Teleconference 2016-03-09

From VTKM
Jump to: navigation, search

Attendees: Ken Moreland (SNL), Rob Maynard (Kitware), Matt Larsen (LLNL), Jeremy Meredith (ORNL), Chris Sewell (LANL)

Jeremy has made some progress on annotation. Dave and Rob have fixed the serious build issues.

Matt is going to send Chris an example he used to load a file and render.

Jeremy is on task to remove "order" from fields.

Chris went to C++ convention meeting last week. Some Thrust technology is going into the C++ standard. That could lead to some opportunities for VTK-m in the future.

Chris and Berk are going to do some VTK-m/Legion test runs.

Matt has got the atomic working. 32 and 64 bit integers add and swap atomics.

Matt trying to integrate VTK-m into Strawman. There are some compile issues. There is two back ends and some rendering code. It is taking 45 minutes to compile and producing binaries on the order of 400 MB. All of Matt's code should be statically typed, so something must be going wrong to observe this behavior. Rob volunteered to help debug what the issue is.

Rob has been working on the filters branch. Was ready last week, but there were some problems with compile time. Shaved off some time by pushing this into a metadata object. Unfortunately, this is not how we want to present the software interface. We may need to take a look at speeding up the resolution of dynamic types.

Chuck has the documentation things all synced up. That should go live tonight.

There were some changes in dashboard flags, and we are tickling some compiler bugs (or perhaps there is some VTK-m code that is invoking undefined behavior).

We talked briefly about the Version 1.0 ToDo List. Here is a snapshot of our current progress.

Feature Status
Settling of the core interface Other than little things and filters, the core interface is pretty well set.
  • Rob has a question about the "dimensionality" of the cell sets. That is supposed to be the topological dimensions of the cells. This would require all the cells having the same dimensionality, which is not really being honored. This is causing some problems. Jeremy and Rob need talk about this.
  • There is a similar question about the "order" on the fields. Right now, it is not used, and we should remove it. In the future, as we implement higher order cells, we will probably have to represent the interpolation as basis functions rather than
Standardization of "filters" Rob is ready to merge his first pass of filters into master.
Support worklet operations that are not one-to-one output Done
Basic file I/O Done.
Basic rendering Rendering is almost ready. Working out some build issues on the dashboard.
Clean up Doxygen Need to get documentation syncing nightly. Chuck looking into this.
User's Guide Behind. Trying to catch up. Making some progress.
Green dashboard
  • Still have problems with some CUDA dashboards
    • Timer issues should be fixed this week.
  • There have been some interop failures. - Rob thinks they are fixed.
    • Probably actually an X-windows issue. Chuck will look into this. Not currently at the top of his priority.
  • The new GCC 5.0 compiler is giving warnings in the testing code.
  • Some timeout issues on renar (e.g. https://open.cdash.org/viewTest.php?onlyfailed&buildid=4123644). Could be caused by JIT of CUDA kernels or might be because multiple things are running.
    • Some issues have been found and will be worked out.
Prototype VTK/ParaView/VisIt integration

(Technically not part of the release, but should indicate we have viable components.)

  • Eric Brugger has made some progress on integrating VTK-m into VisIt. He has an initial implementation and it is working pretty well so far. He has lately added further cell type support (particularly unstructured grids).
  • Rob working on VTK/VTK-m filter integration. Can now move data between VTK and VTK-m