Corrected sign of eps_v.
45 lines of code changed in 5 files:
Updated to eliminate duplicate tags.
46 lines of code changed in 17 files:
Allow convective flux limiters to handle embedded boundaries by defaulting to upwind when close to embedded boundaries.
149 lines of code changed in 8 files:
Fix compile error in RayGPU.cu
0 lines of code changed in 1 file:
Benchmark 1 runs with Arches.
Temperature field has been hardwired.
Answers need to be verified
17 lines of code changed in 2 files:
Add corresponding Label changes to GPU version of Arches ray tracer.
4 lines of code changed in 2 files:
Made VRFLux label local to Ray.cc
8 lines of code changed in 6 files:
- removed plastic_convergence_algo flag.
- removed computePlasticStateBiswajit()
radial return is now the only option
Many of the MPM related RT tests will fail.
7 lines of code changed in 2 files:
Removed obsolete boolean modifies_VRFlux.
Todd: Removal of VRFlux_label from RMCRT.cc
to Ray.cc to come shortly
11 lines of code changed in 4 files:
Third time's a charm. Now we can get the various threaded schedulers under RT.
3 lines of code changed in 1 file:
Use mpirun with GPU RT test.
4 lines of code changed in 1 file:
Add additional command line options for sus (sus_options="string") in RGPU RT scripts.
3 lines of code changed in 2 files:
Compiles again now.
544 lines of code changed in 58 files:
Consolidate the convective flux expressions. Rewrite the flux limiters using a more abstract formulation: phi_face = LowOp(phi) - psi(r)*[ LowOp(phi) - HighOp(phi)] where phi_face is the desired convected value of a scalar, LowOp corresponds to a low order interpolant (e.g. upwind), psi(r) is the flux limiter function that depends on the ratio of successive gradients of phi and the flux limiter type, and finally, HighOp designates a high order interpolant (e.g. central). This changes will allow us to use exisiting operator threading in the SpatialOps library among other things. NOTE: this commit will again break the convection tests because of some round off errors accumulating from splitting algebraic operations. I will update the gold standards as soon as the RT machine fires up.
113 lines of code changed in 6 files:
Small formatting change that was bugging me.
1 lines of code changed in 1 file:
Remove unused variable.
0 lines of code changed in 1 file:
Warning...
1 lines of code changed in 1 file:
Remove offending tests from the regression tester until we find a solution for these random values in ghost cells.
0 lines of code changed in 1 file:
Set default threaded scheduler to centralized version.
2 lines of code changed in 1 file:
Adding information to the output to identify the user defined inputs.
0 lines of code changed in 2 files:
Add two debugstreams to control threaded2 scheduler and cpu affinity.
16 lines of code changed in 4 files:
Fix syntax error in GPU RT script.
2 lines of code changed in 1 file:
Initial version of updateSumI GPU kernel and rayTraceGPU method. Not optimized.
330 lines of code changed in 2 files:
Add GPU version of Ray constructor. This will change when a GPU DataWarehouse is implemented.
106 lines of code changed in 6 files:
fixed conditional surrounding computePlasticStateBiswajit()
1 lines of code changed in 1 file:
- Fix compiler error: "no matching function for call to 'Uintah::Task::Task in RayGPU.cu"
- Also starting to add more GPU tests
7 lines of code changed in 4 files:
Created a label for VRFlux in preparation for
putting the virtual radiometer model under
regression testing.
69 lines of code changed in 6 files:
- Add cuRAND includes to cuda_testdefs.h.in for host and device-side GPU RNG
- Simplify build system CUDA checks
- update configure script
- Turn patch combining back on in GPU scheduler. This was crashing on jaguar GPU partition
11 lines of code changed in 5 files:
Mac port hack script for bigmac
0 lines of code changed in 1 file:
fixed typos
6 lines of code changed in 1 file:
set mesh->logicalBounds to support better rendering with SLIVR in VisIt 2.4.2
20 lines of code changed in 1 file:
computeStressTensor()
- comments
- variable name change:
oldStress-> sigma
devTrialStress -> trialS
computePlasticStateViaRadialReturn()
- computing nn instead of Stilde.
computeEPlasticTangentModulus()
- passing in trialS instead of normTrialS & nn
82 lines of code changed in 2 files:
fix face-centered data problems with udaReader
87 lines of code changed in 4 files:
Dirichlet boundary conditions treatment for pressure projection using the PPE formulation. Also fix a bug in parsing the vorticity expression from the input file. Add regression tests for a lid-driven-cavity flow in all three Cartesian planes, respectively.
