1; RUN: opt < %s -basic-aa -loop-interchange -pass-remarks-missed='loop-interchange' -pass-remarks-output=%t -S \
2; RUN:     -verify-dom-info -verify-loop-info -verify-loop-lcssa 2>&1 | FileCheck -check-prefix=IR %s
3; RUN: FileCheck --input-file=%t %s
4
5; Inner loop only reductions are not supported currently. See discussion at
6; D53027 for more information on the required checks.
7
8target triple = "powerpc64le-unknown-linux-gnu"
9@A = common global [500 x [500 x i32]] zeroinitializer
10@X = common global i32 0
11@B = common global [500 x [500 x i32]] zeroinitializer
12@Y = common global i32 0
13
14;; global X
15
16;;  for( int i=1;i<N;i++)
17;;    for( int j=1;j<N;j++)
18;;      X+=A[j][i];
19
20; CHECK: --- !Missed
21; CHECK-NEXT: Pass:            loop-interchange
22; CHECK-NEXT: Name:            UnsupportedPHI
23; CHECK-NEXT: Function:        reduction_01
24
25; IR-LABEL: @reduction_01(
26; IR-NOT: split
27
28define void @reduction_01(i32 %N) {
29entry:
30  %cmp16 = icmp sgt i32 %N, 1
31  br i1 %cmp16, label %for.body3.lr.ph, label %for.end8
32
33for.body3.lr.ph:                                  ; preds = %for.cond1.for.inc6_crit_edge, %entry
34  %indvars.iv18 = phi i64 [ %indvars.iv.next19, %for.cond1.for.inc6_crit_edge ], [ 1, %entry ]
35  %X.promoted = load i32, i32* @X
36  br label %for.body3
37
38for.body3:                                        ; preds = %for.body3, %for.body3.lr.ph
39  %indvars.iv = phi i64 [ 1, %for.body3.lr.ph ], [ %indvars.iv.next, %for.body3 ]
40  %add15 = phi i32 [ %X.promoted, %for.body3.lr.ph ], [ %add, %for.body3 ]
41  %arrayidx5 = getelementptr inbounds [500 x [500 x i32]], [500 x [500 x i32]]* @A, i64 0, i64 %indvars.iv, i64 %indvars.iv18
42  %0 = load i32, i32* %arrayidx5
43  %add = add nsw i32 %add15, %0
44  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
45  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
46  %exitcond = icmp eq i32 %lftr.wideiv, %N
47  br i1 %exitcond, label %for.cond1.for.inc6_crit_edge, label %for.body3
48
49for.cond1.for.inc6_crit_edge:                     ; preds = %for.body3
50  %add.lcssa = phi i32 [ %add, %for.body3 ]
51  store i32 %add.lcssa, i32* @X
52  %indvars.iv.next19 = add nuw nsw i64 %indvars.iv18, 1
53  %lftr.wideiv20 = trunc i64 %indvars.iv.next19 to i32
54  %exitcond21 = icmp eq i32 %lftr.wideiv20, %N
55  br i1 %exitcond21, label %for.end8, label %for.body3.lr.ph
56
57for.end8:                                         ; preds = %for.cond1.for.inc6_crit_edge, %entry
58  ret void
59}
60
61;; Not tightly nested. Do not interchange.
62;;  for( int i=1;i<N;i++)
63;;    for( int j=1;j<N;j++) {
64;;      for( int k=1;k<N;k++) {
65;;        X+=A[k][j];
66;;      }
67;;      Y+=B[j][i];
68;;    }
69
70;; Not tightly nested. Do not interchange.
71;; Not interchanged hence the phi's in the inner loop will not be split.
72
73; CHECK: --- !Missed
74; CHECK-NEXT: Pass:            loop-interchange
75; CHECK-NEXT: Name:            UnsupportedPHIOuter
76; CHECK-NEXT: Function:        reduction_03
77
78; IR-LABEL: @reduction_03(
79; IR-NOT: split
80
81define void @reduction_03(i32 %N) {
82entry:
83  %cmp35 = icmp sgt i32 %N, 1
84  br i1 %cmp35, label %for.cond4.preheader.lr.ph, label %for.end19
85
86for.cond4.preheader.lr.ph:                        ; preds = %for.cond1.for.inc17_crit_edge, %entry
87  %indvars.iv41 = phi i64 [ %indvars.iv.next42, %for.cond1.for.inc17_crit_edge ], [ 1, %entry ]
88  %Y.promoted = load i32, i32* @Y
89  br label %for.body6.lr.ph
90
91for.body6.lr.ph:                                  ; preds = %for.cond4.for.end_crit_edge, %for.cond4.preheader.lr.ph
92  %indvars.iv37 = phi i64 [ 1, %for.cond4.preheader.lr.ph ], [ %indvars.iv.next38, %for.cond4.for.end_crit_edge ]
93  %add1334 = phi i32 [ %Y.promoted, %for.cond4.preheader.lr.ph ], [ %add13, %for.cond4.for.end_crit_edge ]
94  %X.promoted = load i32, i32* @X
95  br label %for.body6
96
97for.body6:                                        ; preds = %for.body6, %for.body6.lr.ph
98  %indvars.iv = phi i64 [ 1, %for.body6.lr.ph ], [ %indvars.iv.next, %for.body6 ]
99  %arrayidx8 = getelementptr inbounds [500 x [500 x i32]], [500 x [500 x i32]]* @A, i64 0, i64 %indvars.iv, i64 %indvars.iv37
100  %0 = load i32, i32* %arrayidx8
101  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
102  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
103  %exitcond = icmp eq i32 %lftr.wideiv, %N
104  br i1 %exitcond, label %for.cond4.for.end_crit_edge, label %for.body6
105
106for.cond4.for.end_crit_edge:                      ; preds = %for.body6
107  %arrayidx12 = getelementptr inbounds [500 x [500 x i32]], [500 x [500 x i32]]* @B, i64 0, i64 %indvars.iv37, i64 %indvars.iv41
108  %1 = load i32, i32* %arrayidx12
109  %add13 = add nsw i32 %add1334, %1
110  %indvars.iv.next38 = add nuw nsw i64 %indvars.iv37, 1
111  %lftr.wideiv39 = trunc i64 %indvars.iv.next38 to i32
112  %exitcond40 = icmp eq i32 %lftr.wideiv39, %N
113  br i1 %exitcond40, label %for.cond1.for.inc17_crit_edge, label %for.body6.lr.ph
114
115for.cond1.for.inc17_crit_edge:                    ; preds = %for.cond4.for.end_crit_edge
116  %add13.lcssa = phi i32 [ %add13, %for.cond4.for.end_crit_edge ]
117  store i32 %add13.lcssa, i32* @Y
118  %indvars.iv.next42 = add nuw nsw i64 %indvars.iv41, 1
119  %lftr.wideiv43 = trunc i64 %indvars.iv.next42 to i32
120  %exitcond44 = icmp eq i32 %lftr.wideiv43, %N
121  br i1 %exitcond44, label %for.end19, label %for.cond4.preheader.lr.ph
122
123for.end19:                                        ; preds = %for.cond1.for.inc17_crit_edge, %entry
124  ret void
125}
126