Teleconference 2024-06-04
Attendees: Ken Moreland (ORNL), Jeff Amstutz (Nvidia), Mark Bolstad (SNL), Vicente Bolea (Kitware), Silvio Rizzi (ANL)
Viskores / HPSF
We have completed the transition from VTK-m to Viskores. We have released the final version of VTK-m (although there can be patch versions) and the first version of Viskores. These versions are equivalent. There is a script to help convert code from VTK-m to Viskores. Spack package is also available.
As we move away from VTK-m, we will need to transition our communications. Right now we are still using things like the VTK-m mailing list. We plan to move to the Viskores discussions in GitHub. Meeting notes and other things may change. More information will be forthcoming. (Sorry, this is not a big priority right now.)
There was an HPSF Conference in Chicago May 5 - 8. Ken gave a brief update on Viskores. Not a lot to report back on. I had a brief conversation with Barton Fiske who had some interest in Viskores. This was a generally useful event to connect with other HPC developers. We may try to schedule some Viskores activities in the future if there is enough interest.
OASIS Deliverables
The following items are proposed for FY25 (which will end at the end of the fiscal year)
- Address requests for support, improvement, and new features by current and potential applications such as XGC and WarpX as well as other users of ParaView, VisIt, and Ascent.
- Contouring of polygons (such as for contouring slices) added and merged. (Can we get a highlight out of that?)
- Strengthen operation of Viskores on Aurora.
- Improve documentation with automatic generation and web deployment.
- Integrate into High Performance Software Foundation (HPSF) as Viskores.
- Improve integration between Viskores and VTK.
In the future, we will have to think about how to highlight VTK-m/Viskores. Our most visible deployment is usually through Ascent, but we don't want to take Ascent's thunder. I think Ascent gets its main funding through a different stream, so perhaps we could submit complementary highlights through SciDAC and NNSA.
Roundtable
Jeff: The VTK-m graph thing he made a while ago. Should that become part of Viskores proper? This contribution would be specifically graph hookup of Viskores components.
Silvio tried building ParaView with SYCL with CMake settings that Abhishek gave him. The build works, but testing is forthcoming. There was a recent change to Aurora that broke a lot of stuff. Need to circle back and check things.
Nicole has mesh with structured cells and explicit coordinates. External faces is throwing exceptions about a bad cast and call. Ascent has a workaround, but Viskores should fix this issue.
Ascent getting weird ray culling issue. After some investigation, Nicole was able to reproduce at least a similar issue when rendering straight down the y axis.
