[DAG] Enable combineShiftOfShiftedLogic folds after type legalization

[RKSimon]

Summary

This was disabled to prevent regressions, which appear to be just occurring on AMDGPU (at least in our current lit tests).

Fixes #57872

[uweigand]

The SystemZ change is a clear improvement, so that part LGTM. Thanks!

[yonghong-song]

Thanks. The change looks good to me from BPF perspective as it fixed the regression in https://github.com/llvm/llvm-project/issues/57872.

[RKSimon]

@foad What do you think about the AMDGPU diffs? Do you want me to focus on getting D136081 done first?

[foad]

In D136042#3864932, @RKSimon wrote:

@foad What do you think about the AMDGPU diffs? Do you want me to focus on getting D136081 done first?

The diffs look OK to me, modulo one pretty minor regression noted inline.

As for the general approach of putting the smarts in AMDGPUTargetLowering::isDesirableToCommuteWithShift: does AMDGPU have special code to match or(shl(load_zext(),c), load_zext()) and merge it into a wider load? If so, could it be improved to match the case where both loads are shifted? On the other hand, since you've already coded and tested this approach, I think it is fine.

llvm/test/CodeGen/AMDGPU/sdiv.ll

1652 ↗

(On Diff #468086)

Seems like we are losing a couple of bfe formations here, and D136081 does not fix it.

[RKSimon]

Extend AMDGPUTargetLowering::isDesirableToCommuteWithShift to avoid losing SR*(SHL(X,C1),C2) patterns that can fold to BFE instructions

[RKSimon]

@foad I gave up on D136081 and instead just tried harder not to lose patterns that the existing BFE match code can handle.

[RKSimon]

minor cleanup (remove unnecessary check for ISD::SHL)

[RKSimon]

ping? I think everything is covered now

[foad]

AMDGPU parts look fine, thanks, and sorry for the late review.

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

SDValue DAGCombiner::visitShiftByConstant(SDNode *N) {
   if (isBitwiseNot(N->getOperand(0)))
     return SDValue();
 
   // The inner binop must be one-use, since we want to replace it.
   SDValue LHS = N->getOperand(0);
   if (!LHS.hasOneUse() || !TLI.isDesirableToCommuteWithShift(N, Level))
     return SDValue();
 
-  // TODO: This is limited to early combining because it may reveal regressions
-  //       otherwise. But since we just checked a target hook to see if this is
-  //       desirable, that should have filtered out cases where this interferes
-  //       with some other pattern matching.
-  if (!LegalTypes)
-    if (SDValue R = combineShiftOfShiftedLogic(N, DAG))
-      return R;
+  // Fold shift(bitop(shift(x,c1),y), c2) -> bitop(shift(x,c1+c2),shift(y,c2)).
 
   // We want to pull some binops through shifts, so that we have (and (shift))
   // instead of (shift (and)), likewise for add, or, xor, etc.  This sort of
   // thing happens with address calculations, so it's important to canonicalize
   // it.
   switch (LHS.getOpcode()) {
   default:
     return SDValue();

llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h

public:
 
   SDValue getNegatedExpression(SDValue Op, SelectionDAG &DAG,
                                bool LegalOperations, bool ForCodeSize,
                                NegatibleCost &Cost,
                                unsigned Depth) const override;
 
   bool isNarrowingProfitable(EVT VT1, EVT VT2) const override;
 
+  bool isDesirableToCommuteWithShift(const SDNode *N,
+                                     CombineLevel Level) const override;
+
   EVT getTypeForExtReturn(LLVMContext &Context, EVT VT,
                           ISD::NodeType ExtendKind) const override;
 
   MVT getVectorIdxTy(const DataLayout &) const override;
   bool isSelectSupported(SelectSupportKind) const override;
 
   bool isFPImmLegal(const APFloat &Imm, EVT VT,
                     bool ForCodeSize) const override;

llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp

bool AMDGPUTargetLowering::isNarrowingProfitable(EVT SrcVT, EVT DestVT) const {
   // limited number of native 64-bit operations. Shrinking an operation to fit
   // in a single 32-bit register should always be helpful. As currently used,
   // this is much less general than the name suggests, and is only used in
   // places trying to reduce the sizes of loads. Shrinking loads to < 32-bits is
   // not profitable, and may actually be harmful.
   return SrcVT.getSizeInBits() > 32 && DestVT.getSizeInBits() == 32;
 }
 
+bool AMDGPUTargetLowering::isDesirableToCommuteWithShift(
+    const SDNode* N, CombineLevel Level) const {
+  assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
+          N->getOpcode() == ISD::SRL) &&
+         "Expected shift op");
+  // Always commute pre-type legalization and right shifts.
+  // We're looking for shl(or(x,y),z) patterns.
+  if (Level < CombineLevel::AfterLegalizeTypes ||
+      N->getOpcode() != ISD::SHL || N->getOperand(0).getOpcode() != ISD::OR)
+    return true;
+
+  // If only user is a i32 right-shift, then don't destroy a BFE pattern.
+  if (N->getOpcode() == ISD::SHL && N->getValueType(0) == MVT::i32 &&
+      N->use_size() == 1 &&
+      (N->use_begin()->getOpcode() == ISD::SRA ||
+       N->use_begin()->getOpcode() == ISD::SRL))
+    return false;
+
+  // Don't destroy or(shl(load_zext(),c), load_zext()) patterns.
+  auto IsShiftAndLoad = [](SDValue LHS, SDValue RHS) {
+    if (LHS.getOpcode() != ISD::SHL)
+      return false;
+    auto *RHSLd = dyn_cast<LoadSDNode>(RHS);
+    auto *LHS0 = dyn_cast<LoadSDNode>(LHS.getOperand(0));
+    auto *LHS1 = dyn_cast<ConstantSDNode>(LHS.getOperand(1));
+    return LHS0 && LHS1 && RHSLd && LHS0->getExtensionType() == ISD::ZEXTLOAD &&
+           LHS1->getAPIntValue() == LHS0->getMemoryVT().getScalarSizeInBits() &&
+           RHSLd->getExtensionType() == ISD::ZEXTLOAD;
+  };
+  SDValue LHS = N->getOperand(0).getOperand(0);
+  SDValue RHS = N->getOperand(0).getOperand(1);
+  return !(IsShiftAndLoad(LHS, RHS) || IsShiftAndLoad(RHS, LHS));
+}
+
 //===---------------------------------------------------------------------===//
 // TargetLowering Callbacks
 //===---------------------------------------------------------------------===//
 
 CCAssignFn *AMDGPUCallLowering::CCAssignFnForCall(CallingConv::ID CC,
                                                   bool IsVarArg) {
   switch (CC) {
   case CallingConv::AMDGPU_VS:

llvm/test/CodeGen/AMDGPU/cvt_f32_ubyte.ll

 ; SI-NEXT:    s_mov_b32 s2, -1
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    s_mov_b32 s0, s6
 ; SI-NEXT:    s_mov_b32 s1, s7
 ; SI-NEXT:    s_mov_b32 s6, s2
 ; SI-NEXT:    s_mov_b32 s7, s3
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_lshrrev_b32_e32 v5, 16, v4
-; SI-NEXT:    v_lshrrev_b32_e32 v6, 24, v4
-; SI-NEXT:    v_and_b32_e32 v7, 0xff00, v4
 ; SI-NEXT:    v_cvt_f32_ubyte3_e32 v3, v4
 ; SI-NEXT:    v_cvt_f32_ubyte2_e32 v2, v4
 ; SI-NEXT:    v_cvt_f32_ubyte1_e32 v1, v4
 ; SI-NEXT:    v_cvt_f32_ubyte0_e32 v0, v4
-; SI-NEXT:    v_add_i32_e32 v4, vcc, 9, v4
+; SI-NEXT:    v_add_i32_e32 v7, vcc, 9, v4
+; SI-NEXT:    v_and_b32_e32 v6, 0xff00, v4
 ; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; SI-NEXT:    s_waitcnt expcnt(0)
-; SI-NEXT:    v_and_b32_e32 v0, 0xff, v4
-; SI-NEXT:    v_add_i32_e32 v2, vcc, 9, v5
-; SI-NEXT:    v_lshlrev_b32_e32 v1, 8, v6
-; SI-NEXT:    v_or_b32_e32 v0, v7, v0
-; SI-NEXT:    v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT:    v_and_b32_e32 v0, 0xff, v7
+; SI-NEXT:    v_add_i32_e32 v1, vcc, 9, v5
+; SI-NEXT:    v_or_b32_e32 v0, v6, v0
+; SI-NEXT:    v_and_b32_e32 v1, 0xff, v1
+; SI-NEXT:    v_and_b32_e32 v4, 0xff000000, v4
 ; SI-NEXT:    v_add_i32_e32 v0, vcc, 0x900, v0
-; SI-NEXT:    v_or_b32_e32 v1, v1, v2
-; SI-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; SI-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; SI-NEXT:    v_or_b32_e32 v1, v4, v1
 ; SI-NEXT:    v_or_b32_e32 v0, v1, v0
 ; SI-NEXT:    v_add_i32_e32 v0, vcc, 0x9000000, v0
 ; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; SI-NEXT:    s_endpgm
 ;
 ; VI-LABEL: load_v4i8_to_v4f32_2_uses:
 ; VI:       ; %bb.0:
 ; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34

llvm/test/CodeGen/AMDGPU/fast-unaligned-load-store.global.ll

; GFX11-NEXT:    s_endpgm
   ret void
 }
 
 ; Should produce align 1 dword when legal
 define i32 @global_load_2xi16_align1(i16 addrspace(1)* %p) #0 {
 ; GFX7-ALIGNED-LABEL: global_load_2xi16_align1:
 ; GFX7-ALIGNED:       ; %bb.0:
 ; GFX7-ALIGNED-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX7-ALIGNED-NEXT:    v_add_i32_e32 v2, vcc, 1, v0
-; GFX7-ALIGNED-NEXT:    v_addc_u32_e32 v3, vcc, 0, v1, vcc
-; GFX7-ALIGNED-NEXT:    flat_load_ubyte v4, v[2:3]
-; GFX7-ALIGNED-NEXT:    v_add_i32_e32 v2, vcc, 3, v0
+; GFX7-ALIGNED-NEXT:    v_add_i32_e32 v2, vcc, 2, v0
 ; GFX7-ALIGNED-NEXT:    v_addc_u32_e32 v3, vcc, 0, v1, vcc
-; GFX7-ALIGNED-NEXT:    flat_load_ubyte v5, v[0:1]
+; GFX7-ALIGNED-NEXT:    v_add_i32_e32 v4, vcc, 1, v0
+; GFX7-ALIGNED-NEXT:    v_addc_u32_e32 v5, vcc, 0, v1, vcc
+; GFX7-ALIGNED-NEXT:    v_add_i32_e32 v6, vcc, 3, v0
+; GFX7-ALIGNED-NEXT:    v_addc_u32_e32 v7, vcc, 0, v1, vcc
+; GFX7-ALIGNED-NEXT:    flat_load_ubyte v6, v[6:7]
+; GFX7-ALIGNED-NEXT:    flat_load_ubyte v4, v[4:5]
 ; GFX7-ALIGNED-NEXT:    flat_load_ubyte v2, v[2:3]
-; GFX7-ALIGNED-NEXT:    v_add_i32_e32 v0, vcc, 2, v0
-; GFX7-ALIGNED-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
 ; GFX7-ALIGNED-NEXT:    flat_load_ubyte v0, v[0:1]
 ; GFX7-ALIGNED-NEXT:    s_waitcnt vmcnt(3)
-; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v1, 8, v4
+; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v3, 24, v6
 ; GFX7-ALIGNED-NEXT:    s_waitcnt vmcnt(2)
-; GFX7-ALIGNED-NEXT:    v_or_b32_e32 v1, v1, v5
+; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v1, 8, v4
 ; GFX7-ALIGNED-NEXT:    s_waitcnt vmcnt(1)
-; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v2, 8, v2
+; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
 ; GFX7-ALIGNED-NEXT:    s_waitcnt vmcnt(0)
-; GFX7-ALIGNED-NEXT:    v_or_b32_e32 v0, v2, v0
-; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX7-ALIGNED-NEXT:    v_or_b32_e32 v0, v1, v0
+; GFX7-ALIGNED-NEXT:    v_or_b32_e32 v1, v3, v2
+; GFX7-ALIGNED-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX7-ALIGNED-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX7-UNALIGNED-LABEL: global_load_2xi16_align1:
 ; GFX7-UNALIGNED:       ; %bb.0:
 ; GFX7-UNALIGNED-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX7-UNALIGNED-NEXT:    flat_load_dword v0, v[0:1]
 ; GFX7-UNALIGNED-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-UNALIGNED-NEXT:    s_setpc_b64 s[30:31]

llvm/test/CodeGen/AMDGPU/fast-unaligned-load-store.private.ll

 ; Should produce align 1 dword when legal
 define i32 @private_load_2xi16_align1(i16 addrspace(5)* %p) #0 {
 ; GFX7-ALIGNED-LABEL: private_load_2xi16_align1:
 ; GFX7-ALIGNED:       ; %bb.0:
 ; GFX7-ALIGNED-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX7-ALIGNED-NEXT:    v_add_i32_e32 v1, vcc, 2, v0
 ; GFX7-ALIGNED-NEXT:    v_add_i32_e32 v2, vcc, 1, v0
 ; GFX7-ALIGNED-NEXT:    v_add_i32_e32 v3, vcc, 3, v0
-; GFX7-ALIGNED-NEXT:    buffer_load_ubyte v2, v2, s[0:3], 0 offen
 ; GFX7-ALIGNED-NEXT:    buffer_load_ubyte v3, v3, s[0:3], 0 offen
-; GFX7-ALIGNED-NEXT:    buffer_load_ubyte v0, v0, s[0:3], 0 offen
+; GFX7-ALIGNED-NEXT:    buffer_load_ubyte v2, v2, s[0:3], 0 offen
 ; GFX7-ALIGNED-NEXT:    buffer_load_ubyte v1, v1, s[0:3], 0 offen
+; GFX7-ALIGNED-NEXT:    buffer_load_ubyte v0, v0, s[0:3], 0 offen
 ; GFX7-ALIGNED-NEXT:    s_waitcnt vmcnt(3)
-; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v2, 8, v2
+; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v3, 24, v3
 ; GFX7-ALIGNED-NEXT:    s_waitcnt vmcnt(2)
-; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v3, 8, v3
+; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v2, 8, v2
 ; GFX7-ALIGNED-NEXT:    s_waitcnt vmcnt(1)
-; GFX7-ALIGNED-NEXT:    v_or_b32_e32 v0, v2, v0
+; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX7-ALIGNED-NEXT:    s_waitcnt vmcnt(0)
+; GFX7-ALIGNED-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX7-ALIGNED-NEXT:    v_or_b32_e32 v1, v3, v1
-; GFX7-ALIGNED-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX7-ALIGNED-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX7-ALIGNED-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX7-UNALIGNED-LABEL: private_load_2xi16_align1:
 ; GFX7-UNALIGNED:       ; %bb.0:
 ; GFX7-UNALIGNED-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX7-UNALIGNED-NEXT:    buffer_load_dword v0, v0, s[0:3], 0 offen
 ; GFX7-UNALIGNED-NEXT:    s_waitcnt vmcnt(0)

llvm/test/CodeGen/AMDGPU/idot8s.ll

 ; GFX7-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7-NEXT:    buffer_load_dword v2, v[0:1], s[8:11], 0 addr64
 ; GFX7-NEXT:    s_mov_b64 s[8:9], s[6:7]
 ; GFX7-NEXT:    buffer_load_dword v0, v[0:1], s[8:11], 0 addr64
 ; GFX7-NEXT:    s_mov_b32 s2, -1
 ; GFX7-NEXT:    buffer_load_ubyte v1, off, s[0:3], 0
 ; GFX7-NEXT:    s_addc_u32 s13, s13, 0
 ; GFX7-NEXT:    s_waitcnt vmcnt(2)
-; GFX7-NEXT:    v_bfe_i32 v7, v2, 0, 4
-; GFX7-NEXT:    v_bfe_i32 v3, v2, 24, 4
+; GFX7-NEXT:    v_bfe_i32 v8, v2, 0, 4
+; GFX7-NEXT:    v_ashrrev_i32_e32 v3, 28, v2
 ; GFX7-NEXT:    s_waitcnt vmcnt(1)
-; GFX7-NEXT:    v_bfe_i32 v14, v0, 0, 4
-; GFX7-NEXT:    v_bfe_i32 v4, v2, 20, 4
-; GFX7-NEXT:    v_bfe_i32 v5, v2, 16, 4
-; GFX7-NEXT:    v_bfe_i32 v6, v2, 8, 4
-; GFX7-NEXT:    v_ashrrev_i32_e32 v8, 28, v2
+; GFX7-NEXT:    v_bfe_i32 v15, v0, 0, 4
+; GFX7-NEXT:    v_bfe_i32 v4, v2, 24, 4
+; GFX7-NEXT:    v_bfe_i32 v5, v2, 20, 4
+; GFX7-NEXT:    v_bfe_i32 v6, v2, 16, 4
+; GFX7-NEXT:    v_bfe_i32 v7, v2, 8, 4
 ; GFX7-NEXT:    v_bfe_i32 v9, v2, 12, 4
 ; GFX7-NEXT:    v_bfe_i32 v2, v2, 4, 4
-; GFX7-NEXT:    v_and_b32_e32 v7, 0xff, v7
-; GFX7-NEXT:    v_bfe_i32 v10, v0, 24, 4
-; GFX7-NEXT:    v_bfe_i32 v11, v0, 20, 4
-; GFX7-NEXT:    v_bfe_i32 v12, v0, 16, 4
-; GFX7-NEXT:    v_bfe_i32 v13, v0, 8, 4
-; GFX7-NEXT:    v_ashrrev_i32_e32 v15, 28, v0
+; GFX7-NEXT:    v_and_b32_e32 v8, 0xff, v8
+; GFX7-NEXT:    v_ashrrev_i32_e32 v10, 28, v0
+; GFX7-NEXT:    v_bfe_i32 v11, v0, 24, 4
+; GFX7-NEXT:    v_bfe_i32 v12, v0, 20, 4
+; GFX7-NEXT:    v_bfe_i32 v13, v0, 16, 4
+; GFX7-NEXT:    v_bfe_i32 v14, v0, 8, 4
 ; GFX7-NEXT:    v_bfe_i32 v16, v0, 12, 4
 ; GFX7-NEXT:    v_bfe_i32 v0, v0, 4, 4
-; GFX7-NEXT:    v_and_b32_e32 v14, 0xff, v14
+; GFX7-NEXT:    v_and_b32_e32 v15, 0xff, v15
 ; GFX7-NEXT:    v_and_b32_e32 v2, 0xff, v2
 ; GFX7-NEXT:    v_and_b32_e32 v0, 0xff, v0
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
-; GFX7-NEXT:    v_mad_u32_u24 v1, v7, v14, v1
-; GFX7-NEXT:    v_and_b32_e32 v6, 0xff, v6
+; GFX7-NEXT:    v_mad_u32_u24 v1, v8, v15, v1
+; GFX7-NEXT:    v_and_b32_e32 v7, 0xff, v7
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v9, 24, v9
-; GFX7-NEXT:    v_and_b32_e32 v13, 0xff, v13
+; GFX7-NEXT:    v_and_b32_e32 v14, 0xff, v14
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v16, 24, v16
 ; GFX7-NEXT:    v_mad_u32_u24 v0, v2, v0, v1
 ; GFX7-NEXT:    v_alignbit_b32 v9, 0, v9, 24
 ; GFX7-NEXT:    v_alignbit_b32 v16, 0, v16, 24
-; GFX7-NEXT:    v_mad_u32_u24 v0, v6, v13, v0
+; GFX7-NEXT:    v_mad_u32_u24 v0, v7, v14, v0
+; GFX7-NEXT:    v_and_b32_e32 v6, 0xff, v6
+; GFX7-NEXT:    v_and_b32_e32 v13, 0xff, v13
+; GFX7-NEXT:    v_mad_u32_u24 v0, v9, v16, v0
 ; GFX7-NEXT:    v_and_b32_e32 v5, 0xff, v5
 ; GFX7-NEXT:    v_and_b32_e32 v12, 0xff, v12
-; GFX7-NEXT:    v_mad_u32_u24 v0, v9, v16, v0
+; GFX7-NEXT:    v_mad_u32_u24 v0, v6, v13, v0
 ; GFX7-NEXT:    v_and_b32_e32 v4, 0xff, v4
 ; GFX7-NEXT:    v_and_b32_e32 v11, 0xff, v11
 ; GFX7-NEXT:    v_mad_u32_u24 v0, v5, v12, v0
 ; GFX7-NEXT:    v_and_b32_e32 v3, 0xff, v3
 ; GFX7-NEXT:    v_and_b32_e32 v10, 0xff, v10
 ; GFX7-NEXT:    v_mad_u32_u24 v0, v4, v11, v0
-; GFX7-NEXT:    v_and_b32_e32 v8, 0xff, v8
-; GFX7-NEXT:    v_and_b32_e32 v15, 0xff, v15
 ; GFX7-NEXT:    v_mad_u32_u24 v0, v3, v10, v0
-; GFX7-NEXT:    v_mad_u32_u24 v0, v8, v15, v0
 ; GFX7-NEXT:    buffer_store_byte v0, off, s[0:3], 0
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: idot8_acc8_vecMul:
 ; GFX8:       ; %bb.0: ; %entry
 ; GFX8-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
 ; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0

llvm/test/CodeGen/AMDGPU/idot8u.ll

 ; GFX7-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7-NEXT:    buffer_load_dword v2, v[0:1], s[8:11], 0 addr64
 ; GFX7-NEXT:    s_mov_b64 s[8:9], s[6:7]
 ; GFX7-NEXT:    buffer_load_dword v0, v[0:1], s[8:11], 0 addr64
 ; GFX7-NEXT:    s_mov_b32 s2, -1
 ; GFX7-NEXT:    buffer_load_ubyte v1, off, s[0:3], 0
 ; GFX7-NEXT:    s_addc_u32 s13, s13, 0
 ; GFX7-NEXT:    s_waitcnt vmcnt(2)
-; GFX7-NEXT:    v_and_b32_e32 v7, 15, v2
-; GFX7-NEXT:    v_bfe_u32 v6, v2, 4, 4
+; GFX7-NEXT:    v_and_b32_e32 v8, 15, v2
+; GFX7-NEXT:    v_bfe_u32 v3, v2, 24, 4
 ; GFX7-NEXT:    s_waitcnt vmcnt(1)
-; GFX7-NEXT:    v_and_b32_e32 v14, 15, v0
-; GFX7-NEXT:    v_bfe_u32 v8, v2, 12, 4
-; GFX7-NEXT:    v_bfe_u32 v13, v0, 4, 4
-; GFX7-NEXT:    v_bfe_u32 v15, v0, 12, 4
+; GFX7-NEXT:    v_and_b32_e32 v15, 15, v0
+; GFX7-NEXT:    v_bfe_u32 v4, v2, 20, 4
+; GFX7-NEXT:    v_bfe_u32 v5, v2, 16, 4
+; GFX7-NEXT:    v_bfe_u32 v6, v2, 8, 4
+; GFX7-NEXT:    v_bfe_u32 v7, v2, 4, 4
+; GFX7-NEXT:    v_lshrrev_b32_e32 v9, 28, v2
+; GFX7-NEXT:    v_lshlrev_b32_e32 v2, 12, v2
+; GFX7-NEXT:    v_bfe_u32 v10, v0, 24, 4
+; GFX7-NEXT:    v_bfe_u32 v11, v0, 20, 4
+; GFX7-NEXT:    v_bfe_u32 v12, v0, 16, 4
+; GFX7-NEXT:    v_bfe_u32 v13, v0, 8, 4
+; GFX7-NEXT:    v_bfe_u32 v14, v0, 4, 4
+; GFX7-NEXT:    v_lshrrev_b32_e32 v16, 28, v0
+; GFX7-NEXT:    v_lshlrev_b32_e32 v0, 12, v0
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
+; GFX7-NEXT:    v_mad_u32_u24 v1, v8, v15, v1
+; GFX7-NEXT:    v_and_b32_e32 v2, 0xf000000, v2
+; GFX7-NEXT:    v_and_b32_e32 v0, 0xf000000, v0
 ; GFX7-NEXT:    v_mad_u32_u24 v1, v7, v14, v1
-; GFX7-NEXT:    v_bfe_u32 v5, v2, 8, 4
-; GFX7-NEXT:    v_bfe_u32 v12, v0, 8, 4
-; GFX7-NEXT:    v_lshlrev_b32_e32 v8, 24, v8
-; GFX7-NEXT:    v_lshlrev_b32_e32 v15, 24, v15
+; GFX7-NEXT:    v_alignbit_b32 v2, s10, v2, 24
+; GFX7-NEXT:    v_alignbit_b32 v0, 0, v0, 24
 ; GFX7-NEXT:    v_mad_u32_u24 v1, v6, v13, v1
-; GFX7-NEXT:    v_alignbit_b32 v8, 0, v8, 24
-; GFX7-NEXT:    v_alignbit_b32 v14, 0, v15, 24
-; GFX7-NEXT:    v_mad_u32_u24 v1, v5, v12, v1
-; GFX7-NEXT:    v_bfe_u32 v4, v2, 16, 4
-; GFX7-NEXT:    v_lshrrev_b32_e32 v9, 28, v2
-; GFX7-NEXT:    v_bfe_u32 v11, v0, 16, 4
-; GFX7-NEXT:    v_lshrrev_b32_e32 v16, 28, v0
-; GFX7-NEXT:    v_mad_u32_u24 v1, v8, v14, v1
-; GFX7-NEXT:    v_bfe_u32 v3, v2, 20, 4
-; GFX7-NEXT:    v_bfe_u32 v10, v0, 20, 4
-; GFX7-NEXT:    v_alignbit_b32 v2, v9, v2, 24
-; GFX7-NEXT:    v_alignbit_b32 v0, v16, v0, 24
-; GFX7-NEXT:    v_mad_u32_u24 v1, v4, v11, v1
-; GFX7-NEXT:    v_lshrrev_b32_e32 v9, 8, v2
-; GFX7-NEXT:    v_and_b32_e32 v2, 15, v2
-; GFX7-NEXT:    v_lshrrev_b32_e32 v7, 8, v0
-; GFX7-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX7-NEXT:    v_mad_u32_u24 v1, v3, v10, v1
 ; GFX7-NEXT:    v_mad_u32_u24 v0, v2, v0, v1
-; GFX7-NEXT:    v_mad_u32_u24 v0, v9, v7, v0
+; GFX7-NEXT:    v_mad_u32_u24 v0, v5, v12, v0
+; GFX7-NEXT:    v_mad_u32_u24 v0, v4, v11, v0
+; GFX7-NEXT:    v_mad_u32_u24 v0, v3, v10, v0
+; GFX7-NEXT:    v_mad_u32_u24 v0, v9, v16, v0
 ; GFX7-NEXT:    buffer_store_byte v0, off, s[0:3], 0
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: udot8_acc8_vecMul:
 ; GFX8:       ; %bb.0: ; %entry
 ; GFX8-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
 ; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0

llvm/test/CodeGen/AMDGPU/insert_vector_dynelt.ll

 ; GCN-NEXT:    s_lshr_b32 s37, s7, 17
 ; GCN-NEXT:    s_lshr_b32 s38, s7, 18
 ; GCN-NEXT:    s_lshr_b32 s39, s7, 19
 ; GCN-NEXT:    s_lshr_b32 s40, s7, 20
 ; GCN-NEXT:    s_lshr_b32 s41, s7, 21
 ; GCN-NEXT:    s_lshr_b32 s42, s7, 22
 ; GCN-NEXT:    s_lshr_b32 s43, s7, 23
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x77
-; GCN-NEXT:    v_mov_b32_e32 v16, s43
+; GCN-NEXT:    v_mov_b32_e32 v14, s43
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x76
-; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
+; GCN-NEXT:    v_cndmask_b32_e32 v14, 1, v14, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v17, s42
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v14, 3, v14
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x75
-; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
+; GCN-NEXT:    v_or_b32_e32 v14, v14, v17
 ; GCN-NEXT:    v_mov_b32_e32 v17, s41
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x74
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v18, s40
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
 ; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 3, v17
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x73
-; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
+; GCN-NEXT:    v_or_b32_e32 v14, v17, v14
 ; GCN-NEXT:    v_mov_b32_e32 v17, s39
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x72
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v18, s38
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 3, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x71
-; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
+; GCN-NEXT:    v_or_b32_e32 v17, v17, v18
 ; GCN-NEXT:    v_mov_b32_e32 v18, s37
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x70
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v19, s36
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
 ; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 3, v18
 ; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 4, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v14, 4, v14
 ; GCN-NEXT:    v_and_b32_e32 v17, 15, v17
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x7f
-; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
+; GCN-NEXT:    v_or_b32_e32 v14, v17, v14
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 7, s35
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x7e
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 6, s35
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 3, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x7d
-; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
+; GCN-NEXT:    v_or_b32_e32 v17, v17, v18
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 5, s35
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x7c
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 4, s35
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
 ; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 3, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x7b
 ; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 3, s35
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x7a
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 2, s35
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
-; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x78
-; GCN-NEXT:    v_mov_b32_e32 v14, s35
-; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
+; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x78
+; GCN-NEXT:    v_mov_b32_e32 v12, s35
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 3, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v19, 2, v19
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x79
-; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
+; GCN-NEXT:    v_or_b32_e32 v18, v18, v19
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 1, s35
-; GCN-NEXT:    v_cndmask_b32_e32 v14, 1, v14, vcc
+; GCN-NEXT:    v_cndmask_b32_e32 v12, 1, v12, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
-; GCN-NEXT:    v_and_b32_e32 v14, 1, v14
+; GCN-NEXT:    v_and_b32_e32 v12, 1, v12
 ; GCN-NEXT:    v_lshlrev_b16_e32 v19, 1, v19
-; GCN-NEXT:    v_or_b32_e32 v14, v14, v19
-; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
-; GCN-NEXT:    v_and_b32_e32 v14, 3, v14
-; GCN-NEXT:    v_or_b32_e32 v14, v14, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 4, v17
-; GCN-NEXT:    v_and_b32_e32 v14, 15, v14
-; GCN-NEXT:    v_or_b32_sdwa v14, v14, v17 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_or_b32_e32 v12, v12, v19
+; GCN-NEXT:    v_and_b32_e32 v12, 3, v12
+; GCN-NEXT:    v_or_b32_e32 v18, v12, v18
+; GCN-NEXT:    v_mov_b32_e32 v12, 15
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 12, v17
+; GCN-NEXT:    v_and_b32_sdwa v18, v18, v12 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_or_b32_e32 v17, v17, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x6f
-; GCN-NEXT:    v_or_b32_sdwa v14, v16, v14 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GCN-NEXT:    v_lshrrev_b16_e64 v16, 15, s7
+; GCN-NEXT:    v_or_b32_sdwa v14, v14, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GCN-NEXT:    v_lshrrev_b16_e64 v17, 15, s7
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x6e
-; GCN-NEXT:    v_lshrrev_b16_e64 v17, 14, s7
-; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
-; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
-; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
-; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
-; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x6d
-; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
-; GCN-NEXT:    v_lshrrev_b16_e64 v17, 13, s7
-; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
-; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x6c
-; GCN-NEXT:    v_lshrrev_b16_e64 v18, 12, s7
-; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
-; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
-; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
-; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
-; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
-; GCN-NEXT:    v_and_b32_e32 v17, 3, v17
-; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x6b
-; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
-; GCN-NEXT:    v_lshrrev_b16_e64 v17, 11, s7
-; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
-; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x6a
-; GCN-NEXT:    v_lshrrev_b16_e64 v18, 10, s7
+; GCN-NEXT:    v_lshrrev_b16_e64 v18, 14, s7
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
-; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x69
-; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
-; GCN-NEXT:    v_lshrrev_b16_e64 v18, 9, s7
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 3, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
+; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x6d
+; GCN-NEXT:    v_or_b32_e32 v17, v17, v18
+; GCN-NEXT:    v_lshrrev_b16_e64 v18, 13, s7
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
-; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x68
-; GCN-NEXT:    v_lshrrev_b16_e64 v19, 8, s7
+; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x6c
+; GCN-NEXT:    v_lshrrev_b16_e64 v19, 12, s7
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
 ; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 3, v18
+; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x6b
 ; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 4, v16
-; GCN-NEXT:    v_and_b32_e32 v17, 15, v17
+; GCN-NEXT:    v_lshrrev_b16_e64 v18, 11, s7
+; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
+; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x6a
+; GCN-NEXT:    v_lshrrev_b16_e64 v19, 10, s7
+; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
+; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
+; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
+; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 3, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v19, 2, v19
+; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x69
+; GCN-NEXT:    v_or_b32_e32 v18, v18, v19
+; GCN-NEXT:    v_lshrrev_b16_e64 v19, 9, s7
+; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
+; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x68
+; GCN-NEXT:    v_lshrrev_b16_e64 v16, 8, s7
+; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
+; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
+; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
+; GCN-NEXT:    v_lshlrev_b16_e32 v19, 1, v19
+; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
+; GCN-NEXT:    v_or_b32_e32 v16, v16, v19
+; GCN-NEXT:    v_and_b32_e32 v16, 3, v16
+; GCN-NEXT:    v_or_b32_e32 v16, v16, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 12, v17
+; GCN-NEXT:    v_and_b32_sdwa v16, v16, v12 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x67
-; GCN-NEXT:    v_or_b32_sdwa v16, v17, v16 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 7, s7
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x66
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 6, s7
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 3, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x65
-; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
+; GCN-NEXT:    v_or_b32_e32 v17, v17, v18
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 5, s7
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x64
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 4, s7
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
 ; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 3, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x63
 ; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 3, s7
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x62
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 2, s7
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 3, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v19, 2, v19
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x61
-; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
+; GCN-NEXT:    v_or_b32_e32 v18, v18, v19
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 1, s7
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x60
 ; GCN-NEXT:    v_mov_b32_e32 v15, s7
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v19, 1, v19
 ; GCN-NEXT:    v_and_b32_e32 v15, 1, v15
 ; GCN-NEXT:    v_or_b32_e32 v15, v15, v19
-; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
 ; GCN-NEXT:    v_and_b32_e32 v15, 3, v15
 ; GCN-NEXT:    v_or_b32_e32 v15, v15, v18
 ; GCN-NEXT:    v_lshlrev_b16_e32 v17, 4, v17
 ; GCN-NEXT:    v_and_b32_e32 v15, 15, v15
 ; GCN-NEXT:    v_or_b32_e32 v15, v15, v17
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x57
 ; GCN-NEXT:    v_or_b32_sdwa v15, v15, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GCN-NEXT:    v_mov_b32_e32 v16, s34
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x56
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v17, s33
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 3, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x55
-; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
+; GCN-NEXT:    v_or_b32_e32 v16, v16, v17
 ; GCN-NEXT:    v_mov_b32_e32 v17, s31
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x54
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v18, s30
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
 ; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 3, v17
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x53
 ; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
 ; GCN-NEXT:    v_mov_b32_e32 v17, s29
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x52
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v18, s28
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 3, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x51
-; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
+; GCN-NEXT:    v_or_b32_e32 v17, v17, v18
 ; GCN-NEXT:    v_mov_b32_e32 v18, s27
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x50
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v19, s26
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
 ; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 3, v18
 ; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
 ; GCN-NEXT:    v_lshlrev_b16_e32 v16, 4, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 15, v17
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x5f
 ; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 7, s25
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x5e
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 6, s25
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 3, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x5d
-; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
+; GCN-NEXT:    v_or_b32_e32 v17, v17, v18
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 5, s25
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x5c
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 4, s25
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
 ; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 3, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x5b
 ; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 3, s25
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x5a
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 2, s25
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
+; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x58
 ; GCN-NEXT:    v_mov_b32_e32 v3, s25
-; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
-; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 3, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v19, 2, v19
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x59
-; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
+; GCN-NEXT:    v_or_b32_e32 v18, v18, v19
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 1, s25
 ; GCN-NEXT:    v_cndmask_b32_e32 v3, 1, v3, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    v_and_b32_e32 v3, 1, v3
 ; GCN-NEXT:    v_lshlrev_b16_e32 v19, 1, v19
 ; GCN-NEXT:    v_or_b32_e32 v3, v3, v19
-; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
 ; GCN-NEXT:    v_and_b32_e32 v3, 3, v3
 ; GCN-NEXT:    v_or_b32_e32 v3, v3, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 4, v17
-; GCN-NEXT:    v_and_b32_e32 v3, 15, v3
-; GCN-NEXT:    v_or_b32_sdwa v3, v3, v17 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 12, v17
+; GCN-NEXT:    v_and_b32_sdwa v3, v3, v12 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_or_b32_e32 v3, v17, v3
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x4f
 ; GCN-NEXT:    v_or_b32_sdwa v16, v16, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GCN-NEXT:    v_lshrrev_b16_e64 v3, 15, s6
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x4e
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 14, s6
 ; GCN-NEXT:    v_cndmask_b32_e32 v3, 1, v3, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v3, 1, v3
 ; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v3, 3, v3
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x4d
-; GCN-NEXT:    v_or_b32_e32 v3, v17, v3
+; GCN-NEXT:    v_or_b32_e32 v3, v3, v17
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 13, s6
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x4c
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 12, s6
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
 ; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
-; GCN-NEXT:    v_lshlrev_b16_e32 v3, 2, v3
 ; GCN-NEXT:    v_and_b32_e32 v17, 3, v17
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x4b
 ; GCN-NEXT:    v_or_b32_e32 v3, v17, v3
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 11, s6
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x4a
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 10, s6
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 3, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x49
-; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
+; GCN-NEXT:    v_or_b32_e32 v17, v17, v18
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 9, s6
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x48
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 8, s6
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
 ; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    v_and_b32_e32 v18, 3, v18
 ; GCN-NEXT:    v_or_b32_e32 v17, v18, v17
-; GCN-NEXT:    v_lshlrev_b16_e32 v3, 4, v3
-; GCN-NEXT:    v_and_b32_e32 v17, 15, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v3, 12, v3
+; GCN-NEXT:    v_and_b32_sdwa v17, v17, v12 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x47
-; GCN-NEXT:    v_or_b32_sdwa v17, v17, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_or_b32_e32 v17, v3, v17
 ; GCN-NEXT:    v_lshrrev_b16_e64 v3, 7, s6
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x46
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 6, s6
 ; GCN-NEXT:    v_cndmask_b32_e32 v3, 1, v3, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v3, 1, v3
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v3, 3, v3
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x45
-; GCN-NEXT:    v_or_b32_e32 v3, v18, v3
+; GCN-NEXT:    v_or_b32_e32 v3, v3, v18
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 5, s6
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x44
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 4, s6
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
 ; GCN-NEXT:    v_or_b32_e32 v18, v19, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v3, 2, v3
 ; GCN-NEXT:    v_and_b32_e32 v18, 3, v18
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x43
 ; GCN-NEXT:    v_or_b32_e32 v18, v18, v3
 ; GCN-NEXT:    v_lshrrev_b16_e64 v3, 3, s6
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x42
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 2, s6
 ; GCN-NEXT:    v_cndmask_b32_e32 v3, 1, v3, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v3, 1, v3
 ; GCN-NEXT:    v_and_b32_e32 v19, 1, v19
+; GCN-NEXT:    v_lshlrev_b16_e32 v3, 3, v3
+; GCN-NEXT:    v_lshlrev_b16_e32 v19, 2, v19
 ; GCN-NEXT:    s_cmpk_lg_i32 s0, 0x41
-; GCN-NEXT:    v_or_b32_e32 v3, v19, v3
+; GCN-NEXT:    v_or_b32_e32 v3, v3, v19
 ; GCN-NEXT:    v_lshrrev_b16_e64 v19, 1, s6
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 64
 ; GCN-NEXT:    v_mov_b32_e32 v2, s6
 ; GCN-NEXT:    v_cndmask_b32_e32 v19, 1, v19, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v2, 1, v2, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v19, 1, v19
 ; GCN-NEXT:    v_and_b32_e32 v2, 1, v2
 ; GCN-NEXT:    v_or_b32_e32 v2, v2, v19
-; GCN-NEXT:    v_lshlrev_b16_e32 v3, 2, v3
 ; GCN-NEXT:    v_and_b32_e32 v2, 3, v2
 ; GCN-NEXT:    v_or_b32_e32 v2, v2, v3
 ; GCN-NEXT:    v_or_b32_sdwa v3, v15, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GCN-NEXT:    v_lshlrev_b16_e32 v14, 4, v18
 ; GCN-NEXT:    v_and_b32_e32 v2, 15, v2
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 55
 ; GCN-NEXT:    v_or_b32_e32 v2, v2, v14
 ; GCN-NEXT:    v_mov_b32_e32 v14, s24
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 54
 ; GCN-NEXT:    v_cndmask_b32_e32 v14, 1, v14, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v15, s23
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v14, 1, v14
 ; GCN-NEXT:    v_and_b32_e32 v15, 1, v15
+; GCN-NEXT:    v_lshlrev_b16_e32 v14, 3, v14
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 2, v15
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 53
 ; GCN-NEXT:    v_or_b32_sdwa v2, v2, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GCN-NEXT:    v_or_b32_e32 v14, v15, v14
+; GCN-NEXT:    v_or_b32_e32 v14, v14, v15
 ; GCN-NEXT:    v_mov_b32_e32 v15, s22
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 52
 ; GCN-NEXT:    v_or_b32_sdwa v2, v2, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v16, s21
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v15, 1, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
 ; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
-; GCN-NEXT:    v_lshlrev_b16_e32 v14, 2, v14
 ; GCN-NEXT:    v_and_b32_e32 v15, 3, v15
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 51
 ; GCN-NEXT:    v_or_b32_e32 v14, v15, v14
 ; GCN-NEXT:    v_mov_b32_e32 v15, s20
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 50
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v16, s19
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 1, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 3, v15
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 49
-; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
+; GCN-NEXT:    v_or_b32_e32 v15, v15, v16
 ; GCN-NEXT:    v_mov_b32_e32 v16, s18
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 48
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v17, s17
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
 ; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 2, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 3, v16
 ; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
 ; GCN-NEXT:    v_lshlrev_b16_e32 v14, 4, v14
 ; GCN-NEXT:    v_and_b32_e32 v15, 15, v15
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 63
 ; GCN-NEXT:    v_or_b32_e32 v14, v15, v14
 ; GCN-NEXT:    v_lshrrev_b16_e64 v15, 7, s16
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 62
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 6, s16
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 1, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 3, v15
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 61
-; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
+; GCN-NEXT:    v_or_b32_e32 v15, v15, v16
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 5, s16
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 60
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 4, s16
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
 ; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 2, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 3, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 59
 ; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 3, s16
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 58
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 2, s16
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
+; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 56
 ; GCN-NEXT:    v_mov_b32_e32 v13, s16
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
-; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 3, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 57
-; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
+; GCN-NEXT:    v_or_b32_e32 v16, v16, v17
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 1, s16
 ; GCN-NEXT:    v_cndmask_b32_e32 v13, 1, v13, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    v_and_b32_e32 v13, 1, v13
 ; GCN-NEXT:    v_lshlrev_b16_e32 v17, 1, v17
 ; GCN-NEXT:    v_or_b32_e32 v13, v13, v17
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    v_and_b32_e32 v13, 3, v13
 ; GCN-NEXT:    v_or_b32_e32 v13, v13, v16
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 4, v15
-; GCN-NEXT:    v_and_b32_e32 v13, 15, v13
-; GCN-NEXT:    v_or_b32_sdwa v13, v13, v15 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 12, v15
+; GCN-NEXT:    v_and_b32_sdwa v13, v13, v12 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_or_b32_e32 v13, v15, v13
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 47
 ; GCN-NEXT:    v_or_b32_sdwa v14, v14, v13 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GCN-NEXT:    v_lshrrev_b16_e64 v13, 15, s5
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 46
 ; GCN-NEXT:    v_lshrrev_b16_e64 v15, 14, s5
 ; GCN-NEXT:    v_cndmask_b32_e32 v13, 1, v13, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v13, 1, v13
 ; GCN-NEXT:    v_and_b32_e32 v15, 1, v15
+; GCN-NEXT:    v_lshlrev_b16_e32 v13, 3, v13
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 2, v15
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 45
-; GCN-NEXT:    v_or_b32_e32 v13, v15, v13
+; GCN-NEXT:    v_or_b32_e32 v13, v13, v15
 ; GCN-NEXT:    v_lshrrev_b16_e64 v15, 13, s5
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 44
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 12, s5
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v15, 1, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
 ; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
-; GCN-NEXT:    v_lshlrev_b16_e32 v13, 2, v13
 ; GCN-NEXT:    v_and_b32_e32 v15, 3, v15
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 43
 ; GCN-NEXT:    v_or_b32_e32 v13, v15, v13
 ; GCN-NEXT:    v_lshrrev_b16_e64 v15, 11, s5
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 42
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 10, s5
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 1, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 3, v15
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 41
-; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
+; GCN-NEXT:    v_or_b32_e32 v15, v15, v16
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 9, s5
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 40
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 8, s5
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
 ; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 2, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 3, v16
 ; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
-; GCN-NEXT:    v_lshlrev_b16_e32 v13, 4, v13
-; GCN-NEXT:    v_and_b32_e32 v15, 15, v15
+; GCN-NEXT:    v_lshlrev_b16_e32 v13, 12, v13
+; GCN-NEXT:    v_and_b32_sdwa v15, v15, v12 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 39
-; GCN-NEXT:    v_or_b32_sdwa v15, v15, v13 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_or_b32_e32 v15, v13, v15
 ; GCN-NEXT:    v_lshrrev_b16_e64 v13, 7, s5
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 38
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 6, s5
 ; GCN-NEXT:    v_cndmask_b32_e32 v13, 1, v13, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v13, 1, v13
 ; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v13, 3, v13
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 37
-; GCN-NEXT:    v_or_b32_e32 v13, v16, v13
+; GCN-NEXT:    v_or_b32_e32 v13, v13, v16
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 5, s5
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 36
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 4, s5
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
 ; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
-; GCN-NEXT:    v_lshlrev_b16_e32 v13, 2, v13
 ; GCN-NEXT:    v_and_b32_e32 v16, 3, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 35
 ; GCN-NEXT:    v_or_b32_e32 v16, v16, v13
 ; GCN-NEXT:    v_lshrrev_b16_e64 v13, 3, s5
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 34
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 2, s5
 ; GCN-NEXT:    v_cndmask_b32_e32 v13, 1, v13, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v13, 1, v13
 ; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v13, 3, v13
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 33
-; GCN-NEXT:    v_or_b32_e32 v17, v17, v13
+; GCN-NEXT:    v_or_b32_e32 v17, v13, v17
 ; GCN-NEXT:    v_lshrrev_b16_e64 v13, 1, s5
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 32
 ; GCN-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-NEXT:    v_cndmask_b32_e32 v13, 1, v13, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v1, 1, v1, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v13, 1, v13
 ; GCN-NEXT:    v_and_b32_e32 v1, 1, v1
 ; GCN-NEXT:    v_or_b32_e32 v1, v1, v13
-; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
 ; GCN-NEXT:    v_and_b32_e32 v1, 3, v1
 ; GCN-NEXT:    v_or_b32_e32 v1, v1, v17
 ; GCN-NEXT:    v_lshlrev_b16_e32 v16, 4, v16
 ; GCN-NEXT:    v_and_b32_e32 v1, 15, v1
 ; GCN-NEXT:    v_or_b32_e32 v1, v1, v16
 ; GCN-NEXT:    v_or_b32_sdwa v1, v1, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 23
 ; GCN-NEXT:    v_or_b32_sdwa v1, v1, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GCN-NEXT:    v_mov_b32_e32 v14, s15
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 22
 ; GCN-NEXT:    v_cndmask_b32_e32 v14, 1, v14, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v15, s14
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v14, 1, v14
 ; GCN-NEXT:    v_and_b32_e32 v15, 1, v15
+; GCN-NEXT:    v_lshlrev_b16_e32 v14, 3, v14
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 2, v15
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 21
-; GCN-NEXT:    v_or_b32_e32 v14, v15, v14
+; GCN-NEXT:    v_or_b32_e32 v14, v14, v15
 ; GCN-NEXT:    v_mov_b32_e32 v15, s13
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 20
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v16, s12
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v15, 1, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
 ; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
-; GCN-NEXT:    v_lshlrev_b16_e32 v14, 2, v14
 ; GCN-NEXT:    v_and_b32_e32 v15, 3, v15
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 19
 ; GCN-NEXT:    v_or_b32_e32 v14, v15, v14
 ; GCN-NEXT:    v_mov_b32_e32 v15, s11
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 18
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v16, s10
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 1, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 3, v15
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 17
-; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
+; GCN-NEXT:    v_or_b32_e32 v15, v15, v16
 ; GCN-NEXT:    v_mov_b32_e32 v16, s9
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 16
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    v_mov_b32_e32 v18, s8
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
 ; GCN-NEXT:    v_or_b32_e32 v16, v18, v16
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 2, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 3, v16
 ; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
 ; GCN-NEXT:    v_lshlrev_b16_e32 v14, 4, v14
 ; GCN-NEXT:    v_and_b32_e32 v15, 15, v15
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 31
 ; GCN-NEXT:    v_or_b32_e32 v14, v15, v14
 ; GCN-NEXT:    v_lshrrev_b16_e64 v15, 7, s1
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 30
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 6, s1
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 1, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 3, v15
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 29
-; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
+; GCN-NEXT:    v_or_b32_e32 v15, v15, v16
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 5, s1
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 28
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 4, s1
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
 ; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
 ; GCN-NEXT:    v_or_b32_e32 v16, v18, v16
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 2, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 3, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 27
 ; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 3, s1
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 26
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 2, s1
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
+; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 24
 ; GCN-NEXT:    v_mov_b32_e32 v17, s1
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
-; GCN-NEXT:    v_and_b32_e32 v18, 1, v18
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 3, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v18, 2, v18
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 25
-; GCN-NEXT:    v_or_b32_e32 v16, v18, v16
+; GCN-NEXT:    v_or_b32_e32 v16, v16, v18
 ; GCN-NEXT:    v_lshrrev_b16_e64 v18, 1, s1
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v18, 1, v18, vcc
 ; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
 ; GCN-NEXT:    v_lshlrev_b16_e32 v18, 1, v18
 ; GCN-NEXT:    v_or_b32_e32 v17, v17, v18
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 3, v17
 ; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 4, v15
-; GCN-NEXT:    v_and_b32_e32 v16, 15, v16
-; GCN-NEXT:    v_or_b32_sdwa v15, v16, v15 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 12, v15
+; GCN-NEXT:    v_and_b32_sdwa v16, v16, v12 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_or_b32_e32 v15, v15, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 15
 ; GCN-NEXT:    v_or_b32_sdwa v14, v14, v15 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GCN-NEXT:    v_lshrrev_b16_e64 v15, 15, s4
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 14
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 14, s4
 ; GCN-NEXT:    v_cndmask_b32_e32 v15, 1, v15, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 1, v15
 ; GCN-NEXT:    v_and_b32_e32 v16, 1, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 3, v15
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 2, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 13
-; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
+; GCN-NEXT:    v_or_b32_e32 v15, v15, v16
 ; GCN-NEXT:    v_lshrrev_b16_e64 v16, 13, s4
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 12
 ; GCN-NEXT:    v_lshrrev_b16_e64 v17, 12, s4
 ; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v16, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v17, vcc
 ; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
 ; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
 ; GCN-NEXT:    v_or_b32_e32 v16, v17, v16
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 11
-; GCN-NEXT:    v_lshrrev_b16_e64 v18, 11, s4
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 2, v15
+; GCN-NEXT:    v_lshrrev_b16_e64 v17, 11, s4
 ; GCN-NEXT:    v_and_b32_e32 v16, 3, v16
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 10
-; GCN-NEXT:    v_lshrrev_b16_e64 v13, 10, s4
+; GCN-NEXT:    v_lshrrev_b16_e64 v18, 10, s4
 ; GCN-NEXT:    v_or_b32_e32 v15, v16, v15
-; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v18, vcc
+; GCN-NEXT:    v_cndmask_b32_e32 v16, 1, v17, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 9
-; GCN-NEXT:    v_lshrrev_b16_e64 v12, 9, s4
-; GCN-NEXT:    v_cndmask_b32_e32 v13, 1, v13, vcc
+; GCN-NEXT:    v_lshrrev_b16_e64 v13, 9, s4
+; GCN-NEXT:    v_cndmask_b32_e32 v17, 1, v18, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 8
 ; GCN-NEXT:    v_lshrrev_b16_e64 v11, 8, s4
-; GCN-NEXT:    v_cndmask_b32_e32 v12, 1, v12, vcc
+; GCN-NEXT:    v_cndmask_b32_e32 v13, 1, v13, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 7
 ; GCN-NEXT:    v_lshrrev_b16_e64 v10, 7, s4
 ; GCN-NEXT:    v_cndmask_b32_e32 v11, 1, v11, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 6
 ; GCN-NEXT:    v_lshrrev_b16_e64 v9, 6, s4
 ; GCN-NEXT:    v_cndmask_b32_e32 v10, 1, v10, vcc

 ; GCN-NEXT:    v_lshrrev_b16_e64 v4, 1, s4
 ; GCN-NEXT:    v_cndmask_b32_e32 v5, 1, v5, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    s_cmp_lg_u32 s0, 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s4
 ; GCN-NEXT:    v_cndmask_b32_e32 v4, 1, v4, vcc
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, 1, v0, vcc
-; GCN-NEXT:    v_lshlrev_b16_e32 v16, 1, v16
-; GCN-NEXT:    v_and_b32_e32 v13, 1, v13
-; GCN-NEXT:    v_lshlrev_b16_e32 v12, 1, v12
+; GCN-NEXT:    v_and_b32_e32 v17, 1, v17
+; GCN-NEXT:    v_lshlrev_b16_e32 v13, 1, v13
 ; GCN-NEXT:    v_and_b32_e32 v11, 1, v11
-; GCN-NEXT:    v_lshlrev_b16_e32 v10, 1, v10
 ; GCN-NEXT:    v_and_b32_e32 v9, 1, v9
 ; GCN-NEXT:    v_lshlrev_b16_e32 v8, 1, v8
 ; GCN-NEXT:    v_and_b32_e32 v7, 1, v7
-; GCN-NEXT:    v_lshlrev_b16_e32 v6, 1, v6
 ; GCN-NEXT:    v_and_b32_e32 v5, 1, v5
 ; GCN-NEXT:    v_lshlrev_b16_e32 v4, 1, v4
 ; GCN-NEXT:    v_and_b32_e32 v0, 1, v0
-; GCN-NEXT:    v_or_b32_e32 v13, v13, v16
-; GCN-NEXT:    v_or_b32_e32 v11, v11, v12
-; GCN-NEXT:    v_or_b32_e32 v9, v9, v10
+; GCN-NEXT:    v_lshlrev_b16_e32 v16, 3, v16
+; GCN-NEXT:    v_lshlrev_b16_e32 v17, 2, v17
+; GCN-NEXT:    v_or_b32_e32 v11, v11, v13
+; GCN-NEXT:    v_lshlrev_b16_e32 v10, 3, v10
+; GCN-NEXT:    v_lshlrev_b16_e32 v9, 2, v9
 ; GCN-NEXT:    v_or_b32_e32 v7, v7, v8
-; GCN-NEXT:    v_or_b32_e32 v5, v5, v6
+; GCN-NEXT:    v_lshlrev_b16_e32 v6, 3, v6
+; GCN-NEXT:    v_lshlrev_b16_e32 v5, 2, v5
 ; GCN-NEXT:    v_or_b32_e32 v0, v0, v4
-; GCN-NEXT:    v_lshlrev_b16_e32 v13, 2, v13
+; GCN-NEXT:    v_or_b32_e32 v16, v16, v17
 ; GCN-NEXT:    v_and_b32_e32 v11, 3, v11
-; GCN-NEXT:    v_lshlrev_b16_e32 v9, 2, v9
+; GCN-NEXT:    v_or_b32_e32 v9, v10, v9
 ; GCN-NEXT:    v_and_b32_e32 v7, 3, v7
-; GCN-NEXT:    v_lshlrev_b16_e32 v5, 2, v5
+; GCN-NEXT:    v_or_b32_e32 v5, v6, v5
 ; GCN-NEXT:    v_and_b32_e32 v0, 3, v0
-; GCN-NEXT:    v_or_b32_e32 v11, v11, v13
+; GCN-NEXT:    v_or_b32_e32 v11, v11, v16
 ; GCN-NEXT:    v_or_b32_e32 v7, v7, v9
 ; GCN-NEXT:    v_or_b32_e32 v0, v0, v5
-; GCN-NEXT:    v_lshlrev_b16_e32 v15, 4, v15
-; GCN-NEXT:    v_and_b32_e32 v11, 15, v11
+; GCN-NEXT:    v_lshlrev_b16_e32 v15, 12, v15
+; GCN-NEXT:    v_and_b32_sdwa v11, v11, v12 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
 ; GCN-NEXT:    v_lshlrev_b16_e32 v7, 4, v7
 ; GCN-NEXT:    v_and_b32_e32 v0, 15, v0
-; GCN-NEXT:    v_or_b32_sdwa v11, v11, v15 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GCN-NEXT:    v_or_b32_e32 v11, v15, v11
 ; GCN-NEXT:    v_or_b32_e32 v0, v0, v7
 ; GCN-NEXT:    v_or_b32_sdwa v0, v0, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GCN-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-NEXT:    v_or_b32_sdwa v0, v0, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GCN-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-NEXT:    s_endpgm
 entry:

llvm/test/CodeGen/AMDGPU/insert_vector_elt.ll

 ; SI-NEXT:    s_load_dword s6, s[4:5], 0x8
 ; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; SI-NEXT:    s_mov_b32 s3, 0x100f000
 ; SI-NEXT:    s_mov_b32 s2, -1
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    s_lshr_b32 s4, s11, 24
 ; SI-NEXT:    s_cmp_lg_u32 s6, 15
 ; SI-NEXT:    s_cselect_b32 s4, s4, 5
-; SI-NEXT:    s_lshl_b32 s4, s4, 8
+; SI-NEXT:    s_lshl_b32 s4, s4, 24
 ; SI-NEXT:    s_lshr_b32 s5, s11, 16
 ; SI-NEXT:    s_cmp_lg_u32 s6, 14
 ; SI-NEXT:    s_cselect_b32 s5, s5, 5
 ; SI-NEXT:    s_and_b32 s5, s5, 0xff
-; SI-NEXT:    s_or_b32 s4, s5, s4
-; SI-NEXT:    s_lshl_b32 s4, s4, 16
+; SI-NEXT:    s_lshl_b32 s5, s5, 16
+; SI-NEXT:    s_or_b32 s4, s4, s5
 ; SI-NEXT:    s_lshr_b32 s5, s11, 8
 ; SI-NEXT:    s_cmp_lg_u32 s6, 13
 ; SI-NEXT:    s_cselect_b32 s5, s5, 5
 ; SI-NEXT:    s_lshl_b32 s5, s5, 8
 ; SI-NEXT:    s_cmp_lg_u32 s6, 12
 ; SI-NEXT:    s_cselect_b32 s7, s11, 5
 ; SI-NEXT:    s_and_b32 s7, s7, 0xff
 ; SI-NEXT:    s_or_b32 s5, s7, s5
 ; SI-NEXT:    s_and_b32 s5, s5, 0xffff
 ; SI-NEXT:    s_or_b32 s4, s5, s4
 ; SI-NEXT:    s_lshr_b32 s5, s10, 24
 ; SI-NEXT:    s_cmp_lg_u32 s6, 11
 ; SI-NEXT:    s_cselect_b32 s5, s5, 5
-; SI-NEXT:    s_lshl_b32 s5, s5, 8
+; SI-NEXT:    s_lshl_b32 s5, s5, 24
 ; SI-NEXT:    s_lshr_b32 s7, s10, 16
 ; SI-NEXT:    s_cmp_lg_u32 s6, 10
 ; SI-NEXT:    s_cselect_b32 s7, s7, 5
 ; SI-NEXT:    s_and_b32 s7, s7, 0xff
-; SI-NEXT:    s_or_b32 s5, s7, s5
-; SI-NEXT:    s_lshl_b32 s5, s5, 16
+; SI-NEXT:    s_lshl_b32 s7, s7, 16
+; SI-NEXT:    s_or_b32 s5, s5, s7
 ; SI-NEXT:    s_lshr_b32 s7, s10, 8
 ; SI-NEXT:    s_cmp_lg_u32 s6, 9
 ; SI-NEXT:    s_cselect_b32 s7, s7, 5
 ; SI-NEXT:    s_lshl_b32 s7, s7, 8
 ; SI-NEXT:    s_cmp_lg_u32 s6, 8
 ; SI-NEXT:    s_cselect_b32 s10, s10, 5
 ; SI-NEXT:    s_and_b32 s10, s10, 0xff
 ; SI-NEXT:    s_or_b32 s7, s10, s7
 ; SI-NEXT:    s_and_b32 s7, s7, 0xffff
 ; SI-NEXT:    s_or_b32 s5, s7, s5
 ; SI-NEXT:    s_lshr_b32 s7, s9, 24
 ; SI-NEXT:    s_cmp_lg_u32 s6, 7
 ; SI-NEXT:    s_cselect_b32 s7, s7, 5
-; SI-NEXT:    s_lshl_b32 s7, s7, 8
+; SI-NEXT:    s_lshl_b32 s7, s7, 24
 ; SI-NEXT:    s_lshr_b32 s10, s9, 16
 ; SI-NEXT:    s_cmp_lg_u32 s6, 6
 ; SI-NEXT:    s_cselect_b32 s10, s10, 5
 ; SI-NEXT:    s_and_b32 s10, s10, 0xff
-; SI-NEXT:    s_or_b32 s7, s10, s7
-; SI-NEXT:    s_lshl_b32 s7, s7, 16
+; SI-NEXT:    s_lshl_b32 s10, s10, 16
+; SI-NEXT:    s_or_b32 s7, s7, s10
 ; SI-NEXT:    s_lshr_b32 s10, s9, 8
 ; SI-NEXT:    s_cmp_lg_u32 s6, 5
 ; SI-NEXT:    s_cselect_b32 s10, s10, 5
 ; SI-NEXT:    s_lshl_b32 s10, s10, 8
 ; SI-NEXT:    s_cmp_lg_u32 s6, 4
 ; SI-NEXT:    s_cselect_b32 s9, s9, 5
 ; SI-NEXT:    s_and_b32 s9, s9, 0xff
 ; SI-NEXT:    s_or_b32 s9, s9, s10
 ; SI-NEXT:    s_and_b32 s9, s9, 0xffff
 ; SI-NEXT:    s_or_b32 s7, s9, s7
 ; SI-NEXT:    s_lshr_b32 s9, s8, 24
 ; SI-NEXT:    s_cmp_lg_u32 s6, 3
 ; SI-NEXT:    s_cselect_b32 s9, s9, 5
-; SI-NEXT:    s_lshl_b32 s9, s9, 8
+; SI-NEXT:    s_lshl_b32 s9, s9, 24
 ; SI-NEXT:    s_lshr_b32 s10, s8, 16
 ; SI-NEXT:    s_cmp_lg_u32 s6, 2
 ; SI-NEXT:    s_cselect_b32 s10, s10, 5
 ; SI-NEXT:    s_and_b32 s10, s10, 0xff
-; SI-NEXT:    s_or_b32 s9, s10, s9
-; SI-NEXT:    s_lshl_b32 s9, s9, 16
+; SI-NEXT:    s_lshl_b32 s10, s10, 16
+; SI-NEXT:    s_or_b32 s9, s9, s10
 ; SI-NEXT:    s_lshr_b32 s10, s8, 8
 ; SI-NEXT:    s_cmp_lg_u32 s6, 1
 ; SI-NEXT:    s_cselect_b32 s10, s10, 5
 ; SI-NEXT:    s_lshl_b32 s10, s10, 8
 ; SI-NEXT:    s_cmp_lg_u32 s6, 0
 ; SI-NEXT:    s_cselect_b32 s6, s8, 5
 ; SI-NEXT:    s_and_b32 s6, s6, 0xff
 ; SI-NEXT:    s_or_b32 s6, s6, s10

llvm/test/CodeGen/BPF/pr57872.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=bpf-- | FileCheck %s
-; XFAIL: *
 
 %struct.event = type { i8, [84 x i8] }
 
 define void @foo(ptr %g) {
+; CHECK-LABEL: foo:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    r1 = *(u64 *)(r1 + 0)
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 83)
+; CHECK-NEXT:    *(u8 *)(r10 - 4) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 82)
+; CHECK-NEXT:    *(u8 *)(r10 - 5) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 81)
+; CHECK-NEXT:    *(u8 *)(r10 - 6) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 80)
+; CHECK-NEXT:    *(u8 *)(r10 - 7) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 79)
+; CHECK-NEXT:    *(u8 *)(r10 - 8) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 78)
+; CHECK-NEXT:    *(u8 *)(r10 - 9) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 77)
+; CHECK-NEXT:    *(u8 *)(r10 - 10) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 76)
+; CHECK-NEXT:    *(u8 *)(r10 - 11) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 75)
+; CHECK-NEXT:    *(u8 *)(r10 - 12) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 74)
+; CHECK-NEXT:    *(u8 *)(r10 - 13) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 73)
+; CHECK-NEXT:    *(u8 *)(r10 - 14) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 72)
+; CHECK-NEXT:    *(u8 *)(r10 - 15) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 71)
+; CHECK-NEXT:    *(u8 *)(r10 - 16) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 70)
+; CHECK-NEXT:    *(u8 *)(r10 - 17) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 69)
+; CHECK-NEXT:    *(u8 *)(r10 - 18) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 68)
+; CHECK-NEXT:    *(u8 *)(r10 - 19) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 67)
+; CHECK-NEXT:    *(u8 *)(r10 - 20) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 66)
+; CHECK-NEXT:    *(u8 *)(r10 - 21) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 65)
+; CHECK-NEXT:    *(u8 *)(r10 - 22) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 64)
+; CHECK-NEXT:    *(u8 *)(r10 - 23) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 63)
+; CHECK-NEXT:    *(u8 *)(r10 - 24) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 62)
+; CHECK-NEXT:    *(u8 *)(r10 - 25) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 61)
+; CHECK-NEXT:    *(u8 *)(r10 - 26) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 60)
+; CHECK-NEXT:    *(u8 *)(r10 - 27) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 59)
+; CHECK-NEXT:    *(u8 *)(r10 - 28) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 58)
+; CHECK-NEXT:    *(u8 *)(r10 - 29) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 57)
+; CHECK-NEXT:    *(u8 *)(r10 - 30) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 56)
+; CHECK-NEXT:    *(u8 *)(r10 - 31) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 55)
+; CHECK-NEXT:    *(u8 *)(r10 - 32) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 54)
+; CHECK-NEXT:    *(u8 *)(r10 - 33) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 53)
+; CHECK-NEXT:    *(u8 *)(r10 - 34) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 52)
+; CHECK-NEXT:    *(u8 *)(r10 - 35) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 51)
+; CHECK-NEXT:    *(u8 *)(r10 - 36) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 50)
+; CHECK-NEXT:    *(u8 *)(r10 - 37) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 49)
+; CHECK-NEXT:    *(u8 *)(r10 - 38) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 48)
+; CHECK-NEXT:    *(u8 *)(r10 - 39) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 47)
+; CHECK-NEXT:    *(u8 *)(r10 - 40) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 46)
+; CHECK-NEXT:    *(u8 *)(r10 - 41) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 45)
+; CHECK-NEXT:    *(u8 *)(r10 - 42) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 44)
+; CHECK-NEXT:    *(u8 *)(r10 - 43) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 43)
+; CHECK-NEXT:    *(u8 *)(r10 - 44) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 42)
+; CHECK-NEXT:    *(u8 *)(r10 - 45) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 41)
+; CHECK-NEXT:    *(u8 *)(r10 - 46) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 40)
+; CHECK-NEXT:    *(u8 *)(r10 - 47) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 39)
+; CHECK-NEXT:    *(u8 *)(r10 - 48) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 38)
+; CHECK-NEXT:    *(u8 *)(r10 - 49) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 37)
+; CHECK-NEXT:    *(u8 *)(r10 - 50) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 36)
+; CHECK-NEXT:    *(u8 *)(r10 - 51) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 35)
+; CHECK-NEXT:    *(u8 *)(r10 - 52) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 34)
+; CHECK-NEXT:    *(u8 *)(r10 - 53) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 33)
+; CHECK-NEXT:    *(u8 *)(r10 - 54) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 32)
+; CHECK-NEXT:    *(u8 *)(r10 - 55) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 31)
+; CHECK-NEXT:    *(u8 *)(r10 - 56) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 30)
+; CHECK-NEXT:    *(u8 *)(r10 - 57) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 29)
+; CHECK-NEXT:    *(u8 *)(r10 - 58) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 28)
+; CHECK-NEXT:    *(u8 *)(r10 - 59) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 27)
+; CHECK-NEXT:    *(u8 *)(r10 - 60) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 26)
+; CHECK-NEXT:    *(u8 *)(r10 - 61) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 25)
+; CHECK-NEXT:    *(u8 *)(r10 - 62) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 24)
+; CHECK-NEXT:    *(u8 *)(r10 - 63) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 23)
+; CHECK-NEXT:    *(u8 *)(r10 - 64) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 22)
+; CHECK-NEXT:    *(u8 *)(r10 - 65) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 21)
+; CHECK-NEXT:    *(u8 *)(r10 - 66) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 20)
+; CHECK-NEXT:    *(u8 *)(r10 - 67) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 19)
+; CHECK-NEXT:    *(u8 *)(r10 - 68) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 18)
+; CHECK-NEXT:    *(u8 *)(r10 - 69) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 17)
+; CHECK-NEXT:    *(u8 *)(r10 - 70) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 16)
+; CHECK-NEXT:    *(u8 *)(r10 - 71) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 15)
+; CHECK-NEXT:    *(u8 *)(r10 - 72) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 14)
+; CHECK-NEXT:    *(u8 *)(r10 - 73) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 13)
+; CHECK-NEXT:    *(u8 *)(r10 - 74) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 12)
+; CHECK-NEXT:    *(u8 *)(r10 - 75) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 11)
+; CHECK-NEXT:    *(u8 *)(r10 - 76) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 10)
+; CHECK-NEXT:    *(u8 *)(r10 - 77) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 9)
+; CHECK-NEXT:    *(u8 *)(r10 - 78) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 8)
+; CHECK-NEXT:    *(u8 *)(r10 - 79) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 7)
+; CHECK-NEXT:    *(u8 *)(r10 - 80) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 6)
+; CHECK-NEXT:    *(u8 *)(r10 - 81) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 5)
+; CHECK-NEXT:    *(u8 *)(r10 - 82) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 4)
+; CHECK-NEXT:    *(u8 *)(r10 - 83) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 3)
+; CHECK-NEXT:    *(u8 *)(r10 - 84) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 2)
+; CHECK-NEXT:    *(u8 *)(r10 - 85) = r2
+; CHECK-NEXT:    r2 = *(u8 *)(r1 + 1)
+; CHECK-NEXT:    *(u8 *)(r10 - 86) = r2
+; CHECK-NEXT:    r1 = *(u8 *)(r1 + 0)
+; CHECK-NEXT:    *(u8 *)(r10 - 87) = r1
+; CHECK-NEXT:    r1 = r10
+; CHECK-NEXT:    r1 += -88
+; CHECK-NEXT:    call bar
+; CHECK-NEXT:    exit
 entry:
   %event = alloca %struct.event, align 1
   %hostname = getelementptr inbounds %struct.event, ptr %event, i64 0, i32 1
   %0 = load ptr, ptr %g, align 8
   call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 1 dereferenceable(84) %hostname, ptr noundef nonnull align 1 dereferenceable(84) %0, i64 84, i1 false)
   call void @bar(ptr noundef nonnull %event)
   ret void
 }
 
