Teleconference 2016-09-14

Jump to navigation Jump to search

Attendees: Ken Moreland (SNL), Rob Maynard (Kitware), James Kress (ORNL), Chris Sewell (LANL), Sam Li (LANL), Dave Pugmire (ORNL)

Mark Kim officially started at ORNL on Monday.

Mark has a GLEW and OSMesa issue. GLEW in Linux will call GLX_context, which is not available on OSMesa. This is a known problem with GLEW. Berk will ask Markus Hanwell and other VTK developers about it.

Pat has the VTK-m version of the contour tree algorithm working at least with serial backend.

Work continues on streaming device adapter algorithms.

Sam had a weird error with the timer on the CUDA backend. The problem was caused by the case where if you try to invoke the CUDA kernel with an input array of zero elements, that causes the timer to throw an error. Sam will attempt to fix the CUDA schedule algorithm to just return of the schedule is performed on 0 instances.

Rob continues transition away from Boost.

Rob looking more on virtual array handle. It now compiles (without CUDA), but tests are running really slow.