libera/#sicl - IRC Chatlog
Search
11:28:28
scymtym
jcowan: speaking of eclector branches, i think you requested (or maybe suggested) this change: https://github.com/s-expressionists/Eclector/compare/relaxed-sharpsign-s~2...relaxed-sharpsign-s . i'm planning to merge this sometime soon
12:15:49
hayley
One oddity of non-moving collection I observed today (but had to get someone else to come up with reasoning for). When pages are partly filled, it requires more pages to allocate a certain amount of space (spread among many small objects), which necessitates more locking, if one must lock when getting more pages.
12:16:57
hayley
With one benchmark I have, the mutator time is pretty constant with regards to heap size, when using a copying GC, but has a factor which is inversely proportional to heap size when using the mark-region GC.
12:17:56
hayley
The factor is small (about 5% total), but enough to make the mutator faster with mark-region collector than with copying in large heaps, and slower with smaller heaps.
12:18:58
hayley
Fixes would include trying to avoid locking while getting more pages, or compacting so that fewer pages need to be claimed.
13:10:43
Bike
the github branches display compactly shows which branches actually ahve commits that aren't in the main branch. it's pretty nice.
15:33:57
jcowan
scymtym: When I got an error message trying to click on the branch, I assumed the URL was bad, but I now understand the branch is closed. No problem.
15:34:32
jcowan
If I requested the relaxation of #S, I don't remember it, but the change makes sense to me.