972c 100644
--- a/llvm/test/CodeGen/AMDGPU/min.ll
+++ b/llvm/test/CodeGen/AMDGPU/min.ll
@@ -3484,28 +3484,28 @@ define amdgpu_kernel void @s_test_umin_ult_v8i32(ptr addrspace(1) %out, <8 x i32
 ; CI-NEXT:    s_load_dwordx16 s[12:27], s[8:9], 0x8
 ; CI-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
 ; CI-NEXT:    s_waitcnt lgkmcnt(0)
-; CI-NEXT:    s_min_u32 s4, s15, s23
-; CI-NEXT:    s_min_u32 s5, s14, s22
-; CI-NEXT:    s_min_u32 s6, s13, s21
-; CI-NEXT:    s_min_u32 s7, s12, s20
-; CI-NEXT:    s_min_u32 s2, s19, s27
-; CI-NEXT:    s_min_u32 s3, s18, s26
+; CI-NEXT:    s_min_u32 s2, s15, s23
+; CI-NEXT:    s_min_u32 s3, s14, s22
+; CI-NEXT:    s_min_u32 s4, s13, s21
+; CI-NEXT:    s_min_u32 s5, s12, s20
+; CI-NEXT:    s_min_u32 s6, s19, s27
+; CI-NEXT:    s_min_u32 s7, s18, s26
 ; CI-NEXT:    s_min_u32 s8, s17, s25
 ; CI-NEXT:    s_min_u32 s9, s16, s24
-; CI-NEXT:    v_mov_b32_e32 v3, s2
-; CI-NEXT:    s_add_u32 s2, s0, 16
-; CI-NEXT:    v_mov_b32_e32 v2, s3
-; CI-NEXT:    s_addc_u32 s3, s1, 0
+; CI-NEXT:    v_mov_b32_e32 v3, s6
+; CI-NEXT:    s_add_u32 s6, s0, 16
+; CI-NEXT:    v_mov_b32_e32 v2, s7
+; CI-NEXT:    s_addc_u32 s7, s1, 0
 ; CI-NEXT:    v_mov_b32_e32 v0, s9
 ; CI-NEXT:    v_mov_b32_e32 v1, s8
-; CI-NEXT:    v_mov_b32_e32 v5, s3
-; CI-NEXT:    v_mov_b32_e32 v4, s2
+; CI-NEXT:    v_mov_b32_e32 v4, s6
+; CI-NEXT:    v_mov_b32_e32 v5, s7
 ; CI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; CI-NEXT:    v_mov_b32_e32 v5, s1
-; CI-NEXT:    v_mov_b32_e32 v0, s7
-; CI-NEXT:    v_mov_b32_e32 v1, s6
-; CI-NEXT:    v_mov_b32_e32 v2, s5
-; CI-NEXT:    v_mov_b32_e32 v3, s4
+; CI-NEXT:    v_mov_b32_e32 v0, s5
+; CI-NEXT:    v_mov_b32_e32 v1, s4
+; CI-NEXT:    v_mov_b32_e32 v2, s3
+; CI-NEXT:    v_mov_b32_e32 v3, s2
 ; CI-NEXT:    v_mov_b32_e32 v4, s0
 ; CI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; CI-NEXT:    s_endpgm
@@ -3518,28 +3518,28 @@ define amdgpu_kernel void @s_test_umin_ult_v8i32(ptr addrspace(1) %out, <8 x i32
 ; VI-NEXT:    s_load_dwordx16 s[12:27], s[8:9], 0x20
 ; VI-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-NEXT:    s_min_u32 s4, s15, s23
-; VI-NEXT:    s_min_u32 s5, s14, s22
-; VI-NEXT:    s_min_u32 s6, s13, s21
-; VI-NEXT:    s_min_u32 s7, s12, s20
-; VI-NEXT:    s_min_u32 s2, s19, s27
-; VI-NEXT:    s_min_u32 s3, s18, s26
+; VI-NEXT:    s_min_u32 s2, s15, s23
+; VI-NEXT:    s_min_u32 s3, s14, s22
+; VI-NEXT:    s_min_u32 s4, s13, s21
+; VI-NEXT:    s_min_u32 s5, s12, s20
+; VI-NEXT:    s_min_u32 s6, s19, s27
+; VI-NEXT:    s_min_u32 s7, s18, s26
 ; VI-NEXT:    s_min_u32 s8, s17, s25
 ; VI-NEXT:    s_min_u32 s9, s16, s24
-; VI-NEXT:    v_mov_b32_e32 v3, s2
-; VI-NEXT:    s_add_u32 s2, s0, 16
-; VI-NEXT:    v_mov_b32_e32 v2, s3
-; VI-NEXT:    s_addc_u32 s3, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v3, s6
+; VI-NEXT:    s_add_u32 s6, s0, 16
+; VI-NEXT:    v_mov_b32_e32 v2, s7
+; VI-NEXT:    s_addc_u32 s7, s1, 0
 ; VI-NEXT:    v_mov_b32_e32 v0, s9
 ; VI-NEXT:    v_mov_b32_e32 v1, s8
-; VI-NEXT:    v_mov_b32_e32 v5, s3
-; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v4, s6
+; VI-NEXT:    v_mov_b32_e32 v5, s7
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    v_mov_b32_e32 v5, s1
-; VI-NEXT:    v_mov_b32_e32 v0, s7
-; VI-NEXT:    v_mov_b32_e32 v1, s6
-; VI-NEXT:    v_mov_b32_e32 v2, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s4
+; VI-NEXT:    v_mov_b32_e32 v0, s5
+; VI-NEXT:    v_mov_b32_e32 v1, s4
+; VI-NEXT:    v_mov_b32_e32 v2, s3
+; VI-NEXT:    v_mov_b32_e32 v3, s2
 ; VI-NEXT:    v_mov_b32_e32 v4, s0
 ; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; VI-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/packed-fp32.ll b/llvm/test/CodeGen/AMDGPU/packed-fp32.ll
index 87a32de1c1677..d98081ad0c338 100644
--- a/llvm/test/CodeGen/AMDGPU/packed-fp32.ll
+++ b/llvm/test/CodeGen/AMDGPU/packed-fp32.ll
@@ -1191,10 +1191,10 @@ define amdgpu_kernel void @fadd_v2_v_fneg_lo(ptr addrspace(1) %a, float %x) {
 ; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-GISEL-NEXT:    global_load_b64 v[0:1], v4, s[0:1] scale_offset
 ; GFX1250-GISEL-NEXT:    v_max_num_f32_e64 v2, -s2, -s2
-; GFX1250-GISEL-NEXT:    s_mov_b32 s5, s2
-; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
-; GFX1250-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[4:5]
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v3, s2
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s3, v2
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_pk_add_f32 v[0:1], v[0:1], v[2:3]
@@ -1501,10 +1501,9 @@ define amdgpu_kernel void @fadd_v2_v_fneg_hi2(ptr addrspace(1) %a, float %x, flo
 ; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-GISEL-NEXT:    global_load_b64 v[0:1], v4, s[0:1] scale_offset
 ; GFX1250-GISEL-NEXT:    v_max_num_f32_e64 v2, -s2, -s2
-; GFX1250-GISEL-NEXT:    s_mov_b32 s4, s3
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s5, v2
-; GFX1250-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[4:5]
+; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX1250-GISEL-NEXT:    v_dual_mov_b32 v2, s3 :: v_dual_mov_b32 v3, s2
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_pk_add_f32 v[0:1], v[0:1], v[2:3]
@@ -3013,39 +3012,21 @@ define amdgpu_kernel void @fma_v2_v_imm(ptr addrspace(1) %a) {
 ; PACKED-SDAG-NEXT:    global_store_dwordx2 v3, v[0:1], s[0:1]
 ; PACKED-SDAG-NEXT:    s_endpgm
 ;
-; GFX90A-GISEL-LABEL: fma_v2_v_imm:
-; GFX90A-GISEL:       ; %bb.0:
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX90A-GISEL-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s4, 0x43480000
-; GFX90A-GISEL-NEXT:    s_mov_b32 s2, 0x42c80000
-; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-GISEL-NEXT:    global_load_dwordx2 v[0:1], v4, s[0:1]
-; GFX90A-GISEL-NEXT:    s_mov_b32 s5, s4
-; GFX90A-GISEL-NEXT:    s_mov_b32 s3, s2
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-GISEL-NEXT:    v_pk_fma_f32 v[0:1], v[0:1], s[2:3], v[2:3]
-; GFX90A-GISEL-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1]
-; GFX90A-GISEL-NEXT:    s_endpgm
-;
-; GFX942-GISEL-LABEL: fma_v2_v_imm:
-; GFX942-GISEL:       ; %bb.0:
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX942-GISEL-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
-; GFX942-GISEL-NEXT:    s_mov_b32 s4, 0x43480000
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, 0x42c80000
-; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    global_load_dwordx2 v[0:1], v4, s[0:1]
-; GFX942-GISEL-NEXT:    s_mov_b32 s5, s4
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, s2
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[4:5]
-; GFX942-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-GISEL-NEXT:    v_pk_fma_f32 v[0:1], v[0:1], s[2:3], v[2:3]
-; GFX942-GISEL-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    s_endpgm
+; PACKED-GISEL-LABEL: fma_v2_v_imm:
+; PACKED-GISEL:       ; %bb.0:
+; PACKED-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; PACKED-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; PACKED-GISEL-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
+; PACKED-GISEL-NEXT:    s_mov_b32 s2, 0x42c80000
+; PACKED-GISEL-NEXT:    v_mov_b32_e32 v2, 0x43480000
+; PACKED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; PACKED-GISEL-NEXT:    global_load_dwordx2 v[0:1], v4, s[0:1]
+; PACKED-GISEL-NEXT:    s_mov_b32 s3, s2
+; PACKED-GISEL-NEXT:    v_mov_b32_e32 v3, v2
+; PACKED-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; PACKED-GISEL-NEXT:    v_pk_fma_f32 v[0:1], v[0:1], s[2:3], v[2:3]
+; PACKED-GISEL-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1]
+; PACKED-GISEL-NEXT:    s_endpgm
 ;
 ; GFX1250-SDAG-LABEL: fma_v2_v_imm:
 ; GFX1250-SDAG:       ; %bb.0:
@@ -3071,10 +3052,9 @@ define amdgpu_kernel void @fma_v2_v_imm(ptr addrspace(1) %a) {
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
 ; GFX1250-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1250-GISEL-NEXT:    v_and_b32_e32 v4, 0x3ff, v0
-; GFX1250-GISEL-NEXT:    s_mov_b32 s2, 0x43480000
-; GFX1250-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1250-GISEL-NEXT:    s_mov_b32 s3, s2
-; GFX1250-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v2, 0x43480000
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v3, v2
 ; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-GISEL-NEXT:    global_load_b64 v[0:1], v4, s[0:1] scale_offset
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
@@ -3272,39 +3252,21 @@ define amdgpu_kernel void @fma_v2_v_literal32(ptr addrspace(1) %a) {
 ; PACKED-SDAG-NEXT:    global_store_dwordx2 v3, v[0:1], s[0:1]
 ; PACKED-SDAG-NEXT:    s_endpgm
 ;
-; GFX90A-GISEL-LABEL: fma_v2_v_literal32:
-; GFX90A-GISEL:       ; %bb.0:
-; GFX90A-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX90A-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX90A-GISEL-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
-; GFX90A-GISEL-NEXT:    s_mov_b32 s4, 0x43480000
-; GFX90A-GISEL-NEXT:    s_mov_b32 s2, 0x42c80000
-; GFX90A-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-GISEL-NEXT:    global_load_dwordx2 v[0:1], v4, s[0:1]
-; GFX90A-GISEL-NEXT:    s_mov_b32 s5, s4
-; GFX90A-GISEL-NEXT:    s_mov_b32 s3, s2
-; GFX90A-GISEL-NEXT:    v_pk_mov_b32 v[2:3], s[4:5], s[4:5] op_sel:[0,1]
-; GFX90A-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-GISEL-NEXT:    v_pk_fma_f32 v[0:1], v[0:1], s[2:3], v[2:3]
-; GFX90A-GISEL-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1]
-; GFX90A-GISEL-NEXT:    s_endpgm
-;
-; GFX942-GISEL-LABEL: fma_v2_v_literal32:
-; GFX942-GISEL:       ; %bb.0:
-; GFX942-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX942-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX942-GISEL-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
-; GFX942-GISEL-NEXT:    s_mov_b32 s4, 0x43480000
-; GFX942-GISEL-NEXT:    s_mov_b32 s2, 0x42c80000
-; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-GISEL-NEXT:    global_load_dwordx2 v[0:1], v4, s[0:1]
-; GFX942-GISEL-NEXT:    s_mov_b32 s5, s4
-; GFX942-GISEL-NEXT:    s_mov_b32 s3, s2
-; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[4:5]
-; GFX942-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-GISEL-NEXT:    v_pk_fma_f32 v[0:1], v[0:1], s[2:3], v[2:3]
-; GFX942-GISEL-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1]
-; GFX942-GISEL-NEXT:    s_endpgm
+; PACKED-GISEL-LABEL: fma_v2_v_literal32:
+; PACKED-GISEL:       ; %bb.0:
+; PACKED-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; PACKED-GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; PACKED-GISEL-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
+; PACKED-GISEL-NEXT:    s_mov_b32 s2, 0x42c80000
+; PACKED-GISEL-NEXT:    v_mov_b32_e32 v2, 0x43480000
+; PACKED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; PACKED-GISEL-NEXT:    global_load_dwordx2 v[0:1], v4, s[0:1]
+; PACKED-GISEL-NEXT:    s_mov_b32 s3, s2
+; PACKED-GISEL-NEXT:    v_mov_b32_e32 v3, v2
+; PACKED-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; PACKED-GISEL-NEXT:    v_pk_fma_f32 v[0:1], v[0:1], s[2:3], v[2:3]
+; PACKED-GISEL-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1]
+; PACKED-GISEL-NEXT:    s_endpgm
 ;
 ; GFX1250-SDAG-LABEL: fma_v2_v_literal32:
 ; GFX1250-SDAG:       ; %bb.0:
@@ -3330,10 +3292,9 @@ define amdgpu_kernel void @fma_v2_v_literal32(ptr addrspace(1) %a) {
 ; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
 ; GFX1250-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1250-GISEL-NEXT:    v_and_b32_e32 v4, 0x3ff, v0
-; GFX1250-GISEL-NEXT:    s_mov_b32 s2, 0x43480000
-; GFX1250-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1250-GISEL-NEXT:    s_mov_b32 s3, s2
-; GFX1250-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v2, 0x43480000
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_mov_b32_e32 v3, v2
 ; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-GISEL-NEXT:    global_load_b64 v[0:1], v4, s[0:1] scale_offset
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
diff --git a/llvm/test/CodeGen/AMDGPU/packed-fp64.ll b/llvm/test/CodeGen/AMDGPU/packed-fp64.ll
index c9bcd710adb97..8950163f76f04 100644
--- a/llvm/test/CodeGen/AMDGPU/packed-fp64.ll
+++ b/llvm/test/CodeGen/AMDGPU/packed-fp64.ll
@@ -563,17 +563,15 @@ define amdgpu_kernel void @fadd_v2_v_lit_hi0(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 1.0
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 0
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 1.0
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_add_f64 v[0:3], v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x double>, ptr addrspace(1) %a, i32 %id
@@ -607,17 +605,15 @@ define amdgpu_kernel void @fadd_v2_v_lit_lo0(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 1.0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 0
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 1.0
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 0
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_add_f64 v[0:3], v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x double>, ptr addrspace(1) %a, i32 %id
@@ -651,17 +647,15 @@ define amdgpu_kernel void @fadd_v2_v_unfoldable_lit(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 2.0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 1.0
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 2.0
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 1.0
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_add_f64 v[0:3], v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x double>, ptr addrspace(1) %a, i32 %id
@@ -700,11 +694,11 @@ define amdgpu_kernel void @fadd_v2_v_fneg(ptr addrspace(1) %a, double %x) {
 ; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    v_max_num_f64_e64 v[4:5], -s[2:3], -s[2:3]
 ; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s4, v4
-; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s5, v5
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[6:7], s[4:5]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s2, v4
+; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s3, v5
+; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1251-GISEL-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[4:5]
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1251-GISEL-NEXT:    v_pk_add_f64 v[0:3], v[0:3], v[4:7]
@@ -1448,17 +1442,15 @@ define amdgpu_kernel void @fmul_v2_v_unfoldable_lit(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 0x4008000000000000
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 4.0
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 0x4008000000000000
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 4.0
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_mul_f64 v[0:3], v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x double>, ptr addrspace(1) %a, i32 %id
@@ -1497,11 +1489,11 @@ define amdgpu_kernel void @fmul_v2_v_fneg(ptr addrspace(1) %a, double %x) {
 ; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    v_max_num_f64_e64 v[4:5], -s[2:3], -s[2:3]
 ; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s4, v4
-; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s5, v5
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[6:7], s[4:5]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s2, v4
+; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s3, v5
+; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1251-GISEL-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[4:5]
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1251-GISEL-NEXT:    v_pk_mul_f64 v[0:3], v[0:3], v[4:7]
@@ -1933,18 +1925,16 @@ define amdgpu_kernel void @fma_v2_v_imm(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 0x4069000000000000
-; GFX1251-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 0x4069000000000000
+; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[4:5]
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_fma_f64 v[0:3], 0x40590000, v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x double>, ptr addrspace(1) %a, i32 %id
@@ -2033,22 +2023,17 @@ define amdgpu_kernel void @fma_v2_v_unfoldable_lit(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[8:9], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v12, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 0x4008000000000000
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 4.0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[6:7], 2.0
-; GFX1251-GISEL-NEXT:    s_wait_xcnt 0x0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[4:5], 1.0
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[4:5]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 0x4008000000000000
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 4.0
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[10:11], 2.0
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[8:9], 1.0
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v12, s[8:9] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v12, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_fma_f64 v[0:3], v[0:3], v[4:7], v[8:11]
-; GFX1251-GISEL-NEXT:    global_store_b128 v12, v[0:3], s[8:9] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v12, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x double>, ptr addrspace(1) %a, i32 %id
@@ -2087,11 +2072,11 @@ define amdgpu_kernel void @fma_v2_v_fneg(ptr addrspace(1) %a, double %x) {
 ; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    v_max_num_f64_e64 v[4:5], -s[2:3], -s[2:3]
 ; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s4, v4
-; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s5, v5
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[6:7], s[4:5]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s2, v4
+; GFX1251-GISEL-NEXT:    v_readfirstlane_b32 s3, v5
+; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1251-GISEL-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s3
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], v[4:5]
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1251-GISEL-NEXT:    v_pk_fma_f64 v[0:3], v[0:3], v[4:7], v[4:7]
@@ -2762,15 +2747,12 @@ define amdgpu_kernel void @fma_v2_s_imm_imm(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[8:9], s[4:5], 0x24 nv
-; GFX1251-GISEL-NEXT:    s_wait_xcnt 0x0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[4:5], 0x4069000000000000
-; GFX1251-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[6:7], s[4:5]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[4:5]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[6:7]
+; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[2:3], 0x4069000000000000
+; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[2:3]
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[8:9], 0x0
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[6:7], 0x0
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
 ; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
@@ -2787,7 +2769,7 @@ define amdgpu_kernel void @fma_v2_s_imm_imm(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3)
 ; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX1251-GISEL-NEXT:    global_store_b128 v4, v[0:3], s[8:9] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v4, v[0:3], s[6:7] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x double>, ptr addrspace(1) %a, i32 %id
@@ -2825,15 +2807,12 @@ define amdgpu_kernel void @fma_v2_imm_imm_s(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[8:9], s[4:5], 0x24 nv
-; GFX1251-GISEL-NEXT:    s_wait_xcnt 0x0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[4:5], 0x4069000000000000
-; GFX1251-GISEL-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[6:7], s[4:5]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[4:5]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[6:7]
+; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[2:3], 0x4069000000000000
+; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], v[2:3]
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[8:9], 0x0
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[6:7], 0x0
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], s[0:1]
 ; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
@@ -2850,7 +2829,7 @@ define amdgpu_kernel void @fma_v2_imm_imm_s(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX1251-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3)
 ; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX1251-GISEL-NEXT:    global_store_b128 v4, v[0:3], s[8:9] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v4, v[0:3], s[6:7] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x double>, ptr addrspace(1) %a, i32 %id
diff --git a/llvm/test/CodeGen/AMDGPU/packed-u64.ll b/llvm/test/CodeGen/AMDGPU/packed-u64.ll
index 71bc83860ffb2..425e4e3535f0d 100644
--- a/llvm/test/CodeGen/AMDGPU/packed-u64.ll
+++ b/llvm/test/CodeGen/AMDGPU/packed-u64.ll
@@ -563,17 +563,15 @@ define amdgpu_kernel void @add_v2_v_lit_hi0(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 1
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 0
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 1
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_add_nc_u64 v[0:3], v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x i64>, ptr addrspace(1) %a, i32 %id
@@ -607,17 +605,15 @@ define amdgpu_kernel void @add_v2_v_lit_lo0(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 1
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 0
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 1
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 0
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_add_nc_u64 v[0:3], v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x i64>, ptr addrspace(1) %a, i32 %id
@@ -651,17 +647,15 @@ define amdgpu_kernel void @add_v2_v_unfoldable_lit(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 2
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 1
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 2
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 1
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_add_nc_u64 v[0:3], v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x i64>, ptr addrspace(1) %a, i32 %id
@@ -1213,17 +1207,15 @@ define amdgpu_kernel void @sub_v2_v_lit_hi0(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 1
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 0
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 1
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_sub_nc_u64 v[0:3], v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x i64>, ptr addrspace(1) %a, i32 %id
@@ -1257,17 +1249,15 @@ define amdgpu_kernel void @sub_v2_v_lit_lo0(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 1
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 0
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 1
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 0
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_sub_nc_u64 v[0:3], v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x i64>, ptr addrspace(1) %a, i32 %id
@@ -1301,17 +1291,15 @@ define amdgpu_kernel void @sub_v2_v_unfoldable_lit(ptr addrspace(1) %a) {
 ; GFX1251-GISEL-NEXT:    v_nop
 ; GFX1251-GISEL-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; GFX1251-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1251-GISEL-NEXT:    s_load_b64 s[6:7], s[4:5], 0x24 nv
+; GFX1251-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24 nv
 ; GFX1251-GISEL-NEXT:    v_and_b32_e32 v8, 0x3ff, v0
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[2:3], 2
-; GFX1251-GISEL-NEXT:    s_mov_b64 s[0:1], 1
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[2:3]
-; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[6:7], 2
+; GFX1251-GISEL-NEXT:    v_mov_b64_e32 v[4:5], 1
 ; GFX1251-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_load_b128 v[0:3], v8, s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1251-GISEL-NEXT:    v_pk_sub_nc_u64 v[0:3], v[0:3], v[4:7]
-; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[6:7] scale_offset
+; GFX1251-GISEL-NEXT:    global_store_b128 v8, v[0:3], s[0:1] scale_offset
 ; GFX1251-GISEL-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <2 x i64>, ptr addrspace(1) %a, i32 %id
diff --git a/llvm/test/CodeGen/AMDGPU/promote-alloca-vector-dynamic-idx-bitcasts-llc.ll b/llvm/test/CodeGen/AMDGPU/promote-alloca-vector-dynamic-idx-bitcasts-llc.ll
index 5321d0cabf947..daa1f0e589c34 100644
--- a/llvm/test/CodeGen/AMDGPU/promote-alloca-vector-dynamic-idx-bitcasts-llc.ll
+++ b/llvm/test/CodeGen/AMDGPU/promote-alloca-vector-dynamic-idx-bitcasts-llc.ll
@@ -6,80 +6,60 @@
 define amdgpu_kernel void @test_bitcast_llc_v128i8_v16i8(ptr addrspace(1) %out, i32 %idx) {
 ; GFX9-LABEL: test_bitcast_llc_v128i8_v16i8:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x0
-; GFX9-NEXT:    s_load_dword s33, s[4:5], 0x8
-; GFX9-NEXT:    s_lshl_b32 s0, s0, 8
-; GFX9-NEXT:    s_and_b32 s1, s0, 0xff
-; GFX9-NEXT:    s_or_b32 s0, s1, s0
-; GFX9-NEXT:    s_and_b32 s1, s0, 0xffff
-; GFX9-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX9-NEXT:    s_or_b32 s0, s1, s0
-; GFX9-NEXT:    s_mov_b32 s1, s0
-; GFX9-NEXT:    s_mov_b32 s2, s0
-; GFX9-NEXT:    s_mov_b32 s3, s0
-; GFX9-NEXT:    s_mov_b32 s4, s0
-; GFX9-NEXT:    s_mov_b32 s5, s0
-; GFX9-NEXT:    s_mov_b32 s6, s0
-; GFX9-NEXT:    s_mov_b32 s7, s0
-; GFX9-NEXT:    s_mov_b32 s8, s0
-; GFX9-NEXT:    s_mov_b32 s9, s0
-; GFX9-NEXT:    s_mov_b32 s10, s0
-; GFX9-NEXT:    s_mov_b32 s11, s0
-; GFX9-NEXT:    s_mov_b32 s12, s0
-; GFX9-NEXT:    s_mov_b32 s13, s0
-; GFX9-NEXT:    s_mov_b32 s14, s0
-; GFX9-NEXT:    s_mov_b32 s15, s0
-; GFX9-NEXT:    s_mov_b32 s16, s0
-; GFX9-NEXT:    s_mov_b32 s17, s0
-; GFX9-NEXT:    s_mov_b32 s18, s0
-; GFX9-NEXT:    s_mov_b32 s19, s0
-; GFX9-NEXT:    s_mov_b32 s20, s0
-; GFX9-NEXT:    s_mov_b32 s21, s0
-; GFX9-NEXT:    s_mov_b32 s22, s0
-; GFX9-NEXT:    s_mov_b32 s23, s0
-; GFX9-NEXT:    s_mov_b32 s24, s0
-; GFX9-NEXT:    s_mov_b32 s25, s0
-; GFX9-NEXT:    s_mov_b32 s26, s0
-; GFX9-NEXT:    s_mov_b32 s27, s0
-; GFX9-NEXT:    s_mov_b32 s28, s0
-; GFX9-NEXT:    s_mov_b32 s29, s0
-; GFX9-NEXT:    s_mov_b32 s30, s0
-; GFX9-NEXT:    s_mov_b32 s31, s0
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v36, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_lshl_b32 s33, s33, 2
-; GFX9-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
-; GFX9-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
-; GFX9-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
-; GFX9-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
-; GFX9-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
-; GFX9-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
-; GFX9-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
-; GFX9-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
-; GFX9-NEXT:    v_mov_b64_e32 v[16:17], s[16:17]
-; GFX9-NEXT:    v_mov_b64_e32 v[18:19], s[18:19]
-; GFX9-NEXT:    v_mov_b64_e32 v[20:21], s[20:21]
-; GFX9-NEXT:    v_mov_b64_e32 v[22:23], s[22:23]
-; GFX9-NEXT:    v_mov_b64_e32 v[24:25], s[24:25]
-; GFX9-NEXT:    v_mov_b64_e32 v[26:27], s[26:27]
-; GFX9-NEXT:    v_mov_b64_e32 v[28:29], s[28:29]
-; GFX9-NEXT:    v_mov_b64_e32 v[30:31], s[30:31]
-; GFX9-NEXT:    s_set_gpr_idx_on s33, gpr_idx(SRC0)
+; GFX9-NEXT:    s_lshl_b32 s3, s0, 8
+; GFX9-NEXT:    s_and_b32 s4, s0, 0xff
+; GFX9-NEXT:    s_or_b32 s3, s4, s3
+; GFX9-NEXT:    s_and_b32 s4, s3, 0xffff
+; GFX9-NEXT:    s_lshl_b32 s3, s3, 16
+; GFX9-NEXT:    s_or_b32 s3, s4, s3
+; GFX9-NEXT:    s_lshl_b32 s2, s2, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s3
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_mov_b32_e32 v2, s3
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    v_mov_b32_e32 v4, s3
+; GFX9-NEXT:    v_mov_b32_e32 v5, s3
+; GFX9-NEXT:    v_mov_b32_e32 v6, s3
+; GFX9-NEXT:    v_mov_b32_e32 v7, s3
+; GFX9-NEXT:    v_mov_b32_e32 v8, s3
+; GFX9-NEXT:    v_mov_b32_e32 v9, s3
+; GFX9-NEXT:    v_mov_b32_e32 v10, s3
+; GFX9-NEXT:    v_mov_b32_e32 v11, s3
+; GFX9-NEXT:    v_mov_b32_e32 v12, s3
+; GFX9-NEXT:    v_mov_b32_e32 v13, s3
+; GFX9-NEXT:    v_mov_b32_e32 v14, s3
+; GFX9-NEXT:    v_mov_b32_e32 v15, s3
+; GFX9-NEXT:    v_mov_b32_e32 v16, s3
+; GFX9-NEXT:    v_mov_b32_e32 v17, s3
+; GFX9-NEXT:    v_mov_b32_e32 v18, s3
+; GFX9-NEXT:    v_mov_b32_e32 v19, s3
+; GFX9-NEXT:    v_mov_b32_e32 v20, s3
+; GFX9-NEXT:    v_mov_b32_e32 v21, s3
+; GFX9-NEXT:    v_mov_b32_e32 v22, s3
+; GFX9-NEXT:    v_mov_b32_e32 v23, s3
+; GFX9-NEXT:    v_mov_b32_e32 v24, s3
+; GFX9-NEXT:    v_mov_b32_e32 v25, s3
+; GFX9-NEXT:    v_mov_b32_e32 v26, s3
+; GFX9-NEXT:    v_mov_b32_e32 v27, s3
+; GFX9-NEXT:    v_mov_b32_e32 v28, s3
+; GFX9-NEXT:    v_mov_b32_e32 v29, s3
+; GFX9-NEXT:    v_mov_b32_e32 v30, s3
+; GFX9-NEXT:    v_mov_b32_e32 v31, s3
+; GFX9-NEXT:    s_set_gpr_idx_on s2, gpr_idx(SRC0)
 ; GFX9-NEXT:    v_mov_b32_e32 v35, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v34, v2
-; GFX9-NEXT:    s_set_gpr_idx_off
-; GFX9-NEXT:    v_mov_b32_e32 v36, 0
-; GFX9-NEXT:    s_set_gpr_idx_on s33, gpr_idx(SRC0)
 ; GFX9-NEXT:    v_mov_b32_e32 v33, v1
 ; GFX9-NEXT:    v_mov_b32_e32 v32, v0
 ; GFX9-NEXT:    s_set_gpr_idx_off
-; GFX9-NEXT:    global_store_dwordx4 v36, v[32:35], s[34:35]
+; GFX9-NEXT:    global_store_dwordx4 v36, v[32:35], s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX11-LABEL: test_bitcast_llc_v128i8_v16i8:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b64 s[34:35], s[4:5], 0x0
-; GFX11-NEXT:    s_load_b32 s33, s[4:5], 0x8
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 8
 ; GFX11-NEXT:    v_mov_b32_e32 v35, 0
 ; GFX11-NEXT:    s_and_b32 s1, s0, 0xff
@@ -87,67 +67,38 @@ define amdgpu_kernel void @test_bitcast_llc_v128i8_v16i8(ptr addrspace(1) %out,
 ; GFX11-NEXT:    s_or_b32 s0, s1, s0
 ; GFX11-NEXT:    s_and_b32 s1, s0, 0xffff
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT:    s_or_b32 s0, s1, s0
-; GFX11-NEXT:    s_mov_b32 s1, s0
-; GFX11-NEXT:    s_mov_b32 s2, s0
-; GFX11-NEXT:    s_mov_b32 s3, s0
-; GFX11-NEXT:    s_mov_b32 s4, s0
-; GFX11-NEXT:    s_mov_b32 s5, s0
-; GFX11-NEXT:    s_mov_b32 s6, s0
-; GFX11-NEXT:    s_mov_b32 s7, s0
-; GFX11-NEXT:    s_mov_b32 s8, s0
-; GFX11-NEXT:    s_mov_b32 s9, s0
-; GFX11-NEXT:    s_mov_b32 s10, s0
-; GFX11-NEXT:    s_mov_b32 s11, s0
-; GFX11-NEXT:    s_mov_b32 s12, s0
-; GFX11-NEXT:    s_mov_b32 s13, s0
-; GFX11-NEXT:    s_mov_b32 s14, s0
-; GFX11-NEXT:    s_mov_b32 s15, s0
-; GFX11-NEXT:    s_mov_b32 s16, s0
-; GFX11-NEXT:    s_mov_b32 s17, s0
-; GFX11-NEXT:    s_mov_b32 s18, s0
-; GFX11-NEXT:    s_mov_b32 s19, s0
-; GFX11-NEXT:    s_mov_b32 s20, s0
-; GFX11-NEXT:    s_mov_b32 s21, s0
-; GFX11-NEXT:    s_mov_b32 s22, s0
-; GFX11-NEXT:    s_mov_b32 s23, s0
-; GFX11-NEXT:    s_mov_b32 s24, s0
-; GFX11-NEXT:    s_mov_b32 s25, s0
-; GFX11-NEXT:    s_mov_b32 s26, s0
-; GFX11-NEXT:    s_mov_b32 s27, s0
-; GFX11-NEXT:    s_mov_b32 s28, s0
-; GFX11-NEXT:    s_mov_b32 s29, s0
-; GFX11-NEXT:    s_mov_b32 s30, s0
-; GFX11-NEXT:    s_mov_b32 s31, s0
-; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT:    s_or_b32 s2, s1, s0
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX11-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s2
+; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s2
+; GFX11-NEXT:    v_dual_mov_b32 v30, s2 :: v_dual_mov_b32 v31, s2
+; GFX11-NEXT:    v_dual_mov_b32 v4, s2 :: v_dual_mov_b32 v5, s2
+; GFX11-NEXT:    v_dual_mov_b32 v6, s2 :: v_dual_mov_b32 v7, s2
+; GFX11-NEXT:    v_dual_mov_b32 v8, s2 :: v_dual_mov_b32 v9, s2
+; GFX11-NEXT:    v_dual_mov_b32 v10, s2 :: v_dual_mov_b32 v11, s2
+; GFX11-NEXT:    v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v13, s2
+; GFX11-NEXT:    v_dual_mov_b32 v14, s2 :: v_dual_mov_b32 v15, s2
+; GFX11-NEXT:    v_dual_mov_b32 v16, s2 :: v_dual_mov_b32 v17, s2
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    s_lshl_b32 m0, s33, 2
-; GFX11-NEXT:    v_dual_mov_b32 v30, s30 :: v_dual_mov_b32 v31, s31
-; GFX11-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v5, s5
-; GFX11-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
-; GFX11-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
-; GFX11-NEXT:    v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
-; GFX11-NEXT:    v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
-; GFX11-NEXT:    v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s15
-; GFX11-NEXT:    v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
-; GFX11-NEXT:    v_dual_mov_b32 v18, s18 :: v_dual_mov_b32 v19, s19
-; GFX11-NEXT:    v_dual_mov_b32 v20, s20 :: v_dual_mov_b32 v21, s21
-; GFX11-NEXT:    v_dual_mov_b32 v22, s22 :: v_dual_mov_b32 v23, s23
-; GFX11-NEXT:    v_dual_mov_b32 v24, s24 :: v_dual_mov_b32 v25, s25
-; GFX11-NEXT:    v_dual_mov_b32 v26, s26 :: v_dual_mov_b32 v27, s27
-; GFX11-NEXT:    v_dual_mov_b32 v28, s28 :: v_dual_mov_b32 v29, s29
+; GFX11-NEXT:    s_lshl_b32 m0, s3, 2
+; GFX11-NEXT:    v_dual_mov_b32 v18, s2 :: v_dual_mov_b32 v19, s2
+; GFX11-NEXT:    v_dual_mov_b32 v20, s2 :: v_dual_mov_b32 v21, s2
+; GFX11-NEXT:    v_dual_mov_b32 v22, s2 :: v_dual_mov_b32 v23, s2
+; GFX11-NEXT:    v_dual_mov_b32 v24, s2 :: v_dual_mov_b32 v25, s2
+; GFX11-NEXT:    v_dual_mov_b32 v26, s2 :: v_dual_mov_b32 v27, s2
+; GFX11-NEXT:    v_dual_mov_b32 v28, s2 :: v_dual_mov_b32 v29, s2
 ; GFX11-NEXT:    v_movrels_b32_e32 v34, v3
 ; GFX11-NEXT:    v_movrels_b32_e32 v33, v2
 ; GFX11-NEXT:    v_movrels_b32_e32 v32, v1
 ; GFX11-NEXT:    v_movrels_b32_e32 v31, v0
-; GFX11-NEXT:    global_store_b128 v35, v[31:34], s[34:35]
+; GFX11-NEXT:    global_store_b128 v35, v[31:34], s[0:1]
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX12-LABEL: test_bitcast_llc_v128i8_v16i8:
 ; GFX12:       ; %bb.0: ; %entry
-; GFX12-NEXT:    s_load_b96 s[36:38], s[4:5], 0x0
 ; GFX12-NEXT:    s_lshl_b32 s0, s0, 8
 ; GFX12-NEXT:    v_mov_b32_e32 v35, 0
 ; GFX12-NEXT:    s_and_b32 s1, s0, 0xff
@@ -155,62 +106,32 @@ define amdgpu_kernel void @test_bitcast_llc_v128i8_v16i8(ptr addrspace(1) %out,
 ; GFX12-NEXT:    s_or_b32 s0, s1, s0
 ; GFX12-NEXT:    s_and_b32 s1, s0, 0xffff
 ; GFX12-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX12-NEXT:    s_or_b32 s0, s1, s0
-; GFX12-NEXT:    s_mov_b32 s1, s0
-; GFX12-NEXT:    s_mov_b32 s2, s0
-; GFX12-NEXT:    s_mov_b32 s3, s0
-; GFX12-NEXT:    s_mov_b32 s4, s0
-; GFX12-NEXT:    s_mov_b32 s5, s0
-; GFX12-NEXT:    s_mov_b32 s6, s0
-; GFX12-NEXT:    s_mov_b32 s7, s0
-; GFX12-NEXT:    s_mov_b32 s8, s0
-; GFX12-NEXT:    s_mov_b32 s9, s0
-; GFX12-NEXT:    s_mov_b32 s10, s0
-; GFX12-NEXT:    s_mov_b32 s11, s0
-; GFX12-NEXT:    s_mov_b32 s12, s0
-; GFX12-NEXT:    s_mov_b32 s13, s0
-; GFX12-NEXT:    s_mov_b32 s14, s0
-; GFX12-NEXT:    s_mov_b32 s15, s0
-; GFX12-NEXT:    s_mov_b32 s16, s0
-; GFX12-NEXT:    s_mov_b32 s17, s0
-; GFX12-NEXT:    s_mov_b32 s18, s0
-; GFX12-NEXT:    s_mov_b32 s19, s0
-; GFX12-NEXT:    s_mov_b32 s20, s0
-; GFX12-NEXT:    s_mov_b32 s21, s0
-; GFX12-NEXT:    s_mov_b32 s22, s0
-; GFX12-NEXT:    s_mov_b32 s23, s0
-; GFX12-NEXT:    s_mov_b32 s24, s0
-; GFX12-NEXT:    s_mov_b32 s25, s0
-; GFX12-NEXT:    s_mov_b32 s26, s0
-; GFX12-NEXT:    s_mov_b32 s27, s0
-; GFX12-NEXT:    s_mov_b32 s28, s0
-; GFX12-NEXT:    s_mov_b32 s29, s0
-; GFX12-NEXT:    s_mov_b32 s30, s0
-; GFX12-NEXT:    s_mov_b32 s31, s0
-; GFX12-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
-; GFX12-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX12-NEXT:    s_or_b32 s3, s1, s0
+; GFX12-NEXT:    s_load_b96 s[0:2], s[4:5], 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v0, s3 :: v_dual_mov_b32 v1, s3
+; GFX12-NEXT:    v_dual_mov_b32 v2, s3 :: v_dual_mov_b32 v3, s3
+; GFX12-NEXT:    v_dual_mov_b32 v30, s3 :: v_dual_mov_b32 v31, s3
+; GFX12-NEXT:    v_dual_mov_b32 v4, s3 :: v_dual_mov_b32 v5, s3
+; GFX12-NEXT:    v_dual_mov_b32 v6, s3 :: v_dual_mov_b32 v7, s3
+; GFX12-NEXT:    v_dual_mov_b32 v8, s3 :: v_dual_mov_b32 v9, s3
+; GFX12-NEXT:    v_dual_mov_b32 v10, s3 :: v_dual_mov_b32 v11, s3
+; GFX12-NEXT:    v_dual_mov_b32 v12, s3 :: v_dual_mov_b32 v13, s3
+; GFX12-NEXT:    v_dual_mov_b32 v14, s3 :: v_dual_mov_b32 v15, s3
+; GFX12-NEXT:    v_dual_mov_b32 v16, s3 :: v_dual_mov_b32 v17, s3
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    s_lshl_b32 m0, s38, 2
-; GFX12-NEXT:    v_dual_mov_b32 v30, s30 :: v_dual_mov_b32 v31, s31
-; GFX12-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v5, s5
-; GFX12-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v7, s7
-; GFX12-NEXT:    v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
-; GFX12-NEXT:    v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
-; GFX12-NEXT:    v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
-; GFX12-NEXT:    v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s15
-; GFX12-NEXT:    v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
-; GFX12-NEXT:    v_dual_mov_b32 v18, s18 :: v_dual_mov_b32 v19, s19
-; GFX12-NEXT:    v_dual_mov_b32 v20, s20 :: v_dual_mov_b32 v21, s21
-; GFX12-NEXT:    v_dual_mov_b32 v22, s22 :: v_dual_mov_b32 v23, s23
-; GFX12-NEXT:    v_dual_mov_b32 v24, s24 :: v_dual_mov_b32 v25, s25
-; GFX12-NEXT:    v_dual_mov_b32 v26, s26 :: v_dual_mov_b32 v27, s27
-; GFX12-NEXT:    v_dual_mov_b32 v28, s28 :: v_dual_mov_b32 v29, s29
+; GFX12-NEXT:    s_lshl_b32 m0, s2, 2
+; GFX12-NEXT:    v_dual_mov_b32 v18, s3 :: v_dual_mov_b32 v19, s3
+; GFX12-NEXT:    v_dual_mov_b32 v20, s3 :: v_dual_mov_b32 v21, s3
+; GFX12-NEXT:    v_dual_mov_b32 v22, s3 :: v_dual_mov_b32 v23, s3
+; GFX12-NEXT:    v_dual_mov_b32 v24, s3 :: v_dual_mov_b32 v25, s3
+; GFX12-NEXT:    v_dual_mov_b32 v26, s3 :: v_dual_mov_b32 v27, s3
+; GFX12-NEXT:    v_dual_mov_b32 v28, s3 :: v_dual_mov_b32 v29, s3
 ; GFX12-NEXT:    v_movrels_b32_e32 v34, v3
 ; GFX12-NEXT:    v_movrels_b32_e32 v33, v2
 ; GFX12-NEXT:    v_movrels_b32_e32 v32, v1
 ; GFX12-NEXT:    v_movrels_b32_e32 v31, v0
-; GFX12-NEXT:    global_store_b128 v35, v[31:34], s[36:37]
+; GFX12-NEXT:    global_store_b128 v35, v[31:34], s[0:1]
 ; GFX12-NEXT:    s_endpgm
 entry:
   %alloca = freeze <128 x i8> poison
diff --git a/llvm/test/CodeGen/AMDGPU/promote-constOffset-to-imm.ll b/llvm/test/CodeGen/AMDGPU/promote-constOffset-to-imm.ll
index 5b232348c4577..626892694ceb9 100644
--- a/llvm/test/CodeGen/AMDGPU/promote-constOffset-to-imm.ll
+++ b/llvm/test/CodeGen/AMDGPU/promote-constOffset-to-imm.ll
@@ -2582,12 +2582,12 @@ entry:
 define amdgpu_kernel void @negativeoffsetnullptr(ptr %buffer) {
 ; GFX8-LABEL: negativeoffsetnullptr:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dword s1, s[4:5], 0xec
-; GFX8-NEXT:    s_add_u32 s0, 0, -1
+; GFX8-NEXT:    s_load_dword s0, s[4:5], 0xec
+; GFX8-NEXT:    s_add_u32 s1, 0, -1
+; GFX8-NEXT:    v_mov_b32_e32 v0, s1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_addc_u32 s1, s1, -1
-; GFX8-NEXT:    v_mov_b32_e32 v0, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s1
+; GFX8-NEXT:    s_addc_u32 s0, s0, -1
+; GFX8-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX8-NEXT:    flat_load_ubyte v0, v[0:1]
 ; GFX8-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/ptradd-sdag.ll b/llvm/test/CodeGen/AMDGPU/ptradd-sdag.ll
index 42b4282dc2c5c..e5f5a79278bd2 100644
--- a/llvm/test/CodeGen/AMDGPU/ptradd-sdag.ll
+++ b/llvm/test/CodeGen/AMDGPU/ptradd-sdag.ll
@@ -88,8 +88,8 @@ define amdgpu_kernel void @gep_as0_uniform(ptr %p, i64 %offset, ptr %ret) {
 ; GFX8-NEXT:    s_add_u32 s0, s4, 4
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    s_addc_u32 s1, s5, 0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX8-NEXT:    flat_store_dword v[2:3], v5
@@ -124,8 +124,8 @@ define amdgpu_kernel void @gep_as0_uniform(ptr %p, i64 %offset, ptr %ret) {
 ; GFX10-NEXT:    s_add_u32 s0, s0, s2
 ; GFX10-NEXT:    s_addc_u32 s1, s1, s3
 ; GFX10-NEXT:    v_mov_b32_e32 v1, s5
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    v_mov_b32_e32 v2, s0
+; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX10-NEXT:    s_endpgm
 ;
@@ -245,8 +245,8 @@ define amdgpu_kernel void @multi_gep_as0_uniform(ptr %p, i64 %offset, ptr %ret)
 ; GFX8-NEXT:    s_add_u32 s0, s4, 4
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    s_addc_u32 s1, s5, 0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX8-NEXT:    flat_store_dword v[2:3], v5
@@ -285,8 +285,8 @@ define amdgpu_kernel void @multi_gep_as0_uniform(ptr %p, i64 %offset, ptr %ret)
 ; GFX10-NEXT:    s_add_u32 s0, s0, 5
 ; GFX10-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v1, s5
-; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    v_mov_b32_e32 v2, s0
+; GFX10-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX10-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX10-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/rotr.ll b/llvm/test/CodeGen/AMDGPU/rotr.ll
index b25e58604b4e6..f232107841651 100644
--- a/llvm/test/CodeGen/AMDGPU/rotr.ll
+++ b/llvm/test/CodeGen/AMDGPU/rotr.ll
@@ -411,14 +411,14 @@ define amdgpu_kernel void @rotr_v8i32(ptr addrspace(1) %in, <8 x i32> %x, <8 x i
 ; GFX8-NEXT:    s_and_b32 s3, s20, 31
 ; GFX8-NEXT:    s_mov_b32 s13, s12
 ; GFX8-NEXT:    s_lshr_b64 s[12:13], s[12:13], s3
-; GFX8-NEXT:    v_mov_b32_e32 v3, s10
-; GFX8-NEXT:    s_add_u32 s10, s0, 16
-; GFX8-NEXT:    s_addc_u32 s11, s1, 0
+; GFX8-NEXT:    s_add_u32 s3, s0, 16
+; GFX8-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s12
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s16
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s14
-; GFX8-NEXT:    v_mov_b32_e32 v4, s10
-; GFX8-NEXT:    v_mov_b32_e32 v5, s11
+; GFX8-NEXT:    v_mov_b32_e32 v3, s10
+; GFX8-NEXT:    v_mov_b32_e32 v4, s3
+; GFX8-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX8-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v5, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s8
diff --git a/llvm/test/CodeGen/AMDGPU/saddo.ll b/llvm/test/CodeGen/AMDGPU/saddo.ll
index 9d8869c765e29..830d42a41c211 100644
--- a/llvm/test/CodeGen/AMDGPU/saddo.ll
+++ b/llvm/test/CodeGen/AMDGPU/saddo.ll
@@ -58,8 +58,8 @@ define amdgpu_kernel void @saddo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 %b)
 ; VI-NEXT:    s_cselect_b64 s[2:3], 1, 0
 ; VI-NEXT:    s_add_u32 s0, s0, s2
 ; VI-NEXT:    s_addc_u32 s1, s1, s3
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; VI-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/scale-offset-global.ll b/llvm/test/CodeGen/AMDGPU/scale-offset-global.ll
index dd8b27436a622..10cd4fa2ab8f8 100644
--- a/llvm/test/CodeGen/AMDGPU/scale-offset-global.ll
+++ b/llvm/test/CodeGen/AMDGPU/scale-offset-global.ll
@@ -82,8 +82,8 @@ define amdgpu_ps float @global_load_b32_idxprom_wrong_stride(ptr addrspace(1) al
 ;
 ; GFX13-GISEL-LABEL: global_load_b32_idxprom_wrong_stride:
 ; GFX13-GISEL:       ; %bb.0: ; %entry
-; GFX13-GISEL-NEXT:    v_dual_ashrrev_i32 v1, 31, v0 :: v_dual_mov_b32 v3, s1
-; GFX13-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX13-GISEL-NEXT:    v_dual_mov_b32 v2, s0 :: v_dual_ashrrev_i32 v1, 31, v0
+; GFX13-GISEL-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX13-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX13-GISEL-NEXT:    v_lshlrev_b64_e32 v[0:1], 3, v[0:1]
 ; GFX13-GISEL-NEXT:    v_add_co_u32 v0, vcc_lo, v2, v0
diff --git a/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll b/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll
index 7044517b6a432..19fda5f431f61 100644
--- a/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll
+++ b/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll
@@ -21,7 +21,7 @@
 
 ; CHECK-LABEL: {{^}}call_72xi32:
 ; GFX11-PAL:    NumSgprs: 40
-; GFX11-PAL-GCNTRACKERS:    NumSgprs: 37
+; GFX11-PAL-GCNTRACKERS:    NumSgprs: 36
 ; GFX11-PAL:    NumVgprs: 64
 ; GFX11-PAL-GCNTRACKERS:    NumVgprs: 64
 ; GFX11-PAL:    ScratchSize: 2780
diff --git a/llvm/test/CodeGen/AMDGPU/scmp.ll b/llvm/test/CodeGen/AMDGPU/scmp.ll
index 6ad3fee5344cc..a5f4bb5e81d4a 100644
--- a/llvm/test/CodeGen/AMDGPU/scmp.ll
+++ b/llvm/test/CodeGen/AMDGPU/scmp.ll
@@ -601,11 +601,11 @@ define i32 @scmp_i128_uniform(i128 inreg %a, i128 inreg %b) {
 ; GFX7-SDAG-LABEL: scmp_i128_uniform:
 ; GFX7-SDAG:       ; %bb.0:
 ; GFX7-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX7-SDAG-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-SDAG-NEXT:    v_mov_b32_e32 v1, s23
+; GFX7-SDAG-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-SDAG-NEXT:    v_cmp_gt_i64_e32 vcc, s[18:19], v[0:1]
-; GFX7-SDAG-NEXT:    v_mov_b32_e32 v2, s20
 ; GFX7-SDAG-NEXT:    v_mov_b32_e32 v3, s21
+; GFX7-SDAG-NEXT:    v_mov_b32_e32 v2, s20
 ; GFX7-SDAG-NEXT:    s_and_b64 s[6:7], vcc, exec
 ; GFX7-SDAG-NEXT:    v_cmp_gt_u64_e32 vcc, s[16:17], v[2:3]
 ; GFX7-SDAG-NEXT:    v_cmp_eq_u64_e64 s[4:5], s[18:19], v[0:1]
@@ -733,8 +733,8 @@ define i32 @scmp_i64_uniform(i64 inreg %a, i64 inreg %b) {
 ; GFX7-SDAG-LABEL: scmp_i64_uniform:
 ; GFX7-SDAG:       ; %bb.0:
 ; GFX7-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX7-SDAG-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX7-SDAG-NEXT:    v_mov_b32_e32 v1, s19
+; GFX7-SDAG-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX7-SDAG-NEXT:    v_cmp_gt_i64_e32 vcc, s[16:17], v[0:1]
 ; GFX7-SDAG-NEXT:    s_and_b64 s[4:5], vcc, exec
 ; GFX7-SDAG-NEXT:    v_cmp_ge_i64_e32 vcc, s[16:17], v[0:1]
diff --git a/llvm/test/CodeGen/AMDGPU/sdiv64.ll b/llvm/test/CodeGen/AMDGPU/sdiv64.ll
index 38f93f4df18da..92072fc530b61 100644
--- a/llvm/test/CodeGen/AMDGPU/sdiv64.ll
+++ b/llvm/test/CodeGen/AMDGPU/sdiv64.ll
@@ -1559,10 +1559,10 @@ define amdgpu_kernel void @s_test_sdiv_k_num_i64(ptr addrspace(1) %out, i64 %x)
 ; GCN-NEXT:    s_cselect_b32 s7, s14, 0
 ; GCN-NEXT:    s_cselect_b32 s6, s12, s10
 ; GCN-NEXT:    s_xor_b64 s[6:7], s[6:7], s[4:5]
-; GCN-NEXT:    s_sub_u32 s6, s6, s4
-; GCN-NEXT:    s_subb_u32 s7, s7, s4
-; GCN-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NEXT:    v_mov_b32_e32 v1, s7
+; GCN-NEXT:    s_sub_u32 s5, s6, s4
+; GCN-NEXT:    s_subb_u32 s4, s7, s4
+; GCN-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NEXT:    v_mov_b32_e32 v1, s4
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/sext-divergence-driven-isel.ll b/llvm/test/CodeGen/AMDGPU/sext-divergence-driven-isel.ll
index e685d8914b5ca..54f456bbf4822 100644
--- a/llvm/test/CodeGen/AMDGPU/sext-divergence-driven-isel.ll
+++ b/llvm/test/CodeGen/AMDGPU/sext-divergence-driven-isel.ll
@@ -94,11 +94,11 @@ define amdgpu_kernel void @sext_i32_to_i64_uniform(ptr addrspace(1) %out, i32 %a
 ; GCN-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NEXT:    s_ashr_i32 s5, s8, 31
-; GCN-NEXT:    s_add_u32 s4, s6, s8
-; GCN-NEXT:    s_addc_u32 s5, s7, s5
-; GCN-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NEXT:    v_mov_b32_e32 v1, s5
+; GCN-NEXT:    s_ashr_i32 s4, s8, 31
+; GCN-NEXT:    s_add_u32 s5, s6, s8
+; GCN-NEXT:    s_addc_u32 s4, s7, s4
+; GCN-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NEXT:    v_mov_b32_e32 v1, s4
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
   %sext = sext i32 %a to i64
diff --git a/llvm/test/CodeGen/AMDGPU/shl.ll b/llvm/test/CodeGen/AMDGPU/shl.ll
index 2543727a4237f..b13ad70c89979 100644
--- a/llvm/test/CodeGen/AMDGPU/shl.ll
+++ b/llvm/test/CodeGen/AMDGPU/shl.ll
@@ -1096,8 +1096,8 @@ define amdgpu_kernel void @v_shl_32_i64(ptr addrspace(1) %out, ptr addrspace(1)
 ; VI-NEXT:    s_load_dword s2, s[2:3], 0x0
 ; VI-NEXT:    s_add_u32 s0, s0, s4
 ; VI-NEXT:    s_addc_u32 s1, s1, s5
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v1, s2
 ; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
diff --git a/llvm/test/CodeGen/AMDGPU/sra.ll b/llvm/test/CodeGen/AMDGPU/sra.ll
index 66807cc42c12f..943b3023d04d2 100644
--- a/llvm/test/CodeGen/AMDGPU/sra.ll
+++ b/llvm/test/CodeGen/AMDGPU/sra.ll
@@ -716,11 +716,11 @@ define amdgpu_kernel void @s_ashr_32_i64(ptr addrspace(1) %out, [8 x i32], i64 %
 ; SI-NEXT:    s_mov_b32 s3, 0xf000
 ; SI-NEXT:    s_mov_b32 s2, -1
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-NEXT:    s_ashr_i32 s5, s8, 31
-; SI-NEXT:    s_add_u32 s4, s8, s6
-; SI-NEXT:    s_addc_u32 s5, s5, s7
-; SI-NEXT:    v_mov_b32_e32 v0, s4
-; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    s_ashr_i32 s4, s8, 31
+; SI-NEXT:    s_add_u32 s5, s8, s6
+; SI-NEXT:    s_addc_u32 s4, s4, s7
+; SI-NEXT:    v_mov_b32_e32 v0, s5
+; SI-NEXT:    v_mov_b32_e32 v1, s4
 ; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; SI-NEXT:    s_endpgm
 ;
@@ -732,11 +732,11 @@ define amdgpu_kernel void @s_ashr_32_i64(ptr addrspace(1) %out, [8 x i32], i64 %
 ; VI-NEXT:    s_mov_b32 s3, 0xf000
 ; VI-NEXT:    s_mov_b32 s2, -1
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-NEXT:    s_ashr_i32 s5, s8, 31
-; VI-NEXT:    s_add_u32 s4, s8, s6
-; VI-NEXT:    s_addc_u32 s5, s5, s7
-; VI-NEXT:    v_mov_b32_e32 v0, s4
-; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_ashr_i32 s4, s8, 31
+; VI-NEXT:    s_add_u32 s5, s8, s6
+; VI-NEXT:    s_addc_u32 s4, s4, s7
+; VI-NEXT:    v_mov_b32_e32 v0, s5
+; VI-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; VI-NEXT:    s_endpgm
 ;
@@ -1088,11 +1088,11 @@ define amdgpu_kernel void @s_ashr_63_i64(ptr addrspace(1) %out, [8 x i32], i64 %
 ; SI-NEXT:    s_mov_b32 s3, 0xf000
 ; SI-NEXT:    s_mov_b32 s2, -1
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-NEXT:    s_ashr_i32 s5, s8, 31
-; SI-NEXT:    s_add_u32 s4, s5, s6
-; SI-NEXT:    s_addc_u32 s5, s5, s7
-; SI-NEXT:    v_mov_b32_e32 v0, s4
-; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    s_ashr_i32 s4, s8, 31
+; SI-NEXT:    s_add_u32 s5, s4, s6
+; SI-NEXT:    s_addc_u32 s4, s4, s7
+; SI-NEXT:    v_mov_b32_e32 v0, s5
+; SI-NEXT:    v_mov_b32_e32 v1, s4
 ; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; SI-NEXT:    s_endpgm
 ;
@@ -1104,11 +1104,11 @@ define amdgpu_kernel void @s_ashr_63_i64(ptr addrspace(1) %out, [8 x i32], i64 %
 ; VI-NEXT:    s_mov_b32 s3, 0xf000
 ; VI-NEXT:    s_mov_b32 s2, -1
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
-; VI-NEXT:    s_ashr_i32 s5, s8, 31
-; VI-NEXT:    s_add_u32 s4, s5, s6
-; VI-NEXT:    s_addc_u32 s5, s5, s7
-; VI-NEXT:    v_mov_b32_e32 v0, s4
-; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_ashr_i32 s4, s8, 31
+; VI-NEXT:    s_add_u32 s5, s4, s6
+; VI-NEXT:    s_addc_u32 s4, s4, s7
+; VI-NEXT:    v_mov_b32_e32 v0, s5
+; VI-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; VI-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/srem.ll b/llvm/test/CodeGen/AMDGPU/srem.ll
index 56b232d90b4c0..8bfb485805d56 100644
--- a/llvm/test/CodeGen/AMDGPU/srem.ll
+++ b/llvm/test/CodeGen/AMDGPU/srem.ll
@@ -6210,24 +6210,24 @@ define amdgpu_kernel void @srem_v4i64(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; TONGA-NEXT:    v_mov_b32_e32 v0, s6
 ; TONGA-NEXT:    v_mov_b32_e32 v1, s7
 ; TONGA-NEXT:    s_addc_u32 s3, s7, 0
-; TONGA-NEXT:    flat_load_dwordx4 v[14:17], v[0:1]
+; TONGA-NEXT:    flat_load_dwordx4 v[10:13], v[0:1]
 ; TONGA-NEXT:    v_mov_b32_e32 v0, s2
 ; TONGA-NEXT:    v_mov_b32_e32 v1, s3
-; TONGA-NEXT:    flat_load_dwordx4 v[10:13], v[0:1]
+; TONGA-NEXT:    flat_load_dwordx4 v[14:17], v[0:1]
 ; TONGA-NEXT:    v_mov_b32_e32 v0, s0
-; TONGA-NEXT:    v_mov_b32_e32 v1, s1
 ; TONGA-NEXT:    s_add_u32 s0, s6, 16
+; TONGA-NEXT:    v_mov_b32_e32 v1, s1
 ; TONGA-NEXT:    s_addc_u32 s1, s7, 0
-; TONGA-NEXT:    v_mov_b32_e32 v5, s1
 ; TONGA-NEXT:    v_mov_b32_e32 v4, s0
+; TONGA-NEXT:    v_mov_b32_e32 v5, s1
 ; TONGA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; TONGA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; TONGA-NEXT:    s_waitcnt vmcnt(3)
-; TONGA-NEXT:    v_readfirstlane_b32 s3, v15
-; TONGA-NEXT:    v_readfirstlane_b32 s2, v14
+; TONGA-NEXT:    v_readfirstlane_b32 s3, v11
+; TONGA-NEXT:    v_readfirstlane_b32 s2, v10
 ; TONGA-NEXT:    s_waitcnt vmcnt(2)
-; TONGA-NEXT:    v_readfirstlane_b32 s1, v11
-; TONGA-NEXT:    v_readfirstlane_b32 s0, v10
+; TONGA-NEXT:    v_readfirstlane_b32 s1, v15
+; TONGA-NEXT:    v_readfirstlane_b32 s0, v14
 ; TONGA-NEXT:    s_or_b64 s[6:7], s[2:3], s[0:1]
 ; TONGA-NEXT:    s_cmp_lg_u32 s7, 0
 ; TONGA-NEXT:    s_cbranch_scc0 .LBB12_2
@@ -6384,77 +6384,77 @@ define amdgpu_kernel void @srem_v4i64(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; TONGA-NEXT:    v_mov_b32_e32 v9, s7
 ; TONGA-NEXT:    v_mov_b32_e32 v8, s6
 ; TONGA-NEXT:  .LBB12_6: ; %.split
-; TONGA-NEXT:    v_or_b32_e32 v10, v17, v13
+; TONGA-NEXT:    v_or_b32_e32 v10, v13, v17
 ; TONGA-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v10
 ; TONGA-NEXT:    s_cbranch_vccz .LBB12_8
 ; TONGA-NEXT:  ; %bb.7:
-; TONGA-NEXT:    v_ashrrev_i32_e32 v10, 31, v13
-; TONGA-NEXT:    v_add_u32_e32 v11, vcc, v12, v10
-; TONGA-NEXT:    v_addc_u32_e32 v13, vcc, v13, v10, vcc
+; TONGA-NEXT:    v_ashrrev_i32_e32 v10, 31, v17
+; TONGA-NEXT:    v_add_u32_e32 v11, vcc, v16, v10
+; TONGA-NEXT:    v_addc_u32_e32 v14, vcc, v17, v10, vcc
 ; TONGA-NEXT:    v_xor_b32_e32 v11, v11, v10
-; TONGA-NEXT:    v_xor_b32_e32 v10, v13, v10
-; TONGA-NEXT:    v_cvt_f32_u32_e32 v13, v11
-; TONGA-NEXT:    v_cvt_f32_u32_e32 v14, v10
-; TONGA-NEXT:    v_sub_u32_e32 v22, vcc, 0, v11
-; TONGA-NEXT:    v_subb_u32_e32 v23, vcc, 0, v10, vcc
-; TONGA-NEXT:    v_madmk_f32 v13, v14, 0x4f800000, v13
-; TONGA-NEXT:    v_rcp_f32_e32 v13, v13
+; TONGA-NEXT:    v_xor_b32_e32 v10, v14, v10
+; TONGA-NEXT:    v_cvt_f32_u32_e32 v14, v11
+; TONGA-NEXT:    v_cvt_f32_u32_e32 v15, v10
+; TONGA-NEXT:    v_sub_u32_e32 v21, vcc, 0, v11
+; TONGA-NEXT:    v_subb_u32_e32 v22, vcc, 0, v10, vcc
+; TONGA-NEXT:    v_madmk_f32 v14, v15, 0x4f800000, v14
+; TONGA-NEXT:    v_rcp_f32_e32 v14, v14
 ; TONGA-NEXT:    s_mov_b64 s[6:7], 0
-; TONGA-NEXT:    v_mul_f32_e32 v13, 0x5f7ffffc, v13
-; TONGA-NEXT:    v_mul_f32_e32 v14, 0x2f800000, v13
-; TONGA-NEXT:    v_trunc_f32_e32 v14, v14
-; TONGA-NEXT:    v_madmk_f32 v13, v14, 0xcf800000, v13
+; TONGA-NEXT:    v_mul_f32_e32 v14, 0x5f7ffffc, v14
+; TONGA-NEXT:    v_mul_f32_e32 v15, 0x2f800000, v14
+; TONGA-NEXT:    v_trunc_f32_e32 v15, v15
+; TONGA-NEXT:    v_madmk_f32 v14, v15, 0xcf800000, v14
+; TONGA-NEXT:    v_cvt_u32_f32_e32 v19, v15
 ; TONGA-NEXT:    v_cvt_u32_f32_e32 v20, v14
-; TONGA-NEXT:    v_cvt_u32_f32_e32 v21, v13
-; TONGA-NEXT:    v_mul_lo_u32 v15, v22, v20
-; TONGA-NEXT:    v_mad_u64_u32 v[13:14], s[0:1], v22, v21, 0
-; TONGA-NEXT:    v_mul_lo_u32 v18, v23, v21
-; TONGA-NEXT:    v_add_u32_e32 v14, vcc, v14, v15
-; TONGA-NEXT:    v_add_u32_e32 v24, vcc, v14, v18
-; TONGA-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v21, v24, 0
-; TONGA-NEXT:    v_mul_hi_u32 v25, v21, v13
-; TONGA-NEXT:    v_mad_u64_u32 v[18:19], s[0:1], v20, v13, 0
-; TONGA-NEXT:    v_add_u32_e32 v25, vcc, v25, v14
-; TONGA-NEXT:    v_mad_u64_u32 v[13:14], s[0:1], v20, v24, 0
+; TONGA-NEXT:    v_mul_lo_u32 v17, v21, v19
+; TONGA-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v21, v20, 0
+; TONGA-NEXT:    v_mul_lo_u32 v18, v22, v20
+; TONGA-NEXT:    v_add_u32_e32 v15, vcc, v15, v17
+; TONGA-NEXT:    v_add_u32_e32 v23, vcc, v15, v18
+; TONGA-NEXT:    v_mad_u64_u32 v[17:18], s[0:1], v20, v23, 0
+; TONGA-NEXT:    v_mul_hi_u32 v24, v20, v14
+; TONGA-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v19, v14, 0
+; TONGA-NEXT:    v_add_u32_e32 v24, vcc, v24, v17
+; TONGA-NEXT:    v_addc_u32_e32 v25, vcc, 0, v18, vcc
+; TONGA-NEXT:    v_mad_u64_u32 v[17:18], s[0:1], v19, v23, 0
+; TONGA-NEXT:    v_add_u32_e32 v14, vcc, v24, v14
+; TONGA-NEXT:    v_addc_u32_e32 v14, vcc, v25, v15, vcc
+; TONGA-NEXT:    v_addc_u32_e32 v15, vcc, 0, v18, vcc
+; TONGA-NEXT:    v_add_u32_e32 v14, vcc, v14, v17
 ; TONGA-NEXT:    v_addc_u32_e32 v15, vcc, 0, v15, vcc
-; TONGA-NEXT:    v_add_u32_e32 v18, vcc, v25, v18
-; TONGA-NEXT:    v_addc_u32_e32 v15, vcc, v15, v19, vcc
-; TONGA-NEXT:    v_addc_u32_e32 v14, vcc, 0, v14, vcc
-; TONGA-NEXT:    v_add_u32_e32 v13, vcc, v15, v13
-; TONGA-NEXT:    v_addc_u32_e32 v14, vcc, 0, v14, vcc
-; TONGA-NEXT:    v_add_u32_e32 v21, vcc, v21, v13
-; TONGA-NEXT:    v_addc_u32_e32 v20, vcc, v20, v14, vcc
-; TONGA-NEXT:    v_mad_u64_u32 v[13:14], s[0:1], v22, v21, 0
-; TONGA-NEXT:    v_mul_lo_u32 v15, v22, v20
-; TONGA-NEXT:    v_mul_lo_u32 v18, v23, v21
-; TONGA-NEXT:    v_mul_hi_u32 v23, v21, v13
-; TONGA-NEXT:    v_add_u32_e32 v14, vcc, v15, v14
-; TONGA-NEXT:    v_add_u32_e32 v22, vcc, v18, v14
-; TONGA-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v21, v22, 0
-; TONGA-NEXT:    v_mad_u64_u32 v[18:19], s[0:1], v20, v13, 0
-; TONGA-NEXT:    v_add_u32_e32 v23, vcc, v23, v14
-; TONGA-NEXT:    v_mad_u64_u32 v[13:14], s[0:1], v20, v22, 0
+; TONGA-NEXT:    v_add_u32_e32 v20, vcc, v20, v14
+; TONGA-NEXT:    v_addc_u32_e32 v19, vcc, v19, v15, vcc
+; TONGA-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v21, v20, 0
+; TONGA-NEXT:    v_mul_lo_u32 v17, v21, v19
+; TONGA-NEXT:    v_mul_lo_u32 v18, v22, v20
+; TONGA-NEXT:    v_mul_hi_u32 v22, v20, v14
+; TONGA-NEXT:    v_add_u32_e32 v15, vcc, v17, v15
+; TONGA-NEXT:    v_add_u32_e32 v21, vcc, v18, v15
+; TONGA-NEXT:    v_mad_u64_u32 v[17:18], s[0:1], v20, v21, 0
+; TONGA-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v19, v14, 0
+; TONGA-NEXT:    v_add_u32_e32 v22, vcc, v22, v17
+; TONGA-NEXT:    v_addc_u32_e32 v23, vcc, 0, v18, vcc
+; TONGA-NEXT:    v_mad_u64_u32 v[17:18], s[0:1], v19, v21, 0
+; TONGA-NEXT:    v_add_u32_e32 v14, vcc, v22, v14
+; TONGA-NEXT:    v_addc_u32_e32 v14, vcc, v23, v15, vcc
+; TONGA-NEXT:    v_addc_u32_e32 v15, vcc, 0, v18, vcc
+; TONGA-NEXT:    v_add_u32_e32 v14, vcc, v14, v17
 ; TONGA-NEXT:    v_addc_u32_e32 v15, vcc, 0, v15, vcc
-; TONGA-NEXT:    v_add_u32_e32 v18, vcc, v23, v18
-; TONGA-NEXT:    v_addc_u32_e32 v15, vcc, v15, v19, vcc
-; TONGA-NEXT:    v_addc_u32_e32 v14, vcc, 0, v14, vcc
-; TONGA-NEXT:    v_add_u32_e32 v13, vcc, v15, v13
-; TONGA-NEXT:    v_addc_u32_e32 v14, vcc, 0, v14, vcc
-; TONGA-NEXT:    v_add_u32_e32 v15, vcc, v21, v13
-; TONGA-NEXT:    v_addc_u32_e32 v19, vcc, v20, v14, vcc
-; TONGA-NEXT:    v_ashrrev_i32_e32 v20, 31, v17
-; TONGA-NEXT:    v_add_u32_e32 v13, vcc, v16, v20
-; TONGA-NEXT:    v_xor_b32_e32 v21, v13, v20
-; TONGA-NEXT:    v_mad_u64_u32 v[13:14], s[0:1], v21, v19, 0
-; TONGA-NEXT:    v_mul_hi_u32 v22, v21, v15
-; TONGA-NEXT:    v_addc_u32_e32 v17, vcc, v17, v20, vcc
-; TONGA-NEXT:    v_xor_b32_e32 v23, v17, v20
-; TONGA-NEXT:    v_mad_u64_u32 v[17:18], s[0:1], v23, v15, 0
-; TONGA-NEXT:    v_add_u32_e32 v15, vcc, v22, v13
-; TONGA-NEXT:    v_addc_u32_e32 v22, vcc, 0, v14, vcc
-; TONGA-NEXT:    v_mad_u64_u32 v[13:14], s[0:1], v23, v19, 0
-; TONGA-NEXT:    v_add_u32_e32 v15, vcc, v15, v17
-; TONGA-NEXT:    v_addc_u32_e32 v15, vcc, v22, v18, vcc
+; TONGA-NEXT:    v_add_u32_e32 v17, vcc, v20, v14
+; TONGA-NEXT:    v_addc_u32_e32 v15, vcc, v19, v15, vcc
+; TONGA-NEXT:    v_ashrrev_i32_e32 v19, 31, v13
+; TONGA-NEXT:    v_add_u32_e32 v14, vcc, v12, v19
+; TONGA-NEXT:    v_xor_b32_e32 v20, v14, v19
+; TONGA-NEXT:    v_addc_u32_e32 v18, vcc, v13, v19, vcc
+; TONGA-NEXT:    v_mad_u64_u32 v[13:14], s[0:1], v20, v15, 0
+; TONGA-NEXT:    v_mul_hi_u32 v21, v20, v17
+; TONGA-NEXT:    v_xor_b32_e32 v22, v18, v19
+; TONGA-NEXT:    v_mad_u64_u32 v[17:18], s[0:1], v22, v17, 0
+; TONGA-NEXT:    v_add_u32_e32 v21, vcc, v21, v13
+; TONGA-NEXT:    v_addc_u32_e32 v23, vcc, 0, v14, vcc
+; TONGA-NEXT:    v_mad_u64_u32 v[13:14], s[0:1], v22, v15, 0
+; TONGA-NEXT:    v_add_u32_e32 v15, vcc, v21, v17
+; TONGA-NEXT:    v_addc_u32_e32 v15, vcc, v23, v18, vcc
 ; TONGA-NEXT:    v_addc_u32_e32 v14, vcc, 0, v14, vcc
 ; TONGA-NEXT:    v_add_u32_e32 v15, vcc, v15, v13
 ; TONGA-NEXT:    v_addc_u32_e32 v13, vcc, 0, v14, vcc
