Teleconference 2017-10-11

From VTKM
Jump to: navigation, search

Attendees: Ken Moreland (SNL), Matt Letter (SNL), Rob Maynard (Kitware), Sujin Philip (Kitware), Allie Vacanti (Kitware), Berk Geveci (Kitware), James Kress (ORNL), Dave Pugmire (ORNL), Mark Kim (ORNL), Tom Otahal (SNL), Abhishek Yenpure (UO), Matt Larsen (LLNL)

Ken is working on the user's guide in anticipation of a VTK-m release this quarter. Rob proposes targeting the first week of November.

Allie is working on installation set up and VTK-m enabled in ParaView all the time.

Allie is working on improving copy methods. Copy is sometimes slower in parallel than in serial (when running multiple threads on a single processor socket).

Sujin is finishing off the cell locator work. It has been merged in today.

James is working on the ECP showcase demo. He now has access to Sujin's cell locator classes, so that should help.

Tom has been working on the CPU parallel radix sort. Found an implementation based on a paper from 6 years ago. Performs pretty well, but implementation uses OpenMP. The conversion to TBB may be difficult because the algorithm deals with groups of threads. Rob brought up the TBB feature of task arena that might make the conversion easier.

Abhishek has been working on streamlines and has been running into issues with copying a subset of memory in and out of ArrayHandles.

Matt Larsen has been running into trouble with the output of threshold using a cell set that is not discovered. It needs to be copied to an explicit cell set. Right now there is implementation in VTK-h. No branch in VTK-m (yet).

No meeting next week.