Lines of Code
[root]/Packages/Uintah/CCA/Components/Regridder
| Author | Changes | Lines of Code | Lines per Change |
|---|---|---|---|
| Totals | 75 (100.0%) | 1534 (100.0%) | 20.4 |
| luitjens | 29 (38.7%) | 1076 (70.1%) | 37.1 |
| jas | 43 (57.3%) | 450 (29.3%) | 10.4 |
| dav | 2 (2.7%) | 5 (0.3%) | 2.5 |
| harman | 1 (1.3%) | 3 (0.2%) | 3.0 |
Remove Packages/
0 lines of code changed in 17 files:
Added a few debug streams. No longer profile fine tasks seperately from regular tasks.
0 lines of code changed in 2 files:
Use Patch::getFinestRegionsOnPatch()
0 lines of code changed in 2 files:
Updates to Region::difference.
1) Fixed a bug where the function did not properly handle subtracting an empty set
2) No longer includes zero sized regions in the results
3) Changed deque to a vector as we didn't use any of the deque features and vetors are more widely understood.
5 lines of code changed in 1 file:
Turned on .ups validation... You need to update your .ups files and
the ups_spec.xml file now! If your problem specification is not
valid, sus will die (with an error message) immediately. You can
validate an .ups files with "sus -validate filename.ups".
M Core/Containers/StringUtil.h
M Core/Containers/StringUtil.cc
Utility function to print out a vector<string> as a single ',' separated string.
M Packages/Uintah/CCA/Components/MPM/PetscSolver.cc
* Comment.
M Packages/Uintah/Core/GeometryPiece/GeometryPieceFactory.cc
M Packages/Uintah/CCA/Components/SimulationController/SimulationController.cc
* Removed setting of writeMessages.
M Packages/Uintah/CCA/Components/Parent/ComponentFactory.h
M Packages/Uintah/CCA/Components/Parent/ComponentFactory.cc
* Don't pass in the name of (the type of) a component anymore... It is read from the .ups file now.
M Packages/Uintah/CCA/Components/Parent/Switcher.h
* Cosmetics
M Packages/Uintah/CCA/Components/Parent/Switcher.cc
* It is possible that I broke this... I think it is ok, but I'm going
to get this commit in so that the RT can take a look.
* Cosmetics
* Organize #includes.
* The ProblemSpec no longer comes from a port... Just hold onto it when Switcher is initialized.
M Packages/Uintah/CCA/Components/LoadBalancers/DynamicLoadBalancer.h
* Cosmetics - whitespace is your friend. :)
M Packages/Uintah/CCA/Components/LoadBalancers/DynamicLoadBalancer.cc
* Don't need two steps anymore to read a problem spec.
M Packages/Uintah/CCA/Components/DataArchiver/DataArchiver.cc
* Don't need two steps anymore to read a problem spec.
M Packages/Uintah/CCA/Components/Solvers/SolverFactory.cc
* <Solver>
<type>
replaced with:
<Solver type="">
M Packages/Uintah/CCA/Components/ProblemSpecification/ProblemSpecReader.h
* Major updates to ProblemSpec validation.
** Tracking the actual file that specific tags come from to better report errors to user.
** Don't 'create' a PSReader with a filename, just call readInputFile() with the filename.
Might should make readInputFile a static function, but haven't yet as I'm not sure
if it is used multiple times for the same ProbSpec.
** Prob Spec Reader now contains a list of all (xml) files that were read to create the PS.
* Functions used to pass in the tag (parent), but since they are part of the tag class, don't need
to do this. It is implicitly done (parent == this).
* Moved the Tag/Attribute structs out of the ProblemSpecReader class as they aren't really
a part of it, and are not needed anywhere but in ProblemSpecReader.cc.
* Remove #include sgi stuff.
M Packages/Uintah/CCA/Components/ProblemSpecification/ProblemSpecReader.cc
* Display filename and line number when errors are found.
** All tags are added to the main (top level) xml 'doc'ument. Because of this, the filename
of sub-docs was lost. I am catching this and using the *_private (part of an xml node)
to hold onto this info. I think this is ok as the xml documentation says that it is for
application use.
* Added need_applies_to and children requirements parsing.
* Fixed a couple of edge condition errors that could slip through the parsing.
* Consolidate the file name validation. First look for the file specified directly,
then look for it as a relative offest from its parents location,
then look for it as an absolute offset from where the program was run.
I think this covers all the cases that people use and that it makes sense
to do it this way.
M Packages/Uintah/CCA/Components/Arches/Arches.cc
* Made 'whichMMS' an attribute (instead of a separate <Tag>).
M Packages/Uintah/CCA/Components/Regridder/RegridderFactory.cc
* Changed <type> to be an attribute of <Regridder type="">.
M Packages/Uintah/CCA/Ports/ProblemSpecInterface.h
* Pass filename into readInputFile().
* Cosmetics
M Packages/Uintah/scripts/regression_tester
* Allow 'inputs' to be INPUTS to be local as well as in relative src location.
M Packages/Uintah/Core/DataArchive/DataArchive.cc
* One stop readInputFile().
* Catch edge error condition. (d_filebase == "")
M Packages/Uintah/Core/ProblemSpec/ProblemSpec.h
* Removed doWrite... as it wasn't (really) used anywhere, and I'm not sure what its purpose was.
* Cosmetics - whitespace is your friend
M Packages/Uintah/Core/ProblemSpec/ProblemSpec.cc
* Cosmetics - use '{}' and whitespace.
* removed d_write
M Packages/Uintah/StandAlone/tools/compare_mms/compare_mms.cc
M Packages/Uintah/StandAlone/tools/graphview/GV_TaskGraph.cc
M Packages/Uintah/StandAlone/slb.cc
M Packages/Uintah/StandAlone/restart_merger.cc
* One stop readInputFile()
M Packages/Uintah/StandAlone/tools/pfs/pfs.cc
M Packages/Uintah/StandAlone/tools/pfs/pfs2.cc
* One stop readInputFile()
* Removed writeMessages()... don't think it is used anywhere...
M Packages/Uintah/StandAlone/sus.cc
* Clean up 'usage'.
* No more "-mpm" or "-arches", etc... Specify in .ups file.
* FinalizeManager (finalize MPI) was needed in a number of error cases.
M Packages/Uintah/StandAlone/inputs/ups_spec.xml
Lots of updates... Needs a lot more work from those in the know.
Please fix all the FIXMEs. If a parameter is optional, please specify
why and/or its default. Also, make as many things 'REQUIRED' as
possible. Update the children and need_applies_to flags too.
Please put <tags> in ALPHABETICAL order... this will make it much
easier to find and verify them! Try to organize the large <tags> in
sections (as they apply to need_applies_to), but keep each section
alphabetical.
M Packages/Uintah/StandAlone/inputs/*/*.ups
* Updates to be closer to spec.
5 lines of code changed in 2 files:
Changed the volume threshold to 5%.
Optimized the number of global operators. Now only performs 1 all reduce per tiling attempt instead of 1 all gather. The all gather is only done on the final tile set.
0 lines of code changed in 2 files:
Restructured the loops. Changed the growth tolerance to be based off the minimum volume possible instead of the volume of the last tile set.
80 lines of code changed in 1 file:
Moved the basic tiling code into a fuction so it can be used elsewhere.
46 lines of code changed in 3 files:
Minor change to output
1 lines of code changed in 1 file:
Fixed a bug where the wrong numbers of patches were used.
54 lines of code changed in 1 file:
Don't increase the tile size if doing so would signicantly increase the volume of the patch set.
19 lines of code changed in 1 file:
Reduced spam when changing tile size
9 lines of code changed in 1 file:
Don't allow tiles to become larger than the coarser tiles. (This forces tiles to be aligned).
14 lines of code changed in 1 file:
Dynamically resize tiles by factors of 2 depending on if a target number of patches is reached. The default target is 4 patches per processor.
132 lines of code changed in 1 file:
Initial commit of a tiled regridder which is similar to the hierachical regridder without all the performance issues. This regridder should be very fast at making grids and requires very little mpi communication (1 gather and 1 gatherv per level). However, this regridder will tend to make more patches than the BNRRegridder. This should not be a problem as most of our performance/memory problems related to the numbers of patches have been resolved.
To use this regridder specify the type as "Tiled" and specify the <min_patch_size> the same way you would for the BNRRegridder.
In the near future I will dynamically resize the finer level patch sizes according to the target number of patches per processor.
621 lines of code changed in 5 files:
Removed the check for cell deletion dilation
2 lines of code changed in 1 file:
Undid 42948. min boundary cells are used and they are the number of cells that must exist between the boundary of two different levels
5 lines of code changed in 1 file:
Add the MIT License to every *.cc, *.h,*.F, and sub.mk file.
450 lines of code changed in 26 files:
problemSetup_BulletProofing()
- remove min_boundary_cells & cell_deletion_dilation test from bulletproofing
3 lines of code changed in 1 file:
don't do the grid checksums when running in serial.
4 lines of code changed in 1 file:
(3 more)