; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
; RUN: llc -mtriple=amdgcn -mcpu=gfx908 -stop-after=amdgpu-isel < %s | FileCheck -check-prefix=GFX908 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx90a -stop-after=amdgpu-isel < %s | FileCheck -check-prefix=GFX90A_GFX942 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -stop-after=amdgpu-isel < %s | FileCheck -check-prefix=GFX90A_GFX942 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx908 -enable-new-pm -stop-after=amdgpu-isel < %s | FileCheck -check-prefix=GFX908 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx90a -enable-new-pm -stop-after=amdgpu-isel < %s | FileCheck -check-prefix=GFX90A_GFX942 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -enable-new-pm -stop-after=amdgpu-isel < %s | FileCheck -check-prefix=GFX90A_GFX942 %s

define amdgpu_ps void @buffer_atomic_fadd_v2f16_offset_no_rtn(<2 x half> %val, <4 x i32> inreg %rsrc, i32 inreg %soffset) {
  ; GFX908-LABEL: name: buffer_atomic_fadd_v2f16_offset_no_rtn
  ; GFX908: bb.0 (%ir-block.0):
  ; GFX908-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $sgpr4
  ; GFX908-NEXT: {{  $}}
  ; GFX908-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX908-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX908-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX908-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY3]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY1]], %subreg.sub3
  ; GFX908-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_OFFSET [[COPY5]], killed [[REG_SEQUENCE]], [[COPY]], 4095, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 8)
  ; GFX908-NEXT:   S_ENDPGM 0
  ;
  ; GFX90A_GFX942-LABEL: name: buffer_atomic_fadd_v2f16_offset_no_rtn
  ; GFX90A_GFX942: bb.0 (%ir-block.0):
  ; GFX90A_GFX942-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $sgpr4
  ; GFX90A_GFX942-NEXT: {{  $}}
  ; GFX90A_GFX942-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX90A_GFX942-NEXT:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX90A_GFX942-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX90A_GFX942-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX90A_GFX942-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY3]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY1]], %subreg.sub3
  ; GFX90A_GFX942-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_OFFSET [[COPY5]], killed [[REG_SEQUENCE]], [[COPY]], 4095, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 8)
  ; GFX90A_GFX942-NEXT:   S_ENDPGM 0
  %ret = call <2 x half> @llvm.amdgcn.raw.buffer.atomic.fadd.v2f16(<2 x half> %val, <4 x i32> %rsrc, i32 4095, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @buffer_atomic_fadd_v2f16_offen_no_rtn(<2 x half> %val, <4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX908-LABEL: name: buffer_atomic_fadd_v2f16_offen_no_rtn
  ; GFX908: bb.0 (%ir-block.0):
  ; GFX908-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $sgpr4
  ; GFX908-NEXT: {{  $}}
  ; GFX908-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX908-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX908-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX908-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX908-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
  ; GFX908-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_OFFEN [[COPY6]], [[COPY1]], killed [[REG_SEQUENCE]], [[COPY]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 8)
  ; GFX908-NEXT:   S_ENDPGM 0
  ;
  ; GFX90A_GFX942-LABEL: name: buffer_atomic_fadd_v2f16_offen_no_rtn
  ; GFX90A_GFX942: bb.0 (%ir-block.0):
  ; GFX90A_GFX942-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $sgpr4
  ; GFX90A_GFX942-NEXT: {{  $}}
  ; GFX90A_GFX942-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX90A_GFX942-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX90A_GFX942-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX90A_GFX942-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX90A_GFX942-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
  ; GFX90A_GFX942-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_OFFEN [[COPY6]], [[COPY1]], killed [[REG_SEQUENCE]], [[COPY]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 8)
  ; GFX90A_GFX942-NEXT:   S_ENDPGM 0
  %ret = call <2 x half> @llvm.amdgcn.raw.buffer.atomic.fadd.v2f16(<2 x half> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @buffer_atomic_fadd_v2f16_idxen_no_rtn(<2 x half> %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 inreg %soffset) {
  ; GFX908-LABEL: name: buffer_atomic_fadd_v2f16_idxen_no_rtn
  ; GFX908: bb.0 (%ir-block.0):
  ; GFX908-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $sgpr4
  ; GFX908-NEXT: {{  $}}
  ; GFX908-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX908-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX908-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX908-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX908-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
  ; GFX908-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_IDXEN [[COPY6]], [[COPY1]], killed [[REG_SEQUENCE]], [[COPY]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 8)
  ; GFX908-NEXT:   S_ENDPGM 0
  ;
  ; GFX90A_GFX942-LABEL: name: buffer_atomic_fadd_v2f16_idxen_no_rtn
  ; GFX90A_GFX942: bb.0 (%ir-block.0):
  ; GFX90A_GFX942-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $sgpr4
  ; GFX90A_GFX942-NEXT: {{  $}}
  ; GFX90A_GFX942-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX90A_GFX942-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX90A_GFX942-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX90A_GFX942-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX90A_GFX942-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
  ; GFX90A_GFX942-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_IDXEN [[COPY6]], [[COPY1]], killed [[REG_SEQUENCE]], [[COPY]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 8)
  ; GFX90A_GFX942-NEXT:   S_ENDPGM 0
  %ret = call <2 x half> @llvm.amdgcn.struct.buffer.atomic.fadd.v2f16(<2 x half> %val, <4 x i32> %rsrc, i32 %vindex, i32 0, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @buffer_atomic_fadd_v2f16_bothen_no_rtn(<2 x half> %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
  ; GFX908-LABEL: name: buffer_atomic_fadd_v2f16_bothen_no_rtn
  ; GFX908: bb.0 (%ir-block.0):
  ; GFX908-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $vgpr2, $sgpr4
  ; GFX908-NEXT: {{  $}}
  ; GFX908-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX908-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX908-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX908-NEXT:   [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX908-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX908-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY4]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; GFX908-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1
  ; GFX908-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_BOTHEN [[COPY7]], killed [[REG_SEQUENCE1]], killed [[REG_SEQUENCE]], [[COPY]], 0, 2, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 8)
  ; GFX908-NEXT:   S_ENDPGM 0
  ;
  ; GFX90A_GFX942-LABEL: name: buffer_atomic_fadd_v2f16_bothen_no_rtn
  ; GFX90A_GFX942: bb.0 (%ir-block.0):
  ; GFX90A_GFX942-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $vgpr2, $sgpr4
  ; GFX90A_GFX942-NEXT: {{  $}}
  ; GFX90A_GFX942-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX90A_GFX942-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX90A_GFX942-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX90A_GFX942-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX90A_GFX942-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX90A_GFX942-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY4]], %subreg.sub2, [[COPY3]], %subreg.sub3
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64_align2 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1
  ; GFX90A_GFX942-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_BOTHEN [[COPY7]], killed [[REG_SEQUENCE1]], killed [[REG_SEQUENCE]], [[COPY]], 0, 2, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 8)
  ; GFX90A_GFX942-NEXT:   S_ENDPGM 0
  %ret = call <2 x half> @llvm.amdgcn.struct.buffer.atomic.fadd.v2f16(<2 x half> %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 2)
  ret void
}

define amdgpu_ps void @buffer_ptr_atomic_fadd_v2f16_offset_no_rtn(<2 x half> %val, ptr addrspace(8) inreg %rsrc, i32 inreg %soffset) {
  ; GFX908-LABEL: name: buffer_ptr_atomic_fadd_v2f16_offset_no_rtn
  ; GFX908: bb.0 (%ir-block.0):
  ; GFX908-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $sgpr4
  ; GFX908-NEXT: {{  $}}
  ; GFX908-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX908-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX908-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX908-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1
  ; GFX908-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub1
  ; GFX908-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub0
  ; GFX908-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY3]], %subreg.sub1
  ; GFX908-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub1
  ; GFX908-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub0
  ; GFX908-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sgpr_128 = REG_SEQUENCE killed [[COPY9]], %subreg.sub0, killed [[COPY8]], %subreg.sub1, killed [[COPY7]], %subreg.sub2, killed [[COPY6]], %subreg.sub3
  ; GFX908-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_OFFSET [[COPY5]], killed [[REG_SEQUENCE2]], [[COPY]], 4095, 0, implicit $exec :: (volatile dereferenceable load store (s32) on %ir.rsrc, align 1, addrspace 8)
  ; GFX908-NEXT:   S_ENDPGM 0
  ;
  ; GFX90A_GFX942-LABEL: name: buffer_ptr_atomic_fadd_v2f16_offset_no_rtn
  ; GFX90A_GFX942: bb.0 (%ir-block.0):
  ; GFX90A_GFX942-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $sgpr4
  ; GFX90A_GFX942-NEXT: {{  $}}
  ; GFX90A_GFX942-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX90A_GFX942-NEXT:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX90A_GFX942-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX90A_GFX942-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX90A_GFX942-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1
  ; GFX90A_GFX942-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub1
  ; GFX90A_GFX942-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY3]], %subreg.sub1
