Original commit message:
------------------------
Parallel IO: New collated file format
When an OpenFOAM simulation runs in parallel, the data for decomposed fields and
mesh(es) has historically been stored in multiple files within separate
directories for each processor. Processor directories are named 'processorN',
where N is the processor number.
This commit introduces an alternative "collated" file format where the data for
each decomposed field (and mesh) is collated into a single file, which is
written and read on the master processor. The files are stored in a single
directory named 'processors'.
The new format produces significantly fewer files - one per field, instead of N
per field. For large parallel cases, this avoids the restriction on the number
of open files imposed by the operating system limits.
The file writing can be threaded allowing the simulation to continue running
while the data is being written to file. NFS (Network File System) is not
needed when using the the collated format and additionally, there is an option
to run without NFS with the original uncollated approach, known as
"masterUncollated".
The controls for the file handling are in the OptimisationSwitches of
etc/controlDict:
OptimisationSwitches
{
...
//- Parallel IO file handler
// uncollated (default), collated or masterUncollated
fileHandler uncollated;
//- collated: thread buffer size for queued file writes.
// If set to 0 or not sufficient for the file size threading is not used.
// Default: 2e9
maxThreadFileBufferSize 2e9;
//- masterUncollated: non-blocking buffer size.
// If the file exceeds this buffer size scheduled transfer is used.
// Default: 2e9
maxMasterFileBufferSize 2e9;
}
When using the collated file handling, memory is allocated for the data in the
thread. maxThreadFileBufferSize sets the maximum size of memory in bytes that
is allocated. If the data exceeds this size, the write does not use threading.
When using the masterUncollated file handling, non-blocking MPI communication
requires a sufficiently large memory buffer on the master node.
maxMasterFileBufferSize sets the maximum size in bytes of the buffer. If the
data exceeds this size, the system uses scheduled communication.
The installation defaults for the fileHandler choice, maxThreadFileBufferSize
and maxMasterFileBufferSize (set in etc/controlDict) can be over-ridden within
the case controlDict file, like other parameters. Additionally the fileHandler
can be set by:
- the "-fileHandler" command line argument;
- a FOAM_FILEHANDLER environment variable.
A foamFormatConvert utility allows users to convert files between the collated
and uncollated formats, e.g.
mpirun -np 2 foamFormatConvert -parallel -fileHandler uncollated
An example case demonstrating the file handling methods is provided in:
$FOAM_TUTORIALS/IO/fileHandling
The work was undertaken by Mattijs Janssens, in collaboration with Henry Weller.
Updated the tetrahedron and triangle classes to use the barycentric
primitives. Removed duplicate code for generating random positions in
tets and tris, and fixed bug in tri random position.
now possible with level-sets as well as planes. Removed tetPoints class
as this wasn't really used anywhere except for the old tet-cutting
routines. Restored tetPointRef.H to be consistent with other primitive
shapes. Re-wrote tet-overlap mapping in terms of the new cutting.
- use allocator class to wrap the stream pointers instead of passing
them into ISstream, OSstream and using a dynamic cast to delete
then. This is especially important if we will have a bidirectional
stream (can't delete twice!).
STYLE:
- file stream constructors with std::string (C++11)
- for rewind, explicit about in|out direction. This is not currently
important, but avoids surprises with any future bidirectional access.
- combined string streams in StringStream.H header.
Similar to <sstream> include that has both input and output string
streams.
- Remove the unused enums() method since it delivers wholly unreliable
results. It is not guaranteed to cover the full enumeration range,
but only the listed names.
- Remove the unused strings() method.
Duplicated functionality of the words(), but was never used.
- Change access of words() method from static to object.
Better code isolation. Permits the constructor to take over
as the single point of failure for bad input.
- Add values() method
- do not expose internal (HashTable) lookup since it makes it more
difficult to enforce constness and the implementation detail should
not be exposed. However leave toc() and sortedToc() for the interface.
STYLE: relocated NamedEnum under primitives (was containers)
- internal typedef as 'value_type' for some consistency with STL conventions
except turbulence and lagrangian which will also be updated shortly.
For example in the nonNewtonianIcoFoam offsetCylinder tutorial the viscosity
model coefficients may be specified in the corresponding "<type>Coeffs"
sub-dictionary:
transportModel CrossPowerLaw;
CrossPowerLawCoeffs
{
nu0 [0 2 -1 0 0 0 0] 0.01;
nuInf [0 2 -1 0 0 0 0] 10;
m [0 0 1 0 0 0 0] 0.4;
n [0 0 0 0 0 0 0] 3;
}
BirdCarreauCoeffs
{
nu0 [0 2 -1 0 0 0 0] 1e-06;
nuInf [0 2 -1 0 0 0 0] 1e-06;
k [0 0 1 0 0 0 0] 0;
n [0 0 0 0 0 0 0] 1;
}
which allows a quick change between models, or using the simpler
transportModel CrossPowerLaw;
nu0 [0 2 -1 0 0 0 0] 0.01;
nuInf [0 2 -1 0 0 0 0] 10;
m [0 0 1 0 0 0 0] 0.4;
n [0 0 0 0 0 0 0] 3;
if quick switching between models is not required.
To support this more convenient parameter specification the inconsistent
specification of seedSampleSet in the streamLine and wallBoundedStreamLine
functionObjects had to be corrected from
// Seeding method.
seedSampleSet uniform; //cloud; //triSurfaceMeshPointSet;
uniformCoeffs
{
type uniform;
axis x; //distance;
// Note: tracks slightly offset so as not to be on a face
start (-1.001 -0.05 0.0011);
end (-1.001 -0.05 1.0011);
nPoints 20;
}
to the simpler
// Seeding method.
seedSampleSet
{
type uniform;
axis x; //distance;
// Note: tracks slightly offset so as not to be on a face
start (-1.001 -0.05 0.0011);
end (-1.001 -0.05 1.0011);
nPoints 20;
}
which also support the "<type>Coeffs" form
// Seeding method.
seedSampleSet
{
type uniform;
uniformCoeffs
{
axis x; //distance;
// Note: tracks slightly offset so as not to be on a face
start (-1.001 -0.05 0.0011);
end (-1.001 -0.05 1.0011);
nPoints 20;
}
}
- the purpose is more explicit, without needing to check documentation
about what the bool parameter means.
STYLE: improve formatting of fileName documentation
- The code create a box with a (0,0,0) point.
The new definition is more logical and makes it very easy to grow
the bounding box to include new points. It also simplifies much of
the logic in the constructors.
- Use ROOTVGREAT instead of VGREAT for sizing greatBox and invertedBox.
Avoids some overflow issues reported by Mattijs (thus GREAT has been
used in treeBoundBox), but might still need further revision.
- Constructor for bounding box of a single point.
- add(boundBox), add(point) ...
-> Extend box to enclose the second box or point(s).
Eg,
bb.add(pt);
vs.
bb.min() = Foam::min(bb.min(), pt);
bb.max() = Foam::max(bb.max(), pt);
Also works with other bounding boxes.
Eg,
bb.add(bb2);
// OR
bb += bb2;
vs.
bb.min() = Foam::min(bb.min(), bb2.min());
bb.max() = Foam::max(bb.max(), bb2.max());
'+=' operator allows the reduction to be used in parallel
gather/scatter operations.
A global '+' operator is not currently needed.
Note: may be useful in the future to have a 'clear()' method
that resets to a zero-sized (inverted) box.
STYLE: make many bounding box constructors explicit
By default snappyHexMesh writes files relating to the hex-splitting process into
the polyMesh directory: cellLevel level0Edge pointLevel surfaceIndex
but by setting the noRefinement flag:
writeFlags
(
noRefinement
.
.
.
);
these optional files which are generally not needed are not written.
If you run the three stages of snappyHexMesh separately or run a dynamic mesh
solver supporting refinement and unrefinement these files are needed
and "noRefinement" should not be set.
unless the blockMeshDict is in the polyMesh directory or the "-noClean" option
is specified.
This avoids problems running snappyHexMesh without first clearing files from
polyMesh which interfere with the operation of snappyHexMesh.
e.g. in tutorials/heatTransfer/buoyantSimpleFoam/externalCoupledCavity/0/T
hot
{
type externalCoupledTemperature;
commsDir "${FOAM_CASE}/comms";
file "data";
initByExternal yes;
log true;
value uniform 307.75; // 34.6 degC
}
Previously both 'file' and 'fileName' were used inconsistently in different
classes and given that there is no confusion or ambiguity introduced by using
the simpler 'file' rather than 'fileName' this change simplifies the use and
maintenance of OpenFOAM.
- Could be related to interrupted builds.
So if there are any parts of the build that rely on an explicit
'wmakeLnInclude', make sure that the contents are properly updated.
--
ENH: improved feedback from top-level Allwmake
- Report which section (libraries, applications) is being built.
- Provide final summary of date, version, etc, which can be helpful
for later diagnosis or record keeping.
- The -log=XXX option for Allwmake now accepts a directory name
and automatically appends an appropriate log name.
Eg,
./Allwmake -log=logs/ ->> logs/log.linux64GccDPInt32Opt
The default name is built from the value of WM_OPTIONS.
--
BUG: shell not exiting properly in combination with -log option
- the use of 'tee' causes the shell to hang around.
Added an explicit exit to catch this.
--
- Detecting the '-k' (-non-stop) option at the top-level Allwmake, which
may improve robustness.
- Explicit continue-on-error for foamyMesh (as optional component)
- unify format of script messages for better readability
COMP: reduce warnings when building Pstream (old-style casts in openmpi)
- only occurs in combination with distributedTriSurfaceMesh in snappy.
- workaround similar to that previously used for surfaceRedistributePar
(issue #60).
Minor adjustment of incompressible motorBike tutorial to detect use of
distributedTriSurfaceMesh.
