src/llvm-project/llvm/test/CodeGen/AMDGPU/GlobalISel/divergence-temporal-divergent-i1.mir

# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 4
# RUN: llc -global-isel -mtriple=amdgcn-mesa-amdpal -mcpu=gfx1010 -run-pass=amdgpu-global-isel-divergence-lowering %s -o - | FileCheck -check-prefix=GFX10 %s

---
name: temporal_divergent_i1_phi
legalized: true
tracksRegLiveness: true
body: |
  ; GFX10-LABEL: name: temporal_divergent_i1_phi
  ; GFX10: bb.0:
  ; GFX10-NEXT:   successors: %bb.1(0x80000000)
  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
  ; GFX10-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
  ; GFX10-NEXT:   [[C:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
  ; GFX10-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT: bb.1:
  ; GFX10-NEXT:   successors: %bb.2(0x04000000), %bb.1(0x7c000000)
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT:   [[PHI:%[0-9]+]]:_(s32) = G_PHI %7(s32), %bb.1, [[C1]](s32), %bb.0
  ; GFX10-NEXT:   [[PHI1:%[0-9]+]]:_(s32) = G_PHI [[C1]](s32), %bb.0, %9(s32), %bb.1
  ; GFX10-NEXT:   [[PHI2:%[0-9]+]]:_(s1) = G_PHI [[C]](s1), %bb.0, %11(s1), %bb.1
  ; GFX10-NEXT:   [[C2:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
  ; GFX10-NEXT:   [[XOR:%[0-9]+]]:_(s1) = G_XOR [[PHI2]], [[C2]]
  ; GFX10-NEXT:   [[UITOFP:%[0-9]+]]:_(s32) = G_UITOFP [[PHI1]](s32)
  ; GFX10-NEXT:   [[FCMP:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[UITOFP]](s32), [[COPY]]
  ; GFX10-NEXT:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
  ; GFX10-NEXT:   [[ADD:%[0-9]+]]:_(s32) = G_ADD [[PHI1]], [[C3]]
  ; GFX10-NEXT:   [[INTRINSIC_CONVERGENT:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.if.break), [[FCMP]](s1), [[PHI]](s32)
  ; GFX10-NEXT:   SI_LOOP [[INTRINSIC_CONVERGENT]](s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX10-NEXT:   G_BR %bb.2
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT: bb.2:
  ; GFX10-NEXT:   [[PHI3:%[0-9]+]]:_(s1) = G_PHI [[PHI2]](s1), %bb.1
  ; GFX10-NEXT:   [[PHI4:%[0-9]+]]:_(s32) = G_PHI [[INTRINSIC_CONVERGENT]](s32), %bb.1
  ; GFX10-NEXT:   G_INTRINSIC_CONVERGENT_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[PHI4]](s32)
  ; GFX10-NEXT:   [[C4:%[0-9]+]]:_(s32) = G_FCONSTANT float 0.000000e+00
  ; GFX10-NEXT:   [[C5:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
  ; GFX10-NEXT:   [[SELECT:%[0-9]+]]:_(s32) = G_SELECT [[PHI3]](s1), [[C5]], [[C4]]
  ; GFX10-NEXT:   G_STORE [[SELECT]](s32), [[MV]](p0) :: (store (s32))
  ; GFX10-NEXT:   SI_RETURN
  bb.0:
    successors: %bb.1(0x80000000)
    liveins: $vgpr0, $vgpr1, $vgpr2

    %0:_(s32) = COPY $vgpr0
    %1:_(s32) = COPY $vgpr1
    %2:_(s32) = COPY $vgpr2
    %3:_(p0) = G_MERGE_VALUES %1(s32), %2(s32)
    %4:_(s1) = G_CONSTANT i1 true
    %5:_(s32) = G_CONSTANT i32 0

  bb.1:
    successors: %bb.2(0x04000000), %bb.1(0x7c000000)

