diff options
| author | Razya Ladelsky <razya@gcc.gnu.org> | 2009-09-03 08:59:25 +0000 |
|---|---|---|
| committer | Razya Ladelsky <razya@gcc.gnu.org> | 2009-09-03 08:59:25 +0000 |
| commit | 785aa2a70653f35bd38f0880615e7b1d11c2e436 (patch) | |
| tree | 2b9e7a61b5a66c1df82fcc599b1f43406ad9814d /gcc/tree-parloops.c | |
| parent | ee5111a4417b8e62a1d6b8d83ca729338ea1c7c2 (diff) | |
| download | gcc-785aa2a70653f35bd38f0880615e7b1d11c2e436.zip gcc-785aa2a70653f35bd38f0880615e7b1d11c2e436.tar.gz gcc-785aa2a70653f35bd38f0880615e7b1d11c2e436.tar.bz2 | |
re PR tree-optimization/38275 (bootstrap failure when -ftree-parallelize-loops=4 is enabled)
PR tree-optimization/38275
* tree-parloops.c (parallelize_loops): Replace profitability condition
for expected number of iterations.
* testsuite/gcc.dg/autopar/reduc-1char.c: Increase number
of iterations. Adjust the logic accordingly.
* testsuite/gcc.dg/autopar/reduc-2char.c: Ditto.
* testsuite/gcc.dg/autopar/reduc-1.c: Ditto.
* testsuite/gcc.dg/autopar/reduc-2.c: Ditto.
* testsuite/gcc.dg/autopar/reduc-3.c: Ditto.
* testsuite/gcc.dg/autopar/reduc-6.c: Ditto.
* testsuite/gcc.dg/autopar/reduc-7.c: Ditto.
* testsuite/gcc.dg/autopar/reduc-8.c: Ditto.
* testsuite/gcc.dg/autopar/reduc-9.c: Ditto.
* testsuite/gcc.dg/autopar/pr39500-1.c: Ditto.
* testsuite/gcc.dg/autopar/reduc-1short.c: Ditto.
* testsuite/gcc.dg/autopar/reduc-2short.c: Ditto.
* testsuite/gcc.dg/autopar/parallelization-1.c: Ditto.
From-SVN: r151372
Diffstat (limited to 'gcc/tree-parloops.c')
| -rw-r--r-- | gcc/tree-parloops.c | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/gcc/tree-parloops.c b/gcc/tree-parloops.c index a6d8f215..b1ef8355 100644 --- a/gcc/tree-parloops.c +++ b/gcc/tree-parloops.c @@ -1908,7 +1908,7 @@ parallelize_loops (void) /* FIXME: Bypass this check as graphite doesn't update the count and frequency correctly now. */ if (!flag_loop_parallelize_all - && (expected_loop_iterations (loop) <= n_threads + && (estimated_loop_iterations_int (loop, false)<= n_threads * MIN_PER_THREAD /* Do not bother with loops in cold areas. */ || optimize_loop_nest_for_size_p (loop))) continue; |