For anyone interested I made a review request:
https://bugzilla.redhat.com/show_bug.cgi?id=2090823
I'm not 100% sure what to do with some of the debug related rpmlint errors. Any help
would be much appreciated.
Hi Jeremy,
just wanted to thank you for your contributions to ROCm on Fedora + your
upstream work to make the stack more distro-friendly.
I'm pretty much swamped right now. Maybe I find some time to chime in but I hope
someone else could review this.
Debian has a pretty active ROCm packaging team so a potential reviewer could
also double-check with Debian to verify the packaging. Really, upstream is in a
much better shape than it was 3 years ago with hcc so don't be afraid to take
this review.
If anyone is concerned due to the lack of hardware I can offer to do some
functional testing.
Felix
Thanks Felix,
An open concern that I have been discussing with the ROCm guys is that HIP (used for
things like pytorch) requires rocm-opencl source code to compile.
It seems they want to go with the llvm-project approach in the longterm, having opencl,
hip, and the common static lib "ROCclr" as one source tree. I was going to just
package rocm-opencl as-is for now and add hip as a subpackage later.
Let me know if anyone has any concerns with this idea, or if there's a better way to
handle it.
Thanks