Pstream with a single comm, updated parallel ops

src/finiteArea/faMesh/faMesh.C

@@ -1047,6 +1047,18 @@ void Foam::faMesh::addFaPatches(const List<faPatch*>& p)
 }
 
 
+Foam::label Foam::faMesh::comm() const
+{
+    return comm_;
+}
+
+
+Foam::label& Foam::faMesh::comm()
+{
+    return comm_;
+}
+
+
 const Foam::objectRegistry& Foam::faMesh::thisDb() const
 {
     return mesh().thisDb();

src/finiteArea/faMesh/faMesh.H

@@ -110,6 +110,12 @@ class faMesh
             mutable label nFaces_;
 
 
+        // Communication support
+
+            //- Communicator used for parallel communication
+            label comm_;
+
+
     // Demand-driven data
 
         //- Primitive patch
@@ -287,9 +293,8 @@ public:
         );
 
 
-    // Destructor
-
-        virtual ~faMesh();
+    //- Destructor
+    virtual ~faMesh();
 
 
     // Member Functions
@@ -369,6 +374,15 @@ public:
                 }
 
 
+            // Communication support
+
+                //- Return communicator used for parallel communication
+                label comm() const;
+
+                //- Return communicator used for parallel communication
+                label& comm();
+
+
         // Access
 
             //- Return reference to the mesh database

src/finiteArea/faMesh/faPatches/constraint/processor/processorFaPatch.C

@@ -55,6 +55,18 @@ processorFaPatch::~processorFaPatch()
 
 // * * * * * * * * * * * * * * Member Functions  * * * * * * * * * * * * * * //
 
+Foam::label Foam::processorFaPatch::comm() const
+{
+    return boundaryMesh().mesh().comm();
+}
+
+
+int Foam::processorFaPatch::tag() const
+{
+    return Pstream::msgType();
+}
+
+
 void processorFaPatch::makeNonGlobalPatchPoints() const
 {
     // If it is not runing parallel or there are no global points

src/finiteArea/faMesh/faPatches/constraint/processor/processorFaPatch.H

@@ -37,7 +37,6 @@ SourceFiles
 
 #include "coupledFaPatch.H"
 #include "processorLduInterface.H"
-// #include "processorPolyPatch.H"
 
 // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
 
@@ -55,7 +54,10 @@ class processorFaPatch
 {
     // Private data
 
+        //- My processro number
         int myProcNo_;
+
+        //- Neighbour processor number
         int neighbProcNo_;
 
         //- Processor-neighbbour patch edge centres
@@ -75,6 +77,7 @@ class processorFaPatch
         //  non-global, i.e. present in this processor patch
         mutable labelList* nonGlobalPatchPointsPtr_;
 
+
 protected:
 
     // Protected Member functions
@@ -88,27 +91,27 @@ protected:
         //- Find non-globa patch points
         void makeNonGlobalPatchPoints() const;
 
-protected:
 
-    // Protected Member functions
+        // Geometry functions
 
-        //- Initialise the calculation of the patch geometry
-        void initGeometry();
+            //- Initialise the calculation of the patch geometry
+            void initGeometry();
 
-        //- Calculate the patch geometry
-        void calcGeometry();
+            //- Calculate the patch geometry
+            void calcGeometry();
 
-        //- Initialise the patches for moving points
-        void initMovePoints(const pointField&);
+            //- Initialise the patches for moving points
+            void initMovePoints(const pointField&);
 
-        //- Correct patches after moving points
-        void movePoints(const pointField&);
+            //- Correct patches after moving points
+            void movePoints(const pointField&);
 
-        //- Initialise the update of the patch topology
-        virtual void initUpdateMesh();
+            //- Initialise the update of the patch topology
+            virtual void initUpdateMesh();
+
+            //- Update of the patch topology
+            virtual void updateMesh();
 
-        //- Update of the patch topology
-        virtual void updateMesh();
 
 public:
 
@@ -160,9 +163,9 @@ public:
             nonGlobalPatchPointsPtr_(NULL)
         {}
 
-    // Destructor
 
-        virtual ~processorFaPatch();
+    //- Destructor
+    virtual ~processorFaPatch();
 
 
     // Member functions
@@ -192,6 +195,16 @@ public:
             }
         }
 
+
+        // Communications support
+
+            //- Return communicator used for communication
+            virtual label comm() const;
+
+            //- Return message tag to use for communication
+            virtual int tag() const;
+
+
         //- Return face transformation tensor
         virtual const tensorField& forwardT() const
         {

src/finiteVolume/fields/fvPatchFields/constraint/processor/processorFvPatchField.H

@@ -249,7 +249,7 @@ public:
             ) const;
 
 
-        //- Processor coupled interface functions
+        // Processor coupled interface functions
 
             //- Return processor number
             virtual int myProcNo() const
@@ -263,6 +263,16 @@ public:
                 return procPatch_.neighbProcNo();
             }
 
+
+            // Communication support
+
+                //- Return communicator used for parallel communication
+                virtual int comm() const
+                {
+                    return procPatch_.comm();
+                }
+
+
             //- Does the patch field perform the transfromation
             virtual bool doTransform() const
             {

src/finiteVolume/fvMatrices/fvMatrix/fvMatrix.C

@@ -698,7 +698,10 @@ void Foam::fvMatrix<Type>::relax()
 {
     if (psi_.mesh().solutionDict().relaxEquation(psi_.name()))
     {
-        relax(psi_.mesh().solutionDict().equationRelaxationFactor(psi_.name()));
+        relax
+        (
+            psi_.mesh().solutionDict().equationRelaxationFactor(psi_.name())
+        );
     }
     else
     {

src/finiteVolume/fvMesh/fvMesh.H

@@ -216,9 +216,8 @@ public:
         );
 
 
-    // Destructor
-
-        virtual ~fvMesh();
+    //- Destructor
+    virtual ~fvMesh();
 
 
     // Member Functions
@@ -285,6 +284,15 @@ public:
             }
 
 
+            // Communication support
+
+                //- Return communicator used for parallel communication
+                label comm() const
+                {
+                    return polyMesh::comm();
+                }
+
+
             //- Return cell volumes
             const DimensionedField<scalar, volMesh>& V() const;
 

src/finiteVolume/fvMesh/fvPatches/constraint/cyclicGgi/cyclicGgiFvPatch.H

@@ -90,10 +90,9 @@ public:
         {}
 
 
-    // Destructor
-
-        virtual ~cyclicGgiFvPatch()
-        {}
+    //- Destructor
+    virtual ~cyclicGgiFvPatch()
+    {}
 
 
     // Member functions

src/finiteVolume/fvMesh/fvPatches/constraint/ggi/ggiFvPatch.H

@@ -92,13 +92,27 @@ public:
         {}
 
 
-    // Destructor
-
-        virtual ~ggiFvPatch();
+    //- Destructor
+    virtual ~ggiFvPatch();
 
 
     // Member functions
 
+        // Communication support
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const
+            {
+                return ggiPolyPatch_.comm();
+            }
+
+            //- Return message tag used for sending
+            virtual int tag() const
+            {
+                return ggiPolyPatch_.tag();
+            }
+
+
         // Access
 
             //- Return shadow patch

src/finiteVolume/fvMesh/fvPatches/constraint/mixingPlane/mixingPlaneFvPatch.H

@@ -95,13 +95,27 @@ public:
         {}
 
 
-    // Destructor
-
-        virtual ~mixingPlaneFvPatch();
+    //- Destructor
+    virtual ~mixingPlaneFvPatch();
 
 
     // Member functions
 
+        // Communication support
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const
+            {
+                return mixingPlanePolyPatch_.comm();
+            }
+
+            //- Return message tag used for sending
+            virtual int tag() const
+            {
+                return mixingPlanePolyPatch_.tag();
+            }
+
+
         // Access
 
             //- Return shadow patch

src/finiteVolume/fvMesh/fvPatches/constraint/processor/processorFvPatch.H

@@ -86,10 +86,9 @@ public:
         {}
 
 
-    // Destructor
-
-        virtual ~processorFvPatch()
-        {}
+    //- Destructor
+    virtual ~processorFvPatch()
+    {}
 
 
     // Member functions
@@ -119,6 +118,21 @@ public:
             }
         }
 
+        // Communication support
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const
+            {
+                return procPolyPatch_.comm();
+            }
+
+            //- Return message tag used for sending
+            virtual int tag() const
+            {
+                return procPolyPatch_.tag();
+            }
+
+
         //- Return face transformation tensor
         virtual const tensorField& forwardT() const
         {

src/finiteVolume/fvMesh/fvPatches/constraint/regionCouple/regionCoupleFvPatch.H

@@ -95,9 +95,8 @@ public:
         {}
 
 
-    // Destructor
-
-        virtual ~regionCoupleFvPatch();
+    //- Destructor
+    virtual ~regionCoupleFvPatch();
 
 
     // Member functions
@@ -123,6 +122,21 @@ public:
             virtual tmp<vectorField> delta() const;
 
 
+        // Communication support
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const
+            {
+                return rcPolyPatch_.comm();
+            }
+
+            //- Return message tag used for sending
+            virtual int tag() const
+            {
+                return rcPolyPatch_.tag();
+            }
+
+
         // Interpolation
 
             //- Interpolate face field

src/foam/Make/files

@@ -147,7 +147,6 @@ $(Pstreams)/IPstream.C
 $(Pstreams)/OPstream.C
 $(Pstreams)/IPread.C
 $(Pstreams)/OPwrite.C
-$(Pstreams)/PstreamsPrint.C
 
 dictionary = db/dictionary
 $(dictionary)/dictionary.C

src/foam/db/IOstreams/Pstreams/IPread.C

@@ -38,6 +38,8 @@ Foam::IPstream::IPstream
     const commsTypes commsType,
     const int fromProcNo,
     const label bufSize,
+    const int tag,
+    const label comm,
     streamFormat format,
     versionNumber version
 )
@@ -45,6 +47,8 @@ Foam::IPstream::IPstream
     Pstream(commsType, bufSize),
     Istream(format, version),
     fromProcNo_(fromProcNo),
+    tag_(tag),
+    comm_(comm),
     messageSize_(0)
 {
     setOpened();
@@ -52,17 +56,31 @@ Foam::IPstream::IPstream
 
     MPI_Status status;
 
-    // If the buffer size is not specified, probe the incomming message
+    // If the buffer size is not specified, probe the incoming message
     // and set it
     if (!bufSize)
     {
-        MPI_Probe(procID(fromProcNo_), msgType(), MPI_COMM_WORLD, &status);
+        MPI_Probe
+        (
+            fromProcNo_,
+            tag_,
+            PstreamGlobals::MPICommunicators_[comm_],
+            &status
+        );
         MPI_Get_count(&status, MPI_BYTE, &messageSize_);
 
         buf_.setSize(messageSize_);
     }
 
-    messageSize_ = read(commsType, fromProcNo_, buf_.begin(), buf_.size());
+    messageSize_ = IPstream::read
+    (
+        commsType,
+        fromProcNo_,
+        buf_.begin(),
+        buf_.size(),
+        tag_,
+        comm_
+    );
 
     if (!messageSize_)
     {
@@ -83,9 +101,31 @@ Foam::label Foam::IPstream::read
     const commsTypes commsType,
     const int fromProcNo,
     char* buf,
-    const std::streamsize bufSize
+    const std::streamsize bufSize,
+    const int tag,
+    const label comm
 )
 {
+    if (debug)
+    {
+        Pout<< "UIPstream::read : starting read from:" << fromProcNo
+            << " tag:" << tag << " comm:" << comm
+            << " wanted size:" << label(bufSize)
+            << " commsType:" << Pstream::commsTypeNames[commsType]
+            << Foam::endl;
+    }
+
+    if (Pstream::warnComm != -1 && comm != Pstream::warnComm)
+    {
+        Pout<< "UIPstream::read : starting read from:" << fromProcNo
+            << " tag:" << tag << " comm:" << comm
+            << " wanted size:" << label(bufSize)
+            << " commsType:" << Pstream::commsTypeNames[commsType]
+            << " warnComm:" << Pstream::warnComm
+            << Foam::endl;
+        error::printStack(Pout);
+    }
+
     if (commsType == blocking || commsType == scheduled)
     {
         MPI_Status status;
@@ -96,10 +136,10 @@ Foam::label Foam::IPstream::read
             (
                 buf,
                 bufSize,
-                MPI_PACKED,
-                procID(fromProcNo),
-                msgType(),
-                MPI_COMM_WORLD,
+                MPI_BYTE,
+                fromProcNo,
+                tag,
+                PstreamGlobals::MPICommunicators_[comm],
                 &status
             )
         )
@@ -144,10 +184,10 @@ Foam::label Foam::IPstream::read
             (
                 buf,
                 bufSize,
-                MPI_PACKED,
-                procID(fromProcNo),
-                msgType(),
-                MPI_COMM_WORLD,
+                MPI_BYTE,
+                fromProcNo,
+                tag,
+                PstreamGlobals::MPICommunicators_[comm],
                 &request
             )
         )
@@ -162,8 +202,18 @@ Foam::label Foam::IPstream::read
             return 0;
         }
 
-        PstreamGlobals::IPstream_outstandingRequests_.append(request);
+        if (debug)
+        {
+            Pout<< "UIPstream::read : started read from:" << fromProcNo
+                << " tag:" << tag << " read size:" << label(bufSize)
+                << " commsType:" << Pstream::commsTypeNames[commsType]
+                << " request:" << PstreamGlobals::outstandingRequests_.size()
+                << Foam::endl;
+        }
 
+        PstreamGlobals::outstandingRequests_.append(request);
+
+        // Assume the message is completely received.
         return 1;
     }
     else
@@ -180,56 +230,4 @@ Foam::label Foam::IPstream::read
 }
 
 
-void Foam::IPstream::waitRequests()
-{
-    if (PstreamGlobals::IPstream_outstandingRequests_.size())
-    {
-        if
-        (
-            MPI_Waitall
-            (
-                PstreamGlobals::IPstream_outstandingRequests_.size(),
-                PstreamGlobals::IPstream_outstandingRequests_.begin(),
-                MPI_STATUSES_IGNORE
-            )
-        )
-        {
-            FatalErrorIn
-            (
-                "IPstream::waitRequests()"
-            )   << "MPI_Waitall returned with error" << endl;
-        }
-
-        PstreamGlobals::IPstream_outstandingRequests_.clear();
-    }
-}
-
-
-bool Foam::IPstream::finishedRequest(const label i)
-{
-    if (i >= PstreamGlobals::IPstream_outstandingRequests_.size())
-    {
-        FatalErrorIn
-        (
-            "IPstream::finishedRequest(const label)"
-        )   << "There are "
-            << PstreamGlobals::IPstream_outstandingRequests_.size()
-            << " outstanding send requests and you are asking for i=" << i
-            << nl
-            << "Maybe you are mixing blocking/non-blocking comms?"
-            << Foam::abort(FatalError);
-    }
-
-    int flag;
-    MPI_Test
-    (
-        &PstreamGlobals::IPstream_outstandingRequests_[i],
-        &flag,
-        MPI_STATUS_IGNORE
-    );
-
-    return flag != 0;
-}
-
-
 // ************************************************************************* //

src/foam/db/IOstreams/Pstreams/IPreadToken.C

@@ -1,187 +0,0 @@
-/*---------------------------------------------------------------------------*\
-  =========                 |
-  \\      /  F ield         | foam-extend: Open Source CFD
-   \\    /   O peration     | Version:     4.0
-    \\  /    A nd           | Web:         http://www.foam-extend.org
-     \\/     M anipulation  | For copyright notice see file Copyright
--------------------------------------------------------------------------------
-License
-    This file is part of foam-extend.
-
-    foam-extend 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.
-
-    foam-extend 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 foam-extend.  If not, see <http://www.gnu.org/licenses/>.
-
-\*---------------------------------------------------------------------------*/
-
-#include "IPstream.H"
-#include "token.H"
-#include <cctype>
-
-// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
-
-namespace Foam
-{
-
-// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
-
-Istream& IPstream::read(token& t)
-{
-    // Return the put back token if it exists
-    if (Istream::getBack(t))
-    {
-        return *this;
-    }
-
-    char c;
-
-    // return on error
-    if (!read(c))
-    {
-        t.setBad();
-        return *this;
-    }
-
-    // Set the line number of this token to the current stream line number
-    t.lineNumber() = lineNumber();
-
-    // Analyse input starting with this character.
-    switch (c)
-    {
-        // Punctuation
-        case token::END_STATEMENT :
-        case token::BEGIN_LIST :
-        case token::END_LIST :
-        case token::BEGIN_SQR :
-        case token::END_SQR :
-        case token::BEGIN_BLOCK :
-        case token::END_BLOCK :
-        case token::COLON :
-        case token::COMMA :
-        case token::ASSIGN :
-        case token::ADD :
-        case token::SUBTRACT :
-        case token::MULTIPLY :
-        case token::DIVIDE :
-        {
-            t = token::punctuationToken(c);
-            return *this;
-        }
-
-        // Word
-        case token::WORD :
-        {
-            word* wPtr = new word;
-            if (read(*wPtr))
-            {
-                if (token::compound::isCompound(*wPtr))
-                {
-                    t = token::compound::New(*wPtr, *this).ptr();
-                    delete wPtr;
-                }
-                else
-                {
-                    t = wPtr;
-                }
-            }
-            else
-            {
-                delete wPtr;
-                t.setBad();
-            }
-            return *this;
-        }
-
-        // String
-        case token::STRING :
-        {
-            string* sPtr = new string;
-            if (read(*sPtr))
-            {
-                t = sPtr;
-            }
-            else
-            {
-                delete sPtr;
-                t.setBad();
-            }
-            return *this;
-        }
-
-        // Label
-        case token::LABEL :
-        {
-            label l;
-            if (read(l))
-            {
-                t = l;
-            }
-            else
-            {
-                t.setBad();
-            }
-            return *this;
-        }
-
-        // floatScalar
-        case token::FLOAT_SCALAR :
-        {
-            floatScalar s;
-            if (read(s))
-            {
-                t = s;
-            }
-            else
-            {
-                t.setBad();
-            }
-            return *this;
-        }
-
-        // doubleScalar
-        case token::DOUBLE_SCALAR :
-        {
-            doubleScalar s;
-            if (read(s))
-            {
-                t = s;
-            }
-            else
-            {
-                t.setBad();
-            }
-            return *this;
-        }
-
-        // Character (returned as a single character word) or error
-        default:
-        {
-            if (isalpha(c))
-            {
-                t = word(c);
-                return *this;
-            }
-
-            setBad();
-            t.setBad();
-
-            return *this;
-        }
-    }
-}
-
-
-// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
-
-} // End namespace Foam
-
-// ************************************************************************* //

src/foam/db/IOstreams/Pstreams/IPstream.C

@@ -324,4 +324,13 @@ Foam::Istream& Foam::IPstream::rewind()
 }
 
 
+void Foam::IPstream::print(Ostream& os) const
+{
+    os  << "Reading from processor " << fromProcNo_
+        << " using communicator " << comm_
+        <<  " and tag " << tag_
+        << Foam::endl;
+}
+
+
 // ************************************************************************* //

src/foam/db/IOstreams/Pstreams/IPstream.H

@@ -58,6 +58,12 @@ class IPstream
         //- ID of sending processor
         int fromProcNo_;
 
+        //- Message tag
+        const int tag_;
+
+        //- Communicator
+        const label comm_;
+
         //- Message size
         label messageSize_;
 
@@ -86,12 +92,14 @@ public:
             const commsTypes commsType,
             const int fromProcNo,
             const label bufSize = 0,
+            const int tag = Pstream::msgType(),
+            const label comm = Pstream::worldComm,
             streamFormat format = BINARY,
             versionNumber version = currentVersion
         );
 
 
-    // Destructor
+    //- Destructor
     ~IPstream();
 
 
@@ -115,15 +123,11 @@ public:
                 const commsTypes commsType,
                 const int fromProcNo,
                 char* buf,
-                const std::streamsize bufSize
+                const std::streamsize bufSize,
+                const int tag = Pstream::msgType(),
+                const label communicator = 0
             );
 
-            //- Non-blocking receives: wait until all have finished.
-            static void waitRequests();
-
-            //- Non-blocking receives: has request i finished?
-            static bool finishedRequest(const label i);
-
             //- Return next token from stream
             Istream& read(token&);
 

src/foam/db/IOstreams/Pstreams/OPstream.C

@@ -92,13 +92,17 @@ Foam::OPstream::OPstream
     const commsTypes commsType,
     const int toProcNo,
     const label bufSize,
+    const int tag,
+    const label comm,
     streamFormat format,
     versionNumber version
 )
 :
     Pstream(commsType, bufSize),
     Ostream(format, version),
-    toProcNo_(toProcNo)
+    toProcNo_(toProcNo),
+    tag_(tag),
+    comm_(comm)
 {
     setOpened();
     setGood();
@@ -233,4 +237,12 @@ Foam::Ostream& Foam::OPstream::write(const char* data, std::streamsize count)
 }
 
 
+void Foam::OPstream::print(Ostream& os) const
+{
+    os  << "Writing from processor " << toProcNo_
+        << " to processor " << myProcNo() << " in communicator " << comm_
+        << " and tag " << tag_ << Foam::endl;
+}
+
+
 // ************************************************************************* //

src/foam/db/IOstreams/Pstreams/OPstream.H

@@ -58,6 +58,12 @@ class OPstream
          // ID of receiving processor
         int toProcNo_;
 
+        //- Message tag
+        const int tag_;
+
+        //- Communicator
+        const label comm_;
+
 
     // Private member functions
 
@@ -88,14 +94,15 @@ public:
             const commsTypes commsType,
             const int toProcNo,
             const label bufSize = 0,
+            const int tag = Pstream::msgType(),
+            const label comm = Pstream::worldComm,
             streamFormat format = BINARY,
             versionNumber version = currentVersion
         );
 
 
-    // Destructor
-
-        ~OPstream();
+    //- Destructor
+    ~OPstream();
 
 
     // Member functions
@@ -117,15 +124,11 @@ public:
                 const commsTypes commsType,
                 const int toProcNo,
                 const char* buf,
-                const std::streamsize bufSize
+                const std::streamsize bufSize,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
             );
 
-            //- Non-blocking writes: wait until all have finished.
-            static void waitRequests();
-
-            //- Non-blocking writes: has request i finished?
-            static bool finishedRequest(const label i);
-
             //- Write next token to stream
             Ostream& write(const token&);
 

src/foam/db/IOstreams/Pstreams/OPwrite.C

@@ -60,9 +60,33 @@ bool Foam::OPstream::write
     const commsTypes commsType,
     const int toProcNo,
     const char* buf,
-    const std::streamsize bufSize
+    const std::streamsize bufSize,
+    const int tag,
+    const label comm
 )
 {
+    if (debug)
+    {
+        Pout<< "OPstream::write : starting write to:" << toProcNo
+            << " tag:" << tag
+            << " comm:" << comm << " size:" << label(bufSize)
+            << " commsType:" << Pstream::commsTypeNames[commsType]
+            << Foam::endl;
+    }
+    if (Pstream::warnComm != -1 && comm != Pstream::warnComm)
+    {
+        Pout<< "OPstream::write : starting write to:" << toProcNo
+            << " tag:" << tag
+            << " comm:" << comm << " size:" << label(bufSize)
+            << " commsType:" << Pstream::commsTypeNames[commsType]
+            << " warnComm:" << Pstream::warnComm
+            << Foam::endl;
+        error::printStack(Pout);
+    }
+
+
+    PstreamGlobals::checkCommunicator(comm, toProcNo);
+
     bool transferFailed = true;
 
     if (commsType == blocking)
@@ -71,11 +95,19 @@ bool Foam::OPstream::write
         (
             const_cast<char*>(buf),
             bufSize,
-            MPI_PACKED,
-            procID(toProcNo),
-            msgType(),
-            MPI_COMM_WORLD
+            MPI_BYTE,
+            toProcNo,   //procID(toProcNo),
+            tag,
+            PstreamGlobals::MPICommunicators_[comm] // MPI_COMM_WORLD
         );
+
+        if (debug)
+        {
+            Pout<< "OPstream::write : finished write to:" << toProcNo
+                << " tag:" << tag << " size:" << label(bufSize)
+                << " commsType:" << Pstream::commsTypeNames[commsType]
+                << Foam::endl;
+        }
     }
     else if (commsType == scheduled)
     {
@@ -83,11 +115,19 @@ bool Foam::OPstream::write
         (
             const_cast<char*>(buf),
             bufSize,
-            MPI_PACKED,
-            procID(toProcNo),
-            msgType(),
-            MPI_COMM_WORLD
+            MPI_BYTE,
+            toProcNo,   //procID(toProcNo),
+            tag,
+            PstreamGlobals::MPICommunicators_[comm] // MPI_COMM_WORLD
         );
+
+        if (debug)
+        {
+            Pout<< "OPstream::write : finished write to:" << toProcNo
+                << " tag:" << tag << " size:" << label(bufSize)
+                << " commsType:" << Pstream::commsTypeNames[commsType]
+                << Foam::endl;
+        }
     }
     else if (commsType == nonBlocking)
     {
@@ -97,14 +137,23 @@ bool Foam::OPstream::write
         (
             const_cast<char*>(buf),
             bufSize,
-            MPI_PACKED,
-            procID(toProcNo),
-            msgType(),
-            MPI_COMM_WORLD,
+            MPI_BYTE,
+            toProcNo,   //procID(toProcNo),
+            tag,
+            PstreamGlobals::MPICommunicators_[comm],// MPI_COMM_WORLD,
             &request
         );
 
-        PstreamGlobals::OPstream_outstandingRequests_.append(request);
+        if (debug)
+        {
+            Pout<< "OPstream::write : started write to:" << toProcNo
+                << " tag:" << tag << " size:" << label(bufSize)
+                << " commsType:" << Pstream::commsTypeNames[commsType]
+                << " request:" << PstreamGlobals::outstandingRequests_.size()
+                << Foam::endl;
+        }
+
+        PstreamGlobals::outstandingRequests_.append(request);
     }
     else
     {
@@ -120,56 +169,4 @@ bool Foam::OPstream::write
 }
 
 
-void Foam::OPstream::waitRequests()
-{
-    if (PstreamGlobals::OPstream_outstandingRequests_.size())
-    {
-        if
-        (
-            MPI_Waitall
-            (
-                PstreamGlobals::OPstream_outstandingRequests_.size(),
-                PstreamGlobals::OPstream_outstandingRequests_.begin(),
-                MPI_STATUSES_IGNORE
-            )
-        )
-        {
-            FatalErrorIn
-            (
-                "OPstream::waitRequests()"
-            )   << "MPI_Waitall returned with error" << Foam::endl;
-        }
-
-        PstreamGlobals::OPstream_outstandingRequests_.clear();
-    }
-}
-
-
-bool Foam::OPstream::finishedRequest(const label i)
-{
-    if (i >= PstreamGlobals::OPstream_outstandingRequests_.size())
-    {
-        FatalErrorIn
-        (
-            "OPstream::finishedRequest(const label)"
-        )   << "There are "
-            << PstreamGlobals::OPstream_outstandingRequests_.size()
-            << " outstanding send requests and you are asking for i=" << i
-            << nl
-            << "Maybe you are mixing blocking/non-blocking comms?"
-            << Foam::abort(FatalError);
-    }
-
-    int flag;
-    MPI_Test
-    (
-        &PstreamGlobals::OPstream_outstandingRequests_[i],
-        &flag,
-        MPI_STATUS_IGNORE
-    );
-
-    return flag != 0;
-}
-
-
 // ************************************************************************* //

src/foam/db/IOstreams/Pstreams/Pstream.H

@@ -29,11 +29,12 @@ Description
 
 SourceFiles
     Pstream.C
-    PstreamsPrint.C
+    PstreamReduceOps.C
     PstreamCommsStruct.C
     gatherScatter.C
     combineGatherScatter.C
     gatherScatterList.C
+    PstreamExchange.C
 
 \*---------------------------------------------------------------------------*/
 
@@ -47,6 +48,8 @@ SourceFiles
 #include "NamedEnum.H"
 #include "dynamicLabelList.H"
 #include "optimisationSwitch.H"
+#include "ListOps.H"
+#include "LIFOStack.H"
 
 // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
 
@@ -59,7 +62,6 @@ namespace Foam
 
 class Pstream
 {
-
 public:
 
     //- Types of communications
@@ -72,6 +74,7 @@ public:
 
     static const NamedEnum<commsTypes, 3> commsTypeNames;
 
+
     // Public classes
 
         //- Structure for communicating between processors
@@ -85,7 +88,7 @@ public:
                 //- procIDs of processors directly below me
                 labelList below_;
 
-                //- procIDs of all processors below (so not just directly below)
+                //- procIDs of all processors below (not just directly below)
                 labelList allBelow_;
 
                 //- procIDs of all processors not below. (inverse set of
@@ -157,7 +160,7 @@ public:
                 friend Ostream& operator<<(Ostream&, const commsStruct&);
         };
 
-        //- combineReduce operator for lists. Used for counting.
+        //- combineReduce operator for lists.  Used for counting.
         class listEq
         {
         public:
@@ -165,7 +168,7 @@ public:
             template<class T>
             void operator()(T& x, const T& y) const
             {
-                forAll(y, i)
+                forAll (y, i)
                 {
                     if (y[i].size())
                     {
@@ -183,32 +186,38 @@ private:
         //- Is this a parallel run?
         static bool parRun_;
 
-        //- My processor index
-        static int myProcNo_;
-
-        //- Process IDs
-        static List<int> procIDs_;
-
-        //- Default message type
+        //- Default message type info
         static const int msgType_;
 
+        //- Stack of free comms
+        static LIFOStack<label> freeComms_;
+
+        //- My processor index
+        static DynamicList<int> myProcNo_;
+
+        //- Process IDs
+        static DynamicList<List<int> > procIDs_;
+
+        //- List of parent communicators
+        static DynamicList<label> parentCommunicator_;
+
         //- Structure for linear communications
-        static List<commsStruct> linearCommunication_;
+        static DynamicList<List<commsStruct> > linearCommunication_;
 
         //- Structure for tree communications
-        static List<commsStruct> treeCommunication_;
+        static DynamicList<List<commsStruct> > treeCommunication_;
 
 
     // Private member functions
 
         //- Set data for parallel running
-        static void setParRun();
+        static void setParRun(const label nProcs);
 
         //- Calculate linear communication schedule
-        static void calcLinearComm(const label nProcs);
+        static List<commsStruct> calcLinearComm(const label nProcs);
 
         //- Calculate tree communication schedule
-        static void calcTreeComm(const label nProcs);
+        static List<commsStruct> calcTreeComm(const label nProcs);
 
         //- Helper function for tree communication schedule determination
         //  Collects all processorIDs below a processor
@@ -223,6 +232,19 @@ private:
         //  Pstream::init()
         static void initCommunicationSchedule();
 
+        //- Allocate a communicator with index
+        static void allocatePstreamCommunicator
+        (
+            const label parentIndex,
+            const label index
+        );
+
+        //- Free a communicator
+        static void freePstreamCommunicator
+        (
+            const label index
+        );
+
 
 protected:
 
@@ -268,6 +290,15 @@ public:
             );
         }
 
+        //- Number of polling cycles in processor updates
+        static const debug::optimisationSwitch nPollProcInterfaces;
+
+        //- Default communicator (all processors)
+        static label worldComm;
+
+        //- Debugging: warn for use of any communicator differing from warnComm
+        static label warnComm;
+
 
     // Constructors
 
@@ -290,6 +321,79 @@ public:
 
     // Member functions
 
+        //- Allocate a new communicator
+        static label allocateCommunicator
+        (
+            const label parent,
+            const labelList& subRanks,
+            const bool doPstream = true
+        );
+
+        //- Free a previously allocated communicator
+        static void freeCommunicator
+        (
+            const label communicator,
+            const bool doPstream = true
+        );
+
+        //- Free all communicators
+        static void freeCommunicators(const bool doPstream);
+
+
+        //- Helper class for allocating/freeing communicators
+        class communicator
+        {
+            //- Communicator identifier
+            label comm_;
+
+            //- Disallow copy and assignment
+            communicator(const communicator&);
+            void operator=(const communicator&);
+
+        public:
+
+            //- Constructo from components
+            communicator
+            (
+                const label parent,
+                const labelList& subRanks,
+                const bool doPstream
+            )
+            :
+                comm_(allocateCommunicator(parent, subRanks, doPstream))
+            {}
+
+            //- Destructor
+            ~communicator()
+            {
+                freeCommunicator(comm_);
+            }
+
+            //- Cast to label
+            operator label() const
+            {
+                return comm_;
+            }
+        };
+
+
+        //- Return physical processor number (i.e. processor number in
+        //  worldComm) given communicator and processor
+        static int baseProcNo(const label myComm, const int procID);
+
+        //- Return processor number in communicator (given physical processor
+        //  number) (= reverse of baseProcNo)
+        static label procNo(const label comm, const int baseProcID);
+
+        //- Return processor number in communicator (given processor number
+        //  and communicator)
+        static label procNo
+        (
+            const label myComm,
+            const label currentComm,
+            const int currentProcID
+        );
+
         //- Add the valid option this type of communications library
         //  adds/requires on the command line
         static void addValidParOptions(HashTable<string>& validParOptions);
@@ -298,44 +402,78 @@ public:
         //  Spawns slave processes and initialises inter-communication
         static bool init(int& argc, char**& argv);
 
+
+        // Non-blocking comms
+
+            //- Get number of outstanding requests
+            static label nRequests();
+
+            //- Truncate number of outstanding requests
+            static void resetRequests(const label sz);
+
+            //- Wait until all requests (from start onwards) have finished.
+            static void waitRequests(const label start = 0);
+
+            //- Wait until request i has finished.
+            static void waitRequest(const label i);
+
+            //- Non-blocking comms: has request i finished?
+            static bool finishedRequest(const label i);
+
+            static int allocateTag(const char*);
+
+            static int allocateTag(const word&);
+
+            static void freeTag(const char*, const int tag);
+
+            static void freeTag(const word&, const int tag);
+
+
         //- Is this a parallel run?
         static bool& parRun()
         {
             return parRun_;
         }
 
-        //- Number of processes in parallel run
-        static label nProcs()
+        //- Number of processes in parallel run for a given communicator
+        static label nProcs(const label communicator = 0)
         {
-            return procIDs_.size();
+            return procIDs_[communicator].size();
         }
 
-        //- Am I the master process
-        static bool master()
-        {
-            return myProcNo_ == masterNo();
-        }
-
-        //- Process index of the master
+        //- Process index of the master for the global communicator
         static int masterNo()
         {
             return 0;
         }
 
-        //- Number of this process (starting from masterNo() = 0)
-        static int myProcNo()
+        //- Am I the master process
+        static bool master(const label communicator = 0)
         {
-            return myProcNo_;
+            return myProcNo_[communicator] == masterNo();
+        }
+
+
+        //- Number of this process (starting from masterNo() = 0)
+        static int myProcNo(const label communicator = 0)
+        {
+            return myProcNo_[communicator];
+        }
+
+        //- Return parent communicator
+        static label parent(const label communicator)
+        {
+            return parentCommunicator_(communicator);
         }
 
         //- Process IDs
-        static const List<int>& procIDs()
+        static const List<int>& procIDs(const int communicator)
         {
-            return procIDs_;
+            return procIDs_[communicator];
         }
 
         //- Process ID of given process index
-        static int procID(int procNo)
+        static List<int>& procID(const int procNo)
         {
             return procIDs_[procNo];
         }
@@ -347,21 +485,27 @@ public:
         }
 
         //- Process index of last slave
-        static int lastSlave()
+        static int lastSlave(const label communicator = 0)
         {
-            return nProcs() - 1;
+            return nProcs(communicator) - 1;
         }
 
         //- Communication schedule for linear all-to-master (proc 0)
-        static const List<commsStruct>& linearCommunication()
+        static const List<commsStruct>& linearCommunication
+        (
+            const label communicator = 0
+        )
         {
-            return linearCommunication_;
+            return linearCommunication_[communicator];
         }
 
         //- Communication schedule for tree all-to-master (proc 0)
-        static const List<commsStruct>& treeCommunication()
+        static const List<commsStruct>& treeCommunication
+        (
+            const label communicator = 0
+        )
         {
-            return treeCommunication_;
+            return treeCommunication_[communicator];
         }
 
         //- Message tag of standard messages
@@ -400,21 +544,40 @@ public:
             (
                 const List<commsStruct>& comms,
                 T& Value,
-                const BinaryOp& bop
+                const BinaryOp& bop,
+                const int tag,
+                const label comm
             );
 
             //- Like above but switches between linear/tree communication
             template <class T, class BinaryOp>
-            static void gather(T& Value, const BinaryOp& bop);
+            static void gather
+            (
+                T& Value,
+                const BinaryOp& bop,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
+            );
 
             //- Scatter data. Distribute without modification.
             //  Reverse of gather
             template <class T>
-            static void scatter(const List<commsStruct>& comms, T& Value);
+            static void scatter
+            (
+                const List<commsStruct>& comms,
+                T& Value,
+                const int tag,
+                const label comm
+            );
 
             //- Like above but switches between linear/tree communication
             template <class T>
-            static void scatter(T& Value);
+            static void scatter
+            (
+                T& Value,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
+            );
 
 
         // Combine variants. Inplace combine values from processors.
@@ -425,24 +588,39 @@ public:
             (
                 const List<commsStruct>& comms,
                 T& Value,
-                const CombineOp& cop
+                const CombineOp& cop,
+                const int tag,
+                const label comm
             );
 
             //- Like above but switches between linear/tree communication
             template <class T, class CombineOp>
-            static void combineGather(T& Value, const CombineOp& cop);
+            static void combineGather
+            (
+                T& Value,
+                const CombineOp& cop,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
+            );
 
             //- Scatter data. Reverse of combineGather
             template <class T>
             static void combineScatter
             (
                 const List<commsStruct>& comms,
-                T& Value
+                T& Value,
+                const int tag,
+                const label comm
             );
 
             //- Like above but switches between linear/tree communication
             template <class T>
-            static void combineScatter(T& Value);
+            static void combineScatter
+            (
+                T& Value,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
+            );
 
         // Combine variants working on whole List at a time.
 
@@ -451,7 +629,9 @@ public:
             (
                 const List<commsStruct>& comms,
                 List<T>& Value,
-                const CombineOp& cop
+                const CombineOp& cop,
+                const int tag,
+                const label comm
             );
 
             //- Like above but switches between linear/tree communication
@@ -459,7 +639,9 @@ public:
             static void listCombineGather
             (
                 List<T>& Value,
-                const CombineOp& cop
+                const CombineOp& cop,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
             );
 
             //- Scatter data. Reverse of combineGather
@@ -467,12 +649,19 @@ public:
             static void listCombineScatter
             (
                 const List<commsStruct>& comms,
-                List<T>& Value
+                List<T>& Value,
+                const int tag,
+                const label comm
             );
 
             //- Like above but switches between linear/tree communication
             template <class T>
-            static void listCombineScatter(List<T>& Value);
+            static void listCombineScatter
+            (
+                List<T>& Value,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
+            );
 
         // Combine variants working on whole map at a time. Container needs to
         // have iterators and find() defined.
@@ -482,7 +671,9 @@ public:
             (
                 const List<commsStruct>& comms,
                 Container& Values,
-                const CombineOp& cop
+                const CombineOp& cop,
+                const int tag,
+                const label comm
             );
 
             //- Like above but switches between linear/tree communication
@@ -490,7 +681,9 @@ public:
             static void mapCombineGather
             (
                 Container& Values,
-                const CombineOp& cop
+                const CombineOp& cop,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
             );
 
             //- Scatter data. Reverse of combineGather
@@ -498,13 +691,19 @@ public:
             static void mapCombineScatter
             (
                 const List<commsStruct>& comms,
-                Container& Values
+                Container& Values,
+                const int tag,
+                const label comm
             );
 
             //- Like above but switches between linear/tree communication
             template <class Container>
-            static void mapCombineScatter(Container& Values);
-
+            static void mapCombineScatter
+            (
+                Container& Values,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
+            );
 
 
         // Gather/scatter keeping the individual processor data separate.
@@ -516,24 +715,56 @@ public:
             static void gatherList
             (
                 const List<commsStruct>& comms,
-                List<T>& Values
+                List<T>& Values,
+                const int tag,
+                const label comm
             );
 
             //- Like above but switches between linear/tree communication
             template <class T>
-            static void gatherList(List<T>& Values);
+            static void gatherList
+            (
+                List<T>& Values,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
+            );
 
             //- Scatter data. Reverse of gatherList
             template <class T>
             static void scatterList
             (
                 const List<commsStruct>& comms,
-                List<T>& Values
+                List<T>& Values,
+                const int tag,
+                const label comm
             );
 
             //- Like above but switches between linear/tree communication
             template <class T>
-            static void scatterList(List<T>& Values);
+            static void scatterList
+            (
+                List<T>& Values,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm
+            );
+
+
+        // Exchange
+
+            //- Exchange data. Sends sendData, receives into recvData, sets
+            //  sizes (not bytes). sizes[p0][p1] is what processor p0 has
+            //  sent to p1. Continuous data only.
+            //  If block=true will wait for all transfers to finish.
+            template<class Container, class T>
+            static void exchange
+            (
+                const List<Container>&,
+                List<Container>&,
+                labelListList& sizes,
+                const int tag = Pstream::msgType(),
+                const label comm = Pstream::worldComm,
+                const bool block = true
+            );
 };
 
 
@@ -556,6 +787,7 @@ Ostream& operator<<(Ostream&, const Pstream::commsStruct&);
 #   include "gatherScatter.C"
 #   include "combineGatherScatter.C"
 #   include "gatherScatterList.C"
+#   include "PstreamExchange.C"
 #endif
 
 

src/foam/db/IOstreams/Pstreams/PstreamCombineReduceOps.H

@@ -50,26 +50,62 @@ void combineReduce
 (
     const List<Pstream::commsStruct>& comms,
     T& Value,
-    const CombineOp& cop
+    const CombineOp& cop,
+    const int tag,
+    const label comm
 )
 {
-    Pstream::combineGather(comms, Value, cop);
-    Pstream::combineScatter(comms, Value);
+    Pstream::combineGather(comms, Value, cop, tag, comm);
+    Pstream::combineScatter(comms, Value, tag, comm);
 }
 
 
 template <class T, class CombineOp>
-void combineReduce(T& Value, const CombineOp& cop)
+void combineReduce
+(
+    T& Value,
+    const CombineOp& cop,
+    const int tag = Pstream::msgType(),
+    const label comm = Pstream::worldComm
+)
 {
     if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
     {
-        Pstream::combineGather(Pstream::linearCommunication(), Value, cop);
-        Pstream::combineScatter(Pstream::linearCommunication(), Value);
+        Pstream::combineGather
+        (
+            Pstream::linearCommunication(comm),
+            Value,
+            cop,
+            tag,
+            comm
+        );
+
+        Pstream::combineScatter
+        (
+            Pstream::linearCommunication(comm),
+            Value,
+            tag,
+            comm
+        );
     }
     else
     {
-        Pstream::combineGather(Pstream::treeCommunication(), Value, cop);
-        Pstream::combineScatter(Pstream::treeCommunication(), Value);
+        Pstream::combineGather
+        (
+            Pstream::treeCommunication(comm),
+            Value,
+            cop,
+            tag,
+            comm
+        );
+
+        Pstream::combineScatter
+        (
+            Pstream::treeCommunication(comm),
+            Value,
+            tag,
+            comm
+        );
     }
 }
 

src/foam/db/IOstreams/Pstreams/PstreamCommsStruct.C

@@ -68,13 +68,13 @@ Foam::Pstream::commsStruct::commsStruct
 {
     boolList inBelow(nProcs, false);
 
-    forAll(allBelow, belowI)
+    forAll (allBelow, belowI)
     {
         inBelow[allBelow[belowI]] = true;
     }
 
     label notI = 0;
-    forAll(inBelow, procI)
+    forAll (inBelow, procI)
     {
         if ((procI != myProcID) && !inBelow[procI])
         {

src/foam/db/IOstreams/Pstreams/PstreamExchange.C

@@ -0,0 +1,159 @@
+/*---------------------------------------------------------------------------*\
+  =========                 |
+  \\      /  F ield         | foam-extend: Open Source CFD
+   \\    /   O peration     | Version:     4.0
+    \\  /    A nd           | Web:         http://www.foam-extend.org
+     \\/     M anipulation  | For copyright notice see file Copyright
+-------------------------------------------------------------------------------
+License
+    This file is part of foam-extend.
+
+    foam-extend 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.
+
+    foam-extend 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 foam-extend.  If not, see <http://www.gnu.org/licenses/>.
+
+Description
+    Pstream exchange data.
+
+\*---------------------------------------------------------------------------*/
+
+#include "Pstream.H"
+#include "contiguous.H"
+#include "PstreamCombineReduceOps.H"
+
+// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
+
+namespace Foam
+{
+
+// * * * * * * * * * * * * * * * Member Functions  * * * * * * * * * * * * * //
+
+//template<template<class> class ListType, class T>
+template<class Container, class T>
+void Pstream::exchange
+(
+    const List<Container>& sendBufs,
+    List<Container>& recvBufs,
+    labelListList& sizes,
+    const int tag,
+    const label comm,
+    const bool block
+)
+{
+    if (!contiguous<T>())
+    {
+        FatalErrorIn
+        (
+            "Pstream::exchange(..)"
+        )   << "Continuous data only." << Foam::abort(FatalError);
+    }
+
+    if (sendBufs.size() != Pstream::nProcs(comm))
+    {
+        FatalErrorIn
+        (
+            "Pstream::exchange(..)"
+        )   << "Size of list:" << sendBufs.size()
+            << " does not equal the number of processors:"
+            << Pstream::nProcs(comm)
+            << Foam::abort(FatalError);
+    }
+
+    sizes.setSize(Pstream::nProcs(comm));
+    labelList& nsTransPs = sizes[Pstream::myProcNo(comm)];
+    nsTransPs.setSize(Pstream::nProcs(comm));
+
+    forAll (sendBufs, procI)
+    {
+        nsTransPs[procI] = sendBufs[procI].size();
+    }
+
+    // Send sizes across. Note: blocks.
+    combineReduce(sizes, Pstream::listEq(), tag, comm);
+
+    if (Pstream::nProcs(comm) > 1)
+    {
+        label startOfRequests = Pstream::nRequests();
+
+        // Set up receives
+        // ~~~~~~~~~~~~~~~
+
+        recvBufs.setSize(sendBufs.size());
+        forAll (sizes, procI)
+        {
+            label nRecv = sizes[procI][Pstream::myProcNo(comm)];
+
+            if (procI != Pstream::myProcNo(comm) && nRecv > 0)
+            {
+                recvBufs[procI].setSize(nRecv);
+                IPstream::read
+                (
+                    Pstream::nonBlocking,
+                    procI,
+                    reinterpret_cast<char*>(recvBufs[procI].begin()),
+                    nRecv*sizeof(T),
+                    tag,
+                    comm
+                );
+            }
+        }
+
+
+        // Set up sends
+        // ~~~~~~~~~~~~
+
+        forAll (sendBufs, procI)
+        {
+            if (procI != Pstream::myProcNo(comm) && sendBufs[procI].size() > 0)
+            {
+                if
+                (
+                   !OPstream::write
+                    (
+                        Pstream::nonBlocking,
+                        procI,
+                        reinterpret_cast<const char*>(sendBufs[procI].begin()),
+                        sendBufs[procI].size()*sizeof(T),
+                        tag,
+                        comm
+                    )
+                )
+                {
+                    FatalErrorIn("Pstream::exchange(..)")
+                        << "Cannot send outgoing message. "
+                        << "to:" << procI << " nBytes:"
+                        << label(sendBufs[procI].size()*sizeof(T))
+                        << Foam::abort(FatalError);
+                }
+            }
+        }
+
+
+        // Wait for all to finish
+        // ~~~~~~~~~~~~~~~~~~~~~~
+
+        if (block)
+        {
+            Pstream::waitRequests(startOfRequests);
+        }
+    }
+
+    // Do myself
+    recvBufs[Pstream::myProcNo(comm)] = sendBufs[Pstream::myProcNo(comm)];
+}
+
+
+// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
+
+} // End namespace Foam
+
+// ************************************************************************* //

src/foam/db/IOstreams/Pstreams/PstreamGlobals.C

@@ -33,10 +33,50 @@ namespace Foam
 // * * * * * * * * * * * * * * Static Data Members * * * * * * * * * * * * * //
 
 // Outstanding non-blocking operations.
-//! @cond fileScope
-DynamicList<MPI_Request> PstreamGlobals::IPstream_outstandingRequests_;
-DynamicList<MPI_Request> PstreamGlobals::OPstream_outstandingRequests_;
-//! @endcond
+//! \cond fileScope
+DynamicList<MPI_Request> PstreamGlobals::outstandingRequests_;
+//! \endcond
+
+// Max outstanding message tag operations.
+//! \cond fileScope
+int PstreamGlobals::nTags_ = 0;
+//! \endcond
+
+// Free'd message tags
+//! \cond fileScope
+DynamicList<int> PstreamGlobals::freedTags_;
+//! \endcond
+
+
+// Allocated communicators.
+//! \cond fileScope
+DynamicList<MPI_Comm> PstreamGlobals::MPICommunicators_;
+DynamicList<MPI_Group> PstreamGlobals::MPIGroups_;
+//! \endcond
+
+void PstreamGlobals::checkCommunicator
+(
+    const label comm,
+    const label otherProcNo
+)
+{
+    if
+    (
+        comm < 0
+     || comm >= PstreamGlobals::MPICommunicators_.size()
+    )
+    {
+        FatalErrorIn
+        (
+            "PstreamGlobals::checkCommunicator(const label, const label)"
+        )   << "otherProcNo:" << otherProcNo << " : illegal communicator "
+            << comm << endl
+            << "Communicator should be within range 0.."
+            << PstreamGlobals::MPICommunicators_.size() - 1
+            << abort(FatalError);
+    }
+}
+
 
 // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
 

src/foam/db/IOstreams/Pstreams/PstreamGlobals.H

@@ -46,14 +46,24 @@ namespace Foam
 {
 
 /*---------------------------------------------------------------------------*\
-                     Class PstreamGlobals Declaration
+                        Class PstreamGlobals Declaration
 \*---------------------------------------------------------------------------*/
 
 namespace PstreamGlobals
 {
 
-extern DynamicList<MPI_Request> IPstream_outstandingRequests_;
-extern DynamicList<MPI_Request> OPstream_outstandingRequests_;
+extern DynamicList<MPI_Request> outstandingRequests_;
+
+extern int nTags_;
+
+extern DynamicList<int> freedTags_;
+
+
+// Current communicators. First element will be MPI_COMM_WORLD
+extern DynamicList<MPI_Comm> MPICommunicators_;
+extern DynamicList<MPI_Group> MPIGroups_;
+
+void checkCommunicator(const label, const label procNo);
 
 };
 

src/foam/db/IOstreams/Pstreams/PstreamReduceOps.C

@@ -21,28 +21,11 @@ License
     You should have received a copy of the GNU General Public License
     along with foam-extend.  If not, see <http://www.gnu.org/licenses/>.
 
-Description
-    Prints out a description of the streams
-
 \*---------------------------------------------------------------------------*/
 
-#include "IPstream.H"
-#include "OPstream.H"
+#include "PstreamReduceOps.H"
+#include "allReduce.H"
 
 // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
 
-void Foam::IPstream::print(Ostream& os) const
-{
-    os  << "Reading from processor " << fromProcNo_
-        << " to processor " << myProcNo() << Foam::endl;
-}
-
-
-void Foam::OPstream::print(Ostream& os) const
-{
-    os  << "Writing from processor " << toProcNo_
-        << " to processor " << myProcNo() << Foam::endl;
-}
-
-
 // ************************************************************************* //

src/foam/db/IOstreams/Pstreams/PstreamReduceOps.H

@@ -28,6 +28,7 @@ License
 
 #include "Pstream.H"
 #include "ops.H"
+#include "vector2D.H"
 
 // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
 
@@ -42,11 +43,20 @@ void reduce
 (
     const List<Pstream::commsStruct>& comms,
     T& Value,
-    const BinaryOp& bop
+    const BinaryOp& bop,
+    const int tag,
+    const label comm
 )
 {
-    Pstream::gather(comms, Value, bop);
-    Pstream::scatter(comms, Value);
+    if (Pstream::warnComm != -1 && comm != Pstream::warnComm)
+    {
+        Pout<< "** reducing:" << Value << " with comm:" << comm
+            << endl;
+        error::printStack(Pout);
+    }
+
+    Pstream::gather(comms, Value, bop, tag, comm);
+    Pstream::scatter(comms, Value, tag, comm);
 }
 
 
@@ -55,16 +65,18 @@ template <class T, class BinaryOp>
 void reduce
 (
     T& Value,
-    const BinaryOp& bop
+    const BinaryOp& bop,
+    const int tag = Pstream::msgType(),
+    const label comm = Pstream::worldComm
 )
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        reduce(Pstream::linearCommunication(), Value, bop);
+        reduce(Pstream::linearCommunication(comm), Value, bop, tag, comm);
     }
     else
     {
-        reduce(Pstream::treeCommunication(), Value, bop);
+        reduce(Pstream::treeCommunication(comm), Value, bop, tag, comm);
     }
 }
 
@@ -74,26 +86,100 @@ template <class T, class BinaryOp>
 T returnReduce
 (
     const T& Value,
-    const BinaryOp& bop
+    const BinaryOp& bop,
+    const int tag = Pstream::msgType(),
+    const label comm = Pstream::worldComm
 )
 {
     T WorkValue(Value);
 
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        reduce(Pstream::linearCommunication(), WorkValue, bop);
+        reduce(Pstream::linearCommunication(comm), WorkValue, bop, tag, comm);
     }
     else
     {
-        reduce(Pstream::treeCommunication(), WorkValue, bop);
+        reduce(Pstream::treeCommunication(comm), WorkValue, bop, tag, comm);
     }
 
     return WorkValue;
 }
 
 
-// Insist there is a specialisation for the reduction of a scalar
-void reduce(scalar& Value, const sumOp<scalar>& bop);
+// Reduce with sum of both value and count (for averaging)
+template<class T>
+void sumReduce
+(
+    T& Value,
+    label& Count,
+    const int tag = Pstream::msgType(),
+    const label comm = Pstream::worldComm
+)
+{
+    reduce(Value, sumOp<T>(), tag, comm);
+    reduce(Count, sumOp<label>(), tag, comm);
+}
+
+
+// Non-blocking version of reduce. Sets request.
+template<class T, class BinaryOp>
+void reduce
+(
+    T& Value,
+    const BinaryOp& bop,
+    const int tag,
+    const label comm,
+    label& request
+)
+{
+    notImplemented
+    (
+        "reduce(T&, const BinaryOp&, const int, const label, label&"
+    );
+}
+
+
+// Insist there are specialisations for the common reductions of scalar(s)
+void reduce
+(
+    scalar& Value,
+    const sumOp<scalar>& bop,
+    const int tag = Pstream::msgType(),
+    const label comm = Pstream::worldComm
+);
+
+void reduce
+(
+    scalar& Value,
+    const minOp<scalar>& bop,
+    const int tag = Pstream::msgType(),
+    const label comm = Pstream::worldComm
+);
+
+void reduce
+(
+    vector2D& Value,
+    const sumOp<vector2D>& bop,
+    const int tag = Pstream::msgType(),
+    const label comm = Pstream::worldComm
+);
+
+void sumReduce
+(
+    scalar& Value,
+    label& Count,
+    const int tag = Pstream::msgType(),
+    const label comm = Pstream::worldComm
+);
+
+void reduce
+(
+    scalar& Value,
+    const sumOp<scalar>& bop,
+    const int tag,
+    const label comm,
+    label& request
+);
 
 
 // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //

src/foam/db/IOstreams/Pstreams/allReduce.H

@@ -0,0 +1,71 @@
+/*---------------------------------------------------------------------------*\
+  =========                 |
+  \\      /  F ield         | foam-extend: Open Source CFD
+   \\    /   O peration     | Version:     4.0
+    \\  /    A nd           | Web:         http://www.foam-extend.org
+     \\/     M anipulation  | For copyright notice see file Copyright
+-------------------------------------------------------------------------------
+License
+    This file is part of foam-extend.
+
+    foam-extend 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.
+
+    foam-extend 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 foam-extend.  If not, see <http://www.gnu.org/licenses/>.
+
+Description
+    Various functions to wrap MPI_Allreduce
+
+SourceFiles
+    allReduceTemplates.C
+
+\*---------------------------------------------------------------------------*/
+
+#ifndef allReduce_H
+#define allReduce_H
+
+#include "mpi.h"
+
+#include "Pstream.H"
+#include "PstreamGlobals.H"
+
+// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
+
+namespace Foam
+{
+
+template<class Type, class BinaryOp>
+void allReduce
+(
+    Type& Value,
+    int count,
+    MPI_Datatype MPIType,
+    MPI_Op op,
+    const BinaryOp& bop,
+    const int tag,
+    const int communicator
+);
+
+// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
+
+} // End namespace Foam
+
+// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
+
+#ifdef NoRepository
+#   include "allReduceTemplates.C"
+#endif
+
+// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
+
+#endif
+
+// ************************************************************************* //

src/foam/db/IOstreams/Pstreams/allReduceTemplates.C

@@ -0,0 +1,211 @@
+/*---------------------------------------------------------------------------*\
+  =========                 |
+  \\      /  F ield         | foam-extend: Open Source CFD
+   \\    /   O peration     | Version:     4.0
+    \\  /    A nd           | Web:         http://www.foam-extend.org
+     \\/     M anipulation  | For copyright notice see file Copyright
+-------------------------------------------------------------------------------
+License
+    This file is part of foam-extend.
+
+    foam-extend 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.
+
+    foam-extend 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 foam-extend.  If not, see <http://www.gnu.org/licenses/>.
+
+\*---------------------------------------------------------------------------*/
+
+#include "allReduce.H"
+
+// * * * * * * * * * * * * * * * Global Functions  * * * * * * * * * * * * * //
+
+template<class Type, class BinaryOp>
+void Foam::allReduce
+(
+    Type& Value,
+    int MPICount,
+    MPI_Datatype MPIType,
+    MPI_Op MPIOp,
+    const BinaryOp& bop,
+    const int tag,
+    const label comm
+)
+{
+    if (!Pstream::parRun())
+    {
+        return;
+    }
+
+    if (Pstream::nProcs(comm) <= Pstream::nProcsSimpleSum)
+    {
+        if (Pstream::master(comm))
+        {
+            for
+            (
+                int slave = Pstream::firstSlave();
+                slave <= Pstream::lastSlave(comm);
+                slave++
+            )
+            {
+                Type value;
+
+                if
+                (
+                    MPI_Recv
+                    (
+                        &value,
+                        MPICount,
+                        MPIType,
+                        slave,  //Pstream::procID(slave),
+                        tag,
+                        PstreamGlobals::MPICommunicators_[comm],
+                        MPI_STATUS_IGNORE
+                    )
+                )
+                {
+                    FatalErrorIn
+                    (
+                        "void Foam::allReduce\n"
+                        "(\n"
+                        "    Type&,\n"
+                        "    int,\n"
+                        "    MPI_Datatype,\n"
+                        "    MPI_Op,\n"
+                        "    const BinaryOp&,\n"
+                        "    const int\n"
+                        ")\n"
+                    )   << "MPI_Recv failed"
+                        << Foam::abort(FatalError);
+                }
+
+                Value = bop(Value, value);
+            }
+        }
+        else
+        {
+            if
+            (
+                MPI_Send
+                (
+                    &Value,
+                    MPICount,
+                    MPIType,
+                    Pstream::masterNo(),//Pstream::procID(masterNo()),
+                    tag,
+                    PstreamGlobals::MPICommunicators_[comm]
+                )
+            )
+            {
+                FatalErrorIn
+                (
+                    "void Foam::allReduce\n"
+                    "(\n"
+                    "    Type&,\n"
+                    "    int,\n"
+                    "    MPI_Datatype,\n"
+                    "    MPI_Op,\n"
+                    "    const BinaryOp&,\n"
+                    "    const int\n"
+                    ")\n"
+                )   << "MPI_Send failed"
+                    << Foam::abort(FatalError);
+            }
+        }
+
+
+        if (Pstream::master(comm))
+        {
+            for
+            (
+                int slave = Pstream::firstSlave();
+                slave <= Pstream::lastSlave(comm);
+                slave++
+            )
+            {
+                if
+                (
+                    MPI_Send
+                    (
+                        &Value,
+                        MPICount,
+                        MPIType,
+                        slave,      //Pstream::procID(slave),
+                        tag,
+                        PstreamGlobals::MPICommunicators_[comm]
+                    )
+                )
+                {
+                    FatalErrorIn
+                    (
+                        "void Foam::allReduce\n"
+                        "(\n"
+                        "    Type&,\n"
+                        "    int,\n"
+                        "    MPI_Datatype,\n"
+                        "    MPI_Op,\n"
+                        "    const BinaryOp&,\n"
+                        "    const int\n"
+                        ")\n"
+                    )   << "MPI_Send failed"
+                        << Foam::abort(FatalError);
+                }
+            }
+        }
+        else
+        {
+            if
+            (
+                MPI_Recv
+                (
+                    &Value,
+                    MPICount,
+                    MPIType,
+                    Pstream::masterNo(),//Pstream::procID(masterNo()),
+                    tag,
+                    PstreamGlobals::MPICommunicators_[comm],
+                    MPI_STATUS_IGNORE
+                )
+            )
+            {
+                FatalErrorIn
+                (
+                    "void Foam::allReduce\n"
+                    "(\n"
+                    "    Type&,\n"
+                    "    int,\n"
+                    "    MPI_Datatype,\n"
+                    "    MPI_Op,\n"
+                    "    const BinaryOp&,\n"
+                    "    const int\n"
+                    ")\n"
+                )   << "MPI_Recv failed"
+                    << Foam::abort(FatalError);
+            }
+        }
+    }
+    else
+    {
+        Type sum;
+        MPI_Allreduce
+        (
+            &Value,
+            &sum,
+            MPICount,
+            MPIType,
+            MPIOp,
+            PstreamGlobals::MPICommunicators_[comm]
+        );
+        Value = sum;
+    }
+}
+
+
+// ************************************************************************* //

src/foam/db/IOstreams/Pstreams/combineGatherScatter.C

@@ -50,16 +50,18 @@ void Pstream::combineGather
 (
     const List<Pstream::commsStruct>& comms,
     T& Value,
-    const CombineOp& cop
+    const CombineOp& cop,
+    const int tag,
+    const label comm
 )
 {
-    if (Pstream::parRun())
+    if (Pstream::nProcs(comm) > 1)
     {
         // Get my communication order
-        const commsStruct& myComm = comms[Pstream::myProcNo()];
+        const commsStruct& myComm = comms[Pstream::myProcNo(comm)];
 
         // Receive from my downstairs neighbours
-        forAll(myComm.below(), belowI)
+        forAll (myComm.below(), belowI)
         {
             label belowID = myComm.below()[belowI];
 
@@ -71,7 +73,9 @@ void Pstream::combineGather
                     Pstream::scheduled,
                     belowID,
                     reinterpret_cast<char*>(&value),
-                    sizeof(T)
+                    sizeof(T),
+                    tag,
+                    comm
                 );
 
                 if (debug > 1)
@@ -84,7 +88,7 @@ void Pstream::combineGather
             }
             else
             {
-                IPstream fromBelow(Pstream::scheduled, belowID);
+                IPstream fromBelow(Pstream::scheduled, belowID, 0, tag, comm);
                 T value(fromBelow);
 
                 if (debug > 1)
@@ -113,12 +117,21 @@ void Pstream::combineGather
                     Pstream::scheduled,
                     myComm.above(),
                     reinterpret_cast<const char*>(&Value),
-                    sizeof(T)
+                    sizeof(T),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                OPstream toAbove(Pstream::scheduled, myComm.above());
+                OPstream toAbove
+                (
+                    Pstream::scheduled,
+                    myComm.above(),
+                    0,
+                    tag,
+                    comm
+                );
                 toAbove << Value;
             }
         }
@@ -127,26 +140,52 @@ void Pstream::combineGather
 
 
 template <class T, class CombineOp>
-void Pstream::combineGather(T& Value, const CombineOp& cop)
+void Pstream::combineGather
+(
+    T& Value,
+    const CombineOp& cop,
+    const int tag,
+    const label comm
+)
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        combineGather(Pstream::linearCommunication(), Value, cop);
+        combineGather
+        (
+            Pstream::linearCommunication(comm),
+            Value,
+            cop,
+            tag,
+            comm
+        );
     }
     else
     {
-        combineGather(Pstream::treeCommunication(), Value, cop);
+        combineGather
+        (
+            Pstream::treeCommunication(comm),
+            Value,
+            cop,
+            tag,
+            comm
+        );
     }
 }
 
 
 template <class T>
-void Pstream::combineScatter(const List<Pstream::commsStruct>& comms, T& Value)
+void Pstream::combineScatter
+(
+    const List<Pstream::commsStruct>& comms,
+    T& Value,
+    const int tag,
+    const label comm
+)
 {
-    if (Pstream::parRun())
+    if (Pstream::nProcs(comm) > 1)
     {
         // Get my communication order
-        const Pstream::commsStruct& myComm = comms[Pstream::myProcNo()];
+        const Pstream::commsStruct& myComm = comms[Pstream::myProcNo(comm)];
 
         // Receive from up
         if (myComm.above() != -1)
@@ -158,12 +197,21 @@ void Pstream::combineScatter(const List<Pstream::commsStruct>& comms, T& Value)
                     Pstream::scheduled,
                     myComm.above(),
                     reinterpret_cast<char*>(&Value),
-                    sizeof(T)
+                    sizeof(T),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                IPstream fromAbove(Pstream::scheduled, myComm.above());
+                IPstream fromAbove
+                (
+                    Pstream::scheduled,
+                    myComm.above(),
+                    0,
+                    tag,
+                    comm
+                );
                 Value = T(fromAbove);
             }
 
@@ -175,7 +223,7 @@ void Pstream::combineScatter(const List<Pstream::commsStruct>& comms, T& Value)
         }
 
         // Send to my downstairs neighbours
-        forAll(myComm.below(), belowI)
+        forAll (myComm.below(), belowI)
         {
             label belowID = myComm.below()[belowI];
 
@@ -191,12 +239,14 @@ void Pstream::combineScatter(const List<Pstream::commsStruct>& comms, T& Value)
                     Pstream::scheduled,
                     belowID,
                     reinterpret_cast<const char*>(&Value),
-                    sizeof(T)
+                    sizeof(T),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                OPstream toBelow(Pstream::scheduled, belowID);
+                OPstream toBelow(Pstream::scheduled, belowID, 0, tag, comm);
                 toBelow << Value;
             }
         }
@@ -205,15 +255,20 @@ void Pstream::combineScatter(const List<Pstream::commsStruct>& comms, T& Value)
 
 
 template <class T>
-void Pstream::combineScatter(T& Value)
+void Pstream::combineScatter
+(
+    T& Value,
+    const int tag,
+    const label comm
+)
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        combineScatter(Pstream::linearCommunication(), Value);
+        combineScatter(Pstream::linearCommunication(comm), Value, tag, comm);
     }
     else
     {
-        combineScatter(Pstream::treeCommunication(), Value);
+        combineScatter(Pstream::treeCommunication(comm), Value, tag, comm);
     }
 }
 
@@ -227,16 +282,18 @@ void Pstream::listCombineGather
 (
     const List<Pstream::commsStruct>& comms,
     List<T>& Values,
-    const CombineOp& cop
+    const CombineOp& cop,
+    const int tag,
+    const label comm
 )
 {
-    if (Pstream::parRun())
+    if (Pstream::nProcs(comm) > 1)
     {
         // Get my communication order
-        const commsStruct& myComm = comms[Pstream::myProcNo()];
+        const commsStruct& myComm = comms[Pstream::myProcNo(comm)];
 
         // Receive from my downstairs neighbours
-        forAll(myComm.below(), belowI)
+        forAll (myComm.below(), belowI)
         {
             label belowID = myComm.below()[belowI];
 
@@ -249,7 +306,9 @@ void Pstream::listCombineGather
                     Pstream::scheduled,
                     belowID,
                     reinterpret_cast<char*>(receivedValues.begin()),
-                    receivedValues.byteSize()
+                    receivedValues.byteSize(),
+                    tag,
+                    comm
                 );
 
                 if (debug > 1)
@@ -258,14 +317,14 @@ void Pstream::listCombineGather
                         << belowID << " data:" << receivedValues << endl;
                 }
 
-                forAll(Values, i)
+                forAll (Values, i)
                 {
                     cop(Values[i], receivedValues[i]);
                 }
             }
             else
             {
-                IPstream fromBelow(Pstream::scheduled, belowID);
+                IPstream fromBelow(Pstream::scheduled, belowID, 0, tag, comm);
                 List<T> receivedValues(fromBelow);
 
                 if (debug > 1)
@@ -274,7 +333,7 @@ void Pstream::listCombineGather
                         << belowID << " data:" << receivedValues << endl;
                 }
 
-                forAll(Values, i)
+                forAll (Values, i)
                 {
                     cop(Values[i], receivedValues[i]);
                 }
@@ -297,12 +356,21 @@ void Pstream::listCombineGather
                     Pstream::scheduled,
                     myComm.above(),
                     reinterpret_cast<const char*>(Values.begin()),
-                    Values.byteSize()
+                    Values.byteSize(),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                OPstream toAbove(Pstream::scheduled, myComm.above());
+                OPstream toAbove
+                (
+                    Pstream::scheduled,
+                    myComm.above(),
+                    0,
+                    tag,
+                    comm
+                );
                 toAbove << Values;
             }
         }
@@ -311,15 +379,35 @@ void Pstream::listCombineGather
 
 
 template <class T, class CombineOp>
-void Pstream::listCombineGather(List<T>& Values, const CombineOp& cop)
+void Pstream::listCombineGather
+(
+    List<T>& Values,
+    const CombineOp& cop,
+    const int tag,
+    const label comm
+)
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        listCombineGather(Pstream::linearCommunication(), Values, cop);
+        listCombineGather
+        (
+            Pstream::linearCommunication(comm),
+            Values,
+            cop,
+            tag,
+            comm
+        );
     }
     else
     {
-        listCombineGather(Pstream::treeCommunication(), Values, cop);
+        listCombineGather
+        (
+            Pstream::treeCommunication(comm),
+            Values,
+            cop,
+            tag,
+            comm
+        );
     }
 }
 
@@ -328,13 +416,15 @@ template <class T>
 void Pstream::listCombineScatter
 (
     const List<Pstream::commsStruct>& comms,
-    List<T>& Values
+    List<T>& Values,
+    const int tag,
+    const label comm
 )
 {
-    if (Pstream::parRun())
+    if (Pstream::nProcs(comm) > 1)
     {
         // Get my communication order
-        const Pstream::commsStruct& myComm = comms[Pstream::myProcNo()];
+        const Pstream::commsStruct& myComm = comms[Pstream::myProcNo(comm)];
 
         // Receive from up
         if (myComm.above() != -1)
@@ -346,12 +436,21 @@ void Pstream::listCombineScatter
                     Pstream::scheduled,
                     myComm.above(),
                     reinterpret_cast<char*>(Values.begin()),
-                    Values.byteSize()
+                    Values.byteSize(),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                IPstream fromAbove(Pstream::scheduled, myComm.above());
+                IPstream fromAbove
+                (
+                    Pstream::scheduled,
+                    myComm.above(),
+                    0,
+                    tag,
+                    comm
+                );
                 fromAbove >> Values;
             }
 
@@ -363,7 +462,7 @@ void Pstream::listCombineScatter
         }
 
         // Send to my downstairs neighbours
-        forAll(myComm.below(), belowI)
+        forAll (myComm.below(), belowI)
         {
             label belowID = myComm.below()[belowI];
 
@@ -379,12 +478,14 @@ void Pstream::listCombineScatter
                     Pstream::scheduled,
                     belowID,
                     reinterpret_cast<const char*>(Values.begin()),
-                    Values.byteSize()
+                    Values.byteSize(),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                OPstream toBelow(Pstream::scheduled, belowID);
+                OPstream toBelow(Pstream::scheduled, belowID, 0, tag, comm);
                 toBelow << Values;
             }
         }
@@ -393,44 +494,60 @@ void Pstream::listCombineScatter
 
 
 template <class T>
-void Pstream::listCombineScatter(List<T>& Values)
+void Pstream::listCombineScatter
+(
+    List<T>& Values,
+    const int tag,
+    const label comm
+)
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        listCombineScatter(Pstream::linearCommunication(), Values);
+        listCombineScatter
+        (
+            Pstream::linearCommunication(),
+            Values,
+            tag,
+            comm
+        );
     }
     else
     {
-        listCombineScatter(Pstream::treeCommunication(), Values);
+        listCombineScatter
+        (
+            Pstream::treeCommunication(),
+            Values,
+            tag,
+            comm
+        );
     }
 }
 
 
-
-
 // Same thing but for sparse list (map)
 // ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-
 template <class Container, class CombineOp>
 void Pstream::mapCombineGather
 (
     const List<Pstream::commsStruct>& comms,
     Container& Values,
-    const CombineOp& cop
+    const CombineOp& cop,
+    const int tag,
+    const label comm
 )
 {
-    if (Pstream::parRun())
+    if (Pstream::nProcs(comm) > 1)
     {
         // Get my communication order
-        const commsStruct& myComm = comms[Pstream::myProcNo()];
+        const commsStruct& myComm = comms[Pstream::myProcNo(comm)];
 
         // Receive from my downstairs neighbours
-        forAll(myComm.below(), belowI)
+        forAll (myComm.below(), belowI)
         {
             label belowID = myComm.below()[belowI];
 
-            IPstream fromBelow(Pstream::scheduled, belowID);
+            IPstream fromBelow(Pstream::scheduled, belowID, 0, tag, comm);
             Container receivedValues(fromBelow);
 
             if (debug > 1)
@@ -470,7 +587,7 @@ void Pstream::mapCombineGather
                     << " data:" << Values << endl;
             }
 
-            OPstream toAbove(Pstream::scheduled, myComm.above());
+            OPstream toAbove(Pstream::scheduled, myComm.above(), 0, tag, comm);
             toAbove << Values;
         }
     }
@@ -478,15 +595,35 @@ void Pstream::mapCombineGather
 
 
 template <class Container, class CombineOp>
-void Pstream::mapCombineGather(Container& Values, const CombineOp& cop)
+void Pstream::mapCombineGather
+(
+    Container& Values,
+    const CombineOp& cop,
+    const int tag,
+    const label comm
+)
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        mapCombineGather(Pstream::linearCommunication(), Values, cop);
+        mapCombineGather
+        (
+            Pstream::linearCommunication(),
+            Values,
+            cop,
+            tag,
+            comm
+        );
     }
     else
     {
-        mapCombineGather(Pstream::treeCommunication(), Values, cop);
+        mapCombineGather
+        (
+            Pstream::treeCommunication(),
+            Values,
+            cop,
+            tag,
+            comm
+        );
     }
 }
 
@@ -495,18 +632,27 @@ template <class Container>
 void Pstream::mapCombineScatter
 (
     const List<Pstream::commsStruct>& comms,
-    Container& Values
+    Container& Values,
+    const int tag,
+    const label comm
 )
 {
-    if (Pstream::parRun())
+    if (Pstream::nProcs(comm) > 1)
     {
         // Get my communication order
-        const Pstream::commsStruct& myComm = comms[Pstream::myProcNo()];
+        const Pstream::commsStruct& myComm = comms[Pstream::myProcNo(comm)];
 
         // Receive from up
         if (myComm.above() != -1)
         {
-            IPstream fromAbove(Pstream::scheduled, myComm.above());
+            IPstream fromAbove
+            (
+                Pstream::scheduled,
+                myComm.above(),
+                0,
+                tag,
+                comm
+            );
             fromAbove >> Values;
 
             if (debug > 1)
@@ -517,7 +663,7 @@ void Pstream::mapCombineScatter
         }
 
         // Send to my downstairs neighbours
-        forAll(myComm.below(), belowI)
+        forAll (myComm.below(), belowI)
         {
             label belowID = myComm.below()[belowI];
 
@@ -526,7 +672,7 @@ void Pstream::mapCombineScatter
                 Pout<< " sending to " << belowID << " data:" << Values << endl;
             }
 
-            OPstream toBelow(Pstream::scheduled, belowID);
+            OPstream toBelow(Pstream::scheduled, belowID, 0, tag, comm);
             toBelow << Values;
         }
     }
@@ -534,15 +680,32 @@ void Pstream::mapCombineScatter
 
 
 template <class Container>
-void Pstream::mapCombineScatter(Container& Values)
+void Pstream::mapCombineScatter
+(
+    Container& Values,
+    const int tag,
+    const label comm
+)
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        mapCombineScatter(Pstream::linearCommunication(), Values);
+        mapCombineScatter
+        (
+            Pstream::linearCommunication(),
+            Values,
+            tag,
+            comm
+        );
     }
     else
     {
-        mapCombineScatter(Pstream::treeCommunication(), Values);
+        mapCombineScatter
+        (
+            Pstream::treeCommunication(),
+            Values,
+            tag,
+            comm
+        );
     }
 }
 

src/foam/db/IOstreams/Pstreams/gatherScatter.C

@@ -45,16 +45,18 @@ void Pstream::gather
 (
     const List<Pstream::commsStruct>& comms,
     T& Value,
-    const BinaryOp& bop
+    const BinaryOp& bop,
+    const int tag,
+    const label comm
 )
 {
-    if (Pstream::parRun())
+    if (Pstream::nProcs(comm) > 1)
     {
         // Get my communication order
-        const commsStruct& myComm = comms[Pstream::myProcNo()];
+        const commsStruct& myComm = comms[Pstream::myProcNo(comm)];
 
         // Receive from my downstairs neighbours
-        forAll(myComm.below(), belowI)
+        forAll (myComm.below(), belowI)
         {
             T value;
 
@@ -65,12 +67,21 @@ void Pstream::gather
                     Pstream::scheduled,
                     myComm.below()[belowI],
                     reinterpret_cast<char*>(&value),
-                    sizeof(T)
+                    sizeof(T),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                IPstream fromBelow(Pstream::scheduled, myComm.below()[belowI]);
+                IPstream fromBelow
+                (
+                    Pstream::scheduled,
+                    myComm.below()[belowI],
+                    0,
+                    tag,
+                    comm
+                );
                 fromBelow >> value;
             }
 
