; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -o - %s | FileCheck -check-prefix=GFX9 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -o - %s | FileCheck -check-prefix=GFX10 %s

define amdgpu_ps float @atomic_swap_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_swap_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_swap v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_swap_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_swap v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_add_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_add_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.add.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_sub_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_sub_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_sub v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_sub_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_sub v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.sub.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_smin_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_smin_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_smin v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_smin_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_smin v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.smin.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_umin_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_umin_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_umin v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_umin_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_umin v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.umin.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_smax_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_smax_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_smax v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_smax_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_smax v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.smax.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_umax_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_umax_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_umax v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_umax_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_umax v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.umax.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_and_i321d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_and_i321d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_and v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_and_i321d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_and v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.and.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_or_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_or_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_or v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_or_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_or v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.or.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_xor_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_xor_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_xor v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_xor_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_xor v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.xor.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_inc_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_inc_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_inc v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_inc_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_inc v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.inc.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_dec_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_dec_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_dec v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_dec_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_dec v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.dec.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_cmpswap_i32_1d(<8 x i32> inreg %rsrc, i32 %cmp, i32 %swap, i16 %s) {
; GFX9-LABEL: atomic_cmpswap_i32_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_cmpswap v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_cmpswap_i32_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_cmpswap v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.cmpswap.1d.i32.i16(i32 %cmp, i32 %swap, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_add_i32_2d(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t) {
; GFX9-LABEL: atomic_add_i32_2d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    v_mov_b32_e32 v3, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v1, v1, v3, v2
; GFX9-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i32_2d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    v_and_or_b32 v1, v1, 0xffff, v2
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.add.2d.i32.i16(i32 %data, i16 %s, i16 %t, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_add_i32_3d(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t, i16 %r) {
; GFX9-LABEL: atomic_add_i32_3d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX9-NEXT:    s_lshl_b32 s8, s0, 16
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    v_and_or_b32 v1, v1, v4, v2
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v2, v3, v4, s8
; GFX9-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i32_3d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_mov_b32_e32 v4, 0xffff
; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_lshl_b32 s8, s0, 16
; GFX10-NEXT:    v_and_or_b32 v1, v1, v4, v2
; GFX10-NEXT:    v_and_or_b32 v2, v3, v4, s8
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_3D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.add.3d.i32.i16(i32 %data, i16 %s, i16 %t, i16 %r, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_add_i32_cube(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t, i16 %face) {
; GFX9-LABEL: atomic_add_i32_cube:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX9-NEXT:    s_lshl_b32 s8, s0, 16
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    v_and_or_b32 v1, v1, v4, v2
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v2, v3, v4, s8
; GFX9-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 unorm glc a16 da
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i32_cube:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_mov_b32_e32 v4, 0xffff
; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_lshl_b32 s8, s0, 16
; GFX10-NEXT:    v_and_or_b32 v1, v1, v4, v2
; GFX10-NEXT:    v_and_or_b32 v2, v3, v4, s8
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_CUBE unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.add.cube.i32.i16(i32 %data, i16 %s, i16 %t, i16 %face, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_add_i32_1darray(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %slice) {
; GFX9-LABEL: atomic_add_i32_1darray:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    v_mov_b32_e32 v3, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v1, v1, v3, v2
; GFX9-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 unorm glc a16 da
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i32_1darray:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    v_and_or_b32 v1, v1, 0xffff, v2
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D_ARRAY unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.add.1darray.i32.i16(i32 %data, i16 %s, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_add_i32_2darray(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t, i16 %slice) {
; GFX9-LABEL: atomic_add_i32_2darray:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX9-NEXT:    s_lshl_b32 s8, s0, 16
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    v_and_or_b32 v1, v1, v4, v2
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v2, v3, v4, s8
; GFX9-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 unorm glc a16 da
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i32_2darray:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_mov_b32_e32 v4, 0xffff
; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_lshl_b32 s8, s0, 16
; GFX10-NEXT:    v_and_or_b32 v1, v1, v4, v2
; GFX10-NEXT:    v_and_or_b32 v2, v3, v4, s8
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_ARRAY unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.add.2darray.i32.i16(i32 %data, i16 %s, i16 %t, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_add_i32_2dmsaa(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t, i16 %fragid) {
; GFX9-LABEL: atomic_add_i32_2dmsaa:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX9-NEXT:    s_lshl_b32 s8, s0, 16
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    v_and_or_b32 v1, v1, v4, v2
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v2, v3, v4, s8
; GFX9-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i32_2dmsaa:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_mov_b32_e32 v4, 0xffff
; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_lshl_b32 s8, s0, 16
; GFX10-NEXT:    v_and_or_b32 v1, v1, v4, v2
; GFX10-NEXT:    v_and_or_b32 v2, v3, v4, s8
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_MSAA unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.add.2dmsaa.i32.i16(i32 %data, i16 %s, i16 %t, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_add_i32_2darraymsaa(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t, i16 %slice, i16 %fragid) {
; GFX9-LABEL: atomic_add_i32_2darraymsaa:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    v_mov_b32_e32 v5, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX9-NEXT:    v_and_or_b32 v1, v1, v5, v2
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v4
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v2, v3, v5, v2
; GFX9-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 unorm glc a16 da
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i32_2darraymsaa:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_mov_b32_e32 v5, 0xffff
; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 16, v4
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    v_and_or_b32 v1, v1, v5, v2
; GFX10-NEXT:    v_and_or_b32 v2, v3, v5, v4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.add.2darraymsaa.i32.i16(i32 %data, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps float @atomic_add_i32_1d_slc(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX9-LABEL: atomic_add_i32_1d_slc:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 unorm glc slc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i32_1d_slc:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc slc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i32 @llvm.amdgcn.image.atomic.add.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 2)
  %out = bitcast i32 %v to float
  ret float %out
}

define amdgpu_ps <2 x float> @atomic_swap_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_swap_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_swap v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_swap_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_swap v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.swap.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_add_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_add_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.add.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_sub_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_sub_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_sub v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_sub_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_sub v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.sub.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_smin_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_smin_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_smin v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_smin_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_smin v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.smin.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_umin_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_umin_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_umin v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_umin_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_umin v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.umin.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_smax_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_smax_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_smax v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_smax_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_smax v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.smax.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_umax_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_umax_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_umax v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_umax_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_umax v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.umax.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_and_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_and_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_and v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_and_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_and v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.and.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_or_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_or_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_or v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_or_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_or v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.or.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_xor_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_xor_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_xor v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_xor_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_xor v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.xor.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_inc_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_inc_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_inc v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_inc_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_inc v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.inc.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_dec_i64_1d(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_dec_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_dec v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_dec_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_dec v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.dec.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_cmpswap_i64_1d(<8 x i32> inreg %rsrc, i64 %cmp, i64 %swap, i16 %s) {
; GFX9-LABEL: atomic_cmpswap_i64_1d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_cmpswap v[0:3], v4, s[0:7] dmask:0xf unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_cmpswap_i64_1d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_cmpswap v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.cmpswap.1d.i64.i16(i64 %cmp, i64 %swap, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_add_i64_2d(<8 x i32> inreg %rsrc, i64 %data, i16 %s, i16 %t) {
; GFX9-LABEL: atomic_add_i64_2d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v2, v2, v4, v3
; GFX9-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i64_2d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    v_and_or_b32 v2, v2, 0xffff, v3
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.add.2d.i64.i16(i64 %data, i16 %s, i16 %t, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_add_i64_3d(<8 x i32> inreg %rsrc, i64 %data, i16 %s, i16 %t, i16 %r) {
; GFX9-LABEL: atomic_add_i64_3d:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    v_mov_b32_e32 v5, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX9-NEXT:    s_lshl_b32 s8, s0, 16
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    v_and_or_b32 v2, v2, v5, v3
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v3, v4, v5, s8
; GFX9-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i64_3d:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_mov_b32_e32 v5, 0xffff
; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_lshl_b32 s8, s0, 16
; GFX10-NEXT:    v_and_or_b32 v2, v2, v5, v3
; GFX10-NEXT:    v_and_or_b32 v3, v4, v5, s8
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_3D unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.add.3d.i64.i16(i64 %data, i16 %s, i16 %t, i16 %r, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_add_i64_cube(<8 x i32> inreg %rsrc, i64 %data, i16 %s, i16 %t, i16 %face) {
; GFX9-LABEL: atomic_add_i64_cube:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    v_mov_b32_e32 v5, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX9-NEXT:    s_lshl_b32 s8, s0, 16
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    v_and_or_b32 v2, v2, v5, v3
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v3, v4, v5, s8
; GFX9-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 unorm glc a16 da
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i64_cube:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_mov_b32_e32 v5, 0xffff
; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_lshl_b32 s8, s0, 16
; GFX10-NEXT:    v_and_or_b32 v2, v2, v5, v3
; GFX10-NEXT:    v_and_or_b32 v3, v4, v5, s8
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_CUBE unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.add.cube.i64.i16(i64 %data, i16 %s, i16 %t, i16 %face , <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_add_i64_1darray(<8 x i32> inreg %rsrc, i64 %data, i16 %s, i16 %slice) {
; GFX9-LABEL: atomic_add_i64_1darray:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v2, v2, v4, v3
; GFX9-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 unorm glc a16 da
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i64_1darray:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    v_and_or_b32 v2, v2, 0xffff, v3
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D_ARRAY unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.add.1darray.i64.i16(i64 %data, i16 %s, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_add_i64_2darray(<8 x i32> inreg %rsrc, i64 %data, i16 %s, i16 %t, i16 %slice) {
; GFX9-LABEL: atomic_add_i64_2darray:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    v_mov_b32_e32 v5, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX9-NEXT:    s_lshl_b32 s8, s0, 16
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    v_and_or_b32 v2, v2, v5, v3
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v3, v4, v5, s8
; GFX9-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 unorm glc a16 da
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i64_2darray:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_mov_b32_e32 v5, 0xffff
; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_lshl_b32 s8, s0, 16
; GFX10-NEXT:    v_and_or_b32 v2, v2, v5, v3
; GFX10-NEXT:    v_and_or_b32 v3, v4, v5, s8
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D_ARRAY unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.add.2darray.i64.i16(i64 %data, i16 %s, i16 %t, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_add_i64_2dmsaa(<8 x i32> inreg %rsrc, i64 %data, i16 %s, i16 %t, i16 %fragid) {
; GFX9-LABEL: atomic_add_i64_2dmsaa:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    v_mov_b32_e32 v5, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX9-NEXT:    s_lshl_b32 s8, s0, 16
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    v_and_or_b32 v2, v2, v5, v3
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v3, v4, v5, s8
; GFX9-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 unorm glc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i64_2dmsaa:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_mov_b32_e32 v5, 0xffff
; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_lshl_b32 s8, s0, 16
; GFX10-NEXT:    v_and_or_b32 v2, v2, v5, v3
; GFX10-NEXT:    v_and_or_b32 v3, v4, v5, s8
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.add.2dmsaa.i64.i16(i64 %data, i16 %s, i16 %t, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_add_i64_2darraymsaa(<8 x i32> inreg %rsrc, i64 %data, i16 %s, i16 %t, i16 %slice, i16 %fragid) {
; GFX9-LABEL: atomic_add_i64_2darraymsaa:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    v_mov_b32_e32 v6, 0xffff
; GFX9-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX9-NEXT:    v_and_or_b32 v2, v2, v6, v3
; GFX9-NEXT:    v_lshlrev_b32_e32 v3, 16, v5
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    v_and_or_b32 v3, v4, v6, v3
; GFX9-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 unorm glc a16 da
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i64_2darraymsaa:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    v_mov_b32_e32 v6, 0xffff
; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT:    v_lshlrev_b32_e32 v5, 16, v5
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    v_and_or_b32 v2, v2, v6, v3
; GFX10-NEXT:    v_and_or_b32 v3, v4, v6, v5
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm glc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.add.2darraymsaa.i64.i16(i64 %data, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

