xref: /llvm-project/llvm/test/Transforms/LoopFusion/inner_loops.ll (revision 055fb7795aa219a3d274d280ec9129784f169f56)
1; RUN: opt -S -passes=loop-fusion < %s 2>&1 | FileCheck %s
2
3@A = common global [1024 x [1024 x i32]] zeroinitializer, align 16
4@B = common global [1024 x [1024 x i32]] zeroinitializer, align 16
5
6; CHECK: void @dep_free
7; CHECK-NEXT: bb:
8; CHECK-NEXT: br label %[[LOOP1HEADER:bb[0-9]*]]
9; CHECK: [[LOOP1HEADER]]
10; CHECK: br i1 %{{.*}}, label %[[LOOP1BODY:bb[0-9]*]], label %[[LOOP2PREHEADER:bb[0-9]+]]
11; CHECK: [[LOOP1BODY]]
12; CHECK: br label %[[LOOP1LATCH:bb[0-9]*]]
13; CHECK: [[LOOP1LATCH]]
14; CHECK: br label %[[LOOP2PREHEADER:bb[0-9]+]]
15; CHECK: [[LOOP2PREHEADER]]
16; CHECK: br i1 %{{.*}}, label %[[LOOP2BODY:bb[0-9]*]], label %[[LOOP2EXIT:bb[0-9]*]]
17; CHECK: [[LOOP2BODY]]
18; CHECK: br label %[[LOOP2LATCH:bb[0-9]+]]
19; CHECK: [[LOOP2LATCH]]
20; CHECK: br label %[[LOOP1HEADER]]
21; CHECK: ret void
22
23define void @dep_free() {
24bb:
25  br label %bb9
26
27bb9:                                              ; preds = %bb35, %bb
28  %indvars.iv6 = phi i64 [ %indvars.iv.next7, %bb35 ], [ 0, %bb ]
29  %.0 = phi i32 [ 0, %bb ], [ %tmp36, %bb35 ]
30  %exitcond8 = icmp ne i64 %indvars.iv6, 100
31  br i1 %exitcond8, label %bb11, label %bb10
32
33bb10:                                             ; preds = %bb9
34  br label %bb37
35
36bb11:                                             ; preds = %bb9
37  br label %bb12
38
39bb12:                                             ; preds = %bb21, %bb11
40  %indvars.iv = phi i64 [ %indvars.iv.next, %bb21 ], [ 0, %bb11 ]
41  %exitcond = icmp ne i64 %indvars.iv, 100
42  br i1 %exitcond, label %bb14, label %bb23
43
44bb14:                                             ; preds = %bb12
45  %tmp = add nsw i32 %.0, -3
46  %tmp15 = add nuw nsw i64 %indvars.iv6, 3
47  %tmp16 = trunc i64 %tmp15 to i32
48  %tmp17 = mul nsw i32 %tmp, %tmp16
49  %tmp18 = trunc i64 %indvars.iv6 to i32
50  %tmp19 = srem i32 %tmp17, %tmp18
51  %tmp20 = getelementptr inbounds [1024 x [1024 x i32]], ptr @A, i64 0, i64 %indvars.iv6, i64 %indvars.iv
52  store i32 %tmp19, ptr %tmp20, align 4
53  br label %bb21
54
55bb21:                                             ; preds = %bb14
56  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
57  br label %bb12
58
59bb23:                                             ; preds = %bb33, %bb12
60  %indvars.iv3 = phi i64 [ %indvars.iv.next4, %bb33 ], [ 0, %bb12 ]
61  %exitcond5 = icmp ne i64 %indvars.iv3, 100
62  br i1 %exitcond5, label %bb25, label %bb35
63
64bb25:                                             ; preds = %bb23
65  %tmp26 = add nsw i32 %.0, -3
66  %tmp27 = add nuw nsw i64 %indvars.iv6, 3
67  %tmp28 = trunc i64 %tmp27 to i32
68  %tmp29 = mul nsw i32 %tmp26, %tmp28
69  %tmp30 = trunc i64 %indvars.iv6 to i32
70  %tmp31 = srem i32 %tmp29, %tmp30
71  %tmp32 = getelementptr inbounds [1024 x [1024 x i32]], ptr @B, i64 0, i64 %indvars.iv6, i64 %indvars.iv3
72  store i32 %tmp31, ptr %tmp32, align 4
73  br label %bb33
74
75bb33:                                             ; preds = %bb25
76  %indvars.iv.next4 = add nuw nsw i64 %indvars.iv3, 1
77  br label %bb23
78
79bb35:                                             ; preds = %bb23
80  %indvars.iv.next7 = add nuw nsw i64 %indvars.iv6, 1
81  %tmp36 = add nuw nsw i32 %.0, 1
82  br label %bb9
83
84bb37:                                             ; preds = %bb10
85  ret void
86}
87