[Users] Meeting Minutes 2020-11-05

Gabella, William E b.gabella at Vanderbilt.Edu
Thu Nov 5 10:19:26 CST 2020

Present: Yosef Z, Bill G, Steve B, Roland H, Ken S, Peter D, Atul K , 
Zach E, Maria BH, Samuel C, Miguel G

Chair: Yosef
Minutes: Bill

* https://carpetcode.org is gone [RH]

Roland, Erik S has removed carpetcode.org and Erik has canceled the DNS 
entry.  We have to learn what his plans are.
Roland will reach out to Erik, has a phone call with him today.

* upcoming release planning (3.5 weeks to go, 1 weeks to feature freeze, 
1 weeks to have everything reviewed) [RH]

November 12 all codes proposed need to reviewed.  Email for NRPyPN 
today, for official vote next week at the Thursday meeting.

Naming.  Bill suggests Dewitt-Morette.  Roland likes a person who died 
long ago, but Hack was more recent.  No other objections.

** revert of LAPACK and BLAS changes to use bash_utils, Ticket #2476

Roland, shortly after last release changed to use BLAS utils helper 
functions for what to use for /usr/lib64 or similar. LAPACK and
BLAS dir point to the /lib directory, and they fail on some clusters.  
Roland reversed them to the Turing release, and he will take a
look after this release.  Steve, also recalls something about the HDF5 
variables...finding the include directory is a problem, etc,
after its basedir is set.

Steve, has been testing cactus on different compilers and there is a 
version where testAH does not pass nor stop, gcc 4.9.1.
Other gcc's seem fine.

** Gallery test runners search [[Release Details]] [RH]

Single, stable neutron star] [Miguel Gracia]

Finished the test and still has to make the files and figures and upload 
to the repo.  Roland, should also test after the feature freeze.
Sometimes if large changes last minute, test again after release.

Multi Patch Scalar Wave Equation] [Rahime Matur]

Poisson equation] [Bill Gabella]

Ran it last time and does not expect any problems in the test, same HPC, 

Binary neutron star] [Beyhan Karakaş]

Binary black hole GW150914] [Steve Brandt]

Steve, has not done it yet..."might" do it.

Also the state of the cactus.org website.  Seems https://cactus.org is 

** release testsuite 
[http://einsteintoolkit.org/testsuite_results/index.php status]

Page shows a few failing tests.  Roland, run testsuite on all public 
clusters in Simfactory.  See the text after the table for changes from
last release.  Shows the status if different from last release. See "New 
tests that fail" and "Newly passing tests," etc. TwoPunctures,
just initial data maybe one or two time steps.  In Test.ccl has a 
threshold of 1% for TwoPunctures and normally we use 10^-6 or so. Error
goes away if compiled with IEEE conforming flag.  In this release 
replaced ADM Constraints which have been out of the Toolkit for some
time, and this test is using other tests instead---so it is a new 
TwoPunctures test.  Yosef, suggest just putting a grid in between the 
two punctures.

Zach is experiencing non-robust behavior on a certain cluster for Baikal 
and McLachlan.  That HPC uses the modules (LMod?) to change
environment and configuration.

** Tickets marked for the upcoming release

Includes tickets for the Gallery Examples, and major bugs.  The person 
who created the ticket classifies it, so check on ones you have
created and be sure the are "major bugs."

** Open bugs of severity major or worse

* Unanswered Mailing list, 

ld error by Maria is the only real issue, others are informative.

* Open tickets sorted by update time

* Tickets for Review

Always looking for volunteer reviewers.  Steve will work more on #2443.  
Zach will close #2432, it is fixed.  Roland, usually assign
ticket to important persons involved with the code.  Minor ticket or 
enhancement Roland will not hassle people about the ticket and fix.

* Working group updates

Zach, do not have a group around LaTeX parser, but close to having BSSN 
in Cartesian coordinates as LaTeX able to parse into Sympy etc.
Sanitizer for the LaTeX being made more robust and allowing for users to 
specify \xxxx LaTeX defs.  Lie derivatives and covariant derivatives
are supported using the macro language for the parser that show up as 
LaTeX comments, like % define hDD, for a tensor with two down indices.

CarpetX update.  Roland, Sam Cupp is working with CarpetX to learn it 
and add support for grid arrays and enhance existing grid interpolator
interface.  Goal to run binary black hole with carpetX and extract 
gravitational waves.  First runs will be slow, but will check 
compatibility with
current toolkit.

* Spec Benchmarks

Steve, did anyone do anything on this.  Let us not forget about this.  
Add to agenda and copy forward so that we do this.  We need the Spec 

* External Thorns

Steve, we do not like all the external thorns installed a different 
ways, endless other nuisances.  Steve has a short script that installs 
the dependencies
from Spack.  Could do them all this way.  If you have Git access, a user 
can install Spack.  It will want to download code which one HPC blocks 
(?).  In
principle, you could compile elsewhere and move to the HPC.  Steve is 
enthusiastic about Spack.  Some clusters block its functionality, we 
think.  Roland,
we still need the external library thorns, and the all have a "lib" 
variable and that could be installed by Spack or by the usual methods.  
Cactus does
interfere with some of this, the build scripts which show various hacks.

Spack  https://spack.io/about/


William Gabella
Research Assistant Professor
Department of Physics and Astronomy
Nashville, TN USA
(o) 615-343-2713

More information about the Users mailing list