Niko Matsakis wrote:
Here are some relatively simple things we could do to improve performance in type-checking, off the top of my head:
I should add that the overall impact of these (and other) changes might easily be small. I know you have profiles showing malloc to be a relatively minor contribution to overall performance, for example. My feeling is that it's hard to estimate the impact in advance---past experience suggests that sometimes these sorts of changes have an oversized impact relative to the profile and sometimes none at all. Long term I think we should try to tighten up performance and, if we do enough of that, things *will* get faster.
Have we tried to profile memory consumption at all? I'd be curious to know e.g. what portion of our memory is used in the AST representation vs other things. It should be easy enough to use dtrace and get an idea how much is allocated in each pass.
Niko _______________________________________________ Rust-dev mailing list [email protected] https://mail.mozilla.org/listinfo/rust-dev
