➜
This issue trackerhas been migrated toGitHub, and is currentlyread-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.
Created on2019-05-24 05:33 bymethane, last changed2022-04-11 14:59 byadmin. This issue is nowclosed.
| Files | ||||
|---|---|---|---|---|
| File name | Uploaded | Description | Edit | |
| arena.py | tim.peters,2019-05-31 05:15 | Big test case - needs over 80GB of RAM | ||
| result.txt | methane,2019-05-31 15:51 | |||
| free_arenas.txt | fwolff,2019-06-03 05:59 | output with 2x and 4x strings | ||
| Pull Requests | |||
|---|---|---|---|
| URL | Status | Linked | Edit |
| PR 13612 | merged | tim.peters,2019-05-28 06:01 | |
| Messages (8) | |||
|---|---|---|---|
| msg343344 -(view) | Author: Inada Naoki (methane)* | Date: 2019-05-24 05:33 | |
Reported here:*https://stackoverflow.com/questions/56228799/python-hangs-indefinitely-trying-to-delete-deeply-recursive-object*https://mail.python.org/pipermail/python-dev/2019-May/157635.html | |||
| msg343734 -(view) | Author: Tim Peters (tim.peters)* | Date: 2019-05-28 06:05 | |
Created a PR and assigned myself to this bug. | |||
| msg344024 -(view) | Author: Tim Peters (tim.peters)* | Date: 2019-05-31 05:15 | |
Added file arena.py. This adds some code to the OP's original test, to print out build time and teardown time, and display obmalloc stats.You'll need at least 80GB of RAM to run it - I don't have that much. Building the tree may take on the order of an hour, but the time to tear it down is what's interesting here. Apparently 3.7.3 needs more than 4 hours to reclaim the memory. | |||
| msg344082 -(view) | Author: Inada Naoki (methane)* | Date: 2019-05-31 15:51 | |
master:build time 0:48:53.664428teardown time 4:58:20.132930patched:build time 0:48:08.485639teardown time 0:01:10.466707About 250x speedup! | |||
| msg344083 -(view) | Author: Tim Peters (tim.peters)* | Date: 2019-05-31 16:01 | |
Thank you so much, Inada! That's very good to hear :-) | |||
| msg344141 -(view) | Author: Tim Peters (tim.peters)* | Date: 2019-06-01 02:16 | |
New changeset1c263e39c4ed28225a7dc8ca1f24953225ac48ca by Tim Peters in branch 'master':bpo-37029: keep usable_arenas in sorted order without searching (#13612)https://github.com/python/cpython/commit/1c263e39c4ed28225a7dc8ca1f24953225ac48ca | |||
| msg344383 -(view) | Author: Friedel Wolff (fwolff) | Date: 2019-06-03 05:59 | |
I had a bit of time and access to a pretty big machine, so I ran the remove-arena-search branch with the same number of iterations, but also 2x and 4x as many. I attach the output in case it is interesting. The timings seem to be growing linearly with the number of strings. The teardown finishes within minutes. | |||
| msg344474 -(view) | Author: Tim Peters (tim.peters)* | Date: 2019-06-03 20:51 | |
Thank you, Friedl! I appreciate the extra confirmation - and especially on even bigger test cases :-) | |||
| History | |||
|---|---|---|---|
| Date | User | Action | Args |
| 2022-04-11 14:59:15 | admin | set | github: 81210 |
| 2019-06-03 20:51:08 | tim.peters | set | messages: +msg344474 |
| 2019-06-03 05:59:52 | fwolff | set | files: +free_arenas.txt nosy: +fwolff messages: +msg344383 |
| 2019-06-01 02:19:40 | tim.peters | set | status: open -> closed resolution: fixed stage: patch review -> resolved |
| 2019-06-01 02:16:08 | tim.peters | set | messages: +msg344141 |
| 2019-05-31 16:01:53 | tim.peters | set | messages: +msg344083 |
| 2019-05-31 15:51:44 | methane | set | files: +result.txt messages: +msg344082 |
| 2019-05-31 05:15:49 | tim.peters | set | files: +arena.py messages: +msg344024 |
| 2019-05-28 06:05:01 | tim.peters | set | assignee:tim.peters messages: +msg343734 |
| 2019-05-28 06:01:51 | tim.peters | set | keywords: +patch stage: patch review pull_requests: +pull_request13516 |
| 2019-05-24 12:40:43 | methane | set | title: PyObject_Free is O(N^2) where N = # of arenas -> PyObject_Free is O(N) where N = # of arenas |
| 2019-05-24 05:33:57 | methane | create | |