Teleconference 2017-04-26

From VTKM
Jump to navigation Jump to search

Attendees: Ken Moreland (SNL), Abhishek Yenpure (UO), Rob Maynard (Kitware), Sujin Philip (Kitware), Dave Lonie (Kitware), Manish Mathai (UO), Matt Larsen (LLNL), Ollie Lo (LANL), Berk Geveci (Kitware), Tom Fogal (SNL), Will Schroder (Kitware), Pat Fasil (LANL)

Ollie put in merge request for segmented scan by key. Most issues addressed should be fixed. The only issue left is the implementation of exclusive scan by key does a key calculation twice, which slows performance improvements. Also need to add to comments in DeviceAdapterAlgorithm.h. There might also be room for improvement in the TBB implementation. Should add an issue in the issue tracker.

Pat working on extract subset grid.

Matt has been investigating some better quality heuristic-based search structures. Sort by key should enable that. Cleaning up radiography code.

Manish has been familiarizing himself with the VTK-m manual. Now working on implementing hierarchy based acceleration structures. Looking for an example of a recursive search structure.

Abhishek is also getting himself familiar with VTK-m. Now looking at implementing the integration methods that are available in the VisIt PICS code but not yet in VTK-m.

Rob is working on the virtual for-each he started at the hackathon. Has not look at performance, but from library size we have a 50% reduction. Suspect performance will be the same but need to run the benchmarks.

Has an open MR to separate out FunctorBase and WorkletBase and introduce a task based approach.

Dave has been working on reducing the amount of code generated for virtual tables.

Sujin has been working on implicit functions and has been working on issues being raised by Ken and Rob.