Here are the minutes from today's meeting.
Present: Keith, Peter, Allen, Bruno, Zach, Yosef, Leo.
Chair: Roland
Minutes: Beyhan
ET release:
- BBH gallery example: Allen has some issues with his cluster, still needs 24 hours to figure it out.
- Canuda Scalar thorns: documentation and tests are missing. Roland added some documentations for the arrangement. It is too late for test cases, Roland will ask Miguel Zilhao to have some test cases.
- Failures in GRHydro persists in many clusters (Sunrise, SuperMuc, Supermic, QueenBee, db1 and Cori):
Might be related to the intel compiler but stampede that also uses intel compiler seems to be fine (Roland). Might depend on the version of the compiler (Peter). Probably needs to be older than intel 19. Roland will try and see whether it will be overcomed by using an older intel version on QueenBee. It might also be a compiler bug. Yosef thinks he has reported this. They were getting nans in the matter fields after the first iteration with GRHydro (in the gallery example).
- Release will probably not be made today due to these failures (Roland). Roland will try to do it before December.
Unanswered questions
Shamim was in contact with Erik and Roland who provided some suggestions. It would be great to have a volunteer to answer Shahim's further questions due to the release management of Roland.
Open tickets:
- #2664 : Publication webpage is outdated, sphinxcontrib-citations may help keeping it updated automatically:
Suggestion and comments by Roland, Zach, Bruno, Yosef and Leo on a better way to find which papers are using ET.
Roland could give this a try. We are not going to get all of them because Zenodo doi's are not very often cited, people sometimes still cite original ET papers and sometimes do not cite anything just say we used ET.
Zach: script that looks at all thorns in the ET, finding any paper that corresponds to them would really help with this.
Bruno: It would be a good idea to use the script and manually evaluate what the script produced.
Roland: It requires familiarity with the people in the field.
Zach suggested that it would be effective to add this to a regular item in each call. It is similar to the unanswered question in the mailing list.
Roland: We can put it on the agenda. We could also in principle have a web form with just a text box that arxiv links could be dropped in.
Bruno: We can just search for the paper that cites the ET paper.
Zach has some concerns because people who use the ET do not always cite the ET paper, they cite the thorn they use.
Yosef: Most people that use ET just mean it.
- #2663: add a notice about minimum Intel / GCC version required to configure output:
Most of the questions about not compiling ET with intel compiler ends up being wrong standard C++ template library.
For C++11, the error message given is not really helpful. Error message should tell the user "you need to use at least intel 2017 and gcc 6". (Roland)
Something to make simfactory look for a couple places to guess a good email address when setting up a new machine (Roland).