* Richard W. M. Jones: > On Mon, May 06, 2024 at 10:16:45AM +0200, Florian Weimer wrote: >> * Richard W. M. Jones: >> >> > This email is mostly to notify that this is happening. I'm not sure >> > if a single package slowing down compilation means we need to do >> > anything here, but if anyone else sees similar symptoms, let us know. >> >> What blocks you from profiling it and fixing the performance issue? > > I did profile it down to a couple of functions in a very dark corner > of the compiler (see the thread on cocci devel list that I posted). > However in the time available last week I didn't have the time to > understand what it was doing or fix it. It's computing a transitive closure of a graph, but in a strange way. The graph it chokes on has 12,732 vertices and 56,289 edges, and 3,449 vertices are Top. So it's really not very large, and the original graph is even quite sparse. This should fix it: flambda: Improve transitive closure in invariant_params_in_recursion <https://github.com/ocaml/ocaml/pull/13150> Compilation is still not exactly fast, but I think it's bearable (less than 40 seconds over here). My original comment was a bit of a snark, admittedly. This one was slightly more challenging than usual because it takes a while until invariant_params_in_recursion is reached, so if you hit ^C under GDB to early, you won't see where it's really spending its time. But I'm pretty sure profiling with just 10 samples per second would have revealed the problematic area just fine. Thanks, Florian -- _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
