; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -verify-machineinstrs < %s | FileCheck -enable-var-scope  -check-prefixes=GCN,SI %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI %s

declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone

define amdgpu_kernel void @v_uint_to_fp_i64_to_f64(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: v_uint_to_fp_i64_to_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_mov_b32_e32 v1, s3
; SI-NEXT:    v_add_i32_e32 v0, vcc, s2, v0
; SI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; SI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    v_cvt_f64_u32_e32 v[1:2], v1
; SI-NEXT:    v_cvt_f64_u32_e32 v[3:4], v0
; SI-NEXT:    v_ldexp_f64 v[0:1], v[1:2], 32
; SI-NEXT:    v_mov_b32_e32 v2, s0
; SI-NEXT:    v_add_f64 v[0:1], v[0:1], v[3:4]
; SI-NEXT:    v_mov_b32_e32 v3, s1
; SI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: v_uint_to_fp_i64_to_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_mov_b32_e32 v1, s3
; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    v_cvt_f64_u32_e32 v[1:2], v1
; VI-NEXT:    v_cvt_f64_u32_e32 v[3:4], v0
; VI-NEXT:    v_ldexp_f64 v[1:2], v[1:2], 32
; VI-NEXT:    v_add_f64 v[0:1], v[1:2], v[3:4]
; VI-NEXT:    v_mov_b32_e32 v2, s0
; VI-NEXT:    v_mov_b32_e32 v3, s1
; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; VI-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
  %gep = getelementptr i64, ptr addrspace(1) %in, i32 %tid
  %val = load i64, ptr addrspace(1) %gep, align 8
  %result = uitofp i64 %val to double
  store double %result, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @s_uint_to_fp_i64_to_f64(ptr addrspace(1) %out, i64 %in) {
; SI-LABEL: s_uint_to_fp_i64_to_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s3
; SI-NEXT:    v_cvt_f64_u32_e32 v[2:3], s2
; SI-NEXT:    v_mov_b32_e32 v4, s0
; SI-NEXT:    v_mov_b32_e32 v5, s1
; SI-NEXT:    v_ldexp_f64 v[0:1], v[0:1], 32
; SI-NEXT:    v_add_f64 v[0:1], v[0:1], v[2:3]
; SI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: s_uint_to_fp_i64_to_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s3
; VI-NEXT:    v_cvt_f64_u32_e32 v[2:3], s2
; VI-NEXT:    v_ldexp_f64 v[0:1], v[0:1], 32
; VI-NEXT:    v_add_f64 v[0:1], v[0:1], v[2:3]
; VI-NEXT:    v_mov_b32_e32 v2, s0
; VI-NEXT:    v_mov_b32_e32 v3, s1
; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; VI-NEXT:    s_endpgm
  %cast = uitofp i64 %in to double
  store double %cast, ptr addrspace(1) %out, align 8
  ret void
}

define amdgpu_kernel void @s_uint_to_fp_v2i64_to_v2f64(ptr addrspace(1) %out, <2 x i64> %in) {
; SI-LABEL: s_uint_to_fp_v2i64_to_v2f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x4
; SI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s3
; SI-NEXT:    v_cvt_f64_u32_e32 v[2:3], s1
; SI-NEXT:    v_cvt_f64_u32_e32 v[4:5], s2
; SI-NEXT:    v_cvt_f64_u32_e32 v[6:7], s0
; SI-NEXT:    v_ldexp_f64 v[0:1], v[0:1], 32
; SI-NEXT:    v_ldexp_f64 v[8:9], v[2:3], 32
; SI-NEXT:    v_add_f64 v[2:3], v[0:1], v[4:5]
; SI-NEXT:    v_add_f64 v[0:1], v[8:9], v[6:7]
; SI-NEXT:    v_mov_b32_e32 v4, s4
; SI-NEXT:    v_mov_b32_e32 v5, s5
; SI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: s_uint_to_fp_v2i64_to_v2f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x10
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s3
; VI-NEXT:    v_cvt_f64_u32_e32 v[2:3], s1
; VI-NEXT:    v_cvt_f64_u32_e32 v[6:7], s0
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; VI-NEXT:    v_ldexp_f64 v[0:1], v[0:1], 32
; VI-NEXT:    v_ldexp_f64 v[4:5], v[2:3], 32
; VI-NEXT:    v_cvt_f64_u32_e32 v[2:3], s2
; VI-NEXT:    v_add_f64 v[2:3], v[0:1], v[2:3]
; VI-NEXT:    v_add_f64 v[0:1], v[4:5], v[6:7]
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_mov_b32_e32 v5, s1
; VI-NEXT:    v_mov_b32_e32 v4, s0
; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
; VI-NEXT:    s_endpgm
  %cast = uitofp <2 x i64> %in to <2 x double>
  store <2 x double> %cast, ptr addrspace(1) %out, align 16
  ret void
}

define amdgpu_kernel void @s_uint_to_fp_v4i64_to_v4f64(ptr addrspace(1) %out, <4 x i64> %in) {
; SI-LABEL: s_uint_to_fp_v4i64_to_v4f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x8
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s11
; SI-NEXT:    v_cvt_f64_u32_e32 v[4:5], s9
; SI-NEXT:    v_cvt_f64_u32_e32 v[2:3], s10
; SI-NEXT:    v_cvt_f64_u32_e32 v[6:7], s8
; SI-NEXT:    v_cvt_f64_u32_e32 v[8:9], s15
; SI-NEXT:    v_cvt_f64_u32_e32 v[10:11], s13
; SI-NEXT:    v_ldexp_f64 v[0:1], v[0:1], 32
; SI-NEXT:    v_ldexp_f64 v[4:5], v[4:5], 32
; SI-NEXT:    v_add_f64 v[2:3], v[0:1], v[2:3]
; SI-NEXT:    v_add_f64 v[0:1], v[4:5], v[6:7]
; SI-NEXT:    v_cvt_f64_u32_e32 v[4:5], s14
; SI-NEXT:    v_cvt_f64_u32_e32 v[12:13], s12
; SI-NEXT:    v_ldexp_f64 v[6:7], v[8:9], 32
; SI-NEXT:    v_ldexp_f64 v[8:9], v[10:11], 32
; SI-NEXT:    s_add_u32 s2, s0, 16
; SI-NEXT:    s_addc_u32 s3, s1, 0
; SI-NEXT:    v_add_f64 v[6:7], v[6:7], v[4:5]
; SI-NEXT:    v_add_f64 v[4:5], v[8:9], v[12:13]
; SI-NEXT:    v_mov_b32_e32 v9, s3
; SI-NEXT:    v_mov_b32_e32 v8, s2
; SI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
; SI-NEXT:    s_nop 0
; SI-NEXT:    v_mov_b32_e32 v5, s1
; SI-NEXT:    v_mov_b32_e32 v4, s0
; SI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: s_uint_to_fp_v4i64_to_v4f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x20
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_cvt_f64_u32_e32 v[2:3], s15
; VI-NEXT:    v_cvt_f64_u32_e32 v[4:5], s13
; VI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s11
; VI-NEXT:    v_cvt_f64_u32_e32 v[6:7], s9
; VI-NEXT:    v_ldexp_f64 v[8:9], v[2:3], 32
; VI-NEXT:    v_ldexp_f64 v[4:5], v[4:5], 32
; VI-NEXT:    v_ldexp_f64 v[0:1], v[0:1], 32
; VI-NEXT:    v_ldexp_f64 v[10:11], v[6:7], 32
; VI-NEXT:    v_cvt_f64_u32_e32 v[6:7], s14
; VI-NEXT:    v_cvt_f64_u32_e32 v[12:13], s12
; VI-NEXT:    v_cvt_f64_u32_e32 v[2:3], s10
; VI-NEXT:    v_cvt_f64_u32_e32 v[14:15], s8
; VI-NEXT:    v_add_f64 v[6:7], v[8:9], v[6:7]
; VI-NEXT:    v_add_f64 v[4:5], v[4:5], v[12:13]
; VI-NEXT:    v_add_f64 v[2:3], v[0:1], v[2:3]
; VI-NEXT:    v_add_f64 v[0:1], v[10:11], v[14:15]
; VI-NEXT:    s_add_u32 s2, s0, 16
; VI-NEXT:    s_addc_u32 s3, s1, 0
; VI-NEXT:    v_mov_b32_e32 v11, s3
; VI-NEXT:    v_mov_b32_e32 v9, s1
; VI-NEXT:    v_mov_b32_e32 v10, s2
; VI-NEXT:    v_mov_b32_e32 v8, s0
; VI-NEXT:    flat_store_dwordx4 v[10:11], v[4:7]
; VI-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
; VI-NEXT:    s_endpgm
  %cast = uitofp <4 x i64> %in to <4 x double>
  store <4 x double> %cast, ptr addrspace(1) %out, align 16
  ret void
}

define amdgpu_kernel void @s_uint_to_fp_i32_to_f64(ptr addrspace(1) %out, i32 %in) {
; SI-LABEL: s_uint_to_fp_i32_to_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s2, s[4:5], 0x2
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s2
; SI-NEXT:    v_mov_b32_e32 v3, s1
; SI-NEXT:    v_mov_b32_e32 v2, s0
; SI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: s_uint_to_fp_i32_to_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s2, s[4:5], 0x8
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s2
; VI-NEXT:    v_mov_b32_e32 v3, s1
; VI-NEXT:    v_mov_b32_e32 v2, s0
; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; VI-NEXT:    s_endpgm
  %cast = uitofp i32 %in to double
  store double %cast, ptr addrspace(1) %out, align 8
  ret void
}

define amdgpu_kernel void @s_uint_to_fp_v2i32_to_v2f64(ptr addrspace(1) %out, <2 x i32> %in) {
; GCN-LABEL: s_uint_to_fp_v2i32_to_v2f64:
; GCN:       ; %bb.0:
; GCN-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    v_cvt_f64_u32_e32 v[2:3], s3
; GCN-NEXT:    v_cvt_f64_u32_e32 v[0:1], s2
; GCN-NEXT:    v_mov_b32_e32 v5, s1
; GCN-NEXT:    v_mov_b32_e32 v4, s0
; GCN-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
; GCN-NEXT:    s_endpgm
  %cast = uitofp <2 x i32> %in to <2 x double>
  store <2 x double> %cast, ptr addrspace(1) %out, align 16
  ret void
}

define amdgpu_kernel void @s_uint_to_fp_v4i32_to_v4f64(ptr addrspace(1) %out, <4 x i32> %in) {
; SI-LABEL: s_uint_to_fp_v4i32_to_v4f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x4
; SI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s0
; SI-NEXT:    v_cvt_f64_u32_e32 v[6:7], s3
; SI-NEXT:    v_cvt_f64_u32_e32 v[4:5], s2
; SI-NEXT:    s_add_u32 s0, s4, 16
; SI-NEXT:    v_cvt_f64_u32_e32 v[2:3], s1
; SI-NEXT:    s_addc_u32 s1, s5, 0
; SI-NEXT:    v_mov_b32_e32 v9, s1
; SI-NEXT:    v_mov_b32_e32 v8, s0
; SI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
; SI-NEXT:    s_nop 0
; SI-NEXT:    v_mov_b32_e32 v4, s4
; SI-NEXT:    v_mov_b32_e32 v5, s5
; SI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: s_uint_to_fp_v4i32_to_v4f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x10
; VI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s0
; VI-NEXT:    v_cvt_f64_u32_e32 v[6:7], s3
; VI-NEXT:    v_cvt_f64_u32_e32 v[4:5], s2
; VI-NEXT:    s_add_u32 s0, s4, 16
; VI-NEXT:    v_cvt_f64_u32_e32 v[2:3], s1
; VI-NEXT:    s_addc_u32 s1, s5, 0
; VI-NEXT:    v_mov_b32_e32 v9, s1
; VI-NEXT:    v_mov_b32_e32 v8, s0
; VI-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
; VI-NEXT:    s_nop 0
; VI-NEXT:    v_mov_b32_e32 v4, s4
; VI-NEXT:    v_mov_b32_e32 v5, s5
; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
; VI-NEXT:    s_endpgm
  %cast = uitofp <4 x i32> %in to <4 x double>
  store <4 x double> %cast, ptr addrspace(1) %out, align 16
  ret void
}

