gpt4all-cuda: use gcc12 for cuda (fixes build)#434165
gpt4all-cuda: use gcc12 for cuda (fixes build)#434165Titaniumtown wants to merge 3 commits intoNixOS:masterfrom
Conversation
|
@ofborg build gpt4all-cuda |
3f3363f to
df4973b
Compare
| (lib.cmakeBool "LLMODEL_KOMPUTE" false) | ||
| ] | ||
| ++ lib.optionals cudaSupport [ | ||
| "-DCUDA_HOST_COMPILER=${gcc12Stdenv.cc}/bin/c++" |
There was a problem hiding this comment.
have you tried using cudaPackages.backendStdenv.cc ?
There was a problem hiding this comment.
woah. Ok cool I will do that instead.
There was a problem hiding this comment.
Nevermind, it explodes in the same way. Reverting!
| lib, | ||
| config, | ||
| stdenv, | ||
| gcc12Stdenv, |
There was a problem hiding this comment.
how about in all-packages.nix ?
override { stdenv = gcc12Stdenv };
also this should be resolved soon in nvcc
leave a link to the issue in the override
There was a problem hiding this comment.
GCC 12 is EOL and about to be removed. Current versions of CUDA support up to GCC 14. I believe you want to use stdenv' = if cudaSupport then cudaPackages.backendStdenv else stdenv; here.
Edit: Okay, I see that that was tried and there is another issue here. But we shouldn’t be adding more references to GCC 12. Perhaps using another CUDA version will fix this?
|
builds with |
|
@kirillrdy I will test this. I need to investigate why it's so out of date then, thank you! |
Closes #409674
Cuda was not happy that gcc14 was being used as the cuda compiler. For the cuda bits we should be using gcc12.
Things done
passthru.tests.nixpkgs-reviewon this PR. See nixpkgs-review usage../result/bin/.Add a 👍 reaction to pull requests you find important.