Fixed a memory curruption bug in collectParticlesForRegrid
12 lines of code changed in 1 file:
Don't collect particles unless it is enabled.
4 lines of code changed in 1 file:
Readded the collectParticlesOnRegrid code.
161 lines of code changed in 2 files:
Adding the option to re-arrange the order of operations in the MPM code. This
option is currently compile time configurable, by swapping the #defines on USL
and MOM_FORM near the top of SerialMPM.cc As committed, this shouldn't change
answers.
584 lines of code changed in 2 files:
Including Heattransfer.cc
2 lines of code changed in 1 file:
Working version
14 lines of code changed in 1 file:
Devloatization model using particle temperature
53 lines of code changed in 2 files:
Including heat transfer model for particles
589 lines of code changed in 3 files:
Get rid of the half-a$$ed use of d_ref_temp.
1 lines of code changed in 2 files:
Remove reference_temperature as it is evidently unused.
0 lines of code changed in 2 files:
Put the cellCost and particleCost parameters in the correct order.
1 lines of code changed in 1 file:
Resize the cost vector.
2 lines of code changed in 1 file:
added shortcut for fsspeed
3 lines of code changed in 1 file:
pulled a weed and added some comments.
2 lines of code changed in 1 file:
Added a simple cost model forecaster.
Refactored and cleaned up some more of the DLB code that deals with forecasting.
179 lines of code changed in 7 files:
updated commented out debugging statements to match the correct syntax
2 lines of code changed in 1 file:
Initial refactoring of the cost forecasting algorithms in the load balancer.
Created new forecasting base class. Made old profiler inherit from this base class.
More changes incomming soon.
102 lines of code changed in 6 files:
Pulled a weed
0 lines of code changed in 1 file:
Minor refactoring.
14 lines of code changed in 1 file:
Use hash_multimap again. Returned the hash and == functions to their fast implementations. The old problems appeared to be with scrubbing which has been disabled for the simulations that were having problems.
6 lines of code changed in 2 files:
Disable scrubbing for AMR ice. This appears to cause the missing patches in getRegion problem.
Added comments explaining why scrubbing is disabled for those problems.
3 lines of code changed in 1 file:
Fixed some memory leaks
5 lines of code changed in 1 file:
Deleted some more unused code. Refactored some code to help clean up
the DLB.
21 lines of code changed in 2 files:
Removed unused code that was cluttering the DLB.
20 lines of code changed in 2 files:
No longer carry forward NC_CCweight in MPMICE OR in the contact models, just do it in
SerialMPM. Also, the NC_CCweight stuff from MPMICE::refine has been moved to
SerialMPM::refine.
52 lines of code changed in 4 files:
Committing a large set of fixes for the DQMOM and LU classes.
LU now (correctly) implements iterative refinement. It also has some added functionality, with respect to residual calculation, matrix operations, etc.
204 lines of code changed in 4 files:
Don't initialize NC_CCweight, as this is now done in MPM. This will likely
still break when MPMICE is used with FrictionContact as both are trying to carry
this variable forward. Monday...
13 lines of code changed in 1 file:
Fix memory leak (I hope) created by the changes in the contact models.
12 lines of code changed in 2 files:
Same treatment as just committed to FrictionContact. Namely, use NC_CCweights
in determining how full a node is. Also, in both this and FC, get rid of the
unused NumNearParticles junk.
29 lines of code changed in 1 file:
Add initial calculation of NC_CCweights to SerialMPM. In explicit MPM, this is currently only used (for now) in FrictionContact.cc, which carries it forward
each timestep, otherwise it will just get scrubbed.
This shouldn't change the answers.
45 lines of code changed in 2 files:
Added Arches/ChemMix/TabProps.
0 lines of code changed in 2 files:
Add noisyDensityGuess variable.
0 lines of code changed in 2 files:
added a cstyle file output and some flushing to try and avoid write caching
0 lines of code changed in 2 files:
Remove TabProps directory.
1 lines of code changed in 1 file:
Update the release files to include latest changes.
19 lines of code changed in 3 files:
Move documentation pdfs to toplevel of doc/. Remove all other files/directories.
0 lines of code changed in 2 files:
Remove the Angio component from the release.
2 lines of code changed in 9 files:
Made the velocity now a function of the dynamic weights and weighted lengths rather than just the initial values.
4 lines of code changed in 1 file:
flush the write
1 lines of code changed in 1 file:
Time file deletion
13 lines of code changed in 1 file:
Added some error checking.
4 lines of code changed in 1 file:
Added a tool to test the parallel throughput of a filesystem.
To use:
mpirun -np X StandAlone/tools/fspeed 500MB
The filesize is to total amount of data to be written. Each processor will output an equal portion of the total data at the same time.
You can use B, MB, or GB as the unit.
140 lines of code changed in 3 files:
Dynamically allocate the cost profiler so it doesn't take up memory if it is not being used.
38 lines of code changed in 2 files:
Disabled scrubbing when using the DLB.
1 lines of code changed in 1 file:
Don't disable collect particles when profiling.
0 lines of code changed in 1 file:
Use a multimap instead of a hash_multimap to test if this fixes the RT problem.
1 lines of code changed in 1 file:
Initialize h
2 lines of code changed in 1 file:
use variable lable name to compute hash.
8 lines of code changed in 1 file:
Use patch ID to compute hash instead of pointer
4 lines of code changed in 2 files:
Trying out a new type of residual. This residual is normalized component-wise by the solution vector. The norm is the L_inf norm.
4 lines of code changed in 1 file:
Schedule output tasks for all processors involved in the output not just the task that performs the output.
20 lines of code changed in 1 file:
Updated the emit code to use the new multimap correctly.
49 lines of code changed in 1 file:
Specify the bash shell as the interpreter.
12 lines of code changed in 1 file:
Turn off fracture mpm from the -all and -main flags.
2 lines of code changed in 1 file:
Potential fix for some task graph problems that would cause certain variables to not be found.
7 lines of code changed in 1 file:
Create Output and OncePerProc tasks only once on each processor instead of P times.
This will hopefully resolve some of the performance issues we have seen at very large numbers of processors.
This is a fairly major change and I have tested it against an optimized RT. If anyone sees anything odd over the next week please report it to me immediatly.
26 lines of code changed in 1 file:
Removed accidentally added spew.
0 lines of code changed in 1 file:
Added include
3 lines of code changed in 1 file:
Added tau profiling
1 lines of code changed in 1 file:
Removed an error check that is no longer valid.
5 lines of code changed in 2 files:
Repairing DQMOM and LU classes to return source terms for BOTH quadrature nodes.
44 lines of code changed in 4 files:
Add the -jim3 option to puda. This computes histograms of particle values.
191 lines of code changed in 3 files:
Fixed synatx problems with hash_multimap on updraft.
3 lines of code changed in 1 file:
Don't use hash_multimap when HAVE_GNU_HASHMAP is defined until we can figure out the correct syntax.
4 lines of code changed in 1 file:
Added some more debugging lines.
5 lines of code changed in 1 file:
Changed the # of processors to 8.
0 lines of code changed in 2 files:
Switched from a multimap to a hash_multimap.
1 lines of code changed in 1 file:
Moved some debug code into an assertion
2 lines of code changed in 1 file:
Adding residual normalization back in to DQMOM class.
18 lines of code changed in 2 files:
One more commit to fix compile problems
2 lines of code changed in 1 file:
Fixing a few more typos.
7 lines of code changed in 3 files:
Fixing minor typo
1 lines of code changed in 1 file:
Updating DQMOM and LU classes to reflect a few newer changes
533 lines of code changed in 4 files:
Fixed a bug.
2 lines of code changed in 1 file:
Added solver tolerance knob (<solver_tolerance> in <DQMOM>) to control the addition of the source term from the Ax=b solve.
15 lines of code changed in 2 files:
Refined the mixture fraction initialization to only initialize the interior of the domain, not including boundaries.
7 lines of code changed in 1 file:
Fixed a bug in the L_p norm calculator and normalized the residual by the solution vector.
13 lines of code changed in 2 files:
Added the ability to initialize the mixture fraction to a value other than zero by adding the <initial_value> tag into the <MixtureFractionSolver> node.
Committed Charles' new DQMOM solver.
613 lines of code changed in 6 files:
Added part of r44558 back in.
0 lines of code changed in 2 files:
removed r44558 due to issues in the dynamic scheduler.
8 lines of code changed in 2 files:
Enabled scrubbing for MPMICE, AMR, and codes that use a DLB. I'm not sure why it was disabled.
2 lines of code changed in 1 file:
Added support for the depricated gnu hashmap. This fixes updraft compiles.
18 lines of code changed in 1 file:
Added an include which does some hash_map trickery.
1 lines of code changed in 1 file:
Updated the erasing method to ensure we don't invalidate any iterators.
1 lines of code changed in 1 file:
added an end line to some output.
1 lines of code changed in 1 file:
updated some debugging code
6 lines of code changed in 1 file:
Always use version 0 to save scrub count.
21 lines of code changed in 1 file:
Initialize the version.
1 lines of code changed in 1 file:
Added changes for parallel reading of particle data
126 lines of code changed in 7 files:
Fix the order of return vars in getlist() function.
4 lines of code changed in 1 file:
Optimized data structure for VarDB to hash_multimap
26 lines of code changed in 2 files:
Minor optimizations.
14 lines of code changed in 1 file:
Added Julien's residual calculation.
49 lines of code changed in 2 files:
Added some more tau profiling.
Eliminated an unneeded intersection during the task graph compile.
17 lines of code changed in 2 files: