llvm/test/CodeGen/AMDGPU/clamp-modifier.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc -mtriple=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=SI %s
; RUN: llc -mtriple=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX8 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX11 %s

define amdgpu_kernel void @v_clamp_add_src_f32(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_add_src_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_add_f32_e64 v2, v2, 1.0 clamp
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_f32:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f32_e64 v2, v3, 1.0 clamp
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_f32:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f32_e64 v1, v1, 1.0 clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_f32:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f32_e64 v1, v1, 1.0 clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr float, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr float, ptr addrspace(1) %out, i32 %tid
  %a = load float, ptr addrspace(1) %gep0
  %add = fadd float %a, 1.0
  %max = call float @llvm.maxnum.f32(float %add, float 0.0)
  %clamp = call float @llvm.minnum.f32(float %max, float 1.0)
  store float %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_multi_use_src_f32(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_multi_use_src_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_add_f32_e32 v2, 1.0, v2
; SI-NEXT:    v_max_f32_e64 v3, v2, v2 clamp
; SI-NEXT:    buffer_store_dword v3, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    buffer_store_dword v2, off, s[4:7], 0
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_multi_use_src_f32:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f32_e32 v2, 1.0, v3
; GFX8-NEXT:    v_max_f32_e64 v3, v2, v2 clamp
; GFX8-NEXT:    flat_store_dword v[0:1], v3
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_multi_use_src_f32:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f32_e32 v1, 1.0, v1
; GFX9-NEXT:    v_max_f32_e64 v2, v1, v1 clamp
; GFX9-NEXT:    global_store_dword v0, v2, s[0:1]
; GFX9-NEXT:    global_store_dword v[0:1], v1, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_multi_use_src_f32:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f32_e32 v1, 1.0, v1
; GFX11-NEXT:    v_max_f32_e64 v2, v1, v1 clamp
; GFX11-NEXT:    global_store_b32 v0, v2, s[0:1]
; GFX11-NEXT:    global_store_b32 v[0:1], v1, off dlc
; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr float, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr float, ptr addrspace(1) %out, i32 %tid
  %a = load float, ptr addrspace(1) %gep0
  %add = fadd float %a, 1.0
  %max = call float @llvm.maxnum.f32(float %add, float 0.0)
  %clamp = call float @llvm.minnum.f32(float %max, float 1.0)
  store float %clamp, ptr addrspace(1) %out.gep
  store volatile float %add, ptr addrspace(1) undef
  ret void
}

define amdgpu_kernel void @v_clamp_dbg_use_src_f32(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_dbg_use_src_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_add_f32_e64 v2, v2, 1.0 clamp
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_dbg_use_src_f32:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f32_e64 v2, v3, 1.0 clamp
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_dbg_use_src_f32:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f32_e64 v1, v1, 1.0 clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_dbg_use_src_f32:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f32_e64 v1, v1, 1.0 clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr float, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr float, ptr addrspace(1) %out, i32 %tid
  %a = load float, ptr addrspace(1) %gep0
  %add = fadd float %a, 1.0
  call void @llvm.dbg.value(metadata float %add, i64 0, metadata !4, metadata !9), !dbg !10
  %max = call float @llvm.maxnum.f32(float %add, float 0.0)
  %clamp = call float @llvm.minnum.f32(float %max, float 1.0)
  store float %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_neg_src_f32(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_add_neg_src_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_floor_f32_e32 v2, v2
; SI-NEXT:    v_max_f32_e64 v2, -v2, -v2 clamp
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_neg_src_f32:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_floor_f32_e32 v2, v3
; GFX8-NEXT:    v_max_f32_e64 v2, -v2, -v2 clamp
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_neg_src_f32:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_floor_f32_e32 v1, v1
; GFX9-NEXT:    v_max_f32_e64 v1, -v1, -v1 clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_neg_src_f32:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_floor_f32_e32 v1, v1
; GFX11-NEXT:    v_max_f32_e64 v1, -v1, -v1 clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr float, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr float, ptr addrspace(1) %out, i32 %tid
  %a = load float, ptr addrspace(1) %gep0
  %floor = call float @llvm.floor.f32(float %a)
  %neg.floor = fneg float %floor
  %max = call float @llvm.maxnum.f32(float %neg.floor, float 0.0)
  %clamp = call float @llvm.minnum.f32(float %max, float 1.0)
  store float %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_non_clamp_max_f32(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_non_clamp_max_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_add_f32_e32 v2, 1.0, v2
; SI-NEXT:    v_max_f32_e32 v2, 0, v2
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_non_clamp_max_f32:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f32_e32 v2, 1.0, v3
; GFX8-NEXT:    v_max_f32_e32 v2, 0, v2
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_non_clamp_max_f32:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f32_e32 v1, 1.0, v1
; GFX9-NEXT:    v_max_f32_e32 v1, 0, v1
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_non_clamp_max_f32:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f32_e32 v1, 1.0, v1
; GFX11-NEXT:    v_max_f32_e32 v1, 0, v1
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr float, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr float, ptr addrspace(1) %out, i32 %tid
  %a = load float, ptr addrspace(1) %gep0
  %add = fadd float %a, 1.0
  %max = call float @llvm.maxnum.f32(float %add, float 0.0)
  store float %max, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_f32_denormals(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #2 {
; SI-LABEL: v_clamp_add_src_f32_denormals:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_add_f32_e64 v2, v2, 1.0 clamp
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_f32_denormals:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f32_e64 v2, v3, 1.0 clamp
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_f32_denormals:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f32_e64 v1, v1, 1.0 clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_f32_denormals:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f32_e64 v1, v1, 1.0 clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr float, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr float, ptr addrspace(1) %out, i32 %tid
  %a = load float, ptr addrspace(1) %gep0
  %add = fadd float %a, 1.0
  %max = call float @llvm.maxnum.f32(float %add, float 0.0)
  %clamp = call float @llvm.minnum.f32(float %max, float 1.0)
  store float %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_f16_denorm(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_add_src_f16_denorm:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_ushort v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_cvt_f32_f16_e32 v2, v2
; SI-NEXT:    v_add_f32_e64 v2, v2, 1.0 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    buffer_store_short v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_f16_denorm:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 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, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_ushort v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f16_e64 v2, v3, 1.0 clamp
; GFX8-NEXT:    flat_store_short v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_f16_denorm:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], 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[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f16_e64 v1, v1, 1.0 clamp
; GFX9-NEXT:    global_store_short v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_f16_denorm:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_u16 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f16_e64 v1, v1, 1.0 clamp
; GFX11-NEXT:    global_store_b16 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr half, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr half, ptr addrspace(1) %out, i32 %tid
  %a = load half, ptr addrspace(1) %gep0
  %add = fadd half %a, 1.0
  %max = call half @llvm.maxnum.f16(half %add, half 0.0)
  %clamp = call half @llvm.minnum.f16(half %max, half 1.0)
  store half %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_f16_no_denormals(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #3 {
; SI-LABEL: v_clamp_add_src_f16_no_denormals:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_ushort v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_cvt_f32_f16_e32 v2, v2
; SI-NEXT:    v_add_f32_e64 v2, v2, 1.0 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    buffer_store_short v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_f16_no_denormals:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 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, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_ushort v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f16_e64 v2, v3, 1.0 clamp
; GFX8-NEXT:    flat_store_short v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_f16_no_denormals:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], 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[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f16_e64 v1, v1, 1.0 clamp
