llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-enable-uniform-intrinsic-combine=0 -O3 -S < %s | FileCheck %s -check-prefix=CURRENT-CHECK
; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK
; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK
; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O3 -S < %s | FileCheck %s -check-prefix=O3-CHECK

define protected amdgpu_kernel void @trivial_waterfall_eq_zero(ptr addrspace(1) %out) {
; CURRENT-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_eq_zero(
; CURRENT-CHECK-SAME: ptr addrspace(1) writeonly captures(none) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
; CURRENT-CHECK-NEXT:  [[ENTRY:.*:]]
; CURRENT-CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.amdgcn.ballot.i32(i1 true)
; CURRENT-CHECK-NEXT:    [[IS_DONE_PEEL:%.*]] = icmp eq i32 [[TMP0]], 0
; CURRENT-CHECK-NEXT:    br i1 [[IS_DONE_PEEL]], label %[[EXIT:.*]], label %[[IF_PEEL:.*]]
; CURRENT-CHECK:       [[IF_PEEL]]:
; CURRENT-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; CURRENT-CHECK-NEXT:    br label %[[EXIT]]
; CURRENT-CHECK:       [[EXIT]]:
; CURRENT-CHECK-NEXT:    ret void
;
; PASS-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_eq_zero(
; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0:[0-9]+]] {
; PASS-CHECK-NEXT:  [[ENTRY:.*]]:
; PASS-CHECK-NEXT:    br label %[[WHILE:.*]]
; PASS-CHECK:       [[WHILE]]:
; PASS-CHECK-NEXT:    [[DONE:%.*]] = phi i1 [ false, %[[ENTRY]] ], [ true, %[[IF:.*]] ]
; PASS-CHECK-NEXT:    [[NOT_DONE:%.*]] = xor i1 [[DONE]], true
; PASS-CHECK-NEXT:    [[BALLOT:%.*]] = tail call i64 @llvm.amdgcn.ballot.i64(i1 [[NOT_DONE]])
; PASS-CHECK-NEXT:    [[TMP0:%.*]] = xor i1 [[NOT_DONE]], true
; PASS-CHECK-NEXT:    [[IS_DONE:%.*]] = icmp eq i64 [[BALLOT]], 0
; PASS-CHECK-NEXT:    br i1 [[TMP0]], label %[[EXIT:.*]], label %[[IF]]
; PASS-CHECK:       [[IF]]:
; PASS-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; PASS-CHECK-NEXT:    br label %[[WHILE]]
; PASS-CHECK:       [[EXIT]]:
; PASS-CHECK-NEXT:    ret void
;
; O3-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_eq_zero(
; O3-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
; O3-CHECK-NEXT:  [[ENTRY:.*:]]
; O3-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; O3-CHECK-NEXT:    ret void
;
entry:
  br label %while

while:
  %done = phi i1 [ 0, %entry ], [ 1, %if ]
  %not_done = xor i1 %done, true
  %ballot = tail call i64 @llvm.amdgcn.ballot.i64(i1 %not_done)
  %is_done = icmp eq i64 %ballot, 0 ; in this case is_done = !not_done
  br i1 %is_done, label %exit, label %if

if:
  store i32 5, ptr addrspace(1) %out
  br label %while

exit:
  ret void
}

define protected amdgpu_kernel void @trivial_waterfall_eq_zero_swap_op(ptr addrspace(1) %out) {
; CURRENT-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_eq_zero_swap_op(
; CURRENT-CHECK-SAME: ptr addrspace(1) writeonly captures(none) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0]] {
; CURRENT-CHECK-NEXT:  [[ENTRY:.*:]]
; CURRENT-CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.amdgcn.ballot.i32(i1 true)
; CURRENT-CHECK-NEXT:    [[IS_DONE_PEEL:%.*]] = icmp eq i32 [[TMP0]], 0
; CURRENT-CHECK-NEXT:    br i1 [[IS_DONE_PEEL]], label %[[EXIT:.*]], label %[[IF_PEEL:.*]]
; CURRENT-CHECK:       [[IF_PEEL]]:
; CURRENT-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; CURRENT-CHECK-NEXT:    br label %[[EXIT]]
; CURRENT-CHECK:       [[EXIT]]:
; CURRENT-CHECK-NEXT:    ret void
;
; PASS-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_eq_zero_swap_op(
; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] {
; PASS-CHECK-NEXT:  [[ENTRY:.*]]:
; PASS-CHECK-NEXT:    br label %[[WHILE:.*]]
; PASS-CHECK:       [[WHILE]]:
; PASS-CHECK-NEXT:    [[DONE:%.*]] = phi i1 [ false, %[[ENTRY]] ], [ true, %[[IF:.*]] ]
; PASS-CHECK-NEXT:    [[NOT_DONE:%.*]] = xor i1 [[DONE]], true
; PASS-CHECK-NEXT:    [[BALLOT:%.*]] = tail call i64 @llvm.amdgcn.ballot.i64(i1 [[NOT_DONE]])
; PASS-CHECK-NEXT:    [[TMP0:%.*]] = xor i1 [[NOT_DONE]], true
; PASS-CHECK-NEXT:    [[IS_DONE:%.*]] = icmp eq i64 0, [[BALLOT]]
; PASS-CHECK-NEXT:    br i1 [[TMP0]], label %[[EXIT:.*]], label %[[IF]]
; PASS-CHECK:       [[IF]]:
; PASS-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; PASS-CHECK-NEXT:    br label %[[WHILE]]
; PASS-CHECK:       [[EXIT]]:
; PASS-CHECK-NEXT:    ret void
;
; O3-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_eq_zero_swap_op(
; O3-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0]] {
; O3-CHECK-NEXT:  [[ENTRY:.*:]]
; O3-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; O3-CHECK-NEXT:    ret void
;
entry:
  br label %while

while:
  %done = phi i1 [ 0, %entry ], [ 1, %if ]
  %not_done = xor i1 %done, true
  %ballot = tail call i64 @llvm.amdgcn.ballot.i64(i1 %not_done)
  %is_done = icmp eq i64 0, %ballot ; in this case is_done = !not_done
  br i1 %is_done, label %exit, label %if

if:
  store i32 5, ptr addrspace(1) %out
  br label %while

exit:
  ret void
}

define protected amdgpu_kernel void @trivial_waterfall_ne_zero(ptr addrspace(1) %out) {
; CURRENT-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_ne_zero(
; CURRENT-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]]) local_unnamed_addr #[[ATTR1:[0-9]+]] {
; CURRENT-CHECK-NEXT:  [[ENTRY:.*:]]
; CURRENT-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; CURRENT-CHECK-NEXT:    br label %[[WHILE:.*]]
; CURRENT-CHECK:       [[WHILE]]:
; CURRENT-CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.amdgcn.ballot.i32(i1 true)
; CURRENT-CHECK-NEXT:    [[IS_DONE_NOT:%.*]] = icmp eq i32 [[TMP0]], 0
; CURRENT-CHECK-NEXT:    br i1 [[IS_DONE_NOT]], label %[[WHILE]], label %[[EXIT:.*]], !llvm.loop [[LOOP0:![0-9]+]]
; CURRENT-CHECK:       [[EXIT]]:
; CURRENT-CHECK-NEXT:    ret void
;
; PASS-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_ne_zero(
; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] {
; PASS-CHECK-NEXT:  [[ENTRY:.*]]:
; PASS-CHECK-NEXT:    br label %[[WHILE:.*]]
; PASS-CHECK:       [[WHILE]]:
; PASS-CHECK-NEXT:    [[DONE:%.*]] = phi i1 [ false, %[[ENTRY]] ], [ true, %[[IF:.*]] ]
; PASS-CHECK-NEXT:    [[BALLOT:%.*]] = tail call i64 @llvm.amdgcn.ballot.i64(i1 [[DONE]])
; PASS-CHECK-NEXT:    [[IS_DONE:%.*]] = icmp ne i64 0, [[BALLOT]]
; PASS-CHECK-NEXT:    br i1 [[DONE]], label %[[EXIT:.*]], label %[[IF]]
; PASS-CHECK:       [[IF]]:
; PASS-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; PASS-CHECK-NEXT:    br label %[[WHILE]]
; PASS-CHECK:       [[EXIT]]:
; PASS-CHECK-NEXT:    ret void
;
; O3-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_ne_zero(
; O3-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0]] {
; O3-CHECK-NEXT:  [[ENTRY:.*:]]
; O3-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; O3-CHECK-NEXT:    ret void
;
entry:
  br label %while

while:
  %done = phi i1 [ 0, %entry ], [ 1, %if ]
  %ballot = tail call i64 @llvm.amdgcn.ballot.i64(i1 %done)
  %is_done = icmp ne i64 0, %ballot ; in this case is_done = done
  br i1 %is_done, label %exit, label %if

if:
  store i32 5, ptr addrspace(1) %out
  br label %while

exit:
  ret void
}

define protected amdgpu_kernel void @trivial_waterfall_ne_zero_swap(ptr addrspace(1) %out) {
; CURRENT-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_ne_zero_swap(
; CURRENT-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]]) local_unnamed_addr #[[ATTR1]] {
; CURRENT-CHECK-NEXT:  [[ENTRY:.*:]]
; CURRENT-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; CURRENT-CHECK-NEXT:    br label %[[WHILE:.*]]
; CURRENT-CHECK:       [[WHILE]]:
; CURRENT-CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.amdgcn.ballot.i32(i1 true)
; CURRENT-CHECK-NEXT:    [[IS_DONE_NOT:%.*]] = icmp eq i32 [[TMP0]], 0
; CURRENT-CHECK-NEXT:    br i1 [[IS_DONE_NOT]], label %[[WHILE]], label %[[EXIT:.*]], !llvm.loop [[LOOP2:![0-9]+]]
; CURRENT-CHECK:       [[EXIT]]:
; CURRENT-CHECK-NEXT:    ret void
;
; PASS-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_ne_zero_swap(
; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] {
; PASS-CHECK-NEXT:  [[ENTRY:.*]]:
; PASS-CHECK-NEXT:    br label %[[WHILE:.*]]
; PASS-CHECK:       [[WHILE]]:
; PASS-CHECK-NEXT:    [[DONE:%.*]] = phi i1 [ false, %[[ENTRY]] ], [ true, %[[IF:.*]] ]
; PASS-CHECK-NEXT:    [[BALLOT:%.*]] = tail call i64 @llvm.amdgcn.ballot.i64(i1 [[DONE]])
; PASS-CHECK-NEXT:    [[IS_DONE:%.*]] = icmp ne i64 [[BALLOT]], 0
; PASS-CHECK-NEXT:    br i1 [[DONE]], label %[[EXIT:.*]], label %[[IF]]
; PASS-CHECK:       [[IF]]:
; PASS-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; PASS-CHECK-NEXT:    br label %[[WHILE]]
; PASS-CHECK:       [[EXIT]]:
; PASS-CHECK-NEXT:    ret void
;
; O3-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_ne_zero_swap(
; O3-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0]] {
; O3-CHECK-NEXT:  [[ENTRY:.*:]]
; O3-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; O3-CHECK-NEXT:    ret void
;
entry:
  br label %while

while:
  %done = phi i1 [ 0, %entry ], [ 1, %if ]
  %ballot = tail call i64 @llvm.amdgcn.ballot.i64(i1 %done)
  %is_done = icmp ne i64 %ballot, 0 ; in this case is_done = done
  br i1 %is_done, label %exit, label %if

if:
  store i32 5, ptr addrspace(1) %out
  br label %while

exit:
  ret void
}

define protected amdgpu_kernel void @trivial_uniform_waterfall(ptr addrspace(1) %out) {
; CURRENT-CHECK-LABEL: define protected amdgpu_kernel void @trivial_uniform_waterfall(
; CURRENT-CHECK-SAME: ptr addrspace(1) writeonly captures(none) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0]] {
; CURRENT-CHECK-NEXT:  [[ENTRY:.*:]]
; CURRENT-CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.amdgcn.ballot.i32(i1 true)
; CURRENT-CHECK-NEXT:    [[IS_DONE_PEEL:%.*]] = icmp eq i32 [[TMP0]], 0
; CURRENT-CHECK-NEXT:    br i1 [[IS_DONE_PEEL]], label %[[EXIT:.*]], label %[[WORK_PEEL:.*]]
; CURRENT-CHECK:       [[WORK_PEEL]]:
; CURRENT-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; CURRENT-CHECK-NEXT:    br label %[[EXIT]]
; CURRENT-CHECK:       [[EXIT]]:
; CURRENT-CHECK-NEXT:    ret void
;
; PASS-CHECK-LABEL: define protected amdgpu_kernel void @trivial_uniform_waterfall(
; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] {
; PASS-CHECK-NEXT:  [[ENTRY:.*]]:
; PASS-CHECK-NEXT:    br label %[[WHILE:.*]]
; PASS-CHECK:       [[WHILE]]:
; PASS-CHECK-NEXT:    [[DONE:%.*]] = phi i1 [ false, %[[ENTRY]] ], [ [[NEW_DONE:%.*]], %[[TAIL:.*]] ]
; PASS-CHECK-NEXT:    [[NOT_DONE:%.*]] = xor i1 [[DONE]], true
; PASS-CHECK-NEXT:    [[BALLOT:%.*]] = tail call i64 @llvm.amdgcn.ballot.i64(i1 [[NOT_DONE]])
; PASS-CHECK-NEXT:    [[TMP0:%.*]] = xor i1 [[NOT_DONE]], true
; PASS-CHECK-NEXT:    [[IS_DONE:%.*]] = icmp eq i64 [[BALLOT]], 0
; PASS-CHECK-NEXT:    br i1 [[TMP0]], label %[[EXIT:.*]], label %[[IF:.*]]
; PASS-CHECK:       [[IF]]:
; PASS-CHECK-NEXT:    [[IS_FIRST_ACTIVE_ID:%.*]] = icmp eq i32 0, 0
; PASS-CHECK-NEXT:    br i1 [[IS_FIRST_ACTIVE_ID]], label %[[WORK:.*]], label %[[TAIL]]
; PASS-CHECK:       [[WORK]]:
; PASS-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; PASS-CHECK-NEXT:    br label %[[TAIL]]
; PASS-CHECK:       [[TAIL]]:
; PASS-CHECK-NEXT:    [[NEW_DONE]] = phi i1 [ true, %[[WORK]] ], [ false, %[[IF]] ]
; PASS-CHECK-NEXT:    br label %[[WHILE]]
; PASS-CHECK:       [[EXIT]]:
; PASS-CHECK-NEXT:    ret void
;
; O3-CHECK-LABEL: define protected amdgpu_kernel void @trivial_uniform_waterfall(
; O3-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0]] {
; O3-CHECK-NEXT:  [[ENTRY:.*:]]
; O3-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; O3-CHECK-NEXT:    ret void
;
entry:
  br label %while

while:
  %done = phi i1 [ false, %entry ], [ %new_done, %tail ]
  %not_done = xor i1 %done, true
  %ballot = tail call i64 @llvm.amdgcn.ballot.i64(i1 %not_done)
  %is_done = icmp eq i64 %ballot, 0
  br i1 %is_done, label %exit, label %if

if:
  %first_active_id = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 0)
  %is_first_active_id = icmp eq i32 0, %first_active_id
  br i1 %is_first_active_id, label %work, label %tail

work:
  store i32 5, ptr addrspace(1) %out
  br label %tail

tail:
  %new_done = phi i1 [ true, %work ], [ false, %if ]
  br label %while

exit:
  ret void
}

define protected amdgpu_kernel void @uniform_waterfall(ptr addrspace(1) %out, i32 %mymask) {
; CURRENT-CHECK-LABEL: define protected amdgpu_kernel void @uniform_waterfall(
; CURRENT-CHECK-SAME: ptr addrspace(1) writeonly captures(none) [[OUT:%.*]], i32 [[MYMASK:%.*]]) local_unnamed_addr #[[ATTR0]] {
; CURRENT-CHECK-NEXT:  [[ENTRY:.*:]]
; CURRENT-CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.amdgcn.ballot.i32(i1 true)
; CURRENT-CHECK-NEXT:    [[IS_DONE_PEEL:%.*]] = icmp eq i32 [[TMP0]], 0
; CURRENT-CHECK-NEXT:    br i1 [[IS_DONE_PEEL]], label %[[EXIT:.*]], label %[[WORK_PEEL:.*]]
; CURRENT-CHECK:       [[WORK_PEEL]]:
; CURRENT-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; CURRENT-CHECK-NEXT:    br label %[[EXIT]]
; CURRENT-CHECK:       [[EXIT]]:
; CURRENT-CHECK-NEXT:    ret void
;
; PASS-CHECK-LABEL: define protected amdgpu_kernel void @uniform_waterfall(
; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]], i32 [[MYMASK:%.*]]) #[[ATTR0]] {
; PASS-CHECK-NEXT:  [[ENTRY:.*]]:
; PASS-CHECK-NEXT:    br label %[[WHILE:.*]]
; PASS-CHECK:       [[WHILE]]:
; PASS-CHECK-NEXT:    [[DONE:%.*]] = phi i1 [ false, %[[ENTRY]] ], [ [[NEW_DONE:%.*]], %[[TAIL:.*]] ]
; PASS-CHECK-NEXT:    [[NOT_DONE:%.*]] = xor i1 [[DONE]], true
; PASS-CHECK-NEXT:    [[BALLOT:%.*]] = tail call i64 @llvm.amdgcn.ballot.i64(i1 [[NOT_DONE]])
; PASS-CHECK-NEXT:    [[TMP0:%.*]] = xor i1 [[NOT_DONE]], true
; PASS-CHECK-NEXT:    [[IS_DONE:%.*]] = icmp eq i64 [[BALLOT]], 0
; PASS-CHECK-NEXT:    br i1 [[TMP0]], label %[[EXIT:.*]], label %[[IF:.*]]
; PASS-CHECK:       [[IF]]:
; PASS-CHECK-NEXT:    [[IS_FIRST_ACTIVE_ID:%.*]] = icmp eq i32 [[MYMASK]], [[MYMASK]]
; PASS-CHECK-NEXT:    br i1 [[IS_FIRST_ACTIVE_ID]], label %[[WORK:.*]], label %[[TAIL]]
; PASS-CHECK:       [[WORK]]:
; PASS-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; PASS-CHECK-NEXT:    br label %[[TAIL]]
; PASS-CHECK:       [[TAIL]]:
; PASS-CHECK-NEXT:    [[NEW_DONE]] = phi i1 [ true, %[[WORK]] ], [ false, %[[IF]] ]
; PASS-CHECK-NEXT:    br label %[[WHILE]]
; PASS-CHECK:       [[EXIT]]:
; PASS-CHECK-NEXT:    ret void
;
; O3-CHECK-LABEL: define protected amdgpu_kernel void @uniform_waterfall(
; O3-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]], i32 [[MYMASK:%.*]]) local_unnamed_addr #[[ATTR0]] {
; O3-CHECK-NEXT:  [[ENTRY:.*:]]
; O3-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; O3-CHECK-NEXT:    ret void
;
entry:
  br label %while

