Hi,
Hi!
What is the current state with cluster encoding (PVM) in current CVS ?
Last time that we've discussed about this IIRC there was a general consensus on following statements:
1) PVM support, and in general cluster encoding, it's a really nice thing and should be mantained and improved.
2) cluster encoding need a better support into transcode core, not (only, mainly) in modules.
3) we aren't, uhm, enthusiast about current code (see point above, there is a fair amount of duplication and
integration can be greatly improved)
4) we (devs, current maintainers) are interested on this topic but due to severe time constraints we have
other priorities.
Personally speaking improving cluster support isn't in my short (1.2/1.3) nor medium ( 1.4/1.5) plans.
A quick diff of trees between 1.0.2-stable and current CVS shows some changes
made in pvm3/ so am wondering whether anyone is starting to work on this.
Yep, mainly (only?) cleanup, API resync and maintainership work
At present pvm does not compile in CVS (yes I know it's not stable :)), due to
the removal of libioaux/ which at a glance also seems to affect v4l1.
Uh-uh this must be corrected before to go 1.1.0-stable , and I wasn't aware of that. Thanks for report, I'll put this
task in my TODO queue.
Though I think v4l2 is preferred in the latter case now.
Definitively.
In current stable 1.0.2, pvm does not work due to aclib/tcmemcpy
incompatibilities but at the very least does compile.
Can't test this, but of course it should be fixed too. Patches or suggestions of course very welcome.
If PVM or an equivalent is not being actively fixed/developed, then would love
to see this re-implemented and will probably take a poke around myself.
for hotfixes (especially for 1.0.x branch), I can spend some time on this. For new code or improving existing one,
as said before, I can't take care, and one of reasons is that I lack infrastructure to test (yes, I know that PVM can work on a single machine: but I lack even libpvm installed right now ;) ).
Moreover, a redesign of the whole cluster encoding subsystem and better integration with core was definitively a need
for the long term (so for new code).
If you want/like to take care, of you simply want to send some suggestions request, just drop us a note :)
--
Francesco Romani
