| 24 Sep 2022 |
|  peddie joined the room. | 10:45:23 |
 SomeoneSerge (matrix works sometimes) | Isn't showing up where? | 13:28:08 |
 aidalgol | Blender doesn't see it, and python CUDA libraries fail to evaluate CUDA devices. | 20:02:47 |
| aidalgol | I've tried setting hardware.nvidia.nvidiaPersistenced = true; in my NixOS config, and that seems to have resolved it so far. | 20:04:06 |
| 25 Sep 2022 |
 hexa (UTC+1) | fyi https://github.com/NixOS/nixpkgs/pull/192879 | 11:26:48 |
| 29 Sep 2022 |
| SomeoneSerge (matrix works sometimes) | FRidh: Hey, I just don't want to flood the nvidia-ml-py issue, so I post here: the thing is I don't know what you mean by "referring from the stub to the driver library"
I know that if you just run LD_PRELOAD=$(nix-build '<nixpkgs>' -A cudaPackages.cudatoolkit --no-out-link)/lib/stubs/libnvidia-ml.so nvidia-smi (or prepend the stub path in RUNPATH) without doing anything else, you're going to get a failure. If you meant something else (e.g. if you've heard of other ways to use the stubs than to bypass linker errors), then it might that I just lack knowledge
| 10:59:49 |
 FRidh | Yes, and what if you do that after patchelf'ing ${cudatoolkit}/lib/stubs/libnvidia-ml.so to point to /run/current-system/... ? | 11:06:12 |
| FRidh | I have never tried this myself, just thought this should work. | 11:06:31 |
| SomeoneSerge (matrix works sometimes) | Ok, I don't know about anything like it. The way I see it: the downstream app (e.g. nvidia-ml-py) searches for libnvidia-ml.so, resolves that into lib/stubs/libnvidia-ml.so, dlopen's it, and calls a "function" defined there. This now would have to look for a new libnvidia-ml.so? | 11:17:09 |
| SomeoneSerge (matrix works sometimes) | 
Download image.png | 11:28:14 |
| FRidh | Right it's only a convenience for compilation, it's not like you somehow can make it a symlink. | 11:29:42 |
| FRidh | I misunderstood what is possible then. | 11:30:07 |
| 30 Sep 2022 |
| aidalgol | In reply to@aidalgol:matrix.org
I've tried setting hardware.nvidia.nvidiaPersistenced = true; in my NixOS config, and that seems to have resolved it so far. Nope, still happening sporadically. | 06:03:22 |
|  @0b11:matrix.org removed their display name 0x53. | 17:33:24 |
| @0b11:matrix.org left the room. | 17:33:30 |
| 3 Oct 2022 |
| SomeoneSerge (matrix works sometimes) | Btw if anyone uses blender: opensubdiv (and thus blender) has been broken in nixos-22.05 for almost a month now, but got fixed in master and unstable recently. There's hopefully some commit that that hasn't reached the release channel but maybe could be backported... | 10:15:01 |
|  spacesbot - keeps a log of public NixOS channels joined the room. | 14:25:38 |
| 4 Oct 2022 |
 Kevin Mittman (UTC-7) | CUDA 11.8.0 is live. Notably linux-aarch64 tarballs for Jetson are available | 17:58:28 |
| 5 Oct 2022 |
| SomeoneSerge (matrix works sometimes) | Nix usability study mentioning CUDA infra: https://discourse.nixos.org/t/usability-study-session-6/21438 | 16:06:28 |
| 6 Oct 2022 |
| SomeoneSerge (matrix works sometimes) | In reply to @justbrowsing:matrix.org
CUDA 11.8.0 is live. Notably linux-aarch64 tarballs for Jetson are available Missing jetson pieces so far:
- cudnn (needed e.g. for pytorch)
- cuda_nvprof (needed e.g. for
faiss)
| 20:19:44 |
| 7 Oct 2022 |
| Kevin Mittman (UTC-7) | In reply to @ss:someonex.net
Missing jetson pieces so far:
- cudnn (needed e.g. for pytorch)
- cuda_nvprof (needed e.g. for
faiss)
- cuDNN, TensorRT, NCCL, etc. are separate product releases from CUDA SDK.
- Still working on getting sign-off to ship Jetson tarballs in a future release of those products
| 20:42:25 |
| Kevin Mittman (UTC-7) |
- Investigating
cuda_nvprof but seems support on linux-aarch64 platforms was dropped sometime back
- Trying to find any public release notes that mention this
| 21:01:00 |
| SomeoneSerge (matrix works sometimes) |
Investigating cuda_nvprof
Actually, I found today that cuda_profiler.h isn't in cuda_nvprof anymore anyway
| 21:02:10 |
| Kevin Mittman (UTC-7) | Yes, cuda_profiler_api was split from cuda_nvprof in 11.8 | 21:03:46 |
| SomeoneSerge (matrix works sometimes) | Oh, I see. Thanks! | 21:04:05 |
| 8 Oct 2022 |
|  bitmapper joined the room. | 07:51:46 |
| 9 Oct 2022 |
|  mt-caret joined the room. | 01:57:57 |
| 14 Oct 2022 |
 tpw_rules | anybody ever seen glxinfo straight up hang for liek 45 seconds at 100% cpu on nvidia drivers? | 21:16:32 |
| 15 Oct 2022 |
|  underpantsgnome! changed their display name from underpantsgnome to underpantsgnome!. | 00:33:33 |
| aidalgol | The worst I have ever experienced is one of the demo applications, like glxgears running at a horribly low framerate when the drivers are not properly installed. | 19:19:02 |
