xref: /llvm-project/llvm/test/Transforms/LoopVectorize/PowerPC/pr41179.ll (revision 7d7577256b76e4293f455b8093504d5f7044ab4b)
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s -S -passes=loop-vectorize -mtriple=powerpc64-unknown-linux-gnu | FileCheck %s
3
4define void @foo(ptr %start, ptr %end) {
5; CHECK-LABEL: @foo(
6; CHECK-NEXT:  entry:
7; CHECK-NEXT:    [[START2:%.*]] = ptrtoint ptr [[START:%.*]] to i64
8; CHECK-NEXT:    [[END1:%.*]] = ptrtoint ptr [[END:%.*]] to i64
9; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[END1]] to i32
10; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[END1]], -1
11; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[START2]], i64 [[TMP1]])
12; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[UMIN]] to i32
13; CHECK-NEXT:    [[TMP3:%.*]] = sub i32 [[TMP0]], [[TMP2]]
14; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP3]], 2
15; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
16; CHECK:       vector.ph:
17; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP3]], 2
18; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP3]], [[N_MOD_VF]]
19; CHECK-NEXT:    [[IND_END:%.*]] = sub i32 0, [[N_VEC]]
20; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
21; CHECK:       vector.body:
22; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
23; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 0, [[INDEX]]
24; CHECK-NEXT:    [[INDUCTION:%.*]] = add i32 [[OFFSET_IDX]], 0
25; CHECK-NEXT:    [[INDUCTION3:%.*]] = add i32 [[OFFSET_IDX]], -1
26; CHECK-NEXT:    [[TMP4:%.*]] = add nsw i32 -1, [[INDUCTION]]
27; CHECK-NEXT:    [[TMP5:%.*]] = add nsw i32 -1, [[INDUCTION3]]
28; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[END]], i32 [[TMP4]]
29; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[END]], i32 [[TMP5]]
30; CHECK-NEXT:    store i8 0, ptr [[TMP6]], align 1
31; CHECK-NEXT:    store i8 0, ptr [[TMP7]], align 1
32; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
33; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
34; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
35; CHECK:       middle.block:
36; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
37; CHECK-NEXT:    br i1 [[CMP_N]], label [[WHILE_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
38; CHECK:       scalar.ph:
39; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
40; CHECK-NEXT:    br label [[WHILE_BODY:%.*]]
41; CHECK:       while.body:
42; CHECK-NEXT:    [[COUNT_09:%.*]] = phi i32 [ [[ADD:%.*]], [[WHILE_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
43; CHECK-NEXT:    [[ADD]] = add nsw i32 -1, [[COUNT_09]]
44; CHECK-NEXT:    [[G:%.*]] = getelementptr i8, ptr [[END]], i32 [[ADD]]
45; CHECK-NEXT:    store i8 0, ptr [[G]], align 1
46; CHECK-NEXT:    [[CMP:%.*]] = icmp ult ptr [[START]], [[G]]
47; CHECK-NEXT:    br i1 [[CMP]], label [[WHILE_BODY]], label [[WHILE_END_LOOPEXIT]], !llvm.loop [[LOOP2:![0-9]+]]
48; CHECK:       while.end.loopexit:
49; CHECK-NEXT:    ret void
50;
51entry:
52  br label %while.body
53
54while.body:                                       ; preds = %while.body, %entry
55  %count.09 = phi i32 [ %add, %while.body ], [ 0, %entry ]
56  %add = add nsw i32 -1, %count.09
57  %G = getelementptr i8, ptr %end, i32 %add
58  store i8 0, ptr %G
59  %cmp = icmp ult ptr %start, %G
60  br i1 %cmp, label %while.body, label %while.end.loopexit
61
62while.end.loopexit:                               ; preds = %while.body
63  ret void
64}
65