define amdgpu_ps <2 x float> @atomic_add_i64_1d_slc(<8 x i32> inreg %rsrc, i64 %data, i16 %s) {
; GFX9-LABEL: atomic_add_i64_1d_slc:
; GFX9:       ; %bb.0: ; %main_body
; GFX9-NEXT:    s_mov_b32 s0, s2
; GFX9-NEXT:    s_mov_b32 s1, s3
; GFX9-NEXT:    s_mov_b32 s2, s4
; GFX9-NEXT:    s_mov_b32 s3, s5
; GFX9-NEXT:    s_mov_b32 s4, s6
; GFX9-NEXT:    s_mov_b32 s5, s7
; GFX9-NEXT:    s_mov_b32 s6, s8
; GFX9-NEXT:    s_mov_b32 s7, s9
; GFX9-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 unorm glc slc a16
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: atomic_add_i64_1d_slc:
; GFX10:       ; %bb.0: ; %main_body
; GFX10-NEXT:    s_mov_b32 s0, s2
; GFX10-NEXT:    s_mov_b32 s1, s3
; GFX10-NEXT:    s_mov_b32 s2, s4
; GFX10-NEXT:    s_mov_b32 s3, s5
; GFX10-NEXT:    s_mov_b32 s4, s6
; GFX10-NEXT:    s_mov_b32 s5, s7
; GFX10-NEXT:    s_mov_b32 s6, s8
; GFX10-NEXT:    s_mov_b32 s7, s9
; GFX10-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc slc a16
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    ; return to shader part epilog
main_body:
  %v = call i64 @llvm.amdgcn.image.atomic.add.1d.i64.i16(i64 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 2)
  %out = bitcast i64 %v to <2 x float>
  ret <2 x float> %out
}

declare i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.add.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.sub.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.smin.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.umin.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.smax.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.umax.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.and.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.or.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.xor.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.inc.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.dec.1d.i32.i16(i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.cmpswap.1d.i32.i16(i32, i32, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.add.2d.i32.i16(i32, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.add.3d.i32.i16(i32, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.add.cube.i32.i16(i32, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.add.1darray.i32.i16(i32, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.add.2darray.i32.i16(i32, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.add.2dmsaa.i32.i16(i32, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i32 @llvm.amdgcn.image.atomic.add.2darraymsaa.i32.i16(i32, i16, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0

declare i64 @llvm.amdgcn.image.atomic.swap.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.add.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.sub.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.smin.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.umin.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.smax.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.umax.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.and.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.or.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.xor.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.inc.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.dec.1d.i64.i16(i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.cmpswap.1d.i64.i16(i64, i64, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.add.2d.i64.i16(i64, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.add.3d.i64.i16(i64, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.add.cube.i64.i16(i64, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.add.1darray.i64.i16(i64, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.add.2darray.i64.i16(i64, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.add.2dmsaa.i64.i16(i64, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
declare i64 @llvm.amdgcn.image.atomic.add.2darraymsaa.i64.i16(i64, i16, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0

attributes #0 = { nounwind }