!eWOErHSaiddIbsUNsJ:nixos.org

NixOS CUDA

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

You have reached the beginning of time (for this room).


SenderMessageTime
15 Dec 2025
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8) That was evaluating all of the closures attribute in nixos/release.nix with nix eval --json 20:40:10
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)There is some, but it blocks forever, doesn’t do cleanup, and lacks a few other features needed for different types of builtin fetchers and substitutions20:41:22
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8) Robbie Buxton have you run into this (or seen anyone run into it)? You’re one of the like three people I know who use PyTorch from Nixpkgs other than myself 20:42:49
@sporeray:matrix.orgRobbie Buxton
In reply to @connorbaker:matrix.org
Robbie Buxton have you run into this (or seen anyone run into it)? You’re one of the like three people I know who use PyTorch from Nixpkgs other than myself
I haven’t run into that but I have run into libcuda failing on trying to dlopen things it claims to not depend on
21:07:52
@sporeray:matrix.orgRobbie BuxtonRegarding that issue iirc RPATHS are recursive but RUNPATHS aren’t and if a RUNPATH exists the RPATH won’t be used (its replacing RPATHs)21:12:39
@sporeray:matrix.orgRobbie Buxton I think @[Ari Lotter] is right and we should add it as an (undocumented) dep or cudnn 21:15:47
@sporeray:matrix.orgRobbie Buxton * I think @[Ari Lotter] is right and we should add it as an (undocumented) dep of cudnn 21:15:56
@apyh:matrix.orgapyhadding it as a cudnn dep doesn't work still though from my testing21:16:10
@sporeray:matrix.orgRobbie BuxtonHow are you adding it?21:16:20
@apyh:matrix.orgapyhi tried it as a propagatedBuildInput21:16:30
@apyh:matrix.orgapyhthe only thing i managed was to pass it in ld library path manually :/21:16:43
@sporeray:matrix.orgRobbie Buxton Oh you’ll need to patchelf it 21:16:44
@apyh:matrix.orgapyherr.. but it uses dlopen, right?21:17:13
@sporeray:matrix.orgRobbie BuxtonSure but then it’ll know where it is no?21:17:32
@apyh:matrix.orgapyhmy understanding is that bare soname dlopen doesn't care about that21:18:01
@apyh:matrix.orgapyhlike we'd have to patch the *python* binary21:18:13
@apyh:matrix.orgapyh(very possible. I'm completely wrong here though, this is all new to me)21:18:23
@sporeray:matrix.orgRobbie Buxton A bare no name dlopen will check the runpaths for where it’s being called form 21:18:26
@apyh:matrix.orgapyhright, but that's the top level executable, python, not the libraries in the chain, no?21:18:58
@sporeray:matrix.orgRobbie Buxton Nah each library checks its own runpath 21:19:15
@sporeray:matrix.orgRobbie Buxton As it goes down the call chain 21:19:21

Show newer messages


Back to Room ListRoom Version: 9