diff --git a/llvm/lib/Transforms/Scalar/SeparateConstOffsetFromGEP.cpp b/llvm/lib/Transforms/Scalar/SeparateConstOffsetFromGEP.cpp
index 17c466f38c9c33..842a593189b6e6 100644
--- a/llvm/lib/Transforms/Scalar/SeparateConstOffsetFromGEP.cpp
+++ b/llvm/lib/Transforms/Scalar/SeparateConstOffsetFromGEP.cpp
@@ -393,6 +393,11 @@ class SeparateConstOffsetFromGEP {
   /// and returns true if the splitting succeeds.
   bool splitGEP(GetElementPtrInst *GEP);
 
+  /// Tries to reorder the given GEP with the GEP that produces the base if
+  /// doing so results in producing a constant offset as the outermost
+  /// index.
+  bool reorderGEP(GetElementPtrInst *GEP, TargetTransformInfo &TTI);
+
   /// Lower a GEP with multiple indices into multiple GEPs with a single index.
   /// Function splitGEP already split the original GEP into a variadic part and
   /// a constant offset (i.e., AccumulativeByteOffset). This function lowers the
@@ -970,6 +975,66 @@ SeparateConstOffsetFromGEP::lowerToArithmetics(GetElementPtrInst *Variadic,
   Variadic->eraseFromParent();
 }
 
+bool SeparateConstOffsetFromGEP::reorderGEP(GetElementPtrInst *GEP,
+                                            TargetTransformInfo &TTI) {
+  Type *GEPType = GEP->getResultElementType();
+  // TODO: support reordering for non-trivial GEP chains
+  if (GEPType->isAggregateType() || GEP->getNumIndices() != 1)
+    return false;
+
+  auto PtrGEP = dyn_cast<GetElementPtrInst>(GEP->getPointerOperand());
+  if (!PtrGEP)
+    return false;
+  Type *PtrGEPType = PtrGEP->getResultElementType();
+  // TODO: support reordering for non-trivial GEP chains
+  if (PtrGEPType->isAggregateType() || PtrGEP->getNumIndices() != 1)
+    return false;
+
+  // TODO: support reordering for non-trivial GEP chains
+  if (PtrGEPType != GEPType ||
+      PtrGEP->getSourceElementType() != GEP->getSourceElementType())
+    return false;
+
+  bool NestedNeedsExtraction;
+  int64_t NestedByteOffset =
+      accumulateByteOffset(PtrGEP, NestedNeedsExtraction);
+  if (!NestedNeedsExtraction)
+    return false;
+
+  unsigned AddrSpace = PtrGEP->getPointerAddressSpace();
+  if (!TTI.isLegalAddressingMode(GEP->getResultElementType(),
+                                 /*BaseGV=*/nullptr, NestedByteOffset,
+                                 /*HasBaseReg=*/true, /*Scale=*/0, AddrSpace))
+    return false;
+
+  IRBuilder<> Builder(GEP);
+  Builder.SetCurrentDebugLocation(GEP->getDebugLoc());
+  bool GEPInBounds = GEP->isInBounds();
+  bool PtrGEPInBounds = PtrGEP->isInBounds();
+  bool IsChainInBounds = GEPInBounds && PtrGEPInBounds;
+  if (IsChainInBounds) {
+    auto GEPIdx = GEP->indices().begin();
+    auto KnownGEPIdx = computeKnownBits(GEPIdx->get(), *DL);
+    IsChainInBounds &= KnownGEPIdx.isNonNegative();
+    if (IsChainInBounds) {
+      auto PtrGEPIdx = GEP->indices().begin();
+      auto KnownPtrGEPIdx = computeKnownBits(PtrGEPIdx->get(), *DL);
+      IsChainInBounds &= KnownPtrGEPIdx.isNonNegative();
+    }
+  }
+
+  // For trivial GEP chains, we can swap the indicies.
+  auto NewSrc = Builder.CreateGEP(PtrGEPType, PtrGEP->getPointerOperand(),
+                                  SmallVector<Value *, 4>(GEP->indices()));
+  cast<GetElementPtrInst>(NewSrc)->setIsInBounds(IsChainInBounds);
+  auto NewGEP = Builder.CreateGEP(GEPType, NewSrc,
+                                  SmallVector<Value *, 4>(PtrGEP->indices()));
+  cast<GetElementPtrInst>(NewGEP)->setIsInBounds(IsChainInBounds);
+  GEP->replaceAllUsesWith(NewGEP);
+  RecursivelyDeleteTriviallyDeadInstructions(GEP);
+  return true;
+}
+
 bool SeparateConstOffsetFromGEP::splitGEP(GetElementPtrInst *GEP) {
   // Skip vector GEPs.
   if (GEP->getType()->isVectorTy())
@@ -985,11 +1050,13 @@ bool SeparateConstOffsetFromGEP::splitGEP(GetElementPtrInst *GEP) {
   bool NeedsExtraction;
   int64_t AccumulativeByteOffset = accumulateByteOffset(GEP, NeedsExtraction);
 
-  if (!NeedsExtraction)
-    return Changed;
-
   TargetTransformInfo &TTI = GetTTI(*GEP->getFunction());
 
+  if (!NeedsExtraction) {
+    Changed |= reorderGEP(GEP, TTI);
+    return Changed;
+  }
+
   // If LowerGEP is disabled, before really splitting the GEP, check whether the
   // backend supports the addressing mode we are about to produce. If no, this
   // splitting probably won't be beneficial.
diff --git a/llvm/test/CodeGen/AMDGPU/machine-sink-temporal-divergence-swdev407790.ll b/llvm/test/CodeGen/AMDGPU/machine-sink-temporal-divergence-swdev407790.ll
index 4a36b3228ebe3b..d08513d0a250f2 100644
--- a/llvm/test/CodeGen/AMDGPU/machine-sink-temporal-divergence-swdev407790.ll
+++ b/llvm/test/CodeGen/AMDGPU/machine-sink-temporal-divergence-swdev407790.ll
@@ -143,11 +143,11 @@ define protected amdgpu_kernel void @kernel_round1(ptr addrspace(1) nocapture no
 ; CHECK-NEXT:    v_mov_b32_e32 v47, 0
 ; CHECK-NEXT:    s_mov_b32 s49, 0
 ; CHECK-NEXT:    s_branch .LBB0_7
-; CHECK-NEXT:  .LBB0_5: ; %Flow41
+; CHECK-NEXT:  .LBB0_5: ; %Flow43
 ; CHECK-NEXT:    ; in Loop: Header=BB0_7 Depth=1
 ; CHECK-NEXT:    s_inst_prefetch 0x2
 ; CHECK-NEXT:    s_or_b32 exec_lo, exec_lo, s49
-; CHECK-NEXT:  .LBB0_6: ; %Flow42
+; CHECK-NEXT:  .LBB0_6: ; %Flow44
 ; CHECK-NEXT:    ; in Loop: Header=BB0_7 Depth=1
 ; CHECK-NEXT:    v_cmp_ge_u32_e32 vcc_lo, s48, v45
 ; CHECK-NEXT:    v_cmp_lt_u32_e64 s4, 59, v47
@@ -304,7 +304,7 @@ define protected amdgpu_kernel void @kernel_round1(ptr addrspace(1) nocapture no
 ; CHECK-NEXT:    ds_write_b32 v0, v58
 ; CHECK-NEXT:    s_or_b32 exec_lo, exec_lo, s55
 ; CHECK-NEXT:    s_branch .LBB0_9
-; CHECK-NEXT:  .LBB0_18: ; %Flow43
+; CHECK-NEXT:  .LBB0_18: ; %Flow45
 ; CHECK-NEXT:    ; in Loop: Header=BB0_7 Depth=1
 ; CHECK-NEXT:    v_mov_b32_e32 v57, v0
 ; CHECK-NEXT:    s_or_b32 exec_lo, exec_lo, s52
@@ -357,7 +357,7 @@ define protected amdgpu_kernel void @kernel_round1(ptr addrspace(1) nocapture no
 ; CHECK-NEXT:    ds_write_b32 v0, v57
 ; CHECK-NEXT:    s_or_b32 exec_lo, exec_lo, s53
 ; CHECK-NEXT:    s_branch .LBB0_21
-; CHECK-NEXT:  .LBB0_24: ; %Flow47
+; CHECK-NEXT:  .LBB0_24: ; %Flow49
 ; CHECK-NEXT:    s_or_b32 exec_lo, exec_lo, s42
 ; CHECK-NEXT:  .LBB0_25:
 ; CHECK-NEXT:    v_mov_b32_e32 v31, v40
@@ -382,13 +382,11 @@ define protected amdgpu_kernel void @kernel_round1(ptr addrspace(1) nocapture no
 ; CHECK-NEXT:    s_cmov_b32 exec_lo, vcc_lo
 ; CHECK-NEXT:    s_cbranch_scc0 .LBB0_34
 ; CHECK-NEXT:  ; %bb.26:
-; CHECK-NEXT:    s_add_u32 s42, s44, 8
-; CHECK-NEXT:    s_addc_u32 s43, s45, 0
-; CHECK-NEXT:    s_mov_b32 s44, 0
+; CHECK-NEXT:    s_mov_b32 s42, 0
 ; CHECK-NEXT:    s_branch .LBB0_29
-; CHECK-NEXT:  .LBB0_27: ; %Flow38
+; CHECK-NEXT:  .LBB0_27: ; %Flow40
 ; CHECK-NEXT:    ; in Loop: Header=BB0_29 Depth=1
-; CHECK-NEXT:    s_or_b32 exec_lo, exec_lo, s45
+; CHECK-NEXT:    s_or_b32 exec_lo, exec_lo, s43
 ; CHECK-NEXT:  .LBB0_28: ; in Loop: Header=BB0_29 Depth=1
 ; CHECK-NEXT:    v_mov_b32_e32 v31, v40
 ; CHECK-NEXT:    v_mov_b32_e32 v0, 0
@@ -405,13 +403,13 @@ define protected amdgpu_kernel void @kernel_round1(ptr addrspace(1) nocapture no
 ; CHECK-NEXT:    s_swappc_b64 s[30:31], s[6:7]
 ; CHECK-NEXT:    v_add_co_u32 v41, vcc_lo, v0, v41
 ; CHECK-NEXT:    v_cmp_le_u32_e32 vcc_lo, v47, v41
-; CHECK-NEXT:    s_or_b32 s44, vcc_lo, s44
-; CHECK-NEXT:    s_andn2_b32 s4, exec_lo, s44
-; CHECK-NEXT:    s_cselect_b32 exec_lo, s4, s44
+; CHECK-NEXT:    s_or_b32 s42, vcc_lo, s42
+; CHECK-NEXT:    s_andn2_b32 s4, exec_lo, s42
+; CHECK-NEXT:    s_cselect_b32 exec_lo, s4, s42
 ; CHECK-NEXT:    s_cbranch_scc0 .LBB0_34
 ; CHECK-NEXT:  .LBB0_29: ; =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    v_lshlrev_b32_e32 v0, 2, v41
-; CHECK-NEXT:    s_mov_b32 s45, exec_lo
+; CHECK-NEXT:    s_mov_b32 s43, exec_lo
 ; CHECK-NEXT:    ds_read_b32 v0, v0
 ; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
 ; CHECK-NEXT:    v_lshrrev_b32_e32 v63, 10, v0
@@ -420,15 +418,15 @@ define protected amdgpu_kernel void @kernel_round1(ptr addrspace(1) nocapture no
 ; CHECK-NEXT:    v_mul_u32_u24_e32 v1, 0x180, v63
 ; CHECK-NEXT:    v_lshlrev_b32_e32 v0, 5, v62
 ; CHECK-NEXT:    v_lshlrev_b32_e32 v4, 5, v72
-; CHECK-NEXT:    v_add_co_u32 v2, s4, s42, v1
-; CHECK-NEXT:    v_add_co_ci_u32_e64 v3, null, s43, 0, s4
+; CHECK-NEXT:    v_add_co_u32 v2, s4, s44, v1
+; CHECK-NEXT:    v_add_co_ci_u32_e64 v3, null, s45, 0, s4
 ; CHECK-NEXT:    v_add_co_u32 v0, vcc_lo, v2, v0
 ; CHECK-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v3, vcc_lo
 ; CHECK-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
 ; CHECK-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
 ; CHECK-NEXT:    s_clause 0x1
-; CHECK-NEXT:    global_load_dwordx4 v[4:7], v[0:1], off
-; CHECK-NEXT:    global_load_dwordx4 v[8:11], v[2:3], off
+; CHECK-NEXT:    global_load_dwordx4 v[4:7], v[0:1], off offset:8
+; CHECK-NEXT:    global_load_dwordx4 v[8:11], v[2:3], off offset:8
 ; CHECK-NEXT:    s_waitcnt vmcnt(0)
 ; CHECK-NEXT:    v_xor_b32_e32 v46, v9, v5
 ; CHECK-NEXT:    v_xor_b32_e32 v45, v8, v4
@@ -442,8 +440,8 @@ define protected amdgpu_kernel void @kernel_round1(ptr addrspace(1) nocapture no
 ; CHECK-NEXT:    s_cbranch_scc0 .LBB0_28
 ; CHECK-NEXT:  ; %bb.30: ; in Loop: Header=BB0_29 Depth=1
 ; CHECK-NEXT:    s_clause 0x1
-; CHECK-NEXT:    global_load_dwordx2 v[58:59], v[2:3], off offset:16
-; CHECK-NEXT:    global_load_dwordx2 v[60:61], v[0:1], off offset:16
+; CHECK-NEXT:    global_load_dwordx2 v[58:59], v[2:3], off offset:24
+; CHECK-NEXT:    global_load_dwordx2 v[60:61], v[0:1], off offset:24
 ; CHECK-NEXT:    v_lshlrev_b32_e32 v0, 4, v45
 ; CHECK-NEXT:    v_alignbit_b32 v1, v46, v45, 12
 ; CHECK-NEXT:    v_and_b32_e32 v2, 0xf0000, v45
diff --git a/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll b/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll
index 8c806e76bde6ec..b87439a9d6fae7 100644
--- a/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll
+++ b/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll
@@ -31,205 +31,188 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x
   ; CHECK-NEXT:   [[COPY13:%[0-9]+]]:sgpr_32 = COPY $sgpr10
   ; CHECK-NEXT:   [[COPY14:%[0-9]+]]:sgpr_32 = COPY $sgpr8
   ; CHECK-NEXT:   undef [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub0_sub1:sgpr_128 = S_LOAD_DWORDX2_IMM [[COPY]], 232, 0 :: (invariant load (s64) from %ir.39, addrspace 4)
+  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %125:sgpr_128, 0, 0 :: (dereferenceable invariant load (s32))
+  ; CHECK-NEXT:   KILL undef %125:sgpr_128
   ; CHECK-NEXT:   [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY5]], 4, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_LSHL_B32_1:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY4]], 4, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_LSHL_B32_2:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY3]], 4, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ASHR_I32_:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_]], 31, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ASHR_I32_1:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_1]], 31, implicit-def dead $scc
-  ; CHECK-NEXT:   [[S_ASHR_I32_2:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_2]], 31, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub1:sgpr_128 = S_AND_B32 [[S_LOAD_DWORDX2_IMM]].sub1, 65535, implicit-def dead $scc
+  ; CHECK-NEXT:   [[S_ASHR_I32_2:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_2]], 31, implicit-def dead $scc
+  ; CHECK-NEXT:   [[S_SUB_I32_:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM]], 29, implicit-def dead $scc
+  ; CHECK-NEXT:   [[S_SUB_I32_1:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM]], 30, implicit-def dead $scc
   ; CHECK-NEXT:   undef [[S_ADD_U32_:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_2]], implicit-def $scc
   ; CHECK-NEXT:   [[S_ADD_U32_:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
   ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_]], 16, 0 :: (invariant load (s128) from %ir.81, addrspace 4)
   ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM1:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM undef %74:sreg_64, 0, 0 :: (invariant load (s128) from `ptr addrspace(4) undef`, addrspace 4)
-  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %132:sgpr_128, 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK-NEXT:   KILL undef %74:sreg_64
-  ; CHECK-NEXT:   KILL undef %132:sgpr_128
   ; CHECK-NEXT:   KILL [[S_ADD_U32_]].sub0, [[S_ADD_U32_]].sub1
   ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[S_LOAD_DWORDX4_IMM]], 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK-NEXT:   [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
   ; CHECK-NEXT:   undef [[S_MOV_B32_:%[0-9]+]].sub1:sgpr_128 = S_MOV_B32 0
+  ; CHECK-NEXT:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET undef %118:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], undef %89:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   KILL undef %89:sgpr_128
-  ; CHECK-NEXT:   [[S_SUB_I32_:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM]], 29, implicit-def dead $scc
-  ; CHECK-NEXT:   [[S_SUB_I32_1:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM]], 30, implicit-def dead $scc
+  ; CHECK-NEXT:   KILL undef %118:sgpr_128
   ; CHECK-NEXT:   [[S_SUB_I32_2:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM1]], 31, implicit-def dead $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_1:%[0-9]+]]:sreg_32 = S_ADD_U32 [[COPY6]], 64, implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADDC_U32_:%[0-9]+]]:sreg_32 = S_ADDC_U32 undef %54:sreg_32, 0, implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_2:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_1]], [[S_LSHL_B32_]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_2:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_]], [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_3:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_1]], [[S_LSHL_B32_1]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM2:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_2]], 0, 0 :: (invariant load (s128) from %ir.87, addrspace 4)
-  ; CHECK-NEXT:   [[S_ADD_U32_3:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_]], [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_4:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_1]], [[S_LSHL_B32_2]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_4:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_]], [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_ASHR_I32_3:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 undef %171:sreg_32, 31, implicit-def dead $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_5:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_1]], undef %171:sreg_32, implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_5:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_]], [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_6:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_6:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_7:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_1]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_7:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_8:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, undef %171:sreg_32, implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_8:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_9:%[0-9]+]]:sreg_32 = S_ADD_U32 [[COPY7]].sub0, 224, implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADDC_U32_1:%[0-9]+]]:sreg_32 = S_ADDC_U32 undef %51:sreg_32, 0, implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_10:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_9]], [[S_LSHL_B32_]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_10:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_1]], [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_11:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_9]], [[S_LSHL_B32_1]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_11:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_1]], [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_12:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_9]], [[S_LSHL_B32_2]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_12:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_1]], [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_13:%[0-9]+]]:sreg_32 = S_ADD_U32 [[COPY7]].sub0, 576, implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADDC_U32_2:%[0-9]+]]:sreg_32 = S_ADDC_U32 undef %51:sreg_32, 0, implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_14:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_13]], [[S_LSHL_B32_]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_14:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_2]], [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_15:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_13]], [[S_LSHL_B32_2]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_15:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_2]], [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_16:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_13]], undef %171:sreg_32, implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_16:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_2]], [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_17:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY8]], [[S_LSHL_B32_]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_17:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %48:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_18:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_1]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_18:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_19:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_2]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_19:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_1:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_1:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_2:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_1]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_2:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_3:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_2]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM2:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_1]], 64, 0 :: (invariant load (s128) from %ir.87, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM3:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_2]], 64, 0 :: (invariant load (s128) from %ir.93, addrspace 4)
+  ; CHECK-NEXT:   KILL [[S_ADD_U32_1]].sub0, [[S_ADD_U32_1]].sub1
+  ; CHECK-NEXT:   KILL [[S_ADD_U32_2]].sub0, [[S_ADD_U32_2]].sub1
+  ; CHECK-NEXT:   [[S_ADD_U32_3:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   [[S_ASHR_I32_3:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 undef %169:sreg_32, 31, implicit-def dead $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_4:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], undef %169:sreg_32, implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_4:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_5:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_5:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_6:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_1]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_6:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_7:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, undef %169:sreg_32, implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_7:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_8:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_2]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_8:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_9:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY8]], [[S_LSHL_B32_]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_9:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %48:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_10:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_1]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_10:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_11:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_2]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_11:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
   ; CHECK-NEXT:   [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_]], 16, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_1:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_2]], 16, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32))
-  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], undef %312:sreg_32, 0, 0 :: (dereferenceable invariant load (s32))
+  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], undef %301:sreg_32, 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[S_MOV_B32_]], 16, 0 :: (dereferenceable invariant load (s32))
-  ; CHECK-NEXT:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET undef %118:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %367:sgpr_128, undef %368:sreg_32, 0, 0 :: (dereferenceable invariant load (s32))
-  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %378:sgpr_128, 16, 0 :: (dereferenceable invariant load (s32))
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM3:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_3]], 0, 0 :: (invariant load (s128) from %ir.92, addrspace 4)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM4:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 0, 0 :: (invariant load (s128) from %ir.97, addrspace 4)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM5:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 0, 0 :: (invariant load (s128) from %ir.104, addrspace 4)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM6:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 0, 0 :: (invariant load (s128) from %ir.109, addrspace 4)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM7:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 0, 0 :: (invariant load (s128) from %ir.114, addrspace 4)
+  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %356:sgpr_128, undef %357:sreg_32, 0, 0 :: (dereferenceable invariant load (s32))
+  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %367:sgpr_128, 16, 0 :: (dereferenceable invariant load (s32))
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM4:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_3]], 64, 0 :: (invariant load (s128) from %ir.99, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM5:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 64, 0 :: (invariant load (s128) from %ir.107, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM6:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 0, 0 :: (invariant load (s128) from %ir.117, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM7:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 0, 0 :: (invariant load (s128) from %ir.124, addrspace 4)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM2]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %362:sgpr_128, [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32))
-  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM5:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %373:sgpr_128, [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32))
+  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %351:sgpr_128, [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32))
+  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM5:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %362:sgpr_128, [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32))
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM3]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[S_ADD_I32_2:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM]], -98, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_3:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM1]], -114, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_4:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM2]], -130, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_5:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM2]], -178, implicit-def dead $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_20:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY10]], [[S_LSHL_B32_]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_20:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %42:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_21:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_21:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_22:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_1]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM8:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 0, 0 :: (invariant load (s128) from %ir.121, addrspace 4)
-  ; CHECK-NEXT:   [[S_ADD_U32_22:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_23:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_2]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_23:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM8:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 0, 0 :: (invariant load (s128) from %ir.112, addrspace 4)
+  ; CHECK-NEXT:   undef [[S_ADD_U32_12:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY10]], [[S_LSHL_B32_]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_12:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %42:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_13:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_13:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_14:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_1]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_14:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_15:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_2]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_15:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
   ; CHECK-NEXT:   [[S_LSHL_B32_3:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY12]], 4, implicit-def dead $scc
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM3]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[S_ADD_I32_6:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_3]], 16, implicit-def dead $scc
-  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %394:sgpr_128, [[S_ADD_I32_6]], 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN4:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM9:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[COPY7]], 224, 0 :: (invariant load (s128) from %ir.126, addrspace 4)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM10:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_10]], 0, 0 :: (invariant load (s128) from %ir.127, addrspace 4)
+  ; CHECK-NEXT:   [[S_ADD_I32_6:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_3]], 16, implicit-def dead $scc
+  ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_SGPR_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %383:sgpr_128, [[S_ADD_I32_6]], 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN5:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM5]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM11:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_11]], 0, 0 :: (invariant load (s128) from %ir.132, addrspace 4)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM12:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_12]], 0, 0 :: (invariant load (s128) from %ir.137, addrspace 4)
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN6:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM6]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN7:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN8:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM8]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM9:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 224, 0 :: (invariant load (s128) from %ir.129, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM10:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[COPY7]], 224, 0 :: (invariant load (s128) from %ir.145, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM11:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 576, 0 :: (invariant load (s128) from %ir.150, addrspace 4)
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN6:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM8]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM12:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 224, 0 :: (invariant load (s128) from %ir.134, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM13:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 576, 0 :: (invariant load (s128) from %ir.162, addrspace 4)
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN7:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM6]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN8:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[S_ADD_I32_7:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM4]], -217, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_8:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -233, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_9:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM5]], -249, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_10:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM3]], -297, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_11:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -313, implicit-def dead $scc
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM14:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 224, 0 :: (invariant load (s128) from %ir.140, addrspace 4)
   ; CHECK-NEXT:   [[S_ADD_I32_12:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -329, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_13:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -345, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_14:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM6]], -441, implicit-def dead $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_24:%[0-9]+]]:sreg_32 = S_ADD_U32 [[COPY2]], 160, implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADDC_U32_3:%[0-9]+]]:sreg_32 = S_ADDC_U32 undef %36:sreg_32, 0, implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_25:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_24]], [[S_LSHL_B32_2]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_25:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_3]], [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   undef [[S_ADD_U32_16:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_2]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_16:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
   ; CHECK-NEXT:   [[S_LSHL_B32_4:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY13]], 4, implicit-def dead $scc
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN9:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM10]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN9:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[S_ASHR_I32_4:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_4]], 31, implicit-def dead $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_26:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_24]], [[S_LSHL_B32_4]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_26:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_3]], [[S_ASHR_I32_4]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_27:%[0-9]+]]:sreg_32 = S_ADD_U32 [[COPY]].sub0, 168, implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADDC_U32_4:%[0-9]+]]:sreg_32 = S_ADDC_U32 undef %57:sreg_32, 0, implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM13:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_14]], 0, 0 :: (invariant load (s128) from %ir.147, addrspace 4)
+  ; CHECK-NEXT:   undef [[S_ADD_U32_17:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_4]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_17:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_4]], implicit-def dead $scc, implicit $scc
   ; CHECK-NEXT:   [[S_LSHL_B32_5:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY5]], 3, implicit-def dead $scc
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN10:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM11]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN10:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM12]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[S_ASHR_I32_5:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_5]], 31, implicit-def dead $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_28:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_27]], [[S_LSHL_B32_5]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_28:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_4]], [[S_ASHR_I32_5]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_28]], 0, 0 :: (invariant load (s32) from %ir.269, align 8, addrspace 4)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM14:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_15]], 0, 0 :: (invariant load (s128) from %ir.154, addrspace 4)
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN11:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM12]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM15:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_16]], 0, 0 :: (invariant load (s128) from %ir.159, addrspace 4)
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN12:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN13:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM13]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   undef [[S_ADD_U32_18:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_5]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_18:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_5]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   [[S_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_18]], 168, 0 :: (invariant load (s32) from %ir.273, align 8, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM15:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 576, 0 :: (invariant load (s128) from %ir.157, addrspace 4)
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN11:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM14]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN12:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM10]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN13:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM11]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub3:sgpr_128 = S_MOV_B32 553734060
   ; CHECK-NEXT:   [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub2:sgpr_128 = S_MOV_B32 -1
   ; CHECK-NEXT:   [[COPY15:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]]
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM16:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_17]], 0, 0 :: (invariant load (s128) from %ir.167, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM16:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_9]], 0, 0 :: (invariant load (s128) from %ir.170, addrspace 4)
   ; CHECK-NEXT:   [[COPY15:%[0-9]+]].sub1:sgpr_128 = COPY [[S_MOV_B32_]].sub1
   ; CHECK-NEXT:   [[COPY15:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORD_IMM]]
   ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[COPY15]], 0, 0 :: (dereferenceable invariant load (s32))
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN14:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM14]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN15:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM15]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM17:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_18]], 0, 0 :: (invariant load (s128) from %ir.175, addrspace 4)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM18:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_19]], 0, 0 :: (invariant load (s128) from %ir.180, addrspace 4)
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN14:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM15]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN15:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM13]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM17:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_10]], 0, 0 :: (invariant load (s128) from %ir.178, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM18:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_11]], 0, 0 :: (invariant load (s128) from %ir.183, addrspace 4)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN16:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM16]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[S_LSHL_B32_6:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY4]], 3, implicit-def dead $scc
   ; CHECK-NEXT:   [[BUFFER_LOAD_DWORD_OFFSET1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[S_ASHR_I32_6:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_6]], 31, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_15:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM4]], -467, implicit-def dead $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_29:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_27]], [[S_LSHL_B32_6]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_29:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_4]], [[S_ASHR_I32_6]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX2_IMM1:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_29]], 0, 0 :: (invariant load (s64) from %ir.277, addrspace 4)
+  ; CHECK-NEXT:   undef [[S_ADD_U32_19:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_6]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_19:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_6]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX2_IMM1:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_19]], 168, 0 :: (invariant load (s64) from %ir.282, addrspace 4)
   ; CHECK-NEXT:   [[BUFFER_LOAD_DWORD_OFFSET2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM17]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[BUFFER_LOAD_DWORD_OFFSET3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM18]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM19:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_20]], 0, 0 :: (invariant load (s128) from %ir.202, addrspace 4)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM20:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_21]], 0, 0 :: (invariant load (s128) from %ir.208, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM19:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_12]], 0, 0 :: (invariant load (s128) from %ir.205, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM20:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_13]], 0, 0 :: (invariant load (s128) from %ir.211, addrspace 4)
   ; CHECK-NEXT:   [[COPY16:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]]
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM21:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_22]], 0, 0 :: (invariant load (s128) from %ir.213, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM21:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_14]], 0, 0 :: (invariant load (s128) from %ir.216, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM22:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_15]], 0, 0 :: (invariant load (s128) from %ir.221, addrspace 4)
   ; CHECK-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORDX2_IMM1]].sub1, 65535, implicit-def dead $scc
   ; CHECK-NEXT:   [[COPY16:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM1]].sub0
   ; CHECK-NEXT:   [[COPY16:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_]]
   ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_IMM5:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[COPY16]], 0, 0 :: (dereferenceable invariant load (s32))
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM22:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_23]], 0, 0 :: (invariant load (s128) from %ir.218, addrspace 4)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN17:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM19]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN18:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM20]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[S_LSHL_B32_7:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY3]], 3, implicit-def dead $scc
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN19:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM21]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
+  ; CHECK-NEXT:   [[S_LSHL_B32_7:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY3]], 3, implicit-def dead $scc
+  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN20:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM22]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[S_ASHR_I32_7:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_7]], 31, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_16:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM5]], -468, implicit-def dead $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_30:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_27]], [[S_LSHL_B32_7]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_30:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_4]], [[S_ASHR_I32_7]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN20:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM22]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX2_IMM2:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_30]], 0, 0 :: (invariant load (s64) from %ir.287, addrspace 4)
+  ; CHECK-NEXT:   undef [[S_ADD_U32_20:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_7]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_20:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_7]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX2_IMM2:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_20]], 168, 0 :: (invariant load (s64) from %ir.293, addrspace 4)
   ; CHECK-NEXT:   [[COPY17:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]]
   ; CHECK-NEXT:   [[S_AND_B32_1:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORDX2_IMM2]].sub1, 65535, implicit-def dead $scc
   ; CHECK-NEXT:   [[COPY17:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM2]].sub0
   ; CHECK-NEXT:   [[COPY17:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_1]]
   ; CHECK-NEXT:   [[S_BUFFER_LOAD_DWORD_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[COPY17]], 0, 0 :: (dereferenceable invariant load (s32))
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM23:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_25]], 0, 0 :: (invariant load (s128) from %ir.253, addrspace 4)
-  ; CHECK-NEXT:   [[S_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM undef %484:sreg_64, 0, 0 :: (invariant load (s32) from `ptr addrspace(4) undef`, addrspace 4)
-  ; CHECK-NEXT:   KILL [[S_ADD_U32_25]].sub0, [[S_ADD_U32_25]].sub1
-  ; CHECK-NEXT:   KILL undef %484:sreg_64
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM23:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_16]], 160, 0 :: (invariant load (s128) from %ir.256, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM undef %469:sreg_64, 0, 0 :: (invariant load (s32) from `ptr addrspace(4) undef`, addrspace 4)
+  ; CHECK-NEXT:   KILL [[S_ADD_U32_16]].sub0, [[S_ADD_U32_16]].sub1
+  ; CHECK-NEXT:   KILL undef %469:sreg_64
   ; CHECK-NEXT:   KILL [[COPY17]].sub0_sub1_sub2, [[COPY17]].sub3
   ; CHECK-NEXT:   [[S_LSHL_B32_8:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY14]], 3, implicit-def dead $scc
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM24:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_26]], 0, 0 :: (invariant load (s128) from %ir.261, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM24:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_17]], 160, 0 :: (invariant load (s128) from %ir.265, addrspace 4)
   ; CHECK-NEXT:   [[S_ASHR_I32_8:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_8]], 31, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_17:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM6]], -469, implicit-def dead $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_31:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_27]], [[S_LSHL_B32_8]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_31:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_4]], [[S_ASHR_I32_8]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_31]], 0, 0 :: (invariant load (s32) from %ir.298, align 8, addrspace 4)
+  ; CHECK-NEXT:   undef [[S_ADD_U32_21:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_8]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_21:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_8]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   [[S_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_21]], 168, 0 :: (invariant load (s32) from %ir.305, align 8, addrspace 4)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN21:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM23]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN22:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM24]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   KILL [[S_LOAD_DWORDX4_IMM24]]
   ; CHECK-NEXT:   KILL [[S_LOAD_DWORDX4_IMM23]]
+  ; CHECK-NEXT:   KILL [[S_LOAD_DWORDX4_IMM24]]
   ; CHECK-NEXT:   [[S_AND_B32_2:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORD_IMM1]], 65535, implicit-def dead $scc
   ; CHECK-NEXT:   [[COPY18:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]]
   ; CHECK-NEXT:   [[COPY18:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_2]]
@@ -241,24 +224,22 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x
   ; CHECK-NEXT:   [[S_ADD_I32_21:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -507, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_22:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -539, implicit-def dead $scc
   ; CHECK-NEXT:   [[S_ADD_I32_23:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM7]], -473, implicit-def dead $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_32:%[0-9]+]]:sreg_32 = S_ADD_U32 [[COPY1]], 96, implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADDC_U32_5:%[0-9]+]]:sreg_32 = S_ADDC_U32 undef %33:sreg_32, 0, implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   undef [[S_ADD_U32_33:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_32]], [[S_LSHL_B32_]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_33:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_5]], [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM25:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_33]], 0, 0 :: (invariant load (s128) from %ir.316, addrspace 4)
-  ; CHECK-NEXT:   undef [[S_ADD_U32_34:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_32]], [[S_LSHL_B32_1]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_34:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_5]], [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM26:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_34]], 0, 0 :: (invariant load (s128) from %ir.321, addrspace 4)
-  ; CHECK-NEXT:   undef [[S_ADD_U32_35:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[S_ADD_U32_32]], [[S_LSHL_B32_2]], implicit-def $scc
-  ; CHECK-NEXT:   [[S_ADD_U32_35:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 [[S_ADDC_U32_5]], [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM27:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_35]], 0, 0 :: (invariant load (s128) from %ir.326, addrspace 4)
+  ; CHECK-NEXT:   undef [[S_ADD_U32_22:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_22:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM25:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_22]], 96, 0 :: (invariant load (s128) from %ir.323, addrspace 4)
+  ; CHECK-NEXT:   undef [[S_ADD_U32_23:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_1]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_23:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM26:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_23]], 96, 0 :: (invariant load (s128) from %ir.329, addrspace 4)
+  ; CHECK-NEXT:   undef [[S_ADD_U32_24:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_2]], implicit-def $scc
+  ; CHECK-NEXT:   [[S_ADD_U32_24:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX4_IMM27:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_24]], 96, 0 :: (invariant load (s128) from %ir.335, addrspace 4)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN23:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM25]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN24:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM26]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
   ; CHECK-NEXT:   [[BUFFER_LOAD_FORMAT_X_IDXEN25:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM27]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
-  ; CHECK-NEXT:   KILL [[S_LOAD_DWORDX4_IMM25]]
-  ; CHECK-NEXT:   KILL [[V_MOV_B32_e32_]]
   ; CHECK-NEXT:   KILL [[S_LOAD_DWORDX4_IMM26]]
+  ; CHECK-NEXT:   KILL [[V_MOV_B32_e32_]]
   ; CHECK-NEXT:   KILL [[S_LOAD_DWORDX4_IMM27]]
+  ; CHECK-NEXT:   KILL [[S_LOAD_DWORDX4_IMM25]]
   ; CHECK-NEXT:   [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -2, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec
   ; CHECK-NEXT:   [[V_ADD_U32_e64_1:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -1, [[BUFFER_LOAD_FORMAT_X_IDXEN1]], 0, implicit $exec
   ; CHECK-NEXT:   [[V_ADD_U32_e64_2:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -3, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec
@@ -370,13 +351,13 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x
   ; CHECK-NEXT:   [[V_OR_B32_e64_64:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_63]], [[V_ADD_U32_e64_28]], implicit $exec
   ; CHECK-NEXT:   [[V_ADD_U32_e64_30:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -593, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec
   ; CHECK-NEXT:   [[V_OR_B32_e64_65:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_64]], [[V_ADD_U32_e64_29]], implicit $exec
-  ; CHECK-NEXT:   [[S_LOAD_DWORDX8_IMM:%[0-9]+]]:sgpr_256 = S_LOAD_DWORDX8_IMM undef %559:sreg_64, 0, 0 :: (invariant load (s256) from `ptr addrspace(4) undef`, addrspace 4)
+  ; CHECK-NEXT:   [[S_LOAD_DWORDX8_IMM:%[0-9]+]]:sgpr_256 = S_LOAD_DWORDX8_IMM undef %542:sreg_64, 0, 0 :: (invariant load (s256) from `ptr addrspace(4) undef`, addrspace 4)
   ; CHECK-NEXT:   [[V_OR_B32_e64_66:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_65]], [[V_ADD_U32_e64_30]], implicit $exec
   ; CHECK-NEXT:   [[S_ADD_I32_24:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM8]], -594, implicit-def dead $scc
   ; CHECK-NEXT:   [[V_OR_B32_e64_67:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[S_ADD_I32_24]], [[V_OR_B32_e64_66]], implicit $exec
   ; CHECK-NEXT:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 0, [[V_OR_B32_e64_67]], implicit $exec
   ; CHECK-NEXT:   undef [[V_CNDMASK_B32_e64_:%[0-9]+]].sub3:vreg_128 = V_CNDMASK_B32_e64 0, 0, 0, 1, [[V_CMP_EQ_U32_e64_]], implicit $exec
-  ; CHECK-NEXT:   IMAGE_STORE_V4_V2_gfx10 [[V_CNDMASK_B32_e64_]], undef %573:vreg_64, [[S_LOAD_DWORDX8_IMM]], 15, 1, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), addrspace 8)
+  ; CHECK-NEXT:   IMAGE_STORE_V4_V2_gfx10 [[V_CNDMASK_B32_e64_]], undef %556:vreg_64, [[S_LOAD_DWORDX8_IMM]], 15, 1, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), addrspace 8)
   ; CHECK-NEXT:   S_ENDPGM 0
 .expVert:
   %0 = extractelement <31 x i32> %userData, i64 2
diff --git a/llvm/test/CodeGen/PowerPC/licm-remat.ll b/llvm/test/CodeGen/PowerPC/licm-remat.ll
index ffdaf5d2481e3b..cf3e3ac089a498 100644
--- a/llvm/test/CodeGen/PowerPC/licm-remat.ll
+++ b/llvm/test/CodeGen/PowerPC/licm-remat.ll
@@ -21,7 +21,7 @@ define linkonce_odr void @ZN6snappyDecompressor_(ptr %this, ptr %writer) {
 ; CHECK-LABEL: ZN6snappyDecompressor_:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK:       addis 4, 2, .L__ModuleStringPool@toc@ha
-; CHECK:       addi 25, 4, .L__ModuleStringPool@toc@l
+; CHECK:       addi 26, 4, .L__ModuleStringPool@toc@l
 ; CHECK:       .LBB0_2: # %for.cond
 ; CHECK-NOT:   addis {{[0-9]+}}, 2, .L__ModuleStringPool@toc@ha
 ; CHECK:       bctrl
diff --git a/llvm/test/Transforms/SeparateConstOffsetFromGEP/AMDGPU/reorder-gep-inbounds.ll b/llvm/test/Transforms/SeparateConstOffsetFromGEP/AMDGPU/reorder-gep-inbounds.ll
new file mode 100644
index 00000000000000..c24bbd5f658f94
--- /dev/null
+++ b/llvm/test/Transforms/SeparateConstOffsetFromGEP/AMDGPU/reorder-gep-inbounds.ll
@@ -0,0 +1,51 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
+; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a -S -passes=separate-const-offset-from-gep < %s | FileCheck %s
+
+define void @inboundsPossiblyNegative(ptr %in.ptr, i32 %in.idx1) {
+; CHECK-LABEL: define void @inboundsPossiblyNegative(
+; CHECK-SAME: ptr [[IN_PTR:%.*]], i32 [[IN_IDX1:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[IN_IDX1]] to i64
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr <2 x i8>, ptr [[IN_PTR]], i64 [[IDXPROM]]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <2 x i8>, ptr [[TMP0]], i32 1
+; CHECK-NEXT:    ret void
+;
+entry:
+  %const1 = getelementptr inbounds <2 x i8>, ptr %in.ptr, i32 1
+  %idx1 = getelementptr inbounds <2 x i8>, ptr %const1, i32 %in.idx1
+  ret void
+}
+
+define void @inboundsNonNegative(ptr %in.ptr, i32 %in.idx1) {
+; CHECK-LABEL: define void @inboundsNonNegative(
+; CHECK-SAME: ptr [[IN_PTR:%.*]], i32 [[IN_IDX1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[IN_IDX1_NNEG:%.*]] = and i32 [[IN_IDX1]], 2147483647
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[IN_IDX1_NNEG]] to i64
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds <2 x i8>, ptr [[IN_PTR]], i64 [[IDXPROM]]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds <2 x i8>, ptr [[TMP0]], i32 1
+; CHECK-NEXT:    ret void
+;
+entry:
+  %in.idx1.nneg = and i32 %in.idx1, 2147483647
+  %const1 = getelementptr inbounds <2 x i8>, ptr %in.ptr, i32 1
+  %idx1 = getelementptr inbounds <2 x i8>, ptr %const1, i32 %in.idx1.nneg
+  ret void
+}
+
+define void @inboundsNonchained(ptr %in.ptr, i32 %in.idx1) {
+; CHECK-LABEL: define void @inboundsNonchained(
+; CHECK-SAME: ptr [[IN_PTR:%.*]], i32 [[IN_IDX1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[IN_IDX1_NNEG:%.*]] = and i32 [[IN_IDX1]], 2147483647
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[IN_IDX1_NNEG]] to i64
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr <2 x i8>, ptr [[IN_PTR]], i64 [[IDXPROM]]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <2 x i8>, ptr [[TMP0]], i32 1
+; CHECK-NEXT:    ret void
+;
+entry:
+  %in.idx1.nneg = and i32 %in.idx1, 2147483647
+  %const1 = getelementptr inbounds <2 x i8>, ptr %in.ptr, i32 1
+  %idx1 = getelementptr <2 x i8>, ptr %const1, i32 %in.idx1.nneg
+  ret void
+}
diff --git a/llvm/test/Transforms/SeparateConstOffsetFromGEP/AMDGPU/reorder-gep.ll b/llvm/test/Transforms/SeparateConstOffsetFromGEP/AMDGPU/reorder-gep.ll
new file mode 100644
index 00000000000000..7137f0fb66fdb9
--- /dev/null
+++ b/llvm/test/Transforms/SeparateConstOffsetFromGEP/AMDGPU/reorder-gep.ll
@@ -0,0 +1,175 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a --start-before=separate-const-offset-from-gep < %s | FileCheck %s
+
+define protected amdgpu_kernel void @sink_addr(ptr addrspace(3) %in.ptr, i32 %in.idx0, i32 %in.idx1) {
+; CHECK-LABEL: sink_addr:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    s_lshl_b32 s3, s1, 1
+; CHECK-NEXT:    s_add_i32 s0, s0, s3
+; CHECK-NEXT:    s_lshl_b32 s2, s2, 1
+; CHECK-NEXT:    s_add_i32 s0, s0, s2
+; CHECK-NEXT:    s_cmp_lg_u32 s1, 0
+; CHECK-NEXT:    s_cbranch_scc1 .LBB0_2
+; CHECK-NEXT:  ; %bb.1: ; %bb.1
+; CHECK-NEXT:    v_mov_b32_e32 v12, s0
+; CHECK-NEXT:    ds_read_b128 v[0:3], v12
+; CHECK-NEXT:    ds_read_b128 v[4:7], v12 offset:512
+; CHECK-NEXT:    ds_read_b128 v[8:11], v12 offset:1024
+; CHECK-NEXT:    ds_read_b128 v[12:15], v12 offset:1536
+; CHECK-NEXT:    s_waitcnt lgkmcnt(3)
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v[0:3]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    s_waitcnt lgkmcnt(2)
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v[4:7]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    s_waitcnt lgkmcnt(1)
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v[8:11]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v[12:15]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:  .LBB0_2: ; %end
+; CHECK-NEXT:    s_add_i32 s1, s0, 0x200
+; CHECK-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-NEXT:    s_add_i32 s2, s0, 0x400
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v0
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    v_mov_b32_e32 v0, s1
+; CHECK-NEXT:    s_add_i32 s3, s0, 0x600
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v0
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v0
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    v_mov_b32_e32 v0, s3
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v0
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    s_endpgm
+entry:
+  %base = getelementptr half, ptr addrspace(3) %in.ptr, i32 %in.idx0
+  %idx0 = getelementptr half, ptr addrspace(3) %base, i32 %in.idx1
+  %const1 = getelementptr half, ptr addrspace(3) %base, i32 256
+  %idx1 = getelementptr half, ptr addrspace(3) %const1, i32 %in.idx1
+  %const2 = getelementptr half, ptr addrspace(3) %base, i32 512
+  %idx2 = getelementptr half, ptr addrspace(3) %const2, i32 %in.idx1
+  %const3 = getelementptr half, ptr addrspace(3) %base, i32 768
+  %idx3 = getelementptr half, ptr addrspace(3) %const3, i32 %in.idx1
+  %cmp0 = icmp eq i32 %in.idx0, 0
+  br i1 %cmp0, label %bb.1, label %end
+
+bb.1:
+  %val0 = load <8 x half>, ptr addrspace(3) %idx0, align 16
+  %val1 = load <8 x half>, ptr addrspace(3) %idx1, align 16
+  %val2 = load <8 x half>, ptr addrspace(3) %idx2, align 16
+  %val3 = load <8 x half>, ptr addrspace(3) %idx3, align 16
+  call void asm sideeffect "; use $0", "v"(<8 x half> %val0)
+  call void asm sideeffect "; use $0", "v"(<8 x half> %val1)
+  call void asm sideeffect "; use $0", "v"(<8 x half> %val2)
+  call void asm sideeffect "; use $0", "v"(<8 x half> %val3)
+  br label %end
+
+end:
+  call void asm sideeffect "; use $0", "v"(ptr addrspace(3) %idx0)
+  call void asm sideeffect "; use $0", "v"(ptr addrspace(3) %idx1)
+  call void asm sideeffect "; use $0", "v"(ptr addrspace(3) %idx2)
+  call void asm sideeffect "; use $0", "v"(ptr addrspace(3) %idx3)
+  ret void
+}
+
+define protected amdgpu_kernel void @illegal_addr_mode(ptr addrspace(3) %in.ptr, i32 %in.idx0, i32 %in.idx1) {
+; CHECK-LABEL: illegal_addr_mode:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx4 s[4:7], s[6:7], 0x0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    s_lshl_b32 s0, s5, 1
+; CHECK-NEXT:    s_lshl_b32 s1, s6, 1
+; CHECK-NEXT:    s_add_i32 s3, s4, s0
+; CHECK-NEXT:    s_add_i32 s3, s3, s1
+; CHECK-NEXT:    s_add_i32 s2, s3, 0x12a60
+; CHECK-NEXT:    s_add_i32 s1, s3, 0x12c60
+; CHECK-NEXT:    s_add_i32 s0, s3, 0x12ed8
+; CHECK-NEXT:    s_cmp_lg_u32 s5, 0
+; CHECK-NEXT:    s_cbranch_scc1 .LBB1_2
+; CHECK-NEXT:  ; %bb.1: ; %bb.1
+; CHECK-NEXT:    v_mov_b32_e32 v0, s3
+; CHECK-NEXT:    v_mov_b32_e32 v4, s2
+; CHECK-NEXT:    v_mov_b32_e32 v8, s1
+; CHECK-NEXT:    v_mov_b32_e32 v12, s0
+; CHECK-NEXT:    ds_read_b128 v[0:3], v0
+; CHECK-NEXT:    ds_read_b128 v[4:7], v4
+; CHECK-NEXT:    ds_read_b128 v[8:11], v8
+; CHECK-NEXT:    ds_read_b128 v[12:15], v12
+; CHECK-NEXT:    s_waitcnt lgkmcnt(3)
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v[0:3]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    s_waitcnt lgkmcnt(2)
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v[4:7]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    s_waitcnt lgkmcnt(1)
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v[8:11]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v[12:15]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:  .LBB1_2: ; %end
+; CHECK-NEXT:    v_mov_b32_e32 v0, s3
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v0
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v0
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    v_mov_b32_e32 v0, s1
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v0
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use v0
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    s_endpgm
+entry:
+  %base = getelementptr half, ptr addrspace(3) %in.ptr, i32 %in.idx0
+  %idx0 = getelementptr half, ptr addrspace(3) %base, i32 %in.idx1
+  %const1 = getelementptr half, ptr addrspace(3) %base, i32 38192
+  %idx1 = getelementptr half, ptr addrspace(3) %const1, i32 %in.idx1
+  %const2 = getelementptr half, ptr addrspace(3) %base, i32 38448
+  %idx2 = getelementptr half, ptr addrspace(3) %const2, i32 %in.idx1
+  %const3 = getelementptr half, ptr addrspace(3) %base, i32 38764
+  %idx3 = getelementptr half, ptr addrspace(3) %const3, i32 %in.idx1
+  %cmp0 = icmp eq i32 %in.idx0, 0
+  br i1 %cmp0, label %bb.1, label %end
+
+bb.1:
+  %val0 = load <8 x half>, ptr addrspace(3) %idx0, align 16
+  %val1 = load <8 x half>, ptr addrspace(3) %idx1, align 16
+  %val2 = load <8 x half>, ptr addrspace(3) %idx2, align 16
+  %val3 = load <8 x half>, ptr addrspace(3) %idx3, align 16
+  call void asm sideeffect "; use $0", "v"(<8 x half> %val0)
+  call void asm sideeffect "; use $0", "v"(<8 x half> %val1)
+  call void asm sideeffect "; use $0", "v"(<8 x half> %val2)
+  call void asm sideeffect "; use $0", "v"(<8 x half> %val3)
+  br label %end
+
+end:
+  call void asm sideeffect "; use $0", "v"(ptr addrspace(3) %idx0)
+  call void asm sideeffect "; use $0", "v"(ptr addrspace(3) %idx1)
+  call void asm sideeffect "; use $0", "v"(ptr addrspace(3) %idx2)
+  call void asm sideeffect "; use $0", "v"(ptr addrspace(3) %idx3)
+  ret void
+}
diff --git a/llvm/test/Transforms/SeparateConstOffsetFromGEP/NVPTX/lower-gep-reorder.ll b/llvm/test/Transforms/SeparateConstOffsetFromGEP/NVPTX/lower-gep-reorder.ll
new file mode 100644
index 00000000000000..a91c8172177f9d
--- /dev/null
+++ b/llvm/test/Transforms/SeparateConstOffsetFromGEP/NVPTX/lower-gep-reorder.ll
@@ -0,0 +1,65 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
+; RUN: opt -mtriple=nvptx64-nvidia-cuda -S -passes=separate-const-offset-from-gep  < %s | FileCheck %s
+
+define protected amdgpu_kernel void @sink_addr(ptr %in.ptr, i64 %in.idx0, i64 %in.idx1) {
+; CHECK-LABEL: define protected amdgpu_kernel void @sink_addr(
+; CHECK-SAME: ptr [[IN_PTR:%.*]], i64 [[IN_IDX0:%.*]], i64 [[IN_IDX1:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[IDX0:%.*]] = getelementptr [8192 x i64], ptr [[IN_PTR]], i64 [[IN_IDX0]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr [8192 x i64], ptr [[IN_PTR]], i64 [[IN_IDX0]], i64 0
+; CHECK-NEXT:    [[CONST11:%.*]] = getelementptr i8, ptr [[TMP0]], i64 2048
+; CHECK-NEXT:    [[IDX1:%.*]] = getelementptr i64, ptr [[CONST11]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr [8192 x i64], ptr [[IN_PTR]], i64 [[IN_IDX0]], i64 0
+; CHECK-NEXT:    [[CONST22:%.*]] = getelementptr i8, ptr [[TMP1]], i64 4096
+; CHECK-NEXT:    [[IDX2:%.*]] = getelementptr i64, ptr [[CONST22]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr [8192 x i64], ptr [[IN_PTR]], i64 [[IN_IDX0]], i64 0
+; CHECK-NEXT:    [[CONST33:%.*]] = getelementptr i8, ptr [[TMP2]], i64 6144
+; CHECK-NEXT:    [[IDX3:%.*]] = getelementptr i64, ptr [[CONST33]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CMP0:%.*]] = icmp eq i64 [[IN_IDX0]], 0
+; CHECK-NEXT:    br i1 [[CMP0]], label [[BB_1:%.*]], label [[END:%.*]]
+; CHECK:       bb.1:
+; CHECK-NEXT:    [[VAL0:%.*]] = load <8 x i64>, ptr [[IDX0]], align 16
+; CHECK-NEXT:    [[VAL1:%.*]] = load <8 x i64>, ptr [[IDX1]], align 16
+; CHECK-NEXT:    [[VAL2:%.*]] = load <8 x i64>, ptr [[IDX2]], align 16
+; CHECK-NEXT:    [[VAL3:%.*]] = load <8 x i64>, ptr [[IDX3]], align 16
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       end:
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    ret void
+;
+entry:
+  %idx0 = getelementptr [8192 x i64], ptr %in.ptr, i64 %in.idx0, i64 %in.idx1
+  %const1 = getelementptr [8192 x i64], ptr %in.ptr, i64 %in.idx0, i64 256
+  %idx1 = getelementptr i64, ptr  %const1, i64 %in.idx1
+  %const2 = getelementptr [8192 x i64], ptr %in.ptr, i64 %in.idx0, i64 512
+  %idx2 = getelementptr i64, ptr  %const2, i64 %in.idx1
+  %const3 = getelementptr [8192 x i64], ptr %in.ptr, i64 %in.idx0, i64 768
+  %idx3 = getelementptr i64, ptr  %const3, i64 %in.idx1
+  %cmp0 = icmp eq i64 %in.idx0, 0
+  br i1 %cmp0, label %bb.1, label %end
+
+bb.1:
+  %val0 = load <8 x i64>, ptr  %idx0, align 16
+  %val1 = load <8 x i64>, ptr  %idx1, align 16
+  %val2 = load <8 x i64>, ptr  %idx2, align 16
+  %val3 = load <8 x i64>, ptr  %idx3, align 16
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val0)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val1)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val2)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val3)
+  br label %end
+
+end:
+  call void asm sideeffect "; use $0", "v"(ptr  %idx0)
+  call void asm sideeffect "; use $0", "v"(ptr  %idx1)
+  call void asm sideeffect "; use $0", "v"(ptr  %idx2)
+  call void asm sideeffect "; use $0", "v"(ptr  %idx3)
+  ret void
+}
diff --git a/llvm/test/Transforms/SeparateConstOffsetFromGEP/reorder-gep.ll b/llvm/test/Transforms/SeparateConstOffsetFromGEP/reorder-gep.ll
new file mode 100644
index 00000000000000..a15f11a634db5d
--- /dev/null
+++ b/llvm/test/Transforms/SeparateConstOffsetFromGEP/reorder-gep.ll
@@ -0,0 +1,188 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
+; RUN: opt -S -passes=separate-const-offset-from-gep < %s | FileCheck %s
+
+define void @illegal_addr_mode(ptr %in.ptr, i64 %in.idx0, i64 %in.idx1) {
+; CHECK-LABEL: define void @illegal_addr_mode(
+; CHECK-SAME: ptr [[IN_PTR:%.*]], i64 [[IN_IDX0:%.*]], i64 [[IN_IDX1:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[BASE:%.*]] = getelementptr i64, ptr [[IN_PTR]], i64 [[IN_IDX0]]
+; CHECK-NEXT:    [[IDX0:%.*]] = getelementptr i64, ptr [[BASE]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CONST1:%.*]] = getelementptr i64, ptr [[BASE]], i64 256
+; CHECK-NEXT:    [[IDX1:%.*]] = getelementptr i64, ptr [[CONST1]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CONST2:%.*]] = getelementptr i64, ptr [[BASE]], i64 512
+; CHECK-NEXT:    [[IDX2:%.*]] = getelementptr i64, ptr [[CONST2]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CONST3:%.*]] = getelementptr i64, ptr [[BASE]], i64 768
+; CHECK-NEXT:    [[IDX3:%.*]] = getelementptr i64, ptr [[CONST3]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CMP0:%.*]] = icmp eq i64 [[IN_IDX0]], 0
+; CHECK-NEXT:    br i1 [[CMP0]], label [[BB_1:%.*]], label [[END:%.*]]
+; CHECK:       bb.1:
+; CHECK-NEXT:    [[VAL0:%.*]] = load <8 x i64>, ptr [[IDX0]], align 16
+; CHECK-NEXT:    [[VAL1:%.*]] = load <8 x i64>, ptr [[IDX1]], align 16
+; CHECK-NEXT:    [[VAL2:%.*]] = load <8 x i64>, ptr [[IDX2]], align 16
+; CHECK-NEXT:    [[VAL3:%.*]] = load <8 x i64>, ptr [[IDX3]], align 16
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       end:
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    ret void
+;
+entry:
+  %base = getelementptr i64, ptr %in.ptr, i64 %in.idx0
+  %idx0 = getelementptr i64, ptr %base, i64 %in.idx1
+  %const1 = getelementptr i64, ptr %base, i64 256
+  %idx1 = getelementptr i64, ptr %const1, i64 %in.idx1
+  %const2 = getelementptr i64, ptr %base, i64 512
+  %idx2 = getelementptr i64, ptr %const2, i64 %in.idx1
+  %const3 = getelementptr i64, ptr %base, i64 768
+  %idx3 = getelementptr i64, ptr %const3, i64 %in.idx1
+  %cmp0 = icmp eq i64 %in.idx0, 0
+  br i1 %cmp0, label %bb.1, label %end
+
+bb.1:
+  %val0 = load <8 x i64>, ptr %idx0, align 16
+  %val1 = load <8 x i64>, ptr %idx1, align 16
+  %val2 = load <8 x i64>, ptr %idx2, align 16
+  %val3 = load <8 x i64>, ptr %idx3, align 16
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val0)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val1)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val2)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val3)
+  br label %end
+
+end:
+  call void asm sideeffect "; use $0", "v"(ptr %idx0)
+  call void asm sideeffect "; use $0", "v"(ptr %idx1)
+  call void asm sideeffect "; use $0", "v"(ptr %idx2)
+  call void asm sideeffect "; use $0", "v"(ptr %idx3)
+  ret void
+}
+
+
+define void @multi_index_reorder(ptr %in.ptr, i64 %in.idx0, i64 %in.idx1) {
+; CHECK-LABEL: define void @multi_index_reorder(
+; CHECK-SAME: ptr [[IN_PTR:%.*]], i64 [[IN_IDX0:%.*]], i64 [[IN_IDX1:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[IDX0:%.*]] = getelementptr [8192 x i64], ptr [[IN_PTR]], i64 0, i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CONST1:%.*]] = getelementptr [8192 x i64], ptr [[IN_PTR]], i64 0, i64 256
+; CHECK-NEXT:    [[IDX1:%.*]] = getelementptr i64, ptr [[CONST1]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CONST2:%.*]] = getelementptr [8192 x i64], ptr [[IN_PTR]], i64 0, i64 512
+; CHECK-NEXT:    [[IDX2:%.*]] = getelementptr i64, ptr [[CONST2]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CONST3:%.*]] = getelementptr [8192 x i64], ptr [[IN_PTR]], i64 0, i64 768
+; CHECK-NEXT:    [[IDX3:%.*]] = getelementptr i64, ptr [[CONST3]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CMP0:%.*]] = icmp eq i64 [[IN_IDX0]], 0
+; CHECK-NEXT:    br i1 [[CMP0]], label [[BB_1:%.*]], label [[END:%.*]]
+; CHECK:       bb.1:
+; CHECK-NEXT:    [[VAL0:%.*]] = load <8 x i64>, ptr [[IDX0]], align 16
+; CHECK-NEXT:    [[VAL1:%.*]] = load <8 x i64>, ptr [[IDX1]], align 16
+; CHECK-NEXT:    [[VAL2:%.*]] = load <8 x i64>, ptr [[IDX2]], align 16
+; CHECK-NEXT:    [[VAL3:%.*]] = load <8 x i64>, ptr [[IDX3]], align 16
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       end:
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    ret void
+;
+entry:
+  %idx0 = getelementptr [8192 x i64], ptr %in.ptr, i64 0, i64 %in.idx1
+  %const1 = getelementptr [8192 x i64], ptr %in.ptr, i64 0, i64 256
+  %idx1 = getelementptr i64, ptr %const1, i64 %in.idx1
+  %const2 = getelementptr [8192 x i64], ptr %in.ptr, i64 0, i64 512
+  %idx2 = getelementptr i64, ptr %const2, i64 %in.idx1
+  %const3 = getelementptr [8192 x i64], ptr %in.ptr, i64 0, i64 768
+  %idx3 = getelementptr i64, ptr %const3, i64 %in.idx1
+  %cmp0 = icmp eq i64 %in.idx0, 0
+  br i1 %cmp0, label %bb.1, label %end
+
+bb.1:
+  %val0 = load <8 x i64>, ptr %idx0, align 16
+  %val1 = load <8 x i64>, ptr %idx1, align 16
+  %val2 = load <8 x i64>, ptr %idx2, align 16
+  %val3 = load <8 x i64>, ptr %idx3, align 16
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val0)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val1)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val2)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val3)
+  br label %end
+
+end:
+  call void asm sideeffect "; use $0", "v"(ptr %idx0)
+  call void asm sideeffect "; use $0", "v"(ptr %idx1)
+  call void asm sideeffect "; use $0", "v"(ptr %idx2)
+  call void asm sideeffect "; use $0", "v"(ptr %idx3)
+  ret void
+}
+
+
+define void @different_type_reorder(ptr %in.ptr, i64 %in.idx0, i64 %in.idx1) {
+; CHECK-LABEL: define void @different_type_reorder(
+; CHECK-SAME: ptr [[IN_PTR:%.*]], i64 [[IN_IDX0:%.*]], i64 [[IN_IDX1:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[BASE:%.*]] = getelementptr i64, ptr [[IN_PTR]], i64 [[IN_IDX0]]
+; CHECK-NEXT:    [[IDX0:%.*]] = getelementptr i64, ptr [[BASE]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CONST1:%.*]] = getelementptr i8, ptr [[BASE]], i64 256
+; CHECK-NEXT:    [[IDX1:%.*]] = getelementptr i64, ptr [[CONST1]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CONST2:%.*]] = getelementptr i8, ptr [[BASE]], i64 512
+; CHECK-NEXT:    [[IDX2:%.*]] = getelementptr i64, ptr [[CONST2]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CONST3:%.*]] = getelementptr i8, ptr [[BASE]], i64 768
+; CHECK-NEXT:    [[IDX3:%.*]] = getelementptr i64, ptr [[CONST3]], i64 [[IN_IDX1]]
+; CHECK-NEXT:    [[CMP0:%.*]] = icmp eq i64 [[IN_IDX0]], 0
+; CHECK-NEXT:    br i1 [[CMP0]], label [[BB_1:%.*]], label [[END:%.*]]
+; CHECK:       bb.1:
+; CHECK-NEXT:    [[VAL0:%.*]] = load <8 x i64>, ptr [[IDX0]], align 16
+; CHECK-NEXT:    [[VAL1:%.*]] = load <8 x i64>, ptr [[IDX1]], align 16
+; CHECK-NEXT:    [[VAL2:%.*]] = load <8 x i64>, ptr [[IDX2]], align 16
+; CHECK-NEXT:    [[VAL3:%.*]] = load <8 x i64>, ptr [[IDX3]], align 16
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       end:
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    call void asm sideeffect "
+; CHECK-NEXT:    ret void
+;
+entry:
+  %base = getelementptr i64, ptr %in.ptr, i64 %in.idx0
+  %idx0 = getelementptr i64, ptr %base, i64 %in.idx1
+  %const1 = getelementptr i8, ptr %base, i64 256
+  %idx1 = getelementptr i64, ptr %const1, i64 %in.idx1
+  %const2 = getelementptr i8, ptr %base, i64 512
+  %idx2 = getelementptr i64, ptr %const2, i64 %in.idx1
+  %const3 = getelementptr i8, ptr %base, i64 768
+  %idx3 = getelementptr i64, ptr %const3, i64 %in.idx1
+  %cmp0 = icmp eq i64 %in.idx0, 0
+  br i1 %cmp0, label %bb.1, label %end
+
+bb.1:
+  %val0 = load <8 x i64>, ptr %idx0, align 16
+  %val1 = load <8 x i64>, ptr %idx1, align 16
+  %val2 = load <8 x i64>, ptr %idx2, align 16
+  %val3 = load <8 x i64>, ptr %idx3, align 16
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val0)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val1)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val2)
+  call void asm sideeffect "; use $0", "v"(<8 x i64> %val3)
+  br label %end
+
+end:
+  call void asm sideeffect "; use $0", "v"(ptr %idx0)
+  call void asm sideeffect "; use $0", "v"(ptr %idx1)
+  call void asm sideeffect "; use $0", "v"(ptr %idx2)
+  call void asm sideeffect "; use $0", "v"(ptr %idx3)
+  ret void
+}