2 ; RUN: opt < %s -passes=loop-interchange -S -debug 2>&1 | FileCheck %s
4 ; CHECK: Dependency matrix before interchange:
8 ; CHECK-NEXT: Processing InnerLoopId
10 ; This example is taken from github issue #54176
12 define void @foo(i32 noundef %n, i32 noundef %m, ptr nocapture noundef %aa, ptr nocapture noundef readonly %bb, ptr nocapture noundef writeonly %cc) {
14 %arrayidx7 = getelementptr inbounds i8, ptr %aa, i64 512
15 br label %for.cond1.preheader
18 %indvars.iv32 = phi i64 [ 1, %entry ], [ %indvars.iv.next33, %for.cond.cleanup3 ]
19 %0 = add nsw i64 %indvars.iv32, -1
20 %arrayidx9 = getelementptr inbounds [128 x float], ptr %arrayidx7, i64 0, i64 %0
21 %arrayidx12 = getelementptr inbounds [128 x float], ptr %arrayidx7, i64 0, i64 %indvars.iv32
28 %indvars.iv.next33 = add nuw nsw i64 %indvars.iv32, 1
29 %exitcond36 = icmp ne i64 %indvars.iv.next33, 128
30 br i1 %exitcond36, label %for.cond1.preheader, label %for.cond.cleanup
33 %indvars.iv = phi i64 [ 1, %for.cond1.preheader ], [ %indvars.iv.next, %for.body4 ]
34 %arrayidx6 = getelementptr inbounds [128 x float], ptr %bb, i64 %indvars.iv, i64 %indvars.iv32
35 %1 = load float, ptr %arrayidx6, align 4
36 %2 = load float, ptr %arrayidx9, align 4
37 %add = fadd fast float %2, %1
38 store float %add, ptr %arrayidx9, align 4
39 %3 = load float, ptr %arrayidx12, align 4
40 %arrayidx16 = getelementptr inbounds [128 x float], ptr %cc, i64 %indvars.iv, i64 %indvars.iv32
41 store float %3, ptr %arrayidx16, align 4
42 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
43 %exitcond = icmp ne i64 %indvars.iv.next, 1024
44 br i1 %exitcond, label %for.body4, label %for.cond.cleanup3