111
|
1 /* { dg-do compile } */
|
|
2 /* { dg-options "-O2 -ftree-loop-distribution -fdump-tree-ldist-all" } */
|
|
3
|
|
4 int loop1 (int k)
|
|
5 {
|
|
6 unsigned int i;
|
|
7 int a[1000], b[1000], c[1000], d[1000];
|
|
8
|
|
9 for (i = 2; i < (1000-1); i ++) {
|
|
10 a[i] = k * i; /* S1 */
|
|
11 b[i] = a[i-2] + k; /* S2 */
|
|
12 c[i] = b[i-1] + a[i+1]; /* S3 */
|
|
13 d[i] = c[i-1] + k + i; /* S4 */
|
|
14 }
|
|
15 /* Dependences:
|
|
16 S1->S2 (flow, level 1)
|
|
17 S2->S3 (flow, level 1)
|
|
18 S3->S1 (anti, level 1)
|
|
19 S3->S4 (flow, level 1)
|
|
20
|
|
21 There are two partitions: {S1, S2, S3} and {S4}.
|
|
22
|
|
23 {S1, S2, S3} have to be in the same partition because:
|
|
24 - S1 (i) has to be executed before S2 (i+2), as S1 produces a[i] that is then consumed 2 iterations later by S2.
|
|
25 - S2 (i) has to be executed before S3 (i+1), as S2 produces b[i] that is then consumed one iteration later by S3,
|
|
26 - S3 (i) has to be executed before S1 (i+1), as a[i+1] has to execute before the update to a[i],
|
|
27
|
|
28 {S4} is the consumer partition: it consumes the values from array "c" produced in S3.
|
|
29
|
|
30 The cost model should fuse all the tasks together as the cost of
|
|
31 fetching data from caches is too high.
|
|
32 */
|
|
33
|
|
34 return a[1000-2] + b[1000-1] + c[1000-2] + d[1000-2];
|
|
35 }
|
|
36
|
|
37 /* { dg-final { scan-tree-dump-times "distributed: split to 2 loops" 0 "ldist" { xfail *-*-* } } } */
|