@@ -6463,22 +6463,22 @@ define amdgpu_kernel void @srem_v4i64(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; TONGA-NEXT:    v_mul_lo_u32 v15, v10, v15
 ; TONGA-NEXT:    v_add_u32_e32 v14, vcc, v17, v14
 ; TONGA-NEXT:    v_add_u32_e32 v14, vcc, v15, v14
-; TONGA-NEXT:    v_sub_u32_e32 v15, vcc, v23, v14
-; TONGA-NEXT:    v_sub_u32_e32 v13, vcc, v21, v13
+; TONGA-NEXT:    v_sub_u32_e32 v15, vcc, v22, v14
+; TONGA-NEXT:    v_sub_u32_e32 v13, vcc, v20, v13
 ; TONGA-NEXT:    v_subb_u32_e64 v15, s[0:1], v15, v10, vcc
 ; TONGA-NEXT:    v_sub_u32_e64 v17, s[0:1], v13, v11
 ; TONGA-NEXT:    v_subbrev_u32_e64 v18, s[2:3], 0, v15, s[0:1]
 ; TONGA-NEXT:    v_cmp_ge_u32_e64 s[2:3], v18, v10
-; TONGA-NEXT:    v_cndmask_b32_e64 v19, 0, -1, s[2:3]
+; TONGA-NEXT:    v_cndmask_b32_e64 v20, 0, -1, s[2:3]
 ; TONGA-NEXT:    v_cmp_ge_u32_e64 s[2:3], v17, v11
 ; TONGA-NEXT:    v_cndmask_b32_e64 v21, 0, -1, s[2:3]
 ; TONGA-NEXT:    v_cmp_eq_u32_e64 s[2:3], v18, v10
 ; TONGA-NEXT:    v_subb_u32_e64 v15, s[0:1], v15, v10, s[0:1]
-; TONGA-NEXT:    v_cndmask_b32_e64 v19, v19, v21, s[2:3]
+; TONGA-NEXT:    v_cndmask_b32_e64 v20, v20, v21, s[2:3]
 ; TONGA-NEXT:    v_sub_u32_e64 v21, s[0:1], v17, v11
 ; TONGA-NEXT:    v_subbrev_u32_e64 v15, s[0:1], 0, v15, s[0:1]
-; TONGA-NEXT:    v_subb_u32_e32 v14, vcc, v23, v14, vcc
-; TONGA-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v19
+; TONGA-NEXT:    v_subb_u32_e32 v14, vcc, v22, v14, vcc
+; TONGA-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v20
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v14, v10
 ; TONGA-NEXT:    v_cndmask_b32_e64 v15, v18, v15, s[0:1]
 ; TONGA-NEXT:    v_cndmask_b32_e64 v18, 0, -1, vcc
@@ -6490,10 +6490,10 @@ define amdgpu_kernel void @srem_v4i64(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; TONGA-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v10
 ; TONGA-NEXT:    v_cndmask_b32_e32 v11, v13, v17, vcc
 ; TONGA-NEXT:    v_cndmask_b32_e32 v10, v14, v15, vcc
-; TONGA-NEXT:    v_xor_b32_e32 v11, v11, v20
-; TONGA-NEXT:    v_xor_b32_e32 v13, v10, v20
-; TONGA-NEXT:    v_sub_u32_e32 v10, vcc, v11, v20
-; TONGA-NEXT:    v_subb_u32_e32 v11, vcc, v13, v20, vcc
+; TONGA-NEXT:    v_xor_b32_e32 v11, v11, v19
+; TONGA-NEXT:    v_xor_b32_e32 v13, v10, v19
+; TONGA-NEXT:    v_sub_u32_e32 v10, vcc, v11, v19
+; TONGA-NEXT:    v_subb_u32_e32 v11, vcc, v13, v19, vcc
 ; TONGA-NEXT:    s_branch .LBB12_9
 ; TONGA-NEXT:  .LBB12_8:
 ; TONGA-NEXT:    s_mov_b64 s[6:7], -1
@@ -6503,22 +6503,22 @@ define amdgpu_kernel void @srem_v4i64(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; TONGA-NEXT:    s_cmp_lg_u32 s0, 1
 ; TONGA-NEXT:    s_cbranch_scc1 .LBB12_11
 ; TONGA-NEXT:  ; %bb.10:
-; TONGA-NEXT:    v_cvt_f32_u32_e32 v10, v12
-; TONGA-NEXT:    v_sub_u32_e32 v11, vcc, 0, v12
+; TONGA-NEXT:    v_cvt_f32_u32_e32 v10, v16
+; TONGA-NEXT:    v_sub_u32_e32 v11, vcc, 0, v16
 ; TONGA-NEXT:    v_rcp_iflag_f32_e32 v10, v10
 ; TONGA-NEXT:    v_mul_f32_e32 v10, 0x4f7ffffe, v10
 ; TONGA-NEXT:    v_cvt_u32_f32_e32 v10, v10
 ; TONGA-NEXT:    v_mul_lo_u32 v11, v11, v10
 ; TONGA-NEXT:    v_mul_hi_u32 v11, v10, v11
 ; TONGA-NEXT:    v_add_u32_e32 v10, vcc, v10, v11
-; TONGA-NEXT:    v_mul_hi_u32 v10, v16, v10
-; TONGA-NEXT:    v_mul_lo_u32 v10, v10, v12
-; TONGA-NEXT:    v_sub_u32_e32 v10, vcc, v16, v10
-; TONGA-NEXT:    v_subrev_u32_e32 v11, vcc, v12, v10
-; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v10, v12
+; TONGA-NEXT:    v_mul_hi_u32 v10, v12, v10
+; TONGA-NEXT:    v_mul_lo_u32 v10, v10, v16
+; TONGA-NEXT:    v_sub_u32_e32 v10, vcc, v12, v10
+; TONGA-NEXT:    v_subrev_u32_e32 v11, vcc, v16, v10
+; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v10, v16
 ; TONGA-NEXT:    v_cndmask_b32_e32 v10, v10, v11, vcc
-; TONGA-NEXT:    v_subrev_u32_e32 v11, vcc, v12, v10
-; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v10, v12
+; TONGA-NEXT:    v_subrev_u32_e32 v11, vcc, v16, v10
+; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v10, v16
 ; TONGA-NEXT:    v_cndmask_b32_e32 v10, v10, v11, vcc
 ; TONGA-NEXT:    v_mov_b32_e32 v11, 0
 ; TONGA-NEXT:  .LBB12_11: ; %.split.split
@@ -9130,16 +9130,16 @@ define amdgpu_kernel void @srem_v4i64_4(ptr addrspace(1) %out, ptr addrspace(1)
 ; TONGA-NEXT:    v_mov_b32_e32 v1, s3
 ; TONGA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; TONGA-NEXT:    s_add_u32 s4, s2, 16
-; TONGA-NEXT:    s_addc_u32 s5, s3, 0
+; TONGA-NEXT:    s_addc_u32 s2, s3, 0
 ; TONGA-NEXT:    v_mov_b32_e32 v4, s4
-; TONGA-NEXT:    v_mov_b32_e32 v5, s5
+; TONGA-NEXT:    v_mov_b32_e32 v5, s2
 ; TONGA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; TONGA-NEXT:    v_mov_b32_e32 v9, s1
 ; TONGA-NEXT:    v_mov_b32_e32 v8, s0
 ; TONGA-NEXT:    s_add_u32 s0, s0, 16
 ; TONGA-NEXT:    s_addc_u32 s1, s1, 0
-; TONGA-NEXT:    v_mov_b32_e32 v11, s1
 ; TONGA-NEXT:    v_mov_b32_e32 v10, s0
+; TONGA-NEXT:    v_mov_b32_e32 v11, s1
 ; TONGA-NEXT:    s_waitcnt vmcnt(1)
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v12, 31, v1
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v12, 30, v12
diff --git a/llvm/test/CodeGen/AMDGPU/srem64.ll b/llvm/test/CodeGen/AMDGPU/srem64.ll
index fd713082c1f97..8fe159aa034bb 100644
--- a/llvm/test/CodeGen/AMDGPU/srem64.ll
+++ b/llvm/test/CodeGen/AMDGPU/srem64.ll
@@ -1008,10 +1008,10 @@ define amdgpu_kernel void @s_test_srem32_64(ptr addrspace(1) %out, i64 %x, i64 %
 ; GCN-NEXT:    s_cselect_b32 s7, s12, s8
 ; GCN-NEXT:    s_cselect_b32 s6, s13, s10
 ; GCN-NEXT:    s_xor_b64 s[6:7], s[6:7], s[4:5]
-; GCN-NEXT:    s_sub_u32 s6, s6, s4
-; GCN-NEXT:    s_subb_u32 s7, s7, s4
-; GCN-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NEXT:    v_mov_b32_e32 v1, s7
+; GCN-NEXT:    s_sub_u32 s5, s6, s4
+; GCN-NEXT:    s_subb_u32 s4, s7, s4
+; GCN-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NEXT:    v_mov_b32_e32 v1, s4
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/ssubo.ll b/llvm/test/CodeGen/AMDGPU/ssubo.ll
index 7d08ac6222ecc..c1bbf83b62428 100644
--- a/llvm/test/CodeGen/AMDGPU/ssubo.ll
+++ b/llvm/test/CodeGen/AMDGPU/ssubo.ll
@@ -55,8 +55,8 @@ define amdgpu_kernel void @ssubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 %b)
 ; VI-NEXT:    s_cselect_b64 s[0:1], 1, 0
 ; VI-NEXT:    s_add_u32 s0, s4, s0
 ; VI-NEXT:    s_addc_u32 s1, s2, s1
-; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; VI-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/store-weird-sizes.ll b/llvm/test/CodeGen/AMDGPU/store-weird-sizes.ll
index 5fd6b75662efb..3032112ff09bb 100644
--- a/llvm/test/CodeGen/AMDGPU/store-weird-sizes.ll
+++ b/llvm/test/CodeGen/AMDGPU/store-weird-sizes.ll
@@ -81,21 +81,21 @@ define amdgpu_kernel void @local_store_i55(ptr addrspace(3) %ptr, i55 %arg) #0 {
 ; FIJI-NEXT:    s_mov_b32 flat_scratch_lo, s13
 ; FIJI-NEXT:    s_mov_b32 m0, -1
 ; FIJI-NEXT:    s_waitcnt lgkmcnt(0)
-; FIJI-NEXT:    s_and_b32 s4, s1, 0xffff
-; FIJI-NEXT:    s_add_u32 s2, s8, 14
-; FIJI-NEXT:    s_addc_u32 s3, s9, 0
-; FIJI-NEXT:    v_mov_b32_e32 v0, s2
-; FIJI-NEXT:    v_mov_b32_e32 v1, s3
+; FIJI-NEXT:    s_and_b32 s2, s1, 0xffff
+; FIJI-NEXT:    s_add_u32 s3, s8, 14
+; FIJI-NEXT:    s_addc_u32 s4, s9, 0
+; FIJI-NEXT:    v_mov_b32_e32 v0, s3
+; FIJI-NEXT:    v_mov_b32_e32 v1, s4
 ; FIJI-NEXT:    flat_load_ubyte v0, v[0:1]
-; FIJI-NEXT:    s_load_dword s2, s[8:9], 0x0
+; FIJI-NEXT:    s_load_dword s3, s[8:9], 0x0
 ; FIJI-NEXT:    v_mov_b32_e32 v2, s1
 ; FIJI-NEXT:    v_mov_b32_e32 v3, s0
 ; FIJI-NEXT:    s_waitcnt lgkmcnt(0)
-; FIJI-NEXT:    v_mov_b32_e32 v1, s2
+; FIJI-NEXT:    v_mov_b32_e32 v1, s3
 ; FIJI-NEXT:    ds_write_b16 v1, v2 offset:4
 ; FIJI-NEXT:    s_waitcnt vmcnt(0)
 ; FIJI-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
-; FIJI-NEXT:    v_or_b32_e32 v0, s4, v0
+; FIJI-NEXT:    v_or_b32_e32 v0, s2, v0
 ; FIJI-NEXT:    v_bfe_u32 v0, v0, 16, 7
 ; FIJI-NEXT:    ds_write_b8 v1, v0 offset:6
 ; FIJI-NEXT:    ds_write_b32 v1, v3
diff --git a/llvm/test/CodeGen/AMDGPU/sub.ll b/llvm/test/CodeGen/AMDGPU/sub.ll
index 96e1a70d94f19..2aa8072df77dd 100644
--- a/llvm/test/CodeGen/AMDGPU/sub.ll
+++ b/llvm/test/CodeGen/AMDGPU/sub.ll
@@ -324,8 +324,8 @@ define amdgpu_kernel void @test_sub_v4i32(ptr addrspace(1) %out, ptr addrspace(1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    s_add_u32 s2, s2, 16
 ; GFX8-NEXT:    s_addc_u32 s3, s3, 0
-; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v4, s2
+; GFX8-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX8-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GFX8-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GFX8-NEXT:    v_mov_b32_e32 v8, s0
@@ -669,8 +669,8 @@ define amdgpu_kernel void @s_sub_i64(ptr addrspace(1) noalias %out, i64 %a, i64
 ; GFX8-NEXT:    s_sub_u32 s0, s2, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    s_subb_u32 s1, s3, s5
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
+; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
 ; GFX8-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/ucmp.ll b/llvm/test/CodeGen/AMDGPU/ucmp.ll
index 2eef0cd806efb..6a97bc7c06734 100644
--- a/llvm/test/CodeGen/AMDGPU/ucmp.ll
+++ b/llvm/test/CodeGen/AMDGPU/ucmp.ll
@@ -596,11 +596,11 @@ define i32 @ucmp_i128_uniform(i128 inreg %a, i128 inreg %b) {
 ; GFX7-SDAG-LABEL: ucmp_i128_uniform:
 ; GFX7-SDAG:       ; %bb.0:
 ; GFX7-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX7-SDAG-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-SDAG-NEXT:    v_mov_b32_e32 v1, s23
+; GFX7-SDAG-NEXT:    v_mov_b32_e32 v0, s22
 ; GFX7-SDAG-NEXT:    v_cmp_gt_u64_e32 vcc, s[18:19], v[0:1]
-; GFX7-SDAG-NEXT:    v_mov_b32_e32 v2, s20
 ; GFX7-SDAG-NEXT:    v_mov_b32_e32 v3, s21
+; GFX7-SDAG-NEXT:    v_mov_b32_e32 v2, s20
 ; GFX7-SDAG-NEXT:    s_and_b64 s[6:7], vcc, exec
 ; GFX7-SDAG-NEXT:    v_cmp_gt_u64_e32 vcc, s[16:17], v[2:3]
 ; GFX7-SDAG-NEXT:    v_cmp_eq_u64_e64 s[4:5], s[18:19], v[0:1]
