; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -mtriple=amdgcn--amdpal -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GFX6 %s
; RUN: llc -mtriple=amdgcn--amdpal -mcpu=tonga -mattr=-flat-for-global,-xnack -verify-machineinstrs < %s | FileCheck -check-prefix=GFX8 %s

define amdgpu_kernel void @s_mulk_i32_k0(ptr addrspace(1) %out, i32 %b) {
; GFX6-LABEL: s_mulk_i32_k0:
; GFX6:       ; %bb.0:
; GFX6-NEXT:    s_load_dword s6, s[4:5], 0x2
; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX6-NEXT:    s_mov_b32 s3, 0xf000
; GFX6-NEXT:    s_mov_b32 s2, -1
; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
; GFX6-NEXT:    s_mul_i32 s4, s6, 0x41
; GFX6-NEXT:    v_mov_b32_e32 v0, s4
; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; GFX6-NEXT:    s_endpgm
;
; GFX8-LABEL: s_mulk_i32_k0:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dword s6, s[4:5], 0x8
; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX8-NEXT:    s_mov_b32 s3, 0xf000
; GFX8-NEXT:    s_mov_b32 s2, -1
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    s_mul_i32 s4, s6, 0x41
; GFX8-NEXT:    v_mov_b32_e32 v0, s4
; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; GFX8-NEXT:    s_endpgm
  %mul = mul i32 %b, 65
  store i32 %mul, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @s_mulk_i32_k1(ptr addrspace(1) %out, i32 %b) {
; GFX6-LABEL: s_mulk_i32_k1:
; GFX6:       ; %bb.0:
; GFX6-NEXT:    s_load_dword s6, s[4:5], 0x2
; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX6-NEXT:    s_mov_b32 s3, 0xf000
; GFX6-NEXT:    s_mov_b32 s2, -1
; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
; GFX6-NEXT:    s_mul_i32 s4, s6, 0x7fff
; GFX6-NEXT:    v_mov_b32_e32 v0, s4
; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; GFX6-NEXT:    s_endpgm
;
; GFX8-LABEL: s_mulk_i32_k1:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dword s6, s[4:5], 0x8
; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX8-NEXT:    s_mov_b32 s3, 0xf000
; GFX8-NEXT:    s_mov_b32 s2, -1
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    s_mul_i32 s4, s6, 0x7fff
; GFX8-NEXT:    v_mov_b32_e32 v0, s4
; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; GFX8-NEXT:    s_endpgm
  %mul = mul i32 %b, 32767 ; (1 << 15) - 1
  store i32 %mul, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @s_mulk_i32_k2(ptr addrspace(1) %out, i32 %b) {
; GFX6-LABEL: s_mulk_i32_k2:
; GFX6:       ; %bb.0:
; GFX6-NEXT:    s_load_dword s6, s[4:5], 0x2
; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX6-NEXT:    s_mov_b32 s3, 0xf000
; GFX6-NEXT:    s_mov_b32 s2, -1
; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
; GFX6-NEXT:    s_mul_i32 s4, s6, 0xffffffef
; GFX6-NEXT:    v_mov_b32_e32 v0, s4
; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; GFX6-NEXT:    s_endpgm
;
; GFX8-LABEL: s_mulk_i32_k2:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dword s6, s[4:5], 0x8
; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX8-NEXT:    s_mov_b32 s3, 0xf000
; GFX8-NEXT:    s_mov_b32 s2, -1
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    s_mul_i32 s4, s6, 0xffffffef
; GFX8-NEXT:    v_mov_b32_e32 v0, s4
; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; GFX8-NEXT:    s_endpgm
  %mul = mul i32 %b, -17
  store i32 %mul, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @no_s_mulk_i32_k0(ptr addrspace(1) %out, i32 %b) {
; GFX6-LABEL: no_s_mulk_i32_k0:
; GFX6:       ; %bb.0:
; GFX6-NEXT:    s_load_dword s6, s[4:5], 0x2
; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX6-NEXT:    s_mov_b32 s3, 0xf000
; GFX6-NEXT:    s_mov_b32 s2, -1
; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
; GFX6-NEXT:    s_mul_i32 s4, s6, 0x8001
; GFX6-NEXT:    v_mov_b32_e32 v0, s4
; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; GFX6-NEXT:    s_endpgm
;
; GFX8-LABEL: no_s_mulk_i32_k0:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dword s6, s[4:5], 0x8
; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX8-NEXT:    s_mov_b32 s3, 0xf000
; GFX8-NEXT:    s_mov_b32 s2, -1
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    s_mul_i32 s4, s6, 0x8001
; GFX8-NEXT:    v_mov_b32_e32 v0, s4
; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; GFX8-NEXT:    s_endpgm
  %mul = mul i32 %b, 32769 ; 1 << 15 + 1
  store i32 %mul, ptr addrspace(1) %out
  ret void
}

@lds = addrspace(3) global [512 x i32] poison, align 4

define amdgpu_kernel void @commute_s_mulk_i32(ptr addrspace(1) %out, i32 %b) #0 {
; GFX6-LABEL: commute_s_mulk_i32:
; GFX6:       ; %bb.0:
; GFX6-NEXT:    s_load_dword s0, s[4:5], 0x2
; GFX6-NEXT:    v_mov_b32_e32 v0, 0
; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
; GFX6-NEXT:    s_mulk_i32 s0, 0x800
; GFX6-NEXT:    ;;#ASMSTART
; GFX6-NEXT:    ; foo v0, s0
; GFX6-NEXT:    ;;#ASMEND
; GFX6-NEXT:    s_endpgm
;
; GFX8-LABEL: commute_s_mulk_i32:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dword s0, s[4:5], 0x8
; GFX8-NEXT:    v_mov_b32_e32 v0, 0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    s_mulk_i32 s0, 0x800
; GFX8-NEXT:    ;;#ASMSTART
; GFX8-NEXT:    ; foo v0, s0
; GFX8-NEXT:    ;;#ASMEND
; GFX8-NEXT:    s_endpgm
  %size = call i32 @llvm.amdgcn.groupstaticsize()
  %add = mul i32 %size, %b
  call void asm sideeffect "; foo $0, $1", "v,s"(ptr addrspace(3) @lds, i32 %add)
  ret void
}

declare i32 @llvm.amdgcn.groupstaticsize() #1

attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }