d8d6030ab6
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.
504 lines
13 KiB
C
504 lines
13 KiB
C
/*---------------------------------------------------------------------------*\
|
|
========= |
|
|
\\ / F ield | OpenFOAM: The Open Source CFD Toolbox
|
|
\\ / O peration |
|
|
\\ / A nd | Copyright (C) 2015-2017 OpenFOAM Foundation
|
|
\\/ M anipulation | Copyright (C) 2015-2016 OpenCFD Ltd.
|
|
-------------------------------------------------------------------------------
|
|
License
|
|
This file is part of OpenFOAM.
|
|
|
|
OpenFOAM is free software: you can redistribute it and/or modify it
|
|
under the terms of the GNU General Public License as published by
|
|
the Free Software Foundation, either version 3 of the License, or
|
|
(at your option) any later version.
|
|
|
|
OpenFOAM is distributed in the hope that it will be useful, but WITHOUT
|
|
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
|
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
|
for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with OpenFOAM. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
\*---------------------------------------------------------------------------*/
|
|
|
|
#include "checkTools.H"
|
|
#include "polyMesh.H"
|
|
#include "globalMeshData.H"
|
|
#include "hexMatcher.H"
|
|
#include "wedgeMatcher.H"
|
|
#include "prismMatcher.H"
|
|
#include "pyrMatcher.H"
|
|
#include "tetWedgeMatcher.H"
|
|
#include "tetMatcher.H"
|
|
#include "IOmanip.H"
|
|
#include "pointSet.H"
|
|
#include "faceSet.H"
|
|
#include "cellSet.H"
|
|
#include "Time.H"
|
|
#include "surfaceWriter.H"
|
|
#include "syncTools.H"
|
|
#include "globalIndex.H"
|
|
#include "PatchTools.H"
|
|
|
|
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
|
|
|
|
void Foam::printMeshStats(const polyMesh& mesh, const bool allTopology)
|
|
{
|
|
Info<< "Mesh stats" << nl
|
|
<< " points: "
|
|
<< returnReduce(mesh.points().size(), sumOp<label>()) << nl;
|
|
|
|
label nInternalPoints = returnReduce
|
|
(
|
|
mesh.nInternalPoints(),
|
|
sumOp<label>()
|
|
);
|
|
|
|
if (nInternalPoints != -Pstream::nProcs())
|
|
{
|
|
Info<< " internal points: " << nInternalPoints << nl;
|
|
|
|
if (returnReduce(mesh.nInternalPoints(), minOp<label>()) == -1)
|
|
{
|
|
WarningInFunction
|
|
<< "Some processors have their points sorted into internal"
|
|
<< " and external and some do not." << endl
|
|
<< "This can cause problems later on." << endl;
|
|
}
|
|
}
|
|
|
|
if (allTopology && nInternalPoints != -Pstream::nProcs())
|
|
{
|
|
label nEdges = returnReduce(mesh.nEdges(), sumOp<label>());
|
|
label nInternalEdges = returnReduce
|
|
(
|
|
mesh.nInternalEdges(),
|
|
sumOp<label>()
|
|
);
|
|
label nInternal1Edges = returnReduce
|
|
(
|
|
mesh.nInternal1Edges(),
|
|
sumOp<label>()
|
|
);
|
|
label nInternal0Edges = returnReduce
|
|
(
|
|
mesh.nInternal0Edges(),
|
|
sumOp<label>()
|
|
);
|
|
|
|
Info<< " edges: " << nEdges << nl
|
|
<< " internal edges: " << nInternalEdges << nl
|
|
<< " internal edges using one boundary point: "
|
|
<< nInternal1Edges-nInternal0Edges << nl
|
|
<< " internal edges using two boundary points: "
|
|
<< nInternalEdges-nInternal1Edges << nl;
|
|
}
|
|
|
|
label nFaces = returnReduce(mesh.faces().size(), sumOp<label>());
|
|
label nIntFaces = returnReduce(mesh.faceNeighbour().size(), sumOp<label>());
|
|
label nCells = returnReduce(mesh.cells().size(), sumOp<label>());
|
|
|
|
Info<< " faces: " << nFaces << nl
|
|
<< " internal faces: " << nIntFaces << nl
|
|
<< " cells: " << nCells << nl
|
|
<< " faces per cell: "
|
|
<< scalar(nFaces + nIntFaces)/max(1, nCells) << nl
|
|
<< " boundary patches: " << mesh.boundaryMesh().size() << nl
|
|
<< " point zones: " << mesh.pointZones().size() << nl
|
|
<< " face zones: " << mesh.faceZones().size() << nl
|
|
<< " cell zones: " << mesh.cellZones().size() << nl
|
|
<< endl;
|
|
|
|
// Construct shape recognizers
|
|
hexMatcher hex;
|
|
prismMatcher prism;
|
|
wedgeMatcher wedge;
|
|
pyrMatcher pyr;
|
|
tetWedgeMatcher tetWedge;
|
|
tetMatcher tet;
|
|
|
|
// Counters for different cell types
|
|
label nHex = 0;
|
|
label nWedge = 0;
|
|
label nPrism = 0;
|
|
label nPyr = 0;
|
|
label nTet = 0;
|
|
label nTetWedge = 0;
|
|
label nUnknown = 0;
|
|
|
|
Map<label> polyhedralFaces;
|
|
|
|
for (label celli = 0; celli < mesh.nCells(); celli++)
|
|
{
|
|
if (hex.isA(mesh, celli))
|
|
{
|
|
nHex++;
|
|
}
|
|
else if (tet.isA(mesh, celli))
|
|
{
|
|
nTet++;
|
|
}
|
|
else if (pyr.isA(mesh, celli))
|
|
{
|
|
nPyr++;
|
|
}
|
|
else if (prism.isA(mesh, celli))
|
|
{
|
|
nPrism++;
|
|
}
|
|
else if (wedge.isA(mesh, celli))
|
|
{
|
|
nWedge++;
|
|
}
|
|
else if (tetWedge.isA(mesh, celli))
|
|
{
|
|
nTetWedge++;
|
|
}
|
|
else
|
|
{
|
|
nUnknown++;
|
|
polyhedralFaces(mesh.cells()[celli].size())++;
|
|
}
|
|
}
|
|
|
|
reduce(nHex,sumOp<label>());
|
|
reduce(nPrism,sumOp<label>());
|
|
reduce(nWedge,sumOp<label>());
|
|
reduce(nPyr,sumOp<label>());
|
|
reduce(nTetWedge,sumOp<label>());
|
|
reduce(nTet,sumOp<label>());
|
|
reduce(nUnknown,sumOp<label>());
|
|
|
|
Info<< "Overall number of cells of each type:" << nl
|
|
<< " hexahedra: " << nHex << nl
|
|
<< " prisms: " << nPrism << nl
|
|
<< " wedges: " << nWedge << nl
|
|
<< " pyramids: " << nPyr << nl
|
|
<< " tet wedges: " << nTetWedge << nl
|
|
<< " tetrahedra: " << nTet << nl
|
|
<< " polyhedra: " << nUnknown
|
|
<< endl;
|
|
|
|
if (nUnknown > 0)
|
|
{
|
|
Pstream::mapCombineGather(polyhedralFaces, plusEqOp<label>());
|
|
|
|
Info<< " Breakdown of polyhedra by number of faces:" << nl
|
|
<< " faces" << " number of cells" << endl;
|
|
|
|
const labelList sortedKeys = polyhedralFaces.sortedToc();
|
|
|
|
forAll(sortedKeys, keyi)
|
|
{
|
|
const label nFaces = sortedKeys[keyi];
|
|
|
|
Info<< setf(std::ios::right) << setw(13)
|
|
<< nFaces << " " << polyhedralFaces[nFaces] << nl;
|
|
}
|
|
}
|
|
|
|
Info<< endl;
|
|
}
|
|
|
|
|
|
void Foam::mergeAndWrite
|
|
(
|
|
const polyMesh& mesh,
|
|
const surfaceWriter& writer,
|
|
const word& name,
|
|
const indirectPrimitivePatch setPatch,
|
|
const fileName& outputDir
|
|
)
|
|
{
|
|
if (Pstream::parRun())
|
|
{
|
|
labelList pointToGlobal;
|
|
labelList uniqueMeshPointLabels;
|
|
autoPtr<globalIndex> globalPoints;
|
|
autoPtr<globalIndex> globalFaces;
|
|
faceList mergedFaces;
|
|
pointField mergedPoints;
|
|
Foam::PatchTools::gatherAndMerge
|
|
(
|
|
mesh,
|
|
setPatch.localFaces(),
|
|
setPatch.meshPoints(),
|
|
setPatch.meshPointMap(),
|
|
|
|
pointToGlobal,
|
|
uniqueMeshPointLabels,
|
|
globalPoints,
|
|
globalFaces,
|
|
|
|
mergedFaces,
|
|
mergedPoints
|
|
);
|
|
|
|
// Write
|
|
if (Pstream::master())
|
|
{
|
|
writer.write
|
|
(
|
|
outputDir,
|
|
name,
|
|
meshedSurfRef
|
|
(
|
|
mergedPoints,
|
|
mergedFaces
|
|
)
|
|
);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
writer.write
|
|
(
|
|
outputDir,
|
|
name,
|
|
meshedSurfRef
|
|
(
|
|
setPatch.localPoints(),
|
|
setPatch.localFaces()
|
|
)
|
|
);
|
|
}
|
|
}
|
|
|
|
|
|
void Foam::mergeAndWrite
|
|
(
|
|
const surfaceWriter& writer,
|
|
const faceSet& set
|
|
)
|
|
{
|
|
const polyMesh& mesh = refCast<const polyMesh>(set.db());
|
|
|
|
const indirectPrimitivePatch setPatch
|
|
(
|
|
IndirectList<face>(mesh.faces(), set.sortedToc()),
|
|
mesh.points()
|
|
);
|
|
|
|
fileName outputDir
|
|
(
|
|
set.time().path()
|
|
/ (Pstream::parRun() ? ".." : "")
|
|
/ "postProcessing"
|
|
/ mesh.pointsInstance()
|
|
/ set.name()
|
|
);
|
|
outputDir.clean();
|
|
|
|
mergeAndWrite(mesh, writer, set.name(), setPatch, outputDir);
|
|
}
|
|
|
|
|
|
void Foam::mergeAndWrite
|
|
(
|
|
const surfaceWriter& writer,
|
|
const cellSet& set
|
|
)
|
|
{
|
|
const polyMesh& mesh = refCast<const polyMesh>(set.db());
|
|
const polyBoundaryMesh& pbm = mesh.boundaryMesh();
|
|
|
|
|
|
// Determine faces on outside of cellSet
|
|
PackedBoolList isInSet(mesh.nCells());
|
|
forAllConstIter(cellSet, set, iter)
|
|
{
|
|
isInSet[iter.key()] = true;
|
|
}
|
|
|
|
|
|
boolList bndInSet(mesh.nFaces()-mesh.nInternalFaces());
|
|
forAll(pbm, patchi)
|
|
{
|
|
const polyPatch& pp = pbm[patchi];
|
|
const labelList& fc = pp.faceCells();
|
|
forAll(fc, i)
|
|
{
|
|
bndInSet[pp.start()+i-mesh.nInternalFaces()] = isInSet[fc[i]];
|
|
}
|
|
}
|
|
syncTools::swapBoundaryFaceList(mesh, bndInSet);
|
|
|
|
|
|
DynamicList<label> outsideFaces(3*set.size());
|
|
for (label facei = 0; facei < mesh.nInternalFaces(); facei++)
|
|
{
|
|
bool ownVal = isInSet[mesh.faceOwner()[facei]];
|
|
bool neiVal = isInSet[mesh.faceNeighbour()[facei]];
|
|
|
|
if (ownVal != neiVal)
|
|
{
|
|
outsideFaces.append(facei);
|
|
}
|
|
}
|
|
|
|
|
|
forAll(pbm, patchi)
|
|
{
|
|
const polyPatch& pp = pbm[patchi];
|
|
const labelList& fc = pp.faceCells();
|
|
if (pp.coupled())
|
|
{
|
|
forAll(fc, i)
|
|
{
|
|
label facei = pp.start()+i;
|
|
|
|
bool neiVal = bndInSet[facei-mesh.nInternalFaces()];
|
|
if (isInSet[fc[i]] && !neiVal)
|
|
{
|
|
outsideFaces.append(facei);
|
|
}
|
|
}
|
|
}
|
|
else
|
|
{
|
|
forAll(fc, i)
|
|
{
|
|
if (isInSet[fc[i]])
|
|
{
|
|
outsideFaces.append(pp.start()+i);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
const indirectPrimitivePatch setPatch
|
|
(
|
|
IndirectList<face>(mesh.faces(), outsideFaces),
|
|
mesh.points()
|
|
);
|
|
|
|
fileName outputDir
|
|
(
|
|
set.time().path()
|
|
/ (Pstream::parRun() ? ".." : "")
|
|
/ "postProcessing"
|
|
/ mesh.pointsInstance()
|
|
/ set.name()
|
|
);
|
|
outputDir.clean();
|
|
|
|
mergeAndWrite(mesh, writer, set.name(), setPatch, outputDir);
|
|
}
|
|
|
|
|
|
void Foam::mergeAndWrite
|
|
(
|
|
const writer<scalar>& writer,
|
|
const pointSet& set
|
|
)
|
|
{
|
|
const polyMesh& mesh = refCast<const polyMesh>(set.db());
|
|
|
|
pointField mergedPts;
|
|
labelList mergedIDs;
|
|
|
|
if (Pstream::parRun())
|
|
{
|
|
// Note: we explicitly do not merge the points
|
|
// (mesh.globalData().mergePoints etc) since this might
|
|
// hide any synchronisation problem
|
|
|
|
globalIndex globalNumbering(mesh.nPoints());
|
|
|
|
mergedPts.setSize(returnReduce(set.size(), sumOp<label>()));
|
|
mergedIDs.setSize(mergedPts.size());
|
|
|
|
labelList setPointIDs(set.sortedToc());
|
|
|
|
// Get renumbered local data
|
|
pointField myPoints(mesh.points(), setPointIDs);
|
|
labelList myIDs(setPointIDs.size());
|
|
forAll(setPointIDs, i)
|
|
{
|
|
myIDs[i] = globalNumbering.toGlobal(setPointIDs[i]);
|
|
}
|
|
|
|
if (Pstream::master())
|
|
{
|
|
// Insert master data first
|
|
label pOffset = 0;
|
|
SubList<point>(mergedPts, myPoints.size(), pOffset) = myPoints;
|
|
SubList<label>(mergedIDs, myIDs.size(), pOffset) = myIDs;
|
|
pOffset += myPoints.size();
|
|
|
|
// Receive slave ones
|
|
for (int slave=1; slave<Pstream::nProcs(); slave++)
|
|
{
|
|
IPstream fromSlave(Pstream::commsTypes::scheduled, slave);
|
|
|
|
pointField slavePts(fromSlave);
|
|
labelList slaveIDs(fromSlave);
|
|
|
|
SubList<point>(mergedPts, slavePts.size(), pOffset) = slavePts;
|
|
SubList<label>(mergedIDs, slaveIDs.size(), pOffset) = slaveIDs;
|
|
pOffset += slaveIDs.size();
|
|
}
|
|
}
|
|
else
|
|
{
|
|
// Construct processor stream with estimate of size. Could
|
|
// be improved.
|
|
OPstream toMaster
|
|
(
|
|
Pstream::commsTypes::scheduled,
|
|
Pstream::masterNo(),
|
|
myPoints.byteSize() + myIDs.byteSize()
|
|
);
|
|
toMaster << myPoints << myIDs;
|
|
}
|
|
}
|
|
else
|
|
{
|
|
mergedIDs = set.sortedToc();
|
|
mergedPts = pointField(mesh.points(), mergedIDs);
|
|
}
|
|
|
|
|
|
// Write with scalar pointID
|
|
if (Pstream::master())
|
|
{
|
|
scalarField scalarPointIDs(mergedIDs.size());
|
|
forAll(mergedIDs, i)
|
|
{
|
|
scalarPointIDs[i] = 1.0*mergedIDs[i];
|
|
}
|
|
|
|
coordSet points(set.name(), "distance", mergedPts, mag(mergedPts));
|
|
|
|
List<const scalarField*> flds(1, &scalarPointIDs);
|
|
|
|
wordList fldNames(1, "pointID");
|
|
|
|
// Output e.g. pointSet p0 to
|
|
// postProcessing/<time>/p0.vtk
|
|
fileName outputDir
|
|
(
|
|
set.time().path()
|
|
/ (Pstream::parRun() ? ".." : "")
|
|
/ "postProcessing"
|
|
/ mesh.pointsInstance()
|
|
// set.name()
|
|
);
|
|
outputDir.clean();
|
|
mkDir(outputDir);
|
|
|
|
fileName outputFile(outputDir/writer.getFileName(points, wordList()));
|
|
//fileName outputFile(outputDir/set.name());
|
|
|
|
OFstream os(outputFile);
|
|
|
|
writer.write(points, fldNames, flds, os);
|
|
}
|
|
}
|
|
|
|
|
|
// ************************************************************************* //
|