- allocate/free tags.
Not tested.

OK - test blockMesh with cyclics
unitTestCases/singleCyclic/

OK - test cyclics sequential running.
unitTestCases/singleCyclic/

OK - test decomposePar
tested channel395

OK - FaceCellWave
    unitTestCases/twoCavityCyclicForWallDistance/

OK - non-parallel finite volume : channelFoam
    unitTestCases/channel395-splitCyclic vs. channel395-dev

OK - parallel finite volume with processorCyclic: channelFoam
    unitTestCases/channel395-splitCyclic vs. channel395-dev

OK - preProcessing/foamUpgradeCyclics

OK - gamg - sequential.
Tested on channel395-splitCyclic with GAMG.
OK - gamg parallel.
Tested on channel395-splitCyclic with GAMG.

- initTransfer in GAMGprocessorInterfaces using nonblocking+tags
untested.

OK - cyclic baffles.
Tested on t-junction-with-fan

OK. - jumpCyclics/fanFvPatchField. All usages of jump() now need to account
for being owner() or not.
Tested on t-junction-with-fan.

OK - regionSplit
tested on singleCyclic

OK - pointFields on cyclics. volPointInterpolation.
tested on channel395-splitCyclic

OK - fvMeshSubset
tested on singleCyclic

OK - pointEdgeWave (maybe test through inversePointDistanceDiffusivity?)
tested on twoCavityCyclicForWallDistance

OK - scotchDecomp
tested with testCalcCSR on twoCavityCyclicForWallDistance

NOT WORKING - fvMeshDistribute to split cyclic patches into ones
  with different separation.
tested on singleCyclic

OK - test createPatch pointSync
note: only works if face-centre position of 0th faces is ok since uses
this for separation. Should in fact make cyclic planar using patch centre and
normal?
test on twoCavityCyclicForWallDistance with point (0 1 0) set to (0 1.001 0)

NO PROBLEM - renumberMesh
It doesn't do renumbering through cyclics.

OK - rotational cyclics.
Tested on movingCone-with-cyclics

OK - LUscalarMatrix::convert still expects interfaces to be cyclic
tested on channel395 with 'directSolveCoarsest true;' 

OK - grep for size()/2

- all tutorials with cyclics:
    OK - incompressible/DNS/dnsFoam/boxTurb16
    OK - incompressible/channelFoam/channel395
    slight differences due to divergence. combustion/XiFoam/les/pitzDaily3D
    OK - no cyclics. combustion/fireFoam/les/smallPoolFire2D
    discreteMethods/dsmcFoam/freeSpacePeriodic
    discreteMethods/dsmcFoam/wedge15Ma5
    discreteMethods/molecularDynamics/mdEquilibrationFoam/periodicCubeArgon
    OK - incompressible/boundaryFoam/boundaryLaunderSharma
    OK - incompressible/boundaryFoam/boundaryWallFunctions
    OK - incompressible/boundaryFoam/boundaryWallFunctionsProfile
    incompressible/pimpleFoam/t-junction-with-fan
    OK - incompressible/simpleSRFFoam/mixer
    lagrangian/porousExplicitSourceReactingParcelFoam/filter
    lagrangian/reactingParcelFilmFoam/multipleBoxes


- have foamUpgradeCyclics split 'value' field
- createBaffles
- activeBaffleVelocity
- kivaToFoam/readKivaGrid.H sorts cyclics (but in incorrect order?)
- isoSurface.C
- referredCellList.C
- work out scheduled communication?