freenode/#sicl - IRC Chatlog
Search
17:19:21
Bike
beach: i had some more thoughts on unwinding in sicl and cleavir. check it out when you have time, please. https://pastebin.com/raw/b47hbwwP
17:22:23
beach
I need to go cook for my (admittedly small) family in a few minutes, so I'll read it when I am less tired tomorrow morning if that's OK.
19:26:57
drmeister
I've read the proposal once and it sounds good - I've always found funwind-protect and call-with-variable-bound kind of incongruent with the rest of cleavir.
19:58:38
Bike
i think it might not actually be necessary to invalidate parts of the dynamic environment, hm
19:59:22
Bike
and it might be slightly easier if instead of a location for the dynamic environment it was just that calls and blocks/etc tracked their parent, because then the sicl backend could actually do the dedicated register thing easily
23:15:01
jcowan
I meant to ask, but got dragged into a meeting: can't tagbody/go, block, return-from, and catch/throw all be the same at run time and the first two are alphatized at compile time?
3:56:30
beach
jcowan: You could treat them all in a similar way, but then RETURN-FROM and GO would be more expensive than they currently are. Since they are static constructs, the compiler takes advantage of that fact to generate more efficient code.