AMDGPU/GlobalISel: Select flat stores

llvm-svn: 366246

8 files changed, 1650 insertions, 54 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
index f8f89593d08..25e72bbe75a 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
@@ -856,7 +856,7 @@ bool AMDGPUInstructionSelector::selectG_STORE(MachineInstr &I) const {
   unsigned StoreSize = RBI.getSizeInBits(I.getOperand(0).getReg(), MRI, TRI);
   unsigned Opcode;
 
-  // FIXME: Select store instruction based on address space
+  // FIXME: Remove this when integers > s32 naturally selected.
   switch (StoreSize) {
   default:
     return false;
@@ -1363,6 +1363,8 @@ bool AMDGPUInstructionSelector::select(MachineInstr &I,
   case TargetOpcode::G_SELECT:
     return selectG_SELECT(I);
   case TargetOpcode::G_STORE:
+    if (selectImpl(I, CoverageInfo))
+      return true;
     return selectG_STORE(I);
   case TargetOpcode::G_TRUNC:
     return selectG_TRUNC(I);
@@ -1545,7 +1547,7 @@ AMDGPUInstructionSelector::selectSmrdSgpr(MachineOperand &Root) const {
   }};
 }
 
-  template <bool Signed>
+template <bool Signed>
 InstructionSelector::ComplexRendererFns
 AMDGPUInstructionSelector::selectFlatOffsetImpl(MachineOperand &Root) const {
   MachineInstr *MI = Root.getParent();
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy.mir
index 2f2ad31cd0a..558f672c208 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy.mir
@@ -17,13 +17,13 @@ body: |
     ; WAVE64: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr2_sgpr3
     ; WAVE64: [[COPY1:%[0-9]+]]:vreg_64 = COPY [[COPY]]
     ; WAVE64: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
-    ; WAVE64: FLAT_STORE_DWORD [[COPY1]], [[DEF]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; WAVE64: FLAT_STORE_DWORD [[COPY1]], [[DEF]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
     ; WAVE32-LABEL: name: copy
     ; WAVE32: $vcc_hi = IMPLICIT_DEF
     ; WAVE32: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr2_sgpr3
     ; WAVE32: [[COPY1:%[0-9]+]]:vreg_64 = COPY [[COPY]]
     ; WAVE32: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
-    ; WAVE32: FLAT_STORE_DWORD [[COPY1]], [[DEF]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; WAVE32: GLOBAL_STORE_DWORD [[COPY1]], [[DEF]], 0, 0, 0, 0, implicit $exec :: (store 4, addrspace 1)
     %0:sgpr(p1) = COPY $sgpr2_sgpr3
     %1:vgpr(p1) = COPY %0
     %2:vgpr(s32) = G_IMPLICIT_DEF
@@ -46,7 +46,7 @@ body: |
     ; WAVE64: [[COPY3:%[0-9]+]]:sreg_32_xm0 = COPY $scc
     ; WAVE64: [[V_CMP_NE_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_NE_U32_e64 0, [[COPY3]], implicit $exec
     ; WAVE64: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, [[COPY2]], 0, [[COPY1]], [[V_CMP_NE_U32_e64_]], implicit $exec
-    ; WAVE64: FLAT_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; WAVE64: FLAT_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
     ; WAVE32-LABEL: name: copy_vcc_bank_scc_bank
     ; WAVE32: $vcc_hi = IMPLICIT_DEF
     ; WAVE32: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
@@ -55,7 +55,7 @@ body: |
     ; WAVE32: [[COPY3:%[0-9]+]]:sreg_32_xm0 = COPY $scc
     ; WAVE32: [[V_CMP_NE_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_NE_U32_e64 0, [[COPY3]], implicit $exec
     ; WAVE32: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, [[COPY2]], 0, [[COPY1]], [[V_CMP_NE_U32_e64_]], implicit $exec
-    ; WAVE32: FLAT_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; WAVE32: GLOBAL_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_]], 0, 0, 0, 0, implicit $exec :: (store 4, addrspace 1)
     %0:vgpr(p1) = COPY $vgpr0_vgpr1
     %1:vgpr(s32) = COPY $vgpr2
     %2:vgpr(s32) = COPY $vgpr3
@@ -83,7 +83,7 @@ body: |
     ; WAVE64: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, [[COPY2]], 0, [[COPY1]], [[V_CMP_NE_U32_e64_]], implicit $exec
     ; WAVE64: [[V_CMP_NE_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_NE_U32_e64 0, [[COPY3]], implicit $exec
     ; WAVE64: [[V_CNDMASK_B32_e64_1:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, [[V_CNDMASK_B32_e64_]], 0, [[COPY1]], [[V_CMP_NE_U32_e64_1]], implicit $exec
-    ; WAVE64: FLAT_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; WAVE64: FLAT_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
     ; WAVE32-LABEL: name: copy_vcc_bank_scc_bank_2_uses
     ; WAVE32: $vcc_hi = IMPLICIT_DEF
     ; WAVE32: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
@@ -94,7 +94,7 @@ body: |
     ; WAVE32: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, [[COPY2]], 0, [[COPY1]], [[COPY4]], implicit $exec
     ; WAVE32: [[V_CMP_NE_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_NE_U32_e64 0, [[COPY3]], implicit $exec
     ; WAVE32: [[V_CNDMASK_B32_e64_1:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, [[V_CNDMASK_B32_e64_]], 0, [[COPY1]], [[V_CMP_NE_U32_e64_]], implicit $exec
-    ; WAVE32: FLAT_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; WAVE32: GLOBAL_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_1]], 0, 0, 0, 0, implicit $exec :: (store 4, addrspace 1)
     %0:vgpr(p1) = COPY $vgpr0_vgpr1
     %1:vgpr(s32) = COPY $vgpr2
     %2:vgpr(s32) = COPY $vgpr3
@@ -122,7 +122,7 @@ body: |
     ; WAVE64: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; WAVE64: [[COPY3:%[0-9]+]]:sreg_64_xexec = COPY $scc
     ; WAVE64: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, [[COPY2]], 0, [[COPY1]], [[COPY3]], implicit $exec
-    ; WAVE64: FLAT_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; WAVE64: FLAT_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
     ; WAVE32-LABEL: name: copy_vcc_bank_scc_physreg
     ; WAVE32: $vcc_hi = IMPLICIT_DEF
     ; WAVE32: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
@@ -130,7 +130,7 @@ body: |
     ; WAVE32: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; WAVE32: [[COPY3:%[0-9]+]]:sreg_32_xm0_xexec = COPY $scc
     ; WAVE32: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, [[COPY2]], 0, [[COPY1]], [[COPY3]], implicit $exec
-    ; WAVE32: FLAT_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; WAVE32: GLOBAL_STORE_DWORD [[COPY]], [[V_CNDMASK_B32_e64_]], 0, 0, 0, 0, implicit $exec :: (store 4, addrspace 1)
     %0:vgpr(p1) = COPY $vgpr0_vgpr1
     %1:vgpr(s32) = COPY $vgpr2
     %2:vgpr(s32) = COPY $vgpr3
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-implicit-def.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-implicit-def.mir
index 43bd32644ff..5e14f7e5008 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-implicit-def.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-implicit-def.mir
@@ -104,7 +104,7 @@ body: |
     ; GCN-LABEL: name: implicit_def_p1_vgpr
     ; GCN: [[DEF:%[0-9]+]]:vreg_64 = IMPLICIT_DEF
     ; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4, implicit $exec
-    ; GCN: FLAT_STORE_DWORD [[DEF]], [[V_MOV_B32_e32_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GCN: FLAT_STORE_DWORD [[DEF]], [[V_MOV_B32_e32_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
     %0:vgpr(p1) = G_IMPLICIT_DEF
     %1:vgpr(s32) = G_CONSTANT i32 4
     G_STORE %1, %0 :: (store 4, addrspace 1)
@@ -119,9 +119,9 @@ regBankSelected: true
 body: |
   bb.0:
     ; GCN-LABEL: name: implicit_def_p3_vgpr
-    ; GCN: [[DEF:%[0-9]+]]:vgpr(p3) = G_IMPLICIT_DEF
-    ; GCN: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 4
-    ; GCN: G_STORE [[C]](s32), [[DEF]](p3) :: (store 4, addrspace 1)
+    ; GCN: [[DEF:%[0-9]+]]:vreg_64 = IMPLICIT_DEF
+    ; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4, implicit $exec
+    ; GCN: FLAT_STORE_DWORD [[DEF]], [[V_MOV_B32_e32_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
     %0:vgpr(p3) = G_IMPLICIT_DEF
     %1:vgpr(s32) = G_CONSTANT i32 4
     G_STORE %1, %0 :: (store 4, addrspace 1)
@@ -138,7 +138,7 @@ body: |
     ; GCN-LABEL: name: implicit_def_p4_vgpr
     ; GCN: [[DEF:%[0-9]+]]:vreg_64 = IMPLICIT_DEF
     ; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4, implicit $exec
-    ; GCN: FLAT_STORE_DWORD [[DEF]], [[V_MOV_B32_e32_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GCN: FLAT_STORE_DWORD [[DEF]], [[V_MOV_B32_e32_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
     %0:vgpr(p4) = G_IMPLICIT_DEF
     %1:vgpr(s32) = G_CONSTANT i32 4
     G_STORE %1, %0 :: (store 4, addrspace 1)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-flat.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-flat.mir
index eb8e39cd08d..f88d8ee615f 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-flat.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-flat.mir
@@ -1,42 +1,827 @@
-# 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: store_flat_s32_to_4
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_flat_s32_to_4
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
+    ; GFX8-LABEL: name: store_flat_s32_to_4
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
+    ; GFX9-LABEL: name: store_flat_s32_to_4
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
+    ; GFX10-LABEL: name: store_flat_s32_to_4
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = COPY $vgpr2
+    G_STORE %1, %0 :: (store 4, align 4, addrspace 0)
 
---- |
-  define amdgpu_kernel void @global_addrspace(i32 addrspace(1)* %global0,
-                                              i64 addrspace(1)* %global1,
-                                              i96 addrspace(1)* %global2,
-                                              i128 addrspace(1)* %global3) { ret void }
 ...
+
 ---
+name: store_flat_s32_to_2
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_flat_s32_to_2
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_SHORT [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 2)
+    ; GFX8-LABEL: name: store_flat_s32_to_2
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_SHORT [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 2)
+    ; GFX9-LABEL: name: store_flat_s32_to_2
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: FLAT_STORE_SHORT [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 2)
+    ; GFX10-LABEL: name: store_flat_s32_to_2
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: FLAT_STORE_SHORT [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 2)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = COPY $vgpr2
+    G_STORE %1, %0 :: (store 2, align 2, addrspace 0)
+
+...
 
-name:            global_addrspace
+---
+name: store_flat_s32_to_1
 legalized:       true
+tracksRegLiveness: true
 regBankSelected: true
 
-# GCN: global_addrspace
-# GCN: [[PTR:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
-# GCN: [[VAL4:%[0-9]+]]:vgpr_32 = COPY $vgpr2
-# GCN: [[VAL8:%[0-9]+]]:vreg_64 = COPY $vgpr3_vgpr4
-# GCN: [[VAL12:%[0-9]+]]:vreg_96 = COPY $vgpr5_vgpr6_vgpr7
-# GCN: [[VAL16:%[0-9]+]]:vreg_128 = COPY $vgpr8_vgpr9_vgpr10_vgpr11
-# GCN: FLAT_STORE_DWORD [[PTR]], [[VAL4]], 0, 0, 0
-# GCN: FLAT_STORE_DWORDX2 [[PTR]], [[VAL8]], 0, 0, 0
-# GCN: FLAT_STORE_DWORDX3 [[PTR]], [[VAL12]], 0, 0, 0
-# GCN: FLAT_STORE_DWORDX4 [[PTR]], [[VAL16]], 0, 0, 0
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_flat_s32_to_1
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_BYTE [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 1)
+    ; GFX8-LABEL: name: store_flat_s32_to_1
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_BYTE [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 1)
+    ; GFX9-LABEL: name: store_flat_s32_to_1
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: FLAT_STORE_BYTE [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 1)
+    ; GFX10-LABEL: name: store_flat_s32_to_1
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: FLAT_STORE_BYTE [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 1)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = COPY $vgpr2
+    G_STORE %1, %0 :: (store 1, align 1, addrspace 0)
+
+...
