Teleconference 2017-07-12

From VTKM
Jump to navigation Jump to search

Attendees: Ken Moreland (SNL), Alok Hota (SNL), Robert Maynard (Kitware), Sujin Philip (Kitware), David Lonie (Kitware), Mark Kim (ORNL), James Kress (ORNL), David Camp (LBL), Brent Lessley (UO/LBL), Matt Larsen (LLNL), Alberto Villarreal (Intel), Dave Pugmire (ORNL)

Robert has been working on gradients code. Providing gradient calculation within VTK-m. Has checked in a version but has some open questions. Has divergence and q-criterion. When writing out gradients of vectors (i.e. tensors) the ordering was different than VTK. Open issue on whether we should support our "C style."

Dave L. has been refactoring filters and how they manage point and cell data. Also simplified the internal implementation of ArrayHandle to remove code that was duplicated with different value types. This has head to major savings in executable size and compile time.

Sujin has been working on surface normals. Is now moving on to allowing virtual cell sets.

Brent and Matt have been resolving some compiler issues with Intel. Mostly this is mismatches between picking standard libraries and headers from different compilers.

Matt has tracked down some CMake configuration issues.

Matt has been working on VTK-h for the ALPINE project. There has been an issue where Marching Cubes failed for isovalues outside of the scalar range. Rob may have already checked in a fix last week.

Matt has a student that will work on line primitives for ray cast rendering classes. The line primitives will be rasterized.

James was having a couple of issues with rendering. With the 1D and 2D plots, when there is a Y axis with a small range, that axis is squashed down so you can't see anything. Sounds like a bug. Mark reports it is more about having too large of a ratio between X and Y axis.

James reports an issue with in situ integration stakeholders. EAVL had the ability to deliver a simple osmesa library that worked everywhere. We have some porting issues with EGL. Matt has suggested using the upcoming implementation of in-code rasterization of lines to get around the entire dependency.

Dave is working on the PICs branch to merge it in.

Brent reported some potential problems with ArrayHandle ReleaseResources on Xeon Phi. Rob suggested waiting for a fix from Dave L. that might address that.