Teleconference 2017-02-22

From VTKM
Jump to navigation Jump to search

Attendees: Ken Moreland (SNL), Tom Fogal (NVIDIA), Hank Childs (OU), Dave Pugmire (ORNL), James Kress (ORNL), Tom Otahal (SNL), Mark Kim (ORNL), Dave Lonie (Kitware), David Rogers (LANL), Berk Geveci (Kitware)

One XVis milestone for ORNL is to do performance modeling with Oxbow. However, Dave P is having issues with getting time from the Oxbow team. Instead, Dave would like to instead to do profiling with TAU, which supports GPUs as well as CPUs.

Dave P. sent out an email about the data directory structure. Currently the baseline is using a data structure that matches the source directories. This sounds OK. Berk suggested to instead have a flat structure that matches the test name as that would make it easier to find baseline images when a test fails.

Hank has set up all the rooms for the code sprint April 4-6. He will set up a web page and send out an email later today.

Brent pushing on maximal clique algorithm. Currently having an issue with memory consumption.

Tom has been talking with Alberto @ Intel to help with optimizing VTK-m on KNL platforms.

Dave L. working on getting the clip filter installed in VTK.

Berk has started working with strawman. Given our current team, we are considering focusing on this work for "flyweight in situ".

There is an issue on Mac with the rendering library. The rendering code is trying to cast cell sets to their explicit types, and that is failing. Ken guesses this is an RTTI issue with the new exports in the VTK-m libraries. This is becoming a stopper for Berk. He is going to bring in Dave L. to take a look.