Community Packages

Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines

Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.

REPEAT: Do NOT report bugs for outdated packages!
Tasklist

FS#67094 - [magma] Incompatible with CUDA 11

Attached to Project: Community Packages
Opened by Konstantin Gizdov (kgizdov) - Wednesday, 24 June 2020, 09:29 GMT
Last edited by Sven-Hendrik Haase (Svenstaro) - Monday, 13 July 2020, 00:54 GMT
Task Type Bug Report
Category Upstream Bugs
Status Closed
Assigned To Sven-Hendrik Haase (Svenstaro)
Konstantin Gizdov (kgizdov)
Architecture All
Severity Low
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 1
Private No

Details

Description:

CUDA 11 deprecated `cusparseSolveAnalysisInfo` API call which seems to be central to the sparse computation implementation Magma uses. Until this is resolved, Magma will be incompatible with CUDA 11. This issues is to track progress on this.

Additional info:
* package version(s): any
* config and/or log files etc.: irrelevant
* link to upstream bug report, if any: https://bitbucket.org/icl/magma/issues/22/cuda-11-changes-issue

Steps to reproduce:
1. Build against CUDA 11
2. Observe undefined function call error
This task depends upon

Closed by  Sven-Hendrik Haase (Svenstaro)
Monday, 13 July 2020, 00:54 GMT
Reason for closing:  Implemented
Comment by Gao Xiang (zasdfgbnm) - Friday, 10 July 2020, 20:47 GMT
Is it possible to temporarily disable magma-sparse in this package? From https://www.archlinux.org/packages/community/x86_64/magma/, magma is only used by PyTorch, and PyTorch does not use magma-sparse. Doing so can help to unblock upgrading the version and PyTorch to CUDA 11.
Comment by Sven-Hendrik Haase (Svenstaro) - Monday, 13 July 2020, 00:54 GMT
Good idea, thanks for the insight. I pushed a new package that doesn't build magma-sparse and will try to build pytorch against it.

Loading...