; GFX9-NEXT:    global_store_short v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_f16_no_denormals:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_u16 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f16_e64 v1, v1, 1.0 clamp
; GFX11-NEXT:    global_store_b16 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr half, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr half, ptr addrspace(1) %out, i32 %tid
  %a = load half, ptr addrspace(1) %gep0
  %add = fadd half %a, 1.0
  %max = call half @llvm.maxnum.f16(half %add, half 0.0)
  %clamp = call half @llvm.minnum.f16(half %max, half 1.0)
  store half %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_v2f32(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_add_src_v2f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_add_f32_e64 v2, v2, 1.0 clamp
; SI-NEXT:    v_add_f32_e64 v3, v3, 1.0 clamp
; SI-NEXT:    buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_v2f32:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v3, s1
; GFX8-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f32_e64 v0, v0, 1.0 clamp
; GFX8-NEXT:    v_add_f32_e64 v1, v1, 1.0 clamp
; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_v2f32:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dwordx2 v[0:1], v2, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f32_e64 v0, v0, 1.0 clamp
; GFX9-NEXT:    v_add_f32_e64 v1, v1, 1.0 clamp
; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_v2f32:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b64 v[0:1], v2, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f32_e64 v0, v0, 1.0 clamp
; GFX11-NEXT:    v_add_f32_e64 v1, v1, 1.0 clamp
; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <2 x float>, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr <2 x float>, ptr addrspace(1) %out, i32 %tid
  %a = load <2 x float>, ptr addrspace(1) %gep0
  %add = fadd <2 x float> %a, <float 1.0, float 1.0>
  %max = call <2 x float> @llvm.maxnum.v2f32(<2 x float> %add, <2 x float> zeroinitializer)
  %clamp = call <2 x float> @llvm.minnum.v2f32(<2 x float> %max, <2 x float> <float 1.0, float 1.0>)
  store <2 x float> %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_f64(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_add_src_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_add_f64 v[2:3], v[2:3], 1.0 clamp
; SI-NEXT:    buffer_store_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_f64:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v3, s1
; GFX8-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f64 v[0:1], v[0:1], 1.0 clamp
; GFX8-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_f64:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dwordx2 v[0:1], v2, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f64 v[0:1], v[0:1], 1.0 clamp
; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_f64:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b64 v[0:1], v2, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f64 v[0:1], v[0:1], 1.0 clamp
; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr double, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr double, ptr addrspace(1) %out, i32 %tid
  %a = load double, ptr addrspace(1) %gep0
  %add = fadd double %a, 1.0
  %max = call double @llvm.maxnum.f64(double %add, double 0.0)
  %clamp = call double @llvm.minnum.f64(double %max, double 1.0)
  store double %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_mac_to_mad(ptr addrspace(1) %out, ptr addrspace(1) %aptr, float %a) #0 {
; SI-LABEL: v_clamp_mac_to_mad:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_load_dword s8, s[4:5], 0xd
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_mad_f32 v3, s8, s8, v2 clamp
; SI-NEXT:    v_add_f32_e32 v2, v3, v2
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_mac_to_mad:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    s_load_dword s4, s[4:5], 0x34
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_mad_f32 v2, s4, s4, v3 clamp
; GFX8-NEXT:    v_add_f32_e32 v2, v2, v3
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_mac_to_mad:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    s_load_dword s6, s[4:5], 0x34
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_mad_f32 v2, s6, s6, v1 clamp
; GFX9-NEXT:    v_add_f32_e32 v1, v2, v1
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_mac_to_mad:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_load_b32 s4, s[4:5], 0x34
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    v_mul_f32_e64 v2, s4, s4
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f32_e64 v2, v2, v1 clamp
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_add_f32_e32 v1, v2, v1
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr float, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr float, ptr addrspace(1) %out, i32 %tid
  %b = load float, ptr addrspace(1) %gep0

  %mul = fmul float %a, %a
  %add = fadd float %mul, %b
  %max = call float @llvm.maxnum.f32(float %add, float 0.0)
  %clamp = call float @llvm.minnum.f32(float %max, float 1.0)
  %res = fadd float %clamp, %b
