[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: AMD ROCm packaging session followup notes



Hi Étienne,

Thank you very much for sharing the notes!

On 2021-11-24 17:39, Étienne Mollier wrote:
>  Q: In which order to build the different components?
> 
>  A: This is still not entirely clear, but trying to package
>     targets will eventualy reveal dependency trees (hopefully
>     without loops involved).

Indeed. The last time when I try to figure out the hierarchy for
ROCm components, I was confused. But at least the tree structure
can be revealed by reading the code.

>  Q: Where to put llvm related files?  (so it meets FHS [3])
> 
>  A: As changes brought to llvm-amdgpu are upstreamed, they make
>     it to subsequent llvm versions, so making use of the latest
>     upstream llvm/clang might actually be an option.
> 
>     [3]: https://refspecs.linuxfoundation.org/FHS_2.3/fhs-2.3.html

That's what I'm glad to hear. Previously my largest concern on
ROCm is the llvm fork. Our LLVM team will be unhappy to see yet
another llvm fork packaged for Debian. If we can simply use the
upstream llvm, things can't be better.

Out of interest, I have an additional question. Since kfd is already
present in the upstream kernel, does ROCm require any non-free
(per Debian's DFSG definition; e.g. firmware-amd-graphics from
non-free/kernel section) component to run? I have no any hardware
to figure it out by myself. I ask this because dealing things
in the main section is much more comfortable than that in the
contrib or non-free sections.


Reply to: