https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93439
--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> --- I'd say the bug is that we parallelize a loop (loop 7) which has two inner loops (loop 10 and loop 8) and then have some code to avoid trying to parallelize inner loops of the already parallelized loop, but the skip_loop stuff simply assumes that there can be at most one inner loop of a parallelized loop. Now, I don't know if it is incorrect to try to parallelize a loop with more than one inner loop, or if the skip_loop stuff needs adjusting, so that we don't skip just one, but all inner loops of already parallelized loops.
