Present: Roland, Bhavesh, Erik, Jim, Peter, Roberto, Samuel, Steve,
Yosef, Frank, Qian, Ian, 

con2prim workshop at PI:
* Erik had hosted a workshop to learn about current state of the art of
  con2prim methods and eos tables for hydro codes
* Erik gave a quick report on what was learned at the workshop
* good design should have a "cleaning" stage that removes non-phyisical
  points in a well defined manner so that the actual root finding can
  be robust
* people interested in this: Philipp Moesta, David Radice, Wolfgang

big tickets items for the next release:
* want to get them in now to have them well tested for next ET release
* Roland, Ian, Steve have leftover "cleanup" tickets from the release
* need to update trac landing page to point to current release

new thorns:
* GiRAFFE: make sure there is outside interest and push by authors
* Lean-Public, Proca: make sure there is outside interest and push by
* suggest third-party-thorns registration capability, add this to
  discussion items for next week

BNS gallery example:
* Roberto suggests the difference could be due to F90 vs C++ code, can
  check which code was originally used to compute the gallery example
* Roland to re-run with code from time of gallery example creation to
  check on differences between clusters

ET workshop at GT:
* date (June 18th - June 20th) are fixed now
* forming a scientific organization committee right now
* working on registration procedure
* will send url of website to users mailing list once known

OpenMP scalability on Stampede SKX:
* Jim is testing threading changes in OpenMP usage during prolongation
  by Erik
* see this email for results:
* finds ~25% benefit when using newer code
* should update OpenMP settings in Simfactory to have more MPI ranks
  per node (right now uses 2 which is among the worst possible settings)


My email is as private as my paper mail. I therefore support encrypting
and signing email messages. Get my PGP key from .

Attachment: pgpmMxyQDsSTP.pgp
Description: OpenPGP digital signature

Users mailing list

Reply via email to