HIP: Converting CUDA to C++.

[Aniket Pradhan]

Hello everyone!

I was exploring some libraries that can use the AMD GPU for
machine/deep learning. I was able to find some (not-so-promising)
results, but I also found the HIP [1] library for directly converting
the CUDA code to portable C++ code, such that it can run on both
NVidia or AMD GPUs.

I am sure some of you might be interested in packaging this lib for
Fedora. Its backend ROCm-runtime [2] is also available in the
repositories. It might just need a few more deps to be packaged (not
too sure about it), but that's for later.

I wasn't able to find a list of packages the SIG is
maintaining/packaging, and I was wondering if I could push this
library on the list as well. I can also help in the packaging process
if anyone is up for it. Do share your feedback regarding packaging the
lib, whether it should be packaged or not.

[1]: https://github.com/ROCm-Developer-Tools/HIP
[2]: https://apps.fedoraproject.org/packages/rocm-runtime

[Aniket Pradhan]

Hello again!

I found that HIP is already packaged for Fedora, and is available with
the name "hip".

Although I am not able to find the package on apps.fp.o but found the
package on src.fp.o [1].

[1]: https://src.fedoraproject.org/rpms/hip

On Tue, Nov 19, 2019 at 11:26 PM Aniket Pradhan  wrote:
> Hello everyone!
>
> I was exploring some libraries that can use the AMD GPU for
> machine/deep learning. I was able to find some (not-so-promising)
> results, but I also found the HIP [1] library for directly converting
> the CUDA code to portable C++ code, such that it can run on both
> NVidia or AMD GPUs.
>
> I am sure some of you might be interested in packaging this lib for
> Fedora. Its backend ROCm-runtime [2] is also available in the
> repositories. It might just need a few more deps to be packaged (not
> too sure about it), but that's for later.
>
> I wasn't able to find a list of packages the SIG is
> maintaining/packaging, and I was wondering if I could push this
> library on the list as well. I can also help in the packaging process
> if anyone is up for it. Do share your feedback regarding packaging the
> lib, whether it should be packaged or not.
>
> --
> Thanks
> Regards
>
> Aniket Pradhan
> http://home.iiitd.edu.in/~aniket17133/
> Aliases: MeWjOr/major
>
> ()  ascii ribbon campaign - against html e-mail
> /\  www.asciiribbon.org   - against proprietary attachments

[Michael McCune]

On Tue, Nov 19, 2019 at 2:01 PM Aniket Pradhan 
wrote:
> Hello again!
>
> I found that HIP is already packaged for Fedora, and is available with
> the name "hip".
>
> Although I am not able to find the package on apps.fp.o but found the
> package on src.fp.o [1].

very cool, thanks for the update Aniket.

i did like your idea about some sort of package list on the ml-sig wiki or
something. that would be cool, but might be a big load for someone to
maintain it.

peace o/

[1]: https://src.fedoraproject.org/rpms/hip
> On Tue, Nov 19, 2019 at 11:26 PM Aniket Pradhan 
> wrote:
> > Hello everyone!
> >
> > I was exploring some libraries that can use the AMD GPU for
> > machine/deep learning. I was able to find some (not-so-promising)
> > results, but I also found the HIP [1] library for directly converting
> > the CUDA code to portable C++ code, such that it can run on both
> > NVidia or AMD GPUs.
> >
> > I am sure some of you might be interested in packaging this lib for
> > Fedora. Its backend ROCm-runtime [2] is also available in the
> > repositories. It might just need a few more deps to be packaged (not
> > too sure about it), but that's for later.
> >
> > I wasn't able to find a list of packages the SIG is
> > maintaining/packaging, and I was wondering if I could push this
> > library on the list as well. I can also help in the packaging process
> > if anyone is up for it. Do share your feedback regarding packaging the
> > lib, whether it should be packaged or not.
> >
> > --
> > Thanks
> > Regards
> >
> > Aniket Pradhan
> > http://home.iiitd.edu.in/~aniket17133/
> > Aliases: MeWjOr/major
> >
> > ()  ascii ribbon campaign - against html e-mail
> > /\  www.asciiribbon.org   - against proprietary attachments
>
> --
> Thanks
> Regards
>
> Aniket Pradhan
> http://home.iiitd.edu.in/~aniket17133/
> Aliases: MeWjOr/major
>
> ()  ascii ribbon campaign - against html e-mail
> /\  www.asciiribbon.org   - against proprietary attachments
> _______________________________________________
> ml mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
> Fedora Code of Conduct:
> https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives:
> https://lists.fedoraproject.org/archives/list/[e-mail]

[Lumir Balhar]

On 11/19/19 8:26 PM, Michael McCune wrote:
> On Tue, Nov 19, 2019 at 2:01 PM Aniket Pradhan 
> <[e-mail] > wrote:
> Hello again!
>
> I found that HIP is already packaged for Fedora, and is available with
> the name "hip".
>
> Although I am not able to find the package on apps.fp.o but found the
> package on src.fp.o [1].
>
>
> very cool, thanks for the update Aniket.
>
> i did like your idea about some sort of package list on the ml-sig 
> wiki or something. that would be cool, but might be a big load for 
> someone to maintain it.

