llvm/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.tbuffer.store.ll

; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -stop-after=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX10_GFX11 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -stop-after=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX10_GFX11 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -stop-after=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX12 %s

; Natural mapping
define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset(float %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 78, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 78, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
  ret void
}

; Natural mapping
define amdgpu_ps void @raw_tbuffer_store_v2f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset(<2 x float> %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_v2f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY3]], %subreg.sub1, [[COPY4]], %subreg.sub2, [[COPY5]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX10_GFX11-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_XY_OFFEN_exact [[REG_SEQUENCE]], [[COPY6]], [[REG_SEQUENCE1]], [[COPY7]], 0, 78, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_v2f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY3]], %subreg.sub1, [[COPY4]], %subreg.sub2, [[COPY5]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_XY_VBUFFER_OFFEN_exact [[REG_SEQUENCE]], [[COPY6]], [[REG_SEQUENCE1]], [[COPY7]], 0, 78, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.v2f32(<2 x float> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
  ret void
}

; Natural mapping
define amdgpu_ps void @raw_tbuffer_store_v3f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset(<3 x float> %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_v3f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY6]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX10_GFX11-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_XYZ_OFFEN_exact [[REG_SEQUENCE]], [[COPY7]], [[REG_SEQUENCE1]], [[COPY8]], 0, 78, 0, 0, implicit $exec :: (dereferenceable store (<3 x s32>), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_v3f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY6]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_XYZ_VBUFFER_OFFEN_exact [[REG_SEQUENCE]], [[COPY7]], [[REG_SEQUENCE1]], [[COPY8]], 0, 78, 0, 0, implicit $exec :: (dereferenceable store (<3 x s32>), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.v3f32(<3 x float> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
  ret void
}

; Natural mapping
define amdgpu_ps void @raw_tbuffer_store_v4f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset(<4 x float> %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_v4f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY6]], %subreg.sub2, [[COPY7]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX10_GFX11-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_XYZW_OFFEN_exact [[REG_SEQUENCE]], [[COPY8]], [[REG_SEQUENCE1]], [[COPY9]], 0, 78, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_v4f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY6]], %subreg.sub2, [[COPY7]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_XYZW_VBUFFER_OFFEN_exact [[REG_SEQUENCE]], [[COPY8]], [[REG_SEQUENCE1]], [[COPY9]], 0, 78, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.v4f32(<4 x float> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
  ret void
}

; Copies for VGPR arguments
define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__sgpr_voffset__sgpr_soffset(float %val, <4 x i32> inreg %rsrc, i32 inreg %voffset, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__sgpr_voffset__sgpr_soffset
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $vgpr0
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr7
  ; GFX10_GFX11-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]]
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE]], [[COPY6]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__sgpr_voffset__sgpr_soffset
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $vgpr0
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr7
  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]]
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE]], [[COPY6]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
  ret void
}

; Waterfall for rsrc
define amdgpu_ps void @raw_tbuffer_store_f32__vgpr_rsrc__vgpr_voffset__sgpr_soffset(float %val, <4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__vgpr_rsrc__vgpr_voffset__sgpr_soffset
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   successors: %bb.2(0x80000000)
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.2:
  ; GFX10_GFX11-NEXT:   successors: %bb.3(0x80000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
  ; GFX10_GFX11-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
  ; GFX10_GFX11-NEXT:   [[COPY9:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; GFX10_GFX11-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY9]], [[COPY7]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.3:
  ; GFX10_GFX11-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE1]], [[COPY6]], 0, 94, 1, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
  ; GFX10_GFX11-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.4:
  ; GFX10_GFX11-NEXT:   successors: %bb.5(0x80000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_]]
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.5:
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__vgpr_rsrc__vgpr_voffset__sgpr_soffset
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   successors: %bb.2(0x80000000)
  ; GFX12-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.2:
  ; GFX12-NEXT:   successors: %bb.3(0x80000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
  ; GFX12-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; GFX12-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY9]], [[COPY7]], implicit $exec
  ; GFX12-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; GFX12-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; GFX12-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.3:
  ; GFX12-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE1]], [[COPY6]], 0, 94, 1, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
  ; GFX12-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.4:
  ; GFX12-NEXT:   successors: %bb.5(0x80000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_]]
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.5:
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 1)
  ret void
}

; Waterfall for rsrc and soffset
define amdgpu_ps void @raw_tbuffer_store_f32__vgpr_rsrc__vgpr_voffset__vgpr_soffset(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__vgpr_rsrc__vgpr_voffset__vgpr_soffset
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   successors: %bb.2(0x80000000)
  ; GFX10_GFX11-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr6
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.2:
  ; GFX10_GFX11-NEXT:   successors: %bb.3(0x80000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
  ; GFX10_GFX11-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
  ; GFX10_GFX11-NEXT:   [[COPY9:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; GFX10_GFX11-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY9]], [[COPY7]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[S_AND_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[S_AND_B32_]], [[V_CMP_EQ_U32_e64_]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_1]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.3:
  ; GFX10_GFX11-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE1]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
  ; GFX10_GFX11-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.4:
  ; GFX10_GFX11-NEXT:   successors: %bb.5(0x80000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_]]
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.5:
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__vgpr_rsrc__vgpr_voffset__vgpr_soffset
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   successors: %bb.2(0x80000000)
  ; GFX12-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr6
  ; GFX12-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.2:
  ; GFX12-NEXT:   successors: %bb.3(0x80000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
  ; GFX12-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; GFX12-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY9]], [[COPY7]], implicit $exec
  ; GFX12-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; GFX12-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
  ; GFX12-NEXT:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
  ; GFX12-NEXT:   [[S_AND_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[S_AND_B32_]], [[V_CMP_EQ_U32_e64_]], implicit-def dead $scc
  ; GFX12-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_1]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.3:
  ; GFX12-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE1]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
  ; GFX12-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.4:
  ; GFX12-NEXT:   successors: %bb.5(0x80000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_]]
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.5:
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
  ret void
}

; Waterfall for rsrc and soffset, copy for voffset
define amdgpu_ps void @raw_tbuffer_store_f32__vgpr_rsrc__sgpr_voffset__vgpr_soffset(float %val, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__vgpr_rsrc__sgpr_voffset__vgpr_soffset
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   successors: %bb.2(0x80000000)
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; GFX10_GFX11-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]]
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.2:
  ; GFX10_GFX11-NEXT:   successors: %bb.3(0x80000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
  ; GFX10_GFX11-NEXT:   [[COPY9:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
  ; GFX10_GFX11-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; GFX10_GFX11-NEXT:   [[COPY11:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY11]], [[COPY9]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[S_AND_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[S_AND_B32_]], [[V_CMP_EQ_U32_e64_]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_1]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.3:
  ; GFX10_GFX11-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE1]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
  ; GFX10_GFX11-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.4:
  ; GFX10_GFX11-NEXT:   successors: %bb.5(0x80000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_]]
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.5:
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__vgpr_rsrc__sgpr_voffset__vgpr_soffset
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   successors: %bb.2(0x80000000)
  ; GFX12-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]]
  ; GFX12-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.2:
  ; GFX12-NEXT:   successors: %bb.3(0x80000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
  ; GFX12-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; GFX12-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; GFX12-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY11]], [[COPY9]], implicit $exec
  ; GFX12-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
  ; GFX12-NEXT:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
  ; GFX12-NEXT:   [[S_AND_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[S_AND_B32_]], [[V_CMP_EQ_U32_e64_]], implicit-def dead $scc
  ; GFX12-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_1]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.3:
  ; GFX12-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE1]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
  ; GFX12-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.4:
  ; GFX12-NEXT:   successors: %bb.5(0x80000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_]]
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.5:
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
  ret void
}

; Natural mapping + glc
define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_glc(float %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_glc
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 78, 1, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_glc
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 78, 1, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 1)
  ret void
}

; Natural mapping + slc
define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc(float %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 78, 2, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 78, 2, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 2)
  ret void
}

; Natural mapping + glc + slc
define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc_glc(float %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc_glc
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 78, 3, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc_glc
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 78, 3, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 3)
  ret void
}

; Natural mapping + dlc
define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_dlc(float %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_dlc
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 78, 4, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_dlc
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 78, 4, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 4)
  ret void
}



define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vdpr_voffset__sgpr_soffset__voffset0(float %val, <4 x i32> inreg %rsrc, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vdpr_voffset__sgpr_soffset__voffset0
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY]], [[REG_SEQUENCE]], [[COPY5]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vdpr_voffset__sgpr_soffset__voffset0
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFSET_exact [[COPY]], [[REG_SEQUENCE]], [[COPY5]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 0, i32 %soffset, i32 94, i32 0)
  ret void
}

define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset4095(float %val, <4 x i32> inreg %rsrc, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset4095
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY]], [[REG_SEQUENCE]], [[COPY5]], 4095, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset4095
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFSET_exact [[COPY]], [[REG_SEQUENCE]], [[COPY5]], 4095, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 4095, i32 %soffset, i32 94, i32 0)
  ret void
}

define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset4096(float %val, <4 x i32> inreg %rsrc, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset4096
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY6]], [[REG_SEQUENCE]], [[COPY5]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset4096
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFSET_exact [[COPY]], [[REG_SEQUENCE]], [[COPY5]], 4096, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 4096, i32 %soffset, i32 94, i32 0)
  ret void
}

define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_voffset_add16(float %val, <4 x i32> inreg %rsrc, i32 %voffset.base, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_voffset_add16
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 16, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_voffset_add16
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 16, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  %voffset = add i32 %voffset.base, 16
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
  ret void
}

define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset_add4095(float %val, <4 x i32> inreg %rsrc, i32 %voffset.base, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset_add4095
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 4095, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset_add4095
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 4095, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  %voffset = add i32 %voffset.base, 4095
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
  ret void
}

define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset_add4096(float %val, <4 x i32> inreg %rsrc, i32 %voffset.base, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset_add4096
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; GFX10_GFX11-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; GFX10_GFX11-NEXT:   [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[COPY5]], [[COPY7]], 0, implicit $exec
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[V_ADD_U32_e64_]], [[REG_SEQUENCE]], [[COPY6]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset__voffset_add4096
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 4096, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  %voffset = add i32 %voffset.base, 4096
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
  ret void
}

define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset4095(float %val, <4 x i32> inreg %rsrc, i32 %voffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset4095
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4095
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset4095
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4095
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 4095, i32 94, i32 0)
  ret void
}

define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset4096(float %val, <4 x i32> inreg %rsrc, i32 %voffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset4096
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset4096
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 4096, i32 94, i32 0)
  ret void
}

define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add16(float %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset.base) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add16
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; GFX10_GFX11-NEXT:   [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[COPY6]], [[S_MOV_B32_]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[S_ADD_I32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add16
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; GFX12-NEXT:   [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[COPY6]], [[S_MOV_B32_]], implicit-def dead $scc
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[S_ADD_I32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  %soffset = add i32 %soffset.base, 16
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
  ret void
}

define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add4095(float %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset.base) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add4095
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4095
  ; GFX10_GFX11-NEXT:   [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[COPY6]], [[S_MOV_B32_]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[S_ADD_I32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add4095
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4095
  ; GFX12-NEXT:   [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[COPY6]], [[S_MOV_B32_]], implicit-def dead $scc
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[S_ADD_I32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  %soffset = add i32 %soffset.base, 4095
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
  ret void
}

define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add4096(float %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset.base) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add4096
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; GFX10_GFX11-NEXT:   [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[COPY6]], [[S_MOV_B32_]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[S_ADD_I32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add4096
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; GFX12-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; GFX12-NEXT:   [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[COPY6]], [[S_MOV_B32_]], implicit-def dead $scc
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE]], [[S_ADD_I32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   S_ENDPGM 0
  %soffset = add i32 %soffset.base, 4096
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
  ret void
}

; An add of the offset is necessary, with a waterfall loop. Make sure the add is done outside of the waterfall loop.
define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add5000(float %val, <4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset.base) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add5000
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   successors: %bb.2(0x80000000)
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 5000
  ; GFX10_GFX11-NEXT:   [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[COPY6]], [[S_MOV_B32_]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.2:
  ; GFX10_GFX11-NEXT:   successors: %bb.3(0x80000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
  ; GFX10_GFX11-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
  ; GFX10_GFX11-NEXT:   [[COPY9:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; GFX10_GFX11-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY9]], [[COPY7]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.3:
  ; GFX10_GFX11-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE1]], [[S_ADD_I32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
  ; GFX10_GFX11-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.4:
  ; GFX10_GFX11-NEXT:   successors: %bb.5(0x80000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_1]]
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.5:
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add5000
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   successors: %bb.2(0x80000000)
  ; GFX12-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 5000
  ; GFX12-NEXT:   [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[COPY6]], [[S_MOV_B32_]], implicit-def dead $scc
  ; GFX12-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.2:
  ; GFX12-NEXT:   successors: %bb.3(0x80000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
  ; GFX12-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; GFX12-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY9]], [[COPY7]], implicit $exec
  ; GFX12-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; GFX12-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; GFX12-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.3:
  ; GFX12-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE1]], [[S_ADD_I32_]], 0, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
  ; GFX12-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.4:
  ; GFX12-NEXT:   successors: %bb.5(0x80000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_1]]
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.5:
  ; GFX12-NEXT:   S_ENDPGM 0
  %soffset = add i32 %soffset.base, 5000
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
  ret void
}

; An add of the offset is necessary, with a waterfall loop. Make sure the add is done outside of the waterfall loop.
define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_voffset_add5000(float %val, <4 x i32> %rsrc, i32 %voffset.base, i32 inreg %soffset) {
  ; GFX10_GFX11-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_voffset_add5000
  ; GFX10_GFX11: bb.1 (%ir-block.0):
  ; GFX10_GFX11-NEXT:   successors: %bb.2(0x80000000)
  ; GFX10_GFX11-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX10_GFX11-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX10_GFX11-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX10_GFX11-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX10_GFX11-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; GFX10_GFX11-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; GFX10_GFX11-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; GFX10_GFX11-NEXT:   [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[COPY5]], [[COPY7]], 0, implicit $exec
  ; GFX10_GFX11-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.2:
  ; GFX10_GFX11-NEXT:   successors: %bb.3(0x80000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
  ; GFX10_GFX11-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
  ; GFX10_GFX11-NEXT:   [[COPY9:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
  ; GFX10_GFX11-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; GFX10_GFX11-NEXT:   [[COPY11:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY11]], [[COPY9]], implicit $exec
  ; GFX10_GFX11-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; GFX10_GFX11-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.3:
  ; GFX10_GFX11-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[V_ADD_U32_e64_]], [[REG_SEQUENCE1]], [[COPY6]], 904, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX10_GFX11-NEXT:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
  ; GFX10_GFX11-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.4:
  ; GFX10_GFX11-NEXT:   successors: %bb.5(0x80000000)
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_1]]
  ; GFX10_GFX11-NEXT: {{  $}}
  ; GFX10_GFX11-NEXT: bb.5:
  ; GFX10_GFX11-NEXT:   S_ENDPGM 0
  ;
  ; GFX12-LABEL: name: raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_voffset_add5000
  ; GFX12: bb.1 (%ir-block.0):
  ; GFX12-NEXT:   successors: %bb.2(0x80000000)
  ; GFX12-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; GFX12-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; GFX12-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.2:
  ; GFX12-NEXT:   successors: %bb.3(0x80000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; GFX12-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
  ; GFX12-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; GFX12-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY9]], [[COPY7]], implicit $exec
  ; GFX12-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; GFX12-NEXT:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; GFX12-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.3:
  ; GFX12-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   TBUFFER_STORE_FORMAT_X_VBUFFER_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE1]], [[COPY6]], 5000, 94, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 8)
  ; GFX12-NEXT:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
  ; GFX12-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.4:
  ; GFX12-NEXT:   successors: %bb.5(0x80000000)
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_]]
  ; GFX12-NEXT: {{  $}}
  ; GFX12-NEXT: bb.5:
  ; GFX12-NEXT:   S_ENDPGM 0
  %voffset = add i32 %voffset.base, 5000
  call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
  ret void
}

declare void @llvm.amdgcn.raw.tbuffer.store.f32(float, <4 x i32>, i32, i32, i32 immarg, i32 immarg)
declare void @llvm.amdgcn.raw.tbuffer.store.v2f32(<2 x float>, <4 x i32>, i32, i32, i32 immarg, i32 immarg)
declare void @llvm.amdgcn.raw.tbuffer.store.v3f32(<3 x float>, <4 x i32>, i32, i32, i32 immarg, i32 immarg)
declare void @llvm.amdgcn.raw.tbuffer.store.v4f32(<4 x float>, <4 x i32>, i32, i32, i32 immarg, i32 immarg)