https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65594
--- Comment #4 from vries at gcc dot gnu.org --- (In reply to Jakub Jelinek from comment #2) > Well, GOMP_parallel doesn't always create new threads, but still it has some > synchronization overhead etc. IMNSHO the testcase should be scaled down for > the testsuite, IMHO lowering N to 100 would be in order. Lowering N to 400 or lower introduces this failure: ... FAIL: libgomp.graphite/force-parallel-6.c scan-tree-dump-times optimized "loopfn.0" 4 ... When N drops below (MIN_PER_THREAD * N_THREADS) == (100 * 4), the loopfn.0 is still created, but is already unreachable and the call to the function is removed during the following omp expand.