123 lines of code changed in 7 files:
Global:
changed task name
computeStressTensor -> computeStressTensorImplicit
ImplicitCM.cc:
Added bulletproofing. If you land in this stub function then
throw a fit.
*This should not change the answers.
110 lines of code changed in 20 files:
Roll back changes during continued development.
72 lines of code changed in 2 files:
Removed unused variable in IntrusionBC.h.
Passing parent ProblemSpec to scalarInletBase to parse the flux
direction correctly.
Update input files to reflect latest changes in IntrusionBC to fix the
RT tests. This will fix last night's failing tests.
10 lines of code changed in 2 files:
Add calculation of delt to the Wasatch timestepper so that it is accessible by other expressions such as the pressure. NOTE: I added a new member variable to the SimulationState that stores the current deltat. This was needed because of the way we are updating delt in Wasatch. You can now access the current deltat from the SimulationState (e.g. sharedstate). The current_deltat is being set in the AMRSimulationController global loop, right before a component executes its timestep. Finally, subtract the dilatation/dt from the pressure RHS. This is required to filter out any non-divergence-free initial conditions. This will be clearly visible when projecting with boundary conditions. NOTE: THIS WILL BREAK THE REGRESSION TESTER ON THE MULTILEVEL SCALAR TRANSPORT EQUATION. JAMES WILL SOON COMMIT A FIX IN THE EXPRLIB.
123 lines of code changed in 11 files:
Fixing the convergence criteria
4 lines of code changed in 1 file:
Use SpatialOps syntax evaluations in Vorticity.
1 lines of code changed in 1 file:
Silent a warning for older gcc compiler.
2 lines of code changed in 1 file:
Scalar/Velocity input file option for intrusion inlets.
224 lines of code changed in 2 files:
Further improvements on the performance of the Wasatch pressure solver. Use the previous pressure solution as an initial guess - from now on, we will need to specify an initial condition for the pressure from the input file. This is simply a constant BasicExpression. Also, start pushing for the RK3 multiprocessor pressure solve. As of this commit, RK3 pressure solve works with multiprocessor problems. Finally, reduce the number of timesteps on some of the convection regression tests as they are taking a long time to complete. THIS COMMIT WILL BREAK THE WASATCH REGRESSION TESTS.
24 lines of code changed in 4 files:
This is the current buildbot config file.
438 lines of code changed in 1 file:
Update GPUThreadedMPIScheduler to support the same number of threads as
ThreadedMPIScheduler.
6 lines of code changed in 2 files:
Support up to 64 threads per MPI rank.
37 lines of code changed in 12 files:
Update the way Wasatch sets up the matrix for the pressure solve. Since we're using the CG Hypre solver for symmetric matrices, we can substantially improve the solver's performance by solving: {- [A] p = - p_rhs}. In essence we need to supply the hypre CG solver with a positive definite matrix. Also make some solver input settings optional. Finally, update the regression tests for the Taylor-Vortex MMS tests - This will break the regression tests for Wasatch. Will update goldstandards after the RT machine fires up.
17 lines of code changed in 1 file:
Additional regression tests for high order flux limiters with boundary conditions.
69 lines of code changed in 1 file:
Fixed the errors in compilation. I don't know why my compilation at work didn't catch these.
9 lines of code changed in 2 files:
Fixed a bug with the laminar Pr model + a problem in referencing the
incorrect cell when actually using the model in Discretization_new.h.
Also, further refinement on the intrusion class. Seems to be working for
multiple patches now, even when geometry lines up with the patch
boundaries.
686 lines of code changed in 6 files:
Updated the VR code. Now compatible with
the abstracted intensity function.
7 lines of code changed in 1 file:
Fix bug in the z-direction flux limiter interpolant. Also use central interpolation for the convected scalar when the velocity is zero instead of setting the scalar to zero. Regression tests will follow.
6 lines of code changed in 1 file:
Placed ray marching and computation of sumI in
its own function. sumI is used for many calculations
including the virtual radiometer, divQ, and surface
fluxes. The abstraction of sumI allows the intensity
solver to be easily caled without code duplication
for the various applications.
Also changed variable name "SumI" to "sumI" for
consistency.
185 lines of code changed in 2 files:
In iterative solvers added the modified EOS to correctly handle cases where the relative volume is close to 1.
141 lines of code changed in 2 files:
modified a tiny bug in Arenisca.cc
10 lines of code changed in 1 file:
Updated but a long way to go.
142 lines of code changed in 31 files:
Under development. Significant changes are expected in the structure.
18540 lines of code changed in 90 files:
Forgot ElasticPlastic.
0 lines of code changed in 1 file:
Commenting out Biswajit subdir.
1 lines of code changed in 1 file:
Removed ElasticPlastic.
1 lines of code changed in 4 files:
Moved to Biswajit folder.
0 lines of code changed in 30 files:
updated Arenisca: improved its upper limit for kappa
65 lines of code changed in 2 files:
Better handling papi and google performance toolkit configure.
734 lines of code changed in 2 files:
using gforge.sci.utah.edu instead of code.sci.utah.edu for svn repository.
1 lines of code changed in 1 file:
global, variable name change:
tensorSig -> sigma
phi -> flow_rule
sqrtSxS -> normS
doNewtonIterations -> doRadialReturn
added ups option
<plastic_convergence_algo> radialReturn or biswajit </plastic_convergence_algo>
not fully implemented.
computeStressTensor v.1()
- moved code inside of if(sqrtSXS) {} into a new function named
computePlasticStateBiswajit()
- renamed computeStilde() -> computePlasticStateViaRadialReturn()
I'm still working on consolidating this.
* This should not change the answers
245 lines of code changed in 2 files:
Add Stencil4 structure for symmetric matrices. Update the HypreSolver interface to deal with Stencil4 matrices. NOTE: the current workaround does NOT require any of the components to make changes to their linear solver. It would be nice to move to a templated version of the hypre solver with the Stencil type as a template argument, but that will require reworking all the schedul_solver calls. Regression tests passed on ARCHES, ICE, MPM, MPMARCHES, and Wasatch. Please let me know if you get into problems with your linear solver.
261 lines of code changed in 15 files:
Support --with-gperftools in static build.
11 lines of code changed in 3 files:
removed unused deltaGamma().
0 lines of code changed in 2 files:
Multiplying Precipitation Source term by rho within expression. Fixing overlooked compile warning.
21 lines of code changed in 3 files:
Removed unused functions/methods --Scott Bardenhagen
0 lines of code changed in 5 files:
LOCALTESTS == NIGHTLYTESTS
1 lines of code changed in 1 file:
Squashed commit of the following:
Add support for google cpuprofiler heapprofiler heapchecker
The env variables to turn them on:
CPUProfiler HeapProfiler HeapChecker
Note: you cannot do heap profile and heap check at the same time.
61 lines of code changed in 4 files:
- Add build system support for Google Perftools (CPU and heap profiler)
- Add file: include/sci_defs/gperftools_testdefs.h.in
- Update configure script
397 lines of code changed in 6 files:
removed errant ] and spaces
5 lines of code changed in 1 file:
More fallout from swapping hypo_elastic_plastic -> elastic_plastic_hp.
Updated addMPMmaterial machinery.
34 lines of code changed in 1 file:
Bug fix. DivQ at cell 0,0,0 was being affected
by the VR code. This should fix the problem
15 lines of code changed in 1 file:
Updated localTests
localTest = NightlyTest - explode2D_amr
9 lines of code changed in 1 file:
Remove unary field operators in favor of <<= assignment syntax
32 lines of code changed in 4 files:
Changed task name : computeStressTensor -> computeStressTensorImplicit
Most of the input files that use these CMs don't run so cross your fingers.
33 lines of code changed in 8 files:
Added virtual radiometer model. This allows for the
modeling of both wide and narrow-angle radiometers
with user-specified location, orientation, and view
angle. Currently the divQ is also computed for the
location of the radiometer, but with incident
radiation restricted to the view angle. That will
be fixed shortly. Multiple radiometer capability
will also be added shortly.
115 lines of code changed in 2 files:
Constitutive model cleanup:
- moved hypo_elastic_plastic -> land of discarded CMs
- moved small_strain_plastic -> land of discarded CMs
All ups files:
small_strain_plastic -> elastic_plastic_hp
hypoelastic_plastic -> elastic_plastic_hp
This will change the answers
5972 lines of code changed in 10 files:
Rename task:
computeStressTensor -> computeStressTensorImplicit
added printTask() & printSched() throughout.
47 lines of code changed in 2 files:
A few more modifications to the intrusion boundary class.
Also, ensuring that the velocity is set on the face of the intrusion
where the velocity != 0.0.
57 lines of code changed in 3 files:
Modifications on ghost handling in Wasatch
17 lines of code changed in 4 files:
Global:
computeStressTensor()
removed
- if (flag->d_integrator == MPMFlags::Implicit) {
- computeStressTensorImplicit(patches, matl, old_dw, new_dw);
- return;
- }
Now calling computeStressTensorImplicit() directly.
ConstituteModel.cc:
Throw an exception if you land in a empty stub function that is required.
* This should not change the answers
23 lines of code changed in 12 files:
Added new source term to avoid using messy ExprAlgebra from input files. Reformatted some comments. Changed nebo_cond to cond to be consistent with SpatialOps update.
237 lines of code changed in 7 files:
A version of Arenisca which passes verification tests and works for penetration sim with considering cap function
145 lines of code changed in 2 files:
Complete hack to get the localRT to run consistently on hmx & fin.
6 lines of code changed in 1 file:
reverted formatting change that was causing RT to fail.
2 lines of code changed in 1 file:
Fix mistake in last commit with EnthalpyShaddix.
Clean out more warnings.
8 lines of code changed in 5 files:
Warnings...
11 lines of code changed in 3 files:
Changing the map storage to use patchID.
Clean up of input interface.
Changing the way iterators are assined for the intrusion. They now
should be the face index of the intrusion rather than the first interior
index.
142 lines of code changed in 6 files:
Formatting, alignment and demarcations
*This should not change the answers.
287 lines of code changed in 1 file:
Add license & copyright information
2143 lines of code changed in 81 files:
- Fix portability problem related to cudaHostRegister. Now works on both Keeneland and Jaguarpf GPU partition.
- Fix problem where GPUSchdeduler crashes when run without "-gpu", meaning it should be able to run in CPU-only mode too.
- Update message shown when GPUScheduler throws ProblemSetupException.
22 lines of code changed in 1 file:
Backing out addition of Mersenne Twister GPU, opting for a much simpler, and more portable GPU RNG provided by CURAND library.
- Includes an updated configure script.
9 lines of code changed in 6 files:
Added use of new Spatial Ops nebo conditionals. Adding additional error printout for QMOM. Added constant expressions of table headers for wasatch-in-arches build. Removed unecessary ifdef.
30 lines of code changed in 7 files:
Take 2 on incorporating RadProps.
24 lines of code changed in 1 file:
Turn on GPU tests for GPU RT machine. Need to make this more generic.
5 lines of code changed in 1 file:
minor formatting changes
0 lines of code changed in 1 file:
Modify some memory management in QMOM to hold some work arrays as temporaries and avoid reallocation at each execution.
101 lines of code changed in 1 file:
Data onion:
- rays now march over a region of the fine level, not just a fine level patch
The region is defined by user defined thresholds.
Still testing.
98 lines of code changed in 2 files:
Fixing some bad error checking.
0 lines of code changed in 1 file:
Added the new efficiency calculator.