 declare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #2
 declare void @bar(ptr noundef)

llvm/test/CodeGen/Mips/cconv/return-struct.ll

 ; O32-LE-NEXT:    lhu $3, 4($1)
 ; O32-LE-NEXT:    jr $ra
 ; O32-LE-NEXT:    nop
 ;
 ; N32-BE-LABEL: ret_struct_3xi16:
 ; N32-BE:       # %bb.0: # %entry
 ; N32-BE-NEXT:    lui $1, %hi(struct_3xi16)
 ; N32-BE-NEXT:    lw $2, %lo(struct_3xi16)($1)
-; N32-BE-NEXT:    dsll $2, $2, 16
+; N32-BE-NEXT:    dsll $2, $2, 32
 ; N32-BE-NEXT:    addiu $1, $1, %lo(struct_3xi16)
 ; N32-BE-NEXT:    lhu $1, 4($1)
-; N32-BE-NEXT:    or $1, $1, $2
+; N32-BE-NEXT:    dsll $1, $1, 16
 ; N32-BE-NEXT:    jr $ra
-; N32-BE-NEXT:    dsll $2, $1, 16
+; N32-BE-NEXT:    or $2, $2, $1
 ;
 ; N32-LE-LABEL: ret_struct_3xi16:
 ; N32-LE:       # %bb.0: # %entry
 ; N32-LE-NEXT:    lui $1, %hi(struct_3xi16)
 ; N32-LE-NEXT:    lwu $2, %lo(struct_3xi16)($1)
 ; N32-LE-NEXT:    addiu $1, $1, %lo(struct_3xi16)
 ; N32-LE-NEXT:    lh $1, 4($1)
 ; N32-LE-NEXT:    dsll $1, $1, 32
 ; N32-LE-NEXT:    jr $ra
 ; N32-LE-NEXT:    or $2, $2, $1
 ;
 ; N64-BE-LABEL: ret_struct_3xi16:
 ; N64-BE:       # %bb.0: # %entry
 ; N64-BE-NEXT:    lui $1, %highest(struct_3xi16)
 ; N64-BE-NEXT:    daddiu $1, $1, %higher(struct_3xi16)
 ; N64-BE-NEXT:    dsll $1, $1, 16
 ; N64-BE-NEXT:    daddiu $1, $1, %hi(struct_3xi16)
 ; N64-BE-NEXT:    dsll $1, $1, 16
 ; N64-BE-NEXT:    lw $2, %lo(struct_3xi16)($1)
-; N64-BE-NEXT:    dsll $2, $2, 16
+; N64-BE-NEXT:    dsll $2, $2, 32
 ; N64-BE-NEXT:    daddiu $1, $1, %lo(struct_3xi16)
 ; N64-BE-NEXT:    lhu $1, 4($1)
-; N64-BE-NEXT:    or $1, $1, $2
+; N64-BE-NEXT:    dsll $1, $1, 16
 ; N64-BE-NEXT:    jr $ra
-; N64-BE-NEXT:    dsll $2, $1, 16
+; N64-BE-NEXT:    or $2, $2, $1
 ;
 ; N64-LE-LABEL: ret_struct_3xi16:
 ; N64-LE:       # %bb.0: # %entry
 ; N64-LE-NEXT:    lui $1, %highest(struct_3xi16)
 ; N64-LE-NEXT:    daddiu $1, $1, %higher(struct_3xi16)
 ; N64-LE-NEXT:    dsll $1, $1, 16
 ; N64-LE-NEXT:    daddiu $1, $1, %hi(struct_3xi16)
 ; N64-LE-NEXT:    dsll $1, $1, 16

llvm/test/CodeGen/Mips/cconv/vector.ll

 }
 
 define <4 x i8> @i8_4(<4 x i8> %a, <4 x i8> %b) {
 ; MIPS32-LABEL: i8_4:
 ; MIPS32:       # %bb.0:
 ; MIPS32-NEXT:    srl $1, $5, 24
 ; MIPS32-NEXT:    srl $2, $4, 24
 ; MIPS32-NEXT:    addu $1, $2, $1
-; MIPS32-NEXT:    sll $1, $1, 8
-; MIPS32-NEXT:    srl $2, $5, 16
-; MIPS32-NEXT:    srl $3, $4, 16
-; MIPS32-NEXT:    addu $2, $3, $2
-; MIPS32-NEXT:    andi $2, $2, 255
-; MIPS32-NEXT:    or $1, $2, $1
 ; MIPS32-NEXT:    addu $2, $4, $5
-; MIPS32-NEXT:    sll $1, $1, 16
+; MIPS32-NEXT:    sll $1, $1, 24
+; MIPS32-NEXT:    srl $3, $5, 16
+; MIPS32-NEXT:    srl $6, $4, 16
+; MIPS32-NEXT:    addu $3, $6, $3
+; MIPS32-NEXT:    andi $3, $3, 255
+; MIPS32-NEXT:    sll $3, $3, 16
+; MIPS32-NEXT:    or $1, $1, $3
 ; MIPS32-NEXT:    andi $2, $2, 255
 ; MIPS32-NEXT:    srl $3, $5, 8
 ; MIPS32-NEXT:    srl $4, $4, 8
 ; MIPS32-NEXT:    addu $3, $4, $3
 ; MIPS32-NEXT:    sll $3, $3, 8
 ; MIPS32-NEXT:    or $2, $2, $3
 ; MIPS32-NEXT:    andi $2, $2, 65535
 ; MIPS32-NEXT:    or $2, $2, $1
 ; MIPS32-NEXT:    jr $ra
 ; MIPS32-NEXT:    nop
 ;
 ; MIPS64-LABEL: i8_4:
 ; MIPS64:       # %bb.0:
 ; MIPS64-NEXT:    sll $1, $5, 0
 ; MIPS64-NEXT:    srl $2, $1, 24
 ; MIPS64-NEXT:    sll $3, $4, 0
 ; MIPS64-NEXT:    srl $4, $3, 24
 ; MIPS64-NEXT:    addu $2, $4, $2
-; MIPS64-NEXT:    sll $2, $2, 8
-; MIPS64-NEXT:    srl $4, $1, 16
-; MIPS64-NEXT:    srl $5, $3, 16
-; MIPS64-NEXT:    addu $4, $5, $4
-; MIPS64-NEXT:    andi $4, $4, 255
-; MIPS64-NEXT:    or $2, $4, $2
 ; MIPS64-NEXT:    addu $4, $3, $1
-; MIPS64-NEXT:    sll $2, $2, 16
+; MIPS64-NEXT:    sll $2, $2, 24
+; MIPS64-NEXT:    srl $5, $1, 16
+; MIPS64-NEXT:    srl $6, $3, 16
+; MIPS64-NEXT:    addu $5, $6, $5
+; MIPS64-NEXT:    andi $5, $5, 255
+; MIPS64-NEXT:    sll $5, $5, 16
+; MIPS64-NEXT:    or $2, $2, $5
 ; MIPS64-NEXT:    andi $4, $4, 255
 ; MIPS64-NEXT:    srl $1, $1, 8
 ; MIPS64-NEXT:    srl $3, $3, 8
 ; MIPS64-NEXT:    addu $1, $3, $1
 ; MIPS64-NEXT:    sll $1, $1, 8
 ; MIPS64-NEXT:    or $1, $4, $1
 ; MIPS64-NEXT:    andi $1, $1, 65535
 ; MIPS64-NEXT:    or $2, $1, $2

 ; MIPS64R5-NEXT:    nop
   %1 = add <4 x i8> %a, %b
   ret <4 x i8> %1
 }
 
 define <8 x i8> @i8_8(<8 x i8> %a, <8 x i8> %b) {
 ; MIPS32-LABEL: i8_8:
 ; MIPS32:       # %bb.0:
-; MIPS32-NEXT:    srl $1, $6, 24
-; MIPS32-NEXT:    srl $2, $4, 24
-; MIPS32-NEXT:    addu $1, $2, $1
-; MIPS32-NEXT:    sll $1, $1, 8
-; MIPS32-NEXT:    srl $2, $6, 16
-; MIPS32-NEXT:    srl $3, $4, 16
+; MIPS32-NEXT:    addu $1, $4, $6
+; MIPS32-NEXT:    srl $2, $6, 24
+; MIPS32-NEXT:    srl $3, $4, 24
 ; MIPS32-NEXT:    addu $2, $3, $2
-; MIPS32-NEXT:    andi $2, $2, 255
-; MIPS32-NEXT:    srl $3, $7, 24
-; MIPS32-NEXT:    srl $8, $5, 24
-; MIPS32-NEXT:    or $1, $2, $1
-; MIPS32-NEXT:    addu $2, $8, $3
-; MIPS32-NEXT:    addu $3, $4, $6
-; MIPS32-NEXT:    sll $2, $2, 8
-; MIPS32-NEXT:    srl $8, $7, 16
-; MIPS32-NEXT:    srl $9, $5, 16
-; MIPS32-NEXT:    addu $8, $9, $8
-; MIPS32-NEXT:    andi $8, $8, 255
-; MIPS32-NEXT:    or $8, $8, $2
-; MIPS32-NEXT:    sll $1, $1, 16
-; MIPS32-NEXT:    andi $2, $3, 255
+; MIPS32-NEXT:    andi $1, $1, 255
 ; MIPS32-NEXT:    srl $3, $6, 8
-; MIPS32-NEXT:    srl $4, $4, 8
-; MIPS32-NEXT:    addu $3, $4, $3
+; MIPS32-NEXT:    srl $8, $4, 8
+; MIPS32-NEXT:    addu $3, $8, $3
 ; MIPS32-NEXT:    sll $3, $3, 8
-; MIPS32-NEXT:    or $2, $2, $3
-; MIPS32-NEXT:    andi $2, $2, 65535
-; MIPS32-NEXT:    addu $3, $5, $7
-; MIPS32-NEXT:    or $2, $2, $1
-; MIPS32-NEXT:    sll $1, $8, 16
+; MIPS32-NEXT:    srl $6, $6, 16
+; MIPS32-NEXT:    srl $4, $4, 16
+; MIPS32-NEXT:    or $1, $1, $3
+; MIPS32-NEXT:    sll $2, $2, 24
+; MIPS32-NEXT:    addu $3, $4, $6
 ; MIPS32-NEXT:    andi $3, $3, 255
+; MIPS32-NEXT:    sll $3, $3, 16
+; MIPS32-NEXT:    srl $4, $7, 24
+; MIPS32-NEXT:    srl $6, $5, 24
+; MIPS32-NEXT:    or $2, $2, $3
+; MIPS32-NEXT:    andi $1, $1, 65535
+; MIPS32-NEXT:    addu $3, $6, $4
+; MIPS32-NEXT:    addu $4, $5, $7
+; MIPS32-NEXT:    sll $3, $3, 24
+; MIPS32-NEXT:    srl $6, $7, 16
+; MIPS32-NEXT:    srl $8, $5, 16
+; MIPS32-NEXT:    addu $6, $8, $6
+; MIPS32-NEXT:    andi $6, $6, 255
+; MIPS32-NEXT:    sll $6, $6, 16
+; MIPS32-NEXT:    or $2, $1, $2
+; MIPS32-NEXT:    or $1, $3, $6
+; MIPS32-NEXT:    andi $3, $4, 255
 ; MIPS32-NEXT:    srl $4, $7, 8
 ; MIPS32-NEXT:    srl $5, $5, 8
 ; MIPS32-NEXT:    addu $4, $5, $4
 ; MIPS32-NEXT:    sll $4, $4, 8
 ; MIPS32-NEXT:    or $3, $3, $4
 ; MIPS32-NEXT:    andi $3, $3, 65535
 ; MIPS32-NEXT:    or $3, $3, $1
 ; MIPS32-NEXT:    jr $ra
 ; MIPS32-NEXT:    nop
 ;
 ; MIPS64-LABEL: i8_8:
 ; MIPS64:       # %bb.0:
-; MIPS64-NEXT:    dsrl $1, $5, 56
+; MIPS64-NEXT:    dsrl $1, $5, 48
 ; MIPS64-NEXT:    sll $1, $1, 0
-; MIPS64-NEXT:    dsrl $2, $4, 56
+; MIPS64-NEXT:    dsrl $2, $4, 48
 ; MIPS64-NEXT:    sll $2, $2, 0
 ; MIPS64-NEXT:    addu $1, $2, $1
-; MIPS64-NEXT:    dsrl $2, $5, 48
-; MIPS64-NEXT:    sll $1, $1, 8
+; MIPS64-NEXT:    dsrl $2, $5, 56
+; MIPS64-NEXT:    andi $1, $1, 255
 ; MIPS64-NEXT:    sll $2, $2, 0
-; MIPS64-NEXT:    dsrl $3, $4, 48
+; MIPS64-NEXT:    dsrl $3, $4, 56
 ; MIPS64-NEXT:    sll $3, $3, 0
 ; MIPS64-NEXT:    addu $2, $3, $2
-; MIPS64-NEXT:    andi $2, $2, 255
 ; MIPS64-NEXT:    dsrl $3, $5, 40
-; MIPS64-NEXT:    or $1, $2, $1
-; MIPS64-NEXT:    sll $2, $5, 0
+; MIPS64-NEXT:    sll $2, $2, 24
+; MIPS64-NEXT:    sll $1, $1, 16
 ; MIPS64-NEXT:    sll $3, $3, 0
 ; MIPS64-NEXT:    dsrl $6, $4, 40
 ; MIPS64-NEXT:    sll $6, $6, 0
 ; MIPS64-NEXT:    addu $3, $6, $3
-; MIPS64-NEXT:    dsrl $5, $5, 32
-; MIPS64-NEXT:    srl $6, $2, 24
+; MIPS64-NEXT:    dsrl $6, $5, 32
 ; MIPS64-NEXT:    sll $7, $4, 0
-; MIPS64-NEXT:    srl $8, $7, 24
-; MIPS64-NEXT:    addu $6, $8, $6
-; MIPS64-NEXT:    sll $1, $1, 16
-; MIPS64-NEXT:    sll $3, $3, 8
 ; MIPS64-NEXT:    sll $5, $5, 0
+; MIPS64-NEXT:    srl $8, $5, 24
+; MIPS64-NEXT:    srl $9, $7, 24
+; MIPS64-NEXT:    or $1, $2, $1
+; MIPS64-NEXT:    sll $2, $3, 8
+; MIPS64-NEXT:    sll $3, $6, 0
 ; MIPS64-NEXT:    dsrl $4, $4, 32
 ; MIPS64-NEXT:    sll $4, $4, 0
-; MIPS64-NEXT:    addu $4, $4, $5
-; MIPS64-NEXT:    andi $4, $4, 255
-; MIPS64-NEXT:    or $3, $4, $3
-; MIPS64-NEXT:    andi $3, $3, 65535
-; MIPS64-NEXT:    or $1, $3, $1
-; MIPS64-NEXT:    sll $3, $6, 8
-; MIPS64-NEXT:    srl $4, $2, 16
-; MIPS64-NEXT:    srl $5, $7, 16
-; MIPS64-NEXT:    addu $4, $5, $4
+; MIPS64-NEXT:    addu $3, $4, $3
+; MIPS64-NEXT:    andi $3, $3, 255
+; MIPS64-NEXT:    or $2, $3, $2
+; MIPS64-NEXT:    andi $2, $2, 65535
+; MIPS64-NEXT:    or $1, $2, $1
+; MIPS64-NEXT:    addu $2, $9, $8
+; MIPS64-NEXT:    addu $3, $7, $5
+; MIPS64-NEXT:    sll $2, $2, 24
+; MIPS64-NEXT:    srl $4, $5, 16
+; MIPS64-NEXT:    srl $6, $7, 16
+; MIPS64-NEXT:    addu $4, $6, $4
 ; MIPS64-NEXT:    andi $4, $4, 255
-; MIPS64-NEXT:    or $3, $4, $3
-; MIPS64-NEXT:    addu $4, $7, $2
+; MIPS64-NEXT:    sll $4, $4, 16
 ; MIPS64-NEXT:    dsll $1, $1, 32
-; MIPS64-NEXT:    sll $3, $3, 16
-; MIPS64-NEXT:    andi $4, $4, 255
-; MIPS64-NEXT:    srl $2, $2, 8
+; MIPS64-NEXT:    or $2, $2, $4
+; MIPS64-NEXT:    andi $3, $3, 255
+; MIPS64-NEXT:    srl $4, $5, 8
 ; MIPS64-NEXT:    srl $5, $7, 8
-; MIPS64-NEXT:    addu $2, $5, $2
-; MIPS64-NEXT:    sll $2, $2, 8
-; MIPS64-NEXT:    or $2, $4, $2
-; MIPS64-NEXT:    andi $2, $2, 65535
-; MIPS64-NEXT:    or $2, $2, $3
+; MIPS64-NEXT:    addu $4, $5, $4
+; MIPS64-NEXT:    sll $4, $4, 8
+; MIPS64-NEXT:    or $3, $3, $4
+; MIPS64-NEXT:    andi $3, $3, 65535
+; MIPS64-NEXT:    or $2, $3, $2
 ; MIPS64-NEXT:    dsll $2, $2, 32
 ; MIPS64-NEXT:    dsrl $2, $2, 32
 ; MIPS64-NEXT:    or $2, $2, $1
 ; MIPS64-NEXT:    jr $ra
 ; MIPS64-NEXT:    nop
 ;
 ; MIPS32R5EB-LABEL: i8_8:
 ; MIPS32R5EB:       # %bb.0:

