freenode/#clasp - IRC Chatlog

13:09:20 drmeister Hello everyone

14:45:46 Bike so with the find-class macro there's two things - we need a get-cell that creates a cell if it doesn't exist, and we need to be able to detect that a cell doesn't have a class in it. we could probably just put NIL in the cell or something

14:46:03 Bike actually you have it so that (setf (find-class 'foo) nil) removes the cell entirely, which probably isn't the best way to do it i think

14:47:17 beach I would just leave the cell in there.

14:48:22 Bike yeah

14:48:39 Bike technically that means we could get an accumulation of useless cells, but it seems unlikely to happen

14:49:01 Bike if it did happen we could GC them (with them being kept alive by any code find-class-ing those names)

14:49:25 beach Very unlikely, yes.

14:49:52 drmeister I see your point(s) - so I should change (setf (find-class 'foo) nil) to leave the cell in there.

15:05:49 drmeister Kevslinger: Are you online? If you are around can you come see me? Or check Slack.

15:06:05 Kevslinger Yeah I'm responding on slack

16:31:28 drmeister Bike: I added the core:atomic-t-holder class.

16:31:52 drmeister It will have a getter, setter, makunbound, unboundp

16:34:30 drmeister Bike: Regarding the cleavir changes. Should I submit a pull request to your fork of sicl and you can look it over, fix it up and then submit that to beach?

16:35:10 drmeister I added several class predicates and replaced (typep x 'foo) with (foo-p x)

16:36:49 beach drmeister: Bike has write permission to the SICL repository.

16:37:40 drmeister Understood. Clasp uses a fork under Bike's control.

17:09:06 Bike beach: oh yeah, about the predicate thing- did you mean just clasp changing its typep-for-classes to use generic functions, or explicitly writing out generic functions in the cleavir source? the latter is what drmeister is talking about

17:09:27 beach Yes, that's fine with me.

17:09:59 Bike OK, just checking

17:17:24 drmeister Bike: I can't submit a pull request to your fork of sicl because I'd have to have my own fork of sicl.

17:17:34 drmeister Can I just push my changes to your fork? Do I have write access?

17:18:41 Bike I uh... I think it says I invited you but you haven't responded.

17:18:58 Bike https://github.com/Bike/SICL/invitations maybe?

17:19:40 drmeister Done - it's pushed.

17:20:18 Bike ok, i'll figure out how to push that to master later

17:20:24 Bike how did the performance change with the hash table hint?

17:20:24 drmeister There are two commits - one with the predicates and use of predicates and a second where I added :size 1024 :rehash-size 4.0 to the map-instructions-xxx hash-tables.

17:21:16 drmeister I confess - I couldn't tell because of everything else I did. You could drop those ones and I can experiment with it a bit.

18:21:53 drmeister I did look at how much rehashing there was before and after. After - about half as much

18:23:56 drmeister I can’t do anything at the implementation level because I can’t tell these hash tables from others.