Teleconference 2017-10-25

Attendees: Ken Moreland (SNL), Matt Larsen (LLNL), Abhishek Yenpure (UO), James Kress (ORNL), Tom Otahal (SNL), Aaron Knoll (NVIDIA), Rob Maynard (Kitware), Berk Geveci (Kitware), Sujin Philip (Kitware), Oli Lo (LANL)

Planning beginning of November for release of version 1.1. Thinking the first full week. We are still discussing how/if to publish the User's Guide. Berk will ask around for what is needed for printing on demand, including rights, etc.

Issue with VTK-m and XCode 9. The most recent version of XCode 9 appears to be shipped with some C++17 containers, and there seems to be a change in how allocators are specified in containers. One of the tests does not build because we test VTK-m allocators with stl containers. This is not a feature anyone uses, so we will probably comment out the test.

Abishek has been running performance metrics on streamlines on regular grid. Going to try unstructured grids soon using some new cell locators that are coming on line. He is going to compare different search structures as they come online.

James and Dave Pugmire have been working on the ECP demo. Moving to final testing.

Tom has done some more work on the radix sort for TBB. Has an implementation finished that is based on an OpenMP implementation made a while ago by Intel.

Aaron is working on getting the streamlines algorithm accessible in VTK and ParaView.

Rob has been supporting folks like Aaron and Allie who are looking at performance information.

Sujin working on fixing a position issue in the cell locator. It should be fixed and merged into master.

Sujin is back to working on switching from our faux virtual methods to real virtual methods. Currently running into an issue with the Sphere implicit function.

Allie has been looking at performance on CPU backends during the last few weeks. Copy operations have been significantly improved.

Oli submitted a merge request for point locator that is now merged into the main branch.

Rob submitted a GTC submission for VTK-m. We are getting all set for the Intel HPC developer connection.