Teleconference 2016-06-29

Attendees: Ken Moreland (SNL), Rob Maynard (Kitware), Alok Hota (SNL/UT@K), Matt Larsen (LLNL), Dave Pugmire (ORNL), Chris Sewell (LANL), Samual (LANL/UO), Tom Fogal (NVIDIA)

Rom working on VTK/VTK-m bridge. Working through some CUDA compile warnings.

Matt working on explicit grid volume rendering. Generalizing to work on any cell shape. Working through all of the crazy edge cases. Also helping Dave switch from old OpenGL to VBOs.

Dave working on updating OpenGL rendering. Also looking into an issue with OSMesa canvas, which is no longer working. Working on an ISAV paper for XGC project using EAVL (soon to be updated to VTK-m) and ADIOS.

Chris submitted paper on parallel contour trees. Samual is going to be working on worklet compression in VTK-m and move on to streaming algorithms.

Chris will be getting back to Legion/VTK-m scaling tests soon.

Tom looked into how memory fetches behave in virtual methods. If the methods are in the same translation unit and do not need to be virtual, then the compiler gets rid of them. If it actually does virtual methods, there is a slow down, but it still seems to be doing some coalesced fetching. The main overhead is calling the method through the function pointer.