Present: Helvi, Roland, Peter, Yosef
ET tutorial server * cannot currently create account requests via web-form: fixed as of beginning of week, was broken March till June * compiling produces many server down / server up emails: still happens due to latency of frontend responding to http requests, could either ask to move VM to different node or to split system into "login" VM and "worker" VMs for users * Roland to ask admins to move to a different hardware node
Provide compiled version of the ET * https://bitbucket.org/einsteintoolkit/tickets/issues/2260/provide-compiled-e... * useful for people that are less experienced in numerical work, which may also be the same group of people that are likely to have problems setting up docker * Helvi may have time to try this out on her Ubuntu system, serving as a test-case since she has never used docker before
Memory issues: * https://bitbucket.org/einsteintoolkit/tickets/issues/2194/memory-increase-du... * initial idea was that this could be communication buffers that were never shrunk * patch that frees this memory was provided but does not seem to help in all cases * two more cases that fail, small one by Miguel, large one by Gabriele * memory usage in failing case by Miguel is very uneven, possibly due to communication buffers
ET US workshop at RIT * broadcast will likely be done via bluejeans * working on having camera and mike for remote participants to see the screen and the audience * also for lunch discussion * link will be send to mailing list and be available on workshop page https://ccrg.rit.edu/content/events/2019-06-17/north-american-einstein-toolk... * have discussion about changed citation policy for ET during lunch discussion
Gallery examples will be updated * TOV: small version for online/jupyter tutorial needs to be uploaded * BBH: in progress, waiting in queue on local cluster * BNS: in progress, Peter travelling next week
Peter's self-force code: * no updates, will discuss at Capra meeting next week
unanswered question on mailing list / open tickets sorted by update time * none at this point
There will be no ET call next week (just after the ET meeting), next call June 27th.
Yours, Roland
Dear all,
My own thorn runs into NaNs just before evolution starts. To search where exactly those NaNs are appearing I included subroutines which print example values for the corresponding variables and changed the schedule position of these output routines. (See SWTNS_inittest and SWTNS_regridtest inside the schedule.ccl)
The last slot inside the schedule where the values still seemed to work, is inside [CCTK_PREREGRID]. In [CCTK_PRESTEP] the values are already messed up. The only schedule bin that is in between those two is [CCTK_POSTREGRID]. I don't have any regridding and whenever I schedule my output routine inside the [CCTK_POSTREGRID] bin I don't have any output, as you can see in the attached output file.
Maybe I should point out that I use the Llama Multipatch environment for my thorn.
Does anyone have an idea how I can check what happens with my initial data before evolution starts?
I attached the schedule of my thorn and a simulation output file with respect to the (as well attached) parameter file.
Thanks a lot!
Best regards,
Severin
Severin
The major algorithmic step that happens between regridding and evolution is cycling time levels. At the beginning of each evolution step, time levels are cycled, so that the current time level of all evolved variables contain undefined values, to be filled in during evolution. Since you disabled GRHydro, no time evolution happens, and these undefined values are detected by the NanChecker.
-erik
On Fri, Jun 14, 2019 at 12:57 PM Severin Frank severin.frank@uni-tuebingen.de wrote:
Dear all,
My own thorn runs into NaNs just before evolution starts. To search where exactly those NaNs are appearing I included subroutines which print example values for the corresponding variables and changed the schedule position of these output routines. (See SWTNS_inittest and SWTNS_regridtest inside the schedule.ccl)
The last slot inside the schedule where the values still seemed to work, is inside [CCTK_PREREGRID]. In [CCTK_PRESTEP] the values are already messed up. The only schedule bin that is in between those two is [CCTK_POSTREGRID]. I don't have any regridding and whenever I schedule my output routine inside the [CCTK_POSTREGRID] bin I don't have any output, as you can see in the attached output file.
Maybe I should point out that I use the Llama Multipatch environment for my thorn.
Does anyone have an idea how I can check what happens with my initial data before evolution starts?
I attached the schedule of my thorn and a simulation output file with respect to the (as well attached) parameter file.
Thanks a lot!
Best regards,
Severin
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
users@lists.einsteintoolkit.org
-
Erik Schnetter -
Haas, Roland -
Severin Frank