while:
  %done = phi i1 [ false, %entry ], [ %new_done, %tail ]
  %not_done = xor i1 %done, true
  %ballot = tail call i64 @llvm.amdgcn.ballot.i64(i1 %not_done)
  %is_done = icmp eq i64 %ballot, 0
  br i1 %is_done, label %exit, label %if

if:
  %first_active_id = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %mymask)
  %is_first_active_id = icmp eq i32 %mymask, %first_active_id
  br i1 %is_first_active_id, label %work, label %tail

work:
  store i32 5, ptr addrspace(1) %out
  br label %tail

tail:
  %new_done = phi i1 [ true, %work ], [ false, %if ]
  br label %while

exit:
  ret void
}

define protected amdgpu_kernel void @trivial_waterfall_eq_zero_i32(ptr addrspace(1) %out) {
; CURRENT-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_eq_zero_i32(
; CURRENT-CHECK-SAME: ptr addrspace(1) writeonly captures(none) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0]] {
; CURRENT-CHECK-NEXT:  [[ENTRY:.*:]]
; CURRENT-CHECK-NEXT:    [[BALLOT_PEEL:%.*]] = tail call i32 @llvm.amdgcn.ballot.i32(i1 true)
; CURRENT-CHECK-NEXT:    [[IS_DONE_PEEL:%.*]] = icmp eq i32 [[BALLOT_PEEL]], 0
; CURRENT-CHECK-NEXT:    br i1 [[IS_DONE_PEEL]], label %[[EXIT:.*]], label %[[IF_PEEL:.*]]
; CURRENT-CHECK:       [[IF_PEEL]]:
; CURRENT-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; CURRENT-CHECK-NEXT:    br label %[[EXIT]]
; CURRENT-CHECK:       [[EXIT]]:
; CURRENT-CHECK-NEXT:    ret void
;
; PASS-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_eq_zero_i32(
; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] {
; PASS-CHECK-NEXT:  [[ENTRY:.*]]:
; PASS-CHECK-NEXT:    br label %[[WHILE:.*]]
; PASS-CHECK:       [[WHILE]]:
; PASS-CHECK-NEXT:    [[DONE:%.*]] = phi i1 [ false, %[[ENTRY]] ], [ true, %[[IF:.*]] ]
; PASS-CHECK-NEXT:    [[NOT_DONE:%.*]] = xor i1 [[DONE]], true
; PASS-CHECK-NEXT:    [[BALLOT:%.*]] = tail call i32 @llvm.amdgcn.ballot.i32(i1 [[NOT_DONE]])
; PASS-CHECK-NEXT:    [[TMP0:%.*]] = xor i1 [[NOT_DONE]], true
; PASS-CHECK-NEXT:    [[IS_DONE:%.*]] = icmp eq i32 [[BALLOT]], 0
; PASS-CHECK-NEXT:    br i1 [[TMP0]], label %[[EXIT:.*]], label %[[IF]]
; PASS-CHECK:       [[IF]]:
; PASS-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; PASS-CHECK-NEXT:    br label %[[WHILE]]
; PASS-CHECK:       [[EXIT]]:
; PASS-CHECK-NEXT:    ret void
;
; O3-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_eq_zero_i32(
; O3-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0]] {
; O3-CHECK-NEXT:  [[ENTRY:.*:]]
; O3-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; O3-CHECK-NEXT:    ret void
;
entry:
  br label %while

while:
  %done = phi i1 [ 0, %entry ], [ 1, %if ]
  %not_done = xor i1 %done, true
  %ballot = tail call i32 @llvm.amdgcn.ballot.i32(i1 %not_done)
  %is_done = icmp eq i32 %ballot, 0 ; in this case is_done = !not_done
  br i1 %is_done, label %exit, label %if

if:
  store i32 5, ptr addrspace(1) %out
  br label %while

exit:
  ret void
}

define protected amdgpu_kernel void @trivial_waterfall_ne_zero_i32(ptr addrspace(1) %out) {
; CURRENT-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_ne_zero_i32(
; CURRENT-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]]) local_unnamed_addr #[[ATTR1]] {
; CURRENT-CHECK-NEXT:  [[ENTRY:.*:]]
; CURRENT-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; CURRENT-CHECK-NEXT:    br label %[[WHILE:.*]]
; CURRENT-CHECK:       [[WHILE]]:
; CURRENT-CHECK-NEXT:    [[BALLOT:%.*]] = tail call i32 @llvm.amdgcn.ballot.i32(i1 true)
; CURRENT-CHECK-NEXT:    [[IS_DONE_NOT:%.*]] = icmp eq i32 [[BALLOT]], 0
; CURRENT-CHECK-NEXT:    br i1 [[IS_DONE_NOT]], label %[[WHILE]], label %[[EXIT:.*]], !llvm.loop [[LOOP3:![0-9]+]]
; CURRENT-CHECK:       [[EXIT]]:
; CURRENT-CHECK-NEXT:    ret void
;
; PASS-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_ne_zero_i32(
; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] {
; PASS-CHECK-NEXT:  [[ENTRY:.*]]:
; PASS-CHECK-NEXT:    br label %[[WHILE:.*]]
; PASS-CHECK:       [[WHILE]]:
; PASS-CHECK-NEXT:    [[DONE:%.*]] = phi i1 [ false, %[[ENTRY]] ], [ true, %[[IF:.*]] ]
; PASS-CHECK-NEXT:    [[BALLOT:%.*]] = tail call i32 @llvm.amdgcn.ballot.i32(i1 [[DONE]])
; PASS-CHECK-NEXT:    [[IS_DONE:%.*]] = icmp ne i32 0, [[BALLOT]]
; PASS-CHECK-NEXT:    br i1 [[DONE]], label %[[EXIT:.*]], label %[[IF]]
; PASS-CHECK:       [[IF]]:
; PASS-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; PASS-CHECK-NEXT:    br label %[[WHILE]]
; PASS-CHECK:       [[EXIT]]:
; PASS-CHECK-NEXT:    ret void
;
; O3-CHECK-LABEL: define protected amdgpu_kernel void @trivial_waterfall_ne_zero_i32(
; O3-CHECK-SAME: ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[OUT:%.*]]) local_unnamed_addr #[[ATTR0]] {
; O3-CHECK-NEXT:  [[ENTRY:.*:]]
; O3-CHECK-NEXT:    store i32 5, ptr addrspace(1) [[OUT]], align 4
; O3-CHECK-NEXT:    ret void
;
entry:
  br label %while

while:
  %done = phi i1 [ 0, %entry ], [ 1, %if ]
  %ballot = tail call i32 @llvm.amdgcn.ballot.i32(i1 %done)
  %is_done = icmp ne i32 0, %ballot ; in this case is_done = done
  br i1 %is_done, label %exit, label %if

if:
  store i32 5, ptr addrspace(1) %out
  br label %while

exit:
  ret void
}

declare i64 @llvm.amdgcn.ballot.i64(i1) #1
!6 = !{i64 690}
!7 = distinct !{!7, !8}
!8 = !{!"llvm.loop.mustprogress"}
;.
; CURRENT-CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]]}
; CURRENT-CHECK: [[META1]] = !{!"llvm.loop.peeled.count", i32 1}
; CURRENT-CHECK: [[LOOP2]] = distinct !{[[LOOP2]], [[META1]]}
; CURRENT-CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]]}
;.