remove a cout statement
0 lines of code changed in 1 file:
1. Add expression that specifies a (structured) layout of spherical bubbles.
2. Cleanup some code and rename a few variables for clarity.
151 lines of code changed in 8 files:
Expon.cc/h
- added several approximate and fast exponential functions. Consider using these if
there is an exp() in the inner most loop and you don't need 16 digits of precision.
They are best suited for arguments -1 < x < 1.
RMCRT:
- Added machinery to utilize the fast exponent methods. Currently, disabled. We need to
characterize the range of the optical thickness before we can take advantage of them. Preliminary,
tests show a speedup of 1.4 - 1.7 for the Burns & Christon problem while maintaininig ray
convergence.
251 lines of code changed in 7 files:
***changes in answers expected***
* Allow the tangential components of the velocity to be set such that
the interpolated value to the face satisfies the original condition.
Previously, the velocities were set directly on the staggered cell
that resulted in interpolation error to the face but provided the
convenience of not having to reset them every time step. With this
change, they are now being recomputed every time step. (should change
answers for anything with tangential components at a velocity inlet)
* Interpolate the particle source term to the faces of the P-cell, which
are then used in the velocity sources. (changes answers)
* Fix a logic error in the boundary conditions which was causing an
unnecessary loop on outlet BCs. (this shouldn’t change answers)
83 lines of code changed in 6 files:
Fix a bad bug in the SolverParameters base class that essentially set the Matrix setup frequency to 1 - ALWAYS. Unless the user specifies setupFrequency in the input file or the component developer casts the SolverParameters as a HypreSolver2Params, the setupFrequency is always set to 1. This likely impacted all Wasatch results that required a linear solver since Feb 2014. I think the setupFrequency should not be specified by the user from the input file. In addition, if the Matrix stencil is not changing, then there's no need to re-setup the matrix. Instead, I added a new option updateCoefFrequency that allows one to keep the Matrix structure intact but only update the coefficients. Test this via setUpdateCoefFrequency. I also added this option to the Uintah ups spec. This commit should not impact Arches simulations since Arches sets the setupFrequency through the input file. This commit will break many Wasatch RT tests with minor diffs.
43 lines of code changed in 4 files:
Do NOT setup the matrix every timestep for variable density flows.
Turns out that we really don't need to do that since our interface to Hypre will allow changing values while sparsity patterns are fixed.
This SHOULD put us back on track with scalability on the linear solver in Wasatch with varden problems.
1 lines of code changed in 1 file:
One more pidx link
8 lines of code changed in 1 file:
Link in PIDX for static builds.
51 lines of code changed in 15 files:
Fix bug in my previous commit. This should fix the aurora GPU buildbot.
5 lines of code changed in 1 file:
When the pressure solver is disabled, allow users to specify an expression for the pressure. This untangles a few knots related to the fact that the pressure expression requires access to Uintah-related information.
17 lines of code changed in 4 files:
one more try at fixing Arches build
2 lines of code changed in 1 file:
propagate SpatialOps changes of Particle Operators through Arches. This should fix the buildbot (hopefully).
2 lines of code changed in 2 files:
Fix typo in my earlier commit for the 3D compressible flow test
2 lines of code changed in 1 file:
GPU-ready particle operators. Commit on behalf of Dr. Babak Goshayeshi.
29 lines of code changed in 14 files:
add a 3D periodic compressible flow test. This serves as a helpful benchmark for scaling studies on compressible flows.
2 lines of code changed in 1 file:
This has more problems than I have time to fix right now. Turning it off for
the near term.
1 lines of code changed in 1 file:
added new diff_curve option to the NonLinearDiff model. The diff_curve is similar to the LoadCurve used for boundary conditions
42 lines of code changed in 2 files:
Enable CUDA for the analytic density calculator.
6 lines of code changed in 1 file:
Added function stubs to the ScalarDiffusionModel so that particle state variables can be declared and initialized within the a scalar diffusion model
18 lines of code changed in 4 files:
removed PIDX_compression.h include.
0 lines of code changed in 1 file:
Disable one of the GPU RT tests because it is failing checkpointing. We need nightly RT on GPU. We'll disable this test until the bug is fixed.
1 lines of code changed in 1 file:
Revert "BlockRange: restore ability to extract indices from linear index"
This requires c++11
This reverts commit 82ffe5ccd4f75851449b565c01f3a75c5044da30.
0 lines of code changed in 1 file:
BlockRange: restore ability to extract indices from linear index
31 lines of code changed in 1 file:
Unfinalize/Refinalize dws appropriately when doing Host/Device copies.
25 lines of code changed in 1 file:
Remove/add/change some regression tests.
7 lines of code changed in 1 file:
Additional logic for particle splitting, good enough until it isn't.
Fix some logic associated with axisymmetry in SerialMPM.
8 lines of code changed in 2 files:
Changed mass transfer coefficient correction factor to be consistent with derivation.
5 lines of code changed in 2 files:
With the removal of the SCIRun namespace, there were a number of
places where 'SCIRun::' was replaced with 'Uintah::'. However,
some of these were already inside of the "using namespace Uintah {"
sections which causes the icc compiler to emit 1000s of warnings.
I have removed the "Uintah::" from these locations. Basically
this is a purely cosmetic update to remove warnings.
81 lines of code changed in 29 files:
Remove superfulous, outdated and unused classes/files.
1 lines of code changed in 15 files:
Use the functor switch.
8 lines of code changed in 2 files:
Move an allocation outside of the #ifdef.
2 lines of code changed in 1 file:
BUGS:
* Fix a bug with the handoff files. Both heat loss and DQMOM variables
* weren’t being properly setup, resulting in the default value always
* being used.
* Fix a bug wherein the handoff output (showing missed hand-off
* information) wasn’t being
* printed correctly for the velocities.
KOKKOS:
* Cleanup ConvectionHelper/DiscretizationTools to use Dan’s latest
* infrastructure improvements.
* This version of code gives nearly the same answers as the
* non-Kokkoized code. The diffs are understandable from an order of
* operations standpoint.
MODELS:
* For drag, use the magnitude of the difference in velocities
* For drag, use the updated weight rather than the guess. This
* helps stabilize things, BUT it isn’t general enough. —> NEEDS
* WORK<—
* In PartVel, compute the face centered velocities for the
* transport. This should allow us to phase out the
* CCVariable<Vector> velocities.
OTHER
* For MPI_Bcast, send and int rather than an unsigned int to
* remove a warning.
* Consolidated the checkBCs function into BoundaryCondition_new.
3587 lines of code changed in 19 files:
Add new discretization functors. Work in progress...
1400 lines of code changed in 2 files:
* Included Array3.h to support Kokkos-based builds using KokkosView3
4 lines of code changed in 1 file:
Ignore the Eigen dir
0 lines of code changed in 1 file:
Added new functors to the momentum solver. Also, consolidated compiler directives into a new header.
150 lines of code changed in 13 files:
Remove unused files.
0 lines of code changed in 5 files:
Fix compiler error in GPU builds.
1 lines of code changed in 1 file:
Remove empty directories.
0 lines of code changed in 8 files:
Add Uintah namespace.
0 lines of code changed in 2 files:
Add using namespace Uintah.
2 lines of code changed in 1 file:
Add Uintah namespace.
1 lines of code changed in 1 file:
Eliminate SCIRun namespace usage.
6 lines of code changed in 5 files:
Remove old SCIRun TypeDescription and all traces of the SCIRun namespace.
3 lines of code changed in 16 files:
Add Uintah namespace, remove __digital__ reference.
2 lines of code changed in 1 file:
Remove unused classes.
0 lines of code changed in 16 files:
Simplify the SCIRun TypeDescription class.
190 lines of code changed in 11 files:
More changes from SCIRun to Uintah. Delete unused SCIRun build_script files.
26 lines of code changed in 25 files:
Remove Persistent class and all references to it.
68 lines of code changed in 47 files:
Remove CompGeom class.
0 lines of code changed in 9 files:
Remove Transform class and PersistentSTL.
41 lines of code changed in 18 files:
Remove persistent I/O calls from various classes.
65 lines of code changed in 51 files:
Move non-conflicting classes that were in SCIRun namespace to Uintah namespace.
1633 lines of code changed in 903 files:
Functorized Particle Birth model.
67 lines of code changed in 1 file:
Functorized computing the sum of drag, devolitalization, char-oxidation, and enthalpy source terms.
217 lines of code changed in 6 files:
Fix for bizarro ostringstream "bug". While this same code (well, almost the exactly the same, but moved
around a little) has been working since Uintah was created, for some reason, and only on some debug builds,
this "bug" has surfaced. The problem is that the ".str()" operator on an ostringstream returns a "temporary"
string, so that when you use ".c_str()" on it and store the result for later use, the ".str()" can
deallocate itself and thus make the ".c_str()" invalid. The solution to this is to copy the ".str()" out
of the ostringstream and into its own std::string variable and then use the ".c_str()" on that new
variable. Again, I cannot for the life of me imagine why this just started to show up and only in
some situations. However, this should fix it. Note, I say "bug" because in theory this is the defined
behavior for "str()", though according to google, not many people know this. -Dd
8 lines of code changed in 1 file:
This commit changes the way Uintah checks for the file system it
will use to save UDAs. Previously (and still available via a
command line flag to sus) Uintah had every process create a file,
and then every other process look for (some of the files) in order
to determine if everyone was running on the same file system or on
a local system. Currently we don't use any machines that have
local (per node) disks, so we don't need to do this work and,
because on 100K+ jobs, this hammers the file system, can take
minutes, and possibly cause errors, we will now default to assuming
a shared file system (we do a single file check for this).
M CCA/Components/SimulationController/AMRSimulationController.cc
- Remove warning by putting VISIT var into #if HAVE_VISIT.
M CCA/Components/SimulationController/SimulationController.h
M CCA/Components/SimulationController/SimulationController.cc
- Store d_usingLocalFileSystems in the shared state.
M CCA/Components/MPM/ReactionDiffusion/NonLinearDiff1.cc
- Remove warning by putting unused var inside #if.
M CCA/Components/OnTheFlyAnalysis/MinMax.cc
M CCA/Components/ICE/ICE.cc
- Replace GetVisIt() with renamed getVisIt().
M CCA/Components/DataArchiver/DataArchiver.h
M CCA/Components/DataArchiver/DataArchiver.cc
- Main changes to file system check in this file.
- Re-wrote the check for a common FS in setupSharedFileSystem().
- Mostly has process 0 create a file and all the others look for it.
- Put old code that was in initializeOutput() into setupLocalFileSystems().
M CCA/Components/Schedulers/SingleProcessorScheduler.cc
M CCA/Components/Schedulers/UnifiedScheduler.cc
M CCA/Components/Schedulers/MPIScheduler.cc
M CCA/Components/Schedulers/ThreadedMPIScheduler.cc
M CCA/Components/Schedulers/DynamicMPIScheduler.cc
- Minor white space.
M CCA/Components/Schedulers/SchedulerCommon.h
M CCA/Components/Schedulers/SchedulerCommon.cc
- Update names of some class variables to be consistent (_) with
naming convention in file.
- OCD formatting.
M R_Tester/helpers/runSusTests.py
M R_Tester/toplevel/generateGoldStandards.py
- Fix generateGoldStandards check for correct version of python (need 2.7 or better).
- Update "print" statement to be "print()". This allows local RT python scripts
to run under python 3.
M Core/Grid/SimulationState.h
M Core/Grid/SimulationState.cc
- Added d_usingLocalFileSystems var. Few minor whitespace fixes.
- Uintah uses lower case letters to begin functions. Updated setVisIt() and getVisIt()).
M StandAlone/inputs/MPM/disks.ups
- Minor whitespace edit.
M StandAlone/sus.cc
- Check for/handle new command line arg: -local_filesystem
- Most likely no one will ever use this...
- As we do note have access to a system with local file systems, this has not been
tested but the code just passes through to the original code for checking
for a local FS.
- Minor whitespace / formatting.
670 lines of code changed in 20 files:
This is a change to the deposition rate probability that Minmin wasn't able to commit. It should not effect regression tests.
20 lines of code changed in 2 files:
level
- added machinery to return the max number of cells in a patch, over all patch.
nCellsPatch_max(). Needed my PIDX only.
PIDX:
- fixed a corner case bug when computing the outputPatchSize.
47 lines of code changed in 4 files:
Made compiler directives more readable.
4 lines of code changed in 1 file:
A quick bug fix dealing with the GPU memory pool cleanup involving empty placeholder variables in varPointers (entries in varPointers which are there to only hold staging vars)
32 lines of code changed in 1 file:
This fixes a couple of bugs with the GPU memory pool. It also overhauled the CUDA stream pool and fixed all outstanding issues with it.
260 lines of code changed in 8 files:
Functorized sum of momentum from environments.
40 lines of code changed in 1 file:
Remove conditional on pref==0. This was not well thought out.
10 lines of code changed in 1 file:
Global:
changed variable name combinePatches -> outputPatchSize
PIDXOutputContext::computeBoxSize
- added crude logic to determine the best outputPatchSize if it's not specified
by the user. There is an issue that will be rectified in the future.
63 lines of code changed in 2 files:
Added python script that copies Grid and Data from grid.xml into timestep.xml, within each uda and each time-step.
67 lines of code changed in 1 file:
untangle a mess of circular includes that was causing problems for GPU particle operators.
1141 lines of code changed in 14 files:
Fixed compiler warning in constructor
Removed EOL spaces.
query()
- using Parallel.h to get MPI_Comm and pass that to PIDX. -Alan
121 lines of code changed in 1 file:
Modifications to improve scalar flux boundary conditions, in particular with
use of particle splitting.
47 lines of code changed in 7 files:
Change the volume ratio criteria from 1% to 5%. Change the definition of the
area vector that this returns.
7 lines of code changed in 1 file:
DataArchiver::problemSetup()
- fixed logic error when wallTimeInterval is specified. -Alan
- output checkpoint specs for user.
14 lines of code changed in 1 file:
Increase the angular interval for particles if their volume gets smaller than
1 percent of the cell volume. This should probably be an input file option, but, for now, it ain't.
23 lines of code changed in 1 file:
PIDXOutputContext:: initialize()
- only set compression flag if it is a normal output timestep.
PIDXOutputContext::computeBoxSize()
- added infrastructure to compute the size of the pidx box. To be filled in.
85 lines of code changed in 3 files:
Change abs to fabs
3 lines of code changed in 1 file:
Fixes a couple of issues. First is that GPU tasks were being run twice unless the user supplied extra logic to avoid that. Now there is no PostGPU mode that runs tasks a second time. The second fix is an addition of a GPU memory pool. This should help avoid costly cudaMallocs and also give us much better control over checking for GPU memory leaks
313 lines of code changed in 3 files:
added empty methods so you can compile without PIDX enabled.
moved bulletproofing upstream to PIDXOutputContext.
31 lines of code changed in 3 files:
added plumbing PIDX_flags() that's used for
setting PIDX output switches & flags via the ups file.
<DataArchiver type="PIDX">
<PIDX>
<outputRawIO> true </outputRawIO>
<combinePatches> [2,2,2] </combinePatches>
<debugOutput> false </debugOutput>
<compressionType> NONE </compressionType>
</PIDX>
</DataArchiver>
Still need to do something with combinePatches.
153 lines of code changed in 4 files:
The computation of the coal temperature has been functorized. Minor formatting of previous commit.
327 lines of code changed in 3 files:
Added a functor for the CharOxidation Source term. Also cleaned up some warnings.
330 lines of code changed in 5 files:
Functorized the coal devolitalization FOWY model.
174 lines of code changed in 2 files:
In the particle Enthalpy source functor, the function abs( ) was not functioning properly.
Switched abs( ) to std::abs( )
4 lines of code changed in 2 files:
fixed compute timestep for JGConcentration
7 lines of code changed in 2 files:
PIDX output:
- added exception. Only 1 patch per MPI rank.
13 lines of code changed in 1 file:
Stylistic change to functor.
197 lines of code changed in 1 file:
Added a rectangular-annulus boundary condition.
333 lines of code changed in 6 files:
Functorized computation of Coal-Enthalpy source term.
222 lines of code changed in 2 files:
PIDX code:
store the idx filename in pxxxxx.xml, not the full path. Hopefully, this commit will eliminate
some amount of confusion.
25 lines of code changed in 4 files:
Functorized the computation of the radiation fluxes, incident radiation, and heat-flux divergence.
174 lines of code changed in 1 file:
PIDX section:
Throw an exception if the variable type isn't supported, NCVariablables and Particles.
41 lines of code changed in 2 files:
more descriptive variable name
11 lines of code changed in 2 files:
Cleaned up kokkos compiler directives. Use Dan Sunderland's new interface for accessing grid variables.
4 lines of code changed in 1 file:
avoid constructing IntVectors when using operator()(int i, int j, int k) with Array3
13 lines of code changed in 4 files:
protect omp pragma in Uintah::parallel_for with an enable macro
16 lines of code changed in 1 file:
Add operator()(int i, int j, int k) to constVariable and update Poisson1 to always use a functor
160 lines of code changed in 4 files: