| 10 May 2026 |
 Robbie Buxton | In reply to @hexa:lossy.network
I'd say right now the problem is that renting aarch64-linux for cheap is not really possible I don’t think it’s too bad in hyper scalers (graviton isn’t bad) not sure about other places tho | 20:54:26 |
 hexa (UTC+1) | yeah, but cheap :D | 20:57:14 |
| Robbie Buxton | Last time I checked graviton was similar if not cheaper | 20:58:00 |
 neobrain | I'm guessing self-hosting has been considered but isn't an option? nvidia has a pretty generous marketing budget for free hardware | 21:01:09 |
 Gaétan Lepage | aarch64-linux support is definitely on our roadmap. For now, the critical blocker is getting access to relevant hardware. I'll give updates in this channel as soon as we get new opportunities in this regard. | 21:01:46 |
|  NixOS Moderation Botchanged room power levels. | 21:07:38 |
| 11 May 2026 |
| Gaétan Lepage | Looking for an approval on https://github.com/NixOS/nixpkgs/pull/517764 (cudaPackages.cudnn update).
connor (burnt/out) (UTC-8) maybe? | 15:52:02 |
 smudge (she/her) | Redacted or Malformed Event | 19:34:06 |
| smudge (she/her) | should the nixos wiki cuda page's cache section be corrected? it recommends the cache.nixos-cuda.org cache and doesn't mention others which lead me to assume it was the one for public use. is nix-community.cachix.org (per this announcement) the one that should be primarily recommended on the wiki? as well as information about using flox's nixpkgs repo and its cache? Or, if the nix-community.cachix.org isn't actually allowed to distribute should flox be the primary suggested cache?
also should the information for cache.nixos-cuda.org be retained, just with a proper disclaimer that it is for internal use only, or how would you like that to be handled? | 19:40:10 |
| smudge (she/her) | should the nixos wiki cuda page's cache section be corrected? it recommends the cache.nixos-cuda.org cache and doesn't mention others which lead me to assume it was the one for public use. is nix-community.cachix.org (per this announcement) the one that should be primarily recommended on the wiki? as well as information about using flox's nixpkgs repo and its cache? Or, if the nix-community.cachix.org isn't actually allowed to distribute should flox be the primary suggested way to access a cuda cache?
also should the information for cache.nixos-cuda.org be retained, just with a proper disclaimer that it is for internal use only, or how would you like that to be handled? | 19:44:22 |
| smudge (she/her) | should the nixos wiki cuda page's cache section be corrected? it recommends the cache.nixos-cuda.org cache and doesn't mention others which lead me to assume it was the one for public use. is nix-community.cachix.org (per this announcement) the one that should be primarily recommended on the wiki? as well as information about using flox's nixpkgs repo and its cache? Or, if the nix-community.cachix.org isn't technically allowed to distribute should flox be the primary suggested way to access a cuda cache?
also should the information for cache.nixos-cuda.org be retained, just with a proper disclaimer that it is for internal use only, or how would you like that to be handled? | 20:10:58 |
| smudge (she/her) | should the nixos wiki cuda page's cache section be corrected? it recommends the cache.nixos-cuda.org cache and doesn't mention others which lead me to assume it was the one for public use. is nix-community.cachix.org (per this announcement) the one that should be primarily recommended on the wiki? as well as information about using flox's nixpkgs repo and its cache? Or, if the nix-community.cachix.org isn't technically allowed to distribute should flox be the primary suggested way to access a cuda cache?
also should the information for cache.nixos-cuda.org be retained, just with a disclaimer that it is for internal use only? | 20:14:14 |
| 13 May 2026 |
| smudge (she/her) changed their display name from smudge to smudge (she/her). | 08:48:06 |
| 14 May 2026 |
|  weriomat joined the room. | 21:11:04 |
| 15 May 2026 |
|  Grayson Tinker changed their display name from graysontinker to Grayson Tinker. | 03:47:21 |
 ccicnce113424 | I'm currently working on improving the packaging of the NVIDIA driver. My plan, is to first split the namespace and driver extraction into two separate files without changing semantics with the assistance of LLM, and then switch from the current passthru pattern to using scope. The first step is now complete. Semantics should be unchanged. I'm not sure how to verify semantic equivalence (suggestions welcome!), but this PR works fine with my current NixOS configuration. As a side note, I also moved nvidia-modprobe into the nvidiaPackages namespace. Since these changes are mostly structural and don't alter semantics, could they be merged separately, with the scope migration and further work happening in a new PR? | 09:07:14 |
| ccicnce113424 | https://github.com/NixOS/nixpkgs/pull/519313 | 09:07:25 |
| 16 May 2026 |
|  asp345 joined the room. | 13:24:37 |
| 18 May 2026 |
 yorik.sar | Gaétan Lepage: Here's a different fix for the libcusolvermp issue that we had: let's add all of cudaPackages that change by PR to the PR check. https://github.com/nixos-cuda/hydra-jobsets/pull/29 - wdyt? | 14:26:40 |
| 19 May 2026 |
|  hilorioze changed their display name from Yan Hilorioze to hilorioze. | 16:41:32 |
|  Prayag Bhakar joined the room. | 20:15:25 |
| Prayag Bhakar | hey folks, I'm looking for feedback on this PR :: https://github.com/NixOS/nixpkgs/pull/515928
My goal with these changes is to speed up my NVIDIA Jetson Orin Nano builds by adding dGPU architectures targets for Jetson Xavier (sm_72), Jetson Orin (sm_87), Jetson Thor (sm_110). I'm happy to iterate more on the changes if there is feedback :) | 20:22:59 |
| Prayag Bhakar | if/when this gets merged, I was planning on raising a follow-up fix to the nixos-cuda inference package to support these new params. https://github.com/nixos-cuda/infra | 20:25:01 |
| Gaétan Lepage | Hi! Just so you know, aarch64-linux is not supported in our current CI. | 21:24:45 |
| Prayag Bhakar | hi! I did see that there wasn't a callout for an aarch64-linux system in the infra package. I was going to include more thoughts in the follow-up CR, but the few options I see are (a) enable cross compilation via qemu on current hardware (ie :: https://github.com/81reap/jetpack-nixos/blob/d20caea8befd5fdad99efc9fad5527e70124ca98/README.md?plain=1#L339-L351); (b) onboard my own Jetson Nano as a target to build things for the infra pipeline; or (c) I donate $250 for the NVIDIA Jetson Orin Nano Super Developer Kit to be purchased by one of the current maintainers | 22:51:06 |
| Prayag Bhakar | that said, I think the lowest bar to test this on the current CI is to enable cross complication. But that's just my current opinion. Happy to chat more and bar raise these thoughts :) | 22:52:04 |
| 20 May 2026 |
 connor (burnt/out) (UTC-8) | Our infra is fairly limited in terms of capacity and doing builds through QEMU would be prohibitively expensive. Not to say this can't happen, but it's certainly not a platform we build for or test. | 02:05:15 |
| connor (burnt/out) (UTC-8) | Gaétan Lepage: for gpu-burn: https://github.com/NixOS/nixpkgs/pull/522144
I still need to look into CUDA compat/fix the assumption in the logic that cudaVariant only ever has a major version suffix | 05:12:41 |
| connor (burnt/out) (UTC-8) | And then also investigate the NVCC fatal error encountered when using family feature sets with baseline ones | 05:15:40 |
| connor (burnt/out) (UTC-8) | For the CUDA variant stuff, look for changes here:
- https://github.com/NixOS/nixpkgs/blob/e72912dc31edae89f04a12982c684f9652c95348/pkgs/development/cuda-modules/_cuda/lib/cuda.nix#L75-L103
- https://github.com/NixOS/nixpkgs/blob/e72912dc31edae89f04a12982c684f9652c95348/pkgs/development/cuda-modules/buildRedist/default.nix#L70-L97
Ugh so getSupportedReleases needs to support something like desiredCudaVariants being an ordered list
| 05:20:42 |
