I split packaged this into aocl-aocc and aocl-gcc.
Difference is that for example libblis-mt.so from aocl-gcc is linked against libgomp.so (GNU OpenMP) and from aocl-aocc it's linked against libomp.so (LLVM OpenMP).
Search Criteria
Package Base Details: aocl
Package Actions
| Git Clone URL: | https://aur.archlinux.org/aocl.git (read-only, click to copy) |
|---|---|
| Submitter: | disc-kuraudo |
| Maintainer: | disc-kuraudo (jclds139) |
| Last Packager: | disc-kuraudo |
| Votes: | 2 |
| Popularity: | 0.000000 |
| First Submitted: | 2019-10-04 14:02 (UTC) |
| Last Updated: | 2024-03-24 10:36 (UTC) |
disc-kuraudo commented on 2020-02-20 19:03 (UTC)
hfath commented on 2020-02-14 11:56 (UTC) (edited on 2020-02-14 12:09 (UTC) by hfath)
Thanks for the quick update!
Upstream responded with "Due to web portal issues the latest checksums are not updated." -- but what they really should have done is bump the version number...
disc-kuraudo commented on 2020-02-14 10:40 (UTC)
Updated with the new sha256sum.
If you have an old "aocl-linux-gcc-2.1.tar.gz" archive laying around, please re-download!
disc-kuraudo commented on 2020-02-14 10:08 (UTC) (edited on 2020-02-14 10:15 (UTC) by disc-kuraudo)
Weird, they uploaded a different archive under the same name.
The original one with the "e473..." sha256sum has timestamps from 15th of January.
A new downloaded one with the checksum you mention has timestamps from 23th of January... well.
The amd-libm seems to be fixed and split into amdlibm.h and amdlibm_vec.h as well... will update the PKGBUILD.
hfath commented on 2020-02-14 10:01 (UTC) (edited on 2020-02-14 10:02 (UTC) by hfath)
FTR, the distfile as downloaded with Firefox (repeatedly) gives me
% sha256sum aocl-linux-gcc-2.1.tar.gz
8ea9ec16051893e480f7ba09babeaa833897586aa4d0c2c53b22132b5884150c aocl-linux-gcc-2.1.tar.gz
%
-- differing from the checksum on upstream's web page and in PKGBUILD.
I have notified toolchainsupport@amd.com.
hfath commented on 2020-01-08 09:46 (UTC)
Actually, env-modules should put a (t)csh init file in /etc/profile.d, see <https://en.wikipedia.org/wiki/Environment_Modules_(software)#tcsh,_csh_automatic_modules_initialization>
hfath commented on 2020-01-08 09:14 (UTC)
tcsh here, and since I use my own /etc/csh.* config I had to source /etc/modules/init/tcsh (which of course has ${MODULESHOME} in it - catch 22).
I guess it would be helpful to default to /etc/modules, unless ${MODULESHOME} is set.
disc-kuraudo commented on 2020-01-07 14:08 (UTC)
Hmm, it seems that your shell didn't have ${MODULESHOME} set.
Maybe you need to restart the shell after the installation of env-modules and see if the environment variable is set? I think it should be integrated into bash startup scripts, other shells might need manual env-modules sourcing in the ~/.<shell>rc file.
I should check for if this is set in the PKGBUILD, else it kinda malfunctions and indeed will try to install in /modules.
hfath commented on 2020-01-07 14:01 (UTC)
Ah, okay, I see it now. I guess AUR expects you to look out for such things. ;)
The package installs its environment module files under /modules -- is that the canonical location on Arch? The 'env-modules' package is configured for /etc/modules.
And which package would install the 'module' command mentioned in aocl.install - or did you mean what 'env-modules' installs as 'add.modules'?
disc-kuraudo commented on 2020-01-07 12:23 (UTC) (edited on 2020-01-07 12:48 (UTC) by disc-kuraudo)
Manual download of '${pkgname}-${os}-${pkgver}.tar.gz' required from upstream
Had to do it this way because the real download of the tarball is somewhat hidden behind some web stuff that I don't understand.
Pinned Comments
disc-kuraudo commented on 2022-07-19 17:48 (UTC) (edited on 2022-07-19 17:48 (UTC) by disc-kuraudo)