October 2014 Commit Log

Number of Commits:
71
Number of Active Developers:
11
harman 2014-10-31 13:57 Rev.: 52621

DataArchive:reduceUda_ReadUda()

-added logic to ignore a variable if a process doesn't own the patch
that the variable resides on.

20 lines of code changed in 4 files:

u0407585 2014-10-31 13:28 Rev.: 52620

Fixed Char oxidation activation energy sign error. This will break the coal regression testssvn --helpsvn --help

7 lines of code changed in 1 file:

harman 2014-10-31 08:20 Rev.: 52619

Include pressure and viscous forces in the analysis.

- added octave script to analyze the output from puda -ICE_momentum

511 lines of code changed in 5 files:

awabboud 2014-10-30 10:33 Rev.: 52616

Minor fixes to particle sources

3 lines of code changed in 2 files:

jthornoc 2014-10-30 10:29 Rev.: 52615

Clean up a memory leak if the task isnt scheduled.

2 lines of code changed in 1 file:

harman 2014-10-29 14:47 Rev.: 52614

Updated url address.

Changed all instances of SCIRun -> Uintah

Fixed bug if --no_compile option is used don't compile the code but run the tests.

0 lines of code changed in 4 files:

harman 2014-10-29 14:30 Rev.: 52613

replaced tabs with spaces.

Reformatted output when SCIDebug componentTimings:+ is turned on. It's easier to parse and
added a descriptive header.

57 lines of code changed in 1 file:

jthornoc 2014-10-29 13:45 Rev.: 52612

Clean up an issue with the new ignore_invalid_celltype flag.

Also, corners for periodic domains were causing problems. Initialize
these cells to something valid since their value shouldnt matter in the
end.

120 lines of code changed in 5 files:

jthornoc 2014-10-28 09:29 Rev.: 52611

Adding a failure point if invalid cell types are found. One can turn
this off from the input file using the ignore_invalid_celltype from the
ups file.

38 lines of code changed in 3 files:

jthornoc 2014-10-28 08:49 Rev.: 52610

new -> scinew for masks. Still trying to track down the double delete
in SpatialOps for masks.

Try not inserting the mask into the map held by the local mask struct.

0 lines of code changed in 2 files:

amir 2014-10-24 18:10 Rev.: 52607

Fixing a cleaving issue for variable density cases, which was missed before. Adding a 3 dimensional variable density case to the input files.

9 lines of code changed in 1 file:

awabboud 2014-10-24 10:31 Rev.: 52606

supress warning output in cqmom unless turned on to debug

0 lines of code changed in 2 files:

ahumphrey 2014-10-23 17:09 Rev.: 52605

Reanimate per level suppport for maxGhostCells and maxLevelOffset. These are the horizontal and vertical ranges (respectively) considered by the loadbalancer when creating a neighborhood.

Also clean up some compiler warnings in Relocate.cc

69 lines of code changed in 9 files:

ahumphrey 2014-10-22 19:14 Rev.: 52604

More sophisticated verifyChecksum() method for the Unified Scheduler. This will handle the presence of "spatial tasks", e.g. radiometer

20 lines of code changed in 1 file:

ahumphrey 2014-10-22 18:43 Rev.: 52603

Differentiate timingsStats.{min,max,avg} files by MPI_Comm sizes. This way we can gather stats for multiple sized jobs run from the same directory.

8 lines of code changed in 1 file:

ahumphrey 2014-10-22 17:54 Rev.: 52602

Add Mutex to Unified Scheduler for thread-safe MPI stats (e.g. wait times, etc).
Some misc. code cleanup.

23 lines of code changed in 3 files:

jthornoc 2014-10-22 16:22 Rev.: 52601

Fixing a sign error with the swirl BC.

6 lines of code changed in 1 file:

ahumphrey 2014-10-22 12:04 Rev.: 52600

Add significant functionality to MPI stats.
Cleanup and formatting in Unified and MPI schedulers.

559 lines of code changed in 5 files:

jthornoc 2014-10-21 09:51 Rev.: 52599

Dont store masks for use later just yet until the leak/threaded test
problems are sorted.

