Notes on coreutils in Rust (alexgaynor.net)
from ikidd@lemmy.world to linux@lemmy.ml on 24 Mar 14:21
https://lemmy.world/post/27334055

With Ubuntu changing to the Rust implementation of coreutils, what does that mean for performance?

#linux

threaded - newest

catloaf@lemm.ee on 24 Mar 14:33 next collapse

I don’t think anyone is relying on performance of coreutils. If you’re concerned about performance, you’re already writing a program directly, not using a shell script.

ProtonBadger@lemmy.ca on 24 Mar 16:47 collapse

Relying on is perhaps too strong, but I enjoy operations like sort getting faster and I don’t know how they’ve written cp but there’s a cp alternative using async IO with io_uring that’s almost twice as fast, I’m sure it’d interest people if such optimizations made it into coreutils.

mina86@lemmy.wtf on 24 Mar 14:51 next collapse

Why do you think it would affect performance?

satans_methpipe@lemmy.world on 24 Mar 15:34 next collapse

I think your question is relevant as there are unfortunately plenty of shell scripts out there doing critical batch work. But it won’t change the momentum of the Rust push happening right now.

beejjorgensen@lemmy.sdf.org on 24 Mar 15:39 next collapse

Rust has some big binaries due to static linkage, and the Rust coreutils gets around this Busybox-style, compiling everything into one binary that you hard link to. Pretty neat. The project is easy to build and mess with without installing if you’re curious about it. And you could add the build dir to the front of your path if you want to try it out with low risk.

Neptr@lemmy.blahaj.zone on 24 Mar 15:44 next collapse

It is faster, optimization is one of the uutils project’s stated goals.

Xanza@lemm.ee on 24 Mar 16:43 next collapse

The performance you’re dealing with here is in the tens of milliseconds possibly hundreds if you’re lucky. Anyone seriously pursuing this issue from the angle of performance genuinely doesn’t understand the deep rooted issues here.

If you’re so incredibly hard up for compute time that it’s critical for you to squeeze out the extra 1/10 of a second from your system utilities then you need to shut your fucking computer down and go touch grass.

I mean even if this saves you 30 seconds a day 50 weeks a year 5 days a week that’s 2 hours per year it’s saving you… I’d rather slow fuck the two hours and get an extra 2 hours of pay.

cerement@slrpnk.net on 24 Mar 20:25 next collapse

this was me watching some of the cheering when neofetch got archived, people complaining “good, neofetch is too slow” – WTF were you doing with neofetch where speed was a factor?!

QuazarOmega@lemy.lol on 24 Mar 20:42 next collapse

.bashrc greeter? ;-;

Not saying that neofetch going away was a good thing tho

cerement@slrpnk.net on 24 Mar 23:46 collapse

when the last message was “Have taken up farming.”, kinda hard to hold anything against them …

that_leaflet@lemmy.world on 25 Mar 02:35 collapse

On some systems neofetch would actually run quite slow. Even on my fast system it would occasionally take a second because it hung on one step.

noli@lemm.ee on 25 Mar 03:10 collapse

I’d take those tens of milliseconds. That shit scales and I’ve seen infra in the scale of millions more-or-less glued together by shell scripts and coreutils/busybox.

ReakDuck@lemmy.ml on 25 Mar 02:33 collapse

Im more concerned about the stupid license