 ; MIPS32R5EL-NEXT:    nop
   %1 = add <8 x i8> %a, %b
   ret <8 x i8> %1
 }
 
 define <16 x i8> @i8_16(<16 x i8> %a, <16 x i8> %b) {
 ; MIPS32-LABEL: i8_16:
 ; MIPS32:       # %bb.0:
-; MIPS32-NEXT:    lw $1, 24($sp)
-; MIPS32-NEXT:    srl $2, $1, 24
-; MIPS32-NEXT:    srl $3, $6, 24
-; MIPS32-NEXT:    srl $8, $1, 16
-; MIPS32-NEXT:    srl $9, $6, 16
-; MIPS32-NEXT:    srl $10, $1, 8
-; MIPS32-NEXT:    srl $11, $6, 8
-; MIPS32-NEXT:    lw $12, 20($sp)
-; MIPS32-NEXT:    srl $13, $12, 8
-; MIPS32-NEXT:    srl $14, $5, 8
-; MIPS32-NEXT:    addu $13, $14, $13
-; MIPS32-NEXT:    addu $14, $5, $12
-; MIPS32-NEXT:    addu $10, $11, $10
-; MIPS32-NEXT:    addu $1, $6, $1
-; MIPS32-NEXT:    addu $6, $9, $8
-; MIPS32-NEXT:    addu $2, $3, $2
-; MIPS32-NEXT:    srl $3, $12, 24
-; MIPS32-NEXT:    srl $8, $5, 24
-; MIPS32-NEXT:    srl $9, $12, 16
-; MIPS32-NEXT:    srl $5, $5, 16
-; MIPS32-NEXT:    addu $5, $5, $9
-; MIPS32-NEXT:    addu $3, $8, $3
-; MIPS32-NEXT:    sll $2, $2, 8
-; MIPS32-NEXT:    andi $6, $6, 255
-; MIPS32-NEXT:    andi $1, $1, 255
-; MIPS32-NEXT:    sll $8, $10, 8
-; MIPS32-NEXT:    andi $9, $14, 255
-; MIPS32-NEXT:    sll $10, $13, 8
-; MIPS32-NEXT:    lw $11, 28($sp)
-; MIPS32-NEXT:    lw $12, 16($sp)
-; MIPS32-NEXT:    srl $13, $12, 24
-; MIPS32-NEXT:    srl $14, $4, 24
-; MIPS32-NEXT:    srl $15, $11, 24
-; MIPS32-NEXT:    srl $24, $7, 24
-; MIPS32-NEXT:    or $9, $9, $10
-; MIPS32-NEXT:    or $1, $1, $8
-; MIPS32-NEXT:    or $2, $6, $2
-; MIPS32-NEXT:    addu $6, $24, $15
-; MIPS32-NEXT:    sll $3, $3, 8
-; MIPS32-NEXT:    andi $5, $5, 255
-; MIPS32-NEXT:    addu $8, $14, $13
-; MIPS32-NEXT:    sll $8, $8, 8
-; MIPS32-NEXT:    srl $10, $12, 16
+; MIPS32-NEXT:    lw $1, 16($sp)
+; MIPS32-NEXT:    lw $2, 20($sp)
+; MIPS32-NEXT:    lw $3, 24($sp)
+; MIPS32-NEXT:    srl $8, $3, 8
+; MIPS32-NEXT:    srl $9, $6, 8
+; MIPS32-NEXT:    srl $10, $2, 16
+; MIPS32-NEXT:    srl $11, $5, 16
+; MIPS32-NEXT:    srl $12, $1, 16
 ; MIPS32-NEXT:    srl $13, $4, 16
-; MIPS32-NEXT:    addu $10, $13, $10
-; MIPS32-NEXT:    andi $10, $10, 255
-; MIPS32-NEXT:    or $8, $10, $8
-; MIPS32-NEXT:    or $3, $5, $3
-; MIPS32-NEXT:    addu $5, $4, $12
-; MIPS32-NEXT:    sll $6, $6, 8
-; MIPS32-NEXT:    srl $10, $11, 16
-; MIPS32-NEXT:    srl $13, $7, 16
-; MIPS32-NEXT:    addu $10, $13, $10
+; MIPS32-NEXT:    srl $14, $1, 8
+; MIPS32-NEXT:    srl $15, $4, 8
+; MIPS32-NEXT:    addu $24, $6, $3
+; MIPS32-NEXT:    addu $14, $15, $14
+; MIPS32-NEXT:    addu $15, $4, $1
+; MIPS32-NEXT:    addu $12, $13, $12
+; MIPS32-NEXT:    addu $10, $11, $10
+; MIPS32-NEXT:    srl $11, $2, 24
+; MIPS32-NEXT:    addu $13, $5, $2
+; MIPS32-NEXT:    addu $8, $9, $8
+; MIPS32-NEXT:    srl $1, $1, 24
+; MIPS32-NEXT:    srl $4, $4, 24
+; MIPS32-NEXT:    srl $9, $5, 24
+; MIPS32-NEXT:    srl $25, $3, 24
+; MIPS32-NEXT:    srl $gp, $6, 24
+; MIPS32-NEXT:    addu $25, $gp, $25
 ; MIPS32-NEXT:    andi $10, $10, 255
-; MIPS32-NEXT:    or $6, $10, $6
-; MIPS32-NEXT:    sll $10, $2, 16
-; MIPS32-NEXT:    andi $1, $1, 65535
+; MIPS32-NEXT:    addu $9, $9, $11
+; MIPS32-NEXT:    andi $11, $12, 255
+; MIPS32-NEXT:    addu $1, $4, $1
+; MIPS32-NEXT:    andi $4, $15, 255
+; MIPS32-NEXT:    sll $12, $14, 8
+; MIPS32-NEXT:    andi $14, $24, 255
+; MIPS32-NEXT:    sll $8, $8, 8
+; MIPS32-NEXT:    andi $13, $13, 255
+; MIPS32-NEXT:    srl $2, $2, 8
+; MIPS32-NEXT:    srl $5, $5, 8
+; MIPS32-NEXT:    addu $2, $5, $2
+; MIPS32-NEXT:    sll $2, $2, 8
+; MIPS32-NEXT:    srl $3, $3, 16
+; MIPS32-NEXT:    srl $5, $6, 16
+; MIPS32-NEXT:    or $2, $13, $2
+; MIPS32-NEXT:    or $6, $14, $8
+; MIPS32-NEXT:    or $4, $4, $12
+; MIPS32-NEXT:    sll $1, $1, 24
+; MIPS32-NEXT:    sll $8, $11, 16
+; MIPS32-NEXT:    sll $9, $9, 24
+; MIPS32-NEXT:    sll $10, $10, 16
+; MIPS32-NEXT:    sll $11, $25, 24
+; MIPS32-NEXT:    addu $3, $5, $3
+; MIPS32-NEXT:    andi $3, $3, 255
 ; MIPS32-NEXT:    sll $3, $3, 16
-; MIPS32-NEXT:    andi $9, $9, 65535
-; MIPS32-NEXT:    sll $2, $8, 16
-; MIPS32-NEXT:    andi $5, $5, 255
-; MIPS32-NEXT:    srl $8, $12, 8
-; MIPS32-NEXT:    srl $4, $4, 8
-; MIPS32-NEXT:    addu $4, $4, $8
-; MIPS32-NEXT:    sll $4, $4, 8
-; MIPS32-NEXT:    or $4, $5, $4
+; MIPS32-NEXT:    lw $5, 28($sp)
+; MIPS32-NEXT:    srl $12, $5, 24
+; MIPS32-NEXT:    srl $13, $7, 24
+; MIPS32-NEXT:    or $11, $11, $3
+; MIPS32-NEXT:    or $3, $9, $10
+; MIPS32-NEXT:    or $1, $1, $8
 ; MIPS32-NEXT:    andi $4, $4, 65535
-; MIPS32-NEXT:    addu $5, $7, $11
-; MIPS32-NEXT:    or $2, $4, $2
+; MIPS32-NEXT:    addu $8, $13, $12
+; MIPS32-NEXT:    andi $6, $6, 65535
+; MIPS32-NEXT:    andi $9, $2, 65535
+; MIPS32-NEXT:    addu $10, $7, $5
+; MIPS32-NEXT:    sll $8, $8, 24
+; MIPS32-NEXT:    srl $2, $5, 16
+; MIPS32-NEXT:    srl $12, $7, 16
+; MIPS32-NEXT:    addu $2, $12, $2
+; MIPS32-NEXT:    andi $2, $2, 255
+; MIPS32-NEXT:    sll $12, $2, 16
+; MIPS32-NEXT:    or $2, $4, $1
 ; MIPS32-NEXT:    or $3, $9, $3
-; MIPS32-NEXT:    or $4, $1, $10
-; MIPS32-NEXT:    sll $1, $6, 16
-; MIPS32-NEXT:    andi $5, $5, 255
-; MIPS32-NEXT:    srl $6, $11, 8
+; MIPS32-NEXT:    or $4, $6, $11
+; MIPS32-NEXT:    or $1, $8, $12
+; MIPS32-NEXT:    andi $6, $10, 255
+; MIPS32-NEXT:    srl $5, $5, 8
 ; MIPS32-NEXT:    srl $7, $7, 8
-; MIPS32-NEXT:    addu $6, $7, $6
-; MIPS32-NEXT:    sll $6, $6, 8
-; MIPS32-NEXT:    or $5, $5, $6
+; MIPS32-NEXT:    addu $5, $7, $5
+; MIPS32-NEXT:    sll $5, $5, 8
+; MIPS32-NEXT:    or $5, $6, $5
 ; MIPS32-NEXT:    andi $5, $5, 65535
 ; MIPS32-NEXT:    or $5, $5, $1
 ; MIPS32-NEXT:    jr $ra
 ; MIPS32-NEXT:    nop
 ;
 ; MIPS64-LABEL: i8_16:
 ; MIPS64:       # %bb.0:
-; MIPS64-NEXT:    dsrl $1, $7, 56
-; MIPS64-NEXT:    dsrl $2, $5, 56
-; MIPS64-NEXT:    dsrl $3, $7, 48
-; MIPS64-NEXT:    dsrl $8, $5, 48
-; MIPS64-NEXT:    dsrl $9, $6, 56
-; MIPS64-NEXT:    dsrl $10, $4, 56
-; MIPS64-NEXT:    dsrl $11, $7, 32
-; MIPS64-NEXT:    sll $1, $1, 0
+; MIPS64-NEXT:    sll $1, $6, 0
+; MIPS64-NEXT:    dsrl $2, $6, 56
+; MIPS64-NEXT:    dsrl $3, $6, 48
+; MIPS64-NEXT:    dsrl $8, $4, 48
+; MIPS64-NEXT:    srl $9, $1, 16
+; MIPS64-NEXT:    sll $10, $4, 0
+; MIPS64-NEXT:    srl $11, $10, 16
+; MIPS64-NEXT:    dsrl $12, $7, 56
+; MIPS64-NEXT:    addu $13, $10, $1
+; MIPS64-NEXT:    addu $9, $11, $9
 ; MIPS64-NEXT:    sll $2, $2, 0
+; MIPS64-NEXT:    dsrl $11, $7, 48
+; MIPS64-NEXT:    srl $14, $1, 8
+; MIPS64-NEXT:    srl $15, $10, 8
+; MIPS64-NEXT:    addu $14, $15, $14
+; MIPS64-NEXT:    dsrl $15, $4, 56
+; MIPS64-NEXT:    dsrl $24, $7, 40
 ; MIPS64-NEXT:    sll $3, $3, 0
 ; MIPS64-NEXT:    sll $8, $8, 0
-; MIPS64-NEXT:    dsrl $12, $7, 40
-; MIPS64-NEXT:    sll $12, $12, 0
-; MIPS64-NEXT:    dsrl $13, $5, 40
-; MIPS64-NEXT:    sll $13, $13, 0
-; MIPS64-NEXT:    addu $12, $13, $12
-; MIPS64-NEXT:    addu $3, $8, $3
-; MIPS64-NEXT:    addu $1, $2, $1
-; MIPS64-NEXT:    sll $2, $9, 0
-; MIPS64-NEXT:    sll $8, $10, 0
-; MIPS64-NEXT:    dsrl $9, $6, 48
-; MIPS64-NEXT:    sll $9, $9, 0
-; MIPS64-NEXT:    dsrl $10, $4, 48
-; MIPS64-NEXT:    sll $10, $10, 0
-; MIPS64-NEXT:    addu $9, $10, $9
-; MIPS64-NEXT:    addu $2, $8, $2
-; MIPS64-NEXT:    sll $8, $1, 8
-; MIPS64-NEXT:    andi $3, $3, 255
-; MIPS64-NEXT:    sll $1, $12, 8
-; MIPS64-NEXT:    sll $10, $11, 0
-; MIPS64-NEXT:    dsrl $11, $5, 32
-; MIPS64-NEXT:    sll $11, $11, 0
-; MIPS64-NEXT:    addu $10, $11, $10
-; MIPS64-NEXT:    andi $10, $10, 255
-; MIPS64-NEXT:    or $10, $10, $1
-; MIPS64-NEXT:    sll $1, $6, 0
-; MIPS64-NEXT:    or $8, $3, $8
-; MIPS64-NEXT:    sll $2, $2, 8
+; MIPS64-NEXT:    sll $15, $15, 0
 ; MIPS64-NEXT:    andi $9, $9, 255
-; MIPS64-NEXT:    dsrl $11, $6, 40
-; MIPS64-NEXT:    srl $3, $1, 24
-; MIPS64-NEXT:    sll $12, $4, 0
-; MIPS64-NEXT:    srl $13, $12, 24
-; MIPS64-NEXT:    srl $14, $1, 16
-; MIPS64-NEXT:    srl $15, $12, 16
-; MIPS64-NEXT:    andi $10, $10, 65535
-; MIPS64-NEXT:    addu $14, $15, $14
-; MIPS64-NEXT:    addu $13, $13, $3
-; MIPS64-NEXT:    sll $3, $7, 0
-; MIPS64-NEXT:    or $2, $9, $2
-; MIPS64-NEXT:    sll $7, $8, 16
+; MIPS64-NEXT:    addu $2, $15, $2
+; MIPS64-NEXT:    andi $13, $13, 255
+; MIPS64-NEXT:    sll $14, $14, 8
+; MIPS64-NEXT:    addu $3, $8, $3
 ; MIPS64-NEXT:    sll $8, $11, 0
-; MIPS64-NEXT:    dsrl $9, $4, 40
-; MIPS64-NEXT:    sll $9, $9, 0
-; MIPS64-NEXT:    addu $8, $9, $8
+; MIPS64-NEXT:    srl $1, $1, 24
+; MIPS64-NEXT:    sll $11, $12, 0
+; MIPS64-NEXT:    dsrl $12, $5, 56
+; MIPS64-NEXT:    dsrl $15, $5, 48
+; MIPS64-NEXT:    andi $3, $3, 255
+; MIPS64-NEXT:    dsrl $25, $6, 40
+; MIPS64-NEXT:    sll $15, $15, 0
+; MIPS64-NEXT:    srl $10, $10, 24
+; MIPS64-NEXT:    sll $12, $12, 0
+; MIPS64-NEXT:    or $13, $13, $14
+; MIPS64-NEXT:    sll $14, $24, 0
+; MIPS64-NEXT:    sll $2, $2, 24
+; MIPS64-NEXT:    addu $11, $12, $11
+; MIPS64-NEXT:    sll $9, $9, 16
+; MIPS64-NEXT:    addu $1, $10, $1
+; MIPS64-NEXT:    addu $8, $15, $8
+; MIPS64-NEXT:    sll $10, $25, 0
+; MIPS64-NEXT:    dsrl $12, $4, 40
+; MIPS64-NEXT:    sll $12, $12, 0
+; MIPS64-NEXT:    addu $10, $12, $10
+; MIPS64-NEXT:    sll $3, $3, 16
+; MIPS64-NEXT:    andi $8, $8, 255
+; MIPS64-NEXT:    sll $1, $1, 24
+; MIPS64-NEXT:    dsrl $12, $5, 40
+; MIPS64-NEXT:    sll $12, $12, 0
 ; MIPS64-NEXT:    dsrl $6, $6, 32
-; MIPS64-NEXT:    srl $9, $3, 24
-; MIPS64-NEXT:    sll $5, $5, 0
-; MIPS64-NEXT:    srl $11, $5, 24
-; MIPS64-NEXT:    or $7, $10, $7
-; MIPS64-NEXT:    addu $9, $11, $9
-; MIPS64-NEXT:    sll $10, $13, 8
-; MIPS64-NEXT:    andi $11, $14, 255
-; MIPS64-NEXT:    sll $2, $2, 16
-; MIPS64-NEXT:    sll $8, $8, 8
+; MIPS64-NEXT:    or $1, $1, $9
+; MIPS64-NEXT:    addu $9, $12, $14
+; MIPS64-NEXT:    sll $11, $11, 24
+; MIPS64-NEXT:    sll $8, $8, 16
+; MIPS64-NEXT:    dsrl $12, $7, 32
+; MIPS64-NEXT:    andi $13, $13, 65535
+; MIPS64-NEXT:    or $2, $2, $3
+; MIPS64-NEXT:    sll $3, $10, 8
 ; MIPS64-NEXT:    sll $6, $6, 0
 ; MIPS64-NEXT:    dsrl $4, $4, 32
 ; MIPS64-NEXT:    sll $4, $4, 0
 ; MIPS64-NEXT:    addu $4, $4, $6
 ; MIPS64-NEXT:    andi $4, $4, 255
-; MIPS64-NEXT:    or $4, $4, $8
-; MIPS64-NEXT:    andi $4, $4, 65535
-; MIPS64-NEXT:    or $2, $4, $2
-; MIPS64-NEXT:    or $4, $11, $10
-; MIPS64-NEXT:    addu $6, $12, $1
-; MIPS64-NEXT:    sll $8, $9, 8
-; MIPS64-NEXT:    srl $9, $3, 16
-; MIPS64-NEXT:    srl $10, $5, 16
-; MIPS64-NEXT:    addu $9, $10, $9
-; MIPS64-NEXT:    andi $9, $9, 255
-; MIPS64-NEXT:    or $8, $9, $8
-; MIPS64-NEXT:    addu $9, $5, $3
-; MIPS64-NEXT:    dsll $2, $2, 32
-; MIPS64-NEXT:    sll $4, $4, 16
+; MIPS64-NEXT:    or $3, $4, $3
+; MIPS64-NEXT:    andi $3, $3, 65535
+; MIPS64-NEXT:    or $2, $3, $2
+; MIPS64-NEXT:    or $1, $13, $1
+; MIPS64-NEXT:    or $3, $11, $8
+; MIPS64-NEXT:    sll $4, $9, 8
+; MIPS64-NEXT:    sll $6, $12, 0
+; MIPS64-NEXT:    dsrl $8, $5, 32
+; MIPS64-NEXT:    sll $8, $8, 0
+; MIPS64-NEXT:    addu $6, $8, $6
 ; MIPS64-NEXT:    andi $6, $6, 255
-; MIPS64-NEXT:    srl $1, $1, 8
-; MIPS64-NEXT:    srl $10, $12, 8
-; MIPS64-NEXT:    addu $1, $10, $1
-; MIPS64-NEXT:    sll $1, $1, 8
-; MIPS64-NEXT:    or $1, $6, $1
-; MIPS64-NEXT:    andi $1, $1, 65535
-; MIPS64-NEXT:    or $1, $1, $4
+; MIPS64-NEXT:    or $4, $6, $4
+; MIPS64-NEXT:    andi $4, $4, 65535
 ; MIPS64-NEXT:    dsll $1, $1, 32
+; MIPS64-NEXT:    or $3, $4, $3
+; MIPS64-NEXT:    sll $4, $7, 0
+; MIPS64-NEXT:    srl $6, $4, 24
+; MIPS64-NEXT:    sll $5, $5, 0
+; MIPS64-NEXT:    srl $7, $5, 24
+; MIPS64-NEXT:    addu $8, $5, $4
+; MIPS64-NEXT:    dsll $2, $2, 32
 ; MIPS64-NEXT:    dsrl $1, $1, 32
+; MIPS64-NEXT:    addu $6, $7, $6
+; MIPS64-NEXT:    sll $6, $6, 24
+; MIPS64-NEXT:    srl $7, $4, 16
+; MIPS64-NEXT:    srl $9, $5, 16
+; MIPS64-NEXT:    addu $7, $9, $7
+; MIPS64-NEXT:    andi $7, $7, 255
+; MIPS64-NEXT:    sll $7, $7, 16
 ; MIPS64-NEXT:    or $2, $1, $2
-; MIPS64-NEXT:    dsll $1, $7, 32
-; MIPS64-NEXT:    sll $4, $8, 16
-; MIPS64-NEXT:    andi $6, $9, 255
-; MIPS64-NEXT:    srl $3, $3, 8
+; MIPS64-NEXT:    dsll $1, $3, 32
+; MIPS64-NEXT:    or $3, $6, $7
+; MIPS64-NEXT:    andi $6, $8, 255
+; MIPS64-NEXT:    srl $4, $4, 8
 ; MIPS64-NEXT:    srl $5, $5, 8
-; MIPS64-NEXT:    addu $3, $5, $3
-; MIPS64-NEXT:    sll $3, $3, 8
-; MIPS64-NEXT:    or $3, $6, $3
-; MIPS64-NEXT:    andi $3, $3, 65535
-; MIPS64-NEXT:    or $3, $3, $4
+; MIPS64-NEXT:    addu $4, $5, $4
+; MIPS64-NEXT:    sll $4, $4, 8
+; MIPS64-NEXT:    or $4, $6, $4
+; MIPS64-NEXT:    andi $4, $4, 65535
+; MIPS64-NEXT:    or $3, $4, $3
 ; MIPS64-NEXT:    dsll $3, $3, 32
 ; MIPS64-NEXT:    dsrl $3, $3, 32
 ; MIPS64-NEXT:    or $3, $3, $1
 ; MIPS64-NEXT:    jr $ra
 ; MIPS64-NEXT:    nop
 ;
 ; MIPS32R5EB-LABEL: i8_16:
 ; MIPS32R5EB:       # %bb.0:

 ; MIPS32R5EB-NEXT:    addiu $sp, $sp, 48
 ; MIPS32R5EB-NEXT:    jr $ra
 ; MIPS32R5EB-NEXT:    nop
 ;
 ; MIPS64R5EB-LABEL: i24x2:
 ; MIPS64R5EB:       # %bb.0: # %Entry
 ; MIPS64R5EB-NEXT:    daddiu $sp, $sp, -32
 ; MIPS64R5EB-NEXT:    .cfi_def_cfa_offset 32
+; MIPS64R5EB-NEXT:    sh $5, 20($sp)
 ; MIPS64R5EB-NEXT:    dsrl $1, $5, 16
 ; MIPS64R5EB-NEXT:    sw $1, 16($sp)
-; MIPS64R5EB-NEXT:    sh $5, 20($sp)
+; MIPS64R5EB-NEXT:    sh $4, 28($sp)
 ; MIPS64R5EB-NEXT:    dsrl $1, $4, 16
 ; MIPS64R5EB-NEXT:    sw $1, 24($sp)
-; MIPS64R5EB-NEXT:    sh $4, 28($sp)
-; MIPS64R5EB-NEXT:    lb $1, 19($sp)
-; MIPS64R5EB-NEXT:    dsll $1, $1, 8
-; MIPS64R5EB-NEXT:    lbu $2, 20($sp)
-; MIPS64R5EB-NEXT:    or $1, $1, $2
+; MIPS64R5EB-NEXT:    lbu $1, 20($sp)
 ; MIPS64R5EB-NEXT:    dsll $1, $1, 8
-; MIPS64R5EB-NEXT:    lb $2, 27($sp)
+; MIPS64R5EB-NEXT:    lb $2, 19($sp)
+; MIPS64R5EB-NEXT:    dsll $2, $2, 16
+; MIPS64R5EB-NEXT:    or $1, $2, $1
+; MIPS64R5EB-NEXT:    lbu $2, 28($sp)
 ; MIPS64R5EB-NEXT:    dsll $2, $2, 8
-; MIPS64R5EB-NEXT:    lbu $3, 28($sp)
-; MIPS64R5EB-NEXT:    or $2, $2, $3
-; MIPS64R5EB-NEXT:    lbu $3, 21($sp)
-; MIPS64R5EB-NEXT:    dsll $2, $2, 8
-; MIPS64R5EB-NEXT:    or $1, $3, $1
+; MIPS64R5EB-NEXT:    lb $3, 27($sp)
+; MIPS64R5EB-NEXT:    dsll $3, $3, 16
+; MIPS64R5EB-NEXT:    lbu $4, 21($sp)
+; MIPS64R5EB-NEXT:    or $2, $3, $2
+; MIPS64R5EB-NEXT:    or $1, $4, $1
 ; MIPS64R5EB-NEXT:    lh $3, 16($sp)
 ; MIPS64R5EB-NEXT:    dsll $3, $3, 8
 ; MIPS64R5EB-NEXT:    lbu $4, 18($sp)
 ; MIPS64R5EB-NEXT:    or $3, $4, $3
 ; MIPS64R5EB-NEXT:    lbu $4, 29($sp)
 ; MIPS64R5EB-NEXT:    insert.d $w0[0], $3
 ; MIPS64R5EB-NEXT:    insert.d $w0[1], $1
 ; MIPS64R5EB-NEXT:    or $1, $4, $2

llvm/test/CodeGen/Mips/load-store-left-right.ll

 ; MIPS32-EB:       # %bb.0: # %entry
 ; MIPS32-EB-NEXT:    lui $2, %hi(_gp_disp)
 ; MIPS32-EB-NEXT:    addiu $2, $2, %lo(_gp_disp)
 ; MIPS32-EB-NEXT:    addiu $sp, $sp, -24
 ; MIPS32-EB-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
 ; MIPS32-EB-NEXT:    addu $gp, $2, $25
 ; MIPS32-EB-NEXT:    lw $1, %got(arr)($gp)
 ; MIPS32-EB-NEXT:    lwl $4, 0($1)
-; MIPS32-EB-NEXT:    lwr $4, 3($1)
 ; MIPS32-EB-NEXT:    lbu $2, 5($1)
+; MIPS32-EB-NEXT:    lwr $4, 3($1)
+; MIPS32-EB-NEXT:    sll $2, $2, 16
 ; MIPS32-EB-NEXT:    lbu $3, 4($1)
-; MIPS32-EB-NEXT:    sll $3, $3, 8
+; MIPS32-EB-NEXT:    sll $3, $3, 24
 ; MIPS32-EB-NEXT:    or $2, $3, $2
-; MIPS32-EB-NEXT:    sll $2, $2, 16
 ; MIPS32-EB-NEXT:    lbu $1, 6($1)
 ; MIPS32-EB-NEXT:    sll $1, $1, 8
 ; MIPS32-EB-NEXT:    lw $25, %call16(extern_func)($gp)
 ; MIPS32-EB-NEXT:    .reloc ($tmp0), R_MIPS_JALR, extern_func
 ; MIPS32-EB-NEXT:  $tmp0:
 ; MIPS32-EB-NEXT:    jalr $25
 ; MIPS32-EB-NEXT:    or $5, $2, $1
 ; MIPS32-EB-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload

 ; MIPS64-EL:       # %bb.0: # %entry
 ; MIPS64-EL-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64-EL-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64-EL-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64-EL-NEXT:    lui $1, %hi(%neg(%gp_rel(pass_array_byval)))
 ; MIPS64-EL-NEXT:    daddu $1, $1, $25
 ; MIPS64-EL-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(pass_array_byval)))
 ; MIPS64-EL-NEXT:    ld $1, %got_disp(arr)($gp)
-; MIPS64-EL-NEXT:    lwl $2, 3($1)
-; MIPS64-EL-NEXT:    lwr $2, 0($1)
-; MIPS64-EL-NEXT:    daddiu $3, $zero, 1
-; MIPS64-EL-NEXT:    dsll $3, $3, 32
-; MIPS64-EL-NEXT:    daddiu $3, $3, -1
-; MIPS64-EL-NEXT:    and $2, $2, $3
-; MIPS64-EL-NEXT:    lbu $3, 4($1)
-; MIPS64-EL-NEXT:    lbu $4, 5($1)
-; MIPS64-EL-NEXT:    dsll $4, $4, 8
-; MIPS64-EL-NEXT:    or $3, $4, $3
-; MIPS64-EL-NEXT:    dsll $3, $3, 32
-; MIPS64-EL-NEXT:    or $2, $2, $3
+; MIPS64-EL-NEXT:    lbu $2, 4($1)
+; MIPS64-EL-NEXT:    dsll $2, $2, 32
+; MIPS64-EL-NEXT:    lbu $3, 5($1)
+; MIPS64-EL-NEXT:    dsll $3, $3, 40
+; MIPS64-EL-NEXT:    or $2, $3, $2
+; MIPS64-EL-NEXT:    lwl $3, 3($1)
+; MIPS64-EL-NEXT:    lwr $3, 0($1)
+; MIPS64-EL-NEXT:    daddiu $4, $zero, 1
+; MIPS64-EL-NEXT:    dsll $4, $4, 32
+; MIPS64-EL-NEXT:    daddiu $4, $4, -1
+; MIPS64-EL-NEXT:    and $3, $3, $4
+; MIPS64-EL-NEXT:    or $2, $3, $2
 ; MIPS64-EL-NEXT:    lbu $1, 6($1)
 ; MIPS64-EL-NEXT:    dsll $1, $1, 48
 ; MIPS64-EL-NEXT:    ld $25, %call16(extern_func)($gp)
 ; MIPS64-EL-NEXT:    .reloc .Ltmp0, R_MIPS_JALR, extern_func
 ; MIPS64-EL-NEXT:  .Ltmp0:
 ; MIPS64-EL-NEXT:    jalr $25
 ; MIPS64-EL-NEXT:    or $4, $2, $1
 ; MIPS64-EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64-EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
 ; MIPS64-EL-NEXT:    jr $ra
 ; MIPS64-EL-NEXT:    daddiu $sp, $sp, 16
 ;
 ; MIPS64-EB-LABEL: pass_array_byval:
 ; MIPS64-EB:       # %bb.0: # %entry
 ; MIPS64-EB-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64-EB-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64-EB-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64-EB-NEXT:    lui $1, %hi(%neg(%gp_rel(pass_array_byval)))
 ; MIPS64-EB-NEXT:    daddu $1, $1, $25
 ; MIPS64-EB-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(pass_array_byval)))
 ; MIPS64-EB-NEXT:    ld $1, %got_disp(arr)($gp)
-; MIPS64-EB-NEXT:    lwl $2, 0($1)
-; MIPS64-EB-NEXT:    lwr $2, 3($1)
-; MIPS64-EB-NEXT:    dsll $2, $2, 32
-; MIPS64-EB-NEXT:    lbu $3, 5($1)
-; MIPS64-EB-NEXT:    lbu $4, 4($1)
-; MIPS64-EB-NEXT:    dsll $4, $4, 8
-; MIPS64-EB-NEXT:    or $3, $4, $3
-; MIPS64-EB-NEXT:    dsll $3, $3, 16
-; MIPS64-EB-NEXT:    or $2, $2, $3
+; MIPS64-EB-NEXT:    lbu $2, 5($1)
+; MIPS64-EB-NEXT:    dsll $2, $2, 16
+; MIPS64-EB-NEXT:    lbu $3, 4($1)
+; MIPS64-EB-NEXT:    dsll $3, $3, 24
+; MIPS64-EB-NEXT:    or $2, $3, $2
+; MIPS64-EB-NEXT:    lwl $3, 0($1)
+; MIPS64-EB-NEXT:    lwr $3, 3($1)
+; MIPS64-EB-NEXT:    dsll $3, $3, 32
+; MIPS64-EB-NEXT:    or $2, $3, $2
 ; MIPS64-EB-NEXT:    lbu $1, 6($1)
 ; MIPS64-EB-NEXT:    dsll $1, $1, 8
 ; MIPS64-EB-NEXT:    ld $25, %call16(extern_func)($gp)
 ; MIPS64-EB-NEXT:    .reloc .Ltmp0, R_MIPS_JALR, extern_func
 ; MIPS64-EB-NEXT:  .Ltmp0:
 ; MIPS64-EB-NEXT:    jalr $25
 ; MIPS64-EB-NEXT:    or $4, $2, $1
 ; MIPS64-EB-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64-EB-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
 ; MIPS64-EB-NEXT:    jr $ra
 ; MIPS64-EB-NEXT:    daddiu $sp, $sp, 16
 ;
 ; MIPS64R2-EL-LABEL: pass_array_byval:
 ; MIPS64R2-EL:       # %bb.0: # %entry
 ; MIPS64R2-EL-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64R2-EL-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64R2-EL-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64R2-EL-NEXT:    lui $1, %hi(%neg(%gp_rel(pass_array_byval)))
 ; MIPS64R2-EL-NEXT:    daddu $1, $1, $25
 ; MIPS64R2-EL-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(pass_array_byval)))
 ; MIPS64R2-EL-NEXT:    ld $1, %got_disp(arr)($gp)
-; MIPS64R2-EL-NEXT:    lwl $2, 3($1)
-; MIPS64R2-EL-NEXT:    lwr $2, 0($1)
-; MIPS64R2-EL-NEXT:    dext $2, $2, 0, 32
-; MIPS64R2-EL-NEXT:    lbu $3, 4($1)
-; MIPS64R2-EL-NEXT:    lbu $4, 5($1)
-; MIPS64R2-EL-NEXT:    dsll $4, $4, 8
-; MIPS64R2-EL-NEXT:    or $3, $4, $3
-; MIPS64R2-EL-NEXT:    dsll $3, $3, 32
-; MIPS64R2-EL-NEXT:    or $2, $2, $3
+; MIPS64R2-EL-NEXT:    lbu $2, 4($1)
+; MIPS64R2-EL-NEXT:    dsll $2, $2, 32
+; MIPS64R2-EL-NEXT:    lbu $3, 5($1)
+; MIPS64R2-EL-NEXT:    dsll $3, $3, 40
+; MIPS64R2-EL-NEXT:    or $2, $3, $2
+; MIPS64R2-EL-NEXT:    lwl $3, 3($1)
+; MIPS64R2-EL-NEXT:    lwr $3, 0($1)
+; MIPS64R2-EL-NEXT:    dext $3, $3, 0, 32
+; MIPS64R2-EL-NEXT:    or $2, $3, $2
 ; MIPS64R2-EL-NEXT:    lbu $1, 6($1)
 ; MIPS64R2-EL-NEXT:    dsll $1, $1, 48
 ; MIPS64R2-EL-NEXT:    ld $25, %call16(extern_func)($gp)
 ; MIPS64R2-EL-NEXT:    .reloc .Ltmp0, R_MIPS_JALR, extern_func
 ; MIPS64R2-EL-NEXT:  .Ltmp0:
 ; MIPS64R2-EL-NEXT:    jalr $25
 ; MIPS64R2-EL-NEXT:    or $4, $2, $1
 ; MIPS64R2-EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64R2-EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
 ; MIPS64R2-EL-NEXT:    jr $ra
 ; MIPS64R2-EL-NEXT:    daddiu $sp, $sp, 16
 ;
 ; MIPS64R2-EB-LABEL: pass_array_byval:
 ; MIPS64R2-EB:       # %bb.0: # %entry
 ; MIPS64R2-EB-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64R2-EB-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64R2-EB-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64R2-EB-NEXT:    lui $1, %hi(%neg(%gp_rel(pass_array_byval)))
 ; MIPS64R2-EB-NEXT:    daddu $1, $1, $25
 ; MIPS64R2-EB-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(pass_array_byval)))
 ; MIPS64R2-EB-NEXT:    ld $1, %got_disp(arr)($gp)
 ; MIPS64R2-EB-NEXT:    lbu $2, 5($1)
+; MIPS64R2-EB-NEXT:    dsll $2, $2, 16
 ; MIPS64R2-EB-NEXT:    lbu $3, 4($1)
-; MIPS64R2-EB-NEXT:    dsll $3, $3, 8
+; MIPS64R2-EB-NEXT:    dsll $3, $3, 24
 ; MIPS64R2-EB-NEXT:    or $2, $3, $2
-; MIPS64R2-EB-NEXT:    dsll $2, $2, 16
 ; MIPS64R2-EB-NEXT:    lwl $3, 0($1)
 ; MIPS64R2-EB-NEXT:    lwr $3, 3($1)
 ; MIPS64R2-EB-NEXT:    dext $3, $3, 0, 32
 ; MIPS64R2-EB-NEXT:    dsll $3, $3, 32
 ; MIPS64R2-EB-NEXT:    or $2, $3, $2
 ; MIPS64R2-EB-NEXT:    lbu $1, 6($1)
 ; MIPS64R2-EB-NEXT:    dsll $1, $1, 8
 ; MIPS64R2-EB-NEXT:    ld $25, %call16(extern_func)($gp)

llvm/test/CodeGen/Mips/unalignedload.ll

 ; MIPS32-EB:       # %bb.0: # %entry
 ; MIPS32-EB-NEXT:    lui $2, %hi(_gp_disp)
 ; MIPS32-EB-NEXT:    addiu $2, $2, %lo(_gp_disp)
 ; MIPS32-EB-NEXT:    addiu $sp, $sp, -24
 ; MIPS32-EB-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
 ; MIPS32-EB-NEXT:    addu $gp, $2, $25
 ; MIPS32-EB-NEXT:    lw $1, %got(s2)($gp)
 ; MIPS32-EB-NEXT:    lbu $2, 3($1)
+; MIPS32-EB-NEXT:    sll $2, $2, 16
 ; MIPS32-EB-NEXT:    lbu $1, 2($1)
-; MIPS32-EB-NEXT:    sll $1, $1, 8
-; MIPS32-EB-NEXT:    or $1, $1, $2
+; MIPS32-EB-NEXT:    sll $1, $1, 24
 ; MIPS32-EB-NEXT:    lw $25, %call16(foo2)($gp)
 ; MIPS32-EB-NEXT:    .reloc ($tmp0), R_MIPS_JALR, foo2
 ; MIPS32-EB-NEXT:  $tmp0:
 ; MIPS32-EB-NEXT:    jalr $25
-; MIPS32-EB-NEXT:    sll $4, $1, 16
+; MIPS32-EB-NEXT:    or $4, $1, $2
 ; MIPS32-EB-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload
 ; MIPS32-EB-NEXT:    jr $ra
 ; MIPS32-EB-NEXT:    addiu $sp, $sp, 24
 ;
 ; MIPS32R6-EL-LABEL: bar1:
 ; MIPS32R6-EL:       # %bb.0: # %entry
 ; MIPS32R6-EL-NEXT:    lui $2, %hi(_gp_disp)
 ; MIPS32R6-EL-NEXT:    addiu $2, $2, %lo(_gp_disp)

 ; MIPS32-EB:       # %bb.0: # %entry
 ; MIPS32-EB-NEXT:    lui $2, %hi(_gp_disp)
 ; MIPS32-EB-NEXT:    addiu $2, $2, %lo(_gp_disp)
 ; MIPS32-EB-NEXT:    addiu $sp, $sp, -24
 ; MIPS32-EB-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
 ; MIPS32-EB-NEXT:    addu $gp, $2, $25
 ; MIPS32-EB-NEXT:    lw $1, %got(s4)($gp)
 ; MIPS32-EB-NEXT:    lwl $4, 0($1)
-; MIPS32-EB-NEXT:    lwr $4, 3($1)
 ; MIPS32-EB-NEXT:    lbu $2, 5($1)
+; MIPS32-EB-NEXT:    lwr $4, 3($1)
+; MIPS32-EB-NEXT:    sll $2, $2, 16
 ; MIPS32-EB-NEXT:    lbu $3, 4($1)
-; MIPS32-EB-NEXT:    sll $3, $3, 8
+; MIPS32-EB-NEXT:    sll $3, $3, 24
 ; MIPS32-EB-NEXT:    or $2, $3, $2
-; MIPS32-EB-NEXT:    sll $2, $2, 16
 ; MIPS32-EB-NEXT:    lbu $1, 6($1)
 ; MIPS32-EB-NEXT:    sll $1, $1, 8
 ; MIPS32-EB-NEXT:    lw $25, %call16(foo4)($gp)
 ; MIPS32-EB-NEXT:    .reloc ($tmp1), R_MIPS_JALR, foo4
 ; MIPS32-EB-NEXT:  $tmp1:
 ; MIPS32-EB-NEXT:    jalr $25
 ; MIPS32-EB-NEXT:    or $5, $2, $1
 ; MIPS32-EB-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload

llvm/test/CodeGen/RISCV/bswap-bitreverse.ll

; RV64ZBKB-NEXT:    ret
   %tmp = call i64 @llvm.bitreverse.i64(i64 %a)
   %tmp2 = call i64 @llvm.bswap.i64(i64 %tmp)
   ret i64 %tmp2
 }
 
 define i32 @pr55484(i32 %0) {
 ; RV32I-LABEL: pr55484:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    srli a1, a0, 8
-; RV32I-NEXT:    slli a0, a0, 8
-; RV32I-NEXT:    or a0, a1, a0
-; RV32I-NEXT:    slli a0, a0, 16
+; RV32I-NEXT:    slli a1, a0, 8
+; RV32I-NEXT:    slli a0, a0, 24
+; RV32I-NEXT:    or a0, a0, a1
 ; RV32I-NEXT:    srai a0, a0, 16
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: pr55484:
 ; RV64I:       # %bb.0:
-; RV64I-NEXT:    srli a1, a0, 8
-; RV64I-NEXT:    slli a0, a0, 8
-; RV64I-NEXT:    or a0, a1, a0
-; RV64I-NEXT:    slli a0, a0, 48
+; RV64I-NEXT:    slli a1, a0, 40
+; RV64I-NEXT:    slli a0, a0, 56
+; RV64I-NEXT:    or a0, a0, a1
 ; RV64I-NEXT:    srai a0, a0, 48
 ; RV64I-NEXT:    ret
 ;
 ; RV32ZBB-LABEL: pr55484:
 ; RV32ZBB:       # %bb.0:
 ; RV32ZBB-NEXT:    srli a1, a0, 8
 ; RV32ZBB-NEXT:    slli a0, a0, 8
 ; RV32ZBB-NEXT:    or a0, a1, a0
 ; RV32ZBB-NEXT:    sext.h a0, a0
 ; RV32ZBB-NEXT:    ret
 ;
 ; RV64ZBB-LABEL: pr55484:
 ; RV64ZBB:       # %bb.0:
 ; RV64ZBB-NEXT:    srli a1, a0, 8
 ; RV64ZBB-NEXT:    slli a0, a0, 8
 ; RV64ZBB-NEXT:    or a0, a1, a0
 ; RV64ZBB-NEXT:    sext.h a0, a0
 ; RV64ZBB-NEXT:    ret
 ;
 ; RV32ZBKB-LABEL: pr55484:
 ; RV32ZBKB:       # %bb.0:
-; RV32ZBKB-NEXT:    srli a1, a0, 8
-; RV32ZBKB-NEXT:    slli a0, a0, 8
-; RV32ZBKB-NEXT:    or a0, a1, a0
-; RV32ZBKB-NEXT:    slli a0, a0, 16
+; RV32ZBKB-NEXT:    slli a1, a0, 8
+; RV32ZBKB-NEXT:    slli a0, a0, 24
+; RV32ZBKB-NEXT:    or a0, a0, a1
 ; RV32ZBKB-NEXT:    srai a0, a0, 16
 ; RV32ZBKB-NEXT:    ret
 ;
 ; RV64ZBKB-LABEL: pr55484:
 ; RV64ZBKB:       # %bb.0:
-; RV64ZBKB-NEXT:    srli a1, a0, 8
-; RV64ZBKB-NEXT:    slli a0, a0, 8
-; RV64ZBKB-NEXT:    or a0, a1, a0
-; RV64ZBKB-NEXT:    slli a0, a0, 48
+; RV64ZBKB-NEXT:    slli a1, a0, 40
+; RV64ZBKB-NEXT:    slli a0, a0, 56
+; RV64ZBKB-NEXT:    or a0, a0, a1
 ; RV64ZBKB-NEXT:    srai a0, a0, 48
 ; RV64ZBKB-NEXT:    ret
   %2 = lshr i32 %0, 8
   %3 = shl i32 %0, 8
   %4 = or i32 %2, %3
   %5 = trunc i32 %4 to i16
   %6 = sext i16 %5 to i32
   ret i32 %6
 }

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-unaligned.ll

 ; RV32-NEXT:    vmseq.vi v8, v8, 0
 ; RV32-NEXT:    vsetvli zero, zero, e8, mf8, ta, ma
 ; RV32-NEXT:    vmv.x.s a2, v8
 ; RV32-NEXT:    andi a3, a2, 1
 ; RV32-NEXT:    beqz a3, .LBB8_2
 ; RV32-NEXT:  # %bb.1: # %cond.load
 ; RV32-NEXT:    lbu a3, 1(a0)
 ; RV32-NEXT:    lbu a4, 0(a0)
-; RV32-NEXT:    lbu a5, 3(a0)
-; RV32-NEXT:    lbu a6, 2(a0)
+; RV32-NEXT:    lbu a5, 2(a0)
+; RV32-NEXT:    lbu a6, 3(a0)
 ; RV32-NEXT:    slli a3, a3, 8
 ; RV32-NEXT:    or a3, a3, a4
-; RV32-NEXT:    slli a4, a5, 8
-; RV32-NEXT:    or a4, a4, a6
-; RV32-NEXT:    slli a4, a4, 16
+; RV32-NEXT:    slli a4, a5, 16
+; RV32-NEXT:    slli a5, a6, 24
+; RV32-NEXT:    or a4, a5, a4
 ; RV32-NEXT:    or a3, a4, a3
 ; RV32-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
 ; RV32-NEXT:    vmv.v.x v8, a3
 ; RV32-NEXT:    andi a2, a2, 2
 ; RV32-NEXT:    bnez a2, .LBB8_3
 ; RV32-NEXT:    j .LBB8_4
 ; RV32-NEXT:  .LBB8_2:
 ; RV32-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
 ; RV32-NEXT:    vmv.v.i v8, 0
 ; RV32-NEXT:    andi a2, a2, 2
 ; RV32-NEXT:    beqz a2, .LBB8_4
 ; RV32-NEXT:  .LBB8_3: # %cond.load1
 ; RV32-NEXT:    lbu a2, 5(a0)
 ; RV32-NEXT:    lbu a3, 4(a0)
-; RV32-NEXT:    lbu a4, 7(a0)
-; RV32-NEXT:    lbu a0, 6(a0)
+; RV32-NEXT:    lbu a4, 6(a0)
+; RV32-NEXT:    lbu a0, 7(a0)
 ; RV32-NEXT:    slli a2, a2, 8
 ; RV32-NEXT:    or a2, a2, a3
-; RV32-NEXT:    slli a3, a4, 8
-; RV32-NEXT:    or a0, a3, a0
-; RV32-NEXT:    slli a0, a0, 16
+; RV32-NEXT:    slli a3, a4, 16
+; RV32-NEXT:    slli a0, a0, 24
+; RV32-NEXT:    or a0, a0, a3
 ; RV32-NEXT:    or a0, a0, a2
 ; RV32-NEXT:    vmv.s.x v9, a0
 ; RV32-NEXT:    vsetvli zero, zero, e32, mf2, tu, ma
 ; RV32-NEXT:    vslideup.vi v8, v9, 1
 ; RV32-NEXT:  .LBB8_4: # %else2
 ; RV32-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
 ; RV32-NEXT:    vse32.v v8, (a1)
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: masked_load_v2i32_align1:
 ; RV64:       # %bb.0:
 ; RV64-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
 ; RV64-NEXT:    vmseq.vi v8, v8, 0
 ; RV64-NEXT:    vsetvli zero, zero, e8, mf8, ta, ma
 ; RV64-NEXT:    vmv.x.s a2, v8
 ; RV64-NEXT:    andi a3, a2, 1
 ; RV64-NEXT:    beqz a3, .LBB8_2
 ; RV64-NEXT:  # %bb.1: # %cond.load
 ; RV64-NEXT:    lbu a3, 1(a0)
 ; RV64-NEXT:    lbu a4, 0(a0)
-; RV64-NEXT:    lb a5, 3(a0)
-; RV64-NEXT:    lbu a6, 2(a0)
+; RV64-NEXT:    lbu a5, 2(a0)
+; RV64-NEXT:    lb a6, 3(a0)
 ; RV64-NEXT:    slli a3, a3, 8
 ; RV64-NEXT:    or a3, a3, a4
-; RV64-NEXT:    slli a4, a5, 8
-; RV64-NEXT:    or a4, a4, a6
-; RV64-NEXT:    slli a4, a4, 16
+; RV64-NEXT:    slli a4, a5, 16
+; RV64-NEXT:    slli a5, a6, 24
+; RV64-NEXT:    or a4, a5, a4
 ; RV64-NEXT:    or a3, a4, a3
 ; RV64-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
 ; RV64-NEXT:    vmv.v.x v8, a3
 ; RV64-NEXT:    andi a2, a2, 2
 ; RV64-NEXT:    bnez a2, .LBB8_3
 ; RV64-NEXT:    j .LBB8_4
 ; RV64-NEXT:  .LBB8_2:
 ; RV64-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
 ; RV64-NEXT:    vmv.v.i v8, 0
 ; RV64-NEXT:    andi a2, a2, 2
 ; RV64-NEXT:    beqz a2, .LBB8_4
 ; RV64-NEXT:  .LBB8_3: # %cond.load1
 ; RV64-NEXT:    lbu a2, 5(a0)
 ; RV64-NEXT:    lbu a3, 4(a0)
-; RV64-NEXT:    lb a4, 7(a0)
-; RV64-NEXT:    lbu a0, 6(a0)
+; RV64-NEXT:    lbu a4, 6(a0)
+; RV64-NEXT:    lb a0, 7(a0)
 ; RV64-NEXT:    slli a2, a2, 8
 ; RV64-NEXT:    or a2, a2, a3
-; RV64-NEXT:    slli a3, a4, 8
-; RV64-NEXT:    or a0, a3, a0
-; RV64-NEXT:    slli a0, a0, 16
+; RV64-NEXT:    slli a3, a4, 16
+; RV64-NEXT:    slli a0, a0, 24
+; RV64-NEXT:    or a0, a0, a3
 ; RV64-NEXT:    or a0, a0, a2
 ; RV64-NEXT:    vmv.s.x v9, a0
 ; RV64-NEXT:    vsetvli zero, zero, e32, mf2, tu, ma
 ; RV64-NEXT:    vslideup.vi v8, v9, 1
 ; RV64-NEXT:  .LBB8_4: # %else2
 ; RV64-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
 ; RV64-NEXT:    vse32.v v8, (a1)
 ; RV64-NEXT:    ret

llvm/test/CodeGen/RISCV/srem-seteq-illegal-types.ll

 ; RV64-NEXT:    sd s0, 32(sp) # 8-byte Folded Spill
 ; RV64-NEXT:    sd s1, 24(sp) # 8-byte Folded Spill
 ; RV64-NEXT:    sd s2, 16(sp) # 8-byte Folded Spill
 ; RV64-NEXT:    sd s3, 8(sp) # 8-byte Folded Spill
 ; RV64-NEXT:    mv s0, a0
 ; RV64-NEXT:    lb a0, 12(a0)
 ; RV64-NEXT:    lwu a1, 8(s0)
 ; RV64-NEXT:    slli a0, a0, 32
-; RV64-NEXT:    or a0, a1, a0
 ; RV64-NEXT:    ld a2, 0(s0)
+; RV64-NEXT:    or a0, a1, a0
 ; RV64-NEXT:    slli a0, a0, 29
 ; RV64-NEXT:    srai s1, a0, 31
-; RV64-NEXT:    slli a0, a1, 31
-; RV64-NEXT:    srli a1, a2, 33
+; RV64-NEXT:    srli a0, a2, 2
+; RV64-NEXT:    slli a1, a1, 62
 ; RV64-NEXT:    or a0, a1, a0
-; RV64-NEXT:    slli a0, a0, 31
 ; RV64-NEXT:    srai a0, a0, 31
 ; RV64-NEXT:    slli a1, a2, 31
 ; RV64-NEXT:    srai s2, a1, 31
 ; RV64-NEXT:    li a1, 7
 ; RV64-NEXT:    call __moddi3@plt
 ; RV64-NEXT:    mv s3, a0
 ; RV64-NEXT:    li a1, -5
 ; RV64-NEXT:    mv a0, s1