1 lines of code changed in 1 file:

harman 2014-10-20 16:56 Rev.: 52598

Formatting:
- Added {} and method dividers.

120 lines of code changed in 1 file:

harman 2014-10-20 15:01 Rev.: 52597

formatting:
- added {}, method dividers and spacing to improve readability.

106 lines of code changed in 1 file:

harman 2014-10-20 13:56 Rev.: 52596

Example component can compute just the fluxes for the radiometers.
This includes both spatial and temporal scheduling of the task.
It only runs on 1 mpi process due to a problem upstream in
reduceSysVars()

- added regression test.



50 lines of code changed in 3 files:

jsutherland 2014-10-20 13:32 Rev.: 52595

Stylistic consistency changes.

64 lines of code changed in 1 file:

harman 2014-10-20 12:46 Rev.: 52594

Update the example component for the new radiometer. This does not include spatial scheduling of
the radiometer.

29 lines of code changed in 4 files:

jthornoc 2014-10-17 14:41 Rev.: 52593

Make the user specify a velocity vector for the default handoff value.

12 lines of code changed in 1 file:

jthornoc 2014-10-17 14:41 Rev.: 52592

Fixing a bug when reassigning handoff normal index components. The
tangential velocities weren't being reassigned properly.

96 lines of code changed in 1 file:

dav 2014-10-17 11:18 Rev.: 52589

M Makefile.in
M Core/Util/sub.mk

Fix a strange linking error on Prism@CHPC (linux box) where the
tracker program would not link correctly due to missing symbols in
Core/Util (which are found in Core/Malloc) when SCI Malloc is turned
on. While this fix is valid (Core/Util does rely on Core/Malloc), it
should not be necessary as Core/Malloc is linked in other locations.

2 lines of code changed in 2 files:

tsaad 2014-10-17 09:22 Rev.: 52586

silence a few warnings

0 lines of code changed in 2 files:

dav 2014-10-16 18:14 Rev.: 52585

Fix GPU compilation error... Todd please verify this.

0 lines of code changed in 2 files:

dav 2014-10-16 17:48 Rev.: 52584


This fixes the lineextract problem seen on BGQ machine vulcan@llnl.
However, it is likely that this same problem will/does occur on any
strange statically linked architecture. While this isn't a complete
fix, it will allow Jennifer to use the utility for now.

The problem in this case is that static variables are not always
(ever?) created (or at least their constructors are not called) upon
program startup (like they are supposed to be). The Uintah type
system creates a list of all variable types (eg: CCVariable<double>)
that are available to the program, and these are stored in the static
map variables 'types' found in Core/Disclosure/TypeDescription.cc.
However, these types are put into the 'types' variable by the
register_type() function which is called from the TypeDescription
constructor for each type.

Later, when a (textual) type decription is read from an UDA file, that
string is looked up in the 'types' map... If the map was never populated,
then an 'unknown type' type is returned and then an error occurs because
we don't know the real type.

To solve this problem, I just manually instantiate a number of
variable types in the bgq_hack() function in lineextract.cc.

M StandAlone/tools/extractors/lineextract.cc

- Line up variables (and their initialization) for easier viewing.
- Add white space, remove tabs and indent properly.
- Number of coding standard stylistic updates.
- Added bgq_hack() function to manually instantiate needed types.
- Added in warning message in case a needed type hasn't been hacked in yet.

M StandAlone/tools/extractors/timeextract.cc

- White space fixes.
- This file has not been 'hacked' like lineextract, but if it is needed, the
same hack would be easy to copy over.

M Core/DataArchive/DataArchive.cc
M Core/Disclosure/TypeDescription.cc

- Cosmetic updates - white space and coding standard clean ups.



278 lines of code changed in 7 files:

harman 2014-10-16 16:49 Rev.: 52583

updates so RMCRT:float will compile and run

14 lines of code changed in 3 files:

derekhar 2014-10-15 12:59 Rev.: 52582

Fixed swirl so that extra cell values are properly set.

2 lines of code changed in 1 file:

derekhar 2014-10-15 11:03 Rev.: 52581

Fixed bug in swirl boundary conditions when swirl is set on z faces.

3 lines of code changed in 1 file:

tsaad 2014-10-15 09:20 Rev.: 52580

update the spec for reading data files. specify the format in the data file instead of forcing the user to specify it in the ups file. We now support FLAT and XYZ formats. You must specify those (e.g. FLAT or XYZ) in the first line of your input file.

17 lines of code changed in 2 files:

dav 2014-10-14 16:38 Rev.: 52579

Needed to convert the ostringstream to a string...

3 lines of code changed in 1 file:

dav 2014-10-14 16:34 Rev.: 52578


M HyprePrecondBase.cc

- Alphabetize #includes.

- Fix warning message on Mac where a char * is concatenated to an
integer using a +... using a ostringstream instead.


0 lines of code changed in 2 files:

dav 2014-10-14 16:32 Rev.: 52577


M Makefile.in

The Mac is a little temperamental with the SVN info command... this fixes it.

M StandAlone/sub.mk

Removed tabs from incorrect locations.


23 lines of code changed in 3 files:

dav 2014-10-14 15:19 Rev.: 52576


Two main fixes: Checkpointing (parsing of UPS file) fixed. Outputting
of correct SVN revision number for sus runs.

M Makefile.in
A + include/svn_info.h.template
D include/svn_info.h

Place accurate SVN information in svn_info.h each time 'make' is run.
This uses the 'svn info' command and populates the svn_info.h file
(bin/include/) using the svn_info.h.template file (src/include/).

M StandAlone/sub.mk

- sus.o depends on the svn_info.h file being created.

M CCA/Components/DataArchiver/DataArchiver.cc

- Fix checkpointing. Now both walltimeInterval and walltimeIntervalHours work.

- Updated the checkpointing code to try to open the checkpoint file
multiple times if the 'open' system call fails. This issues was
seen (at least) once on vulcan@llnl (bgq machine), and probably
occurred due to a file system glitch. It is likely that this 'fix'
will never be utilized, but it is there just in case.

57 lines of code changed in 5 files:

tsaad 2014-10-14 11:59 Rev.: 52575

fix warches

10 lines of code changed in 1 file:

tsaad 2014-10-14 09:32 Rev.: 52572

change all initial conditions to STATE_NONE. This is a major refactor. If you are using Wasatch, please change all your initial conditions to STATE_NONE. This commit constitutes the first step towards removing state-specification from input files.

32 lines of code changed in 18 files:

jthornoc 2014-10-14 08:56 Rev.: 52571

Switching over to a boost shared_ptr for the mask storage to better
manage memory. Still seeing a little leak here...investigating....

17 lines of code changed in 2 files:

jsutherland 2014-10-13 20:35 Rev.: 52570

Misc. cleanup.

19 lines of code changed in 5 files:

mhomel 2014-10-13 17:19 Rev.: 52569

Added check to avoid state variable iterations when no porosity or fluid effects are present

0 lines of code changed in 2 files:

ahumphrey 2014-10-13 15:11 Rev.: 52568

* Fix static build for IPM library support. For dynamic linking with IPM, report configure message about using LD_PRELOAD
* Generate new configure script
* No headers for IPM, so removed #include from ipm_testdefs.h.in

* Remove a few straggler pieces of windows support in build_scripts directory

43 lines of code changed in 12 files:

awabboud 2014-10-13 12:34 Rev.: 52567

Add a constant source term to the particle models.

243 lines of code changed in 5 files:

ahumphrey 2014-10-10 10:35 Rev.: 52566

Add build system support for IPM profiling library (http://ipm-hpc.org/).
Also generate new configure script.

638 lines of code changed in 4 files:

tsaad 2014-10-09 13:05 Rev.: 52565

add support for reading flat-format initial conditions.

111 lines of code changed in 2 files:

jthornoc 2014-10-08 10:26 Rev.: 52561

Reverse the last commit with a better solution.
The inlet face velocities are no longer being "RK averaged" which was,
in some cases, causing the boundary to lose information.

19 lines of code changed in 3 files:

ahumphrey 2014-10-08 09:58 Rev.: 52560

Fix MAC-specific gcc compiler error in Thread::numProcessors() - "error: taking address of temporary array".
It appears gcc doesn't like the fact that tparams is not bound outside of the sysctl call.

0 lines of code changed in 2 files:

jthornoc 2014-10-07 22:23 Rev.: 52559

Because of the RK averaging and the staggered mesh, the inlet BCs when
read from file must be reset.

2 lines of code changed in 1 file:

jsutherland 2014-10-06 13:55 Rev.: 52556

Turn off one GPU test.

2 lines of code changed in 1 file:

tsaad 2014-10-06 10:41 Rev.: 52555

fix wasatch restarts with particles and particle boundary conditions

30 lines of code changed in 2 files:

ahumphrey 2014-10-05 22:16 Rev.: 52553

* Cleanup CUDA stream creation flow (we should only need 1 stream per GPU task with aggregated H2D & D2H copies).
* Report more & better info for GPUStats DebugStream.
* Add temporary fix for Wasatch case where a field exists on the GPU, but requires a ghost update.
* This should allow James to put some more GPU regression testing in place.
* Some minor cleanup in a few spots

128 lines of code changed in 5 files:

jsutherland 2014-10-04 20:44 Rev.: 52551

Fix an assertion.

0 lines of code changed in 2 files:

jsutherland 2014-10-04 19:41 Rev.: 52550

Remove a superfluous argument to wrap_uintah_field_as_spatialops

10 lines of code changed in 1 file:

ahumphrey 2014-10-03 18:26 Rev.: 52549

Report more meaningful information via GPUStats DebugStream.

21 lines of code changed in 1 file:

dav 2014-10-03 16:14 Rev.: 52547

Fix walltimeStartHours too.

6 lines of code changed in 1 file:

u0407585 2014-10-03 16:04 Rev.: 52546

made wallTimeInterval work again

1 lines of code changed in 1 file:

tsaad 2014-10-03 10:07 Rev.: 52542

Pass the ParticleSubset through the AllocInfo struct. This is consistent with the idea of passing a patch - since the ParticleSubset is to Particles what a Patch is to an Eulerian field. This will fix the threaded Wasatch builds. Thanks to Nathan Yonkee for helping to track this down.

27 lines of code changed in 2 files:

dav 2014-10-02 17:56 Rev.: 52539

(And a few more...) Sub.mks: for macs, need to link in Core/Thread for these libraries. Also alphabetized, removed tabs, and cleaned up.

39 lines of code changed in 5 files:

dav 2014-10-02 17:48 Rev.: 52538

Sub.mks: for macs, need to link in Core/Thread for these libraries. Also alphabetized, removed tabs, and cleaned up.

0 lines of code changed in 8 files:

dav 2014-10-02 17:07 Rev.: 52537

Revert debug version of file that was accidentally committed.

0 lines of code changed in 2 files:

dav 2014-10-02 17:02 Rev.: 52536

Reverting accidental commit of this file.

0 lines of code changed in 1 file:

dav 2014-10-02 16:56 Rev.: 52535

- Cleanups from Vulcan@LLNL. Mostly cosmetic, though a few real fixes
found while porting to BGQ machine.

- Replaced proc0thread0cerr with proc0cerr as proc0cerr is (now
specifically) defined to only print out from thread 0 of process 0.

M CCA/Components/SimulationController/AMRSimulationController.cc

- Cosmetic
- Alphabetize the #includes for better readability and to remove duplicates.
- Put the #ifdef USE_GPERFTOOLS around entire block, instead of having
it multiple times and leaving dead if statements around it.

M CCA/Components/SimulationController/SimulationController.cc

- Cosmetic: Fix indentation (use diff -b to better see differences.)

M CCA/Components/DataArchiver/DataArchiver.cc

- Added the ability to specify checkpoint time via hours (in addition to seconds).
- Cleaned up the code to make it more readable for parsing this
information from input file - ie, use variable names that correspond
to the data they hold.
- Commented out a few lines of code that didn't do anything.
- Indented for readability, lined up some variable assignments.

M CCA/Components/Arches/CompDynamicProcedure.cc

- Alphabetize the #includes for better readability and to remove duplicates.
- Replaced proc0thread0cerr with proc0cerr.

M CCA/Components/Arches/DQMOM.cc
M CCA/Components/Arches/PropertyModels/PropertyModelBase.h
M CCA/Components/Arches/IncDynamicProcedure.cc
M CCA/Components/Arches/Arches.cc

- Replaced proc0thread0cerr with proc0cerr.

M CCA/Components/Arches/ChemMix/ClassicTableInterface.cc

- Throw a ProblemSetupException as that is what the error is.

M Core/Exceptions/InvalidState.cc

- Propagate (user supplied) error message.

M Core/Grid/Patch.cc

- Whitespace cleanup for readability.

M Core/Grid/Grid.h

- Whitespace cleanup for readability.
- Parameter to setExtraCells() is now (explicitly) const.

M Core/Grid/Level.cc
M Core/Grid/Grid.cc

- Coding standard and whitespace cleanup.

M Core/Parallel/Parallel.h

- Updated proc0cout to handle threaded versions of Uintah. If in MPI
only mode, then only prints for proc 0. If in threaded mode, only
prints for proc 0 thread 0.
- Removed proc0thread0cout as now only proc0cout is needed.

M Core/Thread/ThreadPool.h

- Remove tabs, line up code.

M Core/Thread/Thread.h
M Core/Thread/Parallel.h
M testprograms/Thread/parallel.cc

- Added the ThreadNS namespace because there are two classes named
'Parallel' in Uintah, and the one in the new ThreadNS is only used
in a few locations and thus is easier to move into its own NS.

M StandAlone/inputs/UPS_SPEC/ups_spec.xml

- Added 'walltimeStartHours' and 'walltimeIntervalHours' (to augment the
version that uses just seconds).
- Lined up for better readability.

M tools/StackTrace/StackTrace.cc

- Fixes for the stack trace tool for machines that only provide symbol
numbers in their stack traces. However, unfortunately, this still
doesn't seem to provide a valid stack trace on Vulcan.

290 lines of code changed in 31 files:

mhomel 2014-10-02 16:43 Rev.: 52534

Modified the step division, and improved the algorithm for computing the interior point in nonHardeningReturn

134 lines of code changed in 3 files:

dav 2014-10-02 16:14 Rev.: 52532

Updated configure to match configure.ac from previous commit.

100 lines of code changed in 3 files:

dav 2014-10-02 16:13 Rev.: 52531

I am not committing configure at this time as this work was done on a
computer without a new enough version of autoconf. I'll create the
configure soon and commit it. -Dd

M configure.ac

- Fix configure to handle finding fortran library correctly
(specifically when using mpif77). Previously it assumed the gcc
compiler with mpif77, now it checks for intel compiler before
defaulting to gcc.

- Require hypre for Arches and Wasatch now.

- Removed some empty code.

- Properly indented a section.

- Fixed a few AC_MSG_WARN so that the ',' in the output (and following
text) would actually display correctly.

39 lines of code changed in 1 file:

awabboud 2014-10-02 15:17 Rev.: 52530

Get rid of a CQMOM tag and generate values automatically

4 lines of code changed in 1 file:

jthornoc 2014-10-01 22:22 Rev.: 52528

Revert the ability to have tabulated BCs with handoff files. This isn't
a valid option without a bit more work on the tabulated side.

More bug fixing for tabulated BCs.

Add input files for a new RT test.

9 lines of code changed in 3 files:

tsaad 2014-10-01 15:16 Rev.: 52527

* remove a cout statement
* Use the postprocessing graph to compute the cell centered momentum field for one of the Wasatch tests. This will fix the threaded build on this test. Thanks to Nathan Yonkee for his work on this issue.

0 lines of code changed in 2 files:

tsaad 2014-10-01 09:48 Rev.: 52525

* Fix another leak in the recently added interior boundaries
* Allow PostProcessing graphs to be used through the input file

5 lines of code changed in 3 files:

September 2014 »

Generated by StatSVN 0.7.0