zhyang-dev/openfoam
1
0
Fork 0
mirror of https://develop.openfoam.com/Development/openfoam.git synced 2025-11-28 03:28:01 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

INT: Integration of Mattijs' collocated parallel IO additions

Browse Source 
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.
This commit is contained in:
Andrew Heather
2017-07-07 11:39:56 +01:00
parent 03c114010f
commit d8d6030ab6
221 changed files with 11118 additions and 1946 deletions
Download Patch File Download Diff File Expand all files Collapse all files

104
applications/utilities/parallelProcessing/decomposePar/decomposePar.C
View File

@ -104,6 +104,7 @@ Usage
#include "pointFieldDecomposer.H"
#include "lagrangianFieldDecomposer.H"
#include "decompositionModel.H"
#include "collatedFileOperation.H"
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
@ -157,13 +158,14 @@ void decomposeUniform
// Any uniform data to copy/link?
const fileName uniformDir(regionDir/"uniform");
if (isDir(runTime.timePath()/uniformDir))
if (fileHandler().isDir(runTime.timePath()/uniformDir))
{
Info<< "Detected additional non-decomposed files in "
<< runTime.timePath()/uniformDir
<< endl;
const fileName timePath = processorDb.timePath();
const fileName timePath =
fileHandler().filePath(processorDb.timePath());
// If no fields have been decomposed the destination
// directory will not have been created so make sure.
@ -171,11 +173,14 @@ void decomposeUniform
if (copyUniform || mesh.distributed())
{
cp
(
runTime.timePath()/uniformDir,
timePath/uniformDir
);
if (!fileHandler().exists(timePath/uniformDir))
{
fileHandler().cp
(
runTime.timePath()/uniformDir,
timePath/uniformDir
);
}
}
else
{
@ -189,11 +194,15 @@ void decomposeUniform
fileName currentDir(cwd());
chDir(timePath);
ln
(
parentPath/runTime.timeName()/uniformDir,
uniformDir
);
if (!fileHandler().exists(uniformDir))
{
fileHandler().ln
(
parentPath/runTime.timeName()/uniformDir,
uniformDir
);
}
chDir(currentDir);
}
}
@ -329,24 +338,10 @@ int main(int argc, char *argv[])
Info<< "\n\nDecomposing mesh " << regionName << nl << endl;
// determine the existing processor count directly
label nProcs = 0;
while
(
isDir
(
runTime.path()
/ (word("processor") + name(nProcs))
/ runTime.constant()
/ regionDir
/ polyMesh::meshSubDir
)
)
{
++nProcs;
}
// Determine the existing processor count directly
label nProcs = fileHandler().nProcs(runTime.path(), regionDir);
// get requested numberOfSubdomains. Note: have no mesh yet so
// Get requested numberOfSubdomains. Note: have no mesh yet so
// cannot use decompositionModel::New
const label nDomains = readLabel
(
@ -402,6 +397,8 @@ int main(int argc, char *argv[])
Info<< "Removing " << nProcs
<< " existing processor directories" << endl;
fileHandler().rmDir(runTime.path()/word("processors"));
// remove existing processor dirs
// reverse order to avoid gaps if someone interrupts the process
for (label proci = nProcs-1; proci >= 0; --proci)
@ -411,7 +408,7 @@ int main(int argc, char *argv[])
runTime.path()/(word("processor") + name(proci))
);
rmDir(procDir);
fileHandler().rmDir(procDir);
}
procDirsProblem = false;
@ -448,7 +445,11 @@ int main(int argc, char *argv[])
// Decompose the mesh
if (!decomposeFieldsOnly)
{
mesh.decomposeMesh();
// Disable buffering when writing mesh since we need to read
// it later on when decomposing the fields
float bufSz =
fileOperations::collatedFileOperation::maxThreadFileBufferSize;
fileOperations::collatedFileOperation::maxThreadFileBufferSize = 0;
mesh.writeDecomposition(decomposeSets);
@ -505,12 +506,16 @@ int main(int argc, char *argv[])
<< cellDist.name() << " for use in postprocessing."
<< endl;
}
fileOperations::collatedFileOperation::maxThreadFileBufferSize =
bufSz;
}
if (copyZero)
{
// Link the 0 directory into each of the processor directories
// Copy the 0 directory into each of the processor directories
fileName prevTimePath;
for (label proci = 0; proci < mesh.nProcs(); proci++)
{
Time processorDb
@ -521,14 +526,32 @@ int main(int argc, char *argv[])
);
processorDb.setTime(runTime);
if (isDir(runTime.timePath()))
if (fileHandler().isDir(runTime.timePath()))
{
const fileName timePath = processorDb.timePath();
// Get corresponding directory name (to handle processors/)
const fileName timePath
(
fileHandler().objectPath
(
IOobject
(
"",
processorDb.timeName(),
processorDb
),
word::null
)
);
Info<< "Processor " << proci
<< ": linking " << runTime.timePath() << nl
<< " to " << processorDb.timePath() << endl;
cp(runTime.timePath(), processorDb.timePath());
if (timePath != prevTimePath)
{
Info<< "Processor " << proci
<< ": copying " << runTime.timePath() << nl
<< " to " << timePath << endl;
fileHandler().cp(runTime.timePath(), timePath);
prevTimePath = timePath;
}
}
}
}
@ -629,9 +652,10 @@ int main(int argc, char *argv[])
fileNameList cloudDirs
(
readDir
fileHandler().readDir
(
runTime.timePath()/cloud::prefix, fileName::DIRECTORY
runTime.timePath()/cloud::prefix,
fileName::DIRECTORY
)
);

3
applications/utilities/parallelProcessing/decomposePar/domainDecomposition.C
View File

@ -317,9 +317,6 @@ bool Foam::domainDecomposition::writeDecomposition(const bool decomposeSets)
time().caseName()/fileName(word("processor") + Foam::name(proci))
);
// make the processor directory
mkDir(time().rootPath()/processorCasePath);
// create a database
Time processorDb
(

12
applications/utilities/parallelProcessing/decomposePar/lagrangianFieldDecomposerDecomposeFields.C
View File

@ -2,7 +2,7 @@
========= |
\\ / F ield | OpenFOAM: The Open Source CFD Toolbox
\\ / O peration |
\\ / A nd | Copyright (C) 2011-2016 OpenFOAM Foundation
\\ / A nd | Copyright (C) 2011-2017 OpenFOAM Foundation
\\/ M anipulation |
-------------------------------------------------------------------------------
License
@ -188,11 +188,12 @@ void Foam::lagrangianFieldDecomposer::decomposeFields
const PtrList<GeoField>& fields
) const
{
if (particleIndices_.size())
//if (particleIndices_.size())
{
bool valid = particleIndices_.size() > 0;
forAll(fields, fieldi)
{
decomposeField(cloudName, fields[fieldi])().write();
decomposeField(cloudName, fields[fieldi])().write(valid);
}
}
}
@ -205,11 +206,12 @@ void Foam::lagrangianFieldDecomposer::decomposeFieldFields
const PtrList<GeoField>& fields
) const
{
if (particleIndices_.size())
//if (particleIndices_.size())
{
bool valid = particleIndices_.size() > 0;
forAll(fields, fieldi)
{
decomposeFieldField(cloudName, fields[fieldi])().write();
decomposeFieldField(cloudName, fields[fieldi])().write(valid);
}
}
}