Teleconference 2018-01-10

Attendees: Ken Moreland (SNL), Matt Letter (SNL), James Kress (ORNL), Matt Larsen (LLNL), Rob Maynard (Kitware), Sujin Philip (Kitware), Berk Geveci (Kitware), Dave Pugmire (ORNL), Tom Otahal (SNL), Utkarsh Ayachit (Kitware)

OpenGL/X test failures back.

Matt Letter increase worklet arguments from 9 to 20. (Also increased function interface arguments.)

Tom has submited a MR for the radix sort. Rob suggested compiling some of it into a library and also specifying more fast passes. Rob also brought up that we want to use radix sort for sort-by-key, which is not happening right now.

Utkarsh has been making updates to DIY to work well in VTK-m and remove warnings. Working on making DIY build optional. Moving forward to making DIY a ThirdParty library.

Given the mounting pressure to drop MSVC 2013, we plan to drop support at the beginning of February.

Rob is working on infrastructure to support scalars to colors. Making some fixes to dot product to support promoting small ints to 32 ints. Also some fixes to the promotion of data in Magnitude.

Progress begin made on getting the virtual coordinate arrays to build everywhere. Down to problems on only a few dashboards.

Sujin working on a probe filter in VTK-m.

Dave is putting together a VTK-m poster for the ECP meeting.

Dave and James integrating VTK-m in ADIOS services.

Matt Larsen has finished implementing spheres for point rendering.

Matt Larsen created a merge contours method for multiple slice planes.

James brought up a proposal to drop the OpenGL rendering code in favor of the raycast rendering. We had a conversation and could not come up with good use cases for the OpenGL rendering other than simple examples. The main use case of the VTK-m rendering is with batch processing. In this case, interactive contexts do not matter, and the performance is probably close enough between them. There is general agreement that it would be OK (and in some ways beneficial) to remove OpenGL.

No meeting next two weeks.