Teleconference 2018-05-23

From VTKM
Jump to: navigation, search

Attendees: Ken Moreland (SNL), Dave Pugmire (ORNL), Matt Larson (LLNL), Matt Letter (SNL), Ollie Lo (LANL), Robert Maynard (Kitware), Allie Vecanti (Kitware), Sujin Philip (Kitware), Shreeraj Jadhav (Kitware), Haocheng Liu (Kitware), James Kress (ORNL)

ECP miscellany:

  • We need to have "KPP-3" milestones. Basically these are metrics showing the overall value of VTK-m to ECP
    • Currently have count how many VTK-m algorithms in tools (we are distributing our stuff) and how many ECP software products are using VTK-m.
    • The latter metric is problematic. Will probably be static.
    • Jim suggested having a metric on the number of ECP platforms we port to.
  • Updates to milestones should be done
    • Asked for stuff we already made 1.5 years ago
  • Had to defend our spending
    • Have a new spend plan on confluence

Ollie working on virtual objects for point locators.

Allie working on OpenMP backend. Looks like it is working as well as TBB. Also tracking down a weird CUDA bug where we overrun the stack.

Rob has fixed some bugs in the CMake files. Getting back into virtual to how to wrap all arrays in a virtual method as well as cell set.

We added a changelog directory. When you create a nontrivial change, add a changelog file to describe what you did.

Sujin looking into serializing data structures so that they can be transferred across MPI or other distributed memory systems.

New Ph.D. student at Kitware: Shreeraj. Working on common API for cell set. Adding a way to release execution resources for cell sets. Added templated Pi functions to support Pi at different floating point widths.

Mark Kim wrapping up rendering milestone.

Initial point movement work has been implemented.