We do not have to maintain a separated list on wiki because we already 
have the FAS group @machine-lerning-sig which has its own page on 
src.f.o: https://src.fedoraproject.org/group/machine-learning-sig

The list is currently empty and I haven't had time to find important 
AI/ML packages and offer our help with a maintenance but if you know 
some, let me know.

Have a nice day.

Lumír
> peace o/
> [1]: https://src.fedoraproject.org/rpms/hip
>
> On Tue, Nov 19, 2019 at 11:26 PM Aniket Pradhan
> <[e-mail] <mailto:[e-mail]>> wrote:
> >
> > Hello everyone!
> >
> > I was exploring some libraries that can use the AMD GPU for
> > machine/deep learning. I was able to find some (not-so-promising)
> > results, but I also found the HIP [1] library for directly
> converting
> > the CUDA code to portable C++ code, such that it can run on both
> > NVidia or AMD GPUs.
> >
> > I am sure some of you might be interested in packaging this lib for
> > Fedora. Its backend ROCm-runtime [2] is also available in the
> > repositories. It might just need a few more deps to be packaged (not
> > too sure about it), but that's for later.
> >
> > I wasn't able to find a list of packages the SIG is
> > maintaining/packaging, and I was wondering if I could push this
> > library on the list as well. I can also help in the packaging
> process
> > if anyone is up for it. Do share your feedback regarding
> packaging the
> > lib, whether it should be packaged or not.
> >
> > [1]: https://github.com/ROCm-Developer-Tools/HIP
> > [2]: https://apps.fedoraproject.org/packages/rocm-runtime
> >
> > --
> > Thanks
> > Regards
> >
> > Aniket Pradhan
> > http://home.iiitd.edu.in/~aniket17133/
> > Aliases: MeWjOr/major
> >
> > ()  ascii ribbon campaign - against html e-mail
> > /\ www.asciiribbon.org <http://www.asciiribbon.org>  - against
> proprietary attachments
>
>
>
> -- 
> Thanks
> Regards
>
> Aniket Pradhan
> http://home.iiitd.edu.in/~aniket17133/
> Aliases: MeWjOr/major
>
> ()  ascii ribbon campaign - against html e-mail
> /\ www.asciiribbon.org <http://www.asciiribbon.org>  - against
> proprietary attachments
> _______________________________________________
> ml mailing list -- [e-mail]
> <mailto:[e-mail]>
> To unsubscribe send an email to [e-mail]
> <mailto:[e-mail]>
> Fedora Code of Conduct:
> https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> List Guidelines:
> https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives:
> https://lists.fedoraproject.org/archives/list/[e-mail]
>
>
> ml mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
> Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives: https://lists.fedoraproject.org/archives/list/[e-mail]

[Aniket Pradhan]

Hello Michael
> i did like your idea about some sort of package list on the ml-sig wiki or something. that would be cool, but might be a big load for someone to maintain it.

Turns out, it really is not a load to manage it. Have a look at
NeuroFedora-sig's issue board [1].
We just create new tickets to make a list of what all the tasks we
have and what packages we have to package. We also have a
documentation page [2] where we list the packages provided by the
NeuroSig.

I guess the ML-sig can adopt a similar strategy. if the SIG wants to,
I am always here to help. :D

[1]: https://pagure.io/neuro-sig/NeuroFedora/issues
[2]: https://docs.fedoraproject.org/en-US/neurofedora/overview/

\o

--
Thanks
Regards

Aniket Pradhan
http://home.iiitd.edu.in/~aniket17133/
Aliases: MeWjOr/major

()  ascii ribbon campaign - against html e-mail
/\  www.asciiribbon.org   - against proprietary attachments

[Michael McCune]

very cool,

thanks Aniket and Lumir, i hadn't seen those resources before. it /does/
seems quite easy to collect, we just need to identify the packages and get
involved.

peace o/

On Wed, Nov 20, 2019 at 1:56 AM Aniket Pradhan 
wrote:
> Hello Michael
> > i did like your idea about some sort of package list on the ml-sig wiki
>
> or something. that would be cool, but might be a big load for someone to
> maintain it.
>
> Turns out, it really is not a load to manage it. Have a look at
> NeuroFedora-sig's issue board [1].
> We just create new tickets to make a list of what all the tasks we
> have and what packages we have to package. We also have a
> documentation page [2] where we list the packages provided by the
> NeuroSig.
>
> I guess the ML-sig can adopt a similar strategy. if the SIG wants to,
> I am always here to help. :D
>
> \o
>
> --
> Thanks
> Regards
>
> Aniket Pradhan
> http://home.iiitd.edu.in/~aniket17133/
> Aliases: MeWjOr/major
>
> ()  ascii ribbon campaign - against html e-mail
> /\  www.asciiribbon.org   - against proprietary attachments
> _______________________________________________
> ml mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
> Fedora Code of Conduct:
> https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives:
> https://lists.fedoraproject.org/archives/list/[e-mail]