    %6:_(s32) = G_PHI %7(s32), %bb.1, %5(s32), %bb.0
    %8:_(s32) = G_PHI %5(s32), %bb.0, %9(s32), %bb.1
    %10:_(s1) = G_PHI %4(s1), %bb.0, %11(s1), %bb.1
    %12:_(s1) = G_CONSTANT i1 true
    %11:_(s1) = G_XOR %10, %12
    %13:_(s32) = G_UITOFP %8(s32)
    %14:_(s1) = G_FCMP floatpred(ogt), %13(s32), %0
    %15:_(s32) = G_CONSTANT i32 1
    %9:_(s32) = G_ADD %8, %15
    %7:sreg_32_xm0_xexec(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.if.break), %14(s1), %6(s32)
    SI_LOOP %7(s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
    G_BR %bb.2

  bb.2:
    %16:_(s1) = G_PHI %10(s1), %bb.1
    %17:_(s32) = G_PHI %7(s32), %bb.1
    G_INTRINSIC_CONVERGENT_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %17(s32)
    %18:_(s32) = G_FCONSTANT float 0.000000e+00
    %19:_(s32) = G_FCONSTANT float 1.000000e+00
    %20:_(s32) = G_SELECT %16(s1), %19, %18
    G_STORE %20(s32), %3(p0) :: (store (s32))
    SI_RETURN
...

---
name: temporal_divergent_i1_non_phi
legalized: true
tracksRegLiveness: true
body: |
  ; GFX10-LABEL: name: temporal_divergent_i1_non_phi
  ; GFX10: bb.0:
  ; GFX10-NEXT:   successors: %bb.1(0x80000000)
  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
  ; GFX10-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
  ; GFX10-NEXT:   [[C:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
  ; GFX10-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT: bb.1:
  ; GFX10-NEXT:   successors: %bb.2(0x04000000), %bb.1(0x7c000000)
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT:   [[PHI:%[0-9]+]]:_(s32) = G_PHI %7(s32), %bb.1, [[C1]](s32), %bb.0
  ; GFX10-NEXT:   [[PHI1:%[0-9]+]]:_(s32) = G_PHI [[C1]](s32), %bb.0, %9(s32), %bb.1
  ; GFX10-NEXT:   [[PHI2:%[0-9]+]]:_(s1) = G_PHI [[C]](s1), %bb.0, %11(s1), %bb.1
  ; GFX10-NEXT:   [[C2:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
  ; GFX10-NEXT:   [[XOR:%[0-9]+]]:_(s1) = G_XOR [[PHI2]], [[C2]]
  ; GFX10-NEXT:   [[UITOFP:%[0-9]+]]:_(s32) = G_UITOFP [[PHI1]](s32)
  ; GFX10-NEXT:   [[FCMP:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[UITOFP]](s32), [[COPY]]
  ; GFX10-NEXT:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
  ; GFX10-NEXT:   [[ADD:%[0-9]+]]:_(s32) = G_ADD [[PHI1]], [[C3]]
  ; GFX10-NEXT:   [[INTRINSIC_CONVERGENT:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.if.break), [[FCMP]](s1), [[PHI]](s32)
  ; GFX10-NEXT:   SI_LOOP [[INTRINSIC_CONVERGENT]](s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX10-NEXT:   G_BR %bb.2
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT: bb.2:
  ; GFX10-NEXT:   [[PHI3:%[0-9]+]]:_(s1) = G_PHI [[XOR]](s1), %bb.1
  ; GFX10-NEXT:   [[PHI4:%[0-9]+]]:_(s32) = G_PHI [[INTRINSIC_CONVERGENT]](s32), %bb.1
  ; GFX10-NEXT:   G_INTRINSIC_CONVERGENT_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[PHI4]](s32)
  ; GFX10-NEXT:   [[C4:%[0-9]+]]:_(s32) = G_FCONSTANT float 0.000000e+00
  ; GFX10-NEXT:   [[C5:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
  ; GFX10-NEXT:   [[SELECT:%[0-9]+]]:_(s32) = G_SELECT [[PHI3]](s1), [[C5]], [[C4]]
  ; GFX10-NEXT:   G_STORE [[SELECT]](s32), [[MV]](p0) :: (store (s32))
  ; GFX10-NEXT:   SI_RETURN
  bb.0:
    successors: %bb.1(0x80000000)
    liveins: $vgpr0, $vgpr1, $vgpr2

    %0:_(s32) = COPY $vgpr0
    %1:_(s32) = COPY $vgpr1
    %2:_(s32) = COPY $vgpr2
    %3:_(p0) = G_MERGE_VALUES %1(s32), %2(s32)
    %4:_(s1) = G_CONSTANT i1 true
    %5:_(s32) = G_CONSTANT i32 0

  bb.1:
    successors: %bb.2(0x04000000), %bb.1(0x7c000000)

    %6:_(s32) = G_PHI %7(s32), %bb.1, %5(s32), %bb.0
    %8:_(s32) = G_PHI %5(s32), %bb.0, %9(s32), %bb.1
    %10:_(s1) = G_PHI %4(s1), %bb.0, %11(s1), %bb.1
    %12:_(s1) = G_CONSTANT i1 true
    %11:_(s1) = G_XOR %10, %12
    %13:_(s32) = G_UITOFP %8(s32)
    %14:_(s1) = G_FCMP floatpred(ogt), %13(s32), %0
    %15:_(s32) = G_CONSTANT i32 1
    %9:_(s32) = G_ADD %8, %15
    %7:sreg_32_xm0_xexec(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.if.break), %14(s1), %6(s32)
    SI_LOOP %7(s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
    G_BR %bb.2

  bb.2:
    %16:_(s1) = G_PHI %11(s1), %bb.1
    %17:_(s32) = G_PHI %7(s32), %bb.1
    G_INTRINSIC_CONVERGENT_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %17(s32)
    %18:_(s32) = G_FCONSTANT float 0.000000e+00
    %19:_(s32) = G_FCONSTANT float 1.000000e+00
    %20:_(s32) = G_SELECT %16(s1), %19, %18
    G_STORE %20(s32), %3(p0) :: (store (s32))
    SI_RETURN
...

---
name: loop_with_1break
legalized: true
tracksRegLiveness: true
body: |
  ; GFX10-LABEL: name: loop_with_1break
  ; GFX10: bb.0:
  ; GFX10-NEXT:   successors: %bb.1(0x80000000)
  ; GFX10-NEXT:   liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
  ; GFX10-NEXT:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY]](s32), [[COPY1]](s32)
  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr0
  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr1
  ; GFX10-NEXT:   [[MV1:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY3]](s32), [[COPY4]](s32)
  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr2
  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr3
  ; GFX10-NEXT:   [[MV2:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY5]](s32), [[COPY6]](s32)
  ; GFX10-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT: bb.1:
  ; GFX10-NEXT:   successors: %bb.3(0x50000000), %bb.5(0x30000000)
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT:   [[PHI:%[0-9]+]]:_(s32) = G_PHI %13(s32), %bb.5, [[C]](s32), %bb.0
  ; GFX10-NEXT:   [[PHI1:%[0-9]+]]:_(s32) = G_PHI [[C]](s32), %bb.0, %15(s32), %bb.5
  ; GFX10-NEXT:   [[C1:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
  ; GFX10-NEXT:   [[SEXT:%[0-9]+]]:_(s64) = G_SEXT [[PHI1]](s32)
  ; GFX10-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
  ; GFX10-NEXT:   [[SHL:%[0-9]+]]:_(s64) = G_SHL [[SEXT]], [[C2]](s32)
  ; GFX10-NEXT:   [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[MV1]], [[SHL]](s64)
  ; GFX10-NEXT:   [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32), addrspace 1)
  ; GFX10-NEXT:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
  ; GFX10-NEXT:   [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[LOAD]](s32), [[C3]]
  ; GFX10-NEXT:   G_BRCOND [[ICMP]](s1), %bb.3
  ; GFX10-NEXT:   G_BR %bb.5
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT: bb.2:
  ; GFX10-NEXT:   successors: %bb.4(0x80000000)
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT:   [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
  ; GFX10-NEXT:   G_STORE [[C4]](s32), [[MV2]](p1) :: (store (s32), addrspace 1)
  ; GFX10-NEXT:   G_BR %bb.4
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT: bb.3:
  ; GFX10-NEXT:   successors: %bb.5(0x80000000)
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT:   [[C5:%[0-9]+]]:_(s1) = G_CONSTANT i1 false
  ; GFX10-NEXT:   [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
  ; GFX10-NEXT:   [[SHL1:%[0-9]+]]:_(s64) = G_SHL [[SEXT]], [[C6]](s32)
  ; GFX10-NEXT:   [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[MV]], [[SHL1]](s64)
  ; GFX10-NEXT:   [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s32), addrspace 1)
  ; GFX10-NEXT:   [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
  ; GFX10-NEXT:   [[ADD:%[0-9]+]]:_(s32) = G_ADD [[LOAD1]], [[C7]]
  ; GFX10-NEXT:   G_STORE [[ADD]](s32), [[PTR_ADD1]](p1) :: (store (s32), addrspace 1)
  ; GFX10-NEXT:   [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[PHI1]], [[C7]]
  ; GFX10-NEXT:   [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[PHI1]](s32), [[COPY2]]
  ; GFX10-NEXT:   G_BR %bb.5
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT: bb.4:
  ; GFX10-NEXT:   G_INTRINSIC_CONVERGENT_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %34(s32)
  ; GFX10-NEXT:   S_ENDPGM 0
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT: bb.5:
  ; GFX10-NEXT:   successors: %bb.6(0x04000000), %bb.1(0x7c000000)
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT:   [[PHI2:%[0-9]+]]:_(s32) = G_PHI [[ADD1]](s32), %bb.3, [[DEF]](s32), %bb.1
  ; GFX10-NEXT:   [[PHI3:%[0-9]+]]:_(s1) = G_PHI [[C5]](s1), %bb.3, [[C1]](s1), %bb.1
  ; GFX10-NEXT:   [[PHI4:%[0-9]+]]:_(s1) = G_PHI [[ICMP1]](s1), %bb.3, [[C1]](s1), %bb.1
  ; GFX10-NEXT:   [[INTRINSIC_CONVERGENT:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.if.break), [[PHI4]](s1), [[PHI]](s32)
  ; GFX10-NEXT:   SI_LOOP [[INTRINSIC_CONVERGENT]](s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX10-NEXT:   G_BR %bb.6
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT: bb.6:
  ; GFX10-NEXT:   successors: %bb.2(0x40000000), %bb.4(0x40000000)
  ; GFX10-NEXT: {{  $}}
  ; GFX10-NEXT:   [[PHI5:%[0-9]+]]:sreg_32_xm0_xexec(s1) = G_PHI [[PHI3]](s1), %bb.5
  ; GFX10-NEXT:   [[PHI6:%[0-9]+]]:_(s32) = G_PHI [[INTRINSIC_CONVERGENT]](s32), %bb.5
  ; GFX10-NEXT:   G_INTRINSIC_CONVERGENT_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[PHI6]](s32)
  ; GFX10-NEXT:   [[SI_IF:%[0-9]+]]:sreg_32_xm0_xexec(s32) = SI_IF [[PHI5]](s1), %bb.4, implicit-def $exec, implicit-def $scc, implicit $exec
  ; GFX10-NEXT:   G_BR %bb.2
  bb.0:
    successors: %bb.1(0x80000000)
    liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2

    %0:_(s32) = COPY $vgpr0
    %1:_(s32) = COPY $vgpr1
    %2:_(p1) = G_MERGE_VALUES %0(s32), %1(s32)
    %3:_(s32) = COPY $vgpr2
    %4:_(s32) = COPY $sgpr0
    %5:_(s32) = COPY $sgpr1
    %6:_(p1) = G_MERGE_VALUES %4(s32), %5(s32)
    %7:_(s32) = COPY $sgpr2
    %8:_(s32) = COPY $sgpr3
    %9:_(p1) = G_MERGE_VALUES %7(s32), %8(s32)
    %10:_(s32) = G_CONSTANT i32 0
    %11:_(s32) = G_IMPLICIT_DEF

  bb.1:
    successors: %bb.3(0x50000000), %bb.5(0x30000000)

    %12:_(s32) = G_PHI %13(s32), %bb.5, %10(s32), %bb.0
    %14:_(s32) = G_PHI %10(s32), %bb.0, %15(s32), %bb.5
    %16:_(s1) = G_CONSTANT i1 true
    %17:_(s64) = G_SEXT %14(s32)
    %18:_(s32) = G_CONSTANT i32 2
    %19:_(s64) = G_SHL %17, %18(s32)
    %20:_(p1) = G_PTR_ADD %6, %19(s64)
    %21:_(s32) = G_LOAD %20(p1) :: (load (s32), addrspace 1)
    %22:_(s32) = G_CONSTANT i32 0
    %23:_(s1) = G_ICMP intpred(ne), %21(s32), %22
    G_BRCOND %23(s1), %bb.3
    G_BR %bb.5

  bb.2:
    successors: %bb.4(0x80000000)

    %24:_(s32) = G_CONSTANT i32 10
    G_STORE %24(s32), %9(p1) :: (store (s32), addrspace 1)
    G_BR %bb.4

  bb.3:
    successors: %bb.5(0x80000000)

    %25:_(s1) = G_CONSTANT i1 false
    %26:_(s32) = G_CONSTANT i32 2
    %27:_(s64) = G_SHL %17, %26(s32)
    %28:_(p1) = G_PTR_ADD %2, %27(s64)
    %29:_(s32) = G_LOAD %28(p1) :: (load (s32), addrspace 1)
    %30:_(s32) = G_CONSTANT i32 1
    %31:_(s32) = G_ADD %29, %30
    G_STORE %31(s32), %28(p1) :: (store (s32), addrspace 1)
    %32:_(s32) = G_ADD %14, %30
    %33:_(s1) = G_ICMP intpred(ult), %14(s32), %3
    G_BR %bb.5

  bb.4:
    G_INTRINSIC_CONVERGENT_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %34(s32)
    S_ENDPGM 0

  bb.5:
    successors: %bb.6(0x04000000), %bb.1(0x7c000000)

    %15:_(s32) = G_PHI %32(s32), %bb.3, %11(s32), %bb.1
    %35:_(s1) = G_PHI %25(s1), %bb.3, %16(s1), %bb.1
    %36:_(s1) = G_PHI %33(s1), %bb.3, %16(s1), %bb.1
    %13:sreg_32_xm0_xexec(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.if.break), %36(s1), %12(s32)
    SI_LOOP %13(s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
    G_BR %bb.6

  bb.6:
    successors: %bb.2(0x40000000), %bb.4(0x40000000)

    %37:sreg_32_xm0_xexec(s1) = G_PHI %35(s1), %bb.5
    %38:_(s32) = G_PHI %13(s32), %bb.5
    G_INTRINSIC_CONVERGENT_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %38(s32)
    %34:sreg_32_xm0_xexec(s32) = SI_IF %37(s1), %bb.4, implicit-def $exec, implicit-def $scc, implicit $exec
    G_BR %bb.2
...