Teleconference 2017-03-22
Attendees: Ken Moreland (SNL), Matt Larsen (LLNL), Dave Pugmire (ORNL), Rob Maynard (Kitware), Sujin (Kitware), James Kress (ORNL), Hank Childs (UO), Tom Otahal (SNL), Brent Lesly (UO), David Rogers (LANL), Mark Kim (ORNL), Ollie Lo (LANL)
Matt is making lots of changes to raytracing code. Is working on bringing his changes up to master. One change is the ability to add arbitrary buffers/layers to rays (i.e. not just color).
Matt working on DIY sitting on top of VTK-m. Making DIY serialization/deserialization to pass around VTK-m data sets. This is preliminary work for VTK-h.
Discussion of VTK-h: h is for hybrid parallelism. Doing distributed memory things. Relies heavily on VTK-m. Proposed as part of the ECP ALPINE project. DIY will likely play a major role for the MPI piece.
Dave P. looking into multiblock implementation in VTK-m. Considering making a good integration point for the ECP/VTK-m milestone.
For particle advection, talking about duplicating the PIC system in VisIt. Dave P. is throwing together some prototype code for experimentation.
James Kress transitioning old EAVL code in Xlotl to VTK-m. Planning to transition XGC later.
Mark working on an ADIOS/VTK/CFP in situ thing with Dave P. Working on a memory model hierarchy with ZFP compression in ADIOS.
Rob as added MC multi-isocontour support. You can create a single output with several isosurfaces. That is working and all tests are passing. When trying to add this to VTK, some dashboards started to fail.
Sujin implementing dynamic implicit functions that you can pass to functions without templating.
Ollie merged unified memory request.
Brent wrapping up maximal clique work for large scale graphs in VTK-m. Been working on large memory nodes on Pittsburgh supercomputer center.
Tom O. working on vectorization with Alberto.
