On Thu, 23 Mar 2023 16:18:20 PDT (-0700), jeffreya...@gmail.com wrote:


On 3/17/23 10:57, Palmer Dabbelt wrote:


I'm a little bit confused about what the proposal is here: is the idea
to have a branch based on gcc-13 where we coordinate work before it
lands on trunk, or a branch based on gcc-13 where we backport
autovec-related patches once they've landed on trunk?  In my mind those
are actually two different things and I think they're both useful, maybe
we should just do both?
I was thinking it was a branch to coordinate backports.  We could also
have a branch to coordinate development before it lands on the trunk.


The former provides a base for those who might want a stable gcc-13
based compiler, but with RVV support.  The latter is more focused on
ongoing development.

Yep, just two different things.

That implies we need to identify the principals.  I'll suggest Kito,
Juzhe, Michael and myself as the initial list.  I'm certainly open to
others joining.

+Vineet, who's been handling our internal GCC branches.
OK.



Sorry if that throws a bit of a wrench in the works.
No worries at all.


Just for context: in Rivos land we don't have any specific timelines
around 13, so the goal on our end is just to keep the vectorization
stuff progressing smoothly as we spin up more engineering resources on
it.  Our aim is just to get everything on trunk eventually, anything
else is just a stop-gap and we can work around it (though sharing that
work is always a win).
We don't have hard time lines (yet), but I can work backwards from
various plans and conclude that Ventana will need a gcc-13 with vector
backports, hence my original focus on that aspect of the coordination
problem.

OK. We don't have a hard need there, but it'll make life easier so I'm happy to just treat it like a real shipping branch if you guys are going to as well.

Are you OK just having a single "gcc-13 with RISC-V performance backports" branch, or do you want just vector backports? Our internal branch would be all performance-related backports, but no big deal if the upstream stuff is vector-only as that's probably going to be 90%+ of the churn.

Thanks for raising the need for a development coordination branch.

I guess "need" is kind of strong: IMO it's up to the people actually doing the work how to organize the branches. I'm not writing the code here so I'm happy with whatever, just pointing out that there's two different things that could be done ;)

jeff

Reply via email to