| 13 Aug 2025 |
Luke | There is now! #ROCm:nixos.org | 17:37:21 |
| 15 Aug 2025 |
connor (burnt/out) (UTC-8) | Two bugs to be aware of:
- https://github.com/NixOS/nixpkgs/issues/434096
- https://github.com/NixOS/nixpkgs/issues/434101
| 23:31:17 |
| 16 Aug 2025 |
| Martin Charles joined the room. | 04:58:51 |
Martin Charles | curious, why is nvcomp missing from nixpkgs cuda? | 05:15:27 |
connor (burnt/out) (UTC-8) | lack of time, mostly -- there is a redist for it (https://developer.download.nvidia.com/compute/nvcomp/redist/redistrib_5.0.0.6.json) | 05:18:50 |
Martin Charles | hmm ok, i'm happy to package it are you guys open to patches? | 05:25:04 |
Martin Charles | is there some way to consume this? | 05:26:22 |
Martin Charles | it appears this is what the release.nix files are for | 05:26:33 |
Martin Charles | this nvcomp one seems different in that it has multiple cuda versions for each library version which the other packages don't seem to have? | 05:32:51 |
connor (burnt/out) (UTC-8) | You can try doing something similar to the steps listed here: https://github.com/NixOS/nixpkgs/blob/master/doc/languages-frameworks/cuda.section.md#updating-cutensor-cuda-updating-cutensor
I've not had a chance to update that script in a while, so not sure how well/if it still works
Generally we use the script from the docs to take the manfiest provided by NVIDIA (dubbed the redist or redistrib manifest) and create a feature manifest which tells us things relevant to the Nix packaging
Then create a fixup function (https://github.com/NixOS/nixpkgs/tree/master/pkgs/development/cuda-modules/_cuda/fixups) for the packages from the manifest
Lastly add it to the package sets by creating and importing an extension, like https://github.com/NixOS/nixpkgs/blob/3b0feaf8c2160a10a974c890e93f5f5387728784/pkgs/top-level/cuda-packages.nix#L201 | 05:48:12 |
connor (burnt/out) (UTC-8) | It's fairly involved and I've been trying to automate away the entirety of it for the better part of the last year, but there are many sharp edges in many domains :l | 05:49:30 |
| 17 Aug 2025 |
| mastratisi joined the room. | 10:05:31 |
| 21 Aug 2025 |
luke-skywalker | I ran into this issue with the latest nvidia driver (580.76.05) crashing gtk apps and GPU accelerated apps like zed editor on latest linux kernel (6.16.1) on wayland.
Im kind of overwhelmed by all the available packages. Can someone give me a quick hint on which of all the available packages I can safely revert back to?
| 13:23:10 |
luke-skywalker | * I ran into this issue with the latest nvidia driver (580.76.05) crashing gtk apps and GPU accelerated apps like zed editor with latest linux kernel (6.16.1) on wayland.
Im kind of overwhelmed by all the available packages. Can someone give me a quick hint on which of all the available packages I can safely revert back to?
| 13:23:27 |
luke-skywalker | * I ran into this issue with the latest nvidia driver (580.76.05) crashing gtk apps and GPU accelerated apps like zed editor with latest linux kernel (6.16.1) on wayland.
I am installing the nvidia drivers from the nvidia-container-toolkit.
Im kind of overwhelmed by all the available packages. Can someone give me a quick hint on which of all the available packages I can safely revert back to?
| 13:25:20 |
luke-skywalker | * I ran into this issue with the latest nvidia driver (580.76.05) crashing gtk apps and GPU accelerated apps like zed editor with latest linux kernel (6.16.1) on wayland.
I am installing the nvidia drivers from the nvidia-container-toolkit.
Can someone give me a quick hint on which of all the available packages I can safely revert back to? I'm kind of overwhelmed by all the available packages.
| 13:40:31 |
connor (burnt/out) (UTC-8) | What do you mean the drivers from nvidia-container-toolkit? Your NixOS config should have something like this, right? https://github.com/ConnorBaker/nixos-configs/blob/5b9fe63698ee14a52a8187cd34ba769203fae24e/modules/nvidia.nix#L18
You should be able to select the version of the driver you want to use by changing that. | 14:45:00 |
| infinisil joined the room. | 23:36:59 |
infinisil | Does the CUDA team want to add a slide for NixCon's state of the union? The board and SC are reaching out to various teams to prepare a shared slide deck here, and I reserved a slot for the CUDA team :) https://docs.google.com/presentation/d/191-Bgwv6Bdb4KBzIfyAyZBftI0rSuCBo-_tSPrTSrtE/edit?slide=id.g376eb5e3048_2_152#slide=id.g376eb5e3048_2_152 | 23:38:14 |
infinisil | @connor (he/him) (UTC-7) @SomeoneSerge (Ever OOMed by Element) @Samuel Ainsworth ^ | 23:38:31 |
infinisil | You can PM me your google email to get permissions to edit | 23:38:39 |
| 22 Aug 2025 |
connor (burnt/out) (UTC-8) | Sure! I’ll let Serge know as well since I have him on Signal | 00:52:56 |
stick | connor (he/him) (UTC-7): jfyi i tried to update cusparselt to 0.8.0 using your redist scripts, but it failed - it seems the scripts need to be adjusted somehow | 09:42:02 |
| 23 Aug 2025 |
stick | Found the issue
New releases contain files to cuda12 and cuda13, introducing another level in json
See
https://developer.download.nvidia.com/compute/cusparselt/redist/redistrib_0.7.1.json
vs
https://developer.download.nvidia.com/compute/cusparselt/redist/redistrib_0.8.0.json | 12:18:03 |
stick | I manually edited the json to adhere to the old scheme and will create a cusparselt update PR although it is not ideal and i suspect the upcoming cuda releases will use the newer scheme too | 12:23:03 |
stick | cusparselt PR in https://github.com/NixOS/nixpkgs/pull/436186 | 12:42:56 |
| Lun joined the room. | 21:07:38 |
Lun |
TODO: 2. We should probably abandon attributes such as torchWithCuda (etc.)
as they routinely end up consuming the wrong arguments\
(dependencies without cuda support).
Instead we should rely on overlays and nixpkgsFun.
(@SomeoneSerge)
_tritonEffective ?
Is there an issue tracking dropping WithCuda aliases/a discussion thread about this somewhere?
| 21:43:20 |
Lun | *
# TODO: 2. We should probably abandon attributes such as `torchWithCuda` (etc.)
# as they routinely end up consuming the wrong arguments\\
# (dependencies without cuda support).
# Instead we should rely on overlays and nixpkgsFun.
# (@SomeoneSerge)
\_tritonEffective ?
Is there an issue tracking dropping WithCuda aliases/a discussion thread about this somewhere?
| 21:43:40 |
| 24 Aug 2025 |
connor (burnt/out) (UTC-8) | Not that I know of; probably hasn’t been made due to lack of time. Feel free to make one, I can add the CUDA tag to it so it doesn’t get lost. (Well, it’ll be lost in our giant backlog, but whatever.) | 18:31:36 |