freenode/#clasp - IRC Chatlog
Search
20:41:20
karlosz
OK, so i think this algorithm avoids calling reinitialize instance and also only uses 1 hash table. aste.gnome.org/padtgsz8n
20:41:50
karlosz
no generic functions there, but customizing the methods here doesn't really make sense anyway
20:43:08
karlosz
i don't know if there's a better way of doing the ternary marker thing for the hash table
20:45:34
karlosz
i suppose it only happens rarely that the substructure of an ast doesn't coincide with its descendants
23:35:15
karlosz
asts are a little weird, i guess ideally we wouldn't even have circularities in the first place, since its a tree
0:19:43
Bike
don't remember. i think it was fine? i think the only actual circularity is that return-from refers to the block, but there are other ways to represent that
0:35:40
karlosz
the new algorithm for cloning asts works and saves a good deal of consing from building the table
0:36:01
karlosz
i guessif we get rid of the ast circularity we won't have to build hash tables at all
2:37:40
karlosz
well, i found your branch which changed it, so i'm going to try to see if that works
2:41:00
karlosz
https://github.com/robert-strandh/SICL/commit/4061af0cea991c7602cc1d059140a8dc5e0b971b
4:01:18
drmeister
I'm switching back to 8-byte alignment from 16. I think I switched to 16 because I thought I needed another tag for code objects but I don't.
4:02:21
drmeister
It has GC_memalign(sz) but I don't see anything like GC_memalign_xxx for precise GC
4:29:55
karlosz
this should be instance instead of i? https://github.com/clasp-developers/clasp/blob/d9b83576506a9156a3fa2a83d4967f60938cdc9f/src/lisp/kernel/clos/closfastgf.lsp#L196
4:32:17
Bike
i rejiggered that stuff since the reinitialize-instance and change-class things also need to udpate instances sometimes