; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: opt -mtriple=amdgcn-- -amdgpu-codegenprepare -S < %s | FileCheck -check-prefix=OPT %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck %s --check-prefixes=ASM,DAGISEL-ASM
; RUN: llc -mtriple=amdgcn-amd-amdhsa -global-isel -mcpu=gfx900 < %s | FileCheck %s --check-prefixes=ASM,GISEL-ASM

; Tests that we can avoid nullptr checks for addrspacecasts from/to priv/local.
;
; Whenever a testcase is successful, we should see the addrspacecast replaced with the intrinsic
; and the resulting code should have no select/cndmask null check for the pointer.

define void @local_to_flat_nonnull_arg(ptr addrspace(3) nonnull %ptr) {
; OPT-LABEL: define void @local_to_flat_nonnull_arg(
; OPT-SAME: ptr addrspace(3) nonnull [[PTR:%.*]]) {
; OPT-NEXT:    [[TMP1:%.*]] = call ptr @llvm.amdgcn.addrspacecast.nonnull.p0.p3(ptr addrspace(3) [[PTR]])
; OPT-NEXT:    store volatile i32 7, ptr [[TMP1]], align 4
; OPT-NEXT:    ret void
;
; ASM-LABEL: local_to_flat_nonnull_arg:
; ASM:       ; %bb.0:
; ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; ASM-NEXT:    s_mov_b64 s[4:5], src_shared_base
; ASM-NEXT:    v_mov_b32_e32 v1, s5
; ASM-NEXT:    v_mov_b32_e32 v2, 7
; ASM-NEXT:    flat_store_dword v[0:1], v2
; ASM-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ASM-NEXT:    s_setpc_b64 s[30:31]
  %x = addrspacecast ptr addrspace(3) %ptr to ptr
  store volatile i32 7, ptr %x
  ret void
}

define void @private_to_flat_nonnull_arg(ptr addrspace(5) nonnull %ptr) {
; OPT-LABEL: define void @private_to_flat_nonnull_arg(
; OPT-SAME: ptr addrspace(5) nonnull [[PTR:%.*]]) {
; OPT-NEXT:    [[TMP1:%.*]] = call ptr @llvm.amdgcn.addrspacecast.nonnull.p0.p5(ptr addrspace(5) [[PTR]])
; OPT-NEXT:    store volatile i32 7, ptr [[TMP1]], align 4
; OPT-NEXT:    ret void
;
; ASM-LABEL: private_to_flat_nonnull_arg:
; ASM:       ; %bb.0:
; ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; ASM-NEXT:    s_mov_b64 s[4:5], src_private_base
; ASM-NEXT:    v_mov_b32_e32 v1, s5
; ASM-NEXT:    v_mov_b32_e32 v2, 7
; ASM-NEXT:    flat_store_dword v[0:1], v2
; ASM-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ASM-NEXT:    s_setpc_b64 s[30:31]
  %x = addrspacecast ptr addrspace(5) %ptr to ptr
  store volatile i32 7, ptr %x
  ret void
}

define void @flat_to_local_nonnull_arg(ptr nonnull %ptr) {
; OPT-LABEL: define void @flat_to_local_nonnull_arg(
; OPT-SAME: ptr nonnull [[PTR:%.*]]) {
; OPT-NEXT:    [[TMP1:%.*]] = call ptr addrspace(3) @llvm.amdgcn.addrspacecast.nonnull.p3.p0(ptr [[PTR]])
; OPT-NEXT:    store volatile i32 7, ptr addrspace(3) [[TMP1]], align 4
; OPT-NEXT:    ret void
;
; ASM-LABEL: flat_to_local_nonnull_arg:
; ASM:       ; %bb.0:
; ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; ASM-NEXT:    v_mov_b32_e32 v1, 7
; ASM-NEXT:    ds_write_b32 v0, v1
; ASM-NEXT:    s_waitcnt lgkmcnt(0)
; ASM-NEXT:    s_setpc_b64 s[30:31]
  %x = addrspacecast ptr %ptr to ptr addrspace(3)
  store volatile i32 7, ptr addrspace(3) %x
  ret void
}

define void @flat_to_private_nonnull_arg(ptr nonnull %ptr) {
; OPT-LABEL: define void @flat_to_private_nonnull_arg(
; OPT-SAME: ptr nonnull [[PTR:%.*]]) {
; OPT-NEXT:    [[TMP1:%.*]] = call ptr addrspace(5) @llvm.amdgcn.addrspacecast.nonnull.p5.p0(ptr [[PTR]])
; OPT-NEXT:    store volatile i32 7, ptr addrspace(5) [[TMP1]], align 4
; OPT-NEXT:    ret void
;
; ASM-LABEL: flat_to_private_nonnull_arg:
; ASM:       ; %bb.0:
; ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; ASM-NEXT:    v_mov_b32_e32 v1, 7
; ASM-NEXT:    buffer_store_dword v1, v0, s[0:3], 0 offen
; ASM-NEXT:    s_waitcnt vmcnt(0)
; ASM-NEXT:    s_setpc_b64 s[30:31]
  %x = addrspacecast ptr %ptr to ptr addrspace(5)
  store volatile i32 7, ptr addrspace(5) %x
  ret void
}

define void @private_alloca_to_flat(ptr %ptr) {
; OPT-LABEL: define void @private_alloca_to_flat(
; OPT-SAME: ptr [[PTR:%.*]]) {
; OPT-NEXT:    [[ALLOCA:%.*]] = alloca i8, align 1, addrspace(5)
; OPT-NEXT:    [[TMP1:%.*]] = call ptr @llvm.amdgcn.addrspacecast.nonnull.p0.p5(ptr addrspace(5) [[ALLOCA]])
; OPT-NEXT:    store volatile i32 7, ptr [[TMP1]], align 4
; OPT-NEXT:    ret void
;
; DAGISEL-ASM-LABEL: private_alloca_to_flat:
; DAGISEL-ASM:       ; %bb.0:
; DAGISEL-ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; DAGISEL-ASM-NEXT:    s_mov_b64 s[4:5], src_private_base
; DAGISEL-ASM-NEXT:    v_lshrrev_b32_e64 v0, 6, s32
; DAGISEL-ASM-NEXT:    v_mov_b32_e32 v1, s5
; DAGISEL-ASM-NEXT:    v_mov_b32_e32 v2, 7
; DAGISEL-ASM-NEXT:    flat_store_dword v[0:1], v2
; DAGISEL-ASM-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; DAGISEL-ASM-NEXT:    s_setpc_b64 s[30:31]
;
; GISEL-ASM-LABEL: private_alloca_to_flat:
; GISEL-ASM:       ; %bb.0:
; GISEL-ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-ASM-NEXT:    s_lshr_b32 s4, s32, 6
; GISEL-ASM-NEXT:    s_mov_b64 s[6:7], src_private_base
; GISEL-ASM-NEXT:    s_mov_b32 s5, s7
; GISEL-ASM-NEXT:    v_mov_b32_e32 v0, s4
; GISEL-ASM-NEXT:    v_mov_b32_e32 v2, 7
; GISEL-ASM-NEXT:    v_mov_b32_e32 v1, s5
; GISEL-ASM-NEXT:    flat_store_dword v[0:1], v2
; GISEL-ASM-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; GISEL-ASM-NEXT:    s_setpc_b64 s[30:31]
  %alloca = alloca i8, addrspace(5)
  %x = addrspacecast ptr addrspace(5) %alloca to ptr
  store volatile i32 7, ptr %x
  ret void
}

@lds = internal unnamed_addr addrspace(3) global i8 poison, align 4

define void @knownbits_on_flat_to_priv(ptr %ptr) {
; OPT-LABEL: define void @knownbits_on_flat_to_priv(
; OPT-SAME: ptr [[PTR:%.*]]) {
; OPT-NEXT:    [[PTR_INT:%.*]] = ptrtoint ptr [[PTR]] to i64
; OPT-NEXT:    [[PTR_OR:%.*]] = or i64 [[PTR_INT]], 15
; OPT-NEXT:    [[KB_PTR:%.*]] = inttoptr i64 [[PTR_OR]] to ptr
; OPT-NEXT:    [[TMP1:%.*]] = call ptr addrspace(5) @llvm.amdgcn.addrspacecast.nonnull.p5.p0(ptr [[KB_PTR]])
; OPT-NEXT:    store volatile i32 7, ptr addrspace(5) [[TMP1]], align 4
; OPT-NEXT:    ret void
;
; ASM-LABEL: knownbits_on_flat_to_priv:
; ASM:       ; %bb.0:
; ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; ASM-NEXT:    v_or_b32_e32 v0, 15, v0
; ASM-NEXT:    v_mov_b32_e32 v1, 7
; ASM-NEXT:    buffer_store_dword v1, v0, s[0:3], 0 offen
; ASM-NEXT:    s_waitcnt vmcnt(0)
; ASM-NEXT:    s_setpc_b64 s[30:31]
  %ptr.int = ptrtoint ptr %ptr to i64
  %ptr.or = or i64 %ptr.int, 15 ; set some low bits
  %kb.ptr = inttoptr i64 %ptr.or to ptr
  %x = addrspacecast ptr %kb.ptr to ptr addrspace(5)
  store volatile i32 7, ptr addrspace(5) %x
  ret void
}

define void @knownbits_on_priv_to_flat(ptr addrspace(5) %ptr) {
; OPT-LABEL: define void @knownbits_on_priv_to_flat(
; OPT-SAME: ptr addrspace(5) [[PTR:%.*]]) {
; OPT-NEXT:    [[PTR_INT:%.*]] = ptrtoint ptr addrspace(5) [[PTR]] to i32
; OPT-NEXT:    [[PTR_OR:%.*]] = and i32 [[PTR_INT]], 65535
; OPT-NEXT:    [[KB_PTR:%.*]] = inttoptr i32 [[PTR_OR]] to ptr addrspace(5)
; OPT-NEXT:    [[TMP1:%.*]] = call ptr @llvm.amdgcn.addrspacecast.nonnull.p0.p5(ptr addrspace(5) [[KB_PTR]])
; OPT-NEXT:    store volatile i32 7, ptr [[TMP1]], align 4
; OPT-NEXT:    ret void
;
; ASM-LABEL: knownbits_on_priv_to_flat:
; ASM:       ; %bb.0:
; ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; ASM-NEXT:    s_mov_b64 s[4:5], src_private_base
; ASM-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; ASM-NEXT:    v_mov_b32_e32 v1, s5
; ASM-NEXT:    v_mov_b32_e32 v2, 7
; ASM-NEXT:    flat_store_dword v[0:1], v2
; ASM-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ASM-NEXT:    s_setpc_b64 s[30:31]
  %ptr.int = ptrtoint ptr addrspace(5) %ptr to i32
  %ptr.or = and i32 %ptr.int, 65535 ; ensure only lower 16 bits can be set.
  %kb.ptr = inttoptr i32 %ptr.or to ptr addrspace(5)
  %x = addrspacecast ptr addrspace(5) %kb.ptr to ptr
  store volatile i32 7, ptr %x
  ret void
}

define void @recursive_phis(i1 %cond, ptr addrspace(5) %ptr) {
; OPT-LABEL: define void @recursive_phis(
; OPT-SAME: i1 [[COND:%.*]], ptr addrspace(5) [[PTR:%.*]]) {
; OPT-NEXT:  [[ENTRY:.*]]:
; OPT-NEXT:    [[ALLOCA:%.*]] = alloca i8, align 1, addrspace(5)
; OPT-NEXT:    br i1 [[COND]], label %[[THEN:.*]], label %[[ELSE:.*]]
; OPT:       [[THEN]]:
; OPT-NEXT:    [[PTR_INT:%.*]] = ptrtoint ptr addrspace(5) [[PTR]] to i32
; OPT-NEXT:    [[PTR_OR:%.*]] = and i32 [[PTR_INT]], 65535
; OPT-NEXT:    [[KB_PTR:%.*]] = inttoptr i32 [[PTR_OR]] to ptr addrspace(5)
; OPT-NEXT:    br label %[[FINALLY:.*]]
; OPT:       [[ELSE]]:
; OPT-NEXT:    [[OTHER_PHI:%.*]] = phi ptr addrspace(5) [ [[ALLOCA]], %[[ENTRY]] ], [ [[PHI_PTR:%.*]], %[[FINALLY]] ]
; OPT-NEXT:    br label %[[FINALLY]]
; OPT:       [[FINALLY]]:
; OPT-NEXT:    [[PHI_PTR]] = phi ptr addrspace(5) [ [[KB_PTR]], %[[THEN]] ], [ [[OTHER_PHI]], %[[ELSE]] ]
; OPT-NEXT:    [[TMP0:%.*]] = call ptr @llvm.amdgcn.addrspacecast.nonnull.p0.p5(ptr addrspace(5) [[PHI_PTR]])
; OPT-NEXT:    store volatile i32 7, ptr [[TMP0]], align 4
; OPT-NEXT:    br i1 [[COND]], label %[[ELSE]], label %[[END:.*]]
; OPT:       [[END]]:
; OPT-NEXT:    ret void
;
; DAGISEL-ASM-LABEL: recursive_phis:
; DAGISEL-ASM:       ; %bb.0: ; %entry
; DAGISEL-ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; DAGISEL-ASM-NEXT:    v_and_b32_e32 v0, 1, v0
; DAGISEL-ASM-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v0
; DAGISEL-ASM-NEXT:    v_lshrrev_b32_e64 v0, 6, s32
; DAGISEL-ASM-NEXT:    s_and_saveexec_b64 s[4:5], vcc
; DAGISEL-ASM-NEXT:  ; %bb.1: ; %then
; DAGISEL-ASM-NEXT:    v_and_b32_e32 v0, 0xffff, v1
; DAGISEL-ASM-NEXT:  ; %bb.2: ; %finallyendcf.split
; DAGISEL-ASM-NEXT:    s_or_b64 exec, exec, s[4:5]
; DAGISEL-ASM-NEXT:    s_xor_b64 s[6:7], vcc, -1
; DAGISEL-ASM-NEXT:    s_mov_b64 s[4:5], 0
; DAGISEL-ASM-NEXT:    s_mov_b64 s[8:9], src_private_base
; DAGISEL-ASM-NEXT:    v_mov_b32_e32 v2, 7
; DAGISEL-ASM-NEXT:  .LBB7_3: ; %finally
; DAGISEL-ASM-NEXT:    ; =>This Inner Loop Header: Depth=1
; DAGISEL-ASM-NEXT:    s_and_b64 s[10:11], exec, s[6:7]
; DAGISEL-ASM-NEXT:    s_or_b64 s[4:5], s[10:11], s[4:5]
; DAGISEL-ASM-NEXT:    v_mov_b32_e32 v1, s9
; DAGISEL-ASM-NEXT:    flat_store_dword v[0:1], v2
; DAGISEL-ASM-NEXT:    s_waitcnt vmcnt(0)
; DAGISEL-ASM-NEXT:    s_andn2_b64 exec, exec, s[4:5]
; DAGISEL-ASM-NEXT:    s_cbranch_execnz .LBB7_3
; DAGISEL-ASM-NEXT:  ; %bb.4: ; %end
; DAGISEL-ASM-NEXT:    s_or_b64 exec, exec, s[4:5]
; DAGISEL-ASM-NEXT:    s_waitcnt lgkmcnt(0)
; DAGISEL-ASM-NEXT:    s_setpc_b64 s[30:31]
;
; GISEL-ASM-LABEL: recursive_phis:
; GISEL-ASM:       ; %bb.0: ; %entry
; GISEL-ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-ASM-NEXT:    v_and_b32_e32 v0, 1, v0
; GISEL-ASM-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
; GISEL-ASM-NEXT:    s_lshr_b32 s6, s32, 6
; GISEL-ASM-NEXT:    s_xor_b64 s[4:5], vcc, -1
; GISEL-ASM-NEXT:    v_mov_b32_e32 v0, s6
; GISEL-ASM-NEXT:    s_and_saveexec_b64 s[6:7], vcc
; GISEL-ASM-NEXT:  ; %bb.1: ; %then
; GISEL-ASM-NEXT:    v_and_b32_e32 v0, 0xffff, v1
; GISEL-ASM-NEXT:  ; %bb.2: ; %finallyendcf.split
; GISEL-ASM-NEXT:    s_or_b64 exec, exec, s[6:7]
; GISEL-ASM-NEXT:    s_mov_b64 s[8:9], src_private_base
; GISEL-ASM-NEXT:    s_mov_b64 s[6:7], 0
; GISEL-ASM-NEXT:    v_mov_b32_e32 v1, s9
; GISEL-ASM-NEXT:    v_mov_b32_e32 v2, 7
; GISEL-ASM-NEXT:  .LBB7_3: ; %finally
; GISEL-ASM-NEXT:    ; =>This Inner Loop Header: Depth=1
; GISEL-ASM-NEXT:    s_and_b64 s[8:9], exec, s[4:5]
; GISEL-ASM-NEXT:    s_or_b64 s[6:7], s[8:9], s[6:7]
; GISEL-ASM-NEXT:    flat_store_dword v[0:1], v2
; GISEL-ASM-NEXT:    s_waitcnt vmcnt(0)
; GISEL-ASM-NEXT:    s_andn2_b64 exec, exec, s[6:7]
; GISEL-ASM-NEXT:    s_cbranch_execnz .LBB7_3
; GISEL-ASM-NEXT:  ; %bb.4: ; %end
; GISEL-ASM-NEXT:    s_or_b64 exec, exec, s[6:7]
; GISEL-ASM-NEXT:    s_waitcnt lgkmcnt(0)
; GISEL-ASM-NEXT:    s_setpc_b64 s[30:31]
entry:
  %alloca = alloca i8, addrspace(5)
  br i1 %cond, label %then, label %else

then:
  %ptr.int = ptrtoint ptr addrspace(5) %ptr to i32
  %ptr.or = and i32 %ptr.int, 65535 ; ensure low bits are zeroes
  %kb.ptr = inttoptr i32 %ptr.or to ptr addrspace(5)
  br label %finally

else:
  %other.phi = phi ptr addrspace(5) [%alloca, %entry], [%phi.ptr, %finally]
  br label %finally

finally:
  %phi.ptr = phi ptr addrspace(5) [%kb.ptr, %then], [%other.phi, %else]
  %x = addrspacecast ptr addrspace(5) %phi.ptr to ptr
  store volatile i32 7, ptr %x
  br i1 %cond, label %else, label %end

end:
  ret void
}

; This used to assert due to assuming the size of the source address
; space was larger than the destination.

define i32 @cast_private_to_flat_to_private(ptr addrspace(5) %private.ptr) {
; OPT-LABEL: define i32 @cast_private_to_flat_to_private(
; OPT-SAME: ptr addrspace(5) [[PRIVATE_PTR:%.*]]) {
; OPT-NEXT:    [[FLAT_PTR:%.*]] = addrspacecast ptr addrspace(5) [[PRIVATE_PTR]] to ptr
; OPT-NEXT:    [[CAST_BACK:%.*]] = addrspacecast ptr [[FLAT_PTR]] to ptr addrspace(5)
; OPT-NEXT:    [[LOAD:%.*]] = load volatile i32, ptr addrspace(5) [[CAST_BACK]], align 4
; OPT-NEXT:    ret i32 [[LOAD]]
;
; ASM-LABEL: cast_private_to_flat_to_private:
; ASM:       ; %bb.0:
; ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; ASM-NEXT:    buffer_load_dword v0, v0, s[0:3], 0 offen glc
; ASM-NEXT:    s_waitcnt vmcnt(0)
; ASM-NEXT:    s_setpc_b64 s[30:31]
  %flat.ptr = addrspacecast ptr addrspace(5) %private.ptr to ptr
  %cast.back = addrspacecast ptr %flat.ptr to ptr addrspace(5)
  %load = load volatile i32, ptr addrspace(5) %cast.back
  ret i32 %load
}

