Repository navigation
Adding _int.to_str triples time to compile standard library #124
Description
Before jyasskin/rust@684502b:
$ time ./rustboot -L . -shared -o libstd.dylib lib/std.rc
real 0m1.205s; 0m1.168s; 0m1.159s
After:
$ time ./rustboot -L . -shared -o libstd.dylib lib/std.rc
real 0m3.375s; 0m3.571s; 0m3.421s
This is with CFG_NATIVE=1. There must be something pathological in the compiler. A 40-line function shouldn't be able to triple the build time.
Activity
graydon commentedon Jul 25, 2010
Yeah. I think I added a bit too much un-guarded logging in the past week. I'll do a profile-and-shave run on monday and should get it back down a ways.
graydon commentedon Jul 26, 2010
Good news: found the source of pain. Bad news: will take a little bit of time to fix. On it this afternoon.
graydon commentedon Jul 26, 2010
I think this is fixed enough to be tolerable for now. There is still a bit of a sore spot in the RA but it's nothing spectacularly non-local or non-linear like you were seeing. Let me know if you are feeling it slowing down again.
graydon commentedon Jul 27, 2010
Er, for reference sake the commit that did most of the fixing was 3d69407.
move all code accessing vtable internals into the `trait` module
fixed _mm_set_pd and _mm_setr_pd by reversing order (rust-lang#124)
Merge pull request rust-lang#124 from brauliobz/unit_is_not_literal
Add how to run tidy script in isolation to text
Merge pull request rust-lang#124 from JohnTitor/rustup