@@ -728,8 +728,8 @@ define i32 @ucmp_i64_uniform(i64 inreg %a, i64 inreg %b) {
 ; GFX7-SDAG-LABEL: ucmp_i64_uniform:
 ; GFX7-SDAG:       ; %bb.0:
 ; GFX7-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX7-SDAG-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX7-SDAG-NEXT:    v_mov_b32_e32 v1, s19
+; GFX7-SDAG-NEXT:    v_mov_b32_e32 v0, s18
 ; GFX7-SDAG-NEXT:    v_cmp_gt_u64_e32 vcc, s[16:17], v[0:1]
 ; GFX7-SDAG-NEXT:    s_and_b64 s[4:5], vcc, exec
 ; GFX7-SDAG-NEXT:    v_cmp_ge_u64_e32 vcc, s[16:17], v[0:1]
diff --git a/llvm/test/CodeGen/AMDGPU/udiv.ll b/llvm/test/CodeGen/AMDGPU/udiv.ll
index 872b8f6120e96..5796b6975d275 100644
--- a/llvm/test/CodeGen/AMDGPU/udiv.ll
+++ b/llvm/test/CodeGen/AMDGPU/udiv.ll
@@ -1752,18 +1752,18 @@ define amdgpu_kernel void @v_udiv_i23(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; GCN-NEXT:    s_add_u32 s4, s2, 4
 ; GCN-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-NEXT:    s_add_u32 s6, s2, 2
-; GCN-NEXT:    s_addc_u32 s7, s3, 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s4
+; GCN-NEXT:    s_addc_u32 s4, s3, 0
 ; GCN-NEXT:    v_mov_b32_e32 v1, s5
-; GCN-NEXT:    s_add_u32 s4, s2, 6
-; GCN-NEXT:    s_addc_u32 s5, s3, 0
-; GCN-NEXT:    v_mov_b32_e32 v2, s4
-; GCN-NEXT:    v_mov_b32_e32 v3, s5
+; GCN-NEXT:    s_add_u32 s5, s2, 6
+; GCN-NEXT:    s_addc_u32 s7, s3, 0
+; GCN-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-NEXT:    v_mov_b32_e32 v3, s7
 ; GCN-NEXT:    flat_load_ubyte v4, v[2:3]
 ; GCN-NEXT:    flat_load_ushort v5, v[0:1]
 ; GCN-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-NEXT:    v_mov_b32_e32 v2, s6
-; GCN-NEXT:    v_mov_b32_e32 v3, s7
+; GCN-NEXT:    v_mov_b32_e32 v3, s4
 ; GCN-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-NEXT:    flat_load_ubyte v2, v[2:3]
 ; GCN-NEXT:    flat_load_ushort v0, v[0:1]
@@ -2001,18 +2001,18 @@ define amdgpu_kernel void @v_udiv_i24(ptr addrspace(1) %out, ptr addrspace(1) %i
 ; GCN-NEXT:    s_add_u32 s4, s2, 4
 ; GCN-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-NEXT:    s_add_u32 s6, s2, 2
-; GCN-NEXT:    s_addc_u32 s7, s3, 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s4
+; GCN-NEXT:    s_addc_u32 s4, s3, 0
 ; GCN-NEXT:    v_mov_b32_e32 v1, s5
-; GCN-NEXT:    s_add_u32 s4, s2, 6
-; GCN-NEXT:    s_addc_u32 s5, s3, 0
-; GCN-NEXT:    v_mov_b32_e32 v2, s4
-; GCN-NEXT:    v_mov_b32_e32 v3, s5
+; GCN-NEXT:    s_add_u32 s5, s2, 6
+; GCN-NEXT:    s_addc_u32 s7, s3, 0
+; GCN-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-NEXT:    v_mov_b32_e32 v3, s7
 ; GCN-NEXT:    flat_load_ubyte v4, v[2:3]
 ; GCN-NEXT:    flat_load_ushort v5, v[0:1]
 ; GCN-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-NEXT:    v_mov_b32_e32 v2, s6
-; GCN-NEXT:    v_mov_b32_e32 v3, s7
+; GCN-NEXT:    v_mov_b32_e32 v3, s4
 ; GCN-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-NEXT:    flat_load_ubyte v2, v[2:3]
 ; GCN-NEXT:    flat_load_ushort v0, v[0:1]
diff --git a/llvm/test/CodeGen/AMDGPU/uint_to_fp.f64.ll b/llvm/test/CodeGen/AMDGPU/uint_to_fp.f64.ll
index d7aea1bc2c327..3e023eead78a4 100644
--- a/llvm/test/CodeGen/AMDGPU/uint_to_fp.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/uint_to_fp.f64.ll
@@ -214,8 +214,8 @@ define amdgpu_kernel void @s_uint_to_fp_v4i64_to_v4f64(ptr addrspace(1) %out, <4
 ; SI-NEXT:    v_add_f64 v[6:7], v[6:7], v[4:5]
 ; SI-NEXT:    v_add_f64 v[4:5], v[8:9], v[12:13]
 ; SI-NEXT:    s_addc_u32 s1, s9, 0
-; SI-NEXT:    v_mov_b32_e32 v9, s1
 ; SI-NEXT:    v_mov_b32_e32 v8, s0
+; SI-NEXT:    v_mov_b32_e32 v9, s1
 ; SI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; SI-NEXT:    s_nop 0
 ; SI-NEXT:    v_mov_b32_e32 v4, s8
@@ -249,8 +249,8 @@ define amdgpu_kernel void @s_uint_to_fp_v4i64_to_v4f64(ptr addrspace(1) %out, <4
 ; VI-NEXT:    v_add_f64 v[0:1], v[10:11], v[14:15]
 ; VI-NEXT:    s_add_u32 s0, s8, 16
 ; VI-NEXT:    s_addc_u32 s1, s9, 0
-; VI-NEXT:    v_mov_b32_e32 v11, s1
 ; VI-NEXT:    v_mov_b32_e32 v10, s0
+; VI-NEXT:    v_mov_b32_e32 v11, s1
 ; VI-NEXT:    v_mov_b32_e32 v8, s8
 ; VI-NEXT:    v_mov_b32_e32 v9, s9
 ; VI-NEXT:    flat_store_dwordx4 v[10:11], v[4:7]
@@ -374,8 +374,8 @@ define amdgpu_kernel void @s_uint_to_fp_v4i32_to_v4f64(ptr addrspace(1) %out, <4
 ; SI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s0
 ; SI-NEXT:    s_add_u32 s0, s4, 16
 ; SI-NEXT:    s_addc_u32 s1, s5, 0
-; SI-NEXT:    v_mov_b32_e32 v9, s1
 ; SI-NEXT:    v_mov_b32_e32 v8, s0
+; SI-NEXT:    v_mov_b32_e32 v9, s1
 ; SI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; SI-NEXT:    s_nop 0
 ; SI-NEXT:    v_mov_b32_e32 v4, s4
@@ -397,8 +397,8 @@ define amdgpu_kernel void @s_uint_to_fp_v4i32_to_v4f64(ptr addrspace(1) %out, <4
 ; VI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s0
 ; VI-NEXT:    s_add_u32 s0, s4, 16
 ; VI-NEXT:    s_addc_u32 s1, s5, 0
-; VI-NEXT:    v_mov_b32_e32 v9, s1
 ; VI-NEXT:    v_mov_b32_e32 v8, s0
+; VI-NEXT:    v_mov_b32_e32 v9, s1
 ; VI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
 ; VI-NEXT:    s_nop 0
 ; VI-NEXT:    v_mov_b32_e32 v4, s4
diff --git a/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll b/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll
index 06d4c88c03cef..aa03672666262 100644
--- a/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll
+++ b/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll
@@ -1260,7 +1260,7 @@ define amdgpu_gfx_whole_wave void @inreg_args(i1 %active, i32 inreg %i32, <4 x i
 ; GISEL-NEXT:    s_wait_samplecnt 0x0
 ; GISEL-NEXT:    s_wait_bvhcnt 0x0
 ; GISEL-NEXT:    s_wait_kmcnt 0x0
-; GISEL-NEXT:    s_xor_saveexec_b32 s34, -1
+; GISEL-NEXT:    s_xor_saveexec_b32 s0, -1
 ; GISEL-NEXT:    s_clause 0x5 ; 24-byte Folded Spill
 ; GISEL-NEXT:    scratch_store_b32 off, v0, s32
 ; GISEL-NEXT:    scratch_store_b32 off, v1, s32 offset:4
@@ -1269,19 +1269,15 @@ define amdgpu_gfx_whole_wave void @inreg_args(i1 %active, i32 inreg %i32, <4 x i
 ; GISEL-NEXT:    scratch_store_b32 off, v4, s32 offset:16
 ; GISEL-NEXT:    scratch_store_b32 off, v5, s32 offset:20
 ; GISEL-NEXT:    s_mov_b32 exec_lo, -1
-; GISEL-NEXT:    s_mov_b32 s0, s5
-; GISEL-NEXT:    s_mov_b32 s1, s6
-; GISEL-NEXT:    s_mov_b32 s2, s7
-; GISEL-NEXT:    s_mov_b32 s3, s8
-; GISEL-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GISEL-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v1, s1
-; GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
-; GISEL-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v5, s9
+; GISEL-NEXT:    v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v1, s6
+; GISEL-NEXT:    v_dual_mov_b32 v0, s5 :: v_dual_mov_b32 v3, s8
+; GISEL-NEXT:    v_dual_mov_b32 v2, s7 :: v_dual_mov_b32 v5, s9
 ; GISEL-NEXT:    scratch_store_b32 off, v4, s10
 ; GISEL-NEXT:    s_clause 0x1
 ; GISEL-NEXT:    scratch_store_b128 off, v[0:3], s11
 ; GISEL-NEXT:    scratch_store_b32 off, v5, s11
-; GISEL-NEXT:    s_xor_b32 exec_lo, s34, -1
+; GISEL-NEXT:    s_wait_alu depctr_sa_sdst(0)
+; GISEL-NEXT:    s_xor_b32 exec_lo, s0, -1
 ; GISEL-NEXT:    s_clause 0x5 ; 24-byte Folded Reload
 ; GISEL-NEXT:    scratch_load_b32 v0, off, s32
 ; GISEL-NEXT:    scratch_load_b32 v1, off, s32 offset:4
@@ -1289,7 +1285,7 @@ define amdgpu_gfx_whole_wave void @inreg_args(i1 %active, i32 inreg %i32, <4 x i
 ; GISEL-NEXT:    scratch_load_b32 v3, off, s32 offset:12
 ; GISEL-NEXT:    scratch_load_b32 v4, off, s32 offset:16
 ; GISEL-NEXT:    scratch_load_b32 v5, off, s32 offset:20
-; GISEL-NEXT:    s_mov_b32 exec_lo, s34
+; GISEL-NEXT:    s_mov_b32 exec_lo, s0
 ; GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
@@ -1339,7 +1335,7 @@ define amdgpu_gfx_whole_wave void @inreg_args(i1 %active, i32 inreg %i32, <4 x i
 ; GISEL64-NEXT:    s_wait_samplecnt 0x0
 ; GISEL64-NEXT:    s_wait_bvhcnt 0x0
 ; GISEL64-NEXT:    s_wait_kmcnt 0x0
-; GISEL64-NEXT:    s_xor_saveexec_b64 s[34:35], -1
+; GISEL64-NEXT:    s_xor_saveexec_b64 s[0:1], -1
 ; GISEL64-NEXT:    s_clause 0x5 ; 24-byte Folded Spill
 ; GISEL64-NEXT:    scratch_store_b32 off, v0, s32
 ; GISEL64-NEXT:    scratch_store_b32 off, v1, s32 offset:4
@@ -1348,22 +1344,18 @@ define amdgpu_gfx_whole_wave void @inreg_args(i1 %active, i32 inreg %i32, <4 x i
 ; GISEL64-NEXT:    scratch_store_b32 off, v4, s32 offset:16
 ; GISEL64-NEXT:    scratch_store_b32 off, v5, s32 offset:20
 ; GISEL64-NEXT:    s_mov_b64 exec, -1
-; GISEL64-NEXT:    s_mov_b32 s0, s5
-; GISEL64-NEXT:    s_mov_b32 s1, s6
-; GISEL64-NEXT:    s_mov_b32 s2, s7
-; GISEL64-NEXT:    s_mov_b32 s3, s8
 ; GISEL64-NEXT:    v_mov_b32_e32 v4, s4
-; GISEL64-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GISEL64-NEXT:    v_mov_b32_e32 v0, s0
-; GISEL64-NEXT:    v_mov_b32_e32 v1, s1
-; GISEL64-NEXT:    v_mov_b32_e32 v2, s2
-; GISEL64-NEXT:    v_mov_b32_e32 v3, s3
+; GISEL64-NEXT:    v_mov_b32_e32 v0, s5
+; GISEL64-NEXT:    v_mov_b32_e32 v1, s6
+; GISEL64-NEXT:    v_mov_b32_e32 v2, s7
+; GISEL64-NEXT:    v_mov_b32_e32 v3, s8
 ; GISEL64-NEXT:    v_mov_b32_e32 v5, s9
 ; GISEL64-NEXT:    scratch_store_b32 off, v4, s10
 ; GISEL64-NEXT:    s_clause 0x1
 ; GISEL64-NEXT:    scratch_store_b128 off, v[0:3], s11
 ; GISEL64-NEXT:    scratch_store_b32 off, v5, s11
-; GISEL64-NEXT:    s_xor_b64 exec, s[34:35], -1
+; GISEL64-NEXT:    s_wait_alu depctr_sa_sdst(0)
+; GISEL64-NEXT:    s_xor_b64 exec, s[0:1], -1
 ; GISEL64-NEXT:    s_clause 0x5 ; 24-byte Folded Reload
 ; GISEL64-NEXT:    scratch_load_b32 v0, off, s32
 ; GISEL64-NEXT:    scratch_load_b32 v1, off, s32 offset:4
@@ -1371,7 +1363,7 @@ define amdgpu_gfx_whole_wave void @inreg_args(i1 %active, i32 inreg %i32, <4 x i
 ; GISEL64-NEXT:    scratch_load_b32 v3, off, s32 offset:12
 ; GISEL64-NEXT:    scratch_load_b32 v4, off, s32 offset:16
 ; GISEL64-NEXT:    scratch_load_b32 v5, off, s32 offset:20
-; GISEL64-NEXT:    s_mov_b64 exec, s[34:35]
+; GISEL64-NEXT:    s_mov_b64 exec, s[0:1]
 ; GISEL64-NEXT:    s_wait_loadcnt 0x0
 ; GISEL64-NEXT:    s_setpc_b64 s[30:31]
 ;



More information about the llvm-commits mailing list