1. 05 Sep, 2018 2 commits
  2. 04 Sep, 2018 1 commit
  3. 30 Aug, 2018 4 commits
  4. 29 Aug, 2018 2 commits
    • David Thompson's avatar
      The batch system ate my homework. · ab7afe6e
      David Thompson authored
      Deal with logging problems.
      
      + Log event names must be pointers to strings that live for the
        length of the log.
      + Static linking of the timer and senseiCore libraries means
        that each link to the timer library may result in a separate
        set of library-global static variables. So, when finalizing
        the configurable analysis, print the event log for `libsensei`.
        This may cause duplicate printouts of events, but it allows
        us to proceed with timings.
      + Fix compilation on some systems that need `<algorithm>`.
      ab7afe6e
    • David Thompson's avatar
      Get CDF working as a configurable analysis. · 13804aed
      David Thompson authored
      This eliminates all but token dependence on VTKm in favor of
      VTK's vtkSortDataArray class (which uses vtkSMPTools).
      However, it has the advantage of handling different types of
      data arrays, whereas the previous VTK-m worklet insisted on
      being provided doubles.
      13804aed
  5. 27 Aug, 2018 1 commit
  6. 24 Aug, 2018 1 commit
  7. 23 Aug, 2018 1 commit
    • David Thompson's avatar
      Add the ability to time each configurable analysis. · 0ec79055
      David Thompson authored
      Timing is disabled by default.
      To enable timings, add the following to the root node of your
      configurable-analysis XML:
      
      ```xml
      <sensei timing="true" summarize-timing="true" timing-rank-modulus="1">
       ...
      </sensei>
      ```
      
      + The value of the `timing` attribute is evaluated as a boolean
        which defaults to false if the attribute is not present.
        When true, timings are reported.
      + The `summarize-timing` attribute can be used to
        set how timing information is presented.
      + The `timing-rank-modulus` is used to determine which
        ranks of the MPI job will print their timing information.
      0ec79055
  8. 22 Aug, 2018 1 commit
  9. 21 Aug, 2018 1 commit
    • David Thompson's avatar
      Handle VTK-only build with cinema helper. · 6aebe1c8
      David Thompson authored
      The Cinema helper needs to expose different functionality depending
      on what VTK libraries are requested/available.
      
      + When the Cinema helper is expected to generate images, VTK's rendering
        libraries must be available (either through Catalyst or VTK).
      + Other functionality is specific to Catalyst (which provides Ice-T
        parallel, multipass rendering).
      6aebe1c8
  10. 13 Aug, 2018 1 commit
  11. 11 Aug, 2018 2 commits
  12. 10 Aug, 2018 1 commit
    • David Thompson's avatar
      Modernize sensei's VTKm support. · 1cc543d2
      David Thompson authored
      + Do not require a modern VTKm when analyses (such as VTKmContourAnalysis)
        use the vtkAcceleratorsVtkm library in VTK rather than VTKm directly.
      + Make usage of vtkAcceleratorsVtkm a build option (ENABLE_VTK_ACCELERATORS).
      + Add a direct VTKm analysis adaptor: VTKmVolumeReductionAnalysis, which
        performs a Haar wavelet transform and keeps only the low-frequency
        component in order to reduce image-data down to a manageable size.
      + Require a modern VTKm for direct VTKm analysis adaptors.
        Make direct usage of VTK-m a build option (ENABLE_VTKM).
      + Be explicit about installing header files. Using globbing is discouraged
        as it slows CMake down and can produce counterintuitive results.
      + Add a CinemaHelper class used by the VTKmVolumeReductionAnalysis to
        write out data for post-hoc visualization using cinema/arctic-viewer.
      1cc543d2
  13. 06 Aug, 2018 1 commit
  14. 18 Jul, 2018 3 commits
  15. 17 Jul, 2018 1 commit
  16. 12 Jul, 2018 1 commit
  17. 10 Jul, 2018 1 commit
  18. 27 Jun, 2018 4 commits
  19. 01 Jun, 2018 1 commit
    • Burlen Loring's avatar
      unique MPI communication space etc · da17b2f7
      Burlen Loring authored
      * normalize MPI communicator handling by putting some default
        code in DataAdaptor and AnalysisAdaptor base classes. The
        defaults are to use isolate our communciation in a duplicate
        of COMM_WORLD. Adaptors needing to do MPI communicaiton should
        access the communicator by calling GetCommunicator.
      * report errors from ConfigurableAnalysis::Execute and ::Finalize
      * fix a build issue with new VTK AMR writer.
      * clean up warning in parallel 3d miniapp
      
      squash me -- amr reader build fixes"
      da17b2f7
  20. 25 May, 2018 2 commits
  21. 27 Apr, 2018 1 commit
  22. 22 Apr, 2018 1 commit
  23. 16 Apr, 2018 6 commits