; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; FIXME: Test 90a, 940. 908 should fail to select.
; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 < %s | FileCheck -check-prefix=GFX12 %s

define <2 x bfloat> @raw_ptr_buffer_atomic_add_v2bf16_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset_add__sgpr_soffset(<2 x bfloat> %val, ptr addrspace(8) inreg %rsrc, i32 %voffset, i32 inreg %soffset) #0 {
; GFX12-LABEL: raw_ptr_buffer_atomic_add_v2bf16_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset_add__sgpr_soffset:
; GFX12:       ; %bb.0:
; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
; GFX12-NEXT:    s_wait_expcnt 0x0
; GFX12-NEXT:    s_wait_samplecnt 0x0
; GFX12-NEXT:    s_wait_bvhcnt 0x0
; GFX12-NEXT:    s_wait_kmcnt 0x0
; GFX12-NEXT:    buffer_atomic_pk_add_bf16 v0, v1, s[0:3], s16 offen offset:128 th:TH_ATOMIC_RETURN
; GFX12-NEXT:    s_wait_loadcnt 0x0
; GFX12-NEXT:    s_setpc_b64 s[30:31]
  %voffset.add = add i32 %voffset, 128
  %ret = call <2 x bfloat> @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.v2bf16(<2 x bfloat> %val, ptr addrspace(8) %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
  ret <2 x bfloat> %ret
}

define <2 x bfloat> @raw_ptr_buffer_atomic_add_v2bf16_rtn__vgpr_val__sgpr_rsrc__0_voffset__sgpr_soffset__slc(<2 x bfloat> %val, ptr addrspace(8) inreg %rsrc, i32 %voffset, i32 inreg %soffset) #0 {
; GFX12-LABEL: raw_ptr_buffer_atomic_add_v2bf16_rtn__vgpr_val__sgpr_rsrc__0_voffset__sgpr_soffset__slc:
; GFX12:       ; %bb.0:
; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
; GFX12-NEXT:    s_wait_expcnt 0x0
; GFX12-NEXT:    s_wait_samplecnt 0x0
; GFX12-NEXT:    s_wait_bvhcnt 0x0
; GFX12-NEXT:    s_wait_kmcnt 0x0
; GFX12-NEXT:    buffer_atomic_pk_add_bf16 v0, off, s[0:3], s16 offset:92 th:TH_ATOMIC_NT_RETURN
; GFX12-NEXT:    s_wait_loadcnt 0x0
; GFX12-NEXT:    s_setpc_b64 s[30:31]
  %ret = call <2 x bfloat> @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.v2bf16(<2 x bfloat> %val, ptr addrspace(8) %rsrc, i32 92, i32 %soffset, i32 2)
  ret <2 x bfloat> %ret
}

define void @raw_ptr_buffer_atomic_add_v2bf16_noret__vgpr_val__sgpr_rsrc__vgpr_voffset_add__sgpr_soffset(<2 x bfloat> %val, ptr addrspace(8) inreg %rsrc, i32 %voffset, i32 inreg %soffset) #0 {
; GFX12-LABEL: raw_ptr_buffer_atomic_add_v2bf16_noret__vgpr_val__sgpr_rsrc__vgpr_voffset_add__sgpr_soffset:
; GFX12:       ; %bb.0:
; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
; GFX12-NEXT:    s_wait_expcnt 0x0
; GFX12-NEXT:    s_wait_samplecnt 0x0
; GFX12-NEXT:    s_wait_bvhcnt 0x0
; GFX12-NEXT:    s_wait_kmcnt 0x0
; GFX12-NEXT:    buffer_atomic_pk_add_bf16 v0, v1, s[0:3], s16 offen offset:128
; GFX12-NEXT:    s_setpc_b64 s[30:31]
  %voffset.add = add i32 %voffset, 128
  %unused = call <2 x bfloat> @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.v2bf16(<2 x bfloat> %val, ptr addrspace(8) %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
  ret void
}

