Heh, this is fun. By slowing down utf-8 validation in rustc we're seeing up to 44% performance decrease in short programs. This isn't exact science, but if the algorithm were to be *improved* it could have positive improvements for shorter programs. https://perf.rust-lang.org/compare.html?start=ae66171b41b1d05128afd0d666475bc4ba617469&end=c93d8c28b4e86aee203bca6ba8f4d7ecb2045c0d …
-
Show this thread
-
I mean; it could very well be that we hit another bottleneck if we go the other way (~0% perf improvement if already blocked on, say, disk IO). But if that's *not* the case suddenly it becomes rather interesting. At the least worth validating I think.
1 reply 0 retweets 1 likeShow this thread -
Replying to @yoshuawuyts
Hmm, that's quite surprising to me actually. Is the slowdown so severe or is this an indicator that we're validating it more than necessary?
2 replies 0 retweets 0 likes -
Replying to @killercup @yoshuawuyts
In any case, I've been meaning to look at https://github.com/killercup/simd-utf8-check … as well as recent advances again at some point :)
1 reply 0 retweets 1 like -
Replying to @killercup
Oh cool, I found out about the exact same post you're referencing last week. Filed an issue about it too. https://github.com/rust-lang/rust/issues/68455 … Some folks have been messing around with it in the PR linked; that's what the benches above are from!
1 reply 0 retweets 1 like
If you're looking to speed things up; apparently the state of algorithms has progress in the last 2 years; the last comment on the thread includes a python version of an allegedly faster algo:https://github.com/rust-lang/rust/issues/68455#issuecomment-577412509 …
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.