Level::getTotalCellsInRegion()
- Take into account different variable types (CC, NC, SFC(*)) when computing the patch extents.
18 lines of code changed in 3 files:
added some comments for clarity
4 lines of code changed in 2 files:
Variable name change:
tHypreAll_ -> m_tHypreAll
tSolveOnly_ -> m_tSolveOnly
tMatVecSetup -> m_tMatVecSetup
15 lines of code changed in 1 file:
createDetailedTasks()
- set the default value for task->m_max_ghost_cells[level] = 0
2 lines of code changed in 1 file:
Fix reporting of tasks' static vs. scheduled execution order in all schedulers. This is critical for ongoing taskgraph work.
Example output:
-----------------------------------------
Timestep 1 Time=0.01 Next delT=0.01 Wall Time=0.0996022 EMA=0.0507536 Memory Used=71.13 MBs (avg) 71.67 MBs (max on rank: 1 )
Rank-1 Running task: "send old data" (InitialSend) static order: 1 scheduled order: 1
Rank-1 Running task: "ApplicationCommon::updateSystemVars" (OncePerProc) static order: 8 scheduled order: 2
Rank-1 Running task: "Poisson1::timeAdvance" (Normal) static order: 2 scheduled order: 3
Rank-1 Running task: "DataArchiver::outputVariables" (Output) static order: 6 scheduled order: 4
Rank-1 Running task: "Reduction: residual, level -1, dw 1" (Reduction) static order: 3 scheduled order: 5
Rank-1 Running task: "Poisson1::computeStableTimeStep" (Normal) static order: 4 scheduled order: 6
Rank-1 Running task: "DataArchiver::outputVariables(checkpoint)" (Output) static order: 7 scheduled order: 7
Rank-1 Running task: "ApplicationCommon::reduceSystemVars" (OncePerProc) static order: 5 scheduled order: 8
Timestep 2 Time=0.02 Next delT=0.01 Wall Time=0.150609 EMA=0.0509239 Memory Used=71.83 MBs (avg) 72.12 MBs (max on rank: 1 )
-----------------------------------------
78 lines of code changed in 5 files:
Added debugging output statements, commented out.
scheduleInitialize & scheduleRestartInitialize
- set task type to oncePerProc. You must run these tasks on every proc. It's possible
to have no patches for a given proc. For example, restarts with regrids (Arches:regridderTest).
24 lines of code changed in 1 file:
- Added another version of printDebuggingPutInfo(). This one accepts Level* as an argument.
- Finished instrumenting the put() calls so you can see when reductionVars, SoleVars and Grid variables
are put into the DWDataBase.
40 lines of code changed in 2 files:
Added ProcessorGroup * m_myworld.
0 lines of code changed in 4 files:
Removing some old and unused CUDA code. I spoke with Alan on this, he put it in a while ago for a side project, and it hasn't been used since. I had some seg faulting problems on this in branch, and so I'm removing it here in case the seg faults show up in trunk in a GPU build
0 lines of code changed in 2 files:
scheduleAndDoDataCopy()
-Fixed memory leak in SoleVariables.
1 lines of code changed in 1 file:
Added initialization within a geometric object for the heat_loss property model. This won't impact tests.
50 lines of code changed in 2 files:
added a flag to the app to know it is restarting
23 lines of code changed in 4 files:
updated comment.
1 lines of code changed in 1 file:
SoleVariable.h & SchedulerCommon
scheduleAndDoDataCopy()
- Added code to copy SoleVariables to the new_dw when a regrid occurs.
overrideVariableBehavior()
- Improved the comments.
impICE.cc & HypreSolver.cc
- changed overrideVariable(hypre_solver_label) flags to copy data to new_dw.
30 lines of code changed in 4 files:
Arches::scheduleTimeAdvance()
- removed setRegridTimestep() call. An application should not set this flag only the regridder or scheduler.
ExplicitSolver::sched_restartInitialize()
- fixed typo.
2 lines of code changed in 2 files:
removed dup call from merge
0 lines of code changed in 1 file:
minor cleanup for access to the application
14 lines of code changed in 4 files:
white space cleanup
5 lines of code changed in 2 files:
disable debug testing for a few of the implicit tests.
3 lines of code changed in 1 file:
The solvers can now access the applicationInterface methods.
36 lines of code changed in 5 files:
Added code to toggle the isRestart switch. It's enabled in the initialization tasks.
25 lines of code changed in 2 files:
Updates to support PIDX particle output (based on work last week with
Sid and Steve). The PIDX code is passing most local (non-particle)
tests, and the particle data is being saved to some extent at this
point - but the numbers do not compare exactly and there are a few other
strange things going on. I'm committing this so that Steve and Sid
can take a look.
(Note, there are a number of debug prints in the PIDX sections that
need to be removed when the other issues are resolved.)
Updated files:
M Components/DataArchiver/DataArchiver.cc
M Ports/PIDXOutputContext.cc
M Ports/PIDXOutputContext.h
60 lines of code changed in 4 files:
Sticking with the sched_* naming convention.
25 lines of code changed in 6 files:
cleanreally, reallyclean:
Added additional system "find" commands to remove *.o and *.d.
This is a brute force approach.
11 lines of code changed in 1 file:
Arches & Wasatch:
- added calls to
m_solver -> scheduleInitialize()
m_solver -> scheduleRestartInitialize()
Hypre:
- Set the solver and preconditioner type once during initialization. There's no need to set
them every time the solver is executed.
- solver and preconditioner types are defined as enums.
- consolidate redundant conditional blocks.
- Now using switch statements instead of if()-elseif().
- formatting.
The nightly tests pass, except for those that were previously failing.
291 lines of code changed in 10 files:
Create a svn patch and pass it to the buildbot. This eliminates the annoying svn errors.
19 lines of code changed in 1 file:
Changed exception message to be more descriptive.
3 lines of code changed in 1 file:
Replaced tabs with spaces.
30 lines of code changed in 1 file:
comment out debugging output.
3 lines of code changed in 1 file:
removed unneeded interface
0 lines of code changed in 1 file:
This commit updates the code to support a newer version of PIDX, including the beginnings
of particle support. This update passes all of the local RT using UDA output, but
fails on a number of (previously passing) PIDX tests. I am committing so that Sid/Steve
can take a look and fix the issue.
Sid, here is the output from (one of) the failing test:
Running test ---methane_fire_8patch--- (mpi 8 proc) at 01:04:31
Command Line: /usr/bin/time -p /usr/installed/mpich/mpich-3.2-gcc4.9.2-install/bin/mpirun -genvlist MALLOC_STATS -np 8 sus -svnStat -svnDiff methane_fire_8patch.ups
Successfully ran to completion
Comparing udas
*** ERROR, test (methane_fire_8patch) failed uda comparison, tolerances exceeded (256)
See /ssd2/home/dav/SVN/Uintah_Baseline/opt-mpich/local_RT/ARCHES-results/methane_fire_8patch/compare_sus_runs.log.txt for more comparison information.
Changes:
--------
M CCA/Components/DataArchiver/DataArchiver.cc
M CCA/Ports/PIDXOutputContext.h
M CCA/Ports/PIDXOutputContext.cc
- Updates for newer version of PIDX with (I think) the beginnings of particle support.
- Allow for better (finer grained) specification of PIDX flags in .ups input file.
M Core/Util/XMLUtils.h
M Core/Util/XMLUtils.cc
M Core/ProblemSpec/ProblemSpec.h
M Core/ProblemSpec/ProblemSpec.cc
- Added ability to check for unsigned ints.
M StandAlone/tools/extractors/particle2tiff.cc
- Fix compiler warning. White space. Use {} on one line 'if' statements.
M StandAlone/tools/pfs/sub.mk
M StandAlone/Benchmarks/sub.mk
- Link in PIDX.
M configure.ac
M configure
- Newest PIDX requires the zfp library.
282 lines of code changed in 15 files:
formatting
55 lines of code changed in 1 file:
fix upper case issue
7 lines of code changed in 2 files:
Added option to timeout a sus command if the wall clock time exceeds some value.
usage:
<susTimeout_minutes> X </susTimeout_minutes>
77 lines of code changed in 5 files:
Updated the name of the makeMasterUda_index.csh script in the usage and documentation.
0 lines of code changed in 2 files:
- update test input filename
2 lines of code changed in 1 file:
- fix some more header-header includes on Wasatch's GPU linker issue
14 lines of code changed in 5 files:
- get rid of header-header includes with forward declarations, and do some ancillary cleanup. This ought to fix the broken GPU build. If it doesn't I'm going to need some help...
48 lines of code changed in 15 files:
- get rid of duplicate test in the wasatch python script
0 lines of code changed in 1 file:
Adding F1 for psNOx model
4 lines of code changed in 2 files:
Adding aditional parameter to psNox model
3 lines of code changed in 2 files:
- adds a block-implicit pseudotransient continuation method for compressible, multicomponent gases or compressible, calorically perfect gases, or for scalar advection-diffusion-reaction equations with a passive velocity field
- enable fixed-point with local dual time step
- add debug stream for dual time matrix info
- clean up some of the info containers and dual time maps
1332 lines of code changed in 46 files:
Removed: hypre_solver_struct:created_precond_solver & created_solver flags.
They're not needed.
16 lines of code changed in 2 files:
Getting rid of the option to not reset the grid when particles are moved.
Improving the way in which cohesive zone forces are distributed to the grid
nodes. Now using a mass*distance weighted approach, other options are left
in place (commented out) for now.
82 lines of code changed in 2 files:
Replace the compile time configurable #if 0 with an input file option to do
one or two steps in the contact.
6 lines of code changed in 1 file:
Removed the direct solver. It hasn't been used in years and is just more code to wade through....
0 lines of code changed in 6 files:
* Deleted an extra semicolon.
1 lines of code changed in 1 file:
More descriptive variable names:
sol_type -> double_type
const_type -> const_double_type
Tupes -> GridVarType
120 lines of code changed in 9 files:
Removed SolverParameters tangled web from the components/applications. The each solver "owns" the parameters
and the component/application set the parameters via the solver, solver->getParameters()->SetSomething()
- Added virtual function getParameters()
- Removed tabs and other formatting.
- Rename HypreSolver2Params -> HypreParams
- Moved CG and Direct constructors/destructors code so there is consistency in code layout between the different solvers.
360 lines of code changed in 50 files:
Turning off failing test until it can be fixed.
1 lines of code changed in 1 file:
Removing restart from RT 1GW_MFR.ups
274 lines of code changed in 1 file:
- Moved repeative code into new function called "createPopulateHypreVector()
- More explicit variable names:
x -> x_label
guess -> guess_label
A -> A_label
pointers to hypre variables
HB -> HB_p
HA -> HA_p
HX -> HX_p
pointers to solvers and preconditioners
solver -> solver_p
precond -> precond_p
200 lines of code changed in 2 files:
Fixing a bug in the soot model. Not treating intrusions properly was
causing a divergence problem for cases at scale.
70 lines of code changed in 1 file: