We’ve been having discussions over the last few weeks with
stakeholders both inside and outside of AMD about where we could best
have a dedicated and open discussion space for topics around the
AMDGPU target. The conclusion was that we’d like to try the use of a
category in the LLVM Discourse group, which is mostly used for MLIR
discussion so far.
I have started a Discourse topic with more detail, as that seemed
rather fitting:
However, this obviously affects the broader LLVM community, hence this
e-mail on llvm-dev. If people are generally okay with such a move, we
would ask the Discourse admins to please create that new category.
I don't have much personal interest here - but my understanding was
that there was/is a fair bit of pushback to fragmenting the
communications channels to discord before there's a more general
buy-in to switch over across the project? (perhaps I'm misremembering
the previous threads about discourse, though - and we have already
pretty significantly fragmented on the chat-like medium between IRC
and Discord)
I don't have much personal interest here - but my understanding was
that there was/is a fair bit of pushback to fragmenting the
communications channels to discord before there's a more general
buy-in to switch over across the project? (perhaps I'm misremembering
the previous threads about discourse, though - and we have already
pretty significantly fragmented on the chat-like medium between IRC
and Discord)
Discourse is really more of an alternative to mailing lists, not to chat.
For what it's worth, the alternatives we've looked at were:
1. Just use llvm-dev. This avoids the fragmentation you mention, but
it was considered undesirable for several reasons:
1.1. Most people watching llvm-dev wouldn't be interested, and we
don't want to increase the noise unnecessarily.
1.2. Some people in the relevant audience expressed that they prefer
not to subscribe to llvm-dev due to its volume, or are worried that
the topics they're interested in would be drowned out in the general
noise.
1.3. Some people in the relevant audience expressed that posting to
llvm-dev can be scary.
2. Create a new mailing list amdgpu-dev@lists.llvm.org. I think the
relevant audience would be mostly fine with that, but there were two
perceived advantages to the Discourse path:
2.1. If a topic in the amdgpu discussion space come up that turns out
to be interesting to somebody who is not usually following that space,
then the barrier to jumping into that topic is lower on Discourse --
there aren't any issues with posting to a mailing list that one isn't
subscribed to.
2.2. We believe that maintaining a new mailing list is a higher burden
for the LLVM project (I'm happy to be corrected on that point).
Are you saying that you think we should use one of those alternatives,
or do you have yet another alternative in mind?
> I don't have much personal interest here - but my understanding was
> that there was/is a fair bit of pushback to fragmenting the
> communications channels to discord before there's a more general
> buy-in to switch over across the project? (perhaps I'm misremembering
> the previous threads about discourse, though - and we have already
> pretty significantly fragmented on the chat-like medium between IRC
> and Discord)
Discourse is really more of an alternative to mailing lists, not to chat.
I'm not sure what discussion went into MLIR using discourse - whether
it got in sideways by being the way the project already interacted
before it came under the LLVM umbrella. But I have some concerns about
the fragmentation there.
Perhaps most of the push back is about the existing LLVM lists and
doesn't apply as much to new subcommunities, but I have my
doubts/concerns about fragmenting the tools used to interact with LLVM
communities like this - making it harder for folks to move between
different subcommunities when contributing across the boundaries.
it's still not clear to me whether you actually have an alternative?
I think as the discussions pertain to code in the LLVM project proper,
the place to discuss them would be llvm-dev. Though I appreciate the
difficulties in doing so, to be sure (its busy, intimidating (any
particular ideas how it comes off as intimidating? I'd like to help
change that in any way I can - for general improvement whether or not
it contains AMDGPU discussions or not)).
I'm mostly trying to relay the previous discussions/concerns that
other folks expressed - so the conversation can start from there/have
that context. I don't know that I feel especially strongly personally
(not as much as other folks on that previous thread do, at least).