FS#65837 - [cuda] gcc8 dependency can be dropped

Attached to Project: Community Packages
Opened by Jed Brown (jedbrown) - Saturday, 14 March 2020, 19:38 GMT
Last edited by Sven-Hendrik Haase (Svenstaro) - Saturday, 14 March 2020, 22:16 GMT
Task Type Feature Request
Category Packages
Status Closed
Assigned To Sven-Hendrik Haase (Svenstaro)
Felix Yan (felixonmars)
Konstantin Gizdov (kgizdov)
Architecture x86_64
Severity Low
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:

There is no longer a version check blocking recent gcc in /opt/cuda/include/crt/host_config.h, which now reads:

#if defined(__GNUC__)

#if __GNUC__ > 8


#endif /* __GNUC__ > 8 */

Note that there is still a block for clang >= 9, but Arch isn't configured to use that by default.

Additional info:
* package version(s)
cuda-10.2.89-3

Steps to reproduce:

pacman -Rdd gcc8 gcc8-libs

CUDA continues to work.
This task depends upon

Closed by  Sven-Hendrik Haase (Svenstaro)
Saturday, 14 March 2020, 22:16 GMT
Reason for closing:  Won't fix
Additional comments about closing:  See comment
Comment by Sven-Hendrik Haase (Svenstaro) - Saturday, 14 March 2020, 22:15 GMT
True, GCC 9 will work for some simple cases but it'll fail for more complex projects like tensorflow. Of course I'd love to be proven wrong on this but it doesn't seem likely since I tested this for the current cuda release as I do for every release. We patch out the define guard (check the cuda PKGBUILD) so that some daring users such as yourself can use gcc 9 if they so desire but it's neither supported nor working in complex cases.

Reopen this if you manage to compile the tensorflow and pytorch packages using gcc 9. I'd be very happy to drop gcc8, believe me.

Loading...