The next Math meeting is 7/18/2019, 10 am Eastern. If you want to join, please either use this link or DM me a Google handle.
Attendees
@roualdes, @charlesm93, @syclik, @rok_cesnovar, @stevebronder, @wds15
Discussion
We’re still working on a format that works well for us. I thought today’s meeting was quite productive. We went around the room, discussing issues that were blocking, and walked through the existing PRs assigning them to people where appropriate.
-
Flattening refactor. Discussion led by @roualdes. We discussed status, what the next steps are, and where the challenges have been. I believe at the end, we had a path forward with @roulades revisiting the scripts and hopefully not being stuck at the same places. We discussed the merits of breaking apart a large pull request into smaller pieces so it’s easier to review.
-
@charlesm93 mentioned that he’s not blocked and is currently working on experimental work and was here to support the rest of the developers.
-
@rok_cesnovar wanted to walk through the open PRs and assign them so we’re moving on them. We tabled that until the end.
-
@stevebronder mentioned he had some compilation errors on a PR due to the meta traits not being included in the correct order. He’s going to post on Discourse his technical problems.
-
@wds15 had a few blocked issues and led us through them small to big:
-
log gamma function. It’s hard to test for the condition he wants to change the implementation (from
std::lgammatoboost::lgamma). He knows it affects Mac OS, but is uncertain if it affects any of the other OS / compiler compliations. @rok suggested we add to lint checker (make test-math-dependencies) to make sure it isn’t added accidentally. @syclik asked for a summary since it’s hard to tell what’s holding up the PR. -
Intel TBB inclusion. @syclik will respond after finishing the thread local storage PR.
-
Thread local storage PR. Testing is almost done. Need to test Windows. @rok_cesnovar volunteered. @syclik needs to send him his current scripts. Hopefully we can characterize the performance on Windows and we’re done!
-
Design proposal. @wds15 described some of his thought process in his design. It seems like we have a better understanding of what needs to be provided: some examples in how the current Stan language could be extended or how Stan3 could express things and show how that lines up with the C++. It was intentionally left out due to the difficulty of working out what a good design for language is, but it’s starting to be clear that it needs to happen.
-
-
PRs. @rok_cesnovar encouraged us to walk through the PRs and we assigned them to different members of the team that were present. Hopefully we can clear them out.
-
Merge access. @syclik needs to give @roualdes merge access on the repo.
-
Wiki for dependencies. @stevebronder will fill one out for OpenCL. @syclik will fill it out for Boost and Eigen.
-
More discussion on TBB. What it’s useful for: load balancing scheduler.
That’s all the notes I took. Some of this has been editorialized, so please send corrections. Thanks to all that could make it! If you want to join us, please find us next month at the link.
If you have something else that’s blocking, please reach out on the forums.