Stream: t-compiler/wg-nll

Topic: coercions benchmark


simulacrum (Jun 11 2018 at 13:51, on Zulip):

@nikomatsakis I've re-enabled coercions, and it appears to be even worse than inflate currently: https://perf.rust-lang.org/nll-dashboard.html is showing 2,200% (0.8 seconds to 16 seconds)

nikomatsakis (Jun 11 2018 at 13:52, on Zulip):

ok :)

nikomatsakis (Jun 11 2018 at 13:52, on Zulip):

thanks

nikomatsakis (Jun 11 2018 at 13:52, on Zulip):

that is indeed ungreat!

nikomatsakis (Jun 11 2018 at 13:54, on Zulip):

huh, weird. I had git bisect running on https://github.com/rust-lang/rust/pull/51460 — and it claims that the first bad commit is the one that removes the #[inline(never)] annotations...

nikomatsakis (Jun 11 2018 at 13:54, on Zulip):

/me finds that hard to believe

nikomatsakis (Jun 11 2018 at 13:56, on Zulip):

must have done something wrong with the script :)

simulacrum (Jun 11 2018 at 13:58, on Zulip):

Might have inverted good/bad, so maybe check the commit before it manually?

nikomatsakis (Jun 11 2018 at 14:02, on Zulip):

I think it was beacuse my script failed to run cargo clean -p futures in between steps =)

nikomatsakis (Jun 11 2018 at 14:03, on Zulip):

I'm on a plane, so debugging this stuff (over ssh) is painfully slow

nikomatsakis (Jun 11 2018 at 14:03, on Zulip):

but I'll try to start another bisect run...

nikomatsakis (Jun 11 2018 at 14:05, on Zulip):

on the plus side, the results oughta be ready by the time I land ;)

Last update: Nov 21 2019 at 13:10UTC