; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn -mcpu=verde -amdgpu-early-ifcvt=0 -machine-sink-split-probability-threshold=0 -structurizecfg-skip-uniform-regions -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -mtriple=amdgcn -mcpu=tonga -mattr=-flat-for-global -amdgpu-early-ifcvt=0 -machine-sink-split-probability-threshold=0 -structurizecfg-skip-uniform-regions -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=VI %s

define amdgpu_kernel void @uniform_if_scc(i32 %cond, ptr addrspace(1) %out) {
; SI-LABEL: uniform_if_scc:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dword s0, s[4:5], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_eq_u32 s0, 0
; SI-NEXT:    s_mov_b32 s0, 0
; SI-NEXT:    s_cbranch_scc1 .LBB0_2
; SI-NEXT:  ; %bb.1: ; %else
; SI-NEXT:    s_mov_b32 s0, 1
; SI-NEXT:  .LBB0_2: ; %done
; SI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0xb
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    v_mov_b32_e32 v0, s0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_scc:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dword s0, s[4:5], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_eq_u32 s0, 0
; VI-NEXT:    s_mov_b32 s0, 0
; VI-NEXT:    s_cbranch_scc1 .LBB0_2
; VI-NEXT:  ; %bb.1: ; %else
; VI-NEXT:    s_mov_b32 s0, 1
; VI-NEXT:  .LBB0_2: ; %done
; VI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2c
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    v_mov_b32_e32 v0, s0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
entry:
  %cmp0 = icmp eq i32 %cond, 0
  br i1 %cmp0, label %if, label %else

if:
  br label %done

else:
  br label %done

done:
  %value = phi i32 [0, %if], [1, %else]
  store i32 %value, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @uniform_if_vcc(float %cond, ptr addrspace(1) %out) {
; SI-LABEL: uniform_if_vcc:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dword s1, s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s0, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cmp_eq_f32_e64 s[2:3], s1, 0
; SI-NEXT:    s_and_b64 vcc, exec, s[2:3]
; SI-NEXT:    s_cbranch_vccnz .LBB1_2
; SI-NEXT:  ; %bb.1: ; %else
; SI-NEXT:    s_mov_b32 s0, 1
; SI-NEXT:  .LBB1_2: ; %done
; SI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0xb
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    v_mov_b32_e32 v0, s0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_vcc:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dword s1, s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s0, 0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_cmp_eq_f32_e64 s[2:3], s1, 0
; VI-NEXT:    s_and_b64 vcc, exec, s[2:3]
; VI-NEXT:    s_cbranch_vccnz .LBB1_2
; VI-NEXT:  ; %bb.1: ; %else
; VI-NEXT:    s_mov_b32 s0, 1
; VI-NEXT:  .LBB1_2: ; %done
; VI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2c
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    v_mov_b32_e32 v0, s0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
entry:
  %cmp0 = fcmp oeq float %cond, 0.0
  br i1 %cmp0, label %if, label %else

if:
  br label %done

else:
  br label %done

done:
  %value = phi i32 [0, %if], [1, %else]
  store i32 %value, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @uniform_if_swap_br_targets_scc(i32 %cond, ptr addrspace(1) %out) {
; SI-LABEL: uniform_if_swap_br_targets_scc:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dword s0, s[4:5], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_lg_u32 s0, 0
; SI-NEXT:    s_mov_b32 s0, 0
; SI-NEXT:    s_cbranch_scc1 .LBB2_2
; SI-NEXT:  ; %bb.1: ; %else
; SI-NEXT:    s_mov_b32 s0, 1
; SI-NEXT:  .LBB2_2: ; %done
; SI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0xb
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    v_mov_b32_e32 v0, s0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_swap_br_targets_scc:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dword s0, s[4:5], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_lg_u32 s0, 0
; VI-NEXT:    s_mov_b32 s0, 0
; VI-NEXT:    s_cbranch_scc1 .LBB2_2
; VI-NEXT:  ; %bb.1: ; %else
; VI-NEXT:    s_mov_b32 s0, 1
; VI-NEXT:  .LBB2_2: ; %done
; VI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2c
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    v_mov_b32_e32 v0, s0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
entry:
  %cmp0 = icmp eq i32 %cond, 0
  br i1 %cmp0, label %else, label %if

if:
  br label %done

else:
  br label %done

done:
  %value = phi i32 [0, %if], [1, %else]
  store i32 %value, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @uniform_if_swap_br_targets_vcc(float %cond, ptr addrspace(1) %out) {
; SI-LABEL: uniform_if_swap_br_targets_vcc:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dword s1, s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s0, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cmp_neq_f32_e64 s[2:3], s1, 0
; SI-NEXT:    s_and_b64 vcc, exec, s[2:3]
; SI-NEXT:    s_cbranch_vccnz .LBB3_2
; SI-NEXT:  ; %bb.1: ; %else
; SI-NEXT:    s_mov_b32 s0, 1
; SI-NEXT:  .LBB3_2: ; %done
; SI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0xb
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    v_mov_b32_e32 v0, s0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_swap_br_targets_vcc:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dword s1, s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s0, 0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_cmp_neq_f32_e64 s[2:3], s1, 0
; VI-NEXT:    s_and_b64 vcc, exec, s[2:3]
; VI-NEXT:    s_cbranch_vccnz .LBB3_2
; VI-NEXT:  ; %bb.1: ; %else
; VI-NEXT:    s_mov_b32 s0, 1
; VI-NEXT:  .LBB3_2: ; %done
; VI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2c
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    v_mov_b32_e32 v0, s0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
entry:
  %cmp0 = fcmp oeq float %cond, 0.0
  br i1 %cmp0, label %else, label %if

if:
  br label %done

else:
  br label %done

done:
  %value = phi i32 [0, %if], [1, %else]
  store i32 %value, ptr addrspace(1) %out
  ret void
}

; Using a floating-point value in an integer compare will cause the compare to
; be selected for the SALU and then later moved to the VALU.
define amdgpu_kernel void @uniform_if_move_valu(ptr addrspace(1) %out, float %a) {
; SI-LABEL: uniform_if_move_valu:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dword s0, s[4:5], 0xb
; SI-NEXT:    v_mov_b32_e32 v0, 0x41200000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e32 v0, s0, v0
; SI-NEXT:    v_cmp_ne_u32_e32 vcc, 5, v0
; SI-NEXT:    s_cbranch_vccnz .LBB4_2
; SI-NEXT:  ; %bb.1: ; %if
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 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:  .LBB4_2: ; %endif
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_move_valu:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dword s0, s[4:5], 0x2c
; VI-NEXT:    v_mov_b32_e32 v0, 0x41200000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e32 v0, s0, v0
; VI-NEXT:    v_cmp_ne_u32_e32 vcc, 5, v0
; VI-NEXT:    s_cbranch_vccnz .LBB4_2
; VI-NEXT:  ; %bb.1: ; %if
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 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:  .LBB4_2: ; %endif
; VI-NEXT:    s_endpgm
entry:
  %a.0 = fadd float %a, 10.0
  %cond = bitcast float %a.0 to i32
  %cmp = icmp eq i32 %cond, 5
  br i1 %cmp, label %if, label %endif

if:
  store i32 0, ptr addrspace(1) %out
  br label %endif

endif:
  ret void
}

; Using a floating-point value in an integer compare will cause the compare to
; be selected for the SALU and then later moved to the VALU.
define amdgpu_kernel void @uniform_if_move_valu_commute(ptr addrspace(1) %out, float %a) {
; SI-LABEL: uniform_if_move_valu_commute:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dword s0, s[4:5], 0xb
; SI-NEXT:    v_mov_b32_e32 v0, 0x41200000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_add_f32_e32 v0, s0, v0
; SI-NEXT:    v_cmp_gt_u32_e32 vcc, 6, v0
; SI-NEXT:    s_cbranch_vccnz .LBB5_2
; SI-NEXT:  ; %bb.1: ; %if
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 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:  .LBB5_2: ; %endif
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_move_valu_commute:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dword s0, s[4:5], 0x2c
; VI-NEXT:    v_mov_b32_e32 v0, 0x41200000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_add_f32_e32 v0, s0, v0
; VI-NEXT:    v_cmp_gt_u32_e32 vcc, 6, v0
; VI-NEXT:    s_cbranch_vccnz .LBB5_2
; VI-NEXT:  ; %bb.1: ; %if
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 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:  .LBB5_2: ; %endif
; VI-NEXT:    s_endpgm
entry:
  %a.0 = fadd float %a, 10.0
  %cond = bitcast float %a.0 to i32
  %cmp = icmp ugt i32 %cond, 5
  br i1 %cmp, label %if, label %endif

if:
  store i32 0, ptr addrspace(1) %out
  br label %endif

endif:
  ret void
}


define amdgpu_kernel void @uniform_if_else_ret(ptr addrspace(1) nocapture %out, i32 %a) {
; SI-LABEL: uniform_if_else_ret:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dword s2, s[4:5], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_lg_u32 s2, 0
; SI-NEXT:    s_cbranch_scc0 .LBB6_2
; SI-NEXT:  ; %bb.1: ; %if.else
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 2
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
; SI-NEXT:  .LBB6_2: ; %if.then
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 1
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_else_ret:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dword s2, s[4:5], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_lg_u32 s2, 0
; VI-NEXT:    s_cbranch_scc0 .LBB6_2
; VI-NEXT:  ; %bb.1: ; %if.else
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 2
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
; VI-NEXT:  .LBB6_2: ; %if.then
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 1
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
entry:
  %cmp = icmp eq i32 %a, 0
  br i1 %cmp, label %if.then, label %if.else

if.then:                                          ; preds = %entry
  store i32 1, ptr addrspace(1) %out
  br label %if.end

if.else:                                          ; preds = %entry
  store i32 2, ptr addrspace(1) %out
  br label %if.end

if.end:                                           ; preds = %if.else, %if.then
  ret void
}

define amdgpu_kernel void @uniform_if_else(ptr addrspace(1) nocapture %out0, ptr addrspace(1) nocapture %out1, i32 %a) {
; SI-LABEL: uniform_if_else:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dword s6, s[4:5], 0xd
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_lg_u32 s6, 0
; SI-NEXT:    s_cbranch_scc0 .LBB7_2
; SI-NEXT:  ; %bb.1: ; %if.else
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    v_mov_b32_e32 v0, 2
; SI-NEXT:    s_branch .LBB7_3
; SI-NEXT:  .LBB7_2: ; %if.then
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    v_mov_b32_e32 v0, 1
; SI-NEXT:  .LBB7_3: ; %if.end
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_mov_b32 s4, s2
; SI-NEXT:    s_mov_b32 s5, s3
; SI-NEXT:    s_waitcnt expcnt(0)
; SI-NEXT:    v_mov_b32_e32 v0, 3
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_else:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dword s6, s[4:5], 0x34
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_lg_u32 s6, 0
; VI-NEXT:    s_cbranch_scc0 .LBB7_2
; VI-NEXT:  ; %bb.1: ; %if.else
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    v_mov_b32_e32 v0, 2
; VI-NEXT:    s_branch .LBB7_3
; VI-NEXT:  .LBB7_2: ; %if.then
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    v_mov_b32_e32 v0, 1
; VI-NEXT:  .LBB7_3: ; %if.end
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_mov_b32 s4, s2
; VI-NEXT:    s_mov_b32 s5, s3
; VI-NEXT:    v_mov_b32_e32 v0, 3
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
entry:
  %cmp = icmp eq i32 %a, 0
  br i1 %cmp, label %if.then, label %if.else

if.then:                                          ; preds = %entry
  store i32 1, ptr addrspace(1) %out0
  br label %if.end

if.else:                                          ; preds = %entry
  store i32 2, ptr addrspace(1) %out0
  br label %if.end

if.end:                                           ; preds = %if.else, %if.then
  store i32 3, ptr addrspace(1) %out1
  ret void
}

define amdgpu_kernel void @icmp_2_users(ptr addrspace(1) %out, i32 %cond) {
; SI-LABEL: icmp_2_users:
; SI:       ; %bb.0: ; %main_body
; SI-NEXT:    s_load_dword s2, s[4:5], 0xb
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_gt_i32 s2, 0
; SI-NEXT:    s_cselect_b64 s[0:1], -1, 0
; SI-NEXT:    s_cmp_lt_i32 s2, 1
; SI-NEXT:    s_cbranch_scc1 .LBB8_2
; SI-NEXT:  ; %bb.1: ; %IF
; SI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    v_cndmask_b32_e64 v0, 0, -1, s[0:1]
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:  .LBB8_2: ; %ENDIF
; SI-NEXT:    s_endpgm
;
; VI-LABEL: icmp_2_users:
; VI:       ; %bb.0: ; %main_body
; VI-NEXT:    s_load_dword s2, s[4:5], 0x2c
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_gt_i32 s2, 0
; VI-NEXT:    s_cselect_b64 s[0:1], -1, 0
; VI-NEXT:    s_cmp_lt_i32 s2, 1
; VI-NEXT:    s_cbranch_scc1 .LBB8_2
; VI-NEXT:  ; %bb.1: ; %IF
; VI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    v_cndmask_b32_e64 v0, 0, -1, s[0:1]
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:  .LBB8_2: ; %ENDIF
; VI-NEXT:    s_endpgm
main_body:
  %0 = icmp sgt i32 %cond, 0
  %1 = sext i1 %0 to i32
  br i1 %0, label %IF, label %ENDIF

IF:
  store i32 %1, ptr addrspace(1) %out
  br label %ENDIF

ENDIF:                                            ; preds = %IF, %main_body
  ret void
}

define amdgpu_kernel void @icmp_users_different_blocks(i32 %cond0, i32 %cond1, ptr addrspace(1) %out) {
; SI-LABEL: icmp_users_different_blocks:
; SI:       ; %bb.0: ; %bb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_lt_i32 s0, 1
; SI-NEXT:    s_cbranch_scc1 .LBB9_2
; SI-NEXT:  ; %bb.1: ; %bb2
; SI-NEXT:    s_cmp_gt_i32 s1, 0
; SI-NEXT:    s_cselect_b64 s[0:1], -1, 0
; SI-NEXT:    s_and_b64 vcc, exec, s[0:1]
; SI-NEXT:    s_cbranch_vccz .LBB9_3
; SI-NEXT:  .LBB9_2: ; %bb9
; SI-NEXT:    s_endpgm
; SI-NEXT:  .LBB9_3: ; %bb7
; SI-NEXT:    v_cndmask_b32_e64 v1, 0, 1, s[0:1]
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xb
; SI-NEXT:    v_sub_i32_e32 v0, vcc, v0, v1
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: icmp_users_different_blocks:
; VI:       ; %bb.0: ; %bb
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_lt_i32 s0, 1
; VI-NEXT:    s_cbranch_scc1 .LBB9_2
; VI-NEXT:  ; %bb.1: ; %bb2
; VI-NEXT:    s_cmp_gt_i32 s1, 0
; VI-NEXT:    s_cselect_b64 s[0:1], -1, 0
; VI-NEXT:    s_and_b64 vcc, exec, s[0:1]
; VI-NEXT:    s_cbranch_vccz .LBB9_3
; VI-NEXT:  .LBB9_2: ; %bb9
; VI-NEXT:    s_endpgm
; VI-NEXT:  .LBB9_3: ; %bb7
; VI-NEXT:    v_cndmask_b32_e64 v1, 0, 1, s[0:1]
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x2c
; VI-NEXT:    v_sub_u32_e32 v0, vcc, v0, v1
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
bb:
  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x() #0
  %cmp0 = icmp sgt i32 %cond0, 0
  br i1 %cmp0, label %bb2, label %bb9

bb2:                                              ; preds = %bb
  %cmp1 = icmp sgt i32 %cond1, 0
  %tmp2 = sext i1 %cmp1 to i32
  %tmp3 = add i32 %tmp2, %tmp
  br i1 %cmp1, label %bb9, label %bb7

bb7:                                              ; preds = %bb5
  store i32 %tmp3, ptr addrspace(1) %out
  br label %bb9

bb9:                                              ; preds = %bb8, %bb4
  ret void
}

define amdgpu_kernel void @uniform_loop(ptr addrspace(1) %out, i32 %a) {
; SI-LABEL: uniform_loop:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dword s0, s[4:5], 0xb
; SI-NEXT:  .LBB10_1: ; %loop
; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_add_i32 s0, s0, -1
; SI-NEXT:    s_cmp_lg_u32 s0, 0
; SI-NEXT:    s_cbranch_scc1 .LBB10_1
; SI-NEXT:  ; %bb.2: ; %done
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_loop:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dword s0, s[4:5], 0x2c
; VI-NEXT:  .LBB10_1: ; %loop
; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_add_i32 s0, s0, -1
; VI-NEXT:    s_cmp_lg_u32 s0, 0
; VI-NEXT:    s_cbranch_scc1 .LBB10_1
; VI-NEXT:  ; %bb.2: ; %done
; VI-NEXT:    s_endpgm
entry:
  br label %loop

loop:
  %i = phi i32 [0, %entry], [%i.i, %loop]
  %i.i = add i32 %i, 1
  %cmp = icmp eq i32 %a, %i.i
  br i1 %cmp, label %done, label %loop

done:
  ret void
}

; Test uniform and divergent.

define amdgpu_kernel void @uniform_inside_divergent(ptr addrspace(1) %out, i32 %cond) {
; SI-LABEL: uniform_inside_divergent:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    v_cmp_gt_u32_e32 vcc, 16, v0
; SI-NEXT:    s_and_saveexec_b64 s[0:1], vcc
; SI-NEXT:    s_cbranch_execz .LBB11_2
; SI-NEXT:  ; %bb.1: ; %if
; SI-NEXT:    s_load_dword s6, s[4:5], 0xb
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 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:    s_cmp_lg_u32 s6, 0
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_cbranch_scc0 .LBB11_3
; SI-NEXT:  .LBB11_2: ; %endif
; SI-NEXT:    s_endpgm
; SI-NEXT:  .LBB11_3: ; %if_uniform
; SI-NEXT:    s_waitcnt expcnt(0)
; SI-NEXT:    v_mov_b32_e32 v0, 1
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_inside_divergent:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    v_cmp_gt_u32_e32 vcc, 16, v0
; VI-NEXT:    s_and_saveexec_b64 s[0:1], vcc
; VI-NEXT:    s_cbranch_execz .LBB11_2
; VI-NEXT:  ; %bb.1: ; %if
; VI-NEXT:    s_load_dword s6, s[4:5], 0x2c
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 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:    s_cmp_lg_u32 s6, 0
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_cbranch_scc0 .LBB11_3
; VI-NEXT:  .LBB11_2: ; %endif
; VI-NEXT:    s_endpgm
; VI-NEXT:  .LBB11_3: ; %if_uniform
; VI-NEXT:    v_mov_b32_e32 v0, 1
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
entry:
  %tid = call i32 @llvm.amdgcn.workitem.id.x() #0
  %d_cmp = icmp ult i32 %tid, 16
  br i1 %d_cmp, label %if, label %endif

if:
  store i32 0, ptr addrspace(1) %out
  %u_cmp = icmp eq i32 %cond, 0
  br i1 %u_cmp, label %if_uniform, label %endif

if_uniform:
  store i32 1, ptr addrspace(1) %out
  br label %endif

endif:
  ret void
}

define amdgpu_kernel void @divergent_inside_uniform(ptr addrspace(1) %out, i32 %cond) {
; SI-LABEL: divergent_inside_uniform:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dword s0, s[4:5], 0xb
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_lg_u32 s0, 0
; SI-NEXT:    s_cbranch_scc0 .LBB12_2
; SI-NEXT:  .LBB12_1: ; %endif
; SI-NEXT:    s_endpgm
; SI-NEXT:  .LBB12_2: ; %if
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    v_cmp_gt_u32_e32 vcc, 16, v0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v1, off, s[0:3], 0
; SI-NEXT:    s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT:    s_cbranch_execz .LBB12_1
; SI-NEXT:  ; %bb.3: ; %if_uniform
; SI-NEXT:    v_mov_b32_e32 v0, 1
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: divergent_inside_uniform:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dword s0, s[4:5], 0x2c
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_lg_u32 s0, 0
; VI-NEXT:    s_cbranch_scc0 .LBB12_2
; VI-NEXT:  .LBB12_1: ; %endif
; VI-NEXT:    s_endpgm
; VI-NEXT:  .LBB12_2: ; %if
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v1, 0
; VI-NEXT:    v_cmp_gt_u32_e32 vcc, 16, v0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v1, off, s[0:3], 0
; VI-NEXT:    s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT:    s_cbranch_execz .LBB12_1
; VI-NEXT:  ; %bb.3: ; %if_uniform
; VI-NEXT:    v_mov_b32_e32 v0, 1
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
entry:
  %u_cmp = icmp eq i32 %cond, 0
  br i1 %u_cmp, label %if, label %endif

if:
  store i32 0, ptr addrspace(1) %out
  %tid = call i32 @llvm.amdgcn.workitem.id.x() #0
  %d_cmp = icmp ult i32 %tid, 16
  br i1 %d_cmp, label %if_uniform, label %endif

if_uniform:
  store i32 1, ptr addrspace(1) %out
  br label %endif

endif:
  ret void
}

define amdgpu_kernel void @divergent_if_uniform_if(ptr addrspace(1) %out, i32 %cond) {
; SI-LABEL: divergent_if_uniform_if:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
; SI-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
; SI-NEXT:    s_and_saveexec_b64 s[6:7], vcc
; SI-NEXT:    s_cbranch_execz .LBB13_2
; SI-NEXT:  ; %bb.1: ; %if
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v0, 1
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:  .LBB13_2: ; %endif
; SI-NEXT:    s_or_b64 exec, exec, s[6:7]
; SI-NEXT:    s_load_dword s2, s[4:5], 0xb
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_lg_u32 s2, 0
; SI-NEXT:    s_cbranch_scc0 .LBB13_4
; SI-NEXT:  ; %bb.3: ; %exit
; SI-NEXT:    s_endpgm
; SI-NEXT:  .LBB13_4: ; %if_uniform
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    s_waitcnt expcnt(0)
; SI-NEXT:    v_mov_b32_e32 v0, 2
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: divergent_if_uniform_if:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; VI-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
; VI-NEXT:    s_and_saveexec_b64 s[6:7], vcc
; VI-NEXT:    s_cbranch_execz .LBB13_2
; VI-NEXT:  ; %bb.1: ; %if
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 1
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:  .LBB13_2: ; %endif
; VI-NEXT:    s_or_b64 exec, exec, s[6:7]
; VI-NEXT:    s_load_dword s2, s[4:5], 0x2c
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_lg_u32 s2, 0
; VI-NEXT:    s_cbranch_scc0 .LBB13_4
; VI-NEXT:  ; %bb.3: ; %exit
; VI-NEXT:    s_endpgm
; VI-NEXT:  .LBB13_4: ; %if_uniform
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v0, 2
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:    s_endpgm
entry:
  %tid = call i32 @llvm.amdgcn.workitem.id.x() #0
  %d_cmp = icmp eq i32 %tid, 0
  br i1 %d_cmp, label %if, label %endif

if:
  store i32 1, ptr addrspace(1) %out
  br label %endif

endif:
  %u_cmp = icmp eq i32 %cond, 0
  br i1 %u_cmp, label %if_uniform, label %exit

if_uniform:
  store i32 2, ptr addrspace(1) %out
  br label %exit

exit:
  ret void
}

; The condition of the branches in the two blocks are
; uniform. MachineCSE replaces the 2nd condition with the inverse of
; the first, leaving an scc use in a different block than it was
; defed.

define amdgpu_kernel void @cse_uniform_condition_different_blocks(i32 %cond, ptr addrspace(1) %out) {
; SI-LABEL: cse_uniform_condition_different_blocks:
; SI:       ; %bb.0: ; %bb
; SI-NEXT:    s_load_dword s0, s[4:5], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_cmp_lt_i32 s0, 1
; SI-NEXT:    s_cbranch_scc1 .LBB14_2
; SI-NEXT:  ; %bb.1: ; %bb2
; SI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xb
; SI-NEXT:    s_mov_b32 s3, 0xf000
; SI-NEXT:    s_mov_b32 s2, -1
; SI-NEXT:    v_mov_b32_e32 v1, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_load_dword v0, off, s[0:3], 0 glc
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    buffer_store_dword v1, off, s[0:3], 0
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT:  .LBB14_2: ; %bb9
; SI-NEXT:    s_endpgm
;
; VI-LABEL: cse_uniform_condition_different_blocks:
; VI:       ; %bb.0: ; %bb
; VI-NEXT:    s_load_dword s0, s[4:5], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_lt_i32 s0, 1
; VI-NEXT:    s_cbranch_scc1 .LBB14_2
; VI-NEXT:  ; %bb.1: ; %bb2
; VI-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x2c
; VI-NEXT:    s_mov_b32 s3, 0xf000
; VI-NEXT:    s_mov_b32 s2, -1
; VI-NEXT:    v_mov_b32_e32 v1, 0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_load_dword v0, off, s[0:3], 0 glc
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    buffer_store_dword v1, off, s[0:3], 0
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
; VI-NEXT:  .LBB14_2: ; %bb9
; VI-NEXT:    s_endpgm
bb:
  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x() #0
  %tmp1 = icmp sgt i32 %cond, 0
  br i1 %tmp1, label %bb2, label %bb9

bb2:                                              ; preds = %bb
  %tmp3 = load volatile i32, ptr addrspace(1) poison
  store volatile i32 0, ptr addrspace(1) poison
  %tmp9 = icmp sle i32 %cond, 0
  br i1 %tmp9, label %bb9, label %bb7

bb7:                                              ; preds = %bb5
  store i32 %tmp3, ptr addrspace(1) %out
  br label %bb9

bb9:                                              ; preds = %bb8, %bb4
  ret void
}

; Fall-through to the else
define amdgpu_kernel void @uniform_if_scc_i64_eq(i64 %cond, ptr addrspace(1) %out) {
; SI-LABEL: uniform_if_scc_i64_eq:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cmp_eq_u64_e64 s[4:5], s[0:1], 0
; SI-NEXT:    s_mov_b32 s0, 0
; SI-NEXT:    s_and_b64 vcc, exec, s[4:5]
; SI-NEXT:    s_cbranch_vccnz .LBB15_2
; SI-NEXT:  ; %bb.1: ; %else
; SI-NEXT:    s_mov_b32 s0, 1
; SI-NEXT:  .LBB15_2: ; %done
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s4, s2
; SI-NEXT:    s_mov_b32 s5, s3
; SI-NEXT:    v_mov_b32_e32 v0, s0
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_scc_i64_eq:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_eq_u64 s[0:1], 0
; VI-NEXT:    s_mov_b32 s0, 0
; VI-NEXT:    s_cbranch_scc1 .LBB15_2
; VI-NEXT:  ; %bb.1: ; %else
; VI-NEXT:    s_mov_b32 s0, 1
; VI-NEXT:  .LBB15_2: ; %done
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s4, s2
; VI-NEXT:    s_mov_b32 s5, s3
; VI-NEXT:    v_mov_b32_e32 v0, s0
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
entry:
  %cmp0 = icmp eq i64 %cond, 0
  br i1 %cmp0, label %if, label %else

if:
  br label %done

else:
  br label %done

done:
  %value = phi i32 [0, %if], [1, %else]
  store i32 %value, ptr addrspace(1) %out
  ret void
}

; Fall-through to the else
define amdgpu_kernel void @uniform_if_scc_i64_ne(i64 %cond, ptr addrspace(1) %out) {
; SI-LABEL: uniform_if_scc_i64_ne:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cmp_ne_u64_e64 s[4:5], s[0:1], 0
; SI-NEXT:    s_mov_b32 s0, 0
; SI-NEXT:    s_and_b64 vcc, exec, s[4:5]
; SI-NEXT:    s_cbranch_vccnz .LBB16_2
; SI-NEXT:  ; %bb.1: ; %else
; SI-NEXT:    s_mov_b32 s0, 1
; SI-NEXT:  .LBB16_2: ; %done
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s4, s2
; SI-NEXT:    s_mov_b32 s5, s3
; SI-NEXT:    v_mov_b32_e32 v0, s0
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_scc_i64_ne:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_cmp_lg_u64 s[0:1], 0
; VI-NEXT:    s_mov_b32 s0, 0
; VI-NEXT:    s_cbranch_scc1 .LBB16_2
; VI-NEXT:  ; %bb.1: ; %else
; VI-NEXT:    s_mov_b32 s0, 1
; VI-NEXT:  .LBB16_2: ; %done
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s4, s2
; VI-NEXT:    s_mov_b32 s5, s3
; VI-NEXT:    v_mov_b32_e32 v0, s0
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
entry:
  %cmp0 = icmp ne i64 %cond, 0
  br i1 %cmp0, label %if, label %else

if:
  br label %done

else:
  br label %done

done:
  %value = phi i32 [0, %if], [1, %else]
  store i32 %value, ptr addrspace(1) %out
  ret void
}

; Fall-through to the else
define amdgpu_kernel void @uniform_if_scc_i64_sgt(i64 %cond, ptr addrspace(1) %out) {
; SI-LABEL: uniform_if_scc_i64_sgt:
; SI:       ; %bb.0: ; %entry
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cmp_gt_i64_e64 s[4:5], s[0:1], 0
; SI-NEXT:    s_mov_b32 s0, 0
; SI-NEXT:    s_and_b64 vcc, exec, s[4:5]
; SI-NEXT:    s_cbranch_vccnz .LBB17_2
; SI-NEXT:  ; %bb.1: ; %else
; SI-NEXT:    s_mov_b32 s0, 1
; SI-NEXT:  .LBB17_2: ; %done
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s4, s2
; SI-NEXT:    s_mov_b32 s5, s3
; SI-NEXT:    v_mov_b32_e32 v0, s0
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: uniform_if_scc_i64_sgt:
; VI:       ; %bb.0: ; %entry
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_cmp_gt_i64_e64 s[4:5], s[0:1], 0
; VI-NEXT:    s_mov_b32 s0, 0
; VI-NEXT:    s_and_b64 vcc, exec, s[4:5]
; VI-NEXT:    s_cbranch_vccnz .LBB17_2
; VI-NEXT:  ; %bb.1: ; %else
; VI-NEXT:    s_mov_b32 s0, 1
; VI-NEXT:  .LBB17_2: ; %done
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s4, s2
; VI-NEXT:    s_mov_b32 s5, s3
; VI-NEXT:    v_mov_b32_e32 v0, s0
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
entry:
  %cmp0 = icmp sgt i64 %cond, 0
  br i1 %cmp0, label %if, label %else

if:
  br label %done

else:
  br label %done

done:
  %value = phi i32 [0, %if], [1, %else]
  store i32 %value, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @move_to_valu_i64_eq(ptr addrspace(1) %out) {
; SI-LABEL: move_to_valu_i64_eq:
; SI:       ; %bb.0:
; SI-NEXT:    s_mov_b32 m0, -1
; SI-NEXT:    ds_read_b64 v[0:1], v0
; SI-NEXT:    s_mov_b32 s0, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cmp_eq_u64_e32 vcc, 0, v[0:1]
; SI-NEXT:    s_cbranch_vccnz .LBB18_2
; SI-NEXT:  ; %bb.1: ; %else
; SI-NEXT:    s_mov_b32 s0, 1
; SI-NEXT:  .LBB18_2: ; %done
; SI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    v_mov_b32_e32 v0, s0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: move_to_valu_i64_eq:
; VI:       ; %bb.0:
; VI-NEXT:    s_mov_b32 m0, -1
; VI-NEXT:    ds_read_b64 v[0:1], v0
; VI-NEXT:    s_mov_b32 s0, 0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_cmp_eq_u64_e32 vcc, 0, v[0:1]
; VI-NEXT:    s_cbranch_vccnz .LBB18_2
; VI-NEXT:  ; %bb.1: ; %else
; VI-NEXT:    s_mov_b32 s0, 1
; VI-NEXT:  .LBB18_2: ; %done
; VI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    v_mov_b32_e32 v0, s0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %cond = load volatile i64, ptr addrspace(3) poison
  %cmp0 = icmp eq i64 %cond, 0
  br i1 %cmp0, label %if, label %else

if:
  br label %done

else:
  br label %done

done:
  %value = phi i32 [0, %if], [1, %else]
  store i32 %value, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @move_to_valu_i64_ne(ptr addrspace(1) %out) {
; SI-LABEL: move_to_valu_i64_ne:
; SI:       ; %bb.0:
; SI-NEXT:    s_mov_b32 m0, -1
; SI-NEXT:    ds_read_b64 v[0:1], v0
; SI-NEXT:    s_mov_b32 s0, 0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[0:1]
; SI-NEXT:    s_cbranch_vccnz .LBB19_2
; SI-NEXT:  ; %bb.1: ; %else
; SI-NEXT:    s_mov_b32 s0, 1
; SI-NEXT:  .LBB19_2: ; %done
; SI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    v_mov_b32_e32 v0, s0
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: move_to_valu_i64_ne:
; VI:       ; %bb.0:
; VI-NEXT:    s_mov_b32 m0, -1
; VI-NEXT:    ds_read_b64 v[0:1], v0
; VI-NEXT:    s_mov_b32 s0, 0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[0:1]
; VI-NEXT:    s_cbranch_vccnz .LBB19_2
; VI-NEXT:  ; %bb.1: ; %else
; VI-NEXT:    s_mov_b32 s0, 1
; VI-NEXT:  .LBB19_2: ; %done
; VI-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    v_mov_b32_e32 v0, s0
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %cond = load volatile i64, ptr addrspace(3) poison
  %cmp0 = icmp ne i64 %cond, 0
  br i1 %cmp0, label %if, label %else

if:
  br label %done

else:
  br label %done

done:
  %value = phi i32 [0, %if], [1, %else]
  store i32 %value, ptr addrspace(1) %out
  ret void
}

define void @move_to_valu_vgpr_operand_phi(ptr addrspace(3) %out) {
; SI-LABEL: move_to_valu_vgpr_operand_phi:
; SI:       ; %bb.0: ; %bb0
; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT:    v_add_i32_e32 v0, vcc, 28, v0
; SI-NEXT:    v_mov_b32_e32 v1, 1
; SI-NEXT:    s_and_b64 vcc, exec, 0
; SI-NEXT:    s_mov_b32 m0, -1
; SI-NEXT:    s_branch .LBB20_2
; SI-NEXT:  .LBB20_1: ; %bb3
; SI-NEXT:    ; in Loop: Header=BB20_2 Depth=1
; SI-NEXT:    v_add_i32_e64 v0, s[4:5], 8, v0
; SI-NEXT:  .LBB20_2: ; %bb1
; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
; SI-NEXT:    ;;#ASMSTART
; SI-NEXT:    ; def s4
; SI-NEXT:    ;;#ASMEND
; SI-NEXT:    s_cmp_lg_u32 s4, 0
; SI-NEXT:    s_cbranch_scc1 .LBB20_1
; SI-NEXT:  ; %bb.3: ; %bb2
; SI-NEXT:    ; in Loop: Header=BB20_2 Depth=1
; SI-NEXT:    ds_write_b32 v0, v1
; SI-NEXT:    s_mov_b64 vcc, vcc
; SI-NEXT:    s_cbranch_vccz .LBB20_1
; SI-NEXT:  ; %bb.4: ; %DummyReturnBlock
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_setpc_b64 s[30:31]
;
; VI-LABEL: move_to_valu_vgpr_operand_phi:
; VI:       ; %bb.0: ; %bb0
; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT:    v_add_u32_e32 v0, vcc, 28, v0
; VI-NEXT:    v_mov_b32_e32 v1, 1
; VI-NEXT:    s_and_b64 vcc, exec, 0
; VI-NEXT:    s_mov_b32 m0, -1
; VI-NEXT:    s_branch .LBB20_2
; VI-NEXT:  .LBB20_1: ; %bb3
; VI-NEXT:    ; in Loop: Header=BB20_2 Depth=1
; VI-NEXT:    v_add_u32_e64 v0, s[4:5], 8, v0
; VI-NEXT:  .LBB20_2: ; %bb1
; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
; VI-NEXT:    ;;#ASMSTART
; VI-NEXT:    ; def s4
; VI-NEXT:    ;;#ASMEND
; VI-NEXT:    s_cmp_lg_u32 s4, 0
; VI-NEXT:    s_cbranch_scc1 .LBB20_1
; VI-NEXT:  ; %bb.3: ; %bb2
; VI-NEXT:    ; in Loop: Header=BB20_2 Depth=1
; VI-NEXT:    ds_write_b32 v0, v1
; VI-NEXT:    s_mov_b64 vcc, vcc
; VI-NEXT:    s_cbranch_vccz .LBB20_1
; VI-NEXT:  ; %bb.4: ; %DummyReturnBlock
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_setpc_b64 s[30:31]
bb0:
  br label %bb1

bb1:                                              ; preds = %bb3, %bb0
  %tmp0 = phi i32 [ 8, %bb0 ], [ %tmp4, %bb3 ]
  %tmp1 = add nsw i32 %tmp0, -1
  %tmp2 = getelementptr inbounds i32, ptr addrspace(3) %out, i32 %tmp1
  %cond = call i32 asm "; def $0","=s"()
  %cmp = icmp eq i32 %cond, 0
  br i1 %cmp, label %bb2, label %bb3

bb2:                                              ; preds = %bb1
  store volatile i32 1, ptr addrspace(3) %tmp2, align 4
  br label %bb3

bb3:                                              ; preds = %bb2, %bb1
  %tmp4 = add nsw i32 %tmp0, 2
  br label %bb1
}

declare i32 @llvm.amdgcn.workitem.id.x() #0

attributes #0 = { nounwind readnone }
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GCN: {{.*}}