Remove leftover directory from Core/Basis purge.
0 lines of code changed in 42 files:
Add DebugStream reporting on number of tasks executed per TaskWorker thread (ThreadedMPI scheduler).
15 lines of code changed in 2 files:
minor tweaks to take into account input changes
40 lines of code changed in 5 files:
add regression test for a scalability test with an uncoupled, nonlinear, source term
6 lines of code changed in 1 file:
Add support for uncoupled nonlinear source terms in the scalability test. This will allow us to do comprehensive test coverage on GPUs et al.
117 lines of code changed in 3 files:
Back out part of commit r53306 (changes to ThreadedMPIScheduler.cc)
3 lines of code changed in 1 file:
In threaded scheduler cases that (for now) need to be single-threaded, simply handoff the work to MPIScheduler::execute()
These cases are:
* Copy data timestep
* Initial timestep
* Restart initial timestep
22 lines of code changed in 2 files:
For now, turn off per level support for maxGhostCells. This will allow the large-scale DDT simulations on Mira to advance using threaded-schedulers. We can revisit this issue when time permits.
65 lines of code changed in 4 files:
Fix a leak.
13 lines of code changed in 2 files:
Deleting labels here.
6 lines of code changed in 1 file:
Request ghosts for these variables.
7 lines of code changed in 1 file:
fix bug in scalability test rhs when using source terms
1 lines of code changed in 1 file:
Since these objects are not modified, pass by const reference.
24 lines of code changed in 1 file:
Undo a branch commit to keep the Uintah trunk clean.
1 lines of code changed in 1 file:
Fix a bug in the direction.
1 lines of code changed in 1 file:
* BTDevol -> FOWYDevol
* Determine the variable type for the VariableStats model.
* Clean up some RT tests
605 lines of code changed in 8 files:
Move TimeAve -> VariableStats and template the class.
984 lines of code changed in 6 files:
Clean up:
1) get_uintah_const_field -> get_const_uintah_field
2) Combine the bc and standard task interface. Now using enum to
distignuish the two. This will allow other task types to be added (if
needed) in the future.
96 lines of code changed in 7 files:
Better convention.
1 lines of code changed in 1 file:
Add the ability to add on more variables for time averaging OR
reinitialize variables to zero upon restart.
80 lines of code changed in 2 files:
Go over extraCells. Should have been this way from the start.
1 lines of code changed in 1 file:
Remove max_TW spec from the arches_spec.
Remove max_TW from wall model becuase it wasn't being used.
1 lines of code changed in 1 file:
Fix a logic error.
8 lines of code changed in 2 files:
Drop the required density for all time_ave options.
Only need density now if you are computing a flux.
134 lines of code changed in 2 files:
Updates to the timeAve property. This is currently working in the
tradiational Uintah style. The nebo version is not working with
extra-cells.
642 lines of code changed in 7 files:
Pass table (and associated variables) by reference rather than value to
avoid a copy.
Remove thread safety infractions in the find_val functions.
63 lines of code changed in 1 file:
Bandaid problems.
Using wall cell type to only copy in values into the real wall
temperature on non-solve timesteps. The previous method was causing
cases to go unstable.
23 lines of code changed in 2 files:
As a temporary fix to allow restarts, use the old temperature from the
OldDW as the previous solution. This results in the wall temperature
variable to be filled in the gas cells, a feature we don't want in the
long run but it overcomes the restart problem on cases that were run
before this feature was added.
19 lines of code changed in 1 file:
Add a new temperature label that only populates the TRUE wall
temperature value as determined with an emmissivity < 1.
97 lines of code changed in 4 files:
Fix a few bugs in the wall heat transfer model:
1) Change the order of operations in invoking the model
2) Work with patch ghost cells rather than ignore them
3) Dont do full copies on the flux variables. Use variable pointers
instead.
4) Dont allow the user to specify the temperature label for the wallHT
model.
121 lines of code changed in 3 files:
Test commit.
0 lines of code changed in 1 file:
Changed max Temp in WallHT to be (q/sigma)^0.25.
3 lines of code changed in 1 file:
Added maximum wall temperature option.
3 lines of code changed in 1 file:
Test commit.