  store float %res, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_v2f16_denorm(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_add_src_v2f16_denorm:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT:    v_cvt_f32_f16_e32 v3, v3
; SI-NEXT:    v_cvt_f32_f16_e32 v2, v2
; SI-NEXT:    v_add_f32_e64 v3, v3, 1.0 clamp
; SI-NEXT:    v_add_f32_e64 v2, v2, 1.0 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v3, v3
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT:    v_or_b32_e32 v2, v2, v3
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_v2f16_denorm:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    v_mov_b32_e32 v4, 0x3c00
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f16_e64 v2, v3, 1.0 clamp
; GFX8-NEXT:    v_add_f16_sdwa v3, v3, v4 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-NEXT:    v_or_b32_e32 v2, v2, v3
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_v2f16_denorm:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0] clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_v2f16_denorm:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0] clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <2 x half>, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr <2 x half>, ptr addrspace(1) %out, i32 %tid
  %a = load <2 x half>, ptr addrspace(1) %gep0
  %add = fadd <2 x half> %a, <half 1.0, half 1.0>
  %max = call <2 x half> @llvm.maxnum.v2f16(<2 x half> %add, <2 x half> zeroinitializer)
  %clamp = call <2 x half> @llvm.minnum.v2f16(<2 x half> %max, <2 x half> <half 1.0, half 1.0>)
  store <2 x half> %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_v2f16_no_denormals(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #3 {
; SI-LABEL: v_clamp_add_src_v2f16_no_denormals:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT:    v_cvt_f32_f16_e32 v3, v3
; SI-NEXT:    v_cvt_f32_f16_e32 v2, v2
; SI-NEXT:    v_add_f32_e64 v3, v3, 1.0 clamp
; SI-NEXT:    v_add_f32_e64 v2, v2, 1.0 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v3, v3
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT:    v_or_b32_e32 v2, v2, v3
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_v2f16_no_denormals:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    v_mov_b32_e32 v4, 0x3c00
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f16_e64 v2, v3, 1.0 clamp
; GFX8-NEXT:    v_add_f16_sdwa v3, v3, v4 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-NEXT:    v_or_b32_e32 v2, v2, v3
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_v2f16_no_denormals:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0] clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_v2f16_no_denormals:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0] clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <2 x half>, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr <2 x half>, ptr addrspace(1) %out, i32 %tid
  %a = load <2 x half>, ptr addrspace(1) %gep0
  %add = fadd <2 x half> %a, <half 1.0, half 1.0>
  %max = call <2 x half> @llvm.maxnum.v2f16(<2 x half> %add, <2 x half> zeroinitializer)
  %clamp = call <2 x half> @llvm.minnum.v2f16(<2 x half> %max, <2 x half> <half 1.0, half 1.0>)
  store <2 x half> %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_v2f16_denorm_neg(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_add_src_v2f16_denorm_neg:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT:    v_cvt_f32_f16_e32 v3, v3
; SI-NEXT:    v_cvt_f32_f16_e32 v2, v2
; SI-NEXT:    v_add_f32_e32 v3, 1.0, v3
; SI-NEXT:    v_cvt_f16_f32_e32 v3, v3
; SI-NEXT:    v_add_f32_e32 v2, 1.0, v2
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT:    v_or_b32_e32 v2, v2, v3
; SI-NEXT:    v_xor_b32_e32 v2, 0x80008000, v2
; SI-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT:    v_cvt_f32_f16_e64 v3, v3 clamp
; SI-NEXT:    v_cvt_f32_f16_e64 v2, v2 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v3, v3
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT:    v_or_b32_e32 v2, v2, v3
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_v2f16_denorm_neg:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    v_mov_b32_e32 v4, 0x3c00
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f16_e32 v2, 1.0, v3
; GFX8-NEXT:    v_add_f16_sdwa v3, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-NEXT:    v_max_f16_sdwa v3, -v3, -v3 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX8-NEXT:    v_max_f16_e64 v2, -v2, -v2 clamp
; GFX8-NEXT:    v_or_b32_e32 v2, v2, v3
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_v2f16_denorm_neg:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0]
; GFX9-NEXT:    v_pk_max_f16 v1, v1, v1 neg_lo:[1,1] neg_hi:[1,1] clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_v2f16_denorm_neg:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0]
; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 neg_lo:[1,1] neg_hi:[1,1] clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <2 x half>, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr <2 x half>, ptr addrspace(1) %out, i32 %tid
  %a = load <2 x half>, ptr addrspace(1) %gep0
  %add = fadd <2 x half> %a, <half 1.0, half 1.0>
  %neg.add = fsub <2 x half> <half -0.0, half -0.0>, %add
  %max = call <2 x half> @llvm.maxnum.v2f16(<2 x half> %neg.add, <2 x half> zeroinitializer)
  %clamp = call <2 x half> @llvm.minnum.v2f16(<2 x half> %max, <2 x half> <half 1.0, half 1.0>)
  store <2 x half> %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_v2f16_denorm_neg_lo(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_add_src_v2f16_denorm_neg_lo:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT:    v_cvt_f32_f16_e32 v2, v2
; SI-NEXT:    v_cvt_f32_f16_e32 v3, v3
; SI-NEXT:    v_add_f32_e32 v2, 1.0, v2
; SI-NEXT:    v_add_f32_e64 v3, v3, 1.0 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v3, v3
; SI-NEXT:    v_max_f32_e64 v2, -v2, -v2 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT:    v_or_b32_e32 v2, v2, v3
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_v2f16_denorm_neg_lo:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    v_mov_b32_e32 v4, 0x3c00
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f16_e32 v2, 1.0, v3
; GFX8-NEXT:    v_add_f16_sdwa v3, v3, v4 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-NEXT:    v_max_f16_e64 v2, -v2, -v2 clamp
; GFX8-NEXT:    v_or_b32_e32 v2, v2, v3
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_v2f16_denorm_neg_lo:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0]
; GFX9-NEXT:    v_pk_max_f16 v1, v1, v1 neg_lo:[1,1] clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_v2f16_denorm_neg_lo:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0]
; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 neg_lo:[1,1] clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <2 x half>, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr <2 x half>, ptr addrspace(1) %out, i32 %tid
  %a = load <2 x half>, ptr addrspace(1) %gep0
  %add = fadd <2 x half> %a, <half 1.0, half 1.0>
  %lo = extractelement <2 x half> %add, i32 0
  %neg.lo = fsub half -0.0, %lo
  %neg.lo.add = insertelement <2 x half> %add, half %neg.lo, i32 0
  %max = call <2 x half> @llvm.maxnum.v2f16(<2 x half> %neg.lo.add, <2 x half> zeroinitializer)
  %clamp = call <2 x half> @llvm.minnum.v2f16(<2 x half> %max, <2 x half> <half 1.0, half 1.0>)
  store <2 x half> %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_v2f16_denorm_neg_hi(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_add_src_v2f16_denorm_neg_hi:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT:    v_cvt_f32_f16_e32 v3, v3
; SI-NEXT:    v_cvt_f32_f16_e32 v2, v2
; SI-NEXT:    v_add_f32_e32 v3, 1.0, v3
; SI-NEXT:    v_max_f32_e64 v3, -v3, -v3 clamp
; SI-NEXT:    v_add_f32_e64 v2, v2, 1.0 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v3, v3
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT:    v_or_b32_e32 v2, v2, v3
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_v2f16_denorm_neg_hi:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    v_mov_b32_e32 v4, 0x3c00
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f16_e64 v2, v3, 1.0 clamp
; GFX8-NEXT:    v_add_f16_sdwa v3, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-NEXT:    v_max_f16_sdwa v3, -v3, -v3 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX8-NEXT:    v_or_b32_e32 v2, v2, v3
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_v2f16_denorm_neg_hi:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0]
; GFX9-NEXT:    v_pk_max_f16 v1, v1, v1 neg_hi:[1,1] clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_v2f16_denorm_neg_hi:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0]
; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 neg_hi:[1,1] clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <2 x half>, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr <2 x half>, ptr addrspace(1) %out, i32 %tid
  %a = load <2 x half>, ptr addrspace(1) %gep0
  %add = fadd <2 x half> %a, <half 1.0, half 1.0>
  %hi = extractelement <2 x half> %add, i32 1
  %neg.hi = fsub half -0.0, %hi
  %neg.hi.add = insertelement <2 x half> %add, half %neg.hi, i32 1
  %max = call <2 x half> @llvm.maxnum.v2f16(<2 x half> %neg.hi.add, <2 x half> zeroinitializer)
  %clamp = call <2 x half> @llvm.minnum.v2f16(<2 x half> %max, <2 x half> <half 1.0, half 1.0>)
  store <2 x half> %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_clamp_add_src_v2f16_denorm_shuf(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_clamp_add_src_v2f16_denorm_shuf:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_cvt_f32_f16_e32 v3, v2
; SI-NEXT:    v_lshrrev_b32_e32 v2, 16, v2
; SI-NEXT:    v_cvt_f32_f16_e32 v2, v2
; SI-NEXT:    v_add_f32_e64 v3, v3, 1.0 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v3, v3
; SI-NEXT:    v_add_f32_e64 v2, v2, 1.0 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT:    v_or_b32_e32 v2, v2, v3
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_clamp_add_src_v2f16_denorm_shuf:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    v_mov_b32_e32 v4, 0x3c00
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f16_sdwa v2, v3, v4 clamp dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-NEXT:    v_add_f16_sdwa v3, v3, v4 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX8-NEXT:    v_or_b32_e32 v2, v2, v3
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_clamp_add_src_v2f16_denorm_shuf:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0]
; GFX9-NEXT:    v_pk_max_f16 v1, v1, v1 op_sel:[1,1] op_sel_hi:[0,0] clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_clamp_add_src_v2f16_denorm_shuf:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0]
; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 op_sel:[1,1] op_sel_hi:[0,0] clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <2 x half>, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr <2 x half>, ptr addrspace(1) %out, i32 %tid
  %a = load <2 x half>, ptr addrspace(1) %gep0
  %add = fadd <2 x half> %a, <half 1.0, half 1.0>
  %shuf = shufflevector <2 x half> %add, <2 x half> undef, <2 x i32> <i32 1, i32 0>

  %max = call <2 x half> @llvm.maxnum.v2f16(<2 x half> %shuf, <2 x half> zeroinitializer)
  %clamp = call <2 x half> @llvm.minnum.v2f16(<2 x half> %max, <2 x half> <half 1.0, half 1.0>)
  store <2 x half> %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_no_clamp_add_src_v2f16_f32_src(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_no_clamp_add_src_v2f16_f32_src:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_add_f32_e32 v2, 1.0, v2
; SI-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT:    v_cvt_f32_f16_e64 v3, v3 clamp
; SI-NEXT:    v_cvt_f32_f16_e64 v2, v2 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v3, v3
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT:    v_or_b32_e32 v2, v2, v3
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_no_clamp_add_src_v2f16_f32_src:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f32_e32 v2, 1.0, v3
; GFX8-NEXT:    v_max_f16_sdwa v3, v2, v2 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
; GFX8-NEXT:    v_max_f16_e64 v2, v2, v2 clamp
; GFX8-NEXT:    v_or_b32_e32 v2, v2, v3
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_no_clamp_add_src_v2f16_f32_src:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f32_e32 v1, 1.0, v1
; GFX9-NEXT:    v_pk_max_f16 v1, v1, v1 clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_no_clamp_add_src_v2f16_f32_src:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f32_e32 v1, 1.0, v1
; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <2 x half>, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr <2 x half>, ptr addrspace(1) %out, i32 %tid
  %a = load <2 x half>, ptr addrspace(1) %gep0
  %bc = bitcast <2 x half> %a to float
  %f32.op = fadd float %bc, 1.0
  %f32.op.cast = bitcast float %f32.op to <2 x half>
  %max = call <2 x half> @llvm.maxnum.v2f16(<2 x half> %f32.op.cast, <2 x half> zeroinitializer)
  %clamp = call <2 x half> @llvm.minnum.v2f16(<2 x half> %max, <2 x half> <half 1.0, half 1.0>)
  store <2 x half> %clamp, ptr addrspace(1) %out.gep
  ret void
}

define amdgpu_kernel void @v_no_clamp_add_packed_src_f32(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_no_clamp_add_packed_src_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_dword v2, v[0:1], s[4:7], 0 addr64
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT:    v_cvt_f32_f16_e32 v3, v3
; SI-NEXT:    v_cvt_f32_f16_e32 v2, v2
; SI-NEXT:    v_add_f32_e32 v3, 1.0, v3
; SI-NEXT:    v_cvt_f16_f32_e32 v3, v3
; SI-NEXT:    v_add_f32_e32 v2, 1.0, v2
; SI-NEXT:    v_cvt_f16_f32_e32 v2, v2
; SI-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT:    v_or_b32_e32 v2, v2, v3
; SI-NEXT:    v_max_f32_e64 v2, v2, v2 clamp
; SI-NEXT:    buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_no_clamp_add_packed_src_f32:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
; GFX8-NEXT:    v_mov_b32_e32 v4, 0x3c00
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v1, s3
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v2
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    flat_load_dword v3, v[0:1]
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f16_sdwa v2, v3, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-NEXT:    v_add_f16_e32 v3, 1.0, v3
; GFX8-NEXT:    v_or_b32_e32 v2, v3, v2
; GFX8-NEXT:    v_max_f32_e64 v2, v2, v2 clamp
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_no_clamp_add_packed_src_f32:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0]
; GFX9-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_no_clamp_add_packed_src_f32:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_b32 v1, v0, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_pk_add_f16 v1, v1, 1.0 op_sel_hi:[1,0]
; GFX11-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <2 x half>, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr float, ptr addrspace(1) %out, i32 %tid
  %a = load <2 x half>, ptr addrspace(1) %gep0
  %add = fadd <2 x half> %a, <half 1.0, half 1.0>
  %bc.add = bitcast <2 x half> %add to float
  %max = call float @llvm.maxnum.f32(float %bc.add, float 0.0)
  %clamp = call float @llvm.minnum.f32(float %max, float 1.0)
  store float %clamp, ptr addrspace(1) %out.gep
  ret void
}

; Since the high bits are zeroed, it probably would be OK in this case
; to use clamp.
define amdgpu_kernel void @v_no_clamp_add_src_v2f16_f16_src(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #0 {
; SI-LABEL: v_no_clamp_add_src_v2f16_f16_src:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, 0
; SI-NEXT:    v_lshlrev_b32_e32 v1, 1, v0
; SI-NEXT:    v_mov_b32_e32 v2, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b64 s[4:5], s[2:3]
; SI-NEXT:    buffer_load_ushort v1, v[1:2], s[4:7], 0 addr64
; SI-NEXT:    v_cvt_f16_f32_e32 v3, 0
; SI-NEXT:    s_mov_b64 s[2:3], s[6:7]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_cvt_f32_f16_e32 v1, v1
; SI-NEXT:    v_add_f32_e32 v1, 1.0, v1
; SI-NEXT:    v_cvt_f16_f32_e32 v1, v1
; SI-NEXT:    v_cvt_f32_f16_e64 v1, v1 clamp
; SI-NEXT:    v_cvt_f16_f32_e32 v4, v1
; SI-NEXT:    v_lshlrev_b32_e32 v1, 2, v0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 16, v3
; SI-NEXT:    v_or_b32_e32 v0, v4, v0
; SI-NEXT:    buffer_store_dword v0, v[1:2], s[0:3], 0 addr64
; SI-NEXT:    s_endpgm
;
; GFX8-LABEL: v_no_clamp_add_src_v2f16_f16_src:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 1, v0
; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
; GFX8-NEXT:    v_mov_b32_e32 v2, s3
; GFX8-NEXT:    v_add_u32_e32 v1, vcc, s2, v1
; GFX8-NEXT:    v_addc_u32_e32 v2, vcc, 0, v2, vcc
; GFX8-NEXT:    flat_load_ushort v2, v[1:2]
; GFX8-NEXT:    v_mov_b32_e32 v1, s1
; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v0
; GFX8-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; GFX8-NEXT:    s_waitcnt vmcnt(0)
; GFX8-NEXT:    v_add_f16_e64 v2, v2, 1.0 clamp
; GFX8-NEXT:    flat_store_dword v[0:1], v2
; GFX8-NEXT:    s_endpgm
;
; GFX9-LABEL: v_no_clamp_add_src_v2f16_f16_src:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    v_lshlrev_b32_e32 v1, 1, v0
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    global_load_ushort v1, v1, s[2:3]
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_f16_e32 v1, 1.0, v1
; GFX9-NEXT:    v_pk_max_f16 v1, v1, v1 clamp
; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX9-NEXT:    s_endpgm
;
; GFX11-LABEL: v_no_clamp_add_src_v2f16_f16_src:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
; GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT:    v_lshlrev_b32_e32 v1, 1, v0
; GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; GFX11-NEXT:    global_load_u16 v1, v1, s[2:3]
; GFX11-NEXT:    s_waitcnt vmcnt(0)
; GFX11-NEXT:    v_add_f16_e32 v1, 1.0, v1
; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-NEXT:    v_and_b32_e32 v1, 0xffff, v1
; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 clamp
; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX11-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr half, ptr addrspace(1) %aptr, i32 %tid
  %out.gep = getelementptr <2 x half>, ptr addrspace(1) %out, i32 %tid
  %a = load half, ptr addrspace(1) %gep0
  %add = fadd half %a, 1.0
  %bc = bitcast half %add to i16
  %zext = zext i16 %bc to i32
  %v2f16 = bitcast i32 %zext to <2 x half>
  %max = call <2 x half> @llvm.maxnum.v2f16(<2 x half> %v2f16, <2 x half> zeroinitializer)
  %clamp = call <2 x half> @llvm.minnum.v2f16(<2 x half> %max, <2 x half> <half 1.0, half 1.0>)
  store <2 x half> %clamp, ptr addrspace(1) %out.gep
  ret void
}

; FIXME: Worse code pre-gfx9

define <2 x half> @v_clamp_cvt_pkrtz_src_v2f16_denorm(float %a, float %b) #0 {
; SI-LABEL: v_clamp_cvt_pkrtz_src_v2f16_denorm:
; SI:       ; %bb.0:
; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT:    v_cvt_pkrtz_f16_f32_e32 v0, v0, v1
; SI-NEXT:    v_lshrrev_b32_e32 v1, 16, v0
; SI-NEXT:    v_cvt_f32_f16_e64 v0, v0 clamp
; SI-NEXT:    v_cvt_f32_f16_e64 v1, v1 clamp
; SI-NEXT:    s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_clamp_cvt_pkrtz_src_v2f16_denorm:
; GFX8:       ; %bb.0:
; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT:    v_cvt_pkrtz_f16_f32 v0, v0, v1
; GFX8-NEXT:    v_max_f16_sdwa v1, v0, v0 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
; GFX8-NEXT:    v_max_f16_e64 v0, v0, v0 clamp
; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
; GFX8-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_clamp_cvt_pkrtz_src_v2f16_denorm:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_cvt_pkrtz_f16_f32 v0, v0, v1 clamp
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-LABEL: v_clamp_cvt_pkrtz_src_v2f16_denorm:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT:    v_cvt_pk_rtz_f16_f32_e64 v0, v0, v1 clamp
; GFX11-NEXT:    s_setpc_b64 s[30:31]
  %add = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %a, float %b)
  %max = call <2 x half> @llvm.maxnum.v2f16(<2 x half> %add, <2 x half> zeroinitializer)
  %clamp = call <2 x half> @llvm.minnum.v2f16(<2 x half> %max, <2 x half> <half 1.0, half 1.0>)
  ret <2 x half> %clamp
}

declare i32 @llvm.amdgcn.workitem.id.x() #1
declare float @llvm.fabs.f32(float) #1
declare float @llvm.floor.f32(float) #1
declare float @llvm.minnum.f32(float, float) #1
declare float @llvm.maxnum.f32(float, float) #1
declare float @llvm.amdgcn.fmed3.f32(float, float, float) #1
declare double @llvm.fabs.f64(double) #1
declare double @llvm.minnum.f64(double, double) #1
declare double @llvm.maxnum.f64(double, double) #1
declare half @llvm.fabs.f16(half) #1
declare half @llvm.minnum.f16(half, half) #1
declare half @llvm.maxnum.f16(half, half) #1
declare <2 x half> @llvm.minnum.v2f16(<2 x half>, <2 x half>) #1
declare <2 x half> @llvm.maxnum.v2f16(<2 x half>, <2 x half>) #1
declare <2 x float> @llvm.minnum.v2f32(<2 x float>, <2 x float>) #1
declare <2 x float> @llvm.maxnum.v2f32(<2 x float>, <2 x float>) #1
declare <2 x half> @llvm.amdgcn.cvt.pkrtz(float, float) #1


declare void @llvm.dbg.value(metadata, i64, metadata, metadata) #1

attributes #0 = { nounwind "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
attributes #1 = { nounwind readnone }
attributes #2 = { nounwind "denormal-fp-math-f32"="ieee,ieee" }
attributes #3 = { nounwind "denormal-fp-math-f32"="ieee,ieee" "denormal-fp-math"="preserve-sign,preserve-sign" }

!llvm.dbg.cu = !{!0}
!llvm.module.flags = !{!2, !3}

!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, isOptimized: true, runtimeVersion: 0, emissionKind: NoDebug)
!1 = !DIFile(filename: "/tmp/foo.cl", directory: "/dev/null")
!2 = !{i32 2, !"Dwarf Version", i32 4}
!3 = !{i32 2, !"Debug Info Version", i32 3}
!4 = !DILocalVariable(name: "add", arg: 1, scope: !5, file: !1, line: 1)
!5 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !6, isLocal: false, isDefinition: true, scopeLine: 2, flags: DIFlagPrototyped, isOptimized: true, unit: !0)
!6 = !DISubroutineType(types: !7)
!7 = !{null, !8}
!8 = !DIBasicType(name: "float", size: 32, align: 32)
!9 = !DIExpression()
!10 = !DILocation(line: 1, column: 42, scope: !5)