xref: /llvm-project/polly/test/DependenceInfo/reduction_multiple_loops_array_sum.ll (revision e1f056f692d869708c1898d9d65a69ac5584a0ed)
1; RUN: opt -aa-pipeline=basic-aa %loadNPMPolly '-passes=print<polly-dependences>' -disable-output < %s | FileCheck %s
2; RUN: opt -aa-pipeline=basic-aa %loadNPMPolly '-passes=print<polly-dependences>' -polly-dependences-analysis-level=reference-wise -disable-output < %s | FileCheck %s
3; RUN: opt -aa-pipeline=basic-aa %loadNPMPolly '-passes=print<polly-dependences>' -polly-dependences-analysis-level=access-wise -disable-output < %s | FileCheck %s
4;
5; Verify that only the inner reduction like accesses cause reduction dependences
6;
7; CHECK:      Reduction dependences:
8; CHECK-NEXT:     { Stmt_for_body3[i0, i1] -> Stmt_for_body3[i0, 1 + i1] : 0 <= i0 <= 99 and 0 <= i1 <= 98 }
9;
10; void f(int * restrict A, int * restrict sum) {
11;   int i, j, k;
12;   for (i = 0; i < 100; i++) {
13;     *sum *= 7;
14;     for (j = 0; j < 100; j++) {
15;       *sum += A[i+j];
16;       for (k = 0; k< 100; k++) {}
17;     }
18;   }
19; }
20target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-n32-S64"
21
22define void @f(ptr noalias %A, ptr noalias %sum) {
23entry:
24  br label %for.cond
25
26for.cond:                                         ; preds = %for.inc11, %entry
27  %i.0 = phi i32 [ 0, %entry ], [ %inc12, %for.inc11 ]
28  %exitcond2 = icmp ne i32 %i.0, 100
29  br i1 %exitcond2, label %for.body, label %for.end13
30
31for.body:                                         ; preds = %for.cond
32  %tmp = load i32, ptr %sum, align 4
33  %mul = mul nsw i32 %tmp, 7
34  store i32 %mul, ptr %sum, align 4
35  br label %for.cond1
36
37for.cond1:                                        ; preds = %for.inc8, %for.body
38  %j.0 = phi i32 [ 0, %for.body ], [ %inc9, %for.inc8 ]
39  %exitcond1 = icmp ne i32 %j.0, 100
40  br i1 %exitcond1, label %for.body3, label %for.end10
41
42for.body3:                                        ; preds = %for.cond1
43  %add = add nsw i32 %i.0, %j.0
44  %arrayidx = getelementptr inbounds i32, ptr %A, i32 %add
45  %tmp3 = load i32, ptr %arrayidx, align 4
46  %tmp4 = load i32, ptr %sum, align 4
47  %add4 = add nsw i32 %tmp4, %tmp3
48  store i32 %add4, ptr %sum, align 4
49  br label %for.cond5
50
51for.cond5:                                        ; preds = %for.inc, %for.body3
52  %k.0 = phi i32 [ 0, %for.body3 ], [ %inc, %for.inc ]
53  %exitcond = icmp ne i32 %k.0, 100
54  br i1 %exitcond, label %for.body7, label %for.end
55
56for.body7:                                        ; preds = %for.cond5
57  br label %for.inc
58
59for.inc:                                          ; preds = %for.body7
60  %inc = add nsw i32 %k.0, 1
61  br label %for.cond5
62
63for.end:                                          ; preds = %for.cond5
64  br label %for.inc8
65
66for.inc8:                                         ; preds = %for.end
67  %inc9 = add nsw i32 %j.0, 1
68  br label %for.cond1
69
70for.end10:                                        ; preds = %for.cond1
71  br label %for.inc11
72
73for.inc11:                                        ; preds = %for.end10
74  %inc12 = add nsw i32 %i.0, 1
75  br label %for.cond
76
77for.end13:                                        ; preds = %for.cond
78  ret void
79}
80