1 ; RUN: opt %loadPolly -basicaa -polly-ast -polly-ast-detect-parallel -analyze < %s | FileCheck %s
3 ; This loopnest contains a reduction which imposes the same dependences as the
4 ; accesses to the array A. We need to ensure we do __not__ parallelize anything
7 ; CHECK: #pragma minimal dependence distance: 1
11 ; void AandSum(int *restrict sum, int *restrict A) {
12 ; for (int i = 0; i < 1024; i++) {
13 ; A[i] = A[i] + A[i - 1];
14 ; A[i - 1] = A[i] + A[i - 2];
19 target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-n32-S64"
21 define void @AandSum(i32* noalias %sum, i32* noalias %A) {
25 for.cond: ; preds = %for.inc, %entry
26 %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
27 %exitcond = icmp ne i32 %i.0, 1024
28 br i1 %exitcond, label %for.body, label %for.end
30 for.body: ; preds = %for.cond
31 %arrayidx = getelementptr inbounds i32, i32* %A, i32 %i.0
32 %tmp = load i32, i32* %arrayidx, align 4
33 %sub = add nsw i32 %i.0, -1
34 %arrayidx1 = getelementptr inbounds i32, i32* %A, i32 %sub
35 %tmp1 = load i32, i32* %arrayidx1, align 4
36 %add = add nsw i32 %tmp, %tmp1
37 %arrayidx2 = getelementptr inbounds i32, i32* %A, i32 %i.0
38 store i32 %add, i32* %arrayidx2, align 4
39 %sub4 = add nsw i32 %i.0, -2
40 %arrayidx5 = getelementptr inbounds i32, i32* %A, i32 %sub4
41 %tmp2 = load i32, i32* %arrayidx5, align 4
42 %add6 = add nsw i32 %add, %tmp2
43 %sub7 = add nsw i32 %i.0, -1
44 %arrayidx8 = getelementptr inbounds i32, i32* %A, i32 %sub7
45 store i32 %add6, i32* %arrayidx8, align 4
46 %tmp3 = load i32, i32* %sum, align 4
47 %add9 = add nsw i32 %tmp3, %i.0
48 store i32 %add9, i32* %sum, align 4
51 for.inc: ; preds = %for.body
52 %inc = add nsw i32 %i.0, 1
55 for.end: ; preds = %for.cond