Sandia-Kitware Summit, 2009

From ParaQ Wiki
Jump to navigationJump to search
Tuesday, April 7
Time Length Item Who
9:00 0.5 hr Coffee and "A man walks into a bar..." jokes All
9:30 1 hr Multicore Shepherd
10:30 0.5 hr Charting EKG Wylie
11:00 1 hr Annotation Shead
12:00 0.5 hr Working Lunch All
12:30 1 hr Statistics Tutorial Thompson
1:30 1 hr Plugins
  • Loading Management
  • Subtractive Plugins
  • Controlling UI
  • Client/Server Plugins
  • Branding for ParaView
  • Bringing OverView into ParaView fold
Shead
2:30 0.5 hr OverView Release Wylie
3:00 0.5 hr Qt
  • 4.5 Migration
  • LGPL
  • File segregation
  • Font migration
Wylie
3:30 0.5 hr ParaView 3.6 status/release schedule Moreland
4:00 1 hr MARS
  • Fragment generation in exodus
  • Crater identification
  • Visualizing Tracers
  • Milestone status
  • Announcements of feature identification capability
  • Export Control
Ice
5:00 0.5 hr CSCS Particle Stuff Moreland
Wednesday, April 8
Time Length Item Who
9:00 0.5 hr Coffee and "A man walks into a bar..." jokes All
9:30 1 hr New View/Representation Heirarchy Wylie, Baumes
10:30 0.5 hr Time Cleanup
  • Time variable range loading (bug 8596)
  • Issue with filters making inputs temporal
Ayachit
11:00 1 hr In-Situ Fabian, Geveci, Thompson, Moreland
12:00 0.5 hr Working Lunch All
12:30 1.5 hr Informatics Strategery Schroeder, Rogers
2:00 1.5 hr Array Refactoring Shead
3:30 0.5 hr Lookmarks: Major push or jettison? Moreland
4:00 --- Miscellaneous ---
Breakout Sessions
Item Who
ParaView 3.8 Deliverables Moreland, Ayachit, Ice

Action Items

  • QtItemView should go away. -Brian
  • Depricate vtkQtChartRepresentation. -Utkarsh
  • Rename vtkQtChartViewBase to vtkQtChartView. -Utkarsh
  • Add a set representation from input connection convenience function. This function may not always work on views with more complicated inputs (such as hierarchical graph view). -Pat
  • Follow up with looking at Tim's Python wrappings for Qt views and to add wrappings for Chart views. -Pat
  • Make ParaView startup be a branding startup. -Utkarsh
  • Make sure the Kitware as a whole is OK with removal of Qt file segregation. -Berk
    • Move CMake Qt option out of GUI support. - Utkarsh
  • Look into performance of tree ring with VTK directory structure. -Jason
  • Make vtkAbstractArray and subclasses use vtkArray internally. -Kitware (1 month from now)

Plugins

It is a general consensus that it is beneficial to all if the ParaView and OverView main windows merged into one. This would work by making this main window be quite blank. The window would then load in a set of plugins, defined by the branding, which then populate the GUI. In this mode, subtractive plugins become much less important, which is good as this is a pretty dangerous concept.

We identified an open issue where ParaView and OverView have different selection mechanisms. We want to be able to mix sci vis and info vis, so we should find a way to unify the selections. We might be able to capture the difference as an autostart plugin, but that does not help much with respect to mixing sci vis to info vis.

Qt

VTK will officially move to Qt 4.5 after the ParaView 3.6 release. From then we will enforce Qt 4.5 so that we can ensure the LGPL.

Once we adopt the LGPL version of Qt 4.5, the file segregation of Qt classes will go away. There will still be GUI support directory, but databases go to IO, rendering to Rendering, etc.

This should solve the font migration stuff.

ParaView 3.6 Release

Plan: work on bugs for the next six weeks then branch. Release builds in 2-3 weeks from then.

MARS

Charles volunteered to do the fragment generation for Exodus-type data. This basically boils down creating a connectivity filter that works in parallel and then performing statistics (integrating data over the extracted components). This has a higher priority than CTH fragments on the flat mesh.

Crater identification. Kitware needs to give Sandia the code and Sandia needs to try it out.

Charles had the thought that what would be most useful to the user is the ability to perform fragment extraction based on the tracers. That is, we will extract the fragments that contain a tracer, and order them in the output in the same order of the tracer. Assuming the tracer stay in the same fragment, this provides us a simple way to trace the fragment. We need talk to the users to see if that is useful.

We will probably not complete the in/out operation of a tracer in a fragment/crater.

We will draw tracers. Nathan will write the coordinates of the tracer into the spyplot reader global field output. Ken will write a filter to extract the tracers and field information as a poly data and also draw them.

Milestone status is good.

We are considering having Sandia host a repository that could hold export control data. We might ask Bill Goldman if we can host export control on software.sandia.gov. If not, then we can try gforge, which is explicitly hosted on the SRN.

Array Refactoring

  • Performance comparisons between vtkDataArray & vtkDenseArray (Tim Shead)
  • Modify vtkAbstractArray to provide access to an (optional) underlying vtkArray (Tim Shead)
  • Modify vtkDataArray to use vtkDenseArray as an implementation (Andy Bauer)
  • Update vtkFieldData and vtkDataSetAttributes APIs to support vtkArray directly (Andy Bauer)
  • Deprecate vtkAbstractArray and its derivatives (Brian Wylie)

Lookmarks

If lookmarks went away, it would help if it was replaced with:

  • Ability to save out Python script
    • Ability to run script as macros.
  • Ability to save/load camera state separately.