xref: /llvm-project/polly/test/IstAstInfo/reduction_modulo_schedule.ll (revision b332499a94df11870dfc7598645c59656deb933d)
1; RUN: opt %loadPolly -polly-import-jscop -polly-print-ast -polly-ast-detect-parallel -disable-output < %s | FileCheck %s
2; RUN: opt %loadPolly -print-polyhedral-info -polly-check-parallel -disable-output < %s | FileCheck %s -check-prefix=PINFO
3;
4; CHECK:          #pragma known-parallel reduction (+ : MemRef_A)
5; CHECK-NEXT:     for (int c0 = 0; c0 <= 2; c0 += 1) {
6; CHECK-NEXT:       if (c0 == 1) {
7; CHECK-NEXT:         #pragma simd
8; CHECK-NEXT:         for (int c1 = 0; c1 < 2 * n; c1 += 1)
9; CHECK-NEXT:           Stmt_S1(c1);
10; CHECK-NEXT:       } else
11; CHECK-NEXT:         #pragma simd reduction (+ : MemRef_A)
12; CHECK-NEXT:         for (int c1 = c0 / 2; c1 < 2 * n; c1 += 2)
13; CHECK-NEXT:           Stmt_S0(c1);
14; CHECK-NEXT:     }
15;
16; PINFO:      for.cond2: Loop is parallel.
17; PINFO-NEXT: for.cond: Loop is not parallel.
18;
19;    void rms(int *A, long n) {
20;      for (long i = 0; i < 2 * n; i++)
21; S0:    A[0] += i;
22;      for (long i = 0; i < 2 * n; i++)
23; S1:    A[i + 1] = 1;
24;    }
25;
26target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-n32-S64"
27
28define void @rms(ptr %A, i32 %n) {
29entry:
30  br label %for.cond
31
32for.cond:                                         ; preds = %for.inc, %entry
33  %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
34  %mul = shl nsw i32 %n, 1
35  %cmp = icmp slt i32 %i.0, %mul
36  br i1 %cmp, label %for.body, label %for.end
37
38for.body:                                         ; preds = %for.cond
39  br label %S0
40
41S0:                                               ; preds = %for.body
42  %tmp = load i32, ptr %A, align 4
43  %add = add nsw i32 %tmp, %i.0
44  store i32 %add, ptr %A, align 4
45  br label %for.inc
46
47for.inc:                                          ; preds = %S0
48  %inc = add nsw i32 %i.0, 1
49  br label %for.cond
50
51for.end:                                          ; preds = %for.cond
52  br label %for.cond2
53
54for.cond2:                                        ; preds = %for.inc8, %for.end
55  %i1.0 = phi i32 [ 0, %for.end ], [ %inc9, %for.inc8 ]
56  %mul3 = shl nsw i32 %n, 1
57  %cmp4 = icmp slt i32 %i1.0, %mul3
58  br i1 %cmp4, label %for.body5, label %for.end10
59
60for.body5:                                        ; preds = %for.cond2
61  br label %S1
62
63S1:                                               ; preds = %for.body5
64  %add6 = add nsw i32 %i1.0, 1
65  %arrayidx7 = getelementptr inbounds i32, ptr %A, i32 %add6
66  store i32 1, ptr %arrayidx7, align 4
67  br label %for.inc8
68
69for.inc8:                                         ; preds = %S1
70  %inc9 = add nsw i32 %i1.0, 1
71  br label %for.cond2
72
73for.end10:                                        ; preds = %for.cond2
74  ret void
75}
76
77