openfoam/src/OpenFOAM/meshes/ProcessorTopology/ProcessorTopology.C

/*---------------------------------------------------------------------------*\
  =========                 |
  \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox
   \\    /   O peration     |
    \\  /    A nd           | www.openfoam.com
     \\/     M anipulation  |
-------------------------------------------------------------------------------
    Copyright (C) 2011-2017 OpenFOAM Foundation
    Copyright (C) 2022 OpenCFD Ltd.
-------------------------------------------------------------------------------
License
    This file is part of OpenFOAM.

    OpenFOAM is free software: you can redistribute it and/or modify it
    under the terms of the GNU General Public License as published by
    the Free Software Foundation, either version 3 of the License, or
    (at your option) any later version.

    OpenFOAM is distributed in the hope that it will be useful, but WITHOUT
    ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
    FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
    for more details.

    You should have received a copy of the GNU General Public License
    along with OpenFOAM.  If not, see <http://www.gnu.org/licenses/>.

\*---------------------------------------------------------------------------*/

#include "ProcessorTopology.H"
#include "ListOps.H"
#include "Pstream.H"
#include "commSchedule.H"

// * * * * * * * * * * * * * Private Member Functions  * * * * * * * * * * * //

template<class Container, class ProcPatch>
Foam::labelList Foam::ProcessorTopology<Container, ProcPatch>::procNeighbours
(
    const label nProcs,
    const Container& patches
)
{
    // Determine number of processor neighbours and max neighbour id.

    label nNeighbours = 0;

    label maxNb = 0;

    boolList isNeighbourProc(nProcs, false);

    forAll(patches, patchi)
    {
        const auto* cpp = isA<ProcPatch>(patches[patchi]);
        if (cpp)
        {
            const label nbrProci = cpp->neighbProcNo();

            if (!isNeighbourProc[nbrProci])
            {
                isNeighbourProc[nbrProci] = true;
                maxNb = max(maxNb, nbrProci);
                ++nNeighbours;
            }
        }
    }

    labelList neighbours(nNeighbours, -1);

    nNeighbours = 0;

    forAll(isNeighbourProc, proci)
    {
        if (isNeighbourProc[proci])
        {
            neighbours[nNeighbours++] = proci;
        }
    }

    procPatchMap_.setSize(maxNb + 1);
    procPatchMap_ = -1;

    forAll(patches, patchi)
    {
        const auto* cpp = isA<ProcPatch>(patches[patchi]);
        if (cpp)
        {
            const label nbrProci = cpp->neighbProcNo();

            // Reverse map
            procPatchMap_[nbrProci] = patchi;
        }
    }

    return neighbours;
}


// * * * * * * * * * * * * * * * * Constructors  * * * * * * * * * * * * * * //

template<class Container, class ProcPatch>
Foam::ProcessorTopology<Container, ProcPatch>::ProcessorTopology
(
    const Container& patches,
    const label comm
)
:
    labelListList(Pstream::nProcs(comm)),
    patchSchedule_()
{
    if (Pstream::parRun())
    {
        // Fill my 'slot' with my neighbours
        operator[](Pstream::myProcNo(comm)) =
            procNeighbours(this->size(), patches);

        // Distribute to all processors
        Pstream::allGatherList(*this, Pstream::msgType(), comm);
    }

    if
    (
        Pstream::parRun()
     && Pstream::defaultCommsType == Pstream::commsTypes::scheduled
    )
    {
        patchSchedule_.resize(2*patches.size());

        label patchEvali = 0;

        // 1. All non-processor patches
        // ~~~~~~~~~~~~~~~~~~~~~~~~~~~~

        forAll(patches, patchi)
        {
            if (!isA<ProcPatch>(patches[patchi]))
            {
                patchSchedule_[patchEvali].patch = patchi;
                patchSchedule_[patchEvali++].init = true;
                patchSchedule_[patchEvali].patch = patchi;
                patchSchedule_[patchEvali++].init = false;
            }
        }

        // 2. All processor patches
        // ~~~~~~~~~~~~~~~~~~~~~~~~

        // Determine the schedule for all. Insert processor pair once
        // to determine the schedule. Each processor pair stands for both
        // send and receive.
        label nComms = 0;
        for (const labelList& neighbours : *this)
        {
            nComms += neighbours.size();
        }
        DynamicList<labelPair> comms(nComms);

        forAll(*this, proci)
        {
            const labelList& neighbours = operator[](proci);

            for (const label nbrProci : neighbours)
            {
                if (proci < nbrProci)
                {
                    comms.append(labelPair(proci, nbrProci));
                }
            }
        }
        comms.shrink();

        // Determine a schedule.
        labelList mySchedule
        (
            commSchedule
            (
                Pstream::nProcs(comm),
                comms
            ).procSchedule()[Pstream::myProcNo(comm)]
        );

        for (const label commI : mySchedule)
        {
            // Get the other processor
            label nb = comms[commI][0];
            if (nb == Pstream::myProcNo(comm))
            {
                nb = comms[commI][1];
            }
            label patchi = procPatchMap_[nb];

            if (Pstream::myProcNo(comm) > nb)
            {
                patchSchedule_[patchEvali].patch = patchi;
                patchSchedule_[patchEvali++].init = true;
                patchSchedule_[patchEvali].patch = patchi;
                patchSchedule_[patchEvali++].init = false;
            }
            else
            {
                patchSchedule_[patchEvali].patch = patchi;
                patchSchedule_[patchEvali++].init = false;
                patchSchedule_[patchEvali].patch = patchi;
                patchSchedule_[patchEvali++].init = true;
            }
        }
    }
    else
    {
        patchSchedule_ = nonBlockingSchedule(patches);
    }
}


// * * * * * * * * * * * * * * * Member Functions  * * * * * * * * * * * * * //

template<class Container, class ProcPatch>
Foam::lduSchedule
Foam::ProcessorTopology<Container, ProcPatch>::nonBlockingSchedule
(
    const Container& patches
)
{
    lduSchedule patchSchedule(2*patches.size());

    label patchEvali = 0;

    // 1. All non-processor patches
    // ~~~~~~~~~~~~~~~~~~~~~~~~~~~~

    // Have evaluate directly after initEvaluate. Could have them separated
    // as long as they're not intermingled with processor patches since
    // then e.g. any reduce parallel traffic would interfere with the
    // processor swaps.

    forAll(patches, patchi)
    {
        if (!isA<ProcPatch>(patches[patchi]))
        {
            patchSchedule[patchEvali].patch = patchi;
            patchSchedule[patchEvali++].init = true;
            patchSchedule[patchEvali].patch = patchi;
            patchSchedule[patchEvali++].init = false;
        }
    }

    // 2. All processor patches
    // ~~~~~~~~~~~~~~~~~~~~~~~~

    // 2a. initEvaluate
    forAll(patches, patchi)
    {
        if (isA<ProcPatch>(patches[patchi]))
        {
            patchSchedule[patchEvali].patch = patchi;
            patchSchedule[patchEvali++].init = true;
        }
    }

    // 2b. evaluate
    forAll(patches, patchi)
    {
        if (isA<ProcPatch>(patches[patchi]))
        {
            patchSchedule[patchEvali].patch = patchi;
            patchSchedule[patchEvali++].init = false;
        }
    }

    return patchSchedule;
}


// ************************************************************************* //