; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 -mattr=-architected-sgprs -global-isel=0 < %s | FileCheck -check-prefixes=GFX9-SDAG %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 -mattr=-architected-sgprs -global-isel=1 -new-reg-bank-select < %s | FileCheck -check-prefixes=GFX9-GISEL %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 -mattr=+architected-sgprs -global-isel=0 < %s | FileCheck -check-prefixes=GFX9ARCH-SDAG %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 -mattr=+architected-sgprs -global-isel=1 -new-reg-bank-select < %s | FileCheck -check-prefixes=GFX9ARCH-GISEL %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1200 -global-isel=0 < %s | FileCheck -check-prefixes=GFX12,GFX12-SDAG %s
; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1200 -global-isel=1 -new-reg-bank-select < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL %s

define amdgpu_cs void @_amdgpu_cs_main() {
; GFX9-LABEL: _amdgpu_cs_main:
; GFX9:       ; %bb.0: ; %.entry
; GFX9-NEXT:    buffer_store_dwordx3 v[0:2], off, s[0:3], 0
; GFX9-NEXT:    s_endpgm
;
; GFX9-SDAG-LABEL: _amdgpu_cs_main:
; GFX9-SDAG:       ; %bb.0: ; %.entry
; GFX9-SDAG-NEXT:    buffer_store_dwordx3 v[0:2], off, s[0:3], 0
; GFX9-SDAG-NEXT:    s_endpgm
;
; GFX9-GISEL-LABEL: _amdgpu_cs_main:
; GFX9-GISEL:       ; %bb.0: ; %.entry
; GFX9-GISEL-NEXT:    buffer_store_dwordx3 v[0:2], off, s[0:3], 0
; GFX9-GISEL-NEXT:    s_endpgm
;
; GFX9ARCH-SDAG-LABEL: _amdgpu_cs_main:
; GFX9ARCH-SDAG:       ; %bb.0: ; %.entry
; GFX9ARCH-SDAG-NEXT:    s_lshr_b32 s0, ttmp7, 16
; GFX9ARCH-SDAG-NEXT:    s_and_b32 s1, ttmp7, 0xffff
; GFX9ARCH-SDAG-NEXT:    v_mov_b32_e32 v0, ttmp9
; GFX9ARCH-SDAG-NEXT:    v_mov_b32_e32 v1, s1
; GFX9ARCH-SDAG-NEXT:    v_mov_b32_e32 v2, s0
; GFX9ARCH-SDAG-NEXT:    buffer_store_dwordx3 v[0:2], off, s[0:3], 0
; GFX9ARCH-SDAG-NEXT:    s_endpgm
;
; GFX9ARCH-GISEL-LABEL: _amdgpu_cs_main:
; GFX9ARCH-GISEL:       ; %bb.0: ; %.entry
; GFX9ARCH-GISEL-NEXT:    s_mov_b32 s0, ttmp9
; GFX9ARCH-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
; GFX9ARCH-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v0, s0
; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v1, s1
; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v2, s2
; GFX9ARCH-GISEL-NEXT:    buffer_store_dwordx3 v[0:2], off, s[0:3], 0
; GFX9ARCH-GISEL-NEXT:    s_endpgm
;
; GFX12-SDAG-LABEL: _amdgpu_cs_main:
; GFX12-SDAG:       ; %bb.0: ; %.entry
; GFX12-SDAG-NEXT:    s_and_b32 s0, ttmp7, 0xffff
; GFX12-SDAG-NEXT:    s_lshr_b32 s1, ttmp7, 16
; GFX12-SDAG-NEXT:    v_dual_mov_b32 v0, ttmp9 :: v_dual_mov_b32 v1, s0
; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, s1
; GFX12-SDAG-NEXT:    buffer_store_b96 v[0:2], off, s[0:3], null
; GFX12-SDAG-NEXT:    s_endpgm
;
; GFX12-GISEL-LABEL: _amdgpu_cs_main:
; GFX12-GISEL:       ; %bb.0: ; %.entry
; GFX12-GISEL-NEXT:    s_mov_b32 s0, ttmp9
; GFX12-GISEL-NEXT:    s_and_b32 s1, ttmp7, 0xffff
; GFX12-GISEL-NEXT:    s_lshr_b32 s2, ttmp7, 16
; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, s2
; GFX12-GISEL-NEXT:    buffer_store_b96 v[0:2], off, s[0:3], null
; GFX12-GISEL-NEXT:    s_endpgm
.entry:
  %idx = call i32 @llvm.amdgcn.workgroup.id.x()
  %idy = call i32 @llvm.amdgcn.workgroup.id.y()
  %idz = call i32 @llvm.amdgcn.workgroup.id.z()
  %ielemx = insertelement <3 x i32> poison, i32 %idx, i64 0
  %ielemy = insertelement <3 x i32> %ielemx, i32 %idy, i64 1
  %ielemz = insertelement <3 x i32> %ielemy, i32 %idz, i64 2
  call void @llvm.amdgcn.raw.ptr.buffer.store.v3i32(<3 x i32> %ielemz, ptr addrspace(8) poison, i32 0, i32 0, i32 0)
  ret void
}

define amdgpu_cs void @caller() {
; GFX9-SDAG-LABEL: caller:
; GFX9-SDAG:       ; %bb.0:
; GFX9-SDAG-NEXT:    s_getpc_b64 s[8:9]
; GFX9-SDAG-NEXT:    s_mov_b32 s8, s0
; GFX9-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x10
; GFX9-SDAG-NEXT:    s_mov_b32 s5, callee@abs32@hi
; GFX9-SDAG-NEXT:    s_mov_b32 s4, callee@abs32@lo
; GFX9-SDAG-NEXT:    s_mov_b32 s32, 0
; GFX9-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-SDAG-NEXT:    s_add_u32 s8, s8, s0
; GFX9-SDAG-NEXT:    s_addc_u32 s9, s9, 0
; GFX9-SDAG-NEXT:    s_mov_b64 s[0:1], s[8:9]
; GFX9-SDAG-NEXT:    s_mov_b64 s[2:3], s[10:11]
; GFX9-SDAG-NEXT:    s_swappc_b64 s[30:31], s[4:5]
; GFX9-SDAG-NEXT:    s_endpgm
;
; GFX9-GISEL-LABEL: caller:
; GFX9-GISEL:       ; %bb.0:
; GFX9-GISEL-NEXT:    s_getpc_b64 s[8:9]
; GFX9-GISEL-NEXT:    s_mov_b32 s8, s0
; GFX9-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x10
; GFX9-GISEL-NEXT:    s_mov_b32 s4, callee@abs32@lo
; GFX9-GISEL-NEXT:    s_mov_b32 s5, callee@abs32@hi
; GFX9-GISEL-NEXT:    s_mov_b32 s32, 0
; GFX9-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-GISEL-NEXT:    s_add_u32 s8, s8, s0
; GFX9-GISEL-NEXT:    s_addc_u32 s9, s9, 0
; GFX9-GISEL-NEXT:    s_mov_b64 s[0:1], s[8:9]
; GFX9-GISEL-NEXT:    s_mov_b64 s[2:3], s[10:11]
; GFX9-GISEL-NEXT:    s_swappc_b64 s[30:31], s[4:5]
; GFX9-GISEL-NEXT:    s_endpgm
;
; GFX9ARCH-SDAG-LABEL: caller:
; GFX9ARCH-SDAG:       ; %bb.0:
; GFX9ARCH-SDAG-NEXT:    s_getpc_b64 s[8:9]
; GFX9ARCH-SDAG-NEXT:    s_mov_b32 s8, s0
; GFX9ARCH-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x10
; GFX9ARCH-SDAG-NEXT:    s_mov_b32 s5, callee@abs32@hi
; GFX9ARCH-SDAG-NEXT:    s_mov_b32 s4, callee@abs32@lo
; GFX9ARCH-SDAG-NEXT:    v_mov_b32_e32 v0, ttmp9
; GFX9ARCH-SDAG-NEXT:    s_mov_b32 s32, 0
; GFX9ARCH-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9ARCH-SDAG-NEXT:    s_add_u32 s8, s8, s0
; GFX9ARCH-SDAG-NEXT:    s_addc_u32 s9, s9, 0
; GFX9ARCH-SDAG-NEXT:    s_mov_b64 s[0:1], s[8:9]
; GFX9ARCH-SDAG-NEXT:    s_mov_b64 s[2:3], s[10:11]
; GFX9ARCH-SDAG-NEXT:    s_swappc_b64 s[30:31], s[4:5]
; GFX9ARCH-SDAG-NEXT:    s_endpgm
;
; GFX9ARCH-GISEL-LABEL: caller:
; GFX9ARCH-GISEL:       ; %bb.0:
; GFX9ARCH-GISEL-NEXT:    s_getpc_b64 s[8:9]
; GFX9ARCH-GISEL-NEXT:    s_mov_b32 s8, s0
; GFX9ARCH-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[8:9], 0x10
; GFX9ARCH-GISEL-NEXT:    s_mov_b32 s4, callee@abs32@lo
; GFX9ARCH-GISEL-NEXT:    s_mov_b32 s5, callee@abs32@hi
; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v0, ttmp9
; GFX9ARCH-GISEL-NEXT:    s_mov_b32 s32, 0
; GFX9ARCH-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9ARCH-GISEL-NEXT:    s_add_u32 s8, s8, s0
; GFX9ARCH-GISEL-NEXT:    s_addc_u32 s9, s9, 0
; GFX9ARCH-GISEL-NEXT:    s_mov_b64 s[0:1], s[8:9]
; GFX9ARCH-GISEL-NEXT:    s_mov_b64 s[2:3], s[10:11]
; GFX9ARCH-GISEL-NEXT:    s_swappc_b64 s[30:31], s[4:5]
; GFX9ARCH-GISEL-NEXT:    s_endpgm
;
; GFX12-SDAG-LABEL: caller:
; GFX12-SDAG:       ; %bb.0:
; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, ttmp9
; GFX12-SDAG-NEXT:    s_mov_b32 s1, callee@abs32@hi
; GFX12-SDAG-NEXT:    s_mov_b32 s0, callee@abs32@lo
; GFX12-SDAG-NEXT:    s_mov_b32 s32, 0
; GFX12-SDAG-NEXT:    s_wait_alu depctr_sa_sdst(0)
; GFX12-SDAG-NEXT:    s_swappc_b64 s[30:31], s[0:1]
; GFX12-SDAG-NEXT:    s_endpgm
;
; GFX12-GISEL-LABEL: caller:
; GFX12-GISEL:       ; %bb.0:
; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, ttmp9
; GFX12-GISEL-NEXT:    s_mov_b32 s0, callee@abs32@lo
; GFX12-GISEL-NEXT:    s_mov_b32 s1, callee@abs32@hi
; GFX12-GISEL-NEXT:    s_mov_b32 s32, 0
; GFX12-GISEL-NEXT:    s_wait_alu depctr_sa_sdst(0)
; GFX12-GISEL-NEXT:    s_swappc_b64 s[30:31], s[0:1]
; GFX12-GISEL-NEXT:    s_endpgm
  %idx = call i32 @llvm.amdgcn.workgroup.id.x()
  call amdgpu_gfx void @callee(i32 %idx)
  ret void
}

declare amdgpu_gfx void @callee(i32)

define amdgpu_gfx void @workgroup_ids_gfx(ptr addrspace(1) %outx, ptr addrspace(1) %outy, ptr addrspace(1) %outz) {
; GFX9-SDAG-LABEL: workgroup_ids_gfx:
; GFX9-SDAG:       ; %bb.0:
; GFX9-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-SDAG-NEXT:    global_store_dword v[0:1], v0, off
; GFX9-SDAG-NEXT:    s_waitcnt vmcnt(0)
; GFX9-SDAG-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-SDAG-NEXT:    s_waitcnt vmcnt(0)
; GFX9-SDAG-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-SDAG-NEXT:    s_waitcnt vmcnt(0)
; GFX9-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-GISEL-LABEL: workgroup_ids_gfx:
; GFX9-GISEL:       ; %bb.0:
; GFX9-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9ARCH-SDAG-LABEL: workgroup_ids_gfx:
; GFX9ARCH-SDAG:       ; %bb.0:
; GFX9ARCH-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9ARCH-SDAG-NEXT:    v_mov_b32_e32 v6, ttmp9
; GFX9ARCH-SDAG-NEXT:    s_and_b32 s34, ttmp7, 0xffff
; GFX9ARCH-SDAG-NEXT:    global_store_dword v[0:1], v6, off
; GFX9ARCH-SDAG-NEXT:    s_waitcnt vmcnt(0)
; GFX9ARCH-SDAG-NEXT:    v_mov_b32_e32 v0, s34
; GFX9ARCH-SDAG-NEXT:    s_lshr_b32 s34, ttmp7, 16
; GFX9ARCH-SDAG-NEXT:    global_store_dword v[2:3], v0, off
; GFX9ARCH-SDAG-NEXT:    s_waitcnt vmcnt(0)
; GFX9ARCH-SDAG-NEXT:    v_mov_b32_e32 v0, s34
; GFX9ARCH-SDAG-NEXT:    global_store_dword v[4:5], v0, off
; GFX9ARCH-SDAG-NEXT:    s_waitcnt vmcnt(0)
; GFX9ARCH-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9ARCH-GISEL-LABEL: workgroup_ids_gfx:
; GFX9ARCH-GISEL:       ; %bb.0:
; GFX9ARCH-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v6, ttmp9
; GFX9ARCH-GISEL-NEXT:    s_and_b32 s34, ttmp7, 0xffff
; GFX9ARCH-GISEL-NEXT:    s_lshr_b32 s35, ttmp7, 16
; GFX9ARCH-GISEL-NEXT:    global_store_dword v[0:1], v6, off
; GFX9ARCH-GISEL-NEXT:    s_waitcnt vmcnt(0)
; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v0, s34
; GFX9ARCH-GISEL-NEXT:    global_store_dword v[2:3], v0, off
; GFX9ARCH-GISEL-NEXT:    s_waitcnt vmcnt(0)
; GFX9ARCH-GISEL-NEXT:    v_mov_b32_e32 v0, s35
; GFX9ARCH-GISEL-NEXT:    global_store_dword v[4:5], v0, off
; GFX9ARCH-GISEL-NEXT:    s_waitcnt vmcnt(0)
; GFX9ARCH-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-LABEL: workgroup_ids_gfx:
; 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_and_b32 s0, ttmp7, 0xffff
; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
; GFX12-NEXT:    v_dual_mov_b32 v6, ttmp9 :: v_dual_mov_b32 v7, s0
; GFX12-NEXT:    s_lshr_b32 s1, ttmp7, 16
; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
; GFX12-NEXT:    v_mov_b32_e32 v8, s1
; GFX12-NEXT:    s_wait_storecnt 0x0
; GFX12-NEXT:    global_store_b32 v[0:1], v6, off scope:SCOPE_SYS
; GFX12-NEXT:    s_wait_storecnt 0x0
; GFX12-NEXT:    global_store_b32 v[2:3], v7, off scope:SCOPE_SYS
; GFX12-NEXT:    s_wait_storecnt 0x0
; GFX12-NEXT:    global_store_b32 v[4:5], v8, off scope:SCOPE_SYS
; GFX12-NEXT:    s_wait_storecnt 0x0
; GFX12-NEXT:    s_setpc_b64 s[30:31]
  %id.x = call i32 @llvm.amdgcn.workgroup.id.x()
  %id.y = call i32 @llvm.amdgcn.workgroup.id.y()
  %id.z = call i32 @llvm.amdgcn.workgroup.id.z()
  store volatile i32 %id.x, ptr addrspace(1) %outx
  store volatile i32 %id.y, ptr addrspace(1) %outy
  store volatile i32 %id.z, ptr addrspace(1) %outz
  ret void
}