Teleconference 2017-01-25

From VTKM
Jump to navigation Jump to search

Attendees: Ken Moreland (SNL), Tom Otahal (SNL), Alberto Villarreal (Intel), Mark Kim (ORNL), Rob Maynard (Kitware), Dave Lonie (Kitware), James Kress (ORNL), Matt Larsen (LLNL), David Pugmire (ORNL), Berk Geveci (Kitware), Ollie Lo (LANL), David Rogers (LANL)

Ken is working on generalizing the external faces filter. It is slower than the existing version, but adds some features that are necessary for production. Matt was going to send Ken some data that was used...

Tom writing a vector magnitude filter and generalizing the functionality. Working through some compiler warnings.

Running into issues with the raytracer on implicit surfaces. Works on a few implicit surfaces but not for 100s.

James transition some code from EAVL to VTK-m. This comes from funding on the PSI SciDAC.

Dave P. has a branch for checking in some data sets once the git LFS is available.

Dave L. did some work on virtual CUDA. The use of function pointers should work pretty well. There was a 70% overhead for really bad (very small) block size, but 20% for most cases. Overhead goes down as array sizes go up. Last thing left to test is implicit arrays.

Rob has been working on git LFS. There is a snag of using gitlab with LFS. Gitlab has a bug for the https transport. Rob is working with the gitlab developers.

Rob has been working on the CMake code inside of VTK-m. Can't use private includes with cuda, which is a pain.

There is a dashboard now running VTK-m in VTK. All devices are available.

Matt is continuing his refactor of the raytracing stuff. Making the internal raytracing classes available as rendering proxies so that folks can reuse that code for other work. Good for tracing rays, which can be used for things other than rendering or modifying how the rendering works.