freenode/#clasp - IRC Chatlog
Search
23:17:40
karlosz
i thought dynenvs were just markers for things that depended on lexical scope like dynamic variable bindings
23:19:50
karlosz
it just seems weird to me that variables which don't have dynamic extent declarations carry dynenvs
23:30:48
Bike
it's true that the dynamic extent of a leti could be narrowed by looking at liveness. i just wanted to preserve information from the source level if possible
23:31:12
Bike
like with my previous (mostly undeveloped) understanding of dynamic-extent, marking something as dynamic-extent would be only at the whole function scope
23:33:57
karlosz
there should always be an event that causes new variables to become temporary and get deleted
23:34:42
karlosz
like in the wiki page i posted earlier today the events are basically a variable losing SETQs or a readvar of that variable becoming local
23:35:08
karlosz
those are the things which cause lexical variables to become eligible for deletion or better represented
23:36:55
Bike
well, what about the initial pass? i think it would be kind of tricky to check at ast-to-bir time whether there's only one write and one read
23:40:46
karlosz
and of course having lets be converted directly to lexical-bind instructions in ast-to-bir will save a lot of work
2:22:22
drmeister
I'm setting up a loaner machine as a development system so I can get my macbook pro repaired.