From: Matt Arsenault <Matthew.Arsenault@amd.com>
Date: Tue, 16 Jul 2019 18:05:29 +0000 (+0000)
Subject: AMDGPU/GlobalISel: Select flat loads
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=b18edea81b726f1d4db4365475b4b2478eabb4b1;p=llvm

AMDGPU/GlobalISel: Select flat loads

Now that the patterns use the new PatFrag address space support, the
only blocker to importing most load patterns is the addressing mode
complex patterns.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@366237 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/AMDGPU/AMDGPUGISel.td b/lib/Target/AMDGPU/AMDGPUGISel.td
index 6f725d60907..1ccb90b2587 100644
--- a/lib/Target/AMDGPU/AMDGPUGISel.td
+++ b/lib/Target/AMDGPU/AMDGPUGISel.td
@@ -50,6 +50,14 @@ def gi_smrd_sgpr :
     GIComplexOperandMatcher<s64, "selectSmrdSgpr">,
     GIComplexPatternEquiv<SMRDSgpr>;
 
+def gi_flat_offset :
+    GIComplexOperandMatcher<s64, "selectFlatOffset">,
+    GIComplexPatternEquiv<FLATOffset>;
+def gi_flat_offset_signed :
+    GIComplexOperandMatcher<s64, "selectFlatOffsetSigned">,
+    GIComplexPatternEquiv<FLATOffsetSigned>;
+
+
 class GISelSop2Pat <
   SDPatternOperator node,
   Instruction inst,
diff --git a/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index 14ae62968c6..39016ed3719 100644
--- a/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -2937,18 +2937,11 @@ bool AMDGPUTargetLowering::SelectFlatOffset(bool IsSigned,
     SDValue N1 = Addr.getOperand(1);
     int64_t COffsetVal = cast<ConstantSDNode>(N1)->getSExtValue();
 
-    if (ST.getGeneration() >= AMDGPUSubtarget::GFX10) {
-      if ((IsSigned && isInt<12>(COffsetVal)) ||
-          (!IsSigned && isUInt<11>(COffsetVal))) {
-        Addr = N0;
-        OffsetVal = COffsetVal;
-      }
-    } else {
-      if ((IsSigned && isInt<13>(COffsetVal)) ||
-          (!IsSigned && isUInt<12>(COffsetVal))) {
-        Addr = N0;
-        OffsetVal = COffsetVal;
-      }
+    const SIInstrInfo *TII = ST.getInstrInfo();
+    if (TII->isLegalFLATOffset(COffsetVal, findMemSDNode(N)->getAddressSpace(),
+                               IsSigned)) {
+      Addr = N0;
+      OffsetVal = COffsetVal;
     }
   }
 
diff --git a/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp b/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
index aa634e881d8..f8f89593d08 100644
--- a/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
+++ b/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
@@ -1239,47 +1239,9 @@ bool AMDGPUInstructionSelector::hasVgprParts(ArrayRef<GEPInfo> AddrInfo) const {
 }
 
 bool AMDGPUInstructionSelector::selectG_LOAD(MachineInstr &I) const {
-  MachineBasicBlock *BB = I.getParent();
-  MachineFunction *MF = BB->getParent();
-  MachineRegisterInfo &MRI = MF->getRegInfo();
-  const DebugLoc &DL = I.getDebugLoc();
-  Register DstReg = I.getOperand(0).getReg();
-  Register PtrReg = I.getOperand(1).getReg();
-  unsigned LoadSize = RBI.getSizeInBits(DstReg, MRI, TRI);
-  unsigned Opcode;
-
-  if (MRI.getType(I.getOperand(1).getReg()).getSizeInBits() == 32) {
-    LLVM_DEBUG(dbgs() << "Unhandled address space\n");
-    return false;
-  }
-
-  SmallVector<GEPInfo, 4> AddrInfo;
-
-  getAddrModeInfo(I, MRI, AddrInfo);
-
-  switch (LoadSize) {
-  case 32:
-    Opcode = AMDGPU::FLAT_LOAD_DWORD;
-    break;
-  case 64:
-    Opcode = AMDGPU::FLAT_LOAD_DWORDX2;
-    break;
-  default:
-    LLVM_DEBUG(dbgs() << "Unhandled load size\n");
-    return false;
-  }
-
-  MachineInstr *Flat = BuildMI(*BB, &I, DL, TII.get(Opcode))
-                               .add(I.getOperand(0))
-                               .addReg(PtrReg)
-                               .addImm(0)  // offset
-                               .addImm(0)  // glc
-                               .addImm(0)  // slc
-                               .addImm(0); // dlc
-
-  bool Ret = constrainSelectedInstRegOperands(*Flat, TII, TRI, RBI);
-  I.eraseFromParent();
-  return Ret;
+  // TODO: Can/should we insert m0 initialization here for DS instructions and
+  // call the normal selector?
+  return false;
 }
 
 bool AMDGPUInstructionSelector::selectG_BRCOND(MachineInstr &I) const {
@@ -1397,9 +1359,7 @@ bool AMDGPUInstructionSelector::select(MachineInstr &I,
       return true;
     return selectImpl(I, CoverageInfo);
   case TargetOpcode::G_LOAD:
-    if (selectImpl(I, CoverageInfo))
-      return true;
-    return selectG_LOAD(I);
+    return selectImpl(I, CoverageInfo);
   case TargetOpcode::G_SELECT:
     return selectG_SELECT(I);
   case TargetOpcode::G_STORE:
@@ -1584,3 +1544,51 @@ AMDGPUInstructionSelector::selectSmrdSgpr(MachineOperand &Root) const {
     [=](MachineInstrBuilder &MIB) { MIB.addReg(OffsetReg); }
   }};
 }
+
+  template <bool Signed>
+InstructionSelector::ComplexRendererFns
+AMDGPUInstructionSelector::selectFlatOffsetImpl(MachineOperand &Root) const {
+  MachineInstr *MI = Root.getParent();
+  MachineBasicBlock *MBB = MI->getParent();
+  MachineRegisterInfo &MRI = MBB->getParent()->getRegInfo();
+
+  InstructionSelector::ComplexRendererFns Default = {{
+      [=](MachineInstrBuilder &MIB) { MIB.addReg(Root.getReg()); },
+      [=](MachineInstrBuilder &MIB) { MIB.addImm(0); },  // offset
+      [=](MachineInstrBuilder &MIB) { MIB.addImm(0); }  // slc
+    }};
+
+  if (!STI.hasFlatInstOffsets())
+    return Default;
+
+  const MachineInstr *OpDef = MRI.getVRegDef(Root.getReg());
+  if (!OpDef || OpDef->getOpcode() != AMDGPU::G_GEP)
+    return Default;
+
+  Optional<int64_t> Offset =
+    getConstantVRegVal(OpDef->getOperand(2).getReg(), MRI);
+  if (!Offset.hasValue())
+    return Default;
+
+  unsigned AddrSpace = (*MI->memoperands_begin())->getAddrSpace();
+  if (!TII.isLegalFLATOffset(Offset.getValue(), AddrSpace, Signed))
+    return Default;
+
+  Register BasePtr = OpDef->getOperand(1).getReg();
+
+  return {{
+      [=](MachineInstrBuilder &MIB) { MIB.addReg(BasePtr); },
+      [=](MachineInstrBuilder &MIB) { MIB.addImm(Offset.getValue()); },
+      [=](MachineInstrBuilder &MIB) { MIB.addImm(0); }  // slc
+    }};
+}
+
+InstructionSelector::ComplexRendererFns
+AMDGPUInstructionSelector::selectFlatOffset(MachineOperand &Root) const {
+  return selectFlatOffsetImpl<false>(Root);
+}
+
+InstructionSelector::ComplexRendererFns
+AMDGPUInstructionSelector::selectFlatOffsetSigned(MachineOperand &Root) const {
+  return selectFlatOffsetImpl<true>(Root);
+}
diff --git a/lib/Target/AMDGPU/AMDGPUInstructionSelector.h b/lib/Target/AMDGPU/AMDGPUInstructionSelector.h
index 1027a0b5683..e30d745f5cb 100644
--- a/lib/Target/AMDGPU/AMDGPUInstructionSelector.h
+++ b/lib/Target/AMDGPU/AMDGPUInstructionSelector.h
@@ -119,6 +119,15 @@ private:
   InstructionSelector::ComplexRendererFns
   selectSmrdSgpr(MachineOperand &Root) const;
 
+  template <bool Signed>
+  InstructionSelector::ComplexRendererFns
+  selectFlatOffsetImpl(MachineOperand &Root) const;
+  InstructionSelector::ComplexRendererFns
+  selectFlatOffset(MachineOperand &Root) const;
+
+  InstructionSelector::ComplexRendererFns
+  selectFlatOffsetSigned(MachineOperand &Root) const;
+
   const SIInstrInfo &TII;
   const SIRegisterInfo &TRI;
   const AMDGPURegisterBankInfo &RBI;
diff --git a/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 3cf4fbc7524..670f6225fbf 100644
--- a/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -497,6 +497,9 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
       .custom();
   }
 
+  // TODO: Should load to s16 be legal? Most loads extend to 32-bits, but we
+  // handle some operations by just promoting the register during
+  // selection. There are also d16 loads on GFX9+ which preserve the high bits.
   getActionDefinitionsBuilder({G_LOAD, G_STORE})
     .narrowScalarIf([](const LegalityQuery &Query) {
         unsigned Size = Query.Types[0].getSizeInBits();
diff --git a/lib/Target/AMDGPU/SIInstrInfo.cpp b/lib/Target/AMDGPU/SIInstrInfo.cpp
index 34741850f82..ba8ed6993a5 100644
--- a/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -6118,6 +6118,25 @@ bool SIInstrInfo::isBufferSMRD(const MachineInstr &MI) const {
   return RCID == AMDGPU::SReg_128RegClassID;
 }
 
+bool SIInstrInfo::isLegalFLATOffset(int64_t Offset, unsigned AddrSpace,
+                                    bool Signed) const {
+  // TODO: Should 0 be special cased?
+  if (!ST.hasFlatInstOffsets())
+    return false;
+
+  if (ST.hasFlatSegmentOffsetBug() && AddrSpace == AMDGPUAS::FLAT_ADDRESS)
+    return false;
+
+  if (ST.getGeneration() >= AMDGPUSubtarget::GFX10) {
+    return (Signed && isInt<12>(Offset)) ||
+           (!Signed && isUInt<11>(Offset));
+  }
+
+  return (Signed && isInt<13>(Offset)) ||
+         (!Signed && isUInt<12>(Offset));
+}
+
+
 // This must be kept in sync with the SIEncodingFamily class in SIInstrInfo.td
 enum SIEncodingFamily {
   SI = 0,
diff --git a/lib/Target/AMDGPU/SIInstrInfo.h b/lib/Target/AMDGPU/SIInstrInfo.h
index 1f3c659f9d9..3ff35da0b96 100644
--- a/lib/Target/AMDGPU/SIInstrInfo.h
+++ b/lib/Target/AMDGPU/SIInstrInfo.h
@@ -970,6 +970,12 @@ public:
     return isUInt<12>(Imm);
   }
 
+  /// Returns if \p Offset is legal for the subtarget as the offset to a FLAT
+  /// encoded instruction. If \p Signed, this is for an instruction that
+  /// interprets the offset as signed.
+  bool isLegalFLATOffset(int64_t Offset, unsigned AddrSpace,
+                         bool Signed) const;
+
   /// \brief Return a target-specific opcode if Opcode is a pseudo instruction.
   /// Return -1 if the target-specific opcode for the pseudo instruction does
   /// not exist. If Opcode is not a pseudo instruction, this is identity.
diff --git a/test/CodeGen/AMDGPU/GlobalISel/inst-select-load-flat.mir b/test/CodeGen/AMDGPU/GlobalISel/inst-select-load-flat.mir
index 8069dff2634..f579c3ce287 100644
--- a/test/CodeGen/AMDGPU/GlobalISel/inst-select-load-flat.mir
+++ b/test/CodeGen/AMDGPU/GlobalISel/inst-select-load-flat.mir
@@ -1,26 +1,1717 @@
-# RUN: llc -march=amdgcn -mcpu=hawaii -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=GCN
-# RUN: llc -march=amdgcn -mcpu=fiji -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=GCN
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -march=amdgcn -mcpu=hawaii -run-pass=instruction-select -verify-machineinstrs  -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX7 %s
+# RUN: llc -march=amdgcn -mcpu=fiji -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX8 %s
+# RUN: llc -march=amdgcn -mcpu=gfx900 -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX9 %s
+# RUN: llc -march=amdgcn -mcpu=gfx1010 -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX10 %s
+
+
+---
+
+name: load_flat_s32_from_4
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_4
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
+    ; GFX8-LABEL: name: load_flat_s32_from_4
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
+    ; GFX9-LABEL: name: load_flat_s32_from_4
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
+    ; GFX10-LABEL: name: load_flat_s32_from_4
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = G_LOAD %0 :: (load 4, align 4, addrspace 0)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_flat_s32_from_2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_2
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 2)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_USHORT]]
+    ; GFX8-LABEL: name: load_flat_s32_from_2
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 2)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_USHORT]]
+    ; GFX9-LABEL: name: load_flat_s32_from_2
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 2)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_USHORT]]
+    ; GFX10-LABEL: name: load_flat_s32_from_2
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 2)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_USHORT]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = G_LOAD %0 :: (load 2, align 2, addrspace 0)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_flat_s32_from_1
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = G_LOAD %0 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_flat_v2s32
+legalized:       true
+regBankSelected: true
+
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_v2s32
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
+    ; GFX8-LABEL: name: load_flat_v2s32
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
+    ; GFX9-LABEL: name: load_flat_v2s32
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
+    ; GFX10-LABEL: name: load_flat_v2s32
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s32>) = G_LOAD %0 :: (load 8, align 8, addrspace 0)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_flat_v3s32
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_v3s32
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 12, align 4)
+    ; GFX7: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]]
+    ; GFX8-LABEL: name: load_flat_v3s32
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 12, align 4)
+    ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]]
+    ; GFX9-LABEL: name: load_flat_v3s32
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 12, align 4)
+    ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]]
+    ; GFX10-LABEL: name: load_flat_v3s32
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 12, align 4)
+    ; GFX10: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<3 x  s32>) = G_LOAD %0 :: (load 12, align 4, addrspace 0)
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+
+...
+
+---
+
+name: load_flat_v4s32
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_v4s32
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 16, align 4)
+    ; GFX7: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]]
+    ; GFX8-LABEL: name: load_flat_v4s32
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 16, align 4)
+    ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]]
+    ; GFX9-LABEL: name: load_flat_v4s32
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 16, align 4)
+    ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]]
+    ; GFX10-LABEL: name: load_flat_v4s32
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 16, align 4)
+    ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<4 x  s32>) = G_LOAD %0 :: (load 16, align 4, addrspace 0)
+    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+
+...
+
+---
+
+name: load_flat_s64
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s64
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX8-LABEL: name: load_flat_s64
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX9-LABEL: name: load_flat_s64
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX10-LABEL: name: load_flat_s64
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_LOAD %0 :: (load 8, align 8, addrspace 0)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_flat_v2s64
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_v2s64
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_128(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX7: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+    ; GFX8-LABEL: name: load_flat_v2s64
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_128(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+    ; GFX9-LABEL: name: load_flat_v2s64
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_128(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+    ; GFX10-LABEL: name: load_flat_v2s64
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_128(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s64>) = G_LOAD %0 :: (load 16, align 4, addrspace 0)
+    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+
+...
+
+---
+
+name: load_flat_v2p1
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_v2p1
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX7: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
+    ; GFX8-LABEL: name: load_flat_v2p1
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
+    ; GFX9-LABEL: name: load_flat_v2p1
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
+    ; GFX10-LABEL: name: load_flat_v2p1
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x p1>) = G_LOAD %0 :: (load 16, align 4, addrspace 0)
+    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+
+...
+
+---
+
+name: load_flat_s96
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s96
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load 12, align 4)
+    ; GFX7: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96)
+    ; GFX8-LABEL: name: load_flat_s96
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load 12, align 4)
+    ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96)
+    ; GFX9-LABEL: name: load_flat_s96
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load 12, align 4)
+    ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96)
+    ; GFX10-LABEL: name: load_flat_s96
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load 12, align 4)
+    ; GFX10: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s96) = G_LOAD %0 :: (load 12, align 4, addrspace 0)
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+
+...
+
+---
+
+name: load_flat_s128
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s128
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX7: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
+    ; GFX8-LABEL: name: load_flat_s128
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
+    ; GFX9-LABEL: name: load_flat_s128
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
+    ; GFX10-LABEL: name: load_flat_s128
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s128) = G_LOAD %0 :: (load 16, align 4, addrspace 0)
+    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+
+...
+
+---
+
+name: load_flat_p3_from_4
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_p3_from_4
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4)
+    ; GFX7: $vgpr0 = COPY [[LOAD]](p3)
+    ; GFX8-LABEL: name: load_flat_p3_from_4
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4)
+    ; GFX8: $vgpr0 = COPY [[LOAD]](p3)
+    ; GFX9-LABEL: name: load_flat_p3_from_4
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4)
+    ; GFX9: $vgpr0 = COPY [[LOAD]](p3)
+    ; GFX10-LABEL: name: load_flat_p3_from_4
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4)
+    ; GFX10: $vgpr0 = COPY [[LOAD]](p3)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(p3) = G_LOAD %0 :: (load 4, align 4, addrspace 0)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_flat_p1_from_8
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_p1_from_8
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+    ; GFX8-LABEL: name: load_flat_p1_from_8
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+    ; GFX9-LABEL: name: load_flat_p1_from_8
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+    ; GFX10-LABEL: name: load_flat_p1_from_8
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(p1) = G_LOAD %0 :: (load 8, align 8, addrspace 0)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_flat_p999_from_8
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_p999_from_8
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](p999)
+    ; GFX8-LABEL: name: load_flat_p999_from_8
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](p999)
+    ; GFX9-LABEL: name: load_flat_p999_from_8
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](p999)
+    ; GFX10-LABEL: name: load_flat_p999_from_8
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](p999)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(p999) = G_LOAD %0 :: (load 8, align 8, addrspace 0)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_flat_v2p3
+legalized:       true
+regBankSelected: true
+
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_v2p3
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
+    ; GFX8-LABEL: name: load_flat_v2p3
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
+    ; GFX9-LABEL: name: load_flat_v2p3
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
+    ; GFX10-LABEL: name: load_flat_v2p3
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x p3>) = G_LOAD %0 :: (load 8, align 8, addrspace 0)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_flat_v2s16
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_v2s16
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vgpr_32(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load 4)
+    ; GFX7: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+    ; GFX8-LABEL: name: load_flat_v2s16
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vgpr_32(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load 4)
+    ; GFX8: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+    ; GFX9-LABEL: name: load_flat_v2s16
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vgpr_32(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load 4)
+    ; GFX9: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+    ; GFX10-LABEL: name: load_flat_v2s16
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vgpr_32(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load 4)
+    ; GFX10: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s16>) = G_LOAD %0 :: (load 4, align 4, addrspace 0)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_flat_v4s16
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_v4s16
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+    ; GFX8-LABEL: name: load_flat_v4s16
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+    ; GFX9-LABEL: name: load_flat_v4s16
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+    ; GFX10-LABEL: name: load_flat_v4s16
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 8)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<4 x s16>) = G_LOAD %0 :: (load 8, align 8, addrspace 0)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_flat_v6s16
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_v6s16
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load 12, align 4)
+    ; GFX7: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>)
+    ; GFX8-LABEL: name: load_flat_v6s16
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load 12, align 4)
+    ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>)
+    ; GFX9-LABEL: name: load_flat_v6s16
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load 12, align 4)
+    ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>)
+    ; GFX10-LABEL: name: load_flat_v6s16
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load 12, align 4)
+    ; GFX10: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<6 x  s16>) = G_LOAD %0 :: (load 12, align 4, addrspace 0)
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+
+...
+
+---
+
+name: load_flat_v8s16
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_v8s16
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_128(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX7: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
+    ; GFX8-LABEL: name: load_flat_v8s16
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_128(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
+    ; GFX9-LABEL: name: load_flat_v8s16
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_128(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
+    ; GFX10-LABEL: name: load_flat_v8s16
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_128(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load 16, align 4)
+    ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<8 x  s16>) = G_LOAD %0 :: (load 16, align 4, addrspace 0)
+    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+
+...
+
+################################################################################
+### Stress addressing modes
+################################################################################
+
+---
+
+name: load_flat_s32_from_1_gep_2047
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_2047
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2047, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_2047
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2047, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_2047
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 2047, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_2047
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2047, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 2047
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_flat_s32_from_1_gep_2048
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_2048
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2048, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_2048
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2048, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_2048
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 2048, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_2048
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2048, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 2048
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_flat_s32_from_1_gep_m2047
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m2047
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m2047
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m2047
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m2047
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -2047
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_flat_s32_from_1_gep_m2048
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m2048
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m2048
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m2048
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m2048
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -2048
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_flat_s32_from_1_gep_4095
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_4095
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4095, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_4095
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4095, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_4095
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 4095, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_4095
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4095, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 4095
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_flat_s32_from_1_gep_4096
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_4096
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_4096
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_4096
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_4096
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 4096
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_flat_s32_from_1_gep_m4095
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m4095
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m4095
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m4095
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m4095
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -4095
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_flat_s32_from_1_gep_m4096
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m4096
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m4096
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m4096
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m4096
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -4096
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_flat_s32_from_1_gep_8191
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_8191
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_8191
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_8191
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_8191
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 8191
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
 
---- |
-  define amdgpu_kernel void @global_addrspace(i32 addrspace(1)* %global0) { ret void }
 ...
+
 ---
 
-name:            global_addrspace
+name: load_flat_s32_from_1_gep_8192
 legalized:       true
 regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_8192
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_8192
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_8192
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_8192
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 8192
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
 
-# GCN: global_addrspace
-# GCN: [[PTR:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
-# GCN: FLAT_LOAD_DWORD  [[PTR]], 0, 0, 0, 0
+...
+
+---
+
+name: load_flat_s32_from_1_gep_m8191
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
 
 body: |
   bb.0:
     liveins:  $vgpr0_vgpr1
 
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m8191
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m8191
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m8191
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m8191
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
     %0:vgpr(p1) = COPY $vgpr0_vgpr1
-    %1:vgpr(s32) = G_LOAD %0 :: (load 4 from %ir.global0)
+    %1:vgpr(s64) = G_CONSTANT i64 -8191
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_flat_s32_from_1_gep_m8192
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m8192
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m8192
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m8192
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX9: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m8192
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1)
+    ; GFX10: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -8192
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 0)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_atomic_flat_s32
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_atomic_flat_s32
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vgpr_32(s32) = G_LOAD [[COPY]](p1) :: (load monotonic 4)
+    ; GFX7: $vgpr0 = COPY [[LOAD]](s32)
+    ; GFX8-LABEL: name: load_atomic_flat_s32
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vgpr_32(s32) = G_LOAD [[COPY]](p1) :: (load monotonic 4)
+    ; GFX8: $vgpr0 = COPY [[LOAD]](s32)
+    ; GFX9-LABEL: name: load_atomic_flat_s32
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vgpr_32(s32) = G_LOAD [[COPY]](p1) :: (load monotonic 4)
+    ; GFX9: $vgpr0 = COPY [[LOAD]](s32)
+    ; GFX10-LABEL: name: load_atomic_flat_s32
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vgpr_32(s32) = G_LOAD [[COPY]](p1) :: (load monotonic 4)
+    ; GFX10: $vgpr0 = COPY [[LOAD]](s32)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = G_LOAD %0 :: (load monotonic  4, align 4, addrspace 0)
     $vgpr0 = COPY %1
 
 ...
+
 ---
+
+name: load_atomic_flat_s64
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_atomic_flat_s64
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load monotonic 8)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX8-LABEL: name: load_atomic_flat_s64
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load monotonic 8)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX9-LABEL: name: load_atomic_flat_s64
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load monotonic 8)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX10-LABEL: name: load_atomic_flat_s64
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load monotonic 8)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_LOAD %0 :: (load monotonic 8, align 8, addrspace 0)
+    $vgpr0_vgpr1 = COPY %1
+
+...
diff --git a/test/CodeGen/AMDGPU/GlobalISel/inst-select-load-global.mir b/test/CodeGen/AMDGPU/GlobalISel/inst-select-load-global.mir
new file mode 100644
index 00000000000..df86d18c3b3
--- /dev/null
+++ b/test/CodeGen/AMDGPU/GlobalISel/inst-select-load-global.mir
@@ -0,0 +1,1657 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -march=amdgcn -mcpu=hawaii -run-pass=instruction-select -verify-machineinstrs  -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX7 %s
+# RUN: llc -march=amdgcn -mcpu=fiji -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX8 %s
+# RUN: llc -march=amdgcn -mcpu=gfx900 -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX9 %s
+# RUN: llc -march=amdgcn -mcpu=gfx1010 -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX10 %s
+
+# FIXME: global with MUBUF
+
+---
+
+name: load_global_s32_from_4
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_4
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
+    ; GFX8-LABEL: name: load_global_s32_from_4
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
+    ; GFX9-LABEL: name: load_global_s32_from_4
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_DWORD:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 4, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_DWORD]]
+    ; GFX10-LABEL: name: load_global_s32_from_4
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[GLOBAL_LOAD_DWORD:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 4, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_DWORD]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = G_LOAD %0 :: (load 4, align 4, addrspace 1)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_global_s32_from_2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_2
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 2, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_USHORT]]
+    ; GFX8-LABEL: name: load_global_s32_from_2
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 2, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_USHORT]]
+    ; GFX9-LABEL: name: load_global_s32_from_2
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_USHORT:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_USHORT [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 2, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_USHORT]]
+    ; GFX10-LABEL: name: load_global_s32_from_2
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[GLOBAL_LOAD_USHORT:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_USHORT [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 2, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_USHORT]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = G_LOAD %0 :: (load 2, align 2, addrspace 1)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_global_s32_from_1
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = G_LOAD %0 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_global_v2s32
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_v2s32
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
+    ; GFX8-LABEL: name: load_global_v2s32
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
+    ; GFX9-LABEL: name: load_global_v2s32
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = GLOBAL_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 8, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[GLOBAL_LOAD_DWORDX2_]]
+    ; GFX10-LABEL: name: load_global_v2s32
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[GLOBAL_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = GLOBAL_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 8, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[GLOBAL_LOAD_DWORDX2_]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s32>) = G_LOAD %0 :: (load 8, align 8, addrspace 1)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_global_v3s32
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_v3s32
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 12, align 4, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]]
+    ; GFX8-LABEL: name: load_global_v3s32
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 12, align 4, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]]
+    ; GFX9-LABEL: name: load_global_v3s32
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = GLOBAL_LOAD_DWORDX3 [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 12, align 4, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[GLOBAL_LOAD_DWORDX3_]]
+    ; GFX10-LABEL: name: load_global_v3s32
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[GLOBAL_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = GLOBAL_LOAD_DWORDX3 [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 12, align 4, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1_vgpr2 = COPY [[GLOBAL_LOAD_DWORDX3_]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<3 x  s32>) = G_LOAD %0 :: (load 12, align 4, addrspace 1)
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+
+...
+
+---
+
+name: load_global_v4s32
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_v4s32
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 16, align 4, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]]
+    ; GFX8-LABEL: name: load_global_v4s32
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 16, align 4, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]]
+    ; GFX9-LABEL: name: load_global_v4s32
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = GLOBAL_LOAD_DWORDX4 [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 16, align 4, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[GLOBAL_LOAD_DWORDX4_]]
+    ; GFX10-LABEL: name: load_global_v4s32
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[GLOBAL_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = GLOBAL_LOAD_DWORDX4 [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 16, align 4, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[GLOBAL_LOAD_DWORDX4_]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<4 x  s32>) = G_LOAD %0 :: (load 16, align 4, addrspace 1)
+    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+
+...
+
+---
+
+name: load_global_s64
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s64
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX8-LABEL: name: load_global_s64
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX9-LABEL: name: load_global_s64
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX10-LABEL: name: load_global_s64
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_LOAD %0 :: (load 8, align 8, addrspace 1)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_global_v2s64
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_v2s64
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_128(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+    ; GFX8-LABEL: name: load_global_v2s64
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_128(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+    ; GFX9-LABEL: name: load_global_v2s64
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_128(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+    ; GFX10-LABEL: name: load_global_v2s64
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_128(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s64>) = G_LOAD %0 :: (load 16, align 4, addrspace 1)
+    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+
+...
+
+---
+
+name: load_global_v2p1
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_v2p1
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
+    ; GFX8-LABEL: name: load_global_v2p1
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
+    ; GFX9-LABEL: name: load_global_v2p1
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
+    ; GFX10-LABEL: name: load_global_v2p1
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x p1>) = G_LOAD %0 :: (load 16, align 4, addrspace 1)
+    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+
+...
+
+---
+
+name: load_global_s96
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s96
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load 12, align 4, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96)
+    ; GFX8-LABEL: name: load_global_s96
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load 12, align 4, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96)
+    ; GFX9-LABEL: name: load_global_s96
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load 12, align 4, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96)
+    ; GFX10-LABEL: name: load_global_s96
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load 12, align 4, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s96) = G_LOAD %0 :: (load 12, align 4, addrspace 1)
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+
+...
+
+---
+
+name: load_global_s128
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s128
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
+    ; GFX8-LABEL: name: load_global_s128
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
+    ; GFX9-LABEL: name: load_global_s128
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
+    ; GFX10-LABEL: name: load_global_s128
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s128) = G_LOAD %0 :: (load 16, align 4, addrspace 1)
+    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+
+...
+
+---
+
+name: load_global_p3_from_4
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_p3_from_4
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[LOAD]](p3)
+    ; GFX8-LABEL: name: load_global_p3_from_4
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[LOAD]](p3)
+    ; GFX9-LABEL: name: load_global_p3_from_4
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[LOAD]](p3)
+    ; GFX10-LABEL: name: load_global_p3_from_4
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[LOAD]](p3)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(p3) = G_LOAD %0 :: (load 4, align 4, addrspace 1)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_global_p1_from_8
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_p1_from_8
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+    ; GFX8-LABEL: name: load_global_p1_from_8
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+    ; GFX9-LABEL: name: load_global_p1_from_8
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+    ; GFX10-LABEL: name: load_global_p1_from_8
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(p1) = G_LOAD %0 :: (load 8, align 8, addrspace 1)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_global_p999_from_8
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_p999_from_8
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](p999)
+    ; GFX8-LABEL: name: load_global_p999_from_8
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](p999)
+    ; GFX9-LABEL: name: load_global_p999_from_8
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](p999)
+    ; GFX10-LABEL: name: load_global_p999_from_8
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](p999)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(p999) = G_LOAD %0 :: (load 8, align 8, addrspace 1)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_global_v2p3
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_v2p3
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
+    ; GFX8-LABEL: name: load_global_v2p3
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
+    ; GFX9-LABEL: name: load_global_v2p3
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
+    ; GFX10-LABEL: name: load_global_v2p3
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x p3>) = G_LOAD %0 :: (load 8, align 8, addrspace 1)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_global_v2s16
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_v2s16
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vgpr_32(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+    ; GFX8-LABEL: name: load_global_v2s16
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vgpr_32(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+    ; GFX9-LABEL: name: load_global_v2s16
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vgpr_32(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+    ; GFX10-LABEL: name: load_global_v2s16
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vgpr_32(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s16>) = G_LOAD %0 :: (load 4, align 4, addrspace 1)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_global_v4s16
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_v4s16
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+    ; GFX8-LABEL: name: load_global_v4s16
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+    ; GFX9-LABEL: name: load_global_v4s16
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+    ; GFX10-LABEL: name: load_global_v4s16
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<4 x s16>) = G_LOAD %0 :: (load 8, align 8, addrspace 1)
+    $vgpr0_vgpr1 = COPY %1
+
+...
+
+---
+
+name: load_global_v6s16
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_v6s16
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load 12, align 4, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>)
+    ; GFX8-LABEL: name: load_global_v6s16
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load 12, align 4, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>)
+    ; GFX9-LABEL: name: load_global_v6s16
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load 12, align 4, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>)
+    ; GFX10-LABEL: name: load_global_v6s16
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load 12, align 4, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<6 x  s16>) = G_LOAD %0 :: (load 12, align 4, addrspace 1)
+    $vgpr0_vgpr1_vgpr2 = COPY %1
+
+...
+
+---
+
+name: load_global_v8s16
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_v8s16
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_128(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
+    ; GFX8-LABEL: name: load_global_v8s16
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_128(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
+    ; GFX9-LABEL: name: load_global_v8s16
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_128(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
+    ; GFX10-LABEL: name: load_global_v8s16
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_128(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<8 x  s16>) = G_LOAD %0 :: (load 16, align 4, addrspace 1)
+    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
+
+...
+
+################################################################################
+### Stress addressing modes
+################################################################################
+
+---
+
+name: load_global_s32_from_1_gep_2047
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_2047
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2047, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_2047
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2047, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_2047
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], 2047, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_2047
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], 2047, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 2047
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_2048
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_2048
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2048, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_2048
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2048, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_2048
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], 2048, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_2048
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2048, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 2048
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_m2047
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_m2047
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_m2047
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_m2047
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], -2047, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_m2047
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], -2047, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -2047
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_m2048
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_m2048
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_m2048
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_m2048
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], -2048, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_m2048
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], -2048, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -2048
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_4095
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_4095
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4095, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_4095
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4095, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_4095
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], 4095, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_4095
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4095, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 4095
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_4096
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_4096
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_4096
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_4096
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_4096
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 4096
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_m4095
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_m4095
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_m4095
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_m4095
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], -4095, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_m4095
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -4095
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_m4096
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_m4096
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_m4096
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_m4096
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[COPY]], -4096, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_m4096
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -4096
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_8191
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_8191
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_8191
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_8191
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_8191
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 8191
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_8192
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_8192
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_8192
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_8192
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_8192
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 8192
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_m8191
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_m8191
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_m8191
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_m8191
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_m8191
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -8191
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_global_s32_from_1_gep_m8192
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_global_s32_from_1_gep_m8192
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec
+    ; GFX7: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX7: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX7: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX8-LABEL: name: load_global_s32_from_1_gep_m8192
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec
+    ; GFX8: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX8: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX8: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 1, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]]
+    ; GFX9-LABEL: name: load_global_s32_from_1_gep_m8192
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec
+    ; GFX9: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX9: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX9: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX9: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX9: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX9: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX9: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX9: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX9: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    ; GFX10-LABEL: name: load_global_s32_from_1_gep_m8192
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec
+    ; GFX10: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec
+    ; GFX10: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1
+    ; GFX10: [[V_ADD_I32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_I32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_I32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_I32_e64_1]], 0, implicit $exec
+    ; GFX10: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_I32_e64_]], %subreg.sub0, %9, %subreg.sub1
+    ; GFX10: [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load 1, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_UBYTE]]
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_CONSTANT i64 -8192
+    %2:vgpr(p1) = G_GEP %0, %1
+    %3:vgpr(s32) = G_LOAD %2 :: (load 1, align 1, addrspace 1)
+    $vgpr0 = COPY %3
+
+...
+
+---
+
+name: load_atomic_global_s32
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_atomic_global_s32
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vgpr_32(s32) = G_LOAD [[COPY]](p1) :: (load monotonic 4, addrspace 1)
+    ; GFX7: $vgpr0 = COPY [[LOAD]](s32)
+    ; GFX8-LABEL: name: load_atomic_global_s32
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vgpr_32(s32) = G_LOAD [[COPY]](p1) :: (load monotonic 4, addrspace 1)
+    ; GFX8: $vgpr0 = COPY [[LOAD]](s32)
+    ; GFX9-LABEL: name: load_atomic_global_s32
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vgpr_32(s32) = G_LOAD [[COPY]](p1) :: (load monotonic 4, addrspace 1)
+    ; GFX9: $vgpr0 = COPY [[LOAD]](s32)
+    ; GFX10-LABEL: name: load_atomic_global_s32
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vgpr_32(s32) = G_LOAD [[COPY]](p1) :: (load monotonic 4, addrspace 1)
+    ; GFX10: $vgpr0 = COPY [[LOAD]](s32)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = G_LOAD %0 :: (load monotonic  4, align 4, addrspace 1)
+    $vgpr0 = COPY %1
+
+...
+
+---
+
+name: load_atomic_global_s64
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1
+
+    ; GFX7-LABEL: name: load_atomic_global_s64
+    ; GFX7: liveins: $vgpr0_vgpr1
+    ; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX7: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load monotonic 8, addrspace 1)
+    ; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX8-LABEL: name: load_atomic_global_s64
+    ; GFX8: liveins: $vgpr0_vgpr1
+    ; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX8: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load monotonic 8, addrspace 1)
+    ; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX9-LABEL: name: load_atomic_global_s64
+    ; GFX9: liveins: $vgpr0_vgpr1
+    ; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX9: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load monotonic 8, addrspace 1)
+    ; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    ; GFX10-LABEL: name: load_atomic_global_s64
+    ; GFX10: liveins: $vgpr0_vgpr1
+    ; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
+    ; GFX10: [[LOAD:%[0-9]+]]:vreg_64(s64) = G_LOAD [[COPY]](p1) :: (load monotonic 8, addrspace 1)
+    ; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = G_LOAD %0 :: (load monotonic 8, align 8, addrspace 1)
+    $vgpr0_vgpr1 = COPY %1
+
+...