dev-util/nvidia-cuda-toolkit: Bump to 10.1.105

[jvesely]

Other changes:
Bump to EAPI=7
Don't use versionator
Update SRC_URI file name
Don't unpack the source file twice
Use dobin + doins instead of mv

Signed-off-by: Jan Vesely jano.vesely@gmail.com

[gentoo-bot]

Copyright policy change

Please note that on 2018-09-15 Trustees have approved new Gentoo copyright policy. All contributions made to Gentoo need to follow this policy. If you include the Signed-off-by line in your commit message, you indicate that you have read the policy and agree to its terms. For more detailed explanation, please see the new Gentoo copyright policy explained article.

Pull Request assignment

Submitter: @jvesely
Areas affected: ebuilds
Packages affected: dev-util/nvidia-cuda-toolkit

dev-util/nvidia-cuda-toolkit: @gentoo/sci

Linked bugs

No bugs to link found. If your pull request references any of the Gentoo bug reports, please add appropriate GLEP 66 tags to the commit message and request reassignment.

If you do not receive any reply to this pull request, please open or link a bug to attract the attention of maintainers.

---

In order to force reassignment and/or bug reference scan, please append [please reassign] to the pull request title.

Docs: Code of Conduct ● Copyright policy (expl.) ● Devmanual ● GitHub PRs ● Proxy-maint guide

[jvesely]

I've tested the build by rebuilding pycuda and running simple kernel.
Useflags mostly remove content. +doc adds documentation and Oleg confirmed that it works in https://bugs.gentoo.org/679412

[gentoo-repo-qa-bot]

Pull request CI report

Report generated at: 2019-03-24 15:54 UTC
Newest commit scanned: 4acf3a5
Status: ✅ good

No issues found

[hsntgm]

@jvesely did you look to https://bugs.gentoo.org/681694
They changed the folder structure for the install, cublas was moved to a separate folder.
Is this PR fixing it?

Also cudnn depends to cuda toolkit.I think for being sure, this package must be test by calling cudnn directly too.

Let me know if you want to work together for both packages.I need them urgently.

[jvesely]

@jvesely did you look to https://bugs.gentoo.org/681694
They changed the folder structure for the install, cublas was moved to a separate folder.
Is this PR fixing it?

the ebuild changes dir to builds/cuda-toolkit and install stuff from there.
libcublas and libcublas are installed correctly (no broken linkage), but I haven't tested them.

Also cudnn depends to cuda toolkit.I think for being sure, this package must be test by calling cudnn directly too.

Let me know if you want to work together for both packages.I need them urgently.

I don't, sorry. I have no interest beyond getting the basic toolkit and pycuda working on a system with gcc 8.2.
That said, afaict, the toolkit installation should be compatible, so there shouldn't be problems for dependent packages.

[jvesely]

@perfinion, @alexxy you handled the previous bump. Any thoughts on this one?

[perfinion]

This mostly looks good to me, it installs on my system. I bumped cudnn too. I need to test TensorFlow with it first then I'll merge it tomorrow.

[jvesely]

This mostly looks good to me, it installs on my system. I bumped cudnn too. I need to test TensorFlow with it first then I'll merge it tomorrow.

Hi @perfinion , has there been any problem with TensorFlow?

[perfinion]

@jvesely yeah, cuda-10.1 is kind of a mess haha. TensorFlow doesnt work at all with it yet. I just set the deps on TF to force only 10.0 for now. When upstream ports it to cuda 10.1 then i'll backport the patch.

I pushed everything now. Thanks for the patch :).