Hacker News Reader

Comment by ementally | Hacker News Reader

CachyOS (Arch based distro), no.1 on https://distrowatch.com/

That’s just a ranking of subpage hits per day. Not only is that easily gamed, it also says very little about how popular an OS really is.

This is a slight aside, but CachyOS is a great example of the failure of Wikipedia politics.

The "CachyOS" page was deleted[1], and replaced with a redirect to the Arch Linux page. But CachyOS is not mentioned anywhere on that page, nor on the "List of Linux distributions § Arch Linux-based" page.

[1]: https://en.wikipedia.org/wiki/Wikipedia:Articles_for_deletio...

It links to https://en.wikipedia.org/wiki/Arch_Linux#Derivatives which indeed lacks any mention of CachyOS. Luckily, Wikipedia is free to register, and you can just edit pages you feel like could be better. Seems like you found the perfect first edit to make for yourself :)

I have a long Wikipedia history, but that is not the point. There already was a CachyOS page, and it was removed. Why bother contributing stuff that will just be deleted again?

It might have been removed due to the editor's impression that CachyOS is not significantly different from Arch. With proof to the contrary the page may be restored.

There are a lot of derivative(I don't mean it in a negative way) distors out there, not sure if they all need pages.

Most moderated spaces remove content that doesn't fit the community, Wikipedia does take that to the extreme but I still prefer that than the opposite extreme.

It's an endemic issue on wikipedia, and even editing wouldn't fix this one instance, since someone can (and demonstrably already has) remove whatever you add later on. The issue is wikipedias preference for "deletionism", removing perfectly correct information for no particularly good reason. It's especially pernicious when it comes to short articles, which tend to get deleted with impunity, and redirected to sections of articles, which later get renamed, destroying the link, or removed altogether. Nothing can be done by any individual to fix this issue, since it comes from a wikipedia wide policy, which unfortunately is not one of the things that "anyone can edit".

I agree with most of what you wrote, but unless you can demonstrate that someone actually added CachyOS to the "Arch Linux-based" on the "List of Linux distributions" page and it was later removed, I'm not sure how much it matters how Wikipedia generally works.

> Blazingly Fast & Customizable Linux distribution

I love Arch Linux, but please...

(Arch Linux is already "fast" (depends on what you install for your DE, if any) and customizable.)

But their differentiation is that to improve performance they compile all the packages with newer instruction sets as the target as well as enabling more optimizations like LTO. And some are even optimized with PGO.

I find it odd to call a specific Linux distribution blazingly fast.

Gentoo with make.conf (/etc/portage/make.conf[1]) having "CFLAGS="-O3 -march=native -flto"" means that Gentoo, a Linux distribution, is performant?

[1] It is not a good idea to build everything with LTO or PGO enabled because not all packages support LTO / PGO cleanly. Do it on the basis of per-package.

I've seen claims of decent speed improvements when using CachyOS, though I can't say I've ever hunted down solid confirmation. I'm a bit wary of the project because I would have to put a lot of trust in them since they're rebuilding everything themselves and could easily introduce malware somewhere in there. (But I've been scared of distros before only to have it pointed out to me that some very well respected people are involved, so I could be worrying for nothing here too)

Does Gentoo have binpkgs with these compileflags? With CachyOS you don't have to compile, because it's a rolling binary distro. Regarding your [1]: They do that, systemd(or parts of it) are unoptimized, for instance. They don't apply that stuff blindly, only where it works.

For me it feels blazingly fast, even on obsolete KabyLake Core-I5/7(t) forcibly clocked down to about 800Mhz most of the times :)

It fucking flies without much effort. On modern systems even more so. While being rock solid. Without any crashes. Even under Plasma. When I'm reading about bugfixes regarding crashes under Plasma I just shrug and think "Waddya talkin about?". That may be hardware dependent, though, because they are old Lenovo Thinkcentres(1Litre SFF M910q tiny) with excellent firmware.

Using btrfs, profile-sync-demon, zram(Yes. Even with 32GB Ram!). Suspend/Resume working every single time. No glitches, hick-ups, ever. So far. Since 10th of June, 2024.

Edit: Almost always some music out of yt doodling in some bg-tab, in oh-so-slow FF, without any clicks, stuttering, or other breaks.

No need for yt-dlp, mpv at all. Except for dl/saving stuff, sometimes. While FF is rarely under 100 tabs.

Actually, there are binpkgs, now: https://wiki.gentoo.org/wiki/Gentoo_Binary_Host_Quickstart.

My i3 with vim / emacs and even VSCodium flies too, on X Linux. :P

The browser is always the slowest in my case and this has always been my experience, and unfortunately it still is.

Good for you. The 'oh-so-slow' regarding FF was meant as satirical btw. With what I'm using it for, it doesn't lag. Maybe because of https://github.com/graysky2/profile-sync-daemon mostly, and some other fine-tunings, which make it not touching the filesystem over and over again. I really recommend PSD. Not even insane amounts of RAM, it usually takes about 4 to 5GB, rarely going to 8, then shrinking back a while after closing too much tabs. Imagine that!1!! It's all about some sysctl settings :-)

Sideberry (Tree-Style-Tabs like extension) was the ugliest offender there. Though that may have been me misconfiguring it. OTOH I didn't manage to find settings where it didn't do that, and still looked like I wanted it to. At that time, maybe a year ago, I've thought of it as potential 'instant ssd-killer'. Couldn't be bothered. Deinstalled. Now FF has some basic version of vertical tabs by Mozilla itself. It suffices(for now).

Thanks, PSD looks interesting. I only have 8 GB RAM at the moment, I will see if my motherboard supports more (it should support 16 GB at the very least). After the RAM upgrade, I will probably use PSD. Or is there a way to specify how many RAM it can use?

As for tabs, I like the way Vivaldi does it and allows me to customize.

Hrrm. My comment was supoptimal, I guess.

By "Not even insane amounts of RAM, it usually takes about 4 to 5GB, rarely going to 8, then shrinking back a while after closing too much tabs." I meant to say that this applies to the resident size in RAM of the whole browser, not what PSD does, or adds. That would be just what your browser profile is using 'on disk'. Peanuts, so to speak.

With only 8GB it's really hard to tell. It depends on your usage patterns.

First, regarding just PSD, it relies on RuntimeDirectorySize= of https://man.archlinux.org/man/logind.conf.5 which by default is limited to use up to 10% of system memory, but not statically reserved, only "on demand".

Which in turn relies on tmpfs which can use up to half of system RAM by default. Again, on demand only, not statically reserved.

https://wiki.archlinux.org/title/Tmpfs https://wiki.gentoo.org/wiki/Tmpfs

However, I think these are the wrong knobs to turn :-)

PSD can make use of overlayfs, which saves a little bit of used RAM, and is faster to initially sync, but uses more disk space then. But not that much. Which can be further minimized by the number of kept profile backups, or not using backups at all.

Just keep your profile lean and mean, then there's less stuff to shuffle around. Since your'e using Vivaldi, pointing you to wikipages how to make Firefox use less RAM seems pointless ;)

Maybe (carefully) use something like https://www.bleachbit.org for cleaning up Vivalidis profile.

Which leaves things like

https://en.wikipedia.org/wiki/Zram https://wiki.archlinux.org/title/Zram https://wiki.gentoo.org/wiki/Zram https://www.kernel.org/doc/html/latest/admin-guide/blockdev/...

https://en.wikipedia.org/wiki/Zswap https://wiki.archlinux.org/title/Zswap https://wiki.gentoo.org/wiki/Zswap https://docs.kernel.org/admin-guide/mm/zswap.html

to consider.

While it may seem insane to reserve already limited RAM for just another thing, these are worth it. If configured right. I used them, or their predecessors since olden times, when I've just had a Thinkpad T60p with some Centrino and only 4GB.

That made things better in general. Of course it's no silver bullet for everything, but it made the system less sluggish, and it took longer to slow down because of being 'swapped to death'.

From then on I continued to use stuff like that.

On a system with only 8GB, too. ZRAM in this case, because backing device like zswap was impossible, because the HDD blew. So I booted live from USB(2(Arrgh!)) and ran from RAM.

By means of AntiX, which btw. showed my the ways sysctls regarding swappiness, pressure stall information, and related stuff can totally change the behaviour of a system.

Even if it looks strange/ghettoish at first, which can be remastered away easily anyways, the devs really know how to get the most out of older systems with limited RAM and power, in interesting ways. Should be looked at, even if only for 'inspiration', technically.

For instance making things like Firefox shrink back, after having closed too much tabs. And remaining usable, while doing so.

Anyway. Depending on what you do, 8GB only can go a looong way, if configured/used right.

IMO not using ZRAM/ZSWAP, sysctls for swappiness, PSI, etc. is wrong and wasteful.

PSD is just a little icing on the cake.

I am using zram. What would the ideal sysctls be for it for 8 GB (maybe even 16 GB?) RAM?

Depends on your usage patterns. On 8GB I gave it (ZRAM) between 512MB to 768MB RAM, and later on just using whatever values for the relevant sysctls AntiX was using(or their scripts created dynamically during setup/startup) at the time. It isn't just vm.swappiness, there are others, and they changed, or got removed, and replaced by others, over the years. And all of them sort of interwoven, influencing each other via what's happening in vm-subsystem. Can have bad feedback loops if you do it wrong, and be counterproductive. So you have to experiment, and measure with different settings, for what you're running.

Where measure can be anything from htop/atop over https://github.com/cdown/psi-notify , to https://github.com/noiseonwires/memory-pressure and so on.

To understand what these sysctls are about, how they are related to each other, and to ZRAM, I'd recommend reading, or at least skimming https://chrisdown.name/2018/01/02/in-defence-of-swap.html , https://chrisdown.name/2019/07/18/linux-memory-management-at... , https://linuxblog.io/linux-performance-almost-always-add-swa... , https://linuxblog.io/linux-performance-almost-always-add-swa... , https://linuxblog.io/linux-performance-no-swap-space , https://linuxblog.io/running-out-of-ram-linux-add-zram , https://github.com/ValveSoftware/SteamOS/issues/899 , https://lonesysadmin.net/2013/12/22/better-linux-disk-cachin... , https://github.com/CachyOS/CachyOS-Settings/pull/19 , https://github.com/CachyOS/CachyOS-Settings , https://docs.kernel.org/admin-guide/sysctl/vm.html , https://docs.kernel.org/accounting/psi.html , https://facebookmicrosites.github.io/psi/docs/overview , https://unixism.net/2019/08/linux-pressure-stall-information...

even if it seems redundant, OFC skim and discard anything you already know(But I can't know that).

The following not to use them, but for their POVs/concepts regarding the same problem: https://github.com/facebookincubator/oomd , https://github.com/rfjakob/earlyoom , https://github.com/hakavlad/nohang

These, to maybe take inspiration from, to adjust dynamically:

https://www.linuxbash.sh/post/tune-vmswappiness-in-a-script-... , https://github.com/lululoid/LMKD-PSI-Activator (Yes, Android, I know, but still...)

Then there is the whole thing about eBPF, which opens up much more screws to turn dynamically according to demand, according to whichever policies.

https://www.brendangregg.com/ebpf.html / https://deepwiki.com/oracle/bpftune/1-overview

Have fun :)

Reply on news.ycombinator.com