[Users] Meeting Minutes for 2020-01-09
Gabella, William E
b.gabella at Vanderbilt.Edu
Thu Jan 9 10:23:25 CST 2020
Present: Liu Haoyang, Zhao Zhi-Chao, Bill Gabella, Zach Etienne, Steve Brandt, Roland Haas
* ETK Workshop at LSU in August 2020.
o Likely August 3-7, the first week. Let Steve know what you would like to see there. Planning on student tutorial, will have students from LSU.
o Some money for travel support---if you have a student you want to send ask Steve.
o NRPy tutorial would be nice---Zach agrees. NRPy spreading around China; Zach back from a recent visit there.
* Any information about an ETK workshop in Europe? Need to know and coordinate schedules.
* Working group update: Performance Optimization AMRex
o AMRex work with Erik and Steve. AMRex is a mesh refinement library developed by the DoE specifically by LBNL in California. Built-in support
for tiling and multi-threading. Erik, Steve, and Roland, try to use AMRex as an underlying driver for Cactus. Have wave equation working. Roland
visited Erik for a week over break, mesh-refinement code in there, and Zach's Baikal BSSN code that can evolve Minkowski ST without failing.
And can do Schwarzschild BH with two puncture, until it fails. Likely the boundary conditions are wrong. Using two punctures as initial data.
Is a Cactus thorn. Looping over grid points is different, so not backward compatible. Developed as a hydro code. Supports two kinds of variables,
cell-centered and vertex-centered, which is not supported in Cactus at all. Metric is vertex-centered. AMRex uses the PreSync infrastructure of
the Cactus branch Steve has been working on. Code is public. Repo cactusamrex in Erik's account (see below). Is the docker file Erik uses for
testing. Need latest version of cmake. Erik is working on BCs other than periodic, Dirichlet and Reflection. Is a work in progress, but at a point
where we can take Minkowski or Schwarzschild ST and run on a cluster and see how the framework scales---effects on run speed, etc. All
refinements are on the same time step.
o For Minkowski ST when evolve quantities 0 or 1 at least it does not die, but not sure if it is still 0 or 1, responding to Steve's question.
o Took Zach's notebook and re-wrote the glue for it to work. Kept the physics the same.
o Zach suggested a trumpet BH solution with corresponding gauge conditions, guaranteed to be static ST, and is a good test. Also numerically
stable; shows up in the original NRPy paper.
* Steve wants to get rid of cactuscode.org and make it a web page on ETK website. Should have an example web page for next week.
* Next release, 6 months from late October, so maybe April.
o Volunteer for heading the release needs to be determined.
* Roland has some bug fixes to back port to the release.
o System topology thorn will fail if using new (2.X) of hwloc library. Small change to the search.
o In the ticketing system with "backports" in the subject. He would like to do this. Steve will look at the tickets and commits.
o Maybe some updates for simfactory machines, but those get updated all the time.
[edited for readability-my paragraph (non)wrapping looked like hell on lists.einsteintoolkit.org, finally put in hard breaks with shift-return
in Thunderbird, bill]
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Users