1; RUN: opt < %s -basic-aa -loop-interchange -pass-remarks-missed='loop-interchange' \
2; RUN:   -pass-remarks-output=%t -verify-loop-info -verify-dom-info -S | FileCheck -check-prefix=IR %s
3; RUN: FileCheck --input-file=%t %s
4
5; RUN: opt < %s -basic-aa -loop-interchange -pass-remarks-missed='loop-interchange' \
6; RUN:   -da-disable-delinearization-checks -pass-remarks-output=%t             \
7; RUN:   -verify-loop-info -verify-dom-info -S | FileCheck -check-prefix=IR %s
8; RUN: FileCheck --check-prefix=DELIN --input-file=%t %s
9
10target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
11target triple = "x86_64-unknown-linux-gnu"
12
13@A = common global [100 x [100 x i32]] zeroinitializer
14@B = common global [100 x [100 x [100 x i32]]] zeroinitializer
15@C = common global [100 x [100 x i64]] zeroinitializer
16
17;;--------------------------------------Test case 01------------------------------------
18;; This loop can be interchanged with -da-disable-delinearization-checks, otherwise it cannot
19;; be interchanged due to dependence.
20;;  for(int i=0;i<N-1;i++)
21;;    for(int j=1;j<N-1;j++)
22;;      A[j+1][i+1] = A[j+1][i+1] + k;
23
24; CHECK:      Name:            Dependence
25; CHECK-NEXT: Function:        interchange_01
26
27; DELIN:      Name:            Interchanged
28; DELIN-NEXT: Function:        interchange_01
29define void @interchange_01(i32 %k, i32 %N) {
30 entry:
31   %sub = add nsw i32 %N, -1
32   %cmp26 = icmp sgt i32 %N, 1
33   br i1 %cmp26, label %for.cond1.preheader.lr.ph, label %for.end17
34
35 for.cond1.preheader.lr.ph:
36   %cmp324 = icmp sgt i32 %sub, 1
37   %0 = add i32 %N, -2
38   %1 = sext i32 %sub to i64
39   br label %for.cond1.preheader
40
41 for.cond.loopexit:
42   %cmp = icmp slt i64 %indvars.iv.next29, %1
43   br i1 %cmp, label %for.cond1.preheader, label %for.end17
44
45 for.cond1.preheader:
46   %indvars.iv28 = phi i64 [ 0, %for.cond1.preheader.lr.ph ], [ %indvars.iv.next29, %for.cond.loopexit ]
47   %indvars.iv.next29 = add nuw nsw i64 %indvars.iv28, 1
48   br i1 %cmp324, label %for.body4, label %for.cond.loopexit
49
50 for.body4:
51   %indvars.iv = phi i64 [ %indvars.iv.next, %for.body4 ], [ 1, %for.cond1.preheader ]
52   %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
53   %arrayidx7 = getelementptr inbounds [100 x [100 x i32]], [100 x [100 x i32]]* @A, i64 0, i64 %indvars.iv.next, i64 %indvars.iv.next29
54   %2 = load i32, i32* %arrayidx7
55   %add8 = add nsw i32 %2, %k
56   store i32 %add8, i32* %arrayidx7
57   %lftr.wideiv = trunc i64 %indvars.iv to i32
58   %exitcond = icmp eq i32 %lftr.wideiv, %0
59   br i1 %exitcond, label %for.cond.loopexit, label %for.body4
60
61 for.end17:
62   ret void
63}
64
65; When currently cannot interchange this loop, because transform currently
66; expects the latches to be the exiting blocks too.
67
68; IR-LABEL: @interchange_02
69; IR-NOT: split
70;
71; CHECK:      Name:            ExitingNotLatch
72; CHECK-NEXT: Function:        interchange_02
73define void @interchange_02(i64 %k, i64 %N) {
74entry:
75  br label %for1.header
76
77for1.header:
78  %j23 = phi i64 [ 0, %entry ], [ %j.next24, %for1.inc10 ]
79  br label %for2
80
81for2:
82  %j = phi i64 [ %j.next, %latch ], [ 0, %for1.header ]
83  %arrayidx5 = getelementptr inbounds [100 x [100 x i64]], [100 x [100 x i64]]* @C, i64 0, i64 %j, i64 %j23
84  %lv = load i64, i64* %arrayidx5
85  %add = add nsw i64 %lv, %k
86  store i64 %add, i64* %arrayidx5
87  %exitcond = icmp eq i64 %j, 99
88  br i1 %exitcond, label %for1.inc10, label %latch
89latch:
90  %j.next = add nuw nsw i64 %j, 1
91  br label %for2
92
93for1.inc10:
94  %j.next24 = add nuw nsw i64 %j23, 1
95  %exitcond26 = icmp eq i64 %j23, 99
96  br i1 %exitcond26, label %for.end12, label %for1.header
97
98for.end12:
99  ret void
100}
101