!eWOErHSaiddIbsUNsJ:nixos.org

NixOS CUDA

287 Members
CUDA packages maintenance and support in nixpkgs | https://github.com/orgs/NixOS/projects/27/ | https://nixos.org/manual/nixpkgs/unstable/#cuda58 Servers

Load older messages


SenderMessageTime
26 Sep 2025
@lt1379:matrix.orgLun #ROCm:nixos.org 19:17:23
28 Sep 2025
@gjvnq:matrix.orgMir changed their display name from Gabriela Noriega to Mir.18:56:06
29 Sep 2025
@winter:catgirl.cloudWinter so i've learned 🫠 15:18:02
30 Sep 2025
@zhaofeng:zhaofeng.liZhaofeng Li joined the room.00:30:38
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)If anyone has stuff they use to verify or test releases, give it a try with https://github.com/NixOS/nixpkgs/pull/437723, it’s nearing ready for review05:01:50
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8) Also can someone explain to me how the regexes for nixpkgs-review are supposed to work? I don't like rebuilding rocmPackages.aotriton and -P '*.aotriton.*' didn't match it for some reason. 05:25:18
@winter:catgirl.cloudWinterare you sure it’s based on the attrpath?05:26:58
@winter:catgirl.cloudWintertbh it would be weird for it not to be, nvm05:27:10
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8) * Also can someone explain to me how the regexes for nixpkgs-review are supposed to work? I don't like rebuilding rocmPackages.aotriton and -P '.*aotriton.*' didn't match it for some reason.
EDIT: typo in regex name, I meant '.*aotriton.*' not `'.aotriton.'
05:29:19
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8) * Also can someone explain to me how the regexes for nixpkgs-review are supposed to work? I don't like rebuilding rocmPackages.aotriton and -P '.*aotriton.*' didn't match it for some reason.
EDIT: typo in regex name, I meant '.*aotriton.*' not '*.aotriton.*'
05:29:49
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)Unrelated: https://github.com/NixOS/nixpkgs/pull/444123 was merged so there should be a nice closure size reduction coming for CUDA builds of OpenCV05:34:30
@winter:catgirl.cloudWinter(sorry for not having a more useful answer, if nobody gives one by the morning i’ll crack out the source and look lol)05:38:38
@ss:someonex.netSomeoneSerge (back on matrix) Gaétan Lepage: wdyt throwing hydra at it? 08:12:14
@ss:someonex.netSomeoneSerge (back on matrix) * Gaétan Lepage: wdyt throwing hydra at it? I can edit the input, if that's ok 08:36:55
@glepage:matrix.orgGaétan LepageSure, let's go for it.11:24:17
@ss:someonex.netSomeoneSerge (back on matrix) connor (he/him) (UTC-7): https://hydra.nixos-cuda.org/eval/44 12:24:07
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)

I'm a dummy, the option I wanted was one more down and doesn't have a short flag:

  -P, --skip-package SKIP_PACKAGE
                        Packages to not build (can be passed multiple times)
  --skip-package-regex SKIP_PACKAGE_REGEX
                        Regular expression that package attributes have not to match (can be passed multiple times)
12:43:21
@winter:catgirl.cloudWinter
2025-09-30 15:11:37.818161: W external/local_xla/xla/service/gpu/llvm_gpu_backend/default/nvptx_libdevice_path.cc:40] Can't find libdevice directory ${CUDA_DIR}/nvvm/libdevice. This may result in compilation or runtime failures, if the program we try to run uses routines from libdevice.
Searched for CUDA in the following directories:
  ./cuda_sdk_lib
  ...

  /usr/local/cuda
  /opt/cuda
  /nix/store/d2b95k4ysi7822hnxq72np5vvfq7wbbp-python3.12-tensorflow-gpu-2.19.0/lib/python3.12/site-packages/tensorflow/../nvidia/cuda_nvcc
  /nix/store/d2b95k4ysi7822hnxq72np5vvfq7wbbp-python3.12-tensorflow-gpu-2.19.0/lib/python3.12/site-packages/tensorflow/../../nvidia/cuda_nvcc
  /nix/store/d2b95k4ysi7822hnxq72np5vvfq7wbbp-python3.12-tensorflow-gpu-2.19.0/lib/python3.12/site-packages/tensorflow/cuda
  /nix/store/d2b95k4ysi7822hnxq72np5vvfq7wbbp-python3.12-tensorflow-gpu-2.19.0/lib/python3.12/site-packages/tensorflow/../../../..
  /nix/store/d2b95k4ysi7822hnxq72np5vvfq7wbbp-python3.12-tensorflow-gpu-2.19.0/lib/python3.12/site-packages/tensorflow/../../../../..
anyone ever see smth like this before?
19:14:29
@winter:catgirl.cloudWintergrepped around existing issues/prs a bit but no dice19:14:57
1 Oct 2025
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)nvvm is a subdirectory of cuda_nvcc pre-CUDA 13.0; I don’t remember which output it’s in though. Seems like the error is mostly about being unable to find that.01:03:41
@ss:someonex.netSomeoneSerge (back on matrix)I think we put some config next to bin/nvcc that points at the correct libdevice location? Used to be in the overrides15:46:38
@justbrowsing:matrix.orgKevin Mittman (EOY sleep)In CUDA 13, there was a split from cuda_nvcc to cuda_crt, libnvvm, and libnvptxcompiler components15:47:11
@justbrowsing:matrix.orgKevin Mittman (EOY sleep) changed their display name from Kevin Mittman (UTC+9) to Kevin Mittman (UTC-7).15:48:07
@magic_rb:matrix.redalder.org@magic_rb:matrix.redalder.org left the room.18:23:33
@winter:catgirl.cloudWinter this is just using pythonPackages.tensorflow w/ config.cudaSupport on 25.05 -- so that's CUDA 12, right? 19:12:26
@winter:catgirl.cloudWinterdunno why this would be sad then19:12:38
@ss:someonex.netSomeoneSerge (back on matrix)At what stage do you get that error?19:20:29
@winter:catgirl.cloudWinter runtime, after i've successfully imported tf 20:09:22
@winter:catgirl.cloudWinter* runtime, after i've successfully imported tf -- i then get some JIT compilation error20:15:31
@winter:catgirl.cloudWinter seems like XLA_FLAGS may be my friend here 20:16:43

Show newer messages


Back to Room ListRoom Version: 9