; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN:  llc -amdgpu-scalarize-global-loads=false  -mtriple=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SI %s
; RUN:  llc -amdgpu-scalarize-global-loads=false  -mtriple=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI %s

; Use a 64-bit value with lo bits that can be represented as an inline constant
define amdgpu_kernel void @i64_imm_inline_lo(ptr addrspace(1) %out) {
; SI-LABEL: i64_imm_inline_lo:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 5
; SI-NEXT:    v_mov_b32_e32 v1, 0x12345678
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: i64_imm_inline_lo:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 5
; VI-NEXT:    v_mov_b32_e32 v1, 0x12345678
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
entry:
  store i64 1311768464867721221, ptr addrspace(1) %out ; 0x1234567800000005
  ret void
}

; Use a 64-bit value with hi bits that can be represented as an inline constant
define amdgpu_kernel void @i64_imm_inline_hi(ptr addrspace(1) %out) {
; SI-LABEL: i64_imm_inline_hi:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0x12345678
; SI-NEXT:    v_mov_b32_e32 v1, 5
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: i64_imm_inline_hi:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0x12345678
; VI-NEXT:    v_mov_b32_e32 v1, 5
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
entry:
  store i64 21780256376, ptr addrspace(1) %out ; 0x0000000512345678
  ret void
}

define amdgpu_kernel void @store_imm_neg_0.0_i64(ptr addrspace(1) %out) {
; SI-LABEL: store_imm_neg_0.0_i64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_bfrev_b32_e32 v1, 1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_imm_neg_0.0_i64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_bfrev_b32_e32 v1, 1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store i64 -9223372036854775808, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_neg_0.0_i32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_neg_0.0_i32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_bfrev_b32_e32 v0, 1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_neg_0.0_i32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_bfrev_b32_e32 v0, 1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store i32 -2147483648, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_0.0_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_0.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_0.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float 0.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_imm_neg_0.0_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_imm_neg_0.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_bfrev_b32_e32 v0, 1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_imm_neg_0.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_bfrev_b32_e32 v0, 1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float -0.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_0.5_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_0.5_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0.5
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_0.5_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0.5
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float 0.5, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_m_0.5_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_m_0.5_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, -0.5
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_m_0.5_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, -0.5
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float -0.5, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_1.0_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_1.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 1.0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_1.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 1.0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float 1.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_m_1.0_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_m_1.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, -1.0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_m_1.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, -1.0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float -1.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_2.0_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_2.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 2.0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_2.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 2.0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float 2.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_m_2.0_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_m_2.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, -2.0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_m_2.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, -2.0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float -2.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_4.0_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_4.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 4.0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_4.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 4.0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float 4.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_m_4.0_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_m_4.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, -4.0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_m_4.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, -4.0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float -4.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_inv_2pi_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_inv_2pi_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0x3e22f983
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_inv_2pi_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0.15915494
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float 0x3FC45F3060000000, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_m_inv_2pi_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_m_inv_2pi_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0xbe22f983
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_m_inv_2pi_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0xbe22f983
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float 0xBFC45F3060000000, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_literal_imm_f32(ptr addrspace(1) %out) {
; SI-LABEL: store_literal_imm_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0x45800000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_literal_imm_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0x45800000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store float 4096.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_0.0_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_0.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, 0
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_0.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, 0
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, 0.0
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_0.5_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_0.5_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, 0.5
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_0.5_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, 0.5
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, 0.5
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_0.5_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_neg_0.5_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, -0.5
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_0.5_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, -0.5
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, -0.5
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_1.0_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_1.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, 1.0
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_1.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, 1.0
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, 1.0
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_1.0_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_neg_1.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, -1.0
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_1.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, -1.0
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, -1.0
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_2.0_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_2.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, 2.0
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_2.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, 2.0
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, 2.0
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_2.0_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_neg_2.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, -2.0
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_2.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, -2.0
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, -2.0
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_4.0_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_4.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, 4.0
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_4.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, 4.0
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, 4.0
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_4.0_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_neg_4.0_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, -4.0
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_4.0_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, -4.0
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, -4.0
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @commute_add_inline_imm_0.5_f32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: commute_add_inline_imm_0.5_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s10, s6
; SI-NEXT:    s_mov_b32 s11, s7
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b32 s8, s2
; SI-NEXT:    s_mov_b32 s9, s3
; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_add_f32_e32 v0, 0.5, v0
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: commute_add_inline_imm_0.5_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s10, s6
; VI-NEXT:    s_mov_b32 s11, s7
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_mov_b32 s8, s2
; VI-NEXT:    s_mov_b32 s9, s3
; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    v_add_f32_e32 v0, 0.5, v0
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %x = load float, ptr addrspace(1) %in
  %y = fadd float %x, 0.5
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @commute_add_literal_f32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: commute_add_literal_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s10, s6
; SI-NEXT:    s_mov_b32 s11, s7
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b32 s8, s2
; SI-NEXT:    s_mov_b32 s9, s3
; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_add_f32_e32 v0, 0x44800000, v0
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: commute_add_literal_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s10, s6
; VI-NEXT:    s_mov_b32 s11, s7
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_mov_b32 s8, s2
; VI-NEXT:    s_mov_b32 s9, s3
; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    v_add_f32_e32 v0, 0x44800000, v0
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %x = load float, ptr addrspace(1) %in
  %y = fadd float %x, 1024.0
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_1_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_1_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, 1
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_1_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, 1
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, 0x36a0000000000000
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_2_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_2_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, 2
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_2_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, 2
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, 0x36b0000000000000
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_16_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_16_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, 16
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_16_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, 16
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, 0x36e0000000000000
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_1_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_neg_1_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_add_i32 s4, s4, -1
; SI-NEXT:    v_mov_b32_e32 v0, s4
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_1_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_add_i32 s4, s4, -1
; VI-NEXT:    v_mov_b32_e32 v0, s4
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %xbc = bitcast float %x to i32
  %y = add i32 %xbc, -1
  %ybc = bitcast i32 %y to float
  store float %ybc, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_2_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_neg_2_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_add_i32 s4, s4, -2
; SI-NEXT:    v_mov_b32_e32 v0, s4
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_2_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_add_i32 s4, s4, -2
; VI-NEXT:    v_mov_b32_e32 v0, s4
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %xbc = bitcast float %x to i32
  %y = add i32 %xbc, -2
  %ybc = bitcast i32 %y to float
  store float %ybc, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_16_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_neg_16_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_add_i32 s4, s4, -16
; SI-NEXT:    v_mov_b32_e32 v0, s4
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_16_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_add_i32 s4, s4, -16
; VI-NEXT:    v_mov_b32_e32 v0, s4
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %xbc = bitcast float %x to i32
  %y = add i32 %xbc, -16
  %ybc = bitcast i32 %y to float
  store float %ybc, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_63_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_63_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, 63
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_63_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, 63
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, 0x36ff800000000000
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_64_f32(ptr addrspace(1) %out, float %x) {
; SI-LABEL: add_inline_imm_64_f32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e64 v0, s4, 64
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_64_f32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e64 v0, s4, 64
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd float %x, 0x3700000000000000
  store float %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_0.0_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_0.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], 0
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_0.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 0
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0.0
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_0.5_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_0.5_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], 0.5
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_0.5_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 0.5
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0.5
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_0.5_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_neg_0.5_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], -0.5
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_0.5_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], -0.5
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, -0.5
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_1.0_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_1.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], 1.0
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_1.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 1.0
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 1.0
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_1.0_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_neg_1.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], -1.0
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_1.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], -1.0
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, -1.0
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_2.0_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_2.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], 2.0
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_2.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 2.0
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 2.0
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_2.0_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_neg_2.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], -2.0
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_2.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], -2.0
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, -2.0
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_4.0_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_4.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], 4.0
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_4.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 4.0
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 4.0
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_4.0_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_neg_4.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], -4.0
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_4.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], -4.0
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, -4.0
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_inv_2pi_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_inv_2pi_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    v_mov_b32_e32 v0, 0x6dc9c882
; SI-NEXT:    v_mov_b32_e32 v1, 0x3fc45f30
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], v[0:1]
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_inv_2pi_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 0.15915494309189532
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0x3fc45f306dc9c882
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_m_inv_2pi_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_m_inv_2pi_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    v_mov_b32_e32 v0, 0x6dc9c882
; SI-NEXT:    v_mov_b32_e32 v1, 0xbfc45f30
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], v[0:1]
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_m_inv_2pi_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    v_mov_b32_e32 v0, 0x6dc9c882
; VI-NEXT:    v_mov_b32_e32 v1, 0xbfc45f30
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], v[0:1]
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0xbfc45f306dc9c882
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_1_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_1_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], 1
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_1_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 1
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0x0000000000000001
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_2_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_2_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], 2
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_2_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 2
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0x0000000000000002
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_16_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_16_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], 16
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_16_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 16
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0x0000000000000010
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_1_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_neg_1_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    v_mov_b32_e32 v0, -1
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v1, v0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_1_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    v_mov_b32_e32 v0, -1
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v1, v0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0xffffffffffffffff
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_2_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_neg_2_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, -2
; SI-NEXT:    v_mov_b32_e32 v1, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_2_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, -2
; VI-NEXT:    v_mov_b32_e32 v1, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0xfffffffffffffffe
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_neg_16_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_neg_16_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, -16
; SI-NEXT:    v_mov_b32_e32 v1, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_neg_16_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, -16
; VI-NEXT:    v_mov_b32_e32 v1, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0xfffffffffffffff0
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_63_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_63_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], 63
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_63_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 63
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0x000000000000003F
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @add_inline_imm_64_f64(ptr addrspace(1) %out, [8 x i32], double %x) {
; SI-LABEL: add_inline_imm_64_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x13
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f64 v[0:1], s[2:3], 64
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: add_inline_imm_64_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x4c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f64 v[0:1], s[2:3], 64
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  %y = fadd double %x, 0x0000000000000040
  store double %y, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_0.0_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_0.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v1, v0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_0.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v1, v0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double 0.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_literal_imm_neg_0.0_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_literal_imm_neg_0.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_bfrev_b32_e32 v1, 1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_literal_imm_neg_0.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_bfrev_b32_e32 v1, 1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double -0.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_0.5_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_0.5_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_mov_b32_e32 v1, 0x3fe00000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_0.5_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_mov_b32_e32 v1, 0x3fe00000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double 0.5, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_m_0.5_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_m_0.5_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_mov_b32_e32 v1, 0xbfe00000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_m_0.5_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_mov_b32_e32 v1, 0xbfe00000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double -0.5, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_1.0_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_1.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_mov_b32_e32 v1, 0x3ff00000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_1.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_mov_b32_e32 v1, 0x3ff00000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double 1.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_m_1.0_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_m_1.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_mov_b32_e32 v1, 0xbff00000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_m_1.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_mov_b32_e32 v1, 0xbff00000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double -1.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_2.0_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_2.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_mov_b32_e32 v1, 2.0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_2.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_mov_b32_e32 v1, 2.0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double 2.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_m_2.0_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_m_2.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_mov_b32_e32 v1, -2.0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_m_2.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_mov_b32_e32 v1, -2.0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double -2.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_4.0_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_4.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_mov_b32_e32 v1, 0x40100000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_4.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_mov_b32_e32 v1, 0x40100000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double 4.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_m_4.0_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_m_4.0_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_mov_b32_e32 v1, 0xc0100000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_m_4.0_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_mov_b32_e32 v1, 0xc0100000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double -4.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inv_2pi_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inv_2pi_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0x6dc9c882
; SI-NEXT:    v_mov_b32_e32 v1, 0x3fc45f30
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inv_2pi_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0x6dc9c882
; VI-NEXT:    v_mov_b32_e32 v1, 0x3fc45f30
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double 0x3fc45f306dc9c882, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_inline_imm_m_inv_2pi_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_inline_imm_m_inv_2pi_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0x6dc9c882
; SI-NEXT:    v_mov_b32_e32 v1, 0xbfc45f30
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_inline_imm_m_inv_2pi_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0x6dc9c882
; VI-NEXT:    v_mov_b32_e32 v1, 0xbfc45f30
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double 0xbfc45f306dc9c882, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @store_literal_imm_f64(ptr addrspace(1) %out) {
; SI-LABEL: store_literal_imm_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    v_mov_b32_e32 v1, 0x40b00000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: store_literal_imm_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    v_mov_b32_e32 v1, 0x40b00000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
; VI-NEXT:    s_endpgm
  store double 4096.0, ptr addrspace(1) %out
  ret void
}

define amdgpu_vs void @literal_folding(float %arg) {
; GCN-LABEL: literal_folding:
; GCN:       ; %bb.0: ; %main_body
; GCN-NEXT:    v_mul_f32_e32 v1, 0x3f4353f8, v0
; GCN-NEXT:    v_mul_f32_e32 v0, 0xbf4353f8, v0
; GCN-NEXT:    exp pos0 v1, v1, v0, v0 done
; GCN-NEXT:    s_endpgm
main_body:
  %tmp = fmul float %arg, 0x3FE86A7F00000000
  %tmp1 = fmul float %arg, 0xBFE86A7F00000000
  call void @llvm.amdgcn.exp.f32(i32 12, i32 15, float %tmp, float %tmp, float %tmp1, float %tmp1, i1 true, i1 false) #0
  ret void
}

declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #0

attributes #0 = { nounwind }