Teleconference 2017-04-12

Attendees: Ken Moreland (SNL), Brent Lesley (UO), Alberto Villarreal (Intel), Matt Larsen (LLNL), Dave Pugmire (ORNL), Ollie Lo (LANL), James Kress (ORNL), Robert Maynard (Kitware), Tom Otahal (SNL), Mark Kim (ORNL), Dave Lonie (Kitware), Berk Geveci (Kitware), Sujin Philip (Kitware), Manish Mathai (UO), Sudhanshu Sane (UO)

Ken has been working with Brent and Matt on investigating different hash functions for external faces. He has finished some benchmark comparison code and is running on some platforms at Sandia. He will soon compile the results and engage Brent and Matt to write a paper.

Dave and several other folks from ORNL used the code sprint to get a jump start on the particle advection code. The have some initial working code and started doing some scaling studies. Got good scaling to 28 TBB cores.

Brent has also been working on solutions for maximal clique problem.

Manish is looking at putting search structures into the execution environment. It can be part of the data set and be available to all filters.

Tom worked with Rob to make a tiling example to get some data from our KNL node. Should have some results soon.

Alberto has been experimenting with the benchmarks to change the grain size used in the TBB scheduler.

Matt worked on trying to get volume rendering to be way more general. Is trying to finish that this week.

Ollie is working on segmented scan.

Sujin has been working on virtual method calling in VTK-m, particularly with implicit functions. With his current implementation the difference is negligible.

Dave L. is still working on getting the size of the VTK-m library down by compiling in some versions of ArrayHandle.

Rob answered many questions at the hackathon. His main work was looking at the TAU results and found some strange places that VTK-m was spending its time. It was odd that normals were generated before point reduction. Exploring looking at computing normals after points are merged.