Post #4742137 · Elektrine

2026-09-17 23:24 UTC

I compiled my own kernel for a few years when I first started using Linux in the late 90s. “So much faster by removing things I don’t need,” I claimed. I cringe at that now. I’ve been using stock kernels for years now. Any speed improvements from a custom kernel are more than offset by the time it takes to configure and compile it. Making one mistake could wipe out years of theoretical speed improvements as you recover, reconfigure, and recompile. Plus, if you have a problem with a piece of hardware down the road, how do you know the issue isn’t your custom kernel? I’d say, do it once or twice for the experience, or if there’s some specific reason you need it (bleeding edge hardware, for example). Some years back, I realized I’d rather be using my computers instead of administering them.

Replies (2)

@juipeltje@lemmy.world 2026-09-18 08:12

Hmm, yeah in my case the goal would be to actually speed up compile time of the kernel itself rather than speed improvements at runtime, but it does seem like a pain in the ass and your description makes it sound even worse lol. Nonguix substitution servers are working fine now though, and with my server being set up soon enough, if i the substitution servers are ever down again i can just let it compile everything once and then my other devices can just request it from my own server, so it doesn’t feel worth it to mess around with it anymore. But if i do ever feel like learning more about the kernel i might try it just for the hell of it.

Open ##4746041
@vandsjov@feddit.dk 2026-09-18 09:41

“So much faster by removing things I don’t need,” I claimed I never really got to use FreeBSD but I thought that I would, late ‘90s. Found out how to modify and build the kernel. And like you, optimised it by removing stuff i didn’t need, like parallel port drivers. And then let the system compile for hours. So green back then and so used to Windows.

Open ##4746733