Skip to content

V
visit
Switch branch/tag
  1. 24 Nov, 2015 1 commit
  3. 19 Nov, 2015 2 commits
    • camp's avatar
      Internallauncher · 52528801
      camp authored 
          I added a function to return the correct argument for the "process per node"
flag for Open MPI and MPICH. This change affects the mpirun and mpiexe commands.
If your system has a different argument then you can override this function.
This argument is only used if the user has set VisIt's -nn flag.

    I have changed the sbatch job submitter to be like qsub. This was done for
NERSC new Cori system. Now it has many functions that can be overriden to support
any system using sbatch. Also added a new job launcher srun.

    Added the hosttoip variable in MainLauncher to convert host name to IP address.
Had a problem that compute nodes could not convert host name to IP address. By default
this is turned off. You need to set hosttoip=True in your MainLauncher override.
See NERSC custom launcher for an example.

NERSC customlauncher
    Remove references to Carver. That system is gone. Added sbatch support, also 
turned on the hosttoip flag to translate the hostname to an IP address.

gui/QvisHostProfileWindow.C
    Added sbatch/srun option in the list of launchers.

config-site/cori.nersc.gov.cmake
    Added site file for NERSC's Cori system.

Removed NERSC's Carver site config file. System has been retired.



git-svn-id: http://visit.ilight.com/svn/visit/trunk/src@27736 18c085ea-50e0-402c-830e-de6fd14e8384
      52528801
    • brugger's avatar
      Update the revision number of the last test suite pass. · eb7102e0
      brugger authored 
      git-svn-id: http://visit.ilight.com/svn/visit/trunk/src@27735 18c085ea-50e0-402c-830e-de6fd14e8384
      eb7102e0
  5. 18 Nov, 2015 3 commits
  7. 17 Nov, 2015 1 commit
  9. 16 Nov, 2015 1 commit
  11. 15 Nov, 2015 1 commit
  13. 13 Nov, 2015 1 commit
  15. 12 Nov, 2015 3 commits
  17. 11 Nov, 2015 3 commits
  19. 10 Nov, 2015 3 commits
  21. 09 Nov, 2015 3 commits
  23. 08 Nov, 2015 1 commit
  25. 07 Nov, 2015 1 commit
  27. 06 Nov, 2015 3 commits
  29. 05 Nov, 2015 3 commits
  31. 04 Nov, 2015 1 commit
  33. 03 Nov, 2015 3 commits
  35. 02 Nov, 2015 2 commits
  37. 01 Nov, 2015 1 commit
  39. 31 Oct, 2015 1 commit
  41. 30 Oct, 2015 1 commit
  43. 28 Oct, 2015 1 commit
    • loring's avatar
      depth sort update · eaf68626
      loring authored 
      this patch modernizes VTK's depth sorting code. The following improvements
were made:

* transfrom GetCell to GetCellPoints. Building the cells one by one
  is expensive and we only need points to determine the depth. We
  can also then access the points in place.

* transform qsort to std::sort. Comparisons in std::sort get inlined
  so std::sort is much faster. I also reduced the overhead of swaps
  by using a functor so that only cell id is swapped.

* use templates to deal with point types instead of going through
  virtual GetValue API that converts them to double.

* restructure depth computations so that they can be vectorized by
  the compiler.

* I added a Cxx test to improve test coverage of the various sorting
  modes supported by this class.

In tests the improved class is 2 x faster when sorting by first point
and cell bounds, and 1.5 x faster when sorting by parametric center.
The tests made use of a 900M cell iso-surface computed from a cosmology
simulation, and gcc 4.9.2 with -DNDEBUG -Ofast -march=native -mavx


git-svn-id: http://visit.ilight.com/svn/visit/trunk/src@27619 18c085ea-50e0-402c-830e-de6fd14e8384
      eaf68626