Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

cudatoolkit_9: fix vector intrinsic compiler errors #107131

Closed
wants to merge 1 commit into from
Closed

cudatoolkit_9: fix vector intrinsic compiler errors #107131

wants to merge 1 commit into from

Conversation

r-burns
Copy link
Contributor

@r-burns r-burns commented Dec 18, 2020

Fixes #103415

Also adds simple regression test

Motivation for this change
Things done
  • Tested using sandboxing (nix.useSandbox on NixOS, or option sandbox in nix.conf on non-NixOS linux)
  • Built on platform(s)
    • NixOS
    • macOS
    • other Linux distributions
  • Tested via one or more NixOS test(s) if existing and applicable for the change (look inside nixos/tests)
  • Tested compilation of all pkgs that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review wip"
  • Tested execution of all binary files (usually in ./result/bin/)
  • Determined the impact on package closure size (by running nix path-info -S before and after)
  • Ensured that relevant documentation is up to date
  • Fits CONTRIBUTING.md.

@r-burns
Copy link
Contributor Author

r-burns commented Dec 18, 2020 •
edited

Dear reviewers,

The issue this fixes manifests during nvcc invocation so it can be reproduced without needing any nvidia drivers or hardware.

Links to gcc commits reverted:
gcc-mirror/gcc@ec434f4
gcc-mirror/gcc@1992881

The first is just a mass renaming (no functional changes) which needs to be reverted before the older patch applies. The second is the actual function signature fix, which when reverted fixes the incompatiblility with the pre-9.2 cudatoolkit.

@SuperSandro2000
Copy link
Member

ping @danieldk

@danieldk
Copy link
Contributor

Looks fine in principle, but I haven't tested this (my GPU is too new for CUDA 9). Though I wonder how much sense it still makes to ship CUDA 9.

@r-burns
Copy link
Contributor Author

r-burns commented Apr 12, 2021

You can actually repro without a GPU at all as it's a build-time issue. But yeah, maybe we should just drop cuda <= 9 and avoid this entirely.

@danieldk
Copy link
Contributor

You can actually repro without a GPU at all as it's a build-time issue. But yeah, maybe we should just drop cuda <= 9 and avoid this entirely.

Sure, but I think in such cases it's also good to test that the built artifact actually works. (I wouldn't expect differently, since these are renamings.)

danieldk added a commit to danieldk/nixpkgs that referenced this pull request Apr 18, 2021
@danieldk
cudatoolkit-{9,9_0,9_1,9_2}: remove
f75286e 
Remove old CUDA toolkits (and corresponding CuDNN versions).

- Not supported by upstream anymore.
- We do not use them in nixpkgs.
- We do not test or actively maintain them.
- Anything but ancient GPUs is supported by newer toolkits.

Fixes NixOS#107131.
@danieldk danieldk mentioned this pull request Apr 18, 2021
Merged
10 tasks
@r-burns r-burns deleted the cuda branch August 28, 2021 06:13
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@danieldk danieldk Awaiting requested review from danieldk

Assignees
No one assigned
Labels
10.rebuild-darwin: 11-100 10.rebuild-linux: 11-100
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

CUDA compiler fails when including CPU vector intrinsic headers
3 participants
@r-burns @SuperSandro2000 @danieldk