On Fri 25 Nov 2022 08:53:57 PM CST, Lawrence Murray wrote:
Following up on the repository-specific approaches, Arch works with the cuda package in its Community repository, whereas Ubuntu does not (nvidia-cuda-toolkit appears in its Multiverse repository, which contains restricted software and seems unavailable on openSUSE Build Service). I don't believe any of the other distributions have CUDA packages available outside of Nvidia's repositories.
Hi Perhaps leave it to the end user to rebuild locally? You could add a build condition wrapper, eg %bcond_with nvcc Then users can pass this through when running rpmbuild locally? See what is done with libheif, then when built by packman the project config defines x256. https://build.opensuse.org/package/view_file/multimedia:libs/libheif/libheif... -- Cheers Malcolm °¿° (Linux Counter #276890) Tumbleweed 20221123 | GNOME Shell 43.1 | 6.0.8-1-default HP Z440 | Xeon E5-2690 V3 X24 @ 2.60GHz | Nvidia Quadro T400/Nvidia Tesla P4 up 3 days 19:12, 2 users, load average: 0.04, 0.16, 0.25