 ; RV64-NEXT:    sltu a0, a1, a0
 ; RV64-NEXT:    addi a1, s1, -2
 ; RV64-NEXT:    seqz a1, a1
 ; RV64-NEXT:    addi a2, s3, -1
 ; RV64-NEXT:    seqz a2, a2
 ; RV64-NEXT:    neg a0, a0
 ; RV64-NEXT:    addi a2, a2, -1
 ; RV64-NEXT:    addi a1, a1, -1
-; RV64-NEXT:    slli a3, a1, 29
-; RV64-NEXT:    srli a3, a3, 61
-; RV64-NEXT:    sb a3, 12(s0)
-; RV64-NEXT:    slli a1, a1, 2
-; RV64-NEXT:    slli a3, a2, 31
-; RV64-NEXT:    srli a3, a3, 62
-; RV64-NEXT:    or a1, a3, a1
-; RV64-NEXT:    sw a1, 8(s0)
+; RV64-NEXT:    slli a3, a1, 2
+; RV64-NEXT:    slli a4, a2, 31
+; RV64-NEXT:    srli a4, a4, 62
+; RV64-NEXT:    or a3, a4, a3
+; RV64-NEXT:    sw a3, 8(s0)
+; RV64-NEXT:    slli a1, a1, 29
+; RV64-NEXT:    srli a1, a1, 61
+; RV64-NEXT:    sb a1, 12(s0)
 ; RV64-NEXT:    slli a0, a0, 31
 ; RV64-NEXT:    srli a0, a0, 31
 ; RV64-NEXT:    slli a1, a2, 33
 ; RV64-NEXT:    or a0, a0, a1
 ; RV64-NEXT:    sd a0, 0(s0)
 ; RV64-NEXT:    ld ra, 40(sp) # 8-byte Folded Reload
 ; RV64-NEXT:    ld s0, 32(sp) # 8-byte Folded Reload
 ; RV64-NEXT:    ld s1, 24(sp) # 8-byte Folded Reload

 ; RV32M-NEXT:    lw s4, 8(sp) # 4-byte Folded Reload
 ; RV32M-NEXT:    lw s5, 4(sp) # 4-byte Folded Reload
 ; RV32M-NEXT:    lw s6, 0(sp) # 4-byte Folded Reload
 ; RV32M-NEXT:    addi sp, sp, 32
 ; RV32M-NEXT:    ret
 ;
 ; RV64M-LABEL: test_srem_vec:
 ; RV64M:       # %bb.0:
-; RV64M-NEXT:    lb a1, 12(a0)
+; RV64M-NEXT:    ld a1, 0(a0)
 ; RV64M-NEXT:    lwu a2, 8(a0)
-; RV64M-NEXT:    slli a1, a1, 32
-; RV64M-NEXT:    or a1, a2, a1
-; RV64M-NEXT:    ld a3, 0(a0)
-; RV64M-NEXT:    slli a1, a1, 29
-; RV64M-NEXT:    srai a1, a1, 31
-; RV64M-NEXT:    slli a2, a2, 31
-; RV64M-NEXT:    srli a4, a3, 33
+; RV64M-NEXT:    srli a3, a1, 2
+; RV64M-NEXT:    lb a4, 12(a0)
+; RV64M-NEXT:    slli a5, a2, 62
+; RV64M-NEXT:    or a3, a5, a3
+; RV64M-NEXT:    srai a3, a3, 31
+; RV64M-NEXT:    slli a4, a4, 32
 ; RV64M-NEXT:    lui a5, %hi(.LCPI3_0)
 ; RV64M-NEXT:    ld a5, %lo(.LCPI3_0)(a5)
-; RV64M-NEXT:    or a2, a4, a2
-; RV64M-NEXT:    slli a2, a2, 31
+; RV64M-NEXT:    or a2, a2, a4
+; RV64M-NEXT:    slli a2, a2, 29
 ; RV64M-NEXT:    srai a2, a2, 31
 ; RV64M-NEXT:    mulh a4, a2, a5
 ; RV64M-NEXT:    srli a5, a4, 63
 ; RV64M-NEXT:    srai a4, a4, 1
 ; RV64M-NEXT:    add a4, a4, a5
-; RV64M-NEXT:    slli a5, a4, 3
-; RV64M-NEXT:    sub a4, a4, a5
+; RV64M-NEXT:    slli a5, a4, 2
+; RV64M-NEXT:    add a4, a5, a4
 ; RV64M-NEXT:    lui a5, %hi(.LCPI3_1)
 ; RV64M-NEXT:    ld a5, %lo(.LCPI3_1)(a5)
-; RV64M-NEXT:    slli a3, a3, 31
-; RV64M-NEXT:    srai a3, a3, 31
+; RV64M-NEXT:    slli a1, a1, 31
+; RV64M-NEXT:    srai a1, a1, 31
 ; RV64M-NEXT:    add a2, a2, a4
-; RV64M-NEXT:    mulh a4, a1, a5
+; RV64M-NEXT:    mulh a4, a3, a5
 ; RV64M-NEXT:    srli a5, a4, 63
 ; RV64M-NEXT:    srai a4, a4, 1
 ; RV64M-NEXT:    add a4, a4, a5
-; RV64M-NEXT:    slli a5, a4, 2
-; RV64M-NEXT:    add a4, a5, a4
-; RV64M-NEXT:    add a1, a1, a4
-; RV64M-NEXT:    addi a1, a1, -2
-; RV64M-NEXT:    seqz a1, a1
+; RV64M-NEXT:    slli a5, a4, 3
+; RV64M-NEXT:    sub a4, a4, a5
+; RV64M-NEXT:    add a3, a3, a4
+; RV64M-NEXT:    addi a3, a3, -1
+; RV64M-NEXT:    seqz a3, a3
 ; RV64M-NEXT:    lui a4, %hi(.LCPI3_2)
 ; RV64M-NEXT:    ld a4, %lo(.LCPI3_2)(a4)
 ; RV64M-NEXT:    lui a5, %hi(.LCPI3_3)
 ; RV64M-NEXT:    ld a5, %lo(.LCPI3_3)(a5)
-; RV64M-NEXT:    addi a2, a2, -1
+; RV64M-NEXT:    addi a2, a2, -2
 ; RV64M-NEXT:    seqz a2, a2
-; RV64M-NEXT:    mul a3, a3, a4
-; RV64M-NEXT:    add a3, a3, a5
-; RV64M-NEXT:    slli a4, a3, 63
-; RV64M-NEXT:    srli a3, a3, 1
-; RV64M-NEXT:    or a3, a3, a4
-; RV64M-NEXT:    sltu a3, a5, a3
+; RV64M-NEXT:    mul a1, a1, a4
+; RV64M-NEXT:    add a1, a1, a5
+; RV64M-NEXT:    slli a4, a1, 63
+; RV64M-NEXT:    srli a1, a1, 1
+; RV64M-NEXT:    or a1, a1, a4
+; RV64M-NEXT:    sltu a1, a5, a1
 ; RV64M-NEXT:    addi a2, a2, -1
-; RV64M-NEXT:    addi a1, a1, -1
-; RV64M-NEXT:    neg a3, a3
-; RV64M-NEXT:    slli a4, a1, 29
-; RV64M-NEXT:    srli a4, a4, 61
-; RV64M-NEXT:    sb a4, 12(a0)
-; RV64M-NEXT:    slli a4, a2, 33
+; RV64M-NEXT:    addi a3, a3, -1
+; RV64M-NEXT:    neg a1, a1
+; RV64M-NEXT:    slli a4, a3, 33
+; RV64M-NEXT:    slli a1, a1, 31
+; RV64M-NEXT:    srli a1, a1, 31
+; RV64M-NEXT:    or a1, a1, a4
+; RV64M-NEXT:    sd a1, 0(a0)
+; RV64M-NEXT:    slli a1, a2, 2
 ; RV64M-NEXT:    slli a3, a3, 31
-; RV64M-NEXT:    srli a3, a3, 31
-; RV64M-NEXT:    or a3, a3, a4
-; RV64M-NEXT:    sd a3, 0(a0)
-; RV64M-NEXT:    slli a1, a1, 2
-; RV64M-NEXT:    slli a2, a2, 31
-; RV64M-NEXT:    srli a2, a2, 62
-; RV64M-NEXT:    or a1, a2, a1
+; RV64M-NEXT:    srli a3, a3, 62
+; RV64M-NEXT:    or a1, a3, a1
 ; RV64M-NEXT:    sw a1, 8(a0)
+; RV64M-NEXT:    slli a1, a2, 29
+; RV64M-NEXT:    srli a1, a1, 61
+; RV64M-NEXT:    sb a1, 12(a0)
 ; RV64M-NEXT:    ret
 ;
 ; RV32MV-LABEL: test_srem_vec:
 ; RV32MV:       # %bb.0:
 ; RV32MV-NEXT:    addi sp, sp, -64
 ; RV32MV-NEXT:    sw ra, 60(sp) # 4-byte Folded Spill
 ; RV32MV-NEXT:    sw s0, 56(sp) # 4-byte Folded Spill
 ; RV32MV-NEXT:    sw s2, 52(sp) # 4-byte Folded Spill

 ;
 ; RV64MV-LABEL: test_srem_vec:
 ; RV64MV:       # %bb.0:
 ; RV64MV-NEXT:    addi sp, sp, -64
 ; RV64MV-NEXT:    sd ra, 56(sp) # 8-byte Folded Spill
 ; RV64MV-NEXT:    sd s0, 48(sp) # 8-byte Folded Spill
 ; RV64MV-NEXT:    addi s0, sp, 64
 ; RV64MV-NEXT:    andi sp, sp, -32
-; RV64MV-NEXT:    lwu a1, 8(a0)
-; RV64MV-NEXT:    ld a2, 0(a0)
-; RV64MV-NEXT:    slli a3, a1, 31
-; RV64MV-NEXT:    srli a4, a2, 33
-; RV64MV-NEXT:    lb a5, 12(a0)
-; RV64MV-NEXT:    or a3, a4, a3
+; RV64MV-NEXT:    lb a1, 12(a0)
+; RV64MV-NEXT:    lwu a2, 8(a0)
+; RV64MV-NEXT:    slli a1, a1, 32
+; RV64MV-NEXT:    ld a3, 0(a0)
+; RV64MV-NEXT:    or a1, a2, a1
+; RV64MV-NEXT:    slli a1, a1, 29
+; RV64MV-NEXT:    srai a1, a1, 31
+; RV64MV-NEXT:    srli a4, a3, 2
+; RV64MV-NEXT:    slli a2, a2, 62
+; RV64MV-NEXT:    lui a5, %hi(.LCPI3_0)
+; RV64MV-NEXT:    ld a5, %lo(.LCPI3_0)(a5)
+; RV64MV-NEXT:    or a2, a2, a4
 ; RV64MV-NEXT:    slli a3, a3, 31
 ; RV64MV-NEXT:    srai a3, a3, 31
-; RV64MV-NEXT:    slli a4, a5, 32
-; RV64MV-NEXT:    or a1, a1, a4
-; RV64MV-NEXT:    lui a4, %hi(.LCPI3_0)
-; RV64MV-NEXT:    ld a4, %lo(.LCPI3_0)(a4)
-; RV64MV-NEXT:    slli a1, a1, 29
-; RV64MV-NEXT:    slli a2, a2, 31
-; RV64MV-NEXT:    srai a2, a2, 31
-; RV64MV-NEXT:    mulh a4, a2, a4
+; RV64MV-NEXT:    mulh a4, a3, a5
 ; RV64MV-NEXT:    srli a5, a4, 63
 ; RV64MV-NEXT:    add a4, a4, a5
 ; RV64MV-NEXT:    li a5, 6
 ; RV64MV-NEXT:    mul a4, a4, a5
 ; RV64MV-NEXT:    lui a5, %hi(.LCPI3_1)
 ; RV64MV-NEXT:    ld a5, %lo(.LCPI3_1)(a5)
-; RV64MV-NEXT:    srai a1, a1, 31
-; RV64MV-NEXT:    sub a2, a2, a4
-; RV64MV-NEXT:    sd a2, 0(sp)
-; RV64MV-NEXT:    mulh a2, a1, a5
-; RV64MV-NEXT:    srli a4, a2, 63
-; RV64MV-NEXT:    srai a2, a2, 1
-; RV64MV-NEXT:    add a2, a2, a4
-; RV64MV-NEXT:    slli a4, a2, 2
+; RV64MV-NEXT:    srai a2, a2, 31
+; RV64MV-NEXT:    sub a3, a3, a4
+; RV64MV-NEXT:    sd a3, 0(sp)
+; RV64MV-NEXT:    mulh a3, a2, a5
+; RV64MV-NEXT:    srli a4, a3, 63
+; RV64MV-NEXT:    srai a3, a3, 1
+; RV64MV-NEXT:    add a3, a3, a4
+; RV64MV-NEXT:    slli a4, a3, 3
 ; RV64MV-NEXT:    lui a5, %hi(.LCPI3_2)
 ; RV64MV-NEXT:    ld a5, %lo(.LCPI3_2)(a5)
-; RV64MV-NEXT:    add a2, a4, a2
+; RV64MV-NEXT:    sub a3, a3, a4
+; RV64MV-NEXT:    add a2, a2, a3
+; RV64MV-NEXT:    sd a2, 8(sp)
+; RV64MV-NEXT:    mulh a2, a1, a5
+; RV64MV-NEXT:    srli a3, a2, 63
+; RV64MV-NEXT:    srai a2, a2, 1
+; RV64MV-NEXT:    add a2, a2, a3
+; RV64MV-NEXT:    slli a3, a2, 2
+; RV64MV-NEXT:    add a2, a3, a2
 ; RV64MV-NEXT:    add a1, a1, a2
 ; RV64MV-NEXT:    sd a1, 16(sp)
-; RV64MV-NEXT:    mulh a1, a3, a5
-; RV64MV-NEXT:    srli a2, a1, 63
-; RV64MV-NEXT:    srai a1, a1, 1
-; RV64MV-NEXT:    add a1, a1, a2
-; RV64MV-NEXT:    slli a2, a1, 3
-; RV64MV-NEXT:    sub a1, a1, a2
-; RV64MV-NEXT:    add a1, a3, a1
-; RV64MV-NEXT:    sd a1, 8(sp)
 ; RV64MV-NEXT:    mv a1, sp
 ; RV64MV-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
 ; RV64MV-NEXT:    vle64.v v8, (a1)
 ; RV64MV-NEXT:    lui a1, %hi(.LCPI3_3)
 ; RV64MV-NEXT:    addi a1, a1, %lo(.LCPI3_3)
 ; RV64MV-NEXT:    vle64.v v10, (a1)
 ; RV64MV-NEXT:    li a1, -1
 ; RV64MV-NEXT:    srli a1, a1, 31

llvm/test/CodeGen/RISCV/unaligned-load-store.ll

; MISALIGN-NEXT:    ret
   ret i24 %res
 }
 
 define i32 @load_i32(i32* %p) {
 ; RV32I-LABEL: load_i32:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    lbu a1, 1(a0)
 ; RV32I-NEXT:    lbu a2, 0(a0)
-; RV32I-NEXT:    lbu a3, 3(a0)
-; RV32I-NEXT:    lbu a0, 2(a0)
+; RV32I-NEXT:    lbu a3, 2(a0)
+; RV32I-NEXT:    lbu a0, 3(a0)
 ; RV32I-NEXT:    slli a1, a1, 8
 ; RV32I-NEXT:    or a1, a1, a2
-; RV32I-NEXT:    slli a2, a3, 8
-; RV32I-NEXT:    or a0, a2, a0
-; RV32I-NEXT:    slli a0, a0, 16
+; RV32I-NEXT:    slli a2, a3, 16
+; RV32I-NEXT:    slli a0, a0, 24
+; RV32I-NEXT:    or a0, a0, a2
 ; RV32I-NEXT:    or a0, a0, a1
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: load_i32:
 ; RV64I:       # %bb.0:
 ; RV64I-NEXT:    lbu a1, 1(a0)
 ; RV64I-NEXT:    lbu a2, 0(a0)
-; RV64I-NEXT:    lb a3, 3(a0)
-; RV64I-NEXT:    lbu a0, 2(a0)
+; RV64I-NEXT:    lbu a3, 2(a0)
+; RV64I-NEXT:    lb a0, 3(a0)
 ; RV64I-NEXT:    slli a1, a1, 8
 ; RV64I-NEXT:    or a1, a1, a2
-; RV64I-NEXT:    slli a2, a3, 8
-; RV64I-NEXT:    or a0, a2, a0
-; RV64I-NEXT:    slli a0, a0, 16
+; RV64I-NEXT:    slli a2, a3, 16
+; RV64I-NEXT:    slli a0, a0, 24
+; RV64I-NEXT:    or a0, a0, a2
 ; RV64I-NEXT:    or a0, a0, a1
 ; RV64I-NEXT:    ret
 ;
 ; MISALIGN-LABEL: load_i32:
 ; MISALIGN:       # %bb.0:
 ; MISALIGN-NEXT:    lw a0, 0(a0)
 ; MISALIGN-NEXT:    ret
   %res = load i32, i32* %p, align 1
   ret i32 %res
 }
 
 define i64 @load_i64(i64* %p) {
 ; RV32I-LABEL: load_i64:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    lbu a1, 1(a0)
 ; RV32I-NEXT:    lbu a2, 0(a0)
-; RV32I-NEXT:    lbu a3, 3(a0)
-; RV32I-NEXT:    lbu a4, 2(a0)
+; RV32I-NEXT:    lbu a3, 2(a0)
+; RV32I-NEXT:    lbu a4, 3(a0)
 ; RV32I-NEXT:    slli a1, a1, 8
 ; RV32I-NEXT:    or a1, a1, a2
-; RV32I-NEXT:    slli a2, a3, 8
-; RV32I-NEXT:    or a2, a2, a4
-; RV32I-NEXT:    slli a2, a2, 16
+; RV32I-NEXT:    slli a2, a3, 16
+; RV32I-NEXT:    slli a3, a4, 24
+; RV32I-NEXT:    or a2, a3, a2
 ; RV32I-NEXT:    or a2, a2, a1
 ; RV32I-NEXT:    lbu a1, 5(a0)
 ; RV32I-NEXT:    lbu a3, 4(a0)
-; RV32I-NEXT:    lbu a4, 7(a0)
-; RV32I-NEXT:    lbu a0, 6(a0)
+; RV32I-NEXT:    lbu a4, 6(a0)
+; RV32I-NEXT:    lbu a0, 7(a0)
 ; RV32I-NEXT:    slli a1, a1, 8
 ; RV32I-NEXT:    or a1, a1, a3
-; RV32I-NEXT:    slli a3, a4, 8
-; RV32I-NEXT:    or a0, a3, a0
-; RV32I-NEXT:    slli a0, a0, 16
+; RV32I-NEXT:    slli a3, a4, 16
+; RV32I-NEXT:    slli a0, a0, 24
+; RV32I-NEXT:    or a0, a0, a3
 ; RV32I-NEXT:    or a1, a0, a1
 ; RV32I-NEXT:    mv a0, a2
 ; RV32I-NEXT:    ret
 ;
 ; RV64I-LABEL: load_i64:
 ; RV64I:       # %bb.0:
 ; RV64I-NEXT:    lbu a1, 1(a0)
 ; RV64I-NEXT:    lbu a2, 0(a0)
-; RV64I-NEXT:    lbu a3, 3(a0)
-; RV64I-NEXT:    lbu a4, 2(a0)
+; RV64I-NEXT:    lbu a3, 2(a0)
+; RV64I-NEXT:    lbu a4, 3(a0)
 ; RV64I-NEXT:    slli a1, a1, 8
 ; RV64I-NEXT:    or a1, a1, a2
-; RV64I-NEXT:    slli a2, a3, 8
-; RV64I-NEXT:    or a2, a2, a4
-; RV64I-NEXT:    slli a2, a2, 16
+; RV64I-NEXT:    slli a2, a3, 16
+; RV64I-NEXT:    slli a3, a4, 24
+; RV64I-NEXT:    or a2, a3, a2
 ; RV64I-NEXT:    or a1, a2, a1
 ; RV64I-NEXT:    lbu a2, 5(a0)
 ; RV64I-NEXT:    lbu a3, 4(a0)
-; RV64I-NEXT:    lbu a4, 7(a0)
-; RV64I-NEXT:    lbu a0, 6(a0)
+; RV64I-NEXT:    lbu a4, 6(a0)
+; RV64I-NEXT:    lbu a0, 7(a0)
 ; RV64I-NEXT:    slli a2, a2, 8
 ; RV64I-NEXT:    or a2, a2, a3
-; RV64I-NEXT:    slli a3, a4, 8
-; RV64I-NEXT:    or a0, a3, a0
-; RV64I-NEXT:    slli a0, a0, 16
+; RV64I-NEXT:    slli a3, a4, 16
+; RV64I-NEXT:    slli a0, a0, 24
+; RV64I-NEXT:    or a0, a0, a3
 ; RV64I-NEXT:    or a0, a0, a2
 ; RV64I-NEXT:    slli a0, a0, 32
 ; RV64I-NEXT:    or a0, a0, a1
 ; RV64I-NEXT:    ret
 ;
 ; MISALIGN-RV32I-LABEL: load_i64:
 ; MISALIGN-RV32I:       # %bb.0:
 ; MISALIGN-RV32I-NEXT:    lw a2, 0(a0)

llvm/test/CodeGen/SystemZ/store_nonbytesized_vecs.ll

; CHECK-NEXT:    br %r14
   %res = bitcast <16 x i1> %src to i16
   ret i16 %res
 }
 
 ; Truncate a <8 x i32> vector to <8 x i31> and store it (test splitting).
 define void @fun2(<8 x i32> %src, ptr %p)
 ; CHECK-LABEL: fun2:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    stmg %r14, %r15, 112(%r15)
-; CHECK-NEXT:    .cfi_offset %r14, -48
-; CHECK-NEXT:    .cfi_offset %r15, -40
 ; CHECK-NEXT:    vlgvf %r1, %v26, 3
-; CHECK-NEXT:    vlgvf %r0, %v26, 2
+; CHECK-NEXT:    vlgvf %r5, %v24, 0
+; CHECK-NEXT:    vlgvf %r3, %v24, 1
+; CHECK-NEXT:    srlk %r0, %r1, 8
+; CHECK-NEXT:    sllg %r5, %r5, 33
+; CHECK-NEXT:    sth %r0, 28(%r2)
+; CHECK-NEXT:    rosbg %r5, %r3, 31, 55, 2
+; CHECK-NEXT:    vlgvf %r0, %v24, 2
+; CHECK-NEXT:    sllg %r4, %r3, 58
+; CHECK-NEXT:    vlgvf %r3, %v26, 2
 ; CHECK-NEXT:    stc %r1, 30(%r2)
-; CHECK-NEXT:    srlk %r3, %r1, 8
+; CHECK-NEXT:    rosbg %r4, %r0, 6, 36, 27
 ; CHECK-NEXT:    risbgn %r1, %r1, 33, 167, 0
-; CHECK-NEXT:    vlgvf %r5, %v24, 2
-; CHECK-NEXT:    rosbg %r1, %r0, 2, 32, 31
-; CHECK-NEXT:    sth %r3, 28(%r2)
+; CHECK-NEXT:    rosbg %r1, %r3, 2, 32, 31
 ; CHECK-NEXT:    srlg %r1, %r1, 24
-; CHECK-NEXT:    vlgvf %r3, %v24, 3
+; CHECK-NEXT:    rosbg %r5, %r4, 56, 63, 8
+; CHECK-NEXT:    vlgvf %r4, %v24, 3
 ; CHECK-NEXT:    st %r1, 24(%r2)
 ; CHECK-NEXT:    vlgvf %r1, %v26, 0
-; CHECK-NEXT:    risbgn %r14, %r5, 6, 164, 27
-; CHECK-NEXT:    sllg %r4, %r3, 60
-; CHECK-NEXT:    rosbg %r14, %r3, 37, 63, 60
-; CHECK-NEXT:    sllg %r3, %r14, 8
-; CHECK-NEXT:    rosbg %r4, %r1, 4, 34, 29
-; CHECK-NEXT:    rosbg %r3, %r4, 56, 63, 8
-; CHECK-NEXT:    stg %r3, 8(%r2)
-; CHECK-NEXT:    vlgvf %r3, %v24, 1
-; CHECK-NEXT:    sllg %r4, %r3, 58
-; CHECK-NEXT:    rosbg %r4, %r5, 6, 36, 27
-; CHECK-NEXT:    vlgvf %r5, %v24, 0
-; CHECK-NEXT:    sllg %r5, %r5, 25
-; CHECK-NEXT:    rosbg %r5, %r3, 39, 63, 58
-; CHECK-NEXT:    sllg %r3, %r5, 8
-; CHECK-NEXT:    rosbg %r3, %r4, 56, 63, 8
-; CHECK-NEXT:    stg %r3, 0(%r2)
-; CHECK-NEXT:    vlgvf %r3, %v26, 1
-; CHECK-NEXT:    sllg %r4, %r3, 62
-; CHECK-NEXT:    rosbg %r4, %r0, 2, 32, 31
-; CHECK-NEXT:    risbgn %r0, %r1, 4, 162, 29
-; CHECK-NEXT:    rosbg %r0, %r3, 35, 63, 62
+; CHECK-NEXT:    risbgn %r0, %r0, 6, 164, 27
+; CHECK-NEXT:    rosbg %r0, %r4, 37, 63, 60
+; CHECK-NEXT:    stg %r5, 0(%r2)
+; CHECK-NEXT:    sllg %r5, %r4, 60
 ; CHECK-NEXT:    sllg %r0, %r0, 8
+; CHECK-NEXT:    rosbg %r5, %r1, 4, 34, 29
+; CHECK-NEXT:    risbgn %r1, %r1, 4, 162, 29
+; CHECK-NEXT:    rosbg %r0, %r5, 56, 63, 8
+; CHECK-NEXT:    stg %r0, 8(%r2)
+; CHECK-NEXT:    vlgvf %r0, %v26, 1
+; CHECK-NEXT:    sllg %r4, %r0, 62
+; CHECK-NEXT:    rosbg %r1, %r0, 35, 63, 62
+; CHECK-NEXT:    sllg %r0, %r1, 8
+; CHECK-NEXT:    rosbg %r4, %r3, 2, 32, 31
 ; CHECK-NEXT:    rosbg %r0, %r4, 56, 63, 8
 ; CHECK-NEXT:    stg %r0, 16(%r2)
-; CHECK-NEXT:    lmg %r14, %r15, 112(%r15)
 ; CHECK-NEXT:    br %r14
 {
   %tmp = trunc <8 x i32> %src to <8 x i31>
   store <8 x i31> %tmp, ptr %p
   ret void
 }
 
 ; Load and store a <3 x i31> vector (test widening).

llvm/test/CodeGen/Thumb/urem-seteq-illegal-types.ll

; CHECK-NEXT:    .long 859308032 @ 0x33380000
   ret i1 %cmp
 }
 
 define i1 @test_urem_even(i27 %X) nounwind {
 ; CHECK-LABEL: test_urem_even:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    ldr r1, .LCPI1_0
 ; CHECK-NEXT:    muls r1, r0, r1
-; CHECK-NEXT:    lsls r0, r1, #26
+; CHECK-NEXT:    lsls r0, r1, #31
 ; CHECK-NEXT:    ldr r2, .LCPI1_1
 ; CHECK-NEXT:    ands r2, r1
 ; CHECK-NEXT:    lsrs r1, r2, #1
-; CHECK-NEXT:    adds r0, r1, r0
-; CHECK-NEXT:    lsls r0, r0, #5
+; CHECK-NEXT:    lsls r1, r1, #5
+; CHECK-NEXT:    adds r0, r0, r1
 ; CHECK-NEXT:    ldr r1, .LCPI1_2
 ; CHECK-NEXT:    cmp r0, r1
 ; CHECK-NEXT:    blo .LBB1_2
 ; CHECK-NEXT:  @ %bb.1:
 ; CHECK-NEXT:    movs r0, #0
 ; CHECK-NEXT:    bx lr
 ; CHECK-NEXT:  .LBB1_2:
 ; CHECK-NEXT:    movs r0, #1

llvm/test/CodeGen/X86/bool-vector.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=-sse2 | FileCheck %s --check-prefix=X86
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=-sse2 | FileCheck %s --check-prefix=X64
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX2
 
 define i32 @PR15215_bad(<4 x i32> %input) {
 ; X86-LABEL: PR15215_bad:
 ; X86:       # %bb.0: # %entry
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %ah
-; X86-NEXT:    addb %ah, %ah
+; X86-NEXT:    shlb $3, %ah
 ; X86-NEXT:    andb $1, %cl
-; X86-NEXT:    orb %ah, %cl
 ; X86-NEXT:    shlb $2, %cl
+; X86-NEXT:    orb %ah, %cl
 ; X86-NEXT:    addb %dl, %dl
 ; X86-NEXT:    andb $1, %al
 ; X86-NEXT:    orb %dl, %al
 ; X86-NEXT:    andb $3, %al
 ; X86-NEXT:    orb %cl, %al
 ; X86-NEXT:    movzbl %al, %eax
 ; X86-NEXT:    andl $15, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: PR15215_bad:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    addb %cl, %cl
+; X64-NEXT:    shlb $3, %cl
 ; X64-NEXT:    andb $1, %dl
-; X64-NEXT:    orb %cl, %dl
 ; X64-NEXT:    shlb $2, %dl
+; X64-NEXT:    orb %cl, %dl
 ; X64-NEXT:    addb %sil, %sil
 ; X64-NEXT:    andb $1, %dil
 ; X64-NEXT:    orb %sil, %dil
 ; X64-NEXT:    andb $3, %dil
 ; X64-NEXT:    orb %dl, %dil
 ; X64-NEXT:    movzbl %dil, %eax
 ; X64-NEXT:    andl $15, %eax
 ; X64-NEXT:    retq

llvm/test/CodeGen/X86/combine-bitreverse.ll

 ; X86-NEXT:    shll $4, %ecx
 ; X86-NEXT:    shrl $4, %eax
 ; X86-NEXT:    andl $252645135, %eax # imm = 0xF0F0F0F
 ; X86-NEXT:    orl %ecx, %eax
 ; X86-NEXT:    movl %eax, %ecx
 ; X86-NEXT:    andl $858993459, %ecx # imm = 0x33333333
 ; X86-NEXT:    shrl $2, %eax
 ; X86-NEXT:    andl $858993459, %eax # imm = 0x33333333
-; X86-NEXT:    leal (%eax,%ecx,4), %eax
-; X86-NEXT:    movl %eax, %ecx
-; X86-NEXT:    andl $5592405, %ecx # imm = 0x555555
-; X86-NEXT:    shrl %eax
-; X86-NEXT:    andl $22369621, %eax # imm = 0x1555555
-; X86-NEXT:    leal (%eax,%ecx,2), %eax
-; X86-NEXT:    shll $7, %eax
+; X86-NEXT:    leal (%eax,%ecx,4), %ecx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    andl $5592405, %eax # imm = 0x555555
+; X86-NEXT:    shll $6, %ecx
+; X86-NEXT:    andl $-1431655808, %ecx # imm = 0xAAAAAA80
+; X86-NEXT:    shll $8, %eax
+; X86-NEXT:    orl %ecx, %eax
 ; X86-NEXT:    bswapl %eax
 ; X86-NEXT:    movl %eax, %ecx
 ; X86-NEXT:    andl $986895, %ecx # imm = 0xF0F0F
 ; X86-NEXT:    shll $4, %ecx
 ; X86-NEXT:    shrl $4, %eax
 ; X86-NEXT:    andl $135204623, %eax # imm = 0x80F0F0F
 ; X86-NEXT:    orl %ecx, %eax
 ; X86-NEXT:    movl %eax, %ecx

 ; X64-NEXT:    orl %eax, %edi
 ; X64-NEXT:    movl %edi, %eax
 ; X64-NEXT:    andl $858993459, %eax # imm = 0x33333333
 ; X64-NEXT:    shrl $2, %edi
 ; X64-NEXT:    andl $858993459, %edi # imm = 0x33333333
 ; X64-NEXT:    leal (%rdi,%rax,4), %eax
 ; X64-NEXT:    movl %eax, %ecx
 ; X64-NEXT:    andl $5592405, %ecx # imm = 0x555555
-; X64-NEXT:    shrl %eax
-; X64-NEXT:    andl $22369621, %eax # imm = 0x1555555
-; X64-NEXT:    leal (%rax,%rcx,2), %eax
-; X64-NEXT:    shll $7, %eax
-; X64-NEXT:    bswapl %eax
-; X64-NEXT:    movl %eax, %ecx
-; X64-NEXT:    andl $986895, %ecx # imm = 0xF0F0F
-; X64-NEXT:    shll $4, %ecx
-; X64-NEXT:    shrl $4, %eax
-; X64-NEXT:    andl $135204623, %eax # imm = 0x80F0F0F
-; X64-NEXT:    orl %ecx, %eax
-; X64-NEXT:    movl %eax, %ecx
-; X64-NEXT:    andl $3355443, %ecx # imm = 0x333333
-; X64-NEXT:    shrl $2, %eax
-; X64-NEXT:    andl $36909875, %eax # imm = 0x2333333
-; X64-NEXT:    leal (%rax,%rcx,4), %eax
+; X64-NEXT:    shll $6, %eax
+; X64-NEXT:    andl $-1431655808, %eax # imm = 0xAAAAAA80
+; X64-NEXT:    shll $8, %ecx
+; X64-NEXT:    orl %eax, %ecx
+; X64-NEXT:    bswapl %ecx
+; X64-NEXT:    movl %ecx, %eax
+; X64-NEXT:    andl $986895, %eax # imm = 0xF0F0F
+; X64-NEXT:    shll $4, %eax
+; X64-NEXT:    shrl $4, %ecx
+; X64-NEXT:    andl $135204623, %ecx # imm = 0x80F0F0F
+; X64-NEXT:    orl %eax, %ecx
+; X64-NEXT:    movl %ecx, %eax
+; X64-NEXT:    andl $3355443, %eax # imm = 0x333333
+; X64-NEXT:    shrl $2, %ecx
+; X64-NEXT:    andl $36909875, %ecx # imm = 0x2333333
+; X64-NEXT:    leal (%rcx,%rax,4), %eax
 ; X64-NEXT:    movl %eax, %ecx
 ; X64-NEXT:    andl $1431655765, %ecx # imm = 0x55555555
 ; X64-NEXT:    shrl %eax
 ; X64-NEXT:    andl $1431655765, %eax # imm = 0x55555555
 ; X64-NEXT:    leal (%rax,%rcx,2), %eax
 ; X64-NEXT:    retq
   %b = call i32 @llvm.bitreverse.i32(i32 %a0)
   %c = shl i32 %b, 7

 ; X86-NEXT:    orl %ecx, %eax
 ; X86-NEXT:    movl %eax, %ecx
 ; X86-NEXT:    andl $858993459, %ecx # imm = 0x33333333
 ; X86-NEXT:    shrl $2, %eax
 ; X86-NEXT:    andl $858993459, %eax # imm = 0x33333333
 ; X86-NEXT:    leal (%eax,%ecx,4), %eax
 ; X86-NEXT:    movl %eax, %ecx
 ; X86-NEXT:    andl $357913941, %ecx # imm = 0x15555555
-; X86-NEXT:    shrl %eax
-; X86-NEXT:    andl $1431655765, %eax # imm = 0x55555555
-; X86-NEXT:    leal (%eax,%ecx,2), %eax
-; X86-NEXT:    addl %eax, %eax
+; X86-NEXT:    andl $-1431655766, %eax # imm = 0xAAAAAAAA
+; X86-NEXT:    leal (%eax,%ecx,4), %eax
 ; X86-NEXT:    bswapl %eax
 ; X86-NEXT:    movl %eax, %ecx
 ; X86-NEXT:    andl $235867919, %ecx # imm = 0xE0F0F0F
 ; X86-NEXT:    shll $4, %ecx
 ; X86-NEXT:    shrl $4, %eax
 ; X86-NEXT:    andl $252645135, %eax # imm = 0xF0F0F0F
 ; X86-NEXT:    orl %ecx, %eax
 ; X86-NEXT:    movl %eax, %ecx

llvm/test/CodeGen/X86/is_fpclass.ll

 ; CHECK-32-NEXT:    flds {{[0-9]+}}(%esp)
 ; CHECK-32-NEXT:    flds {{[0-9]+}}(%esp)
 ; CHECK-32-NEXT:    flds {{[0-9]+}}(%esp)
 ; CHECK-32-NEXT:    fucomp %st(0)
 ; CHECK-32-NEXT:    fnstsw %ax
 ; CHECK-32-NEXT:    # kill: def $ah killed $ah killed $ax
 ; CHECK-32-NEXT:    sahf
 ; CHECK-32-NEXT:    setp %dh
+; CHECK-32-NEXT:    shlb $2, %dh
 ; CHECK-32-NEXT:    fucomp %st(0)
 ; CHECK-32-NEXT:    fnstsw %ax
 ; CHECK-32-NEXT:    # kill: def $ah killed $ah killed $ax
 ; CHECK-32-NEXT:    sahf
 ; CHECK-32-NEXT:    setp %dl
-; CHECK-32-NEXT:    addb %dl, %dl
+; CHECK-32-NEXT:    shlb $3, %dl
 ; CHECK-32-NEXT:    orb %dh, %dl
 ; CHECK-32-NEXT:    fucomp %st(0)
 ; CHECK-32-NEXT:    fnstsw %ax
 ; CHECK-32-NEXT:    # kill: def $ah killed $ah killed $ax
 ; CHECK-32-NEXT:    sahf
 ; CHECK-32-NEXT:    setp %dh
 ; CHECK-32-NEXT:    fucomp %st(0)
 ; CHECK-32-NEXT:    fnstsw %ax
 ; CHECK-32-NEXT:    # kill: def $ah killed $ah killed $ax
 ; CHECK-32-NEXT:    sahf
 ; CHECK-32-NEXT:    setp %al
 ; CHECK-32-NEXT:    addb %al, %al
 ; CHECK-32-NEXT:    orb %dh, %al
-; CHECK-32-NEXT:    shlb $2, %al
 ; CHECK-32-NEXT:    orb %dl, %al
 ; CHECK-32-NEXT:    movb %al, (%ecx)
 ; CHECK-32-NEXT:    movl %ecx, %eax
 ; CHECK-32-NEXT:    retl $4
 ;
 ; CHECK-64-LABEL: isnan_v4f:
 ; CHECK-64:       # %bb.0: # %entry
 ; CHECK-64-NEXT:    cmpunordps %xmm0, %xmm0

 ; CHECK-32-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-32-NEXT:    .cfi_offset %esi, -8
 ; CHECK-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; CHECK-32-NEXT:    movl $2147483647, %ecx # imm = 0x7FFFFFFF
 ; CHECK-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; CHECK-32-NEXT:    andl %ecx, %edx
 ; CHECK-32-NEXT:    cmpl $2139095041, %edx # imm = 0x7F800001
 ; CHECK-32-NEXT:    setge %dh
+; CHECK-32-NEXT:    shlb $2, %dh
 ; CHECK-32-NEXT:    movl {{[0-9]+}}(%esp), %esi
 ; CHECK-32-NEXT:    andl %ecx, %esi
 ; CHECK-32-NEXT:    cmpl $2139095041, %esi # imm = 0x7F800001
 ; CHECK-32-NEXT:    setge %dl
-; CHECK-32-NEXT:    addb %dl, %dl
+; CHECK-32-NEXT:    shlb $3, %dl
 ; CHECK-32-NEXT:    orb %dh, %dl
 ; CHECK-32-NEXT:    movl {{[0-9]+}}(%esp), %esi
 ; CHECK-32-NEXT:    andl %ecx, %esi
 ; CHECK-32-NEXT:    cmpl $2139095041, %esi # imm = 0x7F800001
 ; CHECK-32-NEXT:    setge %dh
 ; CHECK-32-NEXT:    andl {{[0-9]+}}(%esp), %ecx
 ; CHECK-32-NEXT:    cmpl $2139095041, %ecx # imm = 0x7F800001
 ; CHECK-32-NEXT:    setge %cl
 ; CHECK-32-NEXT:    addb %cl, %cl
 ; CHECK-32-NEXT:    orb %dh, %cl
-; CHECK-32-NEXT:    shlb $2, %cl
 ; CHECK-32-NEXT:    orb %dl, %cl
 ; CHECK-32-NEXT:    movb %cl, (%eax)
 ; CHECK-32-NEXT:    popl %esi
 ; CHECK-32-NEXT:    .cfi_def_cfa_offset 4
 ; CHECK-32-NEXT:    retl $4
 ;
 ; CHECK-64-LABEL: isnan_v4f_strictfp:
 ; CHECK-64:       # %bb.0: # %entry

llvm/test/CodeGen/X86/vector-sext.ll

 ; SSE2-NEXT:    movd %ecx, %xmm0
 ; SSE2-NEXT:    movq %rax, %rcx
 ; SSE2-NEXT:    shrq $17, %rcx
 ; SSE2-NEXT:    shll $15, %ecx
 ; SSE2-NEXT:    sarl $15, %ecx
 ; SSE2-NEXT:    movd %ecx, %xmm1
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE2-NEXT:    movl 8(%rdi), %ecx
-; SSE2-NEXT:    shll $13, %ecx
+; SSE2-NEXT:    shll $28, %ecx
 ; SSE2-NEXT:    movq %rax, %rdx
 ; SSE2-NEXT:    shrq $51, %rdx
-; SSE2-NEXT:    orl %ecx, %edx
 ; SSE2-NEXT:    shll $15, %edx
+; SSE2-NEXT:    orl %ecx, %edx
 ; SSE2-NEXT:    sarl $15, %edx
 ; SSE2-NEXT:    movd %edx, %xmm1
 ; SSE2-NEXT:    shrq $34, %rax
 ; SSE2-NEXT:    shll $15, %eax
 ; SSE2-NEXT:    sarl $15, %eax
 ; SSE2-NEXT:    movd %eax, %xmm2
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: sext_4i17_to_4i32:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    movq (%rdi), %rax
 ; SSSE3-NEXT:    movl %eax, %ecx
 ; SSSE3-NEXT:    shll $15, %ecx
 ; SSSE3-NEXT:    sarl $15, %ecx
 ; SSSE3-NEXT:    movd %ecx, %xmm0
 ; SSSE3-NEXT:    movq %rax, %rcx
 ; SSSE3-NEXT:    shrq $17, %rcx
 ; SSSE3-NEXT:    shll $15, %ecx
 ; SSSE3-NEXT:    sarl $15, %ecx
 ; SSSE3-NEXT:    movd %ecx, %xmm1
 ; SSSE3-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSSE3-NEXT:    movl 8(%rdi), %ecx
-; SSSE3-NEXT:    shll $13, %ecx
+; SSSE3-NEXT:    shll $28, %ecx
 ; SSSE3-NEXT:    movq %rax, %rdx
 ; SSSE3-NEXT:    shrq $51, %rdx
-; SSSE3-NEXT:    orl %ecx, %edx
 ; SSSE3-NEXT:    shll $15, %edx
+; SSSE3-NEXT:    orl %ecx, %edx
 ; SSSE3-NEXT:    sarl $15, %edx
 ; SSSE3-NEXT:    movd %edx, %xmm1
 ; SSSE3-NEXT:    shrq $34, %rax
 ; SSSE3-NEXT:    shll $15, %eax
 ; SSSE3-NEXT:    sarl $15, %eax
 ; SSSE3-NEXT:    movd %eax, %xmm2
 ; SSSE3-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]

 ; SSE41-NEXT:    movd %edx, %xmm0
 ; SSE41-NEXT:    pinsrd $1, %ecx, %xmm0
 ; SSE41-NEXT:    movq %rax, %rcx
 ; SSE41-NEXT:    shrq $34, %rcx
 ; SSE41-NEXT:    shll $15, %ecx
 ; SSE41-NEXT:    sarl $15, %ecx
 ; SSE41-NEXT:    pinsrd $2, %ecx, %xmm0
 ; SSE41-NEXT:    movl 8(%rdi), %ecx
-; SSE41-NEXT:    shll $13, %ecx
+; SSE41-NEXT:    shll $28, %ecx
 ; SSE41-NEXT:    shrq $51, %rax
-; SSE41-NEXT:    orl %ecx, %eax
 ; SSE41-NEXT:    shll $15, %eax
+; SSE41-NEXT:    orl %ecx, %eax
 ; SSE41-NEXT:    sarl $15, %eax
 ; SSE41-NEXT:    pinsrd $3, %eax, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: sext_4i17_to_4i32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    movq (%rdi), %rax
 ; AVX-NEXT:    movq %rax, %rcx
 ; AVX-NEXT:    shrq $17, %rcx
 ; AVX-NEXT:    shll $15, %ecx
 ; AVX-NEXT:    sarl $15, %ecx
 ; AVX-NEXT:    movl %eax, %edx
 ; AVX-NEXT:    shll $15, %edx
 ; AVX-NEXT:    sarl $15, %edx
 ; AVX-NEXT:    vmovd %edx, %xmm0
 ; AVX-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0
 ; AVX-NEXT:    movq %rax, %rcx
 ; AVX-NEXT:    shrq $34, %rcx
 ; AVX-NEXT:    shll $15, %ecx
 ; AVX-NEXT:    sarl $15, %ecx
 ; AVX-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0
 ; AVX-NEXT:    movl 8(%rdi), %ecx
-; AVX-NEXT:    shll $13, %ecx
+; AVX-NEXT:    shll $28, %ecx
 ; AVX-NEXT:    shrq $51, %rax
-; AVX-NEXT:    orl %ecx, %eax
 ; AVX-NEXT:    shll $15, %eax
+; AVX-NEXT:    orl %ecx, %eax
 ; AVX-NEXT:    sarl $15, %eax
 ; AVX-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0
 ; AVX-NEXT:    retq
 ;
 ; X86-SSE2-LABEL: sext_4i17_to_4i32:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-SSE2-NEXT:    movl (%edx), %ecx