1; REQUIRES: asserts
2; RUN: opt < %s -basic-aa -loop-interchange -cache-line-size=64 -verify-dom-info -verify-loop-info \
3; RUN:     -S -debug 2>&1 | FileCheck %s
4
5target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
6
7@A = common global [100 x [100 x i32]] zeroinitializer
8@B = common global [100 x i32] zeroinitializer
9@C = common global [100 x [100 x i32]] zeroinitializer
10@D = common global [100 x [100 x [100 x i32]]] zeroinitializer
11
12;; Test that a flow dependency in outer loop doesn't prevent interchange in
13;; loops i and j.
14;;
15;;  for (int k = 0; k < 100; ++k) {
16;;    T[k] = fn1();
17;;    for (int i = 0; i < 1000; ++i)
18;;      for(int j = 1; j < 1000; ++j)
19;;        Arr[j][i] = Arr[j][i]+k;
20;;    fn2(T[k]);
21;;  }
22
23; CHECK: Processing InnerLoopId = 2 and OuterLoopId = 1
24; CHECK: Loops interchanged.
25
26; CHECK: Processing InnerLoopId = 1 and OuterLoopId = 0
27; CHECK: Not interchanging loops. Cannot prove legality.
28
29@T = internal global [100 x double] zeroinitializer, align 4
30@Arr = internal global [1000 x [1000 x i32]] zeroinitializer, align 4
31
32define void @interchange_09(i32 %k) {
33entry:
34  br label %for.body
35
36for.cond.cleanup:                                 ; preds = %for.cond.cleanup4
37  ret void
38
39for.body:                                         ; preds = %for.cond.cleanup4, %entry
40  %indvars.iv45 = phi i64 [ 0, %entry ], [ %indvars.iv.next46, %for.cond.cleanup4 ]
41  %call = call double @fn1()
42  %arrayidx = getelementptr inbounds [100 x double], [100 x double]* @T, i64 0, i64 %indvars.iv45
43  store double %call, double* %arrayidx, align 8
44  br label %for.cond6.preheader
45
46for.cond6.preheader:                              ; preds = %for.cond.cleanup8, %for.body
47  %indvars.iv42 = phi i64 [ 0, %for.body ], [ %indvars.iv.next43, %for.cond.cleanup8 ]
48  br label %for.body9
49
50for.cond.cleanup4:                                ; preds = %for.cond.cleanup8
51  %tmp = load double, double* %arrayidx, align 8
52  call void @fn2(double %tmp)
53  %indvars.iv.next46 = add nuw nsw i64 %indvars.iv45, 1
54  %exitcond47 = icmp ne i64 %indvars.iv.next46, 100
55  br i1 %exitcond47, label %for.body, label %for.cond.cleanup
56
57for.cond.cleanup8:                                ; preds = %for.body9
58  %indvars.iv.next43 = add nuw nsw i64 %indvars.iv42, 1
59  %exitcond44 = icmp ne i64 %indvars.iv.next43, 1000
60  br i1 %exitcond44, label %for.cond6.preheader, label %for.cond.cleanup4
61
62for.body9:                                        ; preds = %for.body9, %for.cond6.preheader
63  %indvars.iv = phi i64 [ 1, %for.cond6.preheader ], [ %indvars.iv.next, %for.body9 ]
64  %arrayidx13 = getelementptr inbounds [1000 x [1000 x i32]], [1000 x [1000 x i32]]* @Arr, i64 0, i64 %indvars.iv, i64 %indvars.iv42
65  %tmp1 = load i32, i32* %arrayidx13, align 4
66  %tmp2 = trunc i64 %indvars.iv45 to i32
67  %add = add nsw i32 %tmp1, %tmp2
68  store i32 %add, i32* %arrayidx13, align 4
69  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
70  %exitcond = icmp ne i64 %indvars.iv.next, 1000
71  br i1 %exitcond, label %for.body9, label %for.cond.cleanup8
72}
73
74declare double @fn1() readnone
75declare void @fn2(double) readnone
76