@@ -87,12 +98,21 @@ void Pstream::gather
                     Pstream::scheduled,
                     myComm.above(),
                     reinterpret_cast<const char*>(&Value),
-                    sizeof(T)
+                    sizeof(T),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                OPstream toAbove(Pstream::scheduled, myComm.above());
+                OPstream toAbove
+                (
+                    Pstream::scheduled,
+                    myComm.above(),
+                    0,
+                    tag,
+                    comm
+                );
                 toAbove << Value;
             }
         }
@@ -101,26 +121,38 @@ void Pstream::gather
 
 
 template <class T, class BinaryOp>
-void Pstream::gather(T& Value, const BinaryOp& bop)
+void Pstream::gather
+(
+    T& Value,
+    const BinaryOp& bop,
+    const int tag,
+    const label comm
+)
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        gather(Pstream::linearCommunication(), Value, bop);
+        gather(Pstream::linearCommunication(comm), Value, bop, tag, comm);
     }
     else
     {
-        gather(Pstream::treeCommunication(), Value, bop);
+        gather(Pstream::treeCommunication(comm), Value, bop, tag, comm);
     }
 }
 
 
 template <class T>
-void Pstream::scatter(const List<Pstream::commsStruct>& comms, T& Value)
+void Pstream::scatter
+(
+    const List<Pstream::commsStruct>& comms,
+    T& Value,
+    const int tag,
+    const label comm
+)
 {
-    if (Pstream::parRun())
+    if (Pstream::nProcs(comm) > 1)
     {
         // Get my communication order
-        const commsStruct& myComm = comms[Pstream::myProcNo()];
+        const commsStruct& myComm = comms[Pstream::myProcNo(comm)];
 
         // Receive from up
         if (myComm.above() != -1)
@@ -132,18 +164,27 @@ void Pstream::scatter(const List<Pstream::commsStruct>& comms, T& Value)
                     Pstream::scheduled,
                     myComm.above(),
                     reinterpret_cast<char*>(&Value),
-                    sizeof(T)
+                    sizeof(T),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                IPstream fromAbove(Pstream::scheduled, myComm.above());
+                IPstream fromAbove
+                (
+                    Pstream::scheduled,
+                    myComm.above(),
+                    0,
+                    tag,
+                    comm
+                );
                 fromAbove >> Value;
             }
         }
 
         // Send to my downstairs neighbours
-        forAll(myComm.below(), belowI)
+        forAll (myComm.below(), belowI)
         {
             if (contiguous<T>())
             {
@@ -152,12 +193,21 @@ void Pstream::scatter(const List<Pstream::commsStruct>& comms, T& Value)
                     Pstream::scheduled,
                     myComm.below()[belowI],
                     reinterpret_cast<const char*>(&Value),
-                    sizeof(T)
+                    sizeof(T),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                OPstream toBelow(Pstream::scheduled,myComm.below()[belowI]);
+                OPstream toBelow
+                (
+                    Pstream::scheduled,
+                    myComm.below()[belowI],
+                    0,
+                    tag,
+                    comm
+                );
                 toBelow << Value;
             }
         }
@@ -166,15 +216,15 @@ void Pstream::scatter(const List<Pstream::commsStruct>& comms, T& Value)
 
 
 template <class T>
-void Pstream::scatter(T& Value)
+void Pstream::scatter(T& Value, const int tag, const label comm)
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        scatter(Pstream::linearCommunication(), Value);
+        scatter(Pstream::linearCommunication(comm), Value, tag, comm);
     }
     else
     {
-        scatter(Pstream::treeCommunication(), Value);
+        scatter(Pstream::treeCommunication(comm), Value, tag, comm);
     }
 }
 

src/foam/db/IOstreams/Pstreams/gatherScatterList.C

@@ -49,28 +49,35 @@ template <class T>
 void Pstream::gatherList
 (
     const List<Pstream::commsStruct>& comms,
-    List<T>& Values
+    List<T>& Values,
+    const int tag,
+    const label comm
 )
 {
-    if (Pstream::parRun())
+    if (Pstream::nProcs(comm) > 1)
     {
-        if (Values.size() != Pstream::nProcs())
+        if (Values.size() != Pstream::nProcs(comm))
         {
             FatalErrorIn
             (
-                "Pstream::gatherList(const List<Pstream::commsStruct>&"
-                ", List<T>)"
+                "void Pstream::gatherList\n"
+                "(\n"
+                "    const List<Pstream::commsStruct>& comms,\n"
+                "    List<T>& Values,\n"
+                "    const int tag,\n"
+                "    const label comm\n"
+                ")"
             )   << "Size of list:" << Values.size()
                 << " does not equal the number of processors:"
-                << Pstream::nProcs()
+                << Pstream::nProcs(comm)
                 << Foam::abort(FatalError);
         }
 
         // Get my communication order
-        const commsStruct& myComm = comms[Pstream::myProcNo()];
+        const commsStruct& myComm = comms[Pstream::myProcNo(comm)];
 
         // Receive from my downstairs neighbours
-        forAll(myComm.below(), belowI)
+        forAll (myComm.below(), belowI)
         {
             label belowID = myComm.below()[belowI];
             const labelList& belowLeaves = comms[belowID].allBelow();
@@ -84,19 +91,21 @@ void Pstream::gatherList
                     Pstream::scheduled,
                     belowID,
                     reinterpret_cast<char*>(receivedValues.begin()),
-                    receivedValues.byteSize()
+                    receivedValues.byteSize(),
+                    tag,
+                    comm
                 );
 
                 Values[belowID] = receivedValues[0];
 
-                forAll(belowLeaves, leafI)
+                forAll (belowLeaves, leafI)
                 {
                     Values[belowLeaves[leafI]] = receivedValues[leafI + 1];
                 }
             }
             else
             {
-                IPstream fromBelow(Pstream::scheduled, belowID);
+                IPstream fromBelow(Pstream::scheduled, belowID, 0, tag, comm);
                 fromBelow >> Values[belowID];
 
                 if (debug > 1)
@@ -107,7 +116,7 @@ void Pstream::gatherList
                 }
 
                 // Receive from all other processors below belowID
-                forAll(belowLeaves, leafI)
+                forAll (belowLeaves, leafI)
                 {
                     label leafID = belowLeaves[leafI];
                     fromBelow >> Values[leafID];
@@ -132,16 +141,16 @@ void Pstream::gatherList
             if (debug > 1)
             {
                 Pout<< " sending to " << myComm.above()
-                    << " data from: " << Pstream::myProcNo()
-                    << " data: " << Values[Pstream::myProcNo()] << endl;
+                    << " data from: " << Pstream::myProcNo(comm)
+                    << " data: " << Values[Pstream::myProcNo(comm)] << endl;
             }
 
             if (contiguous<T>())
             {
                 List<T> sendingValues(belowLeaves.size() + 1);
-                sendingValues[0] = Values[Pstream::myProcNo()];
+                sendingValues[0] = Values[Pstream::myProcNo(comm)];
 
-                forAll(belowLeaves, leafI)
+                forAll (belowLeaves, leafI)
                 {
                     sendingValues[leafI + 1] = Values[belowLeaves[leafI]];
                 }
@@ -151,15 +160,24 @@ void Pstream::gatherList
                     Pstream::scheduled,
                     myComm.above(),
                     reinterpret_cast<const char*>(sendingValues.begin()),
-                    sendingValues.byteSize()
+                    sendingValues.byteSize(),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                OPstream toAbove(Pstream::scheduled, myComm.above());
-                toAbove << Values[Pstream::myProcNo()];
+                OPstream toAbove
+                (
+                    Pstream::scheduled,
+                    myComm.above(),
+                    0,
+                    tag,
+                    comm
+                );
+                toAbove << Values[Pstream::myProcNo(comm)];
 
-                forAll(belowLeaves, leafI)
+                forAll (belowLeaves, leafI)
                 {
                     label leafID = belowLeaves[leafI];
 
@@ -178,15 +196,15 @@ void Pstream::gatherList
 
 
 template <class T>
-void Pstream::gatherList(List<T>& Values)
+void Pstream::gatherList(List<T>& Values, const int tag, const label comm)
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        gatherList(Pstream::linearCommunication(), Values);
+        gatherList(Pstream::linearCommunication(comm), Values, tag, comm);
     }
     else
     {
-        gatherList(Pstream::treeCommunication(), Values);
+        gatherList(Pstream::treeCommunication(comm), Values, tag, comm);
     }
 }
 
@@ -195,17 +213,24 @@ template <class T>
 void Pstream::scatterList
 (
     const List<Pstream::commsStruct>& comms,
-    List<T>& Values
+    List<T>& Values,
+    const int tag,
+    const label comm
 )
 {
-    if (Pstream::parRun())
+    if (Pstream::nProcs(comm) > 1)
     {
-        if (Values.size() != Pstream::nProcs())
+        if (Values.size() != Pstream::nProcs(comm))
         {
             FatalErrorIn
             (
-                "Pstream::scatterList(const List<Pstream::commsStruct>&"
-                ", List<T>)"
+                "void Pstream::scatterList\n"
+                "(\n"
+                "    const List<Pstream::commsStruct>& comms,\n"
+                "    List<T>& Values,\n"
+                "    const int tag,\n"
+                "    const label comm\n"
+                ")"
             )   << "Size of list:" << Values.size()
                 << " does not equal the number of processors:"
                 << Pstream::nProcs()
@@ -213,7 +238,7 @@ void Pstream::scatterList
         }
 
         // Get my communication order
-        const commsStruct& myComm = comms[Pstream::myProcNo()];
+        const commsStruct& myComm = comms[Pstream::myProcNo(comm)];
 
         // Receive from up
         if (myComm.above() != -1)
@@ -229,19 +254,28 @@ void Pstream::scatterList
                     Pstream::scheduled,
                     myComm.above(),
                     reinterpret_cast<char*>(receivedValues.begin()),
-                    receivedValues.byteSize()
+                    receivedValues.byteSize(),
+                    tag,
+                    comm
                 );
 
-                forAll(notBelowLeaves, leafI)
+                forAll (notBelowLeaves, leafI)
                 {
                     Values[notBelowLeaves[leafI]] = receivedValues[leafI];
                 }
             }
             else
             {
-                IPstream fromAbove(Pstream::scheduled, myComm.above());
+                IPstream fromAbove
+                (
+                    Pstream::scheduled,
+                    myComm.above(),
+                    0,
+                    tag,
+                    comm
+                );
 
-                forAll(notBelowLeaves, leafI)
+                forAll (notBelowLeaves, leafI)
                 {
                     label leafID = notBelowLeaves[leafI];
                     fromAbove >> Values[leafID];
@@ -257,7 +291,7 @@ void Pstream::scatterList
         }
 
         // Send to my downstairs neighbours
-        forAll(myComm.below(), belowI)
+        forAll (myComm.below(), belowI)
         {
             label belowID = myComm.below()[belowI];
             const labelList& notBelowLeaves = comms[belowID].allNotBelow();
@@ -266,7 +300,7 @@ void Pstream::scatterList
             {
                 List<T> sendingValues(notBelowLeaves.size());
 
-                forAll(notBelowLeaves, leafI)
+                forAll (notBelowLeaves, leafI)
                 {
                     sendingValues[leafI] = Values[notBelowLeaves[leafI]];
                 }
@@ -276,15 +310,17 @@ void Pstream::scatterList
                     Pstream::scheduled,
                     belowID,
                     reinterpret_cast<const char*>(sendingValues.begin()),
-                    sendingValues.byteSize()
+                    sendingValues.byteSize(),
+                    tag,
+                    comm
                 );
             }
             else
             {
-                OPstream toBelow(Pstream::scheduled, belowID);
+                OPstream toBelow(Pstream::scheduled, belowID, 0, tag, comm);
 
                 // Send data destined for all other processors below belowID
-                forAll(notBelowLeaves, leafI)
+                forAll (notBelowLeaves, leafI)
                 {
                     label leafID = notBelowLeaves[leafI];
                     toBelow << Values[leafID];
@@ -303,15 +339,15 @@ void Pstream::scatterList
 
 
 template <class T>
-void Pstream::scatterList(List<T>& Values)
+void Pstream::scatterList(List<T>& Values, const int tag, const label comm)
 {
-    if (Pstream::nProcs() < Pstream::nProcsSimpleSum())
+    if (Pstream::nProcs(comm) < Pstream::nProcsSimpleSum())
     {
-        scatterList(Pstream::linearCommunication(), Values);
+        scatterList(Pstream::linearCommunication(comm), Values, tag, comm);
     }
     else
     {
-        scatterList(Pstream::treeCommunication(), Values);
+        scatterList(Pstream::treeCommunication(comm), Values, tag, comm);
     }
 }
 

src/foam/db/regIOobject/regIOobjectRead.C

@@ -28,7 +28,6 @@ License
 #include "objectRegistry.H"
 #include "PstreamReduceOps.H"
 
-
 // * * * * * * * * * * * * * * * Member Functions  * * * * * * * * * * * * * //
 
 Foam::Istream& Foam::regIOobject::readStream()

src/foam/matrices/blockLduMatrix/BlockAMG/BlockMatrixCoarsening/BlockMatrixAgglomeration/BlockMatrixAgglomeration.C

@@ -892,6 +892,7 @@ Foam::BlockMatrixAgglomeration<Type>::restrictMatrix() const
             nCoarseEqns_,
             coarseOwner,
             coarseNeighbour,
+            Pstream::worldComm, //HJ, AMG Comm fineMesh.comm(),
             true
         )
     );
@@ -953,7 +954,8 @@ Foam::BlockMatrixAgglomeration<Type>::restrictMatrix() const
                     coarseInterfaces,
                     fineInterface,
                     fineInterface.interfaceInternalField(agglomIndex_),
-                    fineInterfaceAddr[intI]
+                    fineInterfaceAddr[intI],
+                    Pstream::worldComm //HJ, AMG Comm fineMesh.comm(),
                 ).ptr()
             );
         }

src/foam/matrices/lduMatrix/lduAddressing/lduInterfaceFields/regionCoupleLduInterfaceField/regionCoupleLduInterfaceField.H

@@ -66,7 +66,7 @@ public:
         {}
 
 
-    //-Destructor
+    //- Destructor
     virtual ~regionCoupleLduInterfaceField();
 };
 

src/foam/matrices/lduMatrix/lduAddressing/lduInterfaces/ggiLduInterface/ggiLduInterface.H

@@ -132,6 +132,15 @@ public:
         virtual void expandAddrToZone(labelField&) const = 0;
 
 
+        // Communications support
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const = 0;
+
+            //- Return message tag used for sending
+            virtual int tag() const = 0;
+
+
         // Transfer buffer access
 
             //- Return contents of the label transfer buffer

src/foam/matrices/lduMatrix/lduAddressing/lduInterfaces/mixingPlaneLduInterface/mixingPlaneLduInterface.H

@@ -110,6 +110,15 @@ public:
         virtual const tensorField& reverseT() const = 0;
 
 
+        // Communications support
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const = 0;
+
+            //- Return message tag used for sending
+            virtual int tag() const = 0;
+
+
         // Transfer buffer access
 
             //- Return contents of the label transfer buffer

src/foam/matrices/lduMatrix/lduAddressing/lduInterfaces/processorLduInterface/processorLduInterface.H

@@ -76,9 +76,8 @@ public:
         processorLduInterface();
 
 
-    // Destructor
-
-        virtual ~processorLduInterface();
+    //- Destructor
+    virtual ~processorLduInterface();
 
 
     // Member Functions
@@ -95,6 +94,15 @@ public:
             virtual const tensorField& forwardT() const = 0;
 
 
+        // Communications support
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const = 0;
+
+            //- Return message tag used for sending
+            virtual int tag() const = 0;
+
+
         // Transfer functions
 
             //- Raw send function

src/foam/matrices/lduMatrix/lduAddressing/lduInterfaces/processorLduInterface/processorLduInterfaceTemplates.C

@@ -36,6 +36,8 @@ void Foam::processorLduInterface::send
     const UList<Type>& f
 ) const
 {
+    label nBytes = f.byteSize();
+
     if (commsType == Pstream::blocking || commsType == Pstream::scheduled)
     {
         OPstream::write
@@ -43,7 +45,9 @@ void Foam::processorLduInterface::send
             commsType,
             neighbProcNo(),
             reinterpret_cast<const char*>(f.begin()),
-            f.byteSize()
+            nBytes,
+            tag(),
+            comm()
         );
     }
     else if (commsType == Pstream::nonBlocking)
@@ -55,18 +59,22 @@ void Foam::processorLduInterface::send
             commsType,
             neighbProcNo(),
             receiveBuf_.begin(),
-            receiveBuf_.size()
+            receiveBuf_.size(),
+            tag(),
+            comm()
         );
 
-        resizeBuf(sendBuf_, f.byteSize());
-        memcpy(sendBuf_.begin(), f.begin(), f.byteSize());
+        resizeBuf(sendBuf_, nBytes);
+        memcpy(sendBuf_.begin(), f.begin(), nBytes);
 
         OPstream::write
         (
             commsType,
             neighbProcNo(),
             sendBuf_.begin(),
-            f.byteSize()
+            nBytes,
+            tag(),
+            comm()
         );
     }
     else
@@ -92,7 +100,9 @@ void Foam::processorLduInterface::receive
             commsType,
             neighbProcNo(),
             reinterpret_cast<char*>(f.begin()),
-            f.byteSize()
+            f.byteSize(),
+            tag(),
+            comm()
         );
     }
     else if (commsType == Pstream::nonBlocking)
@@ -141,7 +151,7 @@ void Foam::processorLduInterface::compressedSend
         resizeBuf(sendBuf_, nBytes);
         float *fArray = reinterpret_cast<float*>(sendBuf_.begin());
 
-        for (register label i=0; i<nm1; i++)
+        for (register label i = 0; i < nm1; i++)
         {
             fArray[i] = sArray[i] - slast[i%nCmpts];
         }
@@ -155,7 +165,9 @@ void Foam::processorLduInterface::compressedSend
                 commsType,
                 neighbProcNo(),
                 sendBuf_.begin(),
-                nBytes
+                nBytes,
+                tag(),
+                comm()
             );
         }
         else if (commsType == Pstream::nonBlocking)
@@ -167,7 +179,9 @@ void Foam::processorLduInterface::compressedSend
                 commsType,
                 neighbProcNo(),
                 receiveBuf_.begin(),
-                receiveBuf_.size()
+                receiveBuf_.size(),
+                tag(),
+                comm()
             );
 
             OPstream::write
@@ -175,7 +189,9 @@ void Foam::processorLduInterface::compressedSend
                 commsType,
                 neighbProcNo(),
                 sendBuf_.begin(),
-                nBytes
+                nBytes,
+                tag(),
+                comm()
             );
         }
         else
@@ -215,7 +231,9 @@ void Foam::processorLduInterface::compressedReceive
                 commsType,
                 neighbProcNo(),
                 receiveBuf_.begin(),
-                nBytes
+                nBytes,
+                tag(),
+                comm()
             );
         }
         else if (commsType != Pstream::nonBlocking)
@@ -231,7 +249,7 @@ void Foam::processorLduInterface::compressedReceive
         scalar *sArray = reinterpret_cast<scalar*>(f.begin());
         const scalar *slast = &sArray[nm1];
 
-        for (register label i=0; i<nm1; i++)
+        for (register label i = 0; i < nm1; i++)
         {
             sArray[i] = fArray[i] + slast[i%nCmpts];
         }

src/foam/matrices/lduMatrix/lduAddressing/lduInterfaces/regionCoupleLduInterface/regionCoupleLduInterface.H

@@ -61,7 +61,6 @@ public:
 
     //- Destructor
     virtual ~regionCoupleLduInterface();
-
 };
 
 

src/foam/matrices/lduMatrix/solvers/AMG/AMGAgglomerations/GAMGAgglomeration/GAMGAgglomerateLduAddressing.C

@@ -264,6 +264,7 @@ void Foam::GAMGAgglomeration::agglomerateLduAddressing
             nCoarseCells,
             coarseOwner,
             coarseNeighbour,
+            fineMesh.comm(),
             true
         )
     );
@@ -298,7 +299,8 @@ void Foam::GAMGAgglomeration::agglomerateLduAddressing
                         (
                             restrictMap
                         ),
-                        fineInterfaceAddr[inti]
+                        fineInterfaceAddr[inti],
+                        fineMesh.comm() // Set up comm per level?
                     ).ptr()
                 );
 

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/AMGInterface/AMGInterface.H

@@ -132,14 +132,16 @@ public:
                 const lduInterfacePtrsList& coarseInterfaces,
                 const lduInterface& fineInterface,
                 const labelField& localRestrictAddressing,
-                const labelField& neighbourRestrictAddressing
+                const labelField& neighbourRestrictAddressing,
+                const label coarseComm
             ),
             (
                 lduMesh,
                 coarseInterfaces,
                 fineInterface,
                 localRestrictAddressing,
-                neighbourRestrictAddressing
+                neighbourRestrictAddressing,
+                coarseComm
             )
         );
 
@@ -154,7 +156,8 @@ public:
             const lduInterfacePtrsList& coarseInterfaces,
             const lduInterface& fineInterface,
             const labelField& localRestrictAddressing,
-            const labelField& neighbourRestrictAddressing
+            const labelField& neighbourRestrictAddressing,
+            const label coarseComm
         );
 
 
@@ -162,7 +165,10 @@ public:
 
         //- Construct from fine-level interface,
         //  local and neighbour restrict addressing
-        AMGInterface(const lduPrimitiveMesh& lduMesh)
+        AMGInterface
+        (
+            const lduPrimitiveMesh& lduMesh
+        )
         :
             lduMesh_(lduMesh)
         {}

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/AMGInterface/newAMGInterface.C

@@ -35,7 +35,8 @@ Foam::autoPtr<Foam::AMGInterface> Foam::AMGInterface::New
     const lduInterfacePtrsList& coarseInterfaces,
     const lduInterface& fineInterface,
     const labelField& localRestrictAddressing,
-    const labelField& neighbourRestrictAddressing
+    const labelField& neighbourRestrictAddressing,
+    const label coarseComm
 )
 {
     word coupleType(fineInterface.type());
@@ -53,7 +54,8 @@ Foam::autoPtr<Foam::AMGInterface> Foam::AMGInterface::New
             "    const lduInterfacePtrsList& coarseInterfaces,\n"
             "    const lduInterface& fineInterface,\n"
             "    const labelField& localRestrictAddressing,\n"
-            "    const labelField& neighbourRestrictAddressing\n"
+            "    const labelField& neighbourRestrictAddressing,\n"
+            "    const label coarseComm\n"
             ")"
         )   << "Unknown AMGInterface type " << coupleType << ".\n"
             << "Valid AMGInterface types are :"
@@ -69,7 +71,8 @@ Foam::autoPtr<Foam::AMGInterface> Foam::AMGInterface::New
             coarseInterfaces,
             fineInterface,
             localRestrictAddressing,
-            neighbourRestrictAddressing
+            neighbourRestrictAddressing,
+            coarseComm
         )
     );
 }

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/cyclicAMGInterface/cyclicAMGInterface.C

@@ -48,7 +48,8 @@ Foam::cyclicAMGInterface::cyclicAMGInterface
     const lduInterfacePtrsList& coarseInterfaces,
     const lduInterface& fineInterface,
     const labelField& localRestrictAddressing,
-    const labelField& neighbourRestrictAddressing
+    const labelField& neighbourRestrictAddressing,
+    const label coarseComm // Not needed
 )
 :
     AMGInterface(lduMesh),

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/cyclicAMGInterface/cyclicAMGInterface.H

@@ -84,13 +84,13 @@ public:
             const lduInterfacePtrsList& coarseInterfaces,
             const lduInterface& fineInterface,
             const labelField& localRestrictAddressing,
-            const labelField& neighbourRestrictAddressing
+            const labelField& neighbourRestrictAddressing,
+            const label coarseComm
         );
 
 
-    // Destructor
-
-        virtual ~cyclicAMGInterface();
+    //- Destructor
+    virtual ~cyclicAMGInterface();
 
 
     // Member Functions

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/cyclicGGIAMGInterface/cyclicGGIAMGInterface.C

@@ -48,7 +48,8 @@ Foam::cyclicGGIAMGInterface::cyclicGGIAMGInterface
     const lduInterfacePtrsList& coarseInterfaces,
     const lduInterface& fineInterface,
     const labelField& localRestrictAddressing,
-    const labelField& neighbourRestrictAddressing
+    const labelField& neighbourRestrictAddressing,
+    const label coarseComm
 )
 :
     ggiAMGInterface
@@ -57,7 +58,8 @@ Foam::cyclicGGIAMGInterface::cyclicGGIAMGInterface
         coarseInterfaces,
         fineInterface,
         localRestrictAddressing,
-        neighbourRestrictAddressing
+        neighbourRestrictAddressing,
+        coarseComm
     )
 {}
 

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/cyclicGGIAMGInterface/cyclicGGIAMGInterface.H

@@ -70,7 +70,8 @@ public:
             const lduInterfacePtrsList& coarseInterfaces,
             const lduInterface& fineInterface,
             const labelField& localRestrictAddressing,
-            const labelField& neighbourRestrictAddressing
+            const labelField& neighbourRestrictAddressing,
+            const label coarseComm
         );
 
 

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/ggiAMGInterface/ggiAMGInterface.C

@@ -195,7 +195,8 @@ Foam::ggiAMGInterface::ggiAMGInterface
     const lduInterfacePtrsList& coarseInterfaces,
     const lduInterface& fineInterface,
     const labelField& localRestrictAddressing,
-    const labelField& neighbourRestrictAddressing
+    const labelField& neighbourRestrictAddressing,
+    const label coarseComm
 )
 :
     AMGInterface(lduMesh),
@@ -203,6 +204,8 @@ Foam::ggiAMGInterface::ggiAMGInterface
     zoneSize_(0),
     zoneAddressing_(),
     procMasterFaces_(),
+    comm_(coarseComm),
+    tag_(refCast<const ggiLduInterface>(fineInterface).tag()),
     mapPtr_(NULL)
 {
     // New algorithm will assemble local clusters on the master side and

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/ggiAMGInterface/ggiAMGInterface.H

@@ -74,8 +74,15 @@ class ggiAMGInterface
         //  to allow the slave to insert faces in the same order
         labelListList procMasterFaces_;
 
+
         // Parallel communication
 
+            //- Communicator to use for parallel communication
+            const label comm_;
+
+            //- Message tag used for sending
+            const int tag_;
+
             //- Map-distribute comms tool
             mutable mapDistribute* mapPtr_;
 
@@ -118,7 +125,8 @@ public:
             const lduInterfacePtrsList& coarseInterfaces,
             const lduInterface& fineInterface,
             const labelField& localRestrictAddressing,
-            const labelField& neighbourRestrictAddressing
+            const labelField& neighbourRestrictAddressing,
+            const label coarseComm
         );
 
 
@@ -137,6 +145,21 @@ public:
             }
 
 
+        // Communications support
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const
+            {
+                return comm_;
+            }
+
+            //- Return message tag used for sending
+            virtual int tag() const
+            {
+                return tag_;
+            }
+
+
         // Agglomeration
 
             //- Agglomerating the given fine-level coefficients and return

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/mixingPlaneAMGInterface/mixingPlaneAMGInterface.C

@@ -54,14 +54,17 @@ Foam::mixingPlaneAMGInterface::mixingPlaneAMGInterface
     const lduInterfacePtrsList& coarseInterfaces,
     const lduInterface& fineInterface,
     const labelField& localRestrictAddressing,
-    const labelField& neighbourRestrictAddressing
+    const labelField& neighbourRestrictAddressing,
+    const label coarseComm
 )
 :
     AMGInterface(lduMesh),
     fineMixingPlaneInterface_
     (
         refCast<const mixingPlaneLduInterface>(fineInterface)
-    )
+    ),
+    comm_(coarseComm),
+    tag_(refCast<const mixingPlaneLduInterface>(fineInterface).tag())
 {}
 
 

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/mixingPlaneAMGInterface/mixingPlaneAMGInterface.H

@@ -65,6 +65,15 @@ class mixingPlaneAMGInterface
         const mixingPlaneLduInterface& fineMixingPlaneInterface_;
 
 
+        // Parallel communication
+
+            //- Communicator to use for parallel communication
+            const label comm_;
+
+            //- Message tag used for sending
+            const int tag_;
+
+
     // Private Member Functions
 
         //- Disallow default bitwise copy construct
@@ -96,13 +105,13 @@ public:
             const lduInterfacePtrsList& coarseInterfaces,
             const lduInterface& fineInterface,
             const labelField& localRestrictAddressing,
-            const labelField& neighbourRestrictAddressing
+            const labelField& neighbourRestrictAddressing,
+            const label coarseComm
         );
 
 
-    // Destructor
-
-        virtual ~mixingPlaneAMGInterface();
+    // -Destructor
+    virtual ~mixingPlaneAMGInterface();
 
 
     // Member Functions
@@ -116,6 +125,21 @@ public:
             }
 
 
+        // Communications support
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const
+            {
+                return comm_;
+            }
+
+            //- Return message tag used for sending
+            virtual int tag() const
+            {
+                return tag_;
+            }
+
+
         // Agglomeration
 
             //- Agglomerating the given fine-level coefficients and return

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/processorAMGInterface/processorAMGInterface.C

@@ -48,11 +48,14 @@ Foam::processorAMGInterface::processorAMGInterface
     const lduInterfacePtrsList& coarseInterfaces,
     const lduInterface& fineInterface,
     const labelField& localRestrictAddressing,
-    const labelField& neighbourRestrictAddressing
+    const labelField& neighbourRestrictAddressing,
+    const label coarseComm
 )
 :
     AMGInterface(lduMesh),
-    fineProcInterface_(refCast<const processorLduInterface>(fineInterface))
+    fineProcInterface_(refCast<const processorLduInterface>(fineInterface)),
+    comm_(coarseComm),
+    tag_(refCast<const processorLduInterface>(fineInterface).tag())
 {
     // Make a lookup table of entries for owner/neighbour
     HashTable<SLList<label>, label, Hash<label> > neighboursTable

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/processorAMGInterface/processorAMGInterface.H

@@ -59,6 +59,12 @@ class processorAMGInterface
         //  agglomerated
         const processorLduInterface& fineProcInterface_;
 
+        //- Communicator to use for parallel communication
+        const label comm_;
+
+        //- Message tag used for sending
+        const int tag_;
+
 
     // Private Member Functions
 
@@ -84,13 +90,13 @@ public:
             const lduInterfacePtrsList& coarseInterfaces,
             const lduInterface& fineInterface,
             const labelField& localRestrictAddressing,
-            const labelField& neighbourRestrictAddressing
+            const labelField& neighbourRestrictAddressing,
+            const label coarseComm
         );
 
 
-    // Destructor
-
-        virtual ~processorAMGInterface();
+    //- Destructor
+    virtual ~processorAMGInterface();
 
 
     // Member Functions
@@ -104,6 +110,21 @@ public:
             }
 
 
+        // Communications support
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const
+            {
+                return comm_;
+            }
+
+            //- Return message tag used for sending
+            virtual int tag() const
+            {
+                return tag_;
+            }
+
+
         // Interface transfer functions
 
             //- Initialise interface data transfer

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/regionCoupleAMGInterface/regionCoupleAMGInterface.C

@@ -48,7 +48,8 @@ Foam::regionCoupleAMGInterface::regionCoupleAMGInterface
     const lduInterfacePtrsList& coarseInterfaces,
     const lduInterface& fineInterface,
     const labelField& localRestrictAddressing,
-    const labelField& neighbourRestrictAddressing
+    const labelField& neighbourRestrictAddressing,
+    const label coarseComm
 )
 :
     ggiAMGInterface
@@ -57,7 +58,8 @@ Foam::regionCoupleAMGInterface::regionCoupleAMGInterface
         coarseInterfaces,
         fineInterface,
         localRestrictAddressing,
-        neighbourRestrictAddressing
+        neighbourRestrictAddressing,
+        coarseComm
     ),
     coupled_(fineInterface.coupled())
 {}

src/foam/matrices/lduMatrix/solvers/AMG/interfaces/regionCoupleAMGInterface/regionCoupleAMGInterface.H

@@ -46,7 +46,7 @@ namespace Foam
 {
 
 /*---------------------------------------------------------------------------*\
-                     Class regionCoupleAMGInterface Declaration
+                   Class regionCoupleAMGInterface Declaration
 \*---------------------------------------------------------------------------*/
 
 class regionCoupleAMGInterface
@@ -76,13 +76,13 @@ public:
             const lduInterfacePtrsList& coarseInterfaces,
             const lduInterface& fineInterface,
             const labelField& localRestrictAddressing,
-            const labelField& neighbourRestrictAddressing
+            const labelField& neighbourRestrictAddressing,
+            const label coarseComm
         );
 
 
-    // Destructor
-
-        virtual ~regionCoupleAMGInterface();
+    //- Destructor
+    virtual ~regionCoupleAMGInterface();
 
 
     // Member Functions

src/foam/meshes/lduMesh/lduMesh.H

@@ -59,10 +59,9 @@ public:
 
     // Constructors
 
-    // Destructor
-
-        virtual ~lduMesh()
-        {}
+    //- Destructor
+    virtual ~lduMesh()
+    {}
 
 
     // Member Functions
@@ -78,6 +77,10 @@ public:
             //- Return a list of pointers for each patch
             //  with only those pointing to interfaces being set
             virtual lduInterfacePtrsList interfaces() const = 0;
+
+            //- Return communicator used for parallel communication
+            virtual int comm() const = 0;
+
 };
 
 

src/foam/meshes/lduMesh/lduPrimitiveMesh.H

@@ -69,6 +69,9 @@ class lduPrimitiveMesh
         //  Note this does must be held as a copy.  HJ, 20/Feb/2009
         lduSchedule patchSchedule_;
 
+        //- Communicator to use for parallel communication
+        const label comm_;
+
 
     // Private Member Functions
 
@@ -89,12 +92,14 @@ public:
         (
             const label nCells,
             const unallocLabelList& l,
-            const unallocLabelList& u
+            const unallocLabelList& u,
+            const label comm
         )
         :
             lduAddressing(nCells),
             lowerAddr_(l),
-            upperAddr_(u)
+            upperAddr_(u),
+            comm_(comm)
         {}
 
 
@@ -104,12 +109,14 @@ public:
             const label nCells,
             labelList& l,
             labelList& u,
+            const label comm,
             bool reUse
         )
         :
             lduAddressing(nCells),
             lowerAddr_(l, reUse),
-            upperAddr_(u, reUse)
+            upperAddr_(u, reUse),
+            comm_(comm)
         {}
 
 
@@ -194,6 +201,12 @@ public:
             {
                 return patchSchedule_;
             }
+
+            //- Return communicator used for parallel communication
+            virtual label comm() const
+            {
+                return comm_;
+            }
 };
 
 

src/foam/meshes/pointMesh/pointPatches/constraint/ggi/ggiPointPatch.H

@@ -111,9 +111,8 @@ public:
         );
 
 
-    // Destructor
-
-        virtual ~ggiPointPatch();
+    //- Destructor
+    virtual ~ggiPointPatch();
 
 
     // Member Functions
@@ -122,6 +121,21 @@ public:
 
             //- Return true because this patch is coupled
             virtual bool coupled() const;
+
+
+        // Communications support
+
+            //- Return communicator used for communication
+            virtual label comm() const
+            {
+                return ggiPolyPatch_.comm();
+            }
+
+            //- Return message tag to use for communication
+            virtual int tag() const
+            {
+                return ggiPolyPatch_.tag();
+            }
 };
 
 

src/foam/meshes/pointMesh/pointPatches/constraint/processor/processorPointPatch.H

@@ -124,9 +124,8 @@ public:
         );
 
 
-    // Destructor
-
-        virtual ~processorPointPatch();
+    //- Destructor
+    virtual ~processorPointPatch();
 
 
     // Member functions
@@ -175,6 +174,21 @@ public:
         }
 
 
+        // Communications support
+
+            //- Return communicator used for communication
+            virtual label comm() const
+            {
+                return procPolyPatch_.comm();
+            }
+
+            //- Return message tag to use for communication
+            virtual int tag() const
+            {
+                return procPolyPatch_.tag();
+            }
+
+
         // Access functions for demand driven data
 
             //- Return mesh points

src/foam/meshes/polyMesh/mapPolyMesh/mapDistribute/mapDistribute.C

@@ -33,7 +33,8 @@ License
 Foam::List<Foam::labelPair> Foam::mapDistribute::schedule
 (
     const labelListList& subMap,
-    const labelListList& constructMap
+    const labelListList& constructMap,
+    const int tag
 )
 {
     // Communications: send and receive processor
@@ -74,7 +75,7 @@ Foam::List<Foam::labelPair> Foam::mapDistribute::schedule
             slave++
         )
         {
-            IPstream fromSlave(Pstream::scheduled, slave);
+            IPstream fromSlave(Pstream::scheduled, slave, 0, tag);
             List<labelPair> nbrData(fromSlave);
 
             forAll (nbrData, i)
@@ -95,18 +96,24 @@ Foam::List<Foam::labelPair> Foam::mapDistribute::schedule
             slave++
         )
         {
-            OPstream toSlave(Pstream::scheduled, slave);
+            OPstream toSlave(Pstream::scheduled, slave, 0, tag);
             toSlave << allComms;
         }
     }
     else
     {
         {
-            OPstream toMaster(Pstream::scheduled, Pstream::masterNo());
+            OPstream toMaster(Pstream::scheduled, Pstream::masterNo(), 0, tag);
             toMaster << allComms;
         }
         {
-            IPstream fromMaster(Pstream::scheduled, Pstream::masterNo());
+            IPstream fromMaster
+            (
+                Pstream::scheduled,
+                Pstream::masterNo(),
+                0,
+                tag
+            );
             fromMaster >> allComms;
         }
     }
@@ -156,7 +163,7 @@ const Foam::List<Foam::labelPair>& Foam::mapDistribute::schedule() const
         (
             new List<labelPair>
             (
-                schedule(subMap_, constructMap_)
+                schedule(subMap_, constructMap_, Pstream::msgType())
             )
         );
     }
@@ -311,7 +318,7 @@ Foam::mapDistribute::mapDistribute(const mapDistribute& map)
 
 // * * * * * * * * * * * * * Private Member Functions  * * * * * * * * * * * //
 
-void Foam::mapDistribute::compact(const boolList& elemIsUsed)
+void Foam::mapDistribute::compact(const boolList& elemIsUsed, const int tag)
 {
     // 1. send back to sender. Have him delete the corresponding element
     //    from the submap and do the same to the constructMap locally
@@ -319,7 +326,12 @@ void Foam::mapDistribute::compact(const boolList& elemIsUsed)
 
     // Send elemIsUsed field to neighbour. Use nonblocking code from
     // mapDistribute but in reverse order.
+    if (Pstream::parRun())
     {
+        label startOfRequests = Pstream::nRequests();
+
+        // Set up receives from neighbours
+
         List<boolList> sendFields(Pstream::nProcs());
 
         for (label domain = 0; domain < Pstream::nProcs(); domain++)
@@ -340,7 +352,8 @@ void Foam::mapDistribute::compact(const boolList& elemIsUsed)
                     Pstream::nonBlocking,
                     domain,
                     reinterpret_cast<const char*>(subField.begin()),
-                    subField.size()*sizeof(bool)
+                    subField.size()*sizeof(bool),
+                    tag
                 );
             }
         }
@@ -361,7 +374,8 @@ void Foam::mapDistribute::compact(const boolList& elemIsUsed)
                     Pstream::nonBlocking,
                     domain,
                     reinterpret_cast<char*>(recvFields[domain].begin()),
-                    recvFields[domain].size()*sizeof(bool)
+                    recvFields[domain].size()*sizeof(bool),
+                    tag
                 );
             }
         }
@@ -382,8 +396,7 @@ void Foam::mapDistribute::compact(const boolList& elemIsUsed)
 
         // Wait for all to finish
 
-        OPstream::waitRequests();
-        IPstream::waitRequests();
+        Pstream::waitRequests(startOfRequests);
 
 
         // Compact out all submap entries that are referring to unused elements

src/foam/meshes/polyMesh/mapPolyMesh/mapDistribute/mapDistribute.H

@@ -166,7 +166,8 @@ public:
             static List<labelPair> schedule
             (
                 const labelListList& subMap,
-                const labelListList& constructMap
+                const labelListList& constructMap,
+                const int tag
             );
 
             //- Return a schedule. Demand driven. See above.
@@ -178,8 +179,11 @@ public:
             //- Compact maps. Gets per field a bool whether it is used locally
             //  and works out itself what this side and sender side can remove
             //  from maps.
-            void compact(const boolList& elemIsUsed);
-
+            void compact
+            (
+                const boolList& elemIsUsed,
+                const int tag = Pstream::msgType()
+            );
 
             //- Distribute data. Note:schedule only used for Pstream::scheduled
             //  for now, all others just use send-to-all, receive-from-all.
@@ -191,7 +195,8 @@ public:
                 const label constructSize,
                 const labelListList& subMap,
                 const labelListList& constructMap,
-                List<T>&
+                List<T>&,
+                const int tag = Pstream::msgType()
             );
 
             //- Distribute data. If multiple processors writing to same
@@ -206,12 +211,17 @@ public:
                 const labelListList& constructMap,
                 List<T>&,
                 const CombineOp& cop,
-                const T& nullValue
+                const T& nullValue,
+                const int tag = Pstream::msgType()
             );
 
             //- Distribute data using default commsType.
             template<class T>
-            void distribute(List<T>& fld) const
+            void distribute
+            (
+                 List<T>& fld,
+                const int tag = Pstream::msgType()
+            ) const
             {
                 if
                 (

src/foam/meshes/polyMesh/mapPolyMesh/mapDistribute/mapDistributeTemplates.C

@@ -36,9 +36,35 @@ void Foam::mapDistribute::distribute
     const label constructSize,
     const labelListList& subMap,
     const labelListList& constructMap,
-    List<T>& field
+    List<T>& field,
+    const int tag
 )
 {
+    if (!Pstream::parRun())
+    {
+        // Do only me to me.
+
+        const labelList& mySubMap = subMap[Pstream::myProcNo()];
+
+        List<T> subField(mySubMap.size());
+        forAll(mySubMap, i)
+        {
+            subField[i] = field[mySubMap[i]];
+        }
+
+        // Receive sub field from myself (subField)
+        const labelList& map = constructMap[Pstream::myProcNo()];
+
+        field.setSize(constructSize);
+
+        forAll(map, i)
+        {
+            field[map[i]] = subField[i];
+        }
+
+        return;
+    }
+
     if (commsType == Pstream::blocking)
     {
         // Since buffered sending can reuse the field to collect the
@@ -51,7 +77,7 @@ void Foam::mapDistribute::distribute
 
             if (domain != Pstream::myProcNo() && map.size())
             {
-                OPstream toNbr(Pstream::blocking, domain);
+                OPstream toNbr(Pstream::blocking, domain, 0, tag);
                 toNbr << UIndirectList<T>(field, map);
             }
         }
@@ -82,7 +108,7 @@ void Foam::mapDistribute::distribute
 
             if (domain != Pstream::myProcNo() && map.size())
             {
-                IPstream fromNbr(Pstream::blocking, domain);
+                IPstream fromNbr(Pstream::blocking, domain, 0, tag);
                 List<T> subField(fromNbr);
 
                 checkReceivedSize(domain, map.size(), subField.size());
@@ -116,28 +142,52 @@ void Foam::mapDistribute::distribute
         forAll(schedule, i)
         {
             const labelPair& twoProcs = schedule[i];
+            // twoProcs is a swap pair of processors. The first one is the
+            // one that needs to send first and then receive.
+
             label sendProc = twoProcs[0];
             label recvProc = twoProcs[1];
 
             if (Pstream::myProcNo() == sendProc)
             {
-                // I am sender. Send to recvProc.
-                OPstream toNbr(Pstream::scheduled, recvProc);
-                toNbr << UIndirectList<T>(field, subMap[recvProc]);
+                // I am send first, receive next
+                {
+                    OPstream toNbr(Pstream::scheduled, recvProc, 0, tag);
+                    toNbr << UIndirectList<T>(field, subMap[recvProc]);
+                }
+                {
+                    IPstream fromNbr(Pstream::scheduled, recvProc, 0, tag);
+                    List<T> subField(fromNbr);
+
+                    const labelList& map = constructMap[recvProc];
+
+                    checkReceivedSize(recvProc, map.size(), subField.size());
+
+                    forAll(map, i)
+                    {
+                        newField[map[i]] = subField[i];
+                    }
+                }
             }
             else
             {
-                // I am receiver. Receive from sendProc.
-                IPstream fromNbr(Pstream::scheduled, sendProc);
-                List<T> subField(fromNbr);
-
-                const labelList& map = constructMap[sendProc];
-
-                checkReceivedSize(recvProc, map.size(), subField.size());
-
-                forAll(map, i)
+                // I am receive first, send next
                 {
-                    newField[map[i]] = subField[i];
+                    IPstream fromNbr(Pstream::scheduled, sendProc, 0, tag);
+                    List<T> subField(fromNbr);
+
+                    const labelList& map = constructMap[sendProc];
+
+                    checkReceivedSize(sendProc, map.size(), subField.size());
+
+                    forAll(map, i)
+                    {
+                        newField[map[i]] = subField[i];
+                    }
+                }
+                {
+                    OPstream toNbr(Pstream::scheduled, sendProc, 0, tag);
+                    toNbr << UIndirectList<T>(field, subMap[sendProc]);
                 }
             }
         }
@@ -146,129 +196,173 @@ void Foam::mapDistribute::distribute
     }
     else if (commsType == Pstream::nonBlocking)
     {
+        label nOutstanding = Pstream::nRequests();
+
         if (!contiguous<T>())
         {
-            FatalErrorIn
-            (
-                "template<class T>\n"
-                "void mapDistribute::distribute\n"
-                "(\n"
-                "    const Pstream::commsTypes commsType,\n"
-                "    const List<labelPair>& schedule,\n"
-                "    const label constructSize,\n"
-                "    const labelListList& subMap,\n"
-                "    const labelListList& constructMap,\n"
-                "    List<T>& field\n"
-                ")\n"
-            )   << "Non-blocking only supported for contiguous data."
-                << exit(FatalError);
-        }
-
-        // Set up sends to neighbours
-
-        List<List<T> > sendFields(Pstream::nProcs());
-
-        for (label domain = 0; domain < Pstream::nProcs(); domain++)
-        {
-            const labelList& map = subMap[domain];
-
-            if (domain != Pstream::myProcNo() && map.size())
+            // Stream data into buffer
+            for (label domain = 0; domain < Pstream::nProcs(); domain++)
             {
-                List<T>& subField = sendFields[domain];
+                const labelList& map = subMap[domain];
 
+                if (domain != Pstream::myProcNo() && map.size())
+                {
+                    // Put data into send buffer
+                    OPstream toDomain(Pstream::nonBlocking, domain, 0, tag);
+                    toDomain << UIndirectList<T>(field, map);
+                }
+            }
+
+            // Start receiving. Do not block.
+
+            {
+                // Set up 'send' to myself
+                const labelList& mySubMap = subMap[Pstream::myProcNo()];
+                List<T> mySubField(mySubMap.size());
+                forAll(mySubMap, i)
+                {
+                    mySubField[i] = field[mySubMap[i]];
+                }
+                // Combine bits. Note that can reuse field storage
+                field.setSize(constructSize);
+                // Receive sub field from myself
+                {
+                    const labelList& map = constructMap[Pstream::myProcNo()];
+
+                    forAll(map, i)
+                    {
+                        field[map[i]] = mySubField[i];
+                    }
+                }
+            }
+
+            // Block ourselves, waiting only for the current comms
+            Pstream::waitRequests(nOutstanding);
+
+            // Consume
+            for (label domain = 0; domain < Pstream::nProcs(); domain++)
+            {
+                const labelList& map = constructMap[domain];
+
+                if (domain != Pstream::myProcNo() && map.size())
+                {
+                    IPstream str(Pstream::nonBlocking, domain, 0, tag);
+                    List<T> recvField(str);
+
+                    checkReceivedSize(domain, map.size(), recvField.size());
+
+                    forAll(map, i)
+                    {
+                        field[map[i]] = recvField[i];
+                    }
+                }
+            }
+        }
+        else
+        {
+            // Set up sends to neighbours
+
+            List<List<T > > sendFields(Pstream::nProcs());
+
+            for (label domain = 0; domain < Pstream::nProcs(); domain++)
+            {
+                const labelList& map = subMap[domain];
+
+                if (domain != Pstream::myProcNo() && map.size())
+                {
+                    List<T>& subField = sendFields[domain];
+                    subField.setSize(map.size());
+                    forAll(map, i)
+                    {
+                        subField[i] = field[map[i]];
+                    }
+
+                    OPstream::write
+                    (
+                        Pstream::nonBlocking,
+                        domain,
+                        reinterpret_cast<const char*>(subField.begin()),
+                        subField.byteSize(),
+                        tag
+                    );
+                }
+            }
+
+            // Set up receives from neighbours
+
+            List<List<T > > recvFields(Pstream::nProcs());
+
+            for (label domain = 0; domain < Pstream::nProcs(); domain++)
+            {
+                const labelList& map = constructMap[domain];
+
+                if (domain != Pstream::myProcNo() && map.size())
+                {
+                    recvFields[domain].setSize(map.size());
+                    IPstream::read
+                    (
+                        Pstream::nonBlocking,
+                        domain,
+                        reinterpret_cast<char*>(recvFields[domain].begin()),
+                        recvFields[domain].byteSize(),
+                        tag
+                    );
+                }
+            }
+
+
+            // Set up 'send' to myself
+
+            {
+                const labelList& map = subMap[Pstream::myProcNo()];
+
+                List<T>& subField = sendFields[Pstream::myProcNo()];
                 subField.setSize(map.size());
 
                 forAll(map, i)
                 {
                     subField[i] = field[map[i]];
                 }
-
-                OPstream::write
-                (
-                    Pstream::nonBlocking,
-                    domain,
-                    reinterpret_cast<const char*>(subField.begin()),
-                    subField.size()*sizeof(T)
-                );
             }
-        }
 
-        // Set up receives from neighbours
 
-        List<List<T> > recvFields(Pstream::nProcs());
+            // Combine bits. Note that can reuse field storage
 
-        for (label domain = 0; domain < Pstream::nProcs(); domain++)
-        {
-            const labelList& map = constructMap[domain];
+            field.setSize(constructSize);
 
-            if (domain != Pstream::myProcNo() && map.size())
+            // Receive sub field from myself (sendFields[Pstream::myProcNo()])
             {
-                recvFields[domain].setSize(map.size());
-                IPstream::read
-                (
-                    Pstream::nonBlocking,
-                    domain,
-                    reinterpret_cast<char*>(recvFields[domain].begin()),
-                    recvFields[domain].size()*sizeof(T)
-                );
-            }
-        }
-
-
-        // Set up 'send' to myself
-
-        {
-            const labelList& map = subMap[Pstream::myProcNo()];
-
-            List<T>& subField = sendFields[Pstream::myProcNo()];
-            subField.setSize(map.size());
-
-            forAll(map, i)
-            {
-                subField[i] = field[map[i]];
-            }
-        }
-
-
-        // Combine bits. Note that can reuse field storage
-
-        field.setSize(constructSize);
-
-        // Receive sub field from myself (sendFields[Pstream::myProcNo()])
-        {
-            const labelList& map = constructMap[Pstream::myProcNo()];
-            const List<T>& subField = sendFields[Pstream::myProcNo()];
-
-            forAll(map, i)
-            {
-                field[map[i]] = subField[i];
-            }
-        }
-
-
-        // Wait for all to finish
-
-        OPstream::waitRequests();
-        IPstream::waitRequests();
-
-        // Collect neighbour fields
-
-        for (label domain = 0; domain < Pstream::nProcs(); domain++)
-        {
-            const labelList& map = constructMap[domain];
-
-            if (domain != Pstream::myProcNo() && map.size())
-            {
-                checkReceivedSize
-                (
-                    domain,
-                    map.size(),
-                    recvFields[domain].size()
-                );
+                const labelList& map = constructMap[Pstream::myProcNo()];
+                const List<T>& subField = sendFields[Pstream::myProcNo()];
 
                 forAll(map, i)
                 {
-                    field[map[i]] = recvFields[domain][i];
+                    field[map[i]] = subField[i];
+                }
+            }
+
+
+            // Wait for all to finish
+
+            Pstream::waitRequests(nOutstanding);
+
+
+            // Collect neighbour fields
+
+            for (label domain = 0; domain < Pstream::nProcs(); domain++)
+            {
+                const labelList& map = constructMap[domain];
+
+                if (domain != Pstream::myProcNo() && map.size())
+                {
+                    const List<T>& subField = recvFields[domain];
+
+                    checkReceivedSize(domain, map.size(), subField.size());
+
+                    forAll(map, i)
+                    {
+                        field[map[i]] = subField[i];
+                    }
                 }
             }
         }
@@ -294,9 +388,35 @@ void Foam::mapDistribute::distribute
     const labelListList& constructMap,
     List<T>& field,
     const CombineOp& cop,
-    const T& nullValue
+    const T& nullValue,
+    const int tag
 )
 {
+    if (!Pstream::parRun())
+    {
+        // Do only me to me.
+
+        const labelList& mySubMap = subMap[Pstream::myProcNo()];
+
+        List<T> subField(mySubMap.size());
+        forAll(mySubMap, i)
+        {
+            subField[i] = field[mySubMap[i]];
+        }
+
+        // Receive sub field from myself (subField)
+        const labelList& map = constructMap[Pstream::myProcNo()];
+
+        field.setSize(constructSize);
+        field = nullValue;
+
+        forAll(map, i)
+        {
+            cop(field[map[i]], subField[i]);
+        }
+        return;
+    }
+
     if (commsType == Pstream::blocking)
     {
         // Since buffered sending can reuse the field to collect the
@@ -309,7 +429,7 @@ void Foam::mapDistribute::distribute
 
             if (domain != Pstream::myProcNo() && map.size())
             {
-                OPstream toNbr(Pstream::blocking, domain);
+                OPstream toNbr(Pstream::blocking, domain, 0, tag);
                 toNbr << UIndirectList<T>(field, map);
             }
         }
@@ -341,28 +461,10 @@ void Foam::mapDistribute::distribute
 
             if (domain != Pstream::myProcNo() && map.size())
             {
-                IPstream fromNbr(Pstream::blocking, domain);
+                IPstream fromNbr(Pstream::blocking, domain, 0, tag);
                 List<T> subField(fromNbr);
 
-                if (subField.size() != map.size())
-                {
-                    FatalErrorIn
-                    (
-                        "template<class T>\n"
-                        "void mapDistribute::distribute\n"
-                        "(\n"
-                        "    const Pstream::commsTypes commsType,\n"
-                        "    const List<labelPair>& schedule,\n"
-                        "    const label constructSize,\n"
-                        "    const labelListList& subMap,\n"
-                        "    const labelListList& constructMap,\n"
-                        "    List<T>& field\n"
-                        ")\n"
-                    )   << "Expected from processor " << domain
-                        << " " << map.size() << " but received "
-                        << subField.size() << " elements."
-                        << abort(FatalError);
-                }
+                checkReceivedSize(domain, map.size(), subField.size());
 
                 forAll(map, i)
                 {
@@ -393,195 +495,227 @@ void Foam::mapDistribute::distribute
         forAll(schedule, i)
         {
             const labelPair& twoProcs = schedule[i];
+            // twoProcs is a swap pair of processors. The first one is the
+            // one that needs to send first and then receive.
+
             label sendProc = twoProcs[0];
             label recvProc = twoProcs[1];
 
             if (Pstream::myProcNo() == sendProc)
             {
-                // I am sender. Send to recvProc.
-                OPstream toNbr(Pstream::scheduled, recvProc);
-                toNbr << UIndirectList<T>(field, subMap[recvProc]);
+                // I am send first, receive next
+                {
+                    OPstream toNbr(Pstream::scheduled, recvProc, 0, tag);
+                    toNbr << UIndirectList<T>(field, subMap[recvProc]);
+                }
+                {
+                    IPstream fromNbr(Pstream::scheduled, recvProc, 0, tag);
+                    List<T> subField(fromNbr);
+                    const labelList& map = constructMap[recvProc];
+
+                    checkReceivedSize(recvProc, map.size(), subField.size());
+
+                    forAll(map, i)
+                    {
+                        cop(newField[map[i]], subField[i]);
+                    }
+                }
             }
             else
             {
-                // I am receiver. Receive from sendProc.
-                IPstream fromNbr(Pstream::scheduled, sendProc);
-                List<T> subField(fromNbr);
-
-                const labelList& map = constructMap[sendProc];
-
-                if (subField.size() != map.size())
+                // I am receive first, send next
                 {
-                    FatalErrorIn
-                    (
-                        "template<class T>\n"
-                        "void mapDistribute::distribute\n"
-                        "(\n"
-                        "    const Pstream::commsTypes commsType,\n"
-                        "    const List<labelPair>& schedule,\n"
-                        "    const label constructSize,\n"
-                        "    const labelListList& subMap,\n"
-                        "    const labelListList& constructMap,\n"
-                        "    List<T>& field\n"
-                        ")\n"
-                    )   << "Expected from processor " << sendProc
-                        << " " << map.size() << " but received "
-                        << subField.size() << " elements."
-                        << abort(FatalError);
+                    IPstream fromNbr(Pstream::scheduled, sendProc, 0, tag);
+                    List<T> subField(fromNbr);
+                    const labelList& map = constructMap[sendProc];
+
+                    checkReceivedSize(sendProc, map.size(), subField.size());
+
+                    forAll(map, i)
+                    {
+                        cop(newField[map[i]], subField[i]);
+                    }
                 }
-
-                forAll(map, i)
                 {
-                    cop(newField[map[i]], subField[i]);
+                    OPstream toNbr(Pstream::scheduled, sendProc, 0, tag);
+                    toNbr << UIndirectList<T>(field, subMap[sendProc]);
                 }
             }
         }
+
         field.transfer(newField);
     }
     else if (commsType == Pstream::nonBlocking)
     {
+        label nOutstanding = Pstream::nRequests();
+
         if (!contiguous<T>())
         {
-            FatalErrorIn
-            (
-                "template<class T>\n"
-                "void mapDistribute::distribute\n"
-                "(\n"
-                "    const Pstream::commsTypes commsType,\n"
-                "    const List<labelPair>& schedule,\n"
-                "    const label constructSize,\n"
-                "    const labelListList& subMap,\n"
-                "    const labelListList& constructMap,\n"
-                "    List<T>& field\n"
-                ")\n"
-            )   << "Non-blocking only supported for contiguous data."
-                << exit(FatalError);
-        }
-
-        // Set up sends to neighbours
-
-        List<List<T> > sendFields(Pstream::nProcs());
-
-        for (label domain = 0; domain < Pstream::nProcs(); domain++)
-        {
-            const labelList& map = subMap[domain];
-
-            if (domain != Pstream::myProcNo() && map.size())
+            // Stream data into buffer
+            for (label domain = 0; domain < Pstream::nProcs(); domain++)
             {
-                List<T>& subField = sendFields[domain];
+                const labelList& map = subMap[domain];
+
+                if (domain != Pstream::myProcNo() && map.size())
+                {
+                    // Put data into send buffer
+                    OPstream toDomain(Pstream::nonBlocking, domain, 0, tag);
+                    toDomain << UIndirectList<T>(field, map);
+                }
+            }
+
+            // Start receiving. Do not block.
+
+            {
+                // Set up 'send' to myself
+                List<T> mySubField(field, subMap[Pstream::myProcNo()]);
+                // Combine bits. Note that can reuse field storage
+                field.setSize(constructSize);
+                field = nullValue;
+                // Receive sub field from myself
+                {
+                    const labelList& map = constructMap[Pstream::myProcNo()];
+
+                    forAll(map, i)
+                    {
+                        cop(field[map[i]], mySubField[i]);
+                    }
+                }
+            }
+
+            // Block ourselves, waiting only for the current comms
+            Pstream::waitRequests(nOutstanding);
+
+            // Consume
+            for (label domain = 0; domain < Pstream::nProcs(); domain++)
+            {
+                const labelList& map = constructMap[domain];
+
+                if (domain != Pstream::myProcNo() && map.size())
+                {
+                    IPstream str(Pstream::nonBlocking, domain, 0, tag);
+                    List<T> recvField(str);
+
+                    checkReceivedSize(domain, map.size(), recvField.size());
+
+                    forAll(map, i)
+                    {
+                        cop(field[map[i]], recvField[i]);
+                    }
+                }
+            }
+        }
+        else
+        {
+            // Set up sends to neighbours
+
+            List<List<T> > sendFields(Pstream::nProcs());
+
+            for (label domain = 0; domain < Pstream::nProcs(); domain++)
+            {
+                const labelList& map = subMap[domain];
+
+                if (domain != Pstream::myProcNo() && map.size())
+                {
+                    List<T>& subField = sendFields[domain];
+                    subField.setSize(map.size());
+                    forAll(map, i)
+                    {
+                        subField[i] = field[map[i]];
+                    }
+
+                    OPstream::write
+                    (
+                        Pstream::nonBlocking,
+                        domain,
+                        reinterpret_cast<const char*>(subField.begin()),
+                        subField.size()*sizeof(T),
+                        tag
+                    );
+                }
+            }
+
+            // Set up receives from neighbours
+
+            List<List<T > > recvFields(Pstream::nProcs());
+
+            for (label domain = 0; domain < Pstream::nProcs(); domain++)
+            {
+                const labelList& map = constructMap[domain];
+
+                if (domain != Pstream::myProcNo() && map.size())
+                {
+                    recvFields[domain].setSize(map.size());
+                    IPstream::read
+                    (
+                        Pstream::nonBlocking,
+                        domain,
+                        reinterpret_cast<char*>(recvFields[domain].begin()),
+                        recvFields[domain].size()*sizeof(T),
+                        tag
+                    );
+                }
+            }
+
+            // Set up 'send' to myself
+
+            {
+                const labelList& map = subMap[Pstream::myProcNo()];
+
+                List<T>& subField = sendFields[Pstream::myProcNo()];
                 subField.setSize(map.size());
                 forAll(map, i)
                 {
                     subField[i] = field[map[i]];
                 }
-
-                OPstream::write
-                (
-                    Pstream::nonBlocking,
-                    domain,
-                    reinterpret_cast<const char*>(subField.begin()),
-                    subField.size()*sizeof(T)
-                );
             }
-        }
 
-        // Set up receives from neighbours
 
-        List<List<T> > recvFields(Pstream::nProcs());
+            // Combine bits. Note that can reuse field storage
 
-        for (label domain = 0; domain < Pstream::nProcs(); domain++)
-        {
-            const labelList& map = constructMap[domain];
+            field.setSize(constructSize);
+            field = nullValue;
 
-            if (domain != Pstream::myProcNo() && map.size())
+            // Receive sub field from myself (subField)
             {
-                recvFields[domain].setSize(map.size());
-                IPstream::read
-                (
-                    Pstream::nonBlocking,
-                    domain,
-                    reinterpret_cast<char*>(recvFields[domain].begin()),
-                    recvFields[domain].size()*sizeof(T)
-                );
-            }
-        }
-
-        // Set up 'send' to myself
-
-        {
-            const labelList& map = subMap[Pstream::myProcNo()];
-
-            List<T>& subField = sendFields[Pstream::myProcNo()];
-            subField.setSize(map.size());
-            forAll(map, i)
-            {
-                subField[i] = field[map[i]];
-            }
-        }
-
-
-        // Combine bits. Note that can reuse field storage
-
-        field.setSize(constructSize);
-        field = nullValue;
-
-        // Receive sub field from myself (subField)
-        {
-            const labelList& map = constructMap[Pstream::myProcNo()];
-            const List<T>& subField = sendFields[Pstream::myProcNo()];
-
-            forAll(map, i)
-            {
-                cop(field[map[i]], subField[i]);
-            }
-        }
-
-
-        // Wait for all to finish
-
-        OPstream::waitRequests();
-        IPstream::waitRequests();
-
-        // Collect neighbour fields
-
-        for (label domain = 0; domain < Pstream::nProcs(); domain++)
-        {
-            const labelList& map = constructMap[domain];
-
-            if (domain != Pstream::myProcNo() && map.size())
-            {
-                if (recvFields[domain].size() != map.size())
-                {
-                    FatalErrorIn
-                    (
-                        "template<class T>\n"
-                        "void mapDistribute::distribute\n"
-                        "(\n"
-                        "    const Pstream::commsTypes commsType,\n"
-                        "    const List<labelPair>& schedule,\n"
-                        "    const label constructSize,\n"
-                        "    const labelListList& subMap,\n"
-                        "    const labelListList& constructMap,\n"
-                        "    List<T>& field\n"
-                        ")\n"
-                    )   << "Expected from processor " << domain
-                        << " " << map.size() << " but received "
-                        << recvFields[domain].size() << " elements."
-                        << abort(FatalError);
-                }
+                const labelList& map = constructMap[Pstream::myProcNo()];
+                const List<T>& subField = sendFields[Pstream::myProcNo()];
 
                 forAll(map, i)
                 {
-                    cop(field[map[i]], recvFields[domain][i]);
+                    cop(field[map[i]], subField[i]);
+                }
+            }
+
+
+            // Wait for all to finish
+            Pstream::waitRequests(nOutstanding);
+
+
+            // Collect neighbour fields
+
+            for (label domain = 0; domain < Pstream::nProcs(); domain++)
+            {
+                const labelList& map = constructMap[domain];
+
+                if (domain != Pstream::myProcNo() && map.size())
+                {
+                    const List<T>& subField = recvFields[domain];
+
+                    checkReceivedSize(domain, map.size(), subField.size());
+
+                    forAll(map, i)
+                    {
+                        cop(field[map[i]], subField[i]);
+                    }
                 }
             }
         }
     }
     else
     {
-        // This needs to be cleaned up: temporary solution.  HJ, 15/Jun/2014
         FatalErrorIn("mapDistribute::distribute(..)")
-            << "Unknown communication schedule " << label(commsType)
+            << "Unknown communication schedule " << commsType
             << abort(FatalError);
     }
 }

src/foam/meshes/polyMesh/polyMesh.C

@@ -244,6 +244,7 @@ Foam::polyMesh::polyMesh(const IOobject& io)
     bounds_(allPoints_),
     geometricD_(Vector<label>::zero),
     solutionD_(Vector<label>::zero),
+    comm_(Pstream::worldComm),
     pointZones_
     (
         IOobject
@@ -441,6 +442,7 @@ Foam::polyMesh::polyMesh
     bounds_(allPoints_, syncPar),
     geometricD_(Vector<label>::zero),
     solutionD_(Vector<label>::zero),
+    comm_(Pstream::worldComm),
     pointZones_
     (
         IOobject
@@ -601,6 +603,7 @@ Foam::polyMesh::polyMesh
     bounds_(allPoints_, syncPar),
     geometricD_(Vector<label>::zero),
     solutionD_(Vector<label>::zero),
+    comm_(Pstream::worldComm),
     pointZones_
     (
         IOobject
@@ -925,6 +928,18 @@ Foam::label Foam::polyMesh::nSolutionD() const
 }
 
 
+Foam::label Foam::polyMesh::comm() const
+{
+    return comm_;
+}
+
+
+Foam::label& Foam::polyMesh::comm()
+{
+    return comm_;
+}
+
+
 // Add boundary patches. Constructor helper
 void Foam::polyMesh::addPatches
 (

src/foam/meshes/polyMesh/polyMesh.H

@@ -134,6 +134,9 @@ private:
             //  Defined according to the presence of empty patches
             mutable Vector<label> solutionD_;
 
+            //- Communicator used for parallel communication
+            label comm_;
+
 
         // Zoning information
 
@@ -293,9 +296,8 @@ public:
         );
 
 
-    // Destructor
-
-        virtual ~polyMesh();
+    //- Destructor
+    virtual ~polyMesh();
 
 
     // Member Functions
@@ -386,23 +388,36 @@ public:
             //- Return the number of valid solved-for dimensions in the mesh
             label nSolutionD() const;
 
-            //- Return point zone mesh
-            const pointZoneMesh& pointZones() const
-            {
-                return pointZones_;
-            }
 
-            //- Return face zone mesh
-            const faceZoneMesh& faceZones() const
-            {
-                return faceZones_;
-            }
+            // Communication support
+
+                //- Return communicator used for parallel communication
+                label comm() const;
+
+                //- Return communicator used for parallel communication
+                label& comm();
+
+
+            // Point, face and cell zones
+
+                //- Return point zone mesh
+                const pointZoneMesh& pointZones() const
+                {
+                    return pointZones_;
+                }
+
+                //- Return face zone mesh
+                const faceZoneMesh& faceZones() const
+                {
+                    return faceZones_;
+                }
+
+                //- Return cell zone mesh
+                const cellZoneMesh& cellZones() const
+                {
+                    return cellZones_;
+                }
 
-            //- Return cell zone mesh
-            const cellZoneMesh& cellZones() const
-            {
-                return cellZones_;
-            }
 
             //- Return parallel info
             const globalMeshData& globalData() const;

src/foam/meshes/polyMesh/polyPatches/constraint/ggi/ggiPolyPatch.C

@@ -31,6 +31,7 @@ Contributor
 
 #include "ggiPolyPatch.H"
 #include "polyBoundaryMesh.H"
+#include "polyMesh.H"
 #include "addToRunTimeSelectionTable.H"
 #include "demandDrivenData.H"
 #include "polyPatchID.H"
@@ -65,7 +66,7 @@ bool Foam::ggiPolyPatch::active() const
     // For decomposition and reconstruction
     // If not runing in parallel and the patch is not local, this is a serial
     // operation on a piece of a parallel decomposition and is therefore
-    // inactive.  HJ, 5/Spe/2016
+    // inactive.  HJ, 5/Sep/2016
     if (!Pstream::parRun() && !localParallel())
     {
         return false;
@@ -770,6 +771,18 @@ const Foam::faceZone& Foam::ggiPolyPatch::zone() const
 }
 
 
+Foam::label Foam::ggiPolyPatch::comm() const
+{
+    return boundaryMesh().mesh().comm();
+}
+
+
+int Foam::ggiPolyPatch::tag() const
+{
+    return Pstream::msgType();
+}
+
+
 const Foam::labelList& Foam::ggiPolyPatch::zoneAddressing() const
 {
     if (!zoneAddressingPtr_)

src/foam/meshes/polyMesh/polyPatches/constraint/ggi/ggiPolyPatch.H

@@ -303,6 +303,15 @@ public:
             const faceZone& zone() const;
 
 
+        // Communications support
+
+            //- Return communicator used for communication
+            virtual label comm() const;
+
+            //- Return message tag to use for communication
+            virtual int tag() const;
+
+
         // Interpolation data
 
             //- Is this the master side?

src/foam/meshes/polyMesh/polyPatches/constraint/mixingPlane/mixingPlanePolyPatch.C

@@ -805,6 +805,18 @@ Foam::mixingPlanePolyPatch::patchToPatch() const
 }
 
 
+Foam::label Foam::mixingPlanePolyPatch::comm() const
+{
+    return boundaryMesh().mesh().comm();
+}
+
+
+int Foam::mixingPlanePolyPatch::tag() const
+{
+    return Pstream::msgType();
+}
+
+
 const Foam::vectorField&
 Foam::mixingPlanePolyPatch::reconFaceCellCentres() const
 {

src/foam/meshes/polyMesh/polyPatches/constraint/mixingPlane/mixingPlanePolyPatch.H

@@ -278,9 +278,8 @@ public:
         }
 
 
-    // Destructor
-
-        virtual ~mixingPlanePolyPatch();
+    //- Destructor
+    virtual ~mixingPlanePolyPatch();
 
 
     // Member functions
@@ -346,6 +345,15 @@ public:
             const mixingPlaneZoneInterpolation& patchToPatch() const;
 
 
+        // Communications support
+
+            //- Return communicator used for communication
+            virtual label comm() const;
+
+            //- Return message tag to use for communication
+            virtual int tag() const;
+
+
         // Interpolation functions
 
             //- Interpolate face field to profile: given field on a

src/foam/meshes/polyMesh/polyPatches/constraint/processor/processorPolyPatch.C

@@ -27,12 +27,13 @@ License
 #include "addToRunTimeSelectionTable.H"
 #include "dictionary.H"
 #include "SubField.H"
-#include "demandDrivenData.H"
 #include "matchPoints.H"
 #include "OFstream.H"
+#include "polyBoundaryMesh.H"
 #include "polyMesh.H"
 #include "foamTime.H"
 #include "transformList.H"
+#include "demandDrivenData.H"
 
 // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
 
@@ -134,6 +135,18 @@ Foam::processorPolyPatch::~processorPolyPatch()
 
 // * * * * * * * * * * * * * * * Member Functions  * * * * * * * * * * * * * //
 
+Foam::label Foam::processorPolyPatch::comm() const
+{
+    return boundaryMesh().mesh().comm();
+}
+
+
+int Foam::processorPolyPatch::tag() const
+{
+    return Pstream::msgType();
+}
+
+
 void Foam::processorPolyPatch::initAddressing()
 {
     polyPatch::initAddressing();

src/foam/meshes/polyMesh/polyPatches/constraint/processor/processorPolyPatch.H

@@ -43,6 +43,7 @@ SourceFiles
 #define processorPolyPatch_H
 
 #include "coupledPolyPatch.H"
+#include "polyBoundaryMesh.H"
 
 // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
 
@@ -189,9 +190,8 @@ public:
         }
 
 
-    // Destructor
-
-        virtual ~processorPolyPatch();
+    //- Destructor
+    virtual ~processorPolyPatch();
 
 
     // Member functions
@@ -208,6 +208,15 @@ public:
             return neighbProcNo_;
         }
 
+        // Communications support
+
+            //- Return communicator used for communication
+            virtual label comm() const;
+
+            //- Return message tag to use for communication
+            virtual int tag() const;
+
+
         //- Does the processor own the patch ?
         bool owner() const
         {

src/foam/meshes/polyMesh/polyPatches/constraint/regionCouple/regionCouplePolyPatch.C

@@ -921,6 +921,18 @@ void Foam::regionCouplePolyPatch::detach() const
 }
 
 
+Foam::label Foam::regionCouplePolyPatch::comm() const
+{
+    return boundaryMesh().mesh().comm();
+}
+
+
+int Foam::regionCouplePolyPatch::tag() const
+{
+    return Pstream::msgType();
+}
+
+
 const Foam::labelList& Foam::regionCouplePolyPatch::zoneAddressing() const
 {
     if (!zoneAddressingPtr_)

src/foam/meshes/polyMesh/polyPatches/constraint/regionCouple/regionCouplePolyPatch.H

@@ -291,7 +291,7 @@ public:
         }
 
 
-    //-Destructor
+    //- Destructor
     virtual ~regionCouplePolyPatch();
 
 
@@ -351,6 +351,15 @@ public:
             void detach() const;
 
 
+        // Communications support
+
+            //- Return communicator used for communication
+            virtual label comm() const;
+
+            //- Return message tag to use for communication
+            virtual int tag() const;
+
+
         // Interpolation data
 
             //- Is this the master side?

src/lduSolvers/amg/amgPolicy/pamgPolicy.C

@@ -651,7 +651,8 @@ Foam::autoPtr<Foam::amgMatrix> Foam::pamgPolicy::restrictMatrix
                     coarseInterfaces,
                     fineInterface,
                     fineInterface.interfaceInternalField(child_),
-                    fineInterfaceAddr[intI]
+                    fineInterfaceAddr[intI],
+                    Pstream::worldComm //HJ, AMG Comm fineMesh.comm()
                 ).ptr()
             );
         }

src/tetFiniteElement/tetPolyMesh/tetPolyMesh.C

@@ -61,6 +61,7 @@ void tetPolyMesh::clearOut() const
     clearOutParPointData();
 }
 
+
 // * * * * * * * * * * * * * * * * Constructors  * * * * * * * * * * * * * * //
 
 // Construct from components
@@ -71,6 +72,7 @@ tetPolyMesh::tetPolyMesh(const polyMesh& pMesh)
     boundary_(*this, pMesh.boundaryMesh()),
     faceOffset_(mesh_.nPoints()),
     cellOffset_(faceOffset_ + mesh_.nFaces()),
+    comm_(Pstream::worldComm),
     nPoints_(-1),
     nEdges_(-1),
     nTets_(-1),
@@ -105,6 +107,18 @@ tetPolyMesh::~tetPolyMesh()
 
 // * * * * * * * * * * * * * * * Member Functions  * * * * * * * * * * * * * //
 
+Foam::label Foam::tetPolyMesh::comm() const
+{
+    return comm_;
+}
+
+
+Foam::label& Foam::tetPolyMesh::comm()
+{
+    return comm_;
+}
+
+
 // Return number of edges in decomposition for a face
 label tetPolyMesh::nEdgesForFace(const label faceID) const
 {

src/tetFiniteElement/tetPolyMesh/tetPolyMesh.H

@@ -116,7 +116,7 @@ class tetPolyMeshMapper;
 class tetPolyMeshLduAddressing;
 
 /*---------------------------------------------------------------------------*\
-                     Class tetPolyMesh Declaration
+                           Class tetPolyMesh Declaration
 \*---------------------------------------------------------------------------*/
 
 class tetPolyMesh
@@ -138,6 +138,9 @@ class tetPolyMesh
         //- Offset in numbering to first cell centre
         label cellOffset_;
 
+        //- Communicator used for parallel communication
+        label comm_;
+
 
     // Demand-driven data
 
@@ -205,9 +208,9 @@ public:
         //- Construct from components
         explicit tetPolyMesh(const polyMesh& pMesh);
 
-    // Destructor
 
-        ~tetPolyMesh();
+    //- Destructor
+    virtual ~tetPolyMesh();
 
 
     // Member Functions
@@ -370,6 +373,15 @@ public:
                 const edgeList& parallelEdges() const;
 
 
+            // Communication support
+
+                //- Return communicator used for parallel communication
+                label comm() const;
+
+                //- Return communicator used for parallel communication
+                label& comm();
+
+
         // Edit
 
             //- Update mesh topology using the morph engine