1 lines of code changed in 1 file:
Dereks cell/node iterator fix.
1 lines of code changed in 1 file:
Modified wall heat transfer model.
46 lines of code changed in 1 file:
Merge of Derek/Ben/Jeremy branch into the latest svn head.
822 lines of code changed in 34 files:
Temp. fix for LoadBalancer bug that was causing AMR MPMICE to hang.
This will enable forward progress with threaded schedulers on the Gordon Bell effort related to Jacqueline's DDT problem.
42 lines of code changed in 4 files:
remove a redundant expression related to the Taylor-Green vortex, rename the remaining expression and cleanup related input file.
19 lines of code changed in 4 files:
remove an annoying cout statement that spams the output when running on GPUs
0 lines of code changed in 1 file:
LOCALTESTS:
- turned on threaded tests for Linux operating systems.
NIGHTLYTESTS:
- added rmcrt_bm1_ML_thread
4 lines of code changed in 1 file:
When setting the boundary conditions (setExtraCellScalarValueBC()) use the appropriate label names.
7 lines of code changed in 1 file:
sched_initialize() and initialize task
- initialize abskg on non-arches levels to 0.0. This allows the user to
outputInitTimestep.
26 lines of code changed in 1 file:
GPU DataOnion:
---------------------------
* Package fine level data, ROI and extents of the regions below the fineLevel into levelParams structs.
- Ship this to the GPU kernel.
* Some cleanup and formatting in DWDatabase.
357 lines of code changed in 3 files:
fix Wasatch memory leak for RMCRT.
1 lines of code changed in 1 file:
Moved setExtraCellScalarValueBC () from BoundaryCond_new.h -> .c
Templated the function and added explicit instantions for float and double.
Added getNewBoundaryCondition() to BoundaryCondition.h
ARCHES:RMCRT is now using arches boundary conditions.
Changed input files to reflect the change. User must specify
abskg "ForcedDirichlet"
158 lines of code changed in 6 files:
minor comment changes for cpdi and cpti interpolators
24 lines of code changed in 7 files:
made the output condition on a connection
10 lines of code changed in 1 file:
additional changes for the libsim and the library search
34 lines of code changed in 5 files:
added better options for using visit
70 lines of code changed in 7 files:
A few minor formatting adjustments while looking through CPU rayTrace_dataOnion.
6 lines of code changed in 1 file:
Generating level parameters for DataOnion GPU kernel (multi-level ray marching).
Add convenience class - GPUPoint.
Pushing ever onward....
151 lines of code changed in 4 files:
Add ability for GPUVariables to be differentiated by level (GPU AMR).
Some cleanup and refactoring.
115 lines of code changed in 3 files:
added better lbisim messages
15 lines of code changed in 1 file:
Pass in gridParams to rayTraceDataOnionKernel() so maxLevels and BoundingBox (domain size) is available.
12 lines of code changed in 3 files:
Remove multiple "privates" to make setExtraCell BC public.
10 lines of code changed in 2 files:
GPU DataWarehouse is working with AMR.
Some GPU and CPU DataWarehouse cleanup/reorganization before the big push on DataOnion GPU kernel.
1400 lines of code changed in 3 files:
No longer creating abskgRMCRT when the communication is
based on doubles. It causes a crash.
0 lines of code changed in 1 file:
Added methods so Arches boundary conditions will be used. Currently disabled.
182 lines of code changed in 2 files:
Access function for setExtraScalarValueBC.
17 lines of code changed in 2 files:
fix Wasatch RMCRT in light of Todd's separation of the CellType from the Ray class.
32 lines of code changed in 8 files:
fix a couple of memory bugs in Ray and RMCRTCommon
1 lines of code changed in 2 files:
Adding some computes to scheduleRefine, and a warning for using smooth geometry
pieces with AMR. That needs further testing to ensure the particle size is correct.
12 lines of code changed in 2 files:
Using sched_restartInitialize to enable models to be turned on for
restart when they haven't been turned on previously.
141 lines of code changed in 5 files:
Started filling in the RayTraceDataOnionKernel. With minor exceptions it's identical to
the RayTraceKernel. It gives identical answers to RayTraceKernel, as it should.
283 lines of code changed in 4 files:
M HelperBuildScripts/checkEnv.sh
- Check for cmake.
- Added Surface and Syrah (both @ LLNL).
M README
- Added a few hints.
62 lines of code changed in 2 files:
No longer setting boundary condition on cellType. CellType is initialized and then
carried forward on subsequent timesteps.
This won't change Example RMCRT answers.
0 lines of code changed in 1 file:
Arches is now setting cellType on non-arches levels during initialization. It's then
carried forward on all timesteps.
This will change the answers on all multi-level calculations.
2 lines of code changed in 1 file:
Merge with Todd's change.
1 lines of code changed in 1 file:
Pass BC to source terms through the extraSetup phase.
36 lines of code changed in 12 files:
Tests without RMCRT for the mac.
50 lines of code changed in 1 file:
Initilization of cellType on the coarse levels is now the component's responsibility.
RMCRT_test.cc Arches:RMCRT
- define cellType during the initialization and just carry forward on coarser levels.
RMCRT_test.cc
- Initialize cellType on coarse level to d_flowCell (-1). This is correct and
will changed multi-level answers.
93 lines of code changed in 5 files:
updateSumI_ML()
- moved the call to findRayDirection() upstream. This will fix non-repeatability issues with
all multilevel calcuations. The random seed is now fixed for subsequent calls to RNG.
Yes, the results should change......
1 lines of code changed in 1 file:
Computing the flag in_domain using the bounding box of the computational domain.
When a ray jumps to a coarser cell it can leap outside of the domain.
This is a stop gap commit so the dataOnion tests will run again.
The cellType calculation on the coarser levels is wrong and is there to prevent
the Arches tests from crashing. This must be fixed.
25 lines of code changed in 2 files:
Add another boost library on the link line.
4 lines of code changed in 2 files:
Setup, layout and stubs for DataOnion ray trace kernel.
711 lines of code changed in 7 files:
Add pScaleFactor to the particle refinement code.
8 lines of code changed in 1 file:
* Add initial infrastructure support for scheduleRestartInitialize().
* This should handle empty scheduleRestartInitialize() { } methods gracefully.
* Some other cleanup while under the hood.
106 lines of code changed in 6 files:
Convert one of the more complicated BCs into GPU using the amazing Nebo! Now we have full GPU coverage for Scalar BCs and fluid-flow problems that exclude outflow-type BCs. The bottleneck remains to be the pressure solve.
176 lines of code changed in 15 files:
Added scheduleRestartInitialize() to the framework. It's currently disabled, waiting for Alan's mods.
295 lines of code changed in 78 files:
Move the special treatment of particles under CP(T/D)I from here into the
FileGeometryPiece code where it belongs.
4 lines of code changed in 1 file:
Updates to accomodate CPTI in a manner consistent with the smooth geometry
object types.
41 lines of code changed in 6 files:
Improve the Uintah field wrapper (to SpatialOps) so that it accounts for any discrepancy between extra cells and ghost cells. This will help us move towards variable ghosts.
19 lines of code changed in 1 file:
updateSumI_ML()
- updated to reflect what the single level function is doing.
Answers are identical for 2L, RR:1, benchmark 1 problem.
- added debugging output for direct comparison with 1L algorithm.
75 lines of code changed in 4 files:
- In debugging code limit the cells to output.
- Now wrapping debugStream calls with conditional.
21 lines of code changed in 2 files:
Starting to use cellType the dataOnion tasks. Just pulling it out of the DW. Not being used yet.
33 lines of code changed in 4 files:
* Fix compiler warning, signed and unsigned comparison.
* Report whether or not MPI message combining is turned on, even when <small_messages> is not found in <Scheduler> block.
* Parallel - report number of MPI processes, and threads per process (if any). The notion of "processor" is dated and now confusing. Update section in user guide to reflect this change.
14 lines of code changed in 4 files:
Add support for complex boundary conditions on the GPU! This commit introduces a new Macro named APPLY_COMPLEX_BC that does quite a bit of magic with templates and type inferences.
Here's how to use it:
APPLY_COMPLEX_BC( my_complex_bc)
where my_complex_bc is ANY nebo expression of the same type as the field that is
computed by the expression. So, if you want to apply a BC on a field of type FieldT, then my_complex_bc is of type FieldT.
Examples:
APPLY_COMPLEX_BC( a*x + b);
APPLY_COMPLEX_BC(a_ * (x - x0_)*(x - x0_) + b_ * (x - x0_) + c_);
APPLY_COMPLEX_BC(phic_ * pow( 1.0 - abs(x - x0_) / R_ , 1.0/n_ ));
For expressions that depend on fields of a different type (say F (SVOL) = a* u (XVOL)), then simply use an in-situ interpolant
APPLY_COMPLEX_BC( a* interp<XVol,SVol>(u) );
The rest of this commit simplifies the constant BC application and staggered direct assignment to avoid un-necessary branching and confusion in the evaluate method.
247 lines of code changed in 4 files:
Removed tabs.
722 lines of code changed in 32 files:
allow lineextract to output values only without cell index. This is useful, for example, in unit testing when comparing data in two different tests that are in different directions (x vs y) but are expected to produce the same values (rotational invariance). Simply add -ni (--noindex) to use this feature.
18 lines of code changed in 1 file:
Added destruction of pLastLevel*
2 lines of code changed in 1 file:
Added pLastLevelLabel
8 lines of code changed in 2 files:
1. Added "pLastLevel" for use with AMR. This allows recomputing psize
when a particle changes level, which is necessary since psize is a size relative
to the grid cell dimensions.
2. Hijacked partitionOfUnity a bit, to update the value of psize at the
top of a timestep. As a results, psize must now be retrieved from the new_dw
for all subsequent tasks.
3. Added particle refinement code from SerialMPM.cc.
405 lines of code changed in 5 files:
Change new to scinew.
0 lines of code changed in 2 files:
OnDemandDataWarehouse.cc/.h
Added new method (getLevel()) for getting data for all patches on a level.
This is a specialization of getRegion() and is substantially faster on large patch counts runs.
RMCRT:
Replaced getRegion() call with getLevel()
156 lines of code changed in 5 files:
Add flattened, linearly indexed version of dynamic Array3 type.
371 lines of code changed in 4 files:
Add the script's path to the "path" so this can be run standalone.
9 lines of code changed in 1 file:
missed a comma in my previous commit. Again to fix GPU RT on Aurora
1 lines of code changed in 1 file:
Fix Wasatch GPU RT on Aurora
8 lines of code changed in 1 file:
Add regression tests for Wasatch GPU boundary conditions
17 lines of code changed in 1 file:
Add GPU Support for some boundary conditions.
48 lines of code changed in 4 files:
#define MAX_NAME 64
1 lines of code changed in 1 file:
Increase MAX_NAME length to 32 in GPUDataWarehouse.h
1 lines of code changed in 1 file:
added visit specific flags and better var checking
67 lines of code changed in 3 files:
added visit specific flags
3 lines of code changed in 1 file:
reverted the simstats
11 lines of code changed in 3 files:
Additional stuff to accomodate particle refinement.
Also, commenting out the scaling of psize by "size" in a special case, since it
conflicts with scaling that takes place elsewhere.
25 lines of code changed in 1 file:
Major changes to the boundary conditions masks to allow us to solve boundary value problems on GPUs. Store four volume masks on the (also on the GPU) and use in-situ shifting to apply a boundary condition. This will avoid un-necessary CPU-GPU data transfers. This should not break the RT.
318 lines of code changed in 5 files:
removed debugging output
0 lines of code changed in 1 file:
Refactor mispelled method name.
3 lines of code changed in 2 files:
minor formatting changes & cleanup
0 lines of code changed in 4 files:
Avoid a lot of recalculation in situations where the environment temperature is constant.
3 lines of code changed in 1 file:
Remove superfluous semicolon that triggers compiler warnings on some compilers.
217 lines of code changed in 112 files:
* Some polish on threaded subschedulers. Works with imp ICE/MPM.
* Rearrange members and change access. Don't know why some of this was public.
* Along the above lines,make TaskWorker friend of ThreadedMPIScheduler (same for UnifiedScheduler).
40 lines of code changed in 4 files: