Teleconference 2020-03-11
Attendees: Ken Moreland (SNL), Matt Larsen (LLNL), Abhishek Yenpure (UO), Ollie Lo (LANL), Steven Walton (UO), James Kress (ORNL), Dave Pugmire (ORNL), Nick Davis (SNL), Robert Maynard (Kitware), Vicente Bolea (Kitware)
ECP Updates
- FY21 P6 Activities draft complete
- FY20 Q2 is nearing completion
- We should be starting to complete our tasks for this quarter
- Are there any budget issues?
- Let's think about possible highlights
- App engagement
- Success with porting (e.g. run on Aurora)
- Engagement with other ST teams (e.g. using Kokkos)
ECP task updates
Roundtable
Matt has an issue with WorldCoordiantesToParametericCoordinates where it fails to converge and raises and error. This is an issue.
Matt has been feeling some pain with the fact that control portals can lock access to their ArrayHandle. Suggesting backing out of this behavior (or be able to disable it). We have 3 possible behaviors
- Control portals block invalid access to ArrayHandle while still in scope (current behavior)
- Control portals check with the ArrayHandle to ensure the memory is still available, report error if not
- Error could be assert
- Error could be exception
- Error could go to log
- Control portals ignore access and just assume data are still valid (old behavior)
Steven is working on higher order cells. He is having issues with cell shapes. The problem is that VTK-m (actually LCL) has a fixed list of cell types. Ken suggests looking at polygon cell type, which has a variable number of vertices. That could suggest how to implement an arbitrary order lagrangian cell.
