Package Details: cuda-11.7 11.7.1-3

Git Clone URL: https://aur.archlinux.org/cuda-11.7.git (read-only, click to copy)
Package Base: cuda-11.7
Description: NVIDIA's GPU programming toolkit
Upstream URL: https://developer.nvidia.com/cuda-zone
Licenses: custom:NVIDIA
Provides: cuda-sdk, cuda-toolkit
Replaces: cuda-sdk, cuda-toolkit
Submitter: yodaembedding
Maintainer: yodaembedding
Last Packager: yodaembedding
Votes: 2
Popularity: 0.008123
First Submitted: 2023-06-19 00:50 (UTC)
Last Updated: 2023-11-16 12:39 (UTC)

Dependencies (6)

Required by (1)

Sources (29)

Latest Comments

1 2 Next › Last »

ryukinix commented on 2024-07-30 04:46 (UTC)

How much tweaking I would need to modify this repository installation to have the cuda 11.4 sync to the 470xx nvidia drivers? I didn't find any repository with cuda-11.4

howetuft commented on 2024-07-11 07:23 (UTC) (edited on 2024-07-11 07:24 (UTC) by howetuft)

Hello, Could you also add 'cuda' to provides please? Then it would be recognized as an eligible dependence for some other nvidia packages, like optix... Thanks!

sidicer commented on 2024-05-06 12:59 (UTC) (edited on 2024-05-06 15:34 (UTC) by sidicer)

The package just keeps building endlessly. Just for fun I've left the makepkg -si running in a server and after 2 full days it's still building something (?)

EDIT: It's the gcc11 dependency that's actually endlessly building and not cuda

zyk commented on 2024-02-12 03:44 (UTC)

prepare() { sh cuda_${pkgver}_${_driverver}_linux.run --target "${srcdir}" --noexec

# Fix up samples tht use findgllib_mk # for f in builds/cuda_samples///findgllib.mk; do # patch $f cuda-findgllib_mk.diff # done } cuda_11.7.1_515.65.01_linux.run has no option --target

Alad commented on 2024-01-15 15:53 (UTC) (edited on 2024-01-15 15:54 (UTC) by Alad)

I would also suggest to add the following to provides:

provides=('cuda-toolkit' 'cuda-sdk' 'libcudart.so' 'libcublas.so' 'libcusolver.so' 'libcusparse.so')

The repo package does the same and then namcap will correctly verify dependencies for dependent packages.

In addition to the repo packages, you could also include a versioned provides cuda-toolkit=$pkgver and cuda-sdk=$pkgver (or alternatively cuda=$pkgver).

This would allow AUR packages to directly depend on say, cuda-toolkit<12 without hardcoding a specific older version.

ys-clarry commented on 2023-11-19 10:46 (UTC)

With new package, all the hash validations were passed in my environment. thanks!

yodaembedding commented on 2023-11-16 12:26 (UTC)

Ah, so what must be happening is that *.pc is expanded as:

  nppc.pc
  nppi.pc
  nppial.pc
  ...
  nppist.pc

rather than bash/zsh's behavior:

  nppc.pc
  nppial.pc
  ...
  nppi.pc
  nppist.pc

I have written them all out explicitly to work around this incompatibility.

ys-clarry commented on 2023-11-16 12:05 (UTC)

Hello, I've also run into the hash-mismatch problem and figured out that the .pc file order in the PKGCONFIG file was not good. I've sorted all sha512sums in the PKGCONFIG by .pc names and the hash test were passed.

Alad commented on 2023-10-10 02:13 (UTC)

By the way, it seems that with gcc 13.1, not even basic CUDA programs can be compiled:

https://forums.developer.nvidia.com/t/strange-errors-after-system-gcc-upgraded-to-13-1-1/252441

My suggestion is to use gcc-11 instead in the PKGBUILD

yodaembedding commented on 2023-10-04 01:02 (UTC) (edited on 2023-10-04 01:02 (UTC) by yodaembedding)

Those files are all provided inside the PKGBUILD git repo.

You can confirm the hashes yourself via:

git clone ssh://aur@aur.archlinux.org/cuda-11.7.git
cat PKGBUILD | sed -n -e '/accinj64/, /nvToolsExt/ p' > expected.txt
sha512sum *.pc | sed -e 's:^\([0-9a-f]\+\)\s\+\(.*\):  "\1"  # \2:' > actual.txt
diff expected.txt actual.txt

...So I'm not sure what could be causing the issue.

1 2 Next › Last »