Hello Bruno, Roberto,
I cannot say anything about building on Marconi since I have no access to it. Roberto apparently built the code on the compute nodes (Marconi is not the only machine we end up doing this), though building on the login nodes (if doable without producing a lower performant executable) would be preferable since it would allow one to use simfactory's --remote command.
Please, if you and Roberto could discuss this offline and implement your preferred solution, that would likely be best.
Yours, Roland
Roland, thanks! We were going to do the same in these days. Btw, in order to compile on the login nodes of MarconiA2 we need to change -xMIC-AVX512 into -axMIC-AVX512 and also update the memory request in the submit file which is currently set to 90000, but should be lower (these were the minor changes I mentioned before).
Cheers,
Bruno
2018-09-04 19:07 GMT+02:00 Roland Haas rhaas@illinois.edu:
Hello Bruno, all,
Roberto de Pietri (thank you!) kindly ran the tests on Marconi and things seem to work well there:
https://einsteintoolkit.org/testsuite_results/index.php
Yours, Roland
Hello Bruno,
thank you. That would be great. Ideally if you could run the testsuite
using the --testsuite option of simfactory using 1 node and both 1 and 2 mpi ranks that would be best. The results would ideally be committed to the repo https://bitbucket.org/einsteintoolkit/testsuite_results behind the testsuite status page https://einsteintoolkit.org/ testsuite_results/index.php via the summary.log file.
Yours, Roland
----- Original Message ----- From: Bruno Giacomazzo bruno.giacomazzo@unitn.it Sent: 2018-08-28 - 04:53 To: Roland Haas rhaas@ncsa.illinois.edu Subject: Re: [Users] meeting minutes for 2018-08-27
Hi, just a quick note that I use Marconi A2 and A3 without problems. I
will
check with the latest stable release. With the previous one, I needed
very
minimal (and very easy) changes to the config file and submit script.
Cheers, Bruno
2018-08-27 16:59 GMT+02:00 Roland Haas rhaas@illinois.edu:
Present: Steve, Roland, Yosef, Zach, Helvi, Sam, Bhavesh, Peter, Patrick Nelson, Bill
Working group updates:
- Perforrmance optimization: nothing to report
- Education: nothing to report
- Cosmoparticle: nothing to report
- IllinoisGRMHD: updates are on the wiki page,
https://docs.google.com/presentation/d/1ZBc8VQGZm3Gj3fu5TO75TW8swc4VE TxnKy1Jrj_w2eU/edit?usp=sharing
Gallery examples for next release:
- BBH: comparison between current code and gallery data in Zenodo, results are pretty close. Plots are here
August/006471.html
- poisson: Roland updated code, with this he can reproduce the plots, updated the website
- BNS: Peter found out that change of using EOS_Omni to compute quantities cause most of the initial differences, after merger there are still differences. So far no input from Frank or Roberto, Peter will contact them directly.
Reviews:
- WVUDiagnostics: no progress, want testsuites. Push to next release unless we have tests soon.
- GiRaFFE: main thorn reviewed, most of GiRaFFE_Food, Zach and Roland to prioritize this request over others
- Lean: Erik to review, likely not before the release
- Proca: Peter made some progress, will continue to push it this week
Documentation:
- would like some documentation explaining how to set up checkpoint / restart
IllinoisGRMHD robustness:
- Zach reduced the maximum allowed speed to have a max. speed less
than
c_light
- finds that robustness is greatly increased if one uses 3rd order Lagrange interpolator for IllinoisGRMHD variables. There is a pull request for such an operator:
https://trac.einsteintoolkit.org/ticket/2189
Testsuites:
- remove MPG machines: draco, minerva, hydra are MPG machines.
Wolfgang
said he could maybe run on one of them, if not then we will remove them
- Peter will run smic
- Marconi: a machine in Italy, Roberto de Pietri, Eloisa had access,
if
they cannot run, then we should remove them
Moving website to github:
- may be difficult to move due to multiple repos being linked up
- some pages use php and cgi which are not straightfoward to run from github
- currently have docker image that makes the server much more self-contained than before, docker image is here:
Working group calls:
- IllinoisGRMHD on Wednesday 11:00AM US Eastern Time
- Performance optimization on Friday 12pm US Central time
Yours, Roland
-- My email is as private as my paper mail. I therefore support
encrypting
and signing email messages. Get my PGP key from http://pgp.mit.edu .
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
--
Prof. Bruno Giacomazzo Department of Physics University of Trento via Sommarive 14 38123 Trento Italy
Tel. : +39 0461281631 email : bruno.giacomazzo@unitn.it web: http://www.brunogiacomazzo.org
There are only 10 types of people in the world: Those who understand binary, and those who don't
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
-- My email is as private as my paper mail. I therefore support encrypting and signing email messages. Get my PGP key from http://pgp.mit.edu .