; We can't fold the SGPRs into v_cndmask_b32_e32, because it already
; uses an SGPR (implicit vcc).
define amdgpu_kernel void @uint_to_fp_i1_to_f64(ptr addrspace(1) %out, i32 %in) {
; SI-LABEL: uint_to_fp_i1_to_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s2, s[4:5], 0x2
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_eq_u32 s2, 0
; SI-NEXT:    s_cselect_b32 s2, 0x3ff00000, 0
; SI-NEXT:    v_mov_b32_e32 v3, s1
; SI-NEXT:    v_mov_b32_e32 v1, s2
; SI-NEXT:    v_mov_b32_e32 v2, s0
; SI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uint_to_fp_i1_to_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s2, s[4:5], 0x8
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_eq_u32 s2, 0
; VI-NEXT:    s_cselect_b32 s2, 0x3ff00000, 0
; VI-NEXT:    v_mov_b32_e32 v3, s1
; VI-NEXT:    v_mov_b32_e32 v1, s2
; VI-NEXT:    v_mov_b32_e32 v2, s0
; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; VI-NEXT:    s_endpgm
  %cmp = icmp eq i32 %in, 0
  %fp = uitofp i1 %cmp to double
  store double %fp, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @uint_to_fp_i1_to_f64_load(ptr addrspace(1) %out, i1 %in) {
; SI-LABEL: uint_to_fp_i1_to_f64_load:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s2, s[4:5], 0x2
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_bitcmp1_b32 s2, 0
; SI-NEXT:    s_cselect_b64 s[2:3], -1, 0
; SI-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[2:3]
; SI-NEXT:    v_cvt_f64_u32_e32 v[0:1], v0
; SI-NEXT:    v_mov_b32_e32 v3, s1
; SI-NEXT:    v_mov_b32_e32 v2, s0
; SI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uint_to_fp_i1_to_f64_load:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s2, s[4:5], 0x8
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_bitcmp1_b32 s2, 0
; VI-NEXT:    s_cselect_b64 s[2:3], -1, 0
; VI-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[2:3]
; VI-NEXT:    v_cvt_f64_u32_e32 v[0:1], v0
; VI-NEXT:    v_mov_b32_e32 v3, s1
; VI-NEXT:    v_mov_b32_e32 v2, s0
; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; VI-NEXT:    s_endpgm
  %fp = uitofp i1 %in to double
  store double %fp, ptr addrspace(1) %out, align 8
  ret void
}

define amdgpu_kernel void @s_uint_to_fp_i8_to_f64(ptr addrspace(1) %out, i8 %in) {
; SI-LABEL: s_uint_to_fp_i8_to_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s2, s[4:5], 0x2
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_and_b32 s2, s2, 0xff
; SI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s2
; SI-NEXT:    v_mov_b32_e32 v3, s1
; SI-NEXT:    v_mov_b32_e32 v2, s0
; SI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: s_uint_to_fp_i8_to_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s2, s[4:5], 0x8
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_and_b32 s2, s2, 0xff
; VI-NEXT:    v_cvt_f64_u32_e32 v[0:1], s2
; VI-NEXT:    v_mov_b32_e32 v3, s1
; VI-NEXT:    v_mov_b32_e32 v2, s0
; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; VI-NEXT:    s_endpgm
  %fp = uitofp i8 %in to double
  store double %fp, ptr addrspace(1) %out
  ret void
}

; FIXME: Worse on VI
define double @v_uint_to_fp_i8_to_f64(i8 %in) {
; SI-LABEL: v_uint_to_fp_i8_to_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT:    v_and_b32_e32 v0, 0xff, v0
; SI-NEXT:    v_cvt_f64_u32_e32 v[0:1], v0
; SI-NEXT:    s_setpc_b64 s[30:31]
;
; VI-LABEL: v_uint_to_fp_i8_to_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT:    v_mov_b32_e32 v1, 0xffff
; VI-NEXT:    v_and_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT:    v_cvt_f64_u32_e32 v[0:1], v0
; VI-NEXT:    s_setpc_b64 s[30:31]
  %fp = uitofp i8 %in to double
  ret double %fp
}

define amdgpu_kernel void @s_select_uint_to_fp_i1_vals_f64(ptr addrspace(1) %out, i32 %in) {
; SI-LABEL: s_select_uint_to_fp_i1_vals_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s2, s[4:5], 0x2
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_eq_u32 s2, 0
; SI-NEXT:    s_cselect_b32 s2, 0x3ff00000, 0
; SI-NEXT:    v_mov_b32_e32 v3, s1
; SI-NEXT:    v_mov_b32_e32 v1, s2
; SI-NEXT:    v_mov_b32_e32 v2, s0
; SI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: s_select_uint_to_fp_i1_vals_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s2, s[4:5], 0x8
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_eq_u32 s2, 0
; VI-NEXT:    s_cselect_b32 s2, 0x3ff00000, 0
; VI-NEXT:    v_mov_b32_e32 v3, s1
; VI-NEXT:    v_mov_b32_e32 v1, s2
; VI-NEXT:    v_mov_b32_e32 v2, s0
; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; VI-NEXT:    s_endpgm
  %cmp = icmp eq i32 %in, 0
  %select = select i1 %cmp, double 1.0, double 0.0
  store double %select, ptr addrspace(1) %out, align 8
  ret void
}

define void @v_select_uint_to_fp_i1_vals_f64(ptr addrspace(1) %out, i32 %in) {
; GCN-LABEL: v_select_uint_to_fp_i1_vals_f64:
; GCN:       ; %bb.0:
; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT:    v_mov_b32_e32 v4, 0x3ff00000
; GCN-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
; GCN-NEXT:    v_mov_b32_e32 v3, 0
; GCN-NEXT:    v_cndmask_b32_e32 v4, 0, v4, vcc
; GCN-NEXT:    flat_store_dwordx2 v[0:1], v[3:4]
; GCN-NEXT:    s_waitcnt vmcnt(0)
; GCN-NEXT:    s_setpc_b64 s[30:31]
  %cmp = icmp eq i32 %in, 0
  %select = select i1 %cmp, double 1.0, double 0.0
  store double %select, ptr addrspace(1) %out, align 8
  ret void
}

define amdgpu_kernel void @s_select_uint_to_fp_i1_vals_i64(ptr addrspace(1) %out, i32 %in) {
; SI-LABEL: s_select_uint_to_fp_i1_vals_i64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s2, s[4:5], 0x2
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_eq_u32 s2, 0
; SI-NEXT:    s_cselect_b32 s2, 0x3ff00000, 0
; SI-NEXT:    v_mov_b32_e32 v3, s1
; SI-NEXT:    v_mov_b32_e32 v1, s2
; SI-NEXT:    v_mov_b32_e32 v2, s0
; SI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: s_select_uint_to_fp_i1_vals_i64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s2, s[4:5], 0x8
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_eq_u32 s2, 0
; VI-NEXT:    s_cselect_b32 s2, 0x3ff00000, 0
; VI-NEXT:    v_mov_b32_e32 v3, s1
; VI-NEXT:    v_mov_b32_e32 v1, s2
; VI-NEXT:    v_mov_b32_e32 v2, s0
; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; VI-NEXT:    s_endpgm
  %cmp = icmp eq i32 %in, 0
  %select = select i1 %cmp, i64 u0x3ff0000000000000, i64 0
  store i64 %select, ptr addrspace(1) %out, align 8
  ret void
}

define void @v_select_uint_to_fp_i1_vals_i64(ptr addrspace(1) %out, i32 %in) {
; GCN-LABEL: v_select_uint_to_fp_i1_vals_i64:
; GCN:       ; %bb.0:
; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT:    v_mov_b32_e32 v4, 0x3ff00000
; GCN-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
; GCN-NEXT:    v_mov_b32_e32 v3, 0
; GCN-NEXT:    v_cndmask_b32_e32 v4, 0, v4, vcc
; GCN-NEXT:    flat_store_dwordx2 v[0:1], v[3:4]
; GCN-NEXT:    s_waitcnt vmcnt(0)
; GCN-NEXT:    s_setpc_b64 s[30:31]
  %cmp = icmp eq i32 %in, 0
  %select = select i1 %cmp, i64 u0x3ff0000000000000, i64 0
  store i64 %select, ptr addrspace(1) %out, align 8
  ret void
}

; TODO: This should swap the selected order / invert the compare and do it.
define amdgpu_kernel void @s_swap_select_uint_to_fp_i1_vals_f64(ptr addrspace(1) %out, i32 %in) {
; SI-LABEL: s_swap_select_uint_to_fp_i1_vals_f64:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dword s2, s[4:5], 0x2
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; SI-NEXT:    v_mov_b32_e32 v0, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_eq_u32 s2, 0
; SI-NEXT:    s_cselect_b32 s2, 0, 0x3ff00000
; SI-NEXT:    v_mov_b32_e32 v3, s1
; SI-NEXT:    v_mov_b32_e32 v1, s2
; SI-NEXT:    v_mov_b32_e32 v2, s0
; SI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; SI-NEXT:    s_endpgm
;
; VI-LABEL: s_swap_select_uint_to_fp_i1_vals_f64:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dword s2, s[4:5], 0x8
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
; VI-NEXT:    v_mov_b32_e32 v0, 0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_eq_u32 s2, 0
; VI-NEXT:    s_cselect_b32 s2, 0, 0x3ff00000
; VI-NEXT:    v_mov_b32_e32 v3, s1
; VI-NEXT:    v_mov_b32_e32 v1, s2
; VI-NEXT:    v_mov_b32_e32 v2, s0
; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; VI-NEXT:    s_endpgm
  %cmp = icmp eq i32 %in, 0
  %select = select i1 %cmp, double 0.0, double 1.0
  store double %select, ptr addrspace(1) %out, align 8
  ret void
}

define void @v_swap_select_uint_to_fp_i1_vals_f64(ptr addrspace(1) %out, i32 %in) {
; GCN-LABEL: v_swap_select_uint_to_fp_i1_vals_f64:
; GCN:       ; %bb.0:
; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT:    v_mov_b32_e32 v4, 0x3ff00000
; GCN-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
; GCN-NEXT:    v_mov_b32_e32 v3, 0
; GCN-NEXT:    v_cndmask_b32_e64 v4, v4, 0, vcc
; GCN-NEXT:    flat_store_dwordx2 v[0:1], v[3:4]
; GCN-NEXT:    s_waitcnt vmcnt(0)
; GCN-NEXT:    s_setpc_b64 s[30:31]
  %cmp = icmp eq i32 %in, 0
  %select = select i1 %cmp, double 0.0, double 1.0
  store double %select, ptr addrspace(1) %out, align 8
  ret void
}