  ; GFX90A_GFX942-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub1
  ; GFX90A_GFX942-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sgpr_128 = REG_SEQUENCE killed [[COPY9]], %subreg.sub0, killed [[COPY8]], %subreg.sub1, killed [[COPY7]], %subreg.sub2, killed [[COPY6]], %subreg.sub3
  ; GFX90A_GFX942-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_OFFSET [[COPY5]], killed [[REG_SEQUENCE2]], [[COPY]], 4095, 0, implicit $exec :: (volatile dereferenceable load store (s32) on %ir.rsrc, align 1, addrspace 8)
  ; GFX90A_GFX942-NEXT:   S_ENDPGM 0
  %ret = call <2 x half> @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 4095, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @buffer_ptr_atomic_fadd_v2f16_offen_no_rtn(<2 x half> %val, ptr addrspace(8) inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
  ; GFX908-LABEL: name: buffer_ptr_atomic_fadd_v2f16_offen_no_rtn
  ; GFX908: bb.0 (%ir-block.0):
  ; GFX908-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $sgpr4
  ; GFX908-NEXT: {{  $}}
  ; GFX908-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX908-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX908-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX908-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX908-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
  ; GFX908-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub1
  ; GFX908-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub0
  ; GFX908-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1
  ; GFX908-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub1
  ; GFX908-NEXT:   [[COPY10:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub0
  ; GFX908-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sgpr_128 = REG_SEQUENCE killed [[COPY10]], %subreg.sub0, killed [[COPY9]], %subreg.sub1, killed [[COPY8]], %subreg.sub2, killed [[COPY7]], %subreg.sub3
  ; GFX908-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_OFFEN [[COPY6]], [[COPY1]], killed [[REG_SEQUENCE2]], [[COPY]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32) on %ir.rsrc, align 1, addrspace 8)
  ; GFX908-NEXT:   S_ENDPGM 0
  ;
  ; GFX90A_GFX942-LABEL: name: buffer_ptr_atomic_fadd_v2f16_offen_no_rtn
  ; GFX90A_GFX942: bb.0 (%ir-block.0):
  ; GFX90A_GFX942-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $sgpr4
  ; GFX90A_GFX942-NEXT: {{  $}}
  ; GFX90A_GFX942-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX90A_GFX942-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX90A_GFX942-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX90A_GFX942-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX90A_GFX942-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
  ; GFX90A_GFX942-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub1
  ; GFX90A_GFX942-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1
  ; GFX90A_GFX942-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub1
  ; GFX90A_GFX942-NEXT:   [[COPY10:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sgpr_128 = REG_SEQUENCE killed [[COPY10]], %subreg.sub0, killed [[COPY9]], %subreg.sub1, killed [[COPY8]], %subreg.sub2, killed [[COPY7]], %subreg.sub3
  ; GFX90A_GFX942-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_OFFEN [[COPY6]], [[COPY1]], killed [[REG_SEQUENCE2]], [[COPY]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32) on %ir.rsrc, align 1, addrspace 8)
  ; GFX90A_GFX942-NEXT:   S_ENDPGM 0
  %ret = call <2 x half> @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %voffset, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @buffer_ptr_atomic_fadd_v2f16_idxen_no_rtn(<2 x half> %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 inreg %soffset) {
  ; GFX908-LABEL: name: buffer_ptr_atomic_fadd_v2f16_idxen_no_rtn
  ; GFX908: bb.0 (%ir-block.0):
  ; GFX908-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $sgpr4
  ; GFX908-NEXT: {{  $}}
  ; GFX908-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX908-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX908-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX908-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX908-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
  ; GFX908-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub1
  ; GFX908-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub0
  ; GFX908-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1
  ; GFX908-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub1
  ; GFX908-NEXT:   [[COPY10:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub0
  ; GFX908-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sgpr_128 = REG_SEQUENCE killed [[COPY10]], %subreg.sub0, killed [[COPY9]], %subreg.sub1, killed [[COPY8]], %subreg.sub2, killed [[COPY7]], %subreg.sub3
  ; GFX908-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_IDXEN [[COPY6]], [[COPY1]], killed [[REG_SEQUENCE2]], [[COPY]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32) on %ir.rsrc, align 1, addrspace 8)
  ; GFX908-NEXT:   S_ENDPGM 0
  ;
  ; GFX90A_GFX942-LABEL: name: buffer_ptr_atomic_fadd_v2f16_idxen_no_rtn
  ; GFX90A_GFX942: bb.0 (%ir-block.0):
  ; GFX90A_GFX942-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $sgpr4
  ; GFX90A_GFX942-NEXT: {{  $}}
  ; GFX90A_GFX942-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX90A_GFX942-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX90A_GFX942-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX90A_GFX942-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX90A_GFX942-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
  ; GFX90A_GFX942-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub1
  ; GFX90A_GFX942-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1
  ; GFX90A_GFX942-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub1
  ; GFX90A_GFX942-NEXT:   [[COPY10:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sgpr_128 = REG_SEQUENCE killed [[COPY10]], %subreg.sub0, killed [[COPY9]], %subreg.sub1, killed [[COPY8]], %subreg.sub2, killed [[COPY7]], %subreg.sub3
  ; GFX90A_GFX942-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_IDXEN [[COPY6]], [[COPY1]], killed [[REG_SEQUENCE2]], [[COPY]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32) on %ir.rsrc, align 1, addrspace 8)
  ; GFX90A_GFX942-NEXT:   S_ENDPGM 0
  %ret = call <2 x half> @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 0, i32 %soffset, i32 0)
  ret void
}

define amdgpu_ps void @buffer_ptr_atomic_fadd_v2f16_bothen_no_rtn(<2 x half> %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
  ; GFX908-LABEL: name: buffer_ptr_atomic_fadd_v2f16_bothen_no_rtn
  ; GFX908: bb.0 (%ir-block.0):
  ; GFX908-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $vgpr2, $sgpr4
  ; GFX908-NEXT: {{  $}}
  ; GFX908-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX908-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX908-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX908-NEXT:   [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX908-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX908-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY3]], %subreg.sub1
  ; GFX908-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub1
  ; GFX908-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub0
  ; GFX908-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY5]], %subreg.sub1
  ; GFX908-NEXT:   [[COPY10:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub1
  ; GFX908-NEXT:   [[COPY11:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub0
  ; GFX908-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sgpr_128 = REG_SEQUENCE killed [[COPY11]], %subreg.sub0, killed [[COPY10]], %subreg.sub1, killed [[COPY9]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
  ; GFX908-NEXT:   [[REG_SEQUENCE3:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1
  ; GFX908-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_BOTHEN [[COPY7]], killed [[REG_SEQUENCE3]], killed [[REG_SEQUENCE2]], [[COPY]], 0, 2, implicit $exec :: (volatile dereferenceable load store (s32) on %ir.rsrc, align 1, addrspace 8)
  ; GFX908-NEXT:   S_ENDPGM 0
  ;
  ; GFX90A_GFX942-LABEL: name: buffer_ptr_atomic_fadd_v2f16_bothen_no_rtn
  ; GFX90A_GFX942: bb.0 (%ir-block.0):
  ; GFX90A_GFX942-NEXT:   liveins: $vgpr0, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr1, $vgpr2, $sgpr4
  ; GFX90A_GFX942-NEXT: {{  $}}
  ; GFX90A_GFX942-NEXT:   [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr4
  ; GFX90A_GFX942-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
  ; GFX90A_GFX942-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
  ; GFX90A_GFX942-NEXT:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr2
  ; GFX90A_GFX942-NEXT:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr1
  ; GFX90A_GFX942-NEXT:   [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr0
  ; GFX90A_GFX942-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY3]], %subreg.sub1
  ; GFX90A_GFX942-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub1
  ; GFX90A_GFX942-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE]].sub0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY5]], %subreg.sub1
  ; GFX90A_GFX942-NEXT:   [[COPY10:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub1
  ; GFX90A_GFX942-NEXT:   [[COPY11:%[0-9]+]]:sreg_32 = COPY [[REG_SEQUENCE1]].sub0
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sgpr_128 = REG_SEQUENCE killed [[COPY11]], %subreg.sub0, killed [[COPY10]], %subreg.sub1, killed [[COPY9]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
  ; GFX90A_GFX942-NEXT:   [[REG_SEQUENCE3:%[0-9]+]]:vreg_64_align2 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1
  ; GFX90A_GFX942-NEXT:   BUFFER_ATOMIC_PK_ADD_F16_BOTHEN [[COPY7]], killed [[REG_SEQUENCE3]], killed [[REG_SEQUENCE2]], [[COPY]], 0, 2, implicit $exec :: (volatile dereferenceable load store (s32) on %ir.rsrc, align 1, addrspace 8)
  ; GFX90A_GFX942-NEXT:   S_ENDPGM 0
  %ret = call <2 x half> @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 2)
  ret void
}

declare <2 x half> @llvm.amdgcn.raw.buffer.atomic.fadd.v2f16(<2 x half>, <4 x i32>, i32, i32, i32 immarg)
declare <2 x half> @llvm.amdgcn.struct.buffer.atomic.fadd.v2f16(<2 x half>, <4 x i32>, i32, i32, i32, i32 immarg)

declare <2 x half> @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.v2f16(<2 x half>, ptr addrspace(8), i32, i32, i32 immarg)
declare <2 x half> @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.v2f16(<2 x half>, ptr addrspace(8), i32, i32, i32, i32 immarg)