; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc -mtriple=amdgcn -mcpu=gfx803 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX8 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX10 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=+real-true16 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefixes=GFX11,GFX11-TRUE16 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefixes=GFX11,GFX11-FAKE16 %s

; FIXME: GFX9 should be producing v_mad_u16 instead of v_mad_legacy_u16.

define amdgpu_kernel void @mad_u16(
; GFX8-LABEL: mad_u16:
; GFX8:       ; %bb.0: ; %entry
; GFX8-NEXT:    s_load_dwordx8 s[0:7], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v4, 1, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v4
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    v_mov_b32_e32 v3, s5
; GFX8-NEXT:    v_add_u32_e32 v2, vcc, s4, v4
; GFX8-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
; GFX8-NEXT:    v_mov_b32_e32 v5, s7
; GFX8-NEXT:    v_add_u32_e32 v4, vcc, s6, v4
; GFX8-NEXT:    v_addc_u32_e32 v5, vcc, 0, v5, vcc
; GFX8-NEXT:    flat_load_ushort v6, v[0:1] glc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    flat_load_ushort v2, v[2:3] glc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    flat_load_ushort v3, v[4:5] glc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v0, s0
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_mad_u16 v2, v6, v2, v3
; GFX8-NEXT:    flat_store_short v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: mad_u16:
; GFX9:       ; %bb.0: ; %entry
; GFX9-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_ushort v1, v0, s[10:11] glc
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    global_load_ushort v2, v0, s[12:13] glc
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    global_load_ushort v3, v0, s[14:15] glc
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_mov_b32_e32 v0, 0
; GFX9-NEXT:    v_mad_legacy_u16 v1, v1, v2, v3
; GFX9-NEXT:    global_store_short v0, v1, s[8:9]
; GFX9-NEXT:    s_endpgm
;
; GFX10-LABEL: mad_u16:
; GFX10:       ; %bb.0: ; %entry
; GFX10-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    global_load_ushort v1, v0, s[10:11] glc dlc
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    global_load_ushort v2, v0, s[12:13] glc dlc
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    global_load_ushort v3, v0, s[14:15] glc dlc
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_mov_b32_e32 v0, 0
; GFX10-NEXT:    v_mad_u16 v1, v1, v2, v3
; GFX10-NEXT:    global_store_short v0, v1, s[8:9]
; GFX10-NEXT:    s_endpgm
;
; GFX11-TRUE16-LABEL: mad_u16:
; GFX11-TRUE16:       ; %bb.0: ; %entry
; GFX11-TRUE16-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
; GFX11-TRUE16-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v1, 1, v0
; GFX11-TRUE16-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-TRUE16-NEXT:    global_load_d16_b16 v0, v1, s[2:3] glc dlc
; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT:    global_load_d16_hi_b16 v0, v1, s[4:5] glc dlc
; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT:    global_load_d16_b16 v1, v1, s[6:7] glc dlc
; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT:    v_mad_u16 v0.l, v0.l, v0.h, v1.l
; GFX11-TRUE16-NEXT:    global_store_b16 v2, v0, s[0:1]
; GFX11-TRUE16-NEXT:    s_endpgm
;
; GFX11-FAKE16-LABEL: mad_u16:
; GFX11-FAKE16:       ; %bb.0: ; %entry
; GFX11-FAKE16-NEXT:    s_load_b256 s[0:7], s[4:5], 0x24
; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-FAKE16-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
; GFX11-FAKE16-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-FAKE16-NEXT:    global_load_u16 v1, v0, s[2:3] glc dlc
; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT:    global_load_u16 v2, v0, s[4:5] glc dlc
; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT:    global_load_u16 v0, v0, s[6:7] glc dlc
; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT:    v_mad_u16 v0, v1, v2, v0
; GFX11-FAKE16-NEXT:    global_store_b16 v3, v0, s[0:1]
; GFX11-FAKE16-NEXT:    s_endpgm
    ptr addrspace(1) %r,
    ptr addrspace(1) %a,
    ptr addrspace(1) %b,
    ptr addrspace(1) %c) {
entry:
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %a.gep = getelementptr inbounds i16, ptr addrspace(1) %a, i32 %tid
  %b.gep = getelementptr inbounds i16, ptr addrspace(1) %b, i32 %tid
  %c.gep = getelementptr inbounds i16, ptr addrspace(1) %c, i32 %tid

  %a.val = load volatile i16, ptr addrspace(1) %a.gep
  %b.val = load volatile i16, ptr addrspace(1) %b.gep
  %c.val = load volatile i16, ptr addrspace(1) %c.gep

  %m.val = mul i16 %a.val, %b.val
  %r.val = add i16 %m.val, %c.val

  store i16 %r.val, ptr addrspace(1) %r
  ret void
}

define i16 @v_mad_u16(i16 %arg0, i16 %arg1, i16 %arg2) {
; GFX8-LABEL: v_mad_u16:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT:    v_mad_u16 v0, v0, v1, v2
; GFX8-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_mad_u16:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_mad_legacy_u16 v0, v0, v1, v2
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_mad_u16:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_mad_u16 v0, v0, v1, v2
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-TRUE16-LABEL: v_mad_u16:
; GFX11-TRUE16:       ; %bb.0:
; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT:    v_mad_u16 v0.l, v0.l, v1.l, v2.l
; GFX11-TRUE16-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_mad_u16:
; GFX11-FAKE16:       ; %bb.0:
; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT:    v_mad_u16 v0, v0, v1, v2
; GFX11-FAKE16-NEXT:    s_setpc_b64 s[30:31]
  %mul = mul i16 %arg0, %arg1
  %add = add i16 %mul, %arg2
  ret i16 %add
}

define i32 @v_mad_u16_zext(i16 %arg0, i16 %arg1, i16 %arg2) {
; GFX8-LABEL: v_mad_u16_zext:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT:    v_mad_u16 v0, v0, v1, v2
; GFX8-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_mad_u16_zext:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_mad_legacy_u16 v0, v0, v1, v2
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_mad_u16_zext:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_mad_u16 v0, v0, v1, v2
; GFX10-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-TRUE16-LABEL: v_mad_u16_zext:
; GFX11-TRUE16:       ; %bb.0:
; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT:    v_mad_u16 v0.l, v0.l, v1.l, v2.l
; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; GFX11-TRUE16-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_mad_u16_zext:
; GFX11-FAKE16:       ; %bb.0:
; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT:    v_mad_u16 v0, v0, v1, v2
; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-FAKE16-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; GFX11-FAKE16-NEXT:    s_setpc_b64 s[30:31]
  %mul = mul i16 %arg0, %arg1
  %add = add i16 %mul, %arg2
  %zext = zext i16 %add to i32
  ret i32 %zext
}

define i64 @v_mad_u16_zext64(i16 %arg0, i16 %arg1, i16 %arg2) {
; GFX8-LABEL: v_mad_u16_zext64:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT:    v_mad_u16 v0, v0, v1, v2
; GFX8-NEXT:    v_mov_b32_e32 v1, 0
; GFX8-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_mad_u16_zext64:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_mad_legacy_u16 v0, v0, v1, v2
; GFX9-NEXT:    v_mov_b32_e32 v1, 0
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_mad_u16_zext64:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_mad_u16 v0, v0, v1, v2
; GFX10-NEXT:    v_mov_b32_e32 v1, 0
; GFX10-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-TRUE16-LABEL: v_mad_u16_zext64:
; GFX11-TRUE16:       ; %bb.0:
; GFX11-TRUE16-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT:    v_mad_u16 v0.l, v0.l, v1.l, v2.l
; GFX11-TRUE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0xffff, v0
; GFX11-TRUE16-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_mad_u16_zext64:
; GFX11-FAKE16:       ; %bb.0:
; GFX11-FAKE16-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT:    v_mad_u16 v0, v0, v1, v2
; GFX11-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-FAKE16-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0xffff, v0
; GFX11-FAKE16-NEXT:    s_setpc_b64 s[30:31]
  %mul = mul i16 %arg0, %arg1
  %add = add i16 %mul, %arg2
  %zext = zext i16 %add to i64
  ret i64 %zext
}

define amdgpu_ps i16 @s_mad_u16(i16 inreg %arg0, i16 inreg %arg1, i16 inreg %arg2) {
; GFX8-LABEL: s_mad_u16:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_mul_i32 s0, s0, s1
; GFX8-NEXT:    s_add_i32 s0, s0, s2
; GFX8-NEXT:    ; return to shader part epilog
;
; GFX9-LABEL: s_mad_u16:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_mul_i32 s0, s0, s1
; GFX9-NEXT:    s_add_i32 s0, s0, s2
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: s_mad_u16:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_mul_i32 s0, s0, s1
; GFX10-NEXT:    s_add_i32 s0, s0, s2
; GFX10-NEXT:    ; return to shader part epilog
;
; GFX11-LABEL: s_mad_u16:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_mul_i32 s0, s0, s1
; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT:    s_add_i32 s0, s0, s2
; GFX11-NEXT:    ; return to shader part epilog
  %mul = mul i16 %arg0, %arg1
  %add = add i16 %mul, %arg2
  ret i16 %add
}

define amdgpu_ps i32 @s_mad_u16_zext(i16 inreg %arg0, i16 inreg %arg1, i16 inreg %arg2) {
; GFX8-LABEL: s_mad_u16_zext:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_mul_i32 s0, s0, s1
; GFX8-NEXT:    s_add_i32 s0, s0, s2
; GFX8-NEXT:    s_and_b32 s0, s0, 0xffff
; GFX8-NEXT:    ; return to shader part epilog
;
; GFX9-LABEL: s_mad_u16_zext:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_mul_i32 s0, s0, s1
; GFX9-NEXT:    s_add_i32 s0, s0, s2
; GFX9-NEXT:    s_and_b32 s0, s0, 0xffff
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: s_mad_u16_zext:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_mul_i32 s0, s0, s1
; GFX10-NEXT:    s_add_i32 s0, s0, s2
; GFX10-NEXT:    s_and_b32 s0, s0, 0xffff
; GFX10-NEXT:    ; return to shader part epilog
;
; GFX11-LABEL: s_mad_u16_zext:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_mul_i32 s0, s0, s1
; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-NEXT:    s_add_i32 s0, s0, s2
; GFX11-NEXT:    s_and_b32 s0, s0, 0xffff
; GFX11-NEXT:    ; return to shader part epilog
  %mul = mul i16 %arg0, %arg1
  %add = add i16 %mul, %arg2
  %zext = zext i16 %add to i32
  ret i32 %zext
}

define amdgpu_ps i64 @s_mad_u16_zext64(i16 inreg %arg0, i16 inreg %arg1, i16 inreg %arg2) {
; GFX8-LABEL: s_mad_u16_zext64:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_mul_i32 s0, s0, s1
; GFX8-NEXT:    s_add_i32 s0, s0, s2
; GFX8-NEXT:    s_and_b32 s0, s0, 0xffff
; GFX8-NEXT:    s_mov_b32 s1, 0
; GFX8-NEXT:    ; return to shader part epilog
;
; GFX9-LABEL: s_mad_u16_zext64:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_mul_i32 s0, s0, s1
; GFX9-NEXT:    s_add_i32 s0, s0, s2
; GFX9-NEXT:    s_and_b32 s0, s0, 0xffff
; GFX9-NEXT:    s_mov_b32 s1, 0
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: s_mad_u16_zext64:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_mul_i32 s0, s0, s1
; GFX10-NEXT:    s_mov_b32 s1, 0
; GFX10-NEXT:    s_add_i32 s0, s0, s2
; GFX10-NEXT:    s_and_b32 s0, s0, 0xffff
; GFX10-NEXT:    ; return to shader part epilog
;
; GFX11-LABEL: s_mad_u16_zext64:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_mul_i32 s0, s0, s1
; GFX11-NEXT:    s_mov_b32 s1, 0
; GFX11-NEXT:    s_add_i32 s0, s0, s2
; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT:    s_and_b32 s0, s0, 0xffff
; GFX11-NEXT:    ; return to shader part epilog
  %mul = mul i16 %arg0, %arg1
  %add = add i16 %mul, %arg2
  %zext = zext i16 %add to i64
  ret i64 %zext
}

define amdgpu_ps i32 @s_mad_u16_sext(i16 inreg %arg0, i16 inreg %arg1, i16 inreg %arg2) {
; GFX8-LABEL: s_mad_u16_sext:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_mul_i32 s0, s0, s1
; GFX8-NEXT:    s_add_i32 s0, s0, s2
; GFX8-NEXT:    s_sext_i32_i16 s0, s0
; GFX8-NEXT:    ; return to shader part epilog
;
; GFX9-LABEL: s_mad_u16_sext:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_mul_i32 s0, s0, s1
; GFX9-NEXT:    s_add_i32 s0, s0, s2
; GFX9-NEXT:    s_sext_i32_i16 s0, s0
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10-LABEL: s_mad_u16_sext:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_mul_i32 s0, s0, s1
; GFX10-NEXT:    s_add_i32 s0, s0, s2
; GFX10-NEXT:    s_sext_i32_i16 s0, s0
; GFX10-NEXT:    ; return to shader part epilog
;
; GFX11-LABEL: s_mad_u16_sext:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_mul_i32 s0, s0, s1
; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-NEXT:    s_add_i32 s0, s0, s2
; GFX11-NEXT:    s_sext_i32_i16 s0, s0
; GFX11-NEXT:    ; return to shader part epilog
  %mul = mul i16 %arg0, %arg1
  %add = add i16 %mul, %arg2
  %sext = sext i16 %add to i32
  ret i32 %sext
}

declare i32 @llvm.amdgcn.workitem.id.x()
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GCN: {{.*}}