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;; [FIXME] This loop though valid is currently not interchanged due to the limitation that we cannot split the inner loop latch due to multiple use of inner induction
19;; variable.(used to increment the loop counter and to access A[j+1][i+1]
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; IR-LABEL: @interchange_01
25; IR-NOT: split
26
27; CHECK:      Name:            Dependence
28; CHECK-NEXT: Function:        interchange_01
29
30; DELIN:      Name:            UnsupportedInsBetweenInduction
31; DELIN-NEXT: Function:        interchange_01
32define void @interchange_01(i32 %k, i32 %N) {
33 entry:
34   %sub = add nsw i32 %N, -1
35   %cmp26 = icmp sgt i32 %N, 1
36   br i1 %cmp26, label %for.cond1.preheader.lr.ph, label %for.end17
37
38 for.cond1.preheader.lr.ph:
39   %cmp324 = icmp sgt i32 %sub, 1
40   %0 = add i32 %N, -2
41   %1 = sext i32 %sub to i64
42   br label %for.cond1.preheader
43
44 for.cond.loopexit:
45   %cmp = icmp slt i64 %indvars.iv.next29, %1
46   br i1 %cmp, label %for.cond1.preheader, label %for.end17
47
48 for.cond1.preheader:
49   %indvars.iv28 = phi i64 [ 0, %for.cond1.preheader.lr.ph ], [ %indvars.iv.next29, %for.cond.loopexit ]
50   %indvars.iv.next29 = add nuw nsw i64 %indvars.iv28, 1
51   br i1 %cmp324, label %for.body4, label %for.cond.loopexit
52
53 for.body4:
54   %indvars.iv = phi i64 [ %indvars.iv.next, %for.body4 ], [ 1, %for.cond1.preheader ]
55   %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
56   %arrayidx7 = getelementptr inbounds [100 x [100 x i32]], [100 x [100 x i32]]* @A, i64 0, i64 %indvars.iv.next, i64 %indvars.iv.next29
57   %2 = load i32, i32* %arrayidx7
58   %add8 = add nsw i32 %2, %k
59   store i32 %add8, i32* %arrayidx7
60   %lftr.wideiv = trunc i64 %indvars.iv to i32
61   %exitcond = icmp eq i32 %lftr.wideiv, %0
62   br i1 %exitcond, label %for.cond.loopexit, label %for.body4
63
64 for.end17:
65   ret void
66}
67
68; When currently cannot interchange this loop, because transform currently
69; expects the latches to be the exiting blocks too.
70
71; IR-LABEL: @interchange_02
72; IR-NOT: split
73;
74; CHECK:      Name:            ExitingNotLatch
75; CHECK-NEXT: Function:        interchange_02
76define void @interchange_02(i64 %k, i64 %N) {
77entry:
78  br label %for1.header
79
80for1.header:
81  %j23 = phi i64 [ 0, %entry ], [ %j.next24, %for1.inc10 ]
82  br label %for2
83
84for2:
85  %j = phi i64 [ %j.next, %latch ], [ 0, %for1.header ]
86  %arrayidx5 = getelementptr inbounds [100 x [100 x i64]], [100 x [100 x i64]]* @C, i64 0, i64 %j, i64 %j23
87  %lv = load i64, i64* %arrayidx5
88  %add = add nsw i64 %lv, %k
89  store i64 %add, i64* %arrayidx5
90  %exitcond = icmp eq i64 %j, 99
91  br i1 %exitcond, label %for1.inc10, label %latch
92latch:
93  %j.next = add nuw nsw i64 %j, 1
94  br label %for2
95
96for1.inc10:
97  %j.next24 = add nuw nsw i64 %j23, 1
98  %exitcond26 = icmp eq i64 %j23, 99
99  br i1 %exitcond26, label %for.end12, label %for1.header
100
101for.end12:
102  ret void
103}
104