+
+---
+
+name: store_flat_s64
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
 
 body: |
   bb.0:
-    liveins:  $vgpr0_vgpr1, $vgpr2, $vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10_vgpr11
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
 
+    ; GFX7-LABEL: name: store_flat_s64
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_s64
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_s64
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_s64
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store 8, align 8, addrspace 0)
+
+...
+---
+
+name: store_flat_s96
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+
+    ; GFX7-LABEL: name: store_flat_s96
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX7: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_s96
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX8: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_s96
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX9: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_s96
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX10: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s96) = COPY $vgpr2_vgpr3_vgpr4
+    G_STORE %1, %0 :: (store 12, align 16, addrspace 0)
+
+...
+---
+
+name: store_flat_s128
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+
+    ; GFX7-LABEL: name: store_flat_s128
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_s128
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_s128
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_s128
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    G_STORE %1, %0 :: (store 16, align 16, addrspace 0)
+
+...
+
+---
+
+name: store_flat_v2s32
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_flat_v2s32
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8)
+    ; GFX8-LABEL: name: store_flat_v2s32
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8)
+    ; GFX9-LABEL: name: store_flat_v2s32
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8)
+    ; GFX10-LABEL: name: store_flat_v2s32
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s32>) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store 8, align 8, addrspace 0)
+
+...
+---
+
+name: store_flat_v3s32
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+
+    ; GFX7-LABEL: name: store_flat_v3s32
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX7: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 12, align 16)
+    ; GFX8-LABEL: name: store_flat_v3s32
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX8: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 12, align 16)
+    ; GFX9-LABEL: name: store_flat_v3s32
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX9: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 12, align 16)
+    ; GFX10-LABEL: name: store_flat_v3s32
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX10: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 12, align 16)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+    G_STORE %1, %0 :: (store 12, align 16, addrspace 0)
+
+...
+---
+
+name: store_flat_v4s32
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+
+    ; GFX7-LABEL: name: store_flat_v4s32
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 16)
+    ; GFX8-LABEL: name: store_flat_v4s32
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 16)
+    ; GFX9-LABEL: name: store_flat_v4s32
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 16)
+    ; GFX10-LABEL: name: store_flat_v4s32
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 16)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    G_STORE %1, %0 :: (store 16, align 16, addrspace 0)
+
+...
+
+---
+
+name: store_flat_v2s16
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_flat_v2s16
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_v2s16
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_v2s16
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_v2s16
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s16>) = COPY $vgpr2
+    G_STORE %1, %0 :: (store 4, align 4, addrspace 0)
+
+...
+
+---
+
+name: store_flat_v4s16
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_flat_v4s16
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_v4s16
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_v4s16
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_v4s16
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<4 x s16>) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store 8, align 8, addrspace 0)
+
+...
+
+---
+
+name: store_flat_v6s16
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+
+    ; GFX7-LABEL: name: store_flat_v6s16
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX7: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_v6s16
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX8: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_v6s16
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX9: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_v6s16
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX10: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<6 x s16>) = COPY $vgpr2_vgpr3_vgpr4
+    G_STORE %1, %0 :: (store 12, align 16, addrspace 0)
+
+...
+---
+
+name: store_flat_v8s16
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+
+    ; GFX7-LABEL: name: store_flat_v8s16
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_v8s16
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_v8s16
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_v8s16
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    G_STORE %1, %0 :: (store 16, align 16, addrspace 0)
+
+...
+
+---
+
+name: store_flat_v2s64
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+
+    ; GFX7-LABEL: name: store_flat_v2s64
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_v2s64
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_v2s64
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_v2s64
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    G_STORE %1, %0 :: (store 16, align 16, addrspace 0)
+
+...
+
+---
+
+name: store_flat_p1
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_flat_p1
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_p1
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_p1
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_p1
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(p1) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store 8, align 8, addrspace 0)
+
+...
+
+---
+
+name: store_flat_v2p1
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+
+    ; GFX7-LABEL: name: store_flat_v2p1
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_v2p1
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_v2p1
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_v2p1
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x p1>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    G_STORE %1, %0 :: (store 16, align 16, addrspace 0)
+
+...
+
+---
+
+name: store_flat_p3
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_flat_p3
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_p3
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_p3
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_p3
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(p3) = COPY $vgpr2
+    G_STORE %1, %0 :: (store 4, align 4, addrspace 0)
+
+...
+
+---
+
+name: store_flat_v2p3
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_flat_v2p3
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_flat_v2p3
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_flat_v2p3
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_flat_v2p3
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x p3>) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store 8, align 8, addrspace 0)
+
+...
+---
+
+name: store_atomic_flat_s32
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_atomic_flat_s32
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_atomic_flat_s32
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_atomic_flat_s32
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_atomic_flat_s32
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
     %0:vgpr(p1) = COPY $vgpr0_vgpr1
     %1:vgpr(s32) = COPY $vgpr2
-    %2:vgpr(s64) = COPY $vgpr3_vgpr4
-    %3:vgpr(s96) = COPY $vgpr5_vgpr6_vgpr7
-    %4:vgpr(s128) = COPY $vgpr8_vgpr9_vgpr10_vgpr11
-    G_STORE %1, %0 :: (store 4 into %ir.global0)
-    G_STORE %2, %0 :: (store 8 into %ir.global1)
-    G_STORE %3, %0 :: (store 12 into %ir.global2, align 16)
-    G_STORE %4, %0 :: (store 16 into %ir.global3)
+    G_STORE %1, %0 :: (store monotonic 4, align 4, addrspace 0)
 
 ...
+
 ---
+
+name: store_atomic_flat_s64
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_atomic_flat_s64
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_atomic_flat_s64
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_atomic_flat_s64
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_atomic_flat_s64
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store monotonic 8, align 8, addrspace 0)
+
+...
+
+---
+
+name: store_flat_s32_gep_2047
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_flat_s32_gep_2047
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; 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: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY5:%[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 [[COPY2]], [[COPY3]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY4]], [[COPY5]], 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_STORE_DWORD [[REG_SEQUENCE1]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
+    ; GFX8-LABEL: name: store_flat_s32_gep_2047
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; 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: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY5:%[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 [[COPY2]], [[COPY3]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY4]], [[COPY5]], 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_STORE_DWORD [[REG_SEQUENCE1]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
+    ; GFX9-LABEL: name: store_flat_s32_gep_2047
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 2047, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
+    ; GFX10-LABEL: name: store_flat_s32_gep_2047
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; 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: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX10: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX10: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX10: [[COPY5:%[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 [[COPY2]], [[COPY3]], 0, implicit $exec
+    ; GFX10: %9:vgpr_32, dead %11:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY4]], [[COPY5]], 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_STORE_DWORD [[REG_SEQUENCE1]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = COPY $vgpr2
+    %2:vgpr(s64) = G_CONSTANT i64 2047
+    %3:vgpr(p1) = G_GEP %0, %2
+    G_STORE %1, %3 :: (store 4, align 4, addrspace 0)
+
+...
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-global.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-global.mir
new file mode 100644
index 00000000000..2154d1cfee8
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-global.mir
@@ -0,0 +1,817 @@
+# 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: store_global_s32_to_4
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_global_s32_to_4
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
+    ; GFX8-LABEL: name: store_global_s32_to_4
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
+    ; GFX9-LABEL: name: store_global_s32_to_4
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: GLOBAL_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 4, addrspace 1)
+    ; GFX10-LABEL: name: store_global_s32_to_4
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: GLOBAL_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 4, addrspace 1)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = COPY $vgpr2
+    G_STORE %1, %0 :: (store 4, align 4, addrspace 1)
+
+...
+
+---
+name: store_global_s32_to_2
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_global_s32_to_2
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_SHORT [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 2, addrspace 1)
+    ; GFX8-LABEL: name: store_global_s32_to_2
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_SHORT [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 2, addrspace 1)
+    ; GFX9-LABEL: name: store_global_s32_to_2
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: GLOBAL_STORE_SHORT [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 2, addrspace 1)
+    ; GFX10-LABEL: name: store_global_s32_to_2
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: GLOBAL_STORE_SHORT [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 2, addrspace 1)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = COPY $vgpr2
+    G_STORE %1, %0 :: (store 2, align 2, addrspace 1)
+
+...
+
+---
+name: store_global_s32_to_1
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_global_s32_to_1
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_BYTE [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 1, addrspace 1)
+    ; GFX8-LABEL: name: store_global_s32_to_1
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_BYTE [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 1, addrspace 1)
+    ; GFX9-LABEL: name: store_global_s32_to_1
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: GLOBAL_STORE_BYTE [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 1, addrspace 1)
+    ; GFX10-LABEL: name: store_global_s32_to_1
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: GLOBAL_STORE_BYTE [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 1, addrspace 1)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = COPY $vgpr2
+    G_STORE %1, %0 :: (store 1, align 1, addrspace 1)
+
+...
+
+---
+
+name: store_global_s64
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_global_s64
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_s64
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_s64
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_s64
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store 8, align 8, addrspace 1)
+
+...
+---
+
+name: store_global_s96
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+
+    ; GFX7-LABEL: name: store_global_s96
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX7: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_s96
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX8: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_s96
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX9: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_s96
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX10: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s96) = COPY $vgpr2_vgpr3_vgpr4
+    G_STORE %1, %0 :: (store 12, align 16, addrspace 1)
+
+...
+---
+
+name: store_global_s128
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+
+    ; GFX7-LABEL: name: store_global_s128
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_s128
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_s128
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_s128
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    G_STORE %1, %0 :: (store 16, align 16, addrspace 1)
+
+...
+
+---
+
+name: store_global_v2s32
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_global_v2s32
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8, addrspace 1)
+    ; GFX8-LABEL: name: store_global_v2s32
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8, addrspace 1)
+    ; GFX9-LABEL: name: store_global_v2s32
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: GLOBAL_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 8, addrspace 1)
+    ; GFX10-LABEL: name: store_global_v2s32
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: GLOBAL_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 8, addrspace 1)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s32>) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store 8, align 8, addrspace 1)
+
+...
+---
+
+name: store_global_v3s32
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+
+    ; GFX7-LABEL: name: store_global_v3s32
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX7: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 12, align 16, addrspace 1)
+    ; GFX8-LABEL: name: store_global_v3s32
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX8: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 12, align 16, addrspace 1)
+    ; GFX9-LABEL: name: store_global_v3s32
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX9: GLOBAL_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 12, align 16, addrspace 1)
+    ; GFX10-LABEL: name: store_global_v3s32
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX10: GLOBAL_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 12, align 16, addrspace 1)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+    G_STORE %1, %0 :: (store 12, align 16, addrspace 1)
+
+...
+---
+
+name: store_global_v4s32
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+
+    ; GFX7-LABEL: name: store_global_v4s32
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 16, addrspace 1)
+    ; GFX8-LABEL: name: store_global_v4s32
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 16, addrspace 1)
+    ; GFX9-LABEL: name: store_global_v4s32
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: GLOBAL_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 16, addrspace 1)
+    ; GFX10-LABEL: name: store_global_v4s32
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: GLOBAL_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 16, addrspace 1)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    G_STORE %1, %0 :: (store 16, align 16, addrspace 1)
+
+...
+
+---
+
+name: store_global_v2s16
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_global_v2s16
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_v2s16
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_v2s16
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_v2s16
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s16>) = COPY $vgpr2
+    G_STORE %1, %0 :: (store 4, align 4, addrspace 1)
+
+...
+
+---
+
+name: store_global_v4s16
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_global_v4s16
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_v4s16
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_v4s16
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_v4s16
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<4 x s16>) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store 8, align 8, addrspace 1)
+
+...
+
+---
+
+name: store_global_v6s16
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+
+    ; GFX7-LABEL: name: store_global_v6s16
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX7: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_v6s16
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX8: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_v6s16
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX9: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_v6s16
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_96 = COPY $vgpr2_vgpr3_vgpr4
+    ; GFX10: FLAT_STORE_DWORDX3 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<6 x s16>) = COPY $vgpr2_vgpr3_vgpr4
+    G_STORE %1, %0 :: (store 12, align 16, addrspace 1)
+
+...
+---
+
+name: store_global_v8s16
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+
+    ; GFX7-LABEL: name: store_global_v8s16
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_v8s16
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_v8s16
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_v8s16
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    G_STORE %1, %0 :: (store 16, align 16, addrspace 1)
+
+...
+
+---
+
+name: store_global_v2s64
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+
+    ; GFX7-LABEL: name: store_global_v2s64
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_v2s64
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_v2s64
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_v2s64
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    G_STORE %1, %0 :: (store 16, align 16, addrspace 1)
+
+...
+
+---
+
+name: store_global_p1
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_global_p1
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_p1
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_p1
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_p1
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(p1) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store 8, align 8, addrspace 1)
+
+...
+
+---
+
+name: store_global_v2p1
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+
+    ; GFX7-LABEL: name: store_global_v2p1
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX7: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_v2p1
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX8: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_v2p1
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX9: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_v2p1
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX10: FLAT_STORE_DWORDX4 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x p1>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+    G_STORE %1, %0 :: (store 16, align 16, addrspace 1)
+
+...
+
+---
+
+name: store_global_p3
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_global_p3
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_p3
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_p3
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_p3
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(p3) = COPY $vgpr2
+    G_STORE %1, %0 :: (store 4, align 4, addrspace 1)
+
+...
+
+---
+
+name: store_global_v2p3
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_global_v2p3
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_global_v2p3
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_global_v2p3
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_global_v2p3
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(<2 x p3>) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store 8, align 8, addrspace 1)
+
+...
+---
+
+name: store_atomic_global_s32
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_atomic_global_s32
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX7: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_atomic_global_s32
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX8: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_atomic_global_s32
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_atomic_global_s32
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = COPY $vgpr2
+    G_STORE %1, %0 :: (store monotonic 4, align 4, addrspace 1)
+
+...
+
+---
+
+name: store_atomic_global_s64
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; GFX7-LABEL: name: store_atomic_global_s64
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX8-LABEL: name: store_atomic_global_s64
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX9-LABEL: name: store_atomic_global_s64
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    ; GFX10-LABEL: name: store_atomic_global_s64
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+    ; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s64) = COPY $vgpr2_vgpr3
+    G_STORE %1, %0 :: (store monotonic 8, align 8, addrspace 1)
+
+...
+
+---
+
+name: store_global_s32_gep_2047
+legalized:       true
+tracksRegLiveness: true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins:  $vgpr0_vgpr1, $vgpr2
+
+    ; GFX7-LABEL: name: store_global_s32_gep_2047
+    ; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; 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: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX7: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX7: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX7: [[COPY5:%[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 [[COPY2]], [[COPY3]], 0, implicit $exec
+    ; GFX7: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY4]], [[COPY5]], 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_STORE_DWORD [[REG_SEQUENCE1]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
+    ; GFX8-LABEL: name: store_global_s32_gep_2047
+    ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; 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: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0
+    ; GFX8: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0
+    ; GFX8: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1
+    ; GFX8: [[COPY5:%[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 [[COPY2]], [[COPY3]], 0, implicit $exec
+    ; GFX8: %9:vgpr_32, dead %11:sreg_64_xexec = V_ADDC_U32_e64 [[COPY4]], [[COPY5]], 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_STORE_DWORD [[REG_SEQUENCE1]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
+    ; GFX9-LABEL: name: store_global_s32_gep_2047
+    ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX9: GLOBAL_STORE_DWORD [[COPY]], [[COPY1]], 2047, 0, 0, 0, implicit $exec :: (store 4, addrspace 1)
+    ; GFX10-LABEL: name: store_global_s32_gep_2047
+    ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+    ; GFX10: $vcc_hi = IMPLICIT_DEF
+    ; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+    ; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GFX10: GLOBAL_STORE_DWORD [[COPY]], [[COPY1]], 2047, 0, 0, 0, implicit $exec :: (store 4, addrspace 1)
+    %0:vgpr(p1) = COPY $vgpr0_vgpr1
+    %1:vgpr(s32) = COPY $vgpr2
+    %2:vgpr(s64) = G_CONSTANT i64 2047
+    %3:vgpr(p1) = G_GEP %0, %2
+    G_STORE %1, %3 :: (store 4, align 4, addrspace 1)
+
+...
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i32.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i32.ll
index 8689b650b8f..f35b0b43d36 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i32.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i32.ll
@@ -12,15 +12,11 @@ define amdgpu_kernel void @test_wave32(i32 %arg0, [8 x i32], i32 %saved) {
 ; GCN-NEXT:    s_cbranch_scc0 BB0_2
 ; GCN-NEXT:  ; %bb.1: ; %mid
 ; GCN-NEXT:    v_mov_b32_e32 v0, 0
-; GCN-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GCN-NEXT:    s_waitcnt_vscnt null, 0x0
-; GCN-NEXT:    flat_store_dword v[0:1], v0
+; GCN-NEXT:    global_store_dword v[0:1], v0, off
 ; GCN-NEXT:  BB0_2: ; %bb
 ; GCN-NEXT:    s_or_b32 exec_lo, exec_lo, s0
 ; GCN-NEXT:    v_mov_b32_e32 v0, 0
-; GCN-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GCN-NEXT:    s_waitcnt_vscnt null, 0x0
-; GCN-NEXT:    flat_store_dword v[0:1], v0
+; GCN-NEXT:    global_store_dword v[0:1], v0, off
 ; GCN-NEXT:    s_endpgm
 entry:
   %cond = icmp eq i32 %arg0, 0
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i64.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i64.ll
index 9e19eefab3b..6172c9ceeab 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i64.ll
@@ -11,13 +11,11 @@ define amdgpu_kernel void @test_wave64(i32 %arg0, i64 %saved) {
 ; GCN-NEXT:    s_cbranch_scc0 BB0_2
 ; GCN-NEXT:  ; %bb.1: ; %mid
 ; GCN-NEXT:    v_mov_b32_e32 v0, 0
-; GCN-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GCN-NEXT:    flat_store_dword v[0:1], v0
+; GCN-NEXT:    global_store_dword v[0:1], v0, off
 ; GCN-NEXT:  BB0_2: ; %bb
 ; GCN-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GCN-NEXT:    v_mov_b32_e32 v0, 0
-; GCN-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GCN-NEXT:    flat_store_dword v[0:1], v0
+; GCN-NEXT:    global_store_dword v[0:1], v0, off
 ; GCN-NEXT:    s_endpgm
 entry:
   %cond = icmp eq i32 %arg0, 0
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.if.break.i32.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.if.break.i32.ll
index 282441a2a1d..0f259fcb895 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.if.break.i32.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.if.break.i32.ll
@@ -13,9 +13,7 @@ define amdgpu_kernel void @test_wave32(i32 %arg0, [8 x i32], i32 %saved) {
 ; GCN-NEXT:    v_cmp_ne_u32_e64 s0, 0, s0
 ; GCN-NEXT:    s_or_b32 s0, s0, s1
 ; GCN-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GCN-NEXT:    s_waitcnt_vscnt null, 0x0
-; GCN-NEXT:    flat_store_dword v[0:1], v0
+; GCN-NEXT:    global_store_dword v[0:1], v0, off
 ; GCN-NEXT:    s_endpgm
 entry:
   %cond = icmp eq i32 %arg0, 0