!PbtOpdWBSRFbEZRLIf:numtide.com

Nix Community Projects

636 Members
Meta discussions related to https://nix-community.org. (For project specific discussions use github issues or projects own matrix channel). Need help from an admin? Open an issue on https://github.com/nix-community/infra/issues161 Servers

Load older messages


SenderMessageTime
6 Sep 2024
@zowoq:matrix.orgzowoq

the current aarch64 jobset only builds the variant for normal plug-in pcie GPUs not for jetson boards but we could extend it (which is probably the most common aarch64 user in fact...).

If jetson is the more common use case should we switch to building that instead of the current aarch64 jobset?

The jobset also only chooses the "fat" variants which include code for all available gpu architectures at once. We could spawn variants for individual architectures, which is particularly relevant for embedded systems.

If we built the individual architectures would we still need to build the fat variant? Is building an individual architecture quicker than the "fat" variant?

02:29:43
@zowoq:matrix.orgzowoq *

the current aarch64 jobset only builds the variant for normal plug-in pcie GPUs not for jetson boards but we could extend it (which is probably the most common aarch64 user in fact...).

If jetson is the more common use case should we switch to building that instead of the current aarch64 jobset?

The jobset also only chooses the "fat" variants which include code for all available gpu architectures at once. We could spawn variants for individual architectures, which is particularly relevant for embedded systems.

If we built the individual architectures would we still need to build the "fat" variant? Is building an individual architecture quicker than the "fat" variant?

02:29:58
@ss:someonex.netSomeoneSerge (matrix works sometimes)

If jetson is the more common use case should we switch to building that instead of the current aarch64 jobset?

Imo yes

02:33:46
@zimbatm:numtide.comJonas Chevalier

I wonder what is the general idea for nix-community's sustainability?

The project is already sustainable, thanks to all of our donors <3

The main cost centre is hardware, which we scale depending on the amount of donations.

07:44:37
@zimbatm:numtide.comJonas ChevalierOne thing that could get more love is to add some reporting to the shared build boxes that people SSH into. Some MOTD that shows who else is on the box, and how many builds are currently running.07:49:43
@zimbatm:numtide.comJonas ChevalierI imagine that people step over each other sometimes07:50:02
@emilazy:matrix.orgemily yeah I frequently have to do top to see if someone else is using it 07:52:50
@zimbatm:numtide.comJonas Chevalierwould be cool if you could chat with the person if somebody is hogging the box07:56:31
@zimbatm:numtide.comJonas Chevalier you can try echo hey, who is hogging the box? | wall on the machine, but not many people know you can do that 07:57:20
@magic_rb:matrix.redalder.orgmagic_rbUnix has facilities for that07:57:21
@magic_rb:matrix.redalder.orgmagic_rbExactly07:57:28
@zimbatm:numtide.comJonas ChevalierI guess some instructions can be added to the MOTD as well07:58:07
@magic_rb:matrix.redalder.orgmagic_rbOr maybe we can symlink a markdown file into each of the users home dirs, like a manual thingie08:01:30
@emilazy:matrix.orgemilyeh, my experience is that the builders are unused a large fraction of the time08:01:43
@magic_rb:matrix.redalder.orgmagic_rb(Mark the symlink readonly or smth so it cannot be gotten rid off)08:01:46
@emilazy:matrix.orgemilyI don't mind people "hogging" it because I hog it at other times too :)08:01:50
@emilazy:matrix.orgemilythough some kind of smart automatic scheduling could be cool08:02:04
@zimbatm:numtide.comJonas Chevalieryes that would be real game changer. if we had good scheduling support, with priorities and resource accounting, we could also make the CI machines available, and reversely. But that's quite a bit of R&D08:24:59
@emilazy:matrix.orgemilyit's not so bad to have to manually check in practice since often you need to baby-sit builds anyway08:26:29
@emilazy:matrix.orgemily(periodic "thanks for the builders, they make working on big Nixpkgs changes tolerable" :) )08:26:40
@zimbatm:numtide.comJonas Chevalierif we think crazy in the future, potentially, we could build every project on GitHub with a flake.nix08:32:39
@zimbatm:numtide.comJonas Chevalier:p08:32:50
@gsaurel:laas.frnim65s sounds like a job for slurm or similar 08:41:54
@emilazy:matrix.orgemily
In reply to @zimbatm:numtide.com
if we think crazy in the future, potentially, we could build every project on GitHub with a flake.nix
yeah, like Nixpkgs!
08:43:22
@glepage:matrix.orgGaétan Lepage Hi,
I have to admit that I often find myself heavily using the builders. Many of the packages I maintain are a pain to build and take a lot of resources.
08:44:02
@glepage:matrix.orgGaétan Lepage Also, I very rarely explicitly SSH the builders and simply have them configured as remote-builders on my personal system.
Wouldn't messaging techniques like wall be inefective in this aspect ?
08:45:04
@glepage:matrix.orgGaétan LepageMaybe, there could be a matrix room with all the people that have been allowed to use the builder. What do you think ?08:45:58
@emilazy:matrix.orgemilyI guess probably many people use them as remote builders. I don't because I'm paranoid and don't trust them08:47:39
@zimbatm:numtide.comJonas Chevalier yeah good point. this is the matrix room 08:47:42
@zimbatm:numtide.comJonas Chevalierwe should probably ask everybody to join this room08:48:04

Show newer messages


Back to Room ListRoom Version: 6