647 lines of code changed in 3 files:
Allow boundary geometry objects (circle, ellipse, annulus, rectangle) to be included on boxes with differen upper and lower bounds. The current behavior when multiple boxes are used defaults to those with lowest and highest bounds.
39 lines of code changed in 1 file:
Fix a memory leak that was occurring on the multi-level regression test related to VarLabel creation (thanks, Todd!)
15 lines of code changed in 2 files:
minor cleanup
27 lines of code changed in 2 files:
Improve support for <double> field types in wasatch.
7 lines of code changed in 2 files:
added const explicit template instantiations
8 lines of code changed in 1 file:
is_BC_specified()
-don't check if BCs are specified if the domain is periodic in all directions.
3 lines of code changed in 1 file:
Add field adaptors for <double>
7 lines of code changed in 1 file:
A solid state reaction model with various different constituent pieces.
2195 lines of code changed in 25 files:
Fix last commit. Update configure and configure.ac to really check
"cblas" lib.
64 lines of code changed in 2 files:
Added "-D__STDC_CONSTANT_MACROS" to NVCC_CFLAGS and NVCC_CXXFLAGS (for MersenneTwisterGPU.cu).
__STDC_CONSTANT_MACROS is for the macro UINT64_C() to work correctly from stdint.h and appends
the correct suffix for a 64-bit unsigned integer literal to c. __STDC_CONSTANT_MACROS
is a workaround to allow C++ programs to use macros specified in the C99 standard that
aren't in the C++ standard.
Also updated configure script
25 lines of code changed in 3 files:
The beginnings of a GPU Mersenne Twister (for RMCRT).
12709 lines of code changed in 8 files:
Really check blas library.
27 lines of code changed in 1 file:
Disable super-patch creation in GPU SCheduler (for now).
5 lines of code changed in 1 file:
Prevent name mangling for GPU kernel in GPU scheduler test component.This will enable usage of CUDA driver API at a later point.
65 lines of code changed in 2 files:
set the output base file name for the system. Works for
HypreSolver.cc #define PRINTSYSTEM
8 lines of code changed in 1 file:
tabs!
formatting
5 lines of code changed in 1 file:
Cleaning up how components are GPU-enabled,
138 lines of code changed in 4 files:
Added new reduction variable types minvec and maxvec.
86 lines of code changed in 2 files:
- Cleanup includes for RayGPU.cu
- Add MIT license to CCA/Components/Models/Radiation/RMCRT sources and headers.
149 lines of code changed in 4 files:
Disable multithreading on Wasatch3P for now to get things building again.
1 lines of code changed in 1 file:
carryOverVars()
- manually carryOver reduction variable of type double. Thanks Dav!
28 lines of code changed in 1 file:
SourceTerms/RMCRT.cc
- added single level RMCRT to SourceTerm Framework
90ish% complete
EnthalpySolver & Explicit Solver
- removed RMCRT code.
312 lines of code changed in 5 files:
burns Christon:
define min & max if using grid spatial range.
2 lines of code changed in 1 file:
BurnsChriston()
- user can specify the min & max of a bounding box where abskg will be
set. The default is the entire domain. This will be useful
in multi-level RMCRT tests.
<property_calculator type="burns_christon">
<min> [0,0,0] <min>
<max> [1,1,1] <max>
<property_calculator>
52 lines of code changed in 1 file:
Stubs and necessary modifications to begin the GPU ray tracer.
81 lines of code changed in 4 files:
Unoptimized GPU FirstOrderAdvector. Using new interface... experimental.
158 lines of code changed in 2 files:
Fix Todd's RMCRT soot problem.
Turn back on intrusion inlets.
7 lines of code changed in 3 files:
RadPropertyCalculator.h:
- added missing include files
RMCRT:
- added property calculator
arches_spec.xml:
- added rmcrt_radiation tags
18 lines of code changed in 3 files:
Add the function areAllNodeValuesNumbers. This takes in an NCVariable of
Vectors and returns the first node that it finds with an inf or nan.
43 lines of code changed in 2 files:
Helping Jeremy with a header inclusion problem.
Also improve const correctness
35 lines of code changed in 1 file:
Wrong header in previous commit.
1 lines of code changed in 1 file: