openfoam/applications/utilities/mesh/manipulation/createPatch/createPatch.C
Andrew Heather d8d6030ab6 INT: Integration of Mattijs' collocated parallel IO additions
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.
2017-07-07 11:39:56 +01:00

919 lines
26 KiB
C

/*---------------------------------------------------------------------------*\
========= |
\\ / F ield | OpenFOAM: The Open Source CFD Toolbox
\\ / O peration |
\\ / A nd | Copyright (C) 2011-2017 OpenFOAM Foundation
\\/ M anipulation | Copyright (C) 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/>.
Application
createPatch
Group
grpMeshManipulationUtilities
Description
Utility to create patches out of selected boundary faces. Faces come either
from existing patches or from a faceSet.
More specifically it:
- creates new patches (from selected boundary faces). Synchronise faces
on coupled patches.
- synchronises points on coupled boundaries
- remove patches with 0 faces in them
\*---------------------------------------------------------------------------*/
#include "cyclicPolyPatch.H"
#include "syncTools.H"
#include "argList.H"
#include "polyMesh.H"
#include "Time.H"
#include "SortableList.H"
#include "OFstream.H"
#include "meshTools.H"
#include "faceSet.H"
#include "IOPtrList.H"
#include "polyTopoChange.H"
#include "polyModifyFace.H"
#include "wordReList.H"
#include "processorMeshes.H"
#include "IOdictionary.H"
using namespace Foam;
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
namespace Foam
{
defineTemplateTypeNameAndDebug(IOPtrList<dictionary>, 0);
}
void changePatchID
(
const polyMesh& mesh,
const label faceID,
const label patchID,
polyTopoChange& meshMod
)
{
const label zoneID = mesh.faceZones().whichZone(faceID);
bool zoneFlip = false;
if (zoneID >= 0)
{
const faceZone& fZone = mesh.faceZones()[zoneID];
zoneFlip = fZone.flipMap()[fZone.whichFace(faceID)];
}
meshMod.setAction
(
polyModifyFace
(
mesh.faces()[faceID], // face
faceID, // face ID
mesh.faceOwner()[faceID], // owner
-1, // neighbour
false, // flip flux
patchID, // patch ID
false, // remove from zone
zoneID, // zone ID
zoneFlip // zone flip
)
);
}
// Filter out the empty patches.
void filterPatches(polyMesh& mesh, const HashSet<word>& addedPatchNames)
{
const polyBoundaryMesh& patches = mesh.boundaryMesh();
// Patches to keep
DynamicList<polyPatch*> allPatches(patches.size());
label nOldPatches = returnReduce(patches.size(), sumOp<label>());
// Copy old patches.
forAll(patches, patchi)
{
const polyPatch& pp = patches[patchi];
// Note: reduce possible since non-proc patches guaranteed in same order
if (!isA<processorPolyPatch>(pp))
{
// Add if
// - non zero size
// - or added from the createPatchDict
// - or cyclic (since referred to by other cyclic half or
// proccyclic)
if
(
addedPatchNames.found(pp.name())
|| returnReduce(pp.size(), sumOp<label>()) > 0
|| isA<coupledPolyPatch>(pp)
)
{
allPatches.append
(
pp.clone
(
patches,
allPatches.size(),
pp.size(),
pp.start()
).ptr()
);
}
else
{
Info<< "Removing zero-sized patch " << pp.name()
<< " type " << pp.type()
<< " at position " << patchi << endl;
}
}
}
// Copy non-empty processor patches
forAll(patches, patchi)
{
const polyPatch& pp = patches[patchi];
if (isA<processorPolyPatch>(pp))
{
if (pp.size())
{
allPatches.append
(
pp.clone
(
patches,
allPatches.size(),
pp.size(),
pp.start()
).ptr()
);
}
else
{
Info<< "Removing empty processor patch " << pp.name()
<< " at position " << patchi << endl;
}
}
}
label nAllPatches = returnReduce(allPatches.size(), sumOp<label>());
if (nAllPatches != nOldPatches)
{
Info<< "Removing patches." << endl;
allPatches.shrink();
mesh.removeBoundary();
mesh.addPatches(allPatches);
}
else
{
Info<< "No patches removed." << endl;
forAll(allPatches, i)
{
delete allPatches[i];
}
}
}
// Dump for all patches the current match
void dumpCyclicMatch(const fileName& prefix, const polyMesh& mesh)
{
const polyBoundaryMesh& patches = mesh.boundaryMesh();
forAll(patches, patchi)
{
if
(
isA<cyclicPolyPatch>(patches[patchi])
&& refCast<const cyclicPolyPatch>(patches[patchi]).owner()
)
{
const cyclicPolyPatch& cycPatch =
refCast<const cyclicPolyPatch>(patches[patchi]);
// Dump patches
{
OFstream str(prefix+cycPatch.name()+".obj");
Pout<< "Dumping " << cycPatch.name()
<< " faces to " << str.name() << endl;
meshTools::writeOBJ
(
str,
cycPatch,
cycPatch.points()
);
}
const cyclicPolyPatch& nbrPatch = cycPatch.neighbPatch();
{
OFstream str(prefix+nbrPatch.name()+".obj");
Pout<< "Dumping " << nbrPatch.name()
<< " faces to " << str.name() << endl;
meshTools::writeOBJ
(
str,
nbrPatch,
nbrPatch.points()
);
}
// Lines between corresponding face centres
OFstream str(prefix+cycPatch.name()+nbrPatch.name()+"_match.obj");
label vertI = 0;
Pout<< "Dumping cyclic match as lines between face centres to "
<< str.name() << endl;
forAll(cycPatch, facei)
{
const point& fc0 = mesh.faceCentres()[cycPatch.start()+facei];
meshTools::writeOBJ(str, fc0);
vertI++;
const point& fc1 = mesh.faceCentres()[nbrPatch.start()+facei];
meshTools::writeOBJ(str, fc1);
vertI++;
str<< "l " << vertI-1 << ' ' << vertI << nl;
}
}
}
}
void separateList
(
const vectorField& separation,
UList<vector>& field
)
{
if (separation.size() == 1)
{
// Single value for all.
forAll(field, i)
{
field[i] += separation[0];
}
}
else if (separation.size() == field.size())
{
forAll(field, i)
{
field[i] += separation[i];
}
}
else
{
FatalErrorInFunction
<< "Sizes of field and transformation not equal. field:"
<< field.size() << " transformation:" << separation.size()
<< abort(FatalError);
}
}
// Synchronise points on both sides of coupled boundaries.
template<class CombineOp>
void syncPoints
(
const polyMesh& mesh,
pointField& points,
const CombineOp& cop,
const point& nullValue
)
{
if (points.size() != mesh.nPoints())
{
FatalErrorInFunction
<< "Number of values " << points.size()
<< " is not equal to the number of points in the mesh "
<< mesh.nPoints() << abort(FatalError);
}
const polyBoundaryMesh& patches = mesh.boundaryMesh();
// Is there any coupled patch with transformation?
bool hasTransformation = false;
if (Pstream::parRun())
{
// Send
forAll(patches, patchi)
{
const polyPatch& pp = patches[patchi];
if
(
isA<processorPolyPatch>(pp)
&& pp.nPoints() > 0
&& refCast<const processorPolyPatch>(pp).owner()
)
{
const processorPolyPatch& procPatch =
refCast<const processorPolyPatch>(pp);
// Get data per patchPoint in neighbouring point numbers.
pointField patchInfo(procPatch.nPoints(), nullValue);
const labelList& meshPts = procPatch.meshPoints();
const labelList& nbrPts = procPatch.neighbPoints();
forAll(nbrPts, pointi)
{
label nbrPointi = nbrPts[pointi];
if (nbrPointi >= 0 && nbrPointi < patchInfo.size())
{
patchInfo[nbrPointi] = points[meshPts[pointi]];
}
}
OPstream toNbr
(
Pstream::commsTypes::blocking,
procPatch.neighbProcNo()
);
toNbr << patchInfo;
}
}
// Receive and set.
forAll(patches, patchi)
{
const polyPatch& pp = patches[patchi];
if
(
isA<processorPolyPatch>(pp)
&& pp.nPoints() > 0
&& !refCast<const processorPolyPatch>(pp).owner()
)
{
const processorPolyPatch& procPatch =
refCast<const processorPolyPatch>(pp);
pointField nbrPatchInfo(procPatch.nPoints());
{
// We do not know the number of points on the other side
// so cannot use Pstream::read.
IPstream fromNbr
(
Pstream::commsTypes::blocking,
procPatch.neighbProcNo()
);
fromNbr >> nbrPatchInfo;
}
// Null any value which is not on neighbouring processor
nbrPatchInfo.setSize(procPatch.nPoints(), nullValue);
if (!procPatch.parallel())
{
hasTransformation = true;
transformList(procPatch.forwardT(), nbrPatchInfo);
}
else if (procPatch.separated())
{
hasTransformation = true;
separateList(-procPatch.separation(), nbrPatchInfo);
}
const labelList& meshPts = procPatch.meshPoints();
forAll(meshPts, pointi)
{
label meshPointi = meshPts[pointi];
points[meshPointi] = nbrPatchInfo[pointi];
}
}
}
}
// Do the cyclics.
forAll(patches, patchi)
{
const polyPatch& pp = patches[patchi];
if
(
isA<cyclicPolyPatch>(pp)
&& refCast<const cyclicPolyPatch>(pp).owner()
)
{
const cyclicPolyPatch& cycPatch =
refCast<const cyclicPolyPatch>(pp);
const edgeList& coupledPoints = cycPatch.coupledPoints();
const labelList& meshPts = cycPatch.meshPoints();
const cyclicPolyPatch& nbrPatch = cycPatch.neighbPatch();
const labelList& nbrMeshPts = nbrPatch.meshPoints();
pointField half0Values(coupledPoints.size());
forAll(coupledPoints, i)
{
const edge& e = coupledPoints[i];
label point0 = meshPts[e[0]];
half0Values[i] = points[point0];
}
if (!cycPatch.parallel())
{
hasTransformation = true;
transformList(cycPatch.reverseT(), half0Values);
}
else if (cycPatch.separated())
{
hasTransformation = true;
separateList(cycPatch.separation(), half0Values);
}
forAll(coupledPoints, i)
{
const edge& e = coupledPoints[i];
label point1 = nbrMeshPts[e[1]];
points[point1] = half0Values[i];
}
}
}
//- Note: hasTransformation is only used for warning messages so
// reduction not strictly nessecary.
//reduce(hasTransformation, orOp<bool>());
// Synchronize multiple shared points.
const globalMeshData& pd = mesh.globalData();
if (pd.nGlobalPoints() > 0)
{
if (hasTransformation)
{
WarningInFunction
<< "There are decomposed cyclics in this mesh with"
<< " transformations." << endl
<< "This is not supported. The result will be incorrect"
<< endl;
}
// Values on shared points.
pointField sharedPts(pd.nGlobalPoints(), nullValue);
forAll(pd.sharedPointLabels(), i)
{
label meshPointi = pd.sharedPointLabels()[i];
// Fill my entries in the shared points
sharedPts[pd.sharedPointAddr()[i]] = points[meshPointi];
}
// Combine on master.
Pstream::listCombineGather(sharedPts, cop);
Pstream::listCombineScatter(sharedPts);
// Now we will all have the same information. Merge it back with
// my local information.
forAll(pd.sharedPointLabels(), i)
{
label meshPointi = pd.sharedPointLabels()[i];
points[meshPointi] = sharedPts[pd.sharedPointAddr()[i]];
}
}
}
int main(int argc, char *argv[])
{
#include "addOverwriteOption.H"
#include "addRegionOption.H"
#include "addDictOption.H"
Foam::argList::addBoolOption
(
"writeObj",
"write obj files showing the cyclic matching process"
);
#include "setRootCase.H"
#include "createTime.H"
runTime.functionObjects().off();
Foam::word meshRegionName = polyMesh::defaultRegion;
args.optionReadIfPresent("region", meshRegionName);
const bool overwrite = args.optionFound("overwrite");
#include "createNamedPolyMesh.H"
const bool writeObj = args.optionFound("writeObj");
const word oldInstance = mesh.pointsInstance();
const word dictName("createPatchDict");
#include "setSystemMeshDictionaryIO.H"
Info<< "Reading " << dictIO.instance()/dictIO.name() << nl << endl;
IOdictionary dict(dictIO);
// Whether to synchronise points
const Switch pointSync(dict.lookup("pointSync"));
const polyBoundaryMesh& patches = mesh.boundaryMesh();
// If running parallel check same patches everywhere
patches.checkParallelSync(true);
if (writeObj)
{
dumpCyclicMatch("initial_", mesh);
}
// Read patch construct info from dictionary
PtrList<dictionary> patchSources(dict.lookup("patches"));
HashSet<word> addedPatchNames;
forAll(patchSources, addedI)
{
const dictionary& dict = patchSources[addedI];
addedPatchNames.insert(dict.lookup("name"));
}
// 1. Add all new patches
// ~~~~~~~~~~~~~~~~~~~~~~
if (patchSources.size())
{
// Old and new patches.
DynamicList<polyPatch*> allPatches(patches.size()+patchSources.size());
label startFacei = mesh.nInternalFaces();
// Copy old patches.
forAll(patches, patchi)
{
const polyPatch& pp = patches[patchi];
if (!isA<processorPolyPatch>(pp))
{
allPatches.append
(
pp.clone
(
patches,
patchi,
pp.size(),
startFacei
).ptr()
);
startFacei += pp.size();
}
}
forAll(patchSources, addedI)
{
const dictionary& dict = patchSources[addedI];
word patchName(dict.lookup("name"));
label destPatchi = patches.findPatchID(patchName);
if (destPatchi == -1)
{
dictionary patchDict(dict.subDict("patchInfo"));
destPatchi = allPatches.size();
Info<< "Adding new patch " << patchName
<< " as patch " << destPatchi
<< " from " << patchDict << endl;
patchDict.set("nFaces", 0);
patchDict.set("startFace", startFacei);
// Add an empty patch.
allPatches.append
(
polyPatch::New
(
patchName,
patchDict,
destPatchi,
patches
).ptr()
);
}
else
{
Info<< "Patch '" << patchName << "' already exists. Only "
<< "moving patch faces - type will remain the same" << endl;
}
}
// Copy old patches.
forAll(patches, patchi)
{
const polyPatch& pp = patches[patchi];
if (isA<processorPolyPatch>(pp))
{
allPatches.append
(
pp.clone
(
patches,
patchi,
pp.size(),
startFacei
).ptr()
);
startFacei += pp.size();
}
}
allPatches.shrink();
mesh.removeBoundary();
mesh.addPatches(allPatches);
Info<< endl;
}
// 2. Repatch faces
// ~~~~~~~~~~~~~~~~
polyTopoChange meshMod(mesh);
forAll(patchSources, addedI)
{
const dictionary& dict = patchSources[addedI];
const word patchName(dict.lookup("name"));
label destPatchi = patches.findPatchID(patchName);
if (destPatchi == -1)
{
FatalErrorInFunction
<< "patch " << patchName << " not added. Problem."
<< abort(FatalError);
}
const word sourceType(dict.lookup("constructFrom"));
if (sourceType == "patches")
{
labelHashSet patchSources
(
patches.patchSet
(
wordReList(dict.lookup("patches"))
)
);
// Repatch faces of the patches.
forAllConstIter(labelHashSet, patchSources, iter)
{
const polyPatch& pp = patches[iter.key()];
Info<< "Moving faces from patch " << pp.name()
<< " to patch " << destPatchi << endl;
forAll(pp, i)
{
changePatchID
(
mesh,
pp.start() + i,
destPatchi,
meshMod
);
}
}
}
else if (sourceType == "set")
{
const word setName(dict.lookup("set"));
faceSet faces(mesh, setName);
Info<< "Read " << returnReduce(faces.size(), sumOp<label>())
<< " faces from faceSet " << faces.name() << endl;
// Sort (since faceSet contains faces in arbitrary order)
labelList faceLabels(faces.toc());
SortableList<label> patchFaces(faceLabels);
forAll(patchFaces, i)
{
label facei = patchFaces[i];
if (mesh.isInternalFace(facei))
{
FatalErrorInFunction
<< "Face " << facei << " specified in set "
<< faces.name()
<< " is not an external face of the mesh." << endl
<< "This application can only repatch existing boundary"
<< " faces." << exit(FatalError);
}
changePatchID
(
mesh,
facei,
destPatchi,
meshMod
);
}
}
else
{
FatalErrorInFunction
<< "Invalid source type " << sourceType << endl
<< "Valid source types are 'patches' 'set'" << exit(FatalError);
}
}
Info<< endl;
// Change mesh, use inflation to reforce calculation of transformation
// tensors.
Info<< "Doing topology modification to order faces." << nl << endl;
autoPtr<mapPolyMesh> map = meshMod.changeMesh(mesh, true);
mesh.movePoints(map().preMotionPoints());
if (writeObj)
{
dumpCyclicMatch("coupled_", mesh);
}
// Synchronise points.
if (!pointSync)
{
Info<< "Not synchronising points." << nl << endl;
}
else
{
Info<< "Synchronising points." << nl << endl;
// This is a bit tricky. Both normal and position might be out and
// current separation also includes the normal
// ( separation_ = (nf&(Cr - Cf))*nf ).
// For cyclic patches:
// - for separated ones use user specified offset vector
forAll(mesh.boundaryMesh(), patchi)
{
const polyPatch& pp = mesh.boundaryMesh()[patchi];
if (pp.size() && isA<coupledPolyPatch>(pp))
{
const coupledPolyPatch& cpp =
refCast<const coupledPolyPatch>(pp);
if (cpp.separated())
{
Info<< "On coupled patch " << pp.name()
<< " separation[0] was "
<< cpp.separation()[0] << endl;
if (isA<cyclicPolyPatch>(pp) && pp.size())
{
const cyclicPolyPatch& cycpp =
refCast<const cyclicPolyPatch>(pp);
if (cycpp.transform() == cyclicPolyPatch::TRANSLATIONAL)
{
// Force to wanted separation
Info<< "On cyclic translation patch " << pp.name()
<< " forcing uniform separation of "
<< cycpp.separationVector() << endl;
const_cast<vectorField&>(cpp.separation()) =
pointField(1, cycpp.separationVector());
}
else
{
const cyclicPolyPatch& nbr = cycpp.neighbPatch();
const_cast<vectorField&>(cpp.separation()) =
pointField
(
1,
nbr[0].centre(mesh.points())
- cycpp[0].centre(mesh.points())
);
}
}
Info<< "On coupled patch " << pp.name()
<< " forcing uniform separation of "
<< cpp.separation() << endl;
}
else if (!cpp.parallel())
{
Info<< "On coupled patch " << pp.name()
<< " forcing uniform rotation of "
<< cpp.forwardT()[0] << endl;
const_cast<tensorField&>
(
cpp.forwardT()
).setSize(1);
const_cast<tensorField&>
(
cpp.reverseT()
).setSize(1);
Info<< "On coupled patch " << pp.name()
<< " forcing uniform rotation of "
<< cpp.forwardT() << endl;
}
}
}
Info<< "Synchronising points." << endl;
pointField newPoints(mesh.points());
syncPoints
(
mesh,
newPoints,
minMagSqrEqOp<vector>(),
point(GREAT, GREAT, GREAT)
);
scalarField diff(mag(newPoints-mesh.points()));
Info<< "Points changed by average:" << gAverage(diff)
<< " max:" << gMax(diff) << nl << endl;
mesh.movePoints(newPoints);
}
// 3. Remove zeros-sized patches
// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Info<< "Removing patches with no faces in them." << nl<< endl;
filterPatches(mesh, addedPatchNames);
if (writeObj)
{
dumpCyclicMatch("final_", mesh);
}
// Set the precision of the points data to 10
IOstream::defaultPrecision(max(10u, IOstream::defaultPrecision()));
if (!overwrite)
{
runTime++;
}
else
{
mesh.setInstance(oldInstance);
}
// Write resulting mesh
Info<< "Writing repatched mesh to " << runTime.timeName() << nl << endl;
mesh.write();
topoSet::removeFiles(mesh);
processorMeshes::removeFiles(mesh);
Info<< "End\n" << endl;
return 0;
}
// ************************************************************************* //