xref: /llvm-project/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.ptr.buffer.store.format.f16.ll (revision c3cfbbc4160c3e0284034c98b332b468328458e3)

; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=tonga -stop-after=instruction-select -o - %s | FileCheck -check-prefix=UNPACKED %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx810 -stop-after=instruction-select -o - %s | FileCheck -check-prefix=PACKED %s

define amdgpu_ps void @raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_f16(ptr addrspace(8) inreg %rsrc, half %val, i32 %voffset, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_f16
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_X_gfx80_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 0, 0, implicit $exec :: (dereferenceable store (s16) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_f16
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_X_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 0, 0, implicit $exec :: (dereferenceable store (s16) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.f16(half %val, ptr addrspace(8) %rsrc, i32 %voffset, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__voffset_4095__sgpr_soffset_f16(ptr addrspace(8) inreg %rsrc, half %val, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__voffset_4095__sgpr_soffset_f16
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_X_gfx80_OFFSET_exact [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 4095, 0, 0, implicit $exec :: (dereferenceable store (s16) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__voffset_4095__sgpr_soffset_f16
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_X_OFFSET_exact [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 4095, 0, 0, implicit $exec :: (dereferenceable store (s16) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.f16(half %val, ptr addrspace(8) %rsrc, i32 4095, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16(ptr addrspace(8) inreg %rsrc, <2 x half> %val, i32 %voffset, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; UNPACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY7]], [[COPY4]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[V_LSHRREV_B32_e64_]], %subreg.sub1
  ; UNPACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_gfx80_OFFEN_exact [[REG_SEQUENCE]], [[COPY5]], [[REG_SEQUENCE1]], [[COPY6]], 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %voffset, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16(ptr addrspace(8) inreg %rsrc, <4 x half> %val, i32 %voffset, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; UNPACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY8]], [[COPY4]], implicit $exec
  ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_1:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY9]], [[COPY5]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[V_LSHRREV_B32_e64_]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[V_LSHRREV_B32_e64_1]], %subreg.sub3
  ; UNPACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XYZW_gfx80_OFFEN_exact [[REG_SEQUENCE]], [[COPY6]], [[REG_SEQUENCE1]], [[COPY7]], 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1
  ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; PACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XYZW_OFFEN_exact [[REG_SEQUENCE]], [[COPY6]], [[REG_SEQUENCE1]], [[COPY7]], 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.v4f16(<4 x half> %val, ptr addrspace(8) %rsrc, i32 %voffset, i32 %soffset, i32 0)
  ret void
}

; Make sure unpack code is emitted outside of loop
define amdgpu_ps void @raw_ptr_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16(ptr addrspace(8) %rsrc, <4 x half> %val, i32 %voffset, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   successors: %bb.2(0x80000000)
  ; UNPACKED-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr6
  ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY8]], [[COPY4]], implicit $exec
  ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_1:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY9]], [[COPY5]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[V_LSHRREV_B32_e64_]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[V_LSHRREV_B32_e64_1]], %subreg.sub3
  ; UNPACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT: bb.2:
  ; UNPACKED-NEXT:   successors: %bb.3(0x80000000)
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY]], implicit $exec
  ; UNPACKED-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; UNPACKED-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; UNPACKED-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE2:%[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
  ; UNPACKED-NEXT:   [[COPY10:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; UNPACKED-NEXT:   [[COPY11:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; UNPACKED-NEXT:   [[COPY12:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE2]].sub0_sub1
  ; UNPACKED-NEXT:   [[COPY13:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE2]].sub2_sub3
  ; UNPACKED-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[COPY12]], [[COPY10]], implicit $exec
  ; UNPACKED-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[COPY13]], [[COPY11]], implicit $exec
  ; UNPACKED-NEXT:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; UNPACKED-NEXT:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT: bb.3:
  ; UNPACKED-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XYZW_gfx80_OFFEN_exact [[REG_SEQUENCE]], [[COPY6]], [[REG_SEQUENCE2]], [[COPY7]], 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
  ; UNPACKED-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT: bb.4:
  ; UNPACKED-NEXT:   successors: %bb.5(0x80000000)
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   $exec = S_MOV_B64_term [[S_MOV_B64_]]
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT: bb.5:
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   successors: %bb.2(0x80000000)
  ; PACKED-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1
  ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr6
  ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT: bb.2:
  ; PACKED-NEXT:   successors: %bb.3(0x80000000)
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY]], implicit $exec
  ; PACKED-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; PACKED-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; PACKED-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; PACKED-NEXT:   [[REG_SEQUENCE2:%[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
  ; PACKED-NEXT:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; PACKED-NEXT:   [[COPY9:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; PACKED-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE2]].sub0_sub1
  ; PACKED-NEXT:   [[COPY11:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE2]].sub2_sub3
  ; PACKED-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
  ; PACKED-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[COPY11]], [[COPY9]], implicit $exec
  ; PACKED-NEXT:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; PACKED-NEXT:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT: bb.3:
  ; PACKED-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XYZW_OFFEN_exact [[REG_SEQUENCE]], [[COPY6]], [[REG_SEQUENCE2]], [[COPY7]], 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
  ; PACKED-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT: bb.4:
  ; PACKED-NEXT:   successors: %bb.5(0x80000000)
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   $exec = S_MOV_B64_term [[S_MOV_B64_]]
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT: bb.5:
  ; PACKED-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.v4f16(<4 x half> %val, ptr addrspace(8) %rsrc, i32 %voffset, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_soffset4095(ptr addrspace(8) inreg %rsrc, <2 x half> %val, i32 %voffset, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_soffset4095
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; UNPACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4095
  ; UNPACKED-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_1]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY6]], [[COPY4]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[V_LSHRREV_B32_e64_]], %subreg.sub1
  ; UNPACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_gfx80_OFFEN_exact [[REG_SEQUENCE]], [[COPY5]], [[REG_SEQUENCE1]], [[S_MOV_B32_]], 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_soffset4095
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; PACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4095
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %voffset, i32 4095, i32 0)
  ret void
}

define amdgpu_ps void @raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_soffset4096(ptr addrspace(8) inreg %rsrc, <2 x half> %val, i32 %voffset, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_soffset4096
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; UNPACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; UNPACKED-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_1]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY6]], [[COPY4]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[V_LSHRREV_B32_e64_]], %subreg.sub1
  ; UNPACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_gfx80_OFFEN_exact [[REG_SEQUENCE]], [[COPY5]], [[REG_SEQUENCE1]], [[S_MOV_B32_]], 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_soffset4096
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; PACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   S_ENDPGM 0
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %voffset, i32 4096, i32 0)
  ret void
}

define amdgpu_ps void @raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_voffset_add_16(ptr addrspace(8) inreg %rsrc, <2 x half> %val, i32 %voffset, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_voffset_add_16
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; UNPACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY7]], [[COPY4]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[V_LSHRREV_B32_e64_]], %subreg.sub1
  ; UNPACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_gfx80_OFFEN_exact [[REG_SEQUENCE]], [[COPY5]], [[REG_SEQUENCE1]], [[COPY6]], 16, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_voffset_add_16
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 16, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   S_ENDPGM 0
  %voffset.add = add i32 %voffset, 16
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_voffset_add_4095(ptr addrspace(8) inreg %rsrc, <2 x half> %val, i32 %voffset, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_voffset_add_4095
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; UNPACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY7]], [[COPY4]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[V_LSHRREV_B32_e64_]], %subreg.sub1
  ; UNPACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_gfx80_OFFEN_exact [[REG_SEQUENCE]], [[COPY5]], [[REG_SEQUENCE1]], [[COPY6]], 4095, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_voffset_add_4095
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 4095, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   S_ENDPGM 0
  %voffset.add = add i32 %voffset, 4095
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_voffset_add_4096(ptr addrspace(8) inreg %rsrc, <2 x half> %val, i32 %voffset, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_voffset_add_4096
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; UNPACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; UNPACKED-NEXT:   [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64 = V_ADD_CO_U32_e64 [[COPY5]], [[COPY7]], 0, implicit $exec
  ; UNPACKED-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_1]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY8]], [[COPY4]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[V_LSHRREV_B32_e64_]], %subreg.sub1
  ; UNPACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_gfx80_OFFEN_exact [[REG_SEQUENCE]], [[V_ADD_CO_U32_e64_]], [[REG_SEQUENCE1]], [[COPY6]], 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f16_voffset_add_4096
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
  ; PACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; PACKED-NEXT:   [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64 = V_ADD_CO_U32_e64 [[COPY5]], [[COPY7]], 0, implicit $exec
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XY_OFFEN_exact [[COPY4]], [[V_ADD_CO_U32_e64_]], [[REG_SEQUENCE]], [[COPY6]], 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   S_ENDPGM 0
  %voffset.add = add i32 %voffset, 4096
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
  ret void
}


; Check what happens with offset add inside a waterfall loop
define amdgpu_ps void @raw_ptr_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16_add_4096(ptr addrspace(8) %rsrc, <4 x half> %val, i32 %voffset, i32 inreg %soffset) {
  ; UNPACKED-LABEL: name: raw_ptr_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16_add_4096
  ; UNPACKED: bb.1 (%ir-block.0):
  ; UNPACKED-NEXT:   successors: %bb.2(0x80000000)
  ; UNPACKED-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr6
  ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; UNPACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; UNPACKED-NEXT:   [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64 = V_ADD_CO_U32_e64 [[COPY6]], [[COPY8]], 0, implicit $exec
  ; UNPACKED-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 16
  ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_1]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY9]], [[COPY4]], implicit $exec
  ; UNPACKED-NEXT:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_1]]
  ; UNPACKED-NEXT:   [[V_LSHRREV_B32_e64_1:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 [[COPY10]], [[COPY5]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[V_LSHRREV_B32_e64_]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[V_LSHRREV_B32_e64_1]], %subreg.sub3
  ; UNPACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; UNPACKED-NEXT:   [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT: bb.2:
  ; UNPACKED-NEXT:   successors: %bb.3(0x80000000)
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY]], implicit $exec
  ; UNPACKED-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; UNPACKED-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; UNPACKED-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; UNPACKED-NEXT:   [[REG_SEQUENCE2:%[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
  ; UNPACKED-NEXT:   [[COPY11:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; UNPACKED-NEXT:   [[COPY12:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; UNPACKED-NEXT:   [[COPY13:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE2]].sub0_sub1
  ; UNPACKED-NEXT:   [[COPY14:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE2]].sub2_sub3
  ; UNPACKED-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[COPY13]], [[COPY11]], implicit $exec
  ; UNPACKED-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[COPY14]], [[COPY12]], implicit $exec
  ; UNPACKED-NEXT:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; UNPACKED-NEXT:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT: bb.3:
  ; UNPACKED-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XYZW_gfx80_OFFEN_exact [[REG_SEQUENCE]], [[V_ADD_CO_U32_e64_]], [[REG_SEQUENCE2]], [[COPY7]], 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; UNPACKED-NEXT:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
  ; UNPACKED-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT: bb.4:
  ; UNPACKED-NEXT:   successors: %bb.5(0x80000000)
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT:   $exec = S_MOV_B64_term [[S_MOV_B64_]]
  ; UNPACKED-NEXT: {{  $}}
  ; UNPACKED-NEXT: bb.5:
  ; UNPACKED-NEXT:   S_ENDPGM 0
  ;
  ; PACKED-LABEL: name: raw_ptr_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16_add_4096
  ; PACKED: bb.1 (%ir-block.0):
  ; PACKED-NEXT:   successors: %bb.2(0x80000000)
  ; PACKED-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
  ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
  ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
  ; PACKED-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1
  ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr6
  ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr2
  ; PACKED-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
  ; PACKED-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
  ; PACKED-NEXT:   [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64 = V_ADD_CO_U32_e64 [[COPY6]], [[COPY8]], 0, implicit $exec
  ; PACKED-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; PACKED-NEXT:   [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT: bb.2:
  ; PACKED-NEXT:   successors: %bb.3(0x80000000)
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY]], implicit $exec
  ; PACKED-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
  ; PACKED-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
  ; PACKED-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
  ; PACKED-NEXT:   [[REG_SEQUENCE2:%[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
  ; PACKED-NEXT:   [[COPY9:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
  ; PACKED-NEXT:   [[COPY10:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
  ; PACKED-NEXT:   [[COPY11:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE2]].sub0_sub1
  ; PACKED-NEXT:   [[COPY12:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE2]].sub2_sub3
  ; PACKED-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[COPY11]], [[COPY9]], implicit $exec
  ; PACKED-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[COPY12]], [[COPY10]], implicit $exec
  ; PACKED-NEXT:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
  ; PACKED-NEXT:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT: bb.3:
  ; PACKED-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   BUFFER_STORE_FORMAT_D16_XYZW_OFFEN_exact [[REG_SEQUENCE]], [[V_ADD_CO_U32_e64_]], [[REG_SEQUENCE2]], [[COPY7]], 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>) into %ir.rsrc, align 1, addrspace 8)
  ; PACKED-NEXT:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
  ; PACKED-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT: bb.4:
  ; PACKED-NEXT:   successors: %bb.5(0x80000000)
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT:   $exec = S_MOV_B64_term [[S_MOV_B64_]]
  ; PACKED-NEXT: {{  $}}
  ; PACKED-NEXT: bb.5:
  ; PACKED-NEXT:   S_ENDPGM 0
  %voffset.add = add i32 %voffset, 4096
  call void @llvm.amdgcn.raw.ptr.buffer.store.format.v4f16(<4 x half> %val, ptr addrspace(8) %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
  ret void
}

declare void @llvm.amdgcn.raw.ptr.buffer.store.format.f16(half, ptr addrspace(8), i32, i32, i32 immarg)
declare void @llvm.amdgcn.raw.ptr.buffer.store.format.v2f16(<2 x half>, ptr addrspace(8), i32, i32, i32 immarg)
declare void @llvm.amdgcn.raw.ptr.buffer.store.format.v4f16(<4 x half>, ptr addrspace(8), i32, i32, i32 immarg)