LAMMPS developer workshop discussion

I think that a hackaton is a good idea and I would definitely attend if I get any funding (which is likely in my opinion). But I want to back up Oleg’s idea. It’s true that all of us being locked down and no distractions shoots up productivity (for me, at least). But working on things that /could/ be done remotely has a threefold benefit: 1. We don’t have to get any funding; 2. We don’t have to wait to get at least some things done; 3. For the hackaton, everyone should be more comfortable with LAMMPS inner works.

I do realize though that remote work means more work for the mentors, but I think it should be reconsidered. Maybe the best thing to do is write a list of proposed modifications, filter it down and then see if any of them make sense, to the mentors-to-be, to do remotely.

I think that a hackaton is a good idea and I would definitely attend if I
get any funding (which is likely in my opinion). But I want to back up
Oleg's idea. It's true that all of us being locked down and no distractions
shoots up productivity (for me, at least). But working on things that
/could/ be done remotely has a threefold benefit: 1. We don't have to get
any funding; 2. We don't have to wait to get at least some things done; 3.
For the hackaton, everyone should be more comfortable with LAMMPS inner
works.

yes, it *should* work. but in my experience there is a difference
between the theory and the reality of (scientific) software
development in academia. that said, i also have seen and personally
experienced that remote work becomes considerably more efficient and
effective after an in-person session: after i sat next to steve while
he was merging the USER-OMP package and i could see how he works and
immediately explain details that would have otherwise taken us many
e-mail exchanges to figure out, i am now able to provide him with
patches and updates in a way that it much easier for him to digest and
process (and he has gradually become used to my preferences, too).

so i firmly believe that we need something that bootstraps the
process. ideally, this leads to expanding the pool of people that can
help less experienced (in how to do things the LAMMPS way)
developers. from that nucleus, it should be easier to gradually
incorporate more people actively into LAMMPS development and turn this
into a much more distributed process, which in turn will make
organizing future events easier, too.

over the years, we also experimented with merging new code first into
LAMMPS-ICMS, stabilize and clean it up there and then feed it in
digestible chunks to steve. that has worked quite well and recently
the differences between LAMMPS-ICMS and upstream are quite small and
synchronization is done more frequently.
so for example, we could take an initial hack-a-thon as starting point
for a LAMMPS-NEXT repository that is maintained by a group of people
and works in the same way. and for *that* we could then explore the
various ideas that were already mentioned or that people may add.

I do realize though that remote work means more work for the mentors, but I

which is the resource that is currently the most scarce and thus
should be used most effectively. that is my largest concern.

think it should be reconsidered. Maybe the best thing to do is write a list
of proposed modifications, filter it down and then see if any of them make
sense, to the mentors-to-be, to do remotely.

"just do it, man". seriously! i don't have the monopoly of getting
such a thing off the ground. right now, we are exploring ideas and
having some competing concurrent proposals floating around is a good
thing, for as long as it doesn't get out of hand. at some point
decisions will have to be made, but doesn't even have to result in
just one approach.

only, for as long as i am taking the lead, i require to be convinced
to do things different from how i envision it. as you know, i am an
opinionated and egotistical pinhead, so some effort is required. on
the other hand, if there is significant support for alternatives, i am
more than willing to support that as well. only i believe that the
lead would have to be a person that fully believes that his (or her)
way is the best way. i am outright horrible as a manager, when i don't
believe in something. :frowning:

axel.