; This is UB but shouldn't assert.
define i32 @cast_private_to_flat_to_local(ptr addrspace(5) %private.ptr) {
; OPT-LABEL: define i32 @cast_private_to_flat_to_local(
; OPT-SAME: ptr addrspace(5) [[PRIVATE_PTR:%.*]]) {
; OPT-NEXT:    [[FLAT_PTR:%.*]] = addrspacecast ptr addrspace(5) [[PRIVATE_PTR]] to ptr
; OPT-NEXT:    [[CAST_BACK:%.*]] = addrspacecast ptr [[FLAT_PTR]] to ptr addrspace(3)
; OPT-NEXT:    [[LOAD:%.*]] = load volatile i32, ptr addrspace(3) [[CAST_BACK]], align 4
; OPT-NEXT:    ret i32 [[LOAD]]
;
; DAGISEL-ASM-LABEL: cast_private_to_flat_to_local:
; DAGISEL-ASM:       ; %bb.0:
; DAGISEL-ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; DAGISEL-ASM-NEXT:    s_mov_b64 s[4:5], src_private_base
; DAGISEL-ASM-NEXT:    v_mov_b32_e32 v1, s5
; DAGISEL-ASM-NEXT:    v_cmp_ne_u32_e32 vcc, -1, v0
; DAGISEL-ASM-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
; DAGISEL-ASM-NEXT:    v_cndmask_b32_e32 v0, 0, v0, vcc
; DAGISEL-ASM-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[0:1]
; DAGISEL-ASM-NEXT:    v_cndmask_b32_e32 v0, -1, v0, vcc
; DAGISEL-ASM-NEXT:    ds_read_b32 v0, v0
; DAGISEL-ASM-NEXT:    s_waitcnt lgkmcnt(0)
; DAGISEL-ASM-NEXT:    s_setpc_b64 s[30:31]
;
; GISEL-ASM-LABEL: cast_private_to_flat_to_local:
; GISEL-ASM:       ; %bb.0:
; GISEL-ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-ASM-NEXT:    s_mov_b64 s[4:5], src_private_base
; GISEL-ASM-NEXT:    v_mov_b32_e32 v1, s5
; GISEL-ASM-NEXT:    v_cmp_ne_u32_e32 vcc, -1, v0
; GISEL-ASM-NEXT:    v_cndmask_b32_e32 v0, 0, v0, vcc
; GISEL-ASM-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
; GISEL-ASM-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[0:1]
; GISEL-ASM-NEXT:    v_cndmask_b32_e32 v0, -1, v0, vcc
; GISEL-ASM-NEXT:    ds_read_b32 v0, v0
; GISEL-ASM-NEXT:    s_waitcnt lgkmcnt(0)
; GISEL-ASM-NEXT:    s_setpc_b64 s[30:31]
  %flat.ptr = addrspacecast ptr addrspace(5) %private.ptr to ptr
  %cast.back = addrspacecast ptr %flat.ptr to ptr addrspace(3)
  %load = load volatile i32, ptr addrspace(3) %cast.back
  ret i32 %load
}

; This is UB but shouldn't assert.
define i32 @cast_private_to_flat_to_global(ptr addrspace(6) %const32.ptr) {
; OPT-LABEL: define i32 @cast_private_to_flat_to_global(
; OPT-SAME: ptr addrspace(6) [[CONST32_PTR:%.*]]) {
; OPT-NEXT:    [[FLAT_PTR:%.*]] = addrspacecast ptr addrspace(6) [[CONST32_PTR]] to ptr
; OPT-NEXT:    [[LOCAL_PTR:%.*]] = addrspacecast ptr [[FLAT_PTR]] to ptr addrspace(3)
; OPT-NEXT:    [[LOAD:%.*]] = load volatile i32, ptr addrspace(3) [[LOCAL_PTR]], align 4
; OPT-NEXT:    ret i32 [[LOAD]]
;
; ASM-LABEL: cast_private_to_flat_to_global:
; ASM:       ; %bb.0:
; ASM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; ASM-NEXT:    v_mov_b32_e32 v1, 0
; ASM-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[0:1]
; ASM-NEXT:    v_cndmask_b32_e32 v0, -1, v0, vcc
; ASM-NEXT:    ds_read_b32 v0, v0
; ASM-NEXT:    s_waitcnt lgkmcnt(0)
; ASM-NEXT:    s_setpc_b64 s[30:31]
  %flat.ptr = addrspacecast ptr addrspace(6) %const32.ptr to ptr
  %local.ptr = addrspacecast ptr %flat.ptr to ptr addrspace(3)
  %load = load volatile i32, ptr addrspace(3) %local.ptr
  ret i32 %load
}