[Users] Meeting Minutes 2020-08-13

Gabella, William E b.gabella at Vanderbilt.Edu
Thu Aug 13 10:07:52 CDT 2020

Attending:  Steve B, Bill G, Peter D, Atul K, Roland H, Zach E, Yosef Z,
Maria H, Miguel G, Helvi W

Chair: Steve B
Minutes: Bill G

* LSU Online ET Workshop  https://www.cct.lsu.edu/Einsteintoolkitworkshop 

  Seemed to work out okay.  Need the tutorial server to re-boot
automatically, some web pages do not really exit.  Currently three
users are using 123% CPU right now and likely not real.  Did work well. 
Peters tutorial needs to work better than having an extra compiled

** Transitioning from NDSLabs to just the LSU machine.  Ask to re-point
the DNS entry to LSU from NDSLabs.  Also the core flexibility up to
48 cores is easy with the LSU system...on a single node.  To use more
you have to span multiple nodes, and that is possible. 

** Post workshop survey?  Have we done that with the other workshops? 
No one recalls a post workshop survey.  Sounds like a good idea.

** Steve has money for in-person meeting next year...assuming things
loosen up with the pandemic.

* Other Workshops

** No one knows about a European ET workshop and seems late to get one

** Barry Wardell Advances in Computational Relativity workshop at Brown
University, Sep 9 to Dec 11, 2020.  ET should be represented there.
Zach and Peter are attending.  https://icerm.brown.edu/programs/sp-f20/

** Helvi is organizing LISA Waveform Working Group, sometime 7-11
September, 2 or 3 days long. Online with Zoom.

** KITP Santa Barbara Probing Effective Theories of Gravity in Strong
Fields and Cosmology, Monday for three weeks, every day from 8 to 10am.
Open and free since online.  https://www.kitp.ucsb.edu/activities/greft20


Especially want to know how the GPU support works.  It can but down not
have to use CUDA unified memory.  Do not have to worry about keeping
track of the data movement.  They have a heat equation example to can
run on the GPU and does so seemlessly.  You can tell, because you give
them a lambda funciton that tells them how to run.  Multiple MPI
processes with one GPU all run ont he one GPU, but does try to spread
them out. 
IT is possible to have some on the CPU and others on GPUs but not
recommended by AMREX because of load balancing.

Zach, GPU related.  The student that wrote a Latex to C code parser will
be re-joining the efforts funded out of the gran and he will be working on
GPU kernel output starting with CUDA and NRPy.  Original plan is to look
at NVidia sample codes, wave equation in 3D and build off of that.  Steve,
likes the idea of building on top of AMRex to have the most impact down
the road.  Zack, worried about speed-up if you do not get the finite
equations right.  Steve, the AMR is the hard part, need to do
prolongation, restriction, and populate ghost zones and that is
functionality you do not
want to replicate.  Zach, on the Kernel side generating the EEQs so that
there is good performance seems the hard, important part.

* Other topics - None

* Unanswered Questions - None

* Tickets - No discussion 

For Review, no discussion

* Volunteers

Chairs: Peter (Aug 20), Zach (Aug 27)
Minutes: Maria (Aug 20), Roland (Aug 27)

* Anything else

Maria, do analysis and post-processing work?
Roland, It is just a list and gets
updated.  Wolfgang K's PostCactus was just linked there, should be
good.  SimulationTools works.  But we do not test these tools and even
that they still exist.  Steve, should we check those tools?  Roland,
No.  Bill, At least check the links maybe.

Atul, Binary BH gallery, made a branch on GIT and updated the index file
for that page.  Atul, needs to be checked?  Ticket? Maybe not, but link to
ET bitbucket, as a branch,
, click on View source, upper right.

William Gabella
Research Assistant Professor
Department of Physics and Astronomy
Vanderbilt University
Nashville, TN  USA

b.gabella at vanderbilt.edu
(o) 615-343-2713

-------------- next part --------------
A non-text attachment was scrubbed...
Name: pEpkey.asc
Type: application/pgp-keys
Size: 2271 bytes
Desc: pEpkey.asc
Url : http://lists.einsteintoolkit.org/pipermail/users/attachments/20200813/498442c0/attachment.bin 

More information about the Users mailing list