/*---------------------------------------------------------------------------*\
========= |
\\ / F ield | OpenFOAM: The Open Source CFD Toolbox
\\ / O peration |
\\ / A nd | Copyright (C) 2011-2016 OpenFOAM Foundation
\\/ M anipulation |
-------------------------------------------------------------------------------
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 .
Class
Foam::mapDistribute
Description
Class containing processor-to-processor mapping information.
We store mapping from the bits-to-send to the complete starting list
(subXXXMap) and from the received bits to their location in the new
list (constructXXXMap).
Note:
Schedule is a list of processor pairs (one send, one receive. One of
them will be myself) which forms a scheduled (i.e. non-buffered) exchange.
See distribute on how to use it.
Note2: number of items sent on one processor have to equal the number
of items received on the other processor.
To aid constructing these maps there are the constructors from global
numbering, either with or without transforms.
- without transforms:
Constructors using compact numbering: layout is
- all my own elements first (whether used or not)
- followed by used-only remote elements sorted by remote processor.
So e.g 4 procs and on proc 1 the compact
table will first have all globalIndex.localSize() elements from proc1
followed by used-only elements of proc0, proc2, proc3.
The constructed mapDistribute sends the local elements from and
receives the remote elements into their compact position.
compactMap[procI] is the position of elements from procI in the compact
map. compactMap[myProcNo()] is empty since trivial addressing.
It rewrites the input global indices into indices into the constructed
data.
- with transforms:
This requires the precalculated set of possible transforms
(globalIndexAndTransform). These are given as permutations (+, -, or none)
of up to 3 independent transforms.
The layout of the data is
- all my own elements first (whether used or not)
- followed by used-only remote elements sorted by remote processor.
- followed by - for each transformation index - the set of local or
remote elements with that transformation.
The inputs for the constructor are
- the set of untransformed local or remote indices in globalIndex
numbering. These get rewritten to be indices into the layout of the data.
- the set of transformed local or remote indices in globalIndexAndTransform
encoding. These are labelPairs.
Any distribute with transforms is now done as:
1. exchange data with other processors and receive these into the
slots for that processor
2. for all transformations transform a subset of the data according
to transformElements_[transformI] and store this starting from
transformStart_[transformI]
In the same way a reverse distribute will
1. apply the inverse transform to the data starting at
transformStart_[transformI] and copy the result back into the
transformElements_[transformI]. These might be local or remote slots.
2. the data in the remote slots will now be sent back to the correct
location in the originating processor.
E.g. a map to handle
- mesh points on a mesh with
- 1 cyclic so 3 permutations (+,-,none) will have layout
- on e.g. processor 1 out of 2:
+------+ <- transformStart[2]
| |
| | <- transform2 applied to data in local or remote slots
| |
+------+ <- transformStart[1]
| |
| | <- transform1 applied to data in local or remote slots
| |
+------+ <- transformStart[1]
| |
| | <- transform0 applied to data in local or remote slots
| |
+------+ <- transformStart[0]
| |
| | <- data from proc2
| |
+------+
| |
| | <- data from proc0
| |
+------+ <- mesh.nPoints()
| |
| |
| |
+------+ 0
SourceFiles
mapDistribute.C
mapDistributeTemplates.C
\*---------------------------------------------------------------------------*/
#ifndef mapDistribute_H
#define mapDistribute_H
#include "transformList.H"
#include "labelList.H"
#include "labelPair.H"
#include "Pstream.H"
#include "boolList.H"
#include "Map.H"
#include "vectorTensorTransform.H"
#include "coupledPolyPatch.H"
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
namespace Foam
{
class mapPolyMesh;
class globalIndex;
class PstreamBuffers;
class globalIndexAndTransform;
/*---------------------------------------------------------------------------*\
Class mapDistribute Declaration
\*---------------------------------------------------------------------------*/
class mapDistribute
{
// Private data
//- Size of reconstructed data
label constructSize_;
//- Maps from subsetted data back to original data
labelListList subMap_;
//- Maps from subsetted data to new reconstructed data
labelListList constructMap_;
// Optional transformation
//- For every globalIndexAndTransform::transformPermutations
// gives the elements that need to be transformed
labelListList transformElements_;
//- Destination in constructMap for transformed elements
labelList transformStart_;
//- Schedule
mutable autoPtr> schedulePtr_;
// Private Member Functions
static void checkReceivedSize
(
const label procI,
const label expectedSize,
const label receivedSize
);
//- Construct per processor compact addressing of the global elements
// needed. The ones from the local processor are not included since
// these are always all needed.
void calcCompactAddressing
(
const globalIndex& globalNumbering,
const labelList& elements,
List