freenode/#clasp - IRC Chatlog
Search
3:03:14
beach
Common Lisp signals are a model of simplicity compared to the description of C++ exception handling in that article.
3:04:38
drmeister
Yeah - unfortunately everything rests on me figuring out the C++ exception handling. It's amazing that I got it almost right but not quite.
3:05:07
beach
Bike: Yes, you are right, some optimizations are possible because we always have all the code at our disposal, so they resemble static optimizations. When the code changes, we just adapt to the new version of the entire code.
3:05:44
beach
Bike: I am pretty sure there are many other such optimization opportunities to explore.
13:58:16
Bike
drmeister: i'm trying to elide the vaslist construction if it's unneeded. I thought this would be pretty easy because there's already a flag for whether the effective methods need a valist i can use. However, codegen-slot-reader/writer call vaslist-end for some reason. What does it do?
14:00:05
Bike
looks like it does va_end. but i don't see how that would be necessary only for slot readers and writers.
14:00:34
Bike
if it's necessary for anything that doesn't use the vaslist, fast method calls should be doing it too.
14:03:33
Bike
but even for readers and writers it only does it for the instance-slot case. what the heck.
14:04:14
Bike
okay i'm just going to guess it's needed for everything but calls that use the vaslist.