ParMETIS is a parallel version of METIS and can be used as an alternative to
ptScotch or Zoltan, supporting multi-constraints and redistribution:
Description
ParMetis redistribution in parallel
Note: parMetis methods do not support serial operation.
Parameters
- Method of decomposition
- kWay: multilevel k-way
- geomKway: combined coordinate-based and multi-level k-way
- adaptiveRepart: balances the work load of a graph
- Options
- options[0]: The specified options are used if options[0] = 1
- options[1]: Specifies the level of information to be returned during
the execution of the algorithm. Timing information can be obtained by
setting this to 1. Additional options for this parameter can be obtained
by looking at parmetis.h. Default: 0.
- options[2]: Random number seed for the routine
- options[3]: Specifies whether the sub-domains and processors are coupled
or un-coupled. If the number of sub-domains desired (i.e., nparts) and
the number of processors that are being used is not the same, then these
must be un-coupled. However, if nparts equals the number of processors,
these can either be coupled or de-coupled. If sub-domains and processors
are coupled, then the initial partitioning will be obtained implicitly
from the graph distribution. However, if sub-domains are un-coupled from
processors, then the initial partitioning needs to be obtained from the
initial values assigned to the part array.
- itr: Parameter which describes the ratio of inter-processor communication
time compared to data redistribution time. Should be set between 0.000001
and 1000000.0. If set high, a repartitioning with a low edge-cut will be
computed. If it is set low, a repartitioning that requires little data
redistribution will be computed. Good values for this parameter can be
obtained by dividing inter-processor communication time by data
redistribution time. Otherwise, a value of 1000.0 is recommended.
Default: 1000.
The ParMETIS sources can be downloaded and compiled in ThirdParty-dev using the
link in the README file and the compilation commands in Allwmake.
Note the specific license under which ParMETIS is released:
Copyright & License Notice
--------------------------
The ParMETIS package is copyrighted by the Regents of the
University of Minnesota. It can be freely used for educational and
research purposes by non-profit institutions and US government
agencies only. Other organizations are allowed to use ParMETIS
only for evaluation purposes, and any further uses will require prior
approval. The software may not be sold or redistributed without prior
approval. One may make copies of the software for their use provided
that the copies, are not sold or distributed, are used under the same
terms and conditions.
As unestablished research software, this code is provided on an
``as is'' basis without warranty of any kind, either expressed or
implied. The downloading, or executing any part of this software
constitutes an implicit agreement to these terms. These terms and
conditions are subject to change at any time without prior notice.
Optional CPU load caching can be switched-on for Lagrangian cloud