define void @raw_ptr_buffer_atomic_add_v2bf16_noret__vgpr_val__sgpr_rsrc__0_voffset__sgpr_soffset__slc(<2 x bfloat> %val, ptr addrspace(8) inreg %rsrc, i32 %voffset, i32 inreg %soffset) #0 {
; GFX12-LABEL: raw_ptr_buffer_atomic_add_v2bf16_noret__vgpr_val__sgpr_rsrc__0_voffset__sgpr_soffset__slc:
; GFX12:       ; %bb.0:
; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
; GFX12-NEXT:    s_wait_expcnt 0x0
; GFX12-NEXT:    s_wait_samplecnt 0x0
; GFX12-NEXT:    s_wait_bvhcnt 0x0
; GFX12-NEXT:    s_wait_kmcnt 0x0
; GFX12-NEXT:    buffer_atomic_pk_add_bf16 v0, off, s[0:3], s16 offset:92 th:TH_ATOMIC_NT
; GFX12-NEXT:    s_setpc_b64 s[30:31]
  %unused = call <2 x bfloat> @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.v2bf16(<2 x bfloat> %val, ptr addrspace(8) %rsrc, i32 92, i32 %soffset, i32 2)
  ret void
}

; Test waterfall loop
define <2 x bfloat> @raw_ptr_buffer_atomic_add_v2bf16_rtn__vgpr_val__vgpr_rsrc__vgpr_voffset_add__vgpr_soffset(<2 x bfloat> %val, ptr addrspace(8) %rsrc, i32 %voffset, i32 %soffset) #0 {
; GFX12-LABEL: raw_ptr_buffer_atomic_add_v2bf16_rtn__vgpr_val__vgpr_rsrc__vgpr_voffset_add__vgpr_soffset:
; GFX12:       ; %bb.0:
; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
; GFX12-NEXT:    s_wait_expcnt 0x0
; GFX12-NEXT:    s_wait_samplecnt 0x0
; GFX12-NEXT:    s_wait_bvhcnt 0x0
; GFX12-NEXT:    s_wait_kmcnt 0x0
; GFX12-NEXT:    s_mov_b32 s2, exec_lo
; GFX12-NEXT:  .LBB4_1: ; =>This Inner Loop Header: Depth=1
; GFX12-NEXT:    v_readfirstlane_b32 s4, v1
; GFX12-NEXT:    v_readfirstlane_b32 s5, v2
; GFX12-NEXT:    v_readfirstlane_b32 s6, v3
; GFX12-NEXT:    v_readfirstlane_b32 s7, v4
; GFX12-NEXT:    v_readfirstlane_b32 s3, v6
; GFX12-NEXT:    s_wait_alu 0xf1ff
; GFX12-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[1:2]
; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX12-NEXT:    v_cmp_eq_u64_e64 s0, s[6:7], v[3:4]
; GFX12-NEXT:    v_cmp_eq_u32_e64 s1, s3, v6
; GFX12-NEXT:    s_and_b32 s0, vcc_lo, s0
; GFX12-NEXT:    s_wait_alu 0xfffe
; GFX12-NEXT:    s_and_b32 s0, s0, s1
; GFX12-NEXT:    s_wait_alu 0xfffe
; GFX12-NEXT:    s_and_saveexec_b32 s0, s0
; GFX12-NEXT:    s_wait_loadcnt 0x0
; GFX12-NEXT:    buffer_atomic_pk_add_bf16 v0, v5, s[4:7], s3 offen offset:128 th:TH_ATOMIC_RETURN
; GFX12-NEXT:    ; implicit-def: $vgpr1_vgpr2_vgpr3_vgpr4
; GFX12-NEXT:    ; implicit-def: $vgpr6
; GFX12-NEXT:    ; implicit-def: $vgpr5
; GFX12-NEXT:    s_xor_b32 exec_lo, exec_lo, s0
; GFX12-NEXT:    s_cbranch_execnz .LBB4_1
; GFX12-NEXT:  ; %bb.2:
; GFX12-NEXT:    s_mov_b32 exec_lo, s2
; GFX12-NEXT:    s_wait_loadcnt 0x0
; GFX12-NEXT:    s_setpc_b64 s[30:31]
  %voffset.add = add i32 %voffset, 128
  %ret = call <2 x bfloat> @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.v2bf16(<2 x bfloat> %val, ptr addrspace(8) %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
  ret <2 x bfloat> %ret
}

declare <2 x bfloat> @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.v2bf16(<2 x bfloat>, ptr addrspace(8), i32, i32, i32 immarg)

attributes #0 = { nounwind }