I am considering installing Clear Linux on my laptop, but Julia seems not included in its bundle while R and Python are included. I was wondering how I can install Julia on Clear Linux.
https://clearlinux.org/documentation/clear-linux/reference/bundles/available-bundles
Build from source?
Install a desktop environment and browser, and try running the binaries from Julia’s download page?
They have git and all the build tools.
I’m curious if you’ll get any performance improvements. Clear Linux is consistently impressive in benchmarks. Although a bit party if that is probably shipping optimized libraries. Check out the R benchmark here:
Over five times faster than the competition! IIRC, this was mostly testing a few machine learning libraries.
Not sure if you’d see any difference running software you compiled yourself.
But I’m curious! I’d consider switching if it’s 5% faster somehow, and I can’t easily replicate whatever those changes are via rebuilding the kernel.
Lots of people here know way more than I do on this, though.
Anyone have insight?
As far as Julia is concerned, changing distributions shouldn’t make a big performance difference. What matters is to use a fast BLAS and to have Julia generate code targeted to your CPU rather than to a generic x86_64. The former is done automatically by OpenBLAS, which dynamically chooses the appropriate kernels for your CPU (though you can also use another BLAS like MKL). The latter can be achieved by rebuilding the sysimage or by compiling Julia from source.
In that case, Clear Linux markets themselves for server and cloud applications – not desktop or laptop.
Feel free to try it, but there won’t be as much support, so you may find yourself having to build packages often.
Like, Atom or VSCode – their download pages have links for Debian and Fedora based distributions, and Arch’s AUR covers them. What about Clear Linux?
Many things that take little effort on mainstream distros (eg, their package managers cover almost everything) that you would want to do on a laptop/desktop may sudenly take a lot more effort.
You can get it to work, and have fun playing with it. But, in terms of having all the tools there to make you as productive as possible, all the usual choices developed with desktops/laptops in mind will serve you better.
On that note, if you want to optimize your laptop experience, the lowest hanging fruit may be on preserving your battery life.
Which will, again, likely be harder on clear Linux, since this is not the intended application.
Let’s say I download JuliaPro to my desktop.
What should I do to make it generate code targeting my specific CPU?
I’d be happy to know what should I do for Windows, macOS and Linux.
Just rebuild the sysimage. The code snippet I have in #7 here might be old now? But it’s just a quick one-liner.