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

Re: rocm-hipamd_5.0.0-1~exp1_amd64.changes is NEW



Étienne Mollier, on 2022-07-03:
> M. Zhou, on 2022-06-30:
> > Cool. I quickly went through the packaging scripts, and updated
> > the ROCm dependency tree:
> > https://salsa.debian.org/rocm-team/community/team-project
> > 
> > I have to admit it is really easy to forget the whole dependency
> > tree if not touching the sources for a while. So it is nice to
> > have a documentation.
> 
> I concur, this is very convenient to check the status of
> packages, and see what change is likely to affect which package.
> I pulled updates to ROCm 5.2.0 for the level 1, trying to make
> sure I was not throwing too much entropy to the entire graph.
> Things seem to look good for the moment.

Level 2 / rocm-device-libs may have to wait a bit though,
because otherwise I end up with missing symbols in a rocm-hipamd
5.0.0 build attempt:

	lld: error: undefined symbol: __oclc_ABI_version
	>>> referenced by /tmp/hipTextureMipmapObj2D-cae27e/hipTextureMipmapObj2D-gfx803.o:(tex2DKernel(float*, __hip_texture*, int, int, float))
	>>> referenced by /tmp/hipTextureMipmapObj2D-cae27e/hipTextureMipmapObj2D-gfx803.o:(tex2DKernel(float*, __hip_texture*, int, int, float))
	clang: error: amdgcn-link command failed with exit code 1 (use -v to see invocation)

I have not tried to bump rocm-hipamd version yet, but I suppose
transitionning the remaining packages to ROCm 5.2.0 would solve
the issue.  I guess it would be preferable to keep rocm-hipamd
5.0.0 ftbfs free while it is still in the New queue?

Kind Regards,
-- 
Étienne Mollier <emollier@emlwks999.eu>
Fingerprint:  8f91 b227 c7d6 f2b1 948c  8236 793c f67e 8f0d 11da
Sent from /dev/pts/2, please excuse my verbosity.

Attachment: signature.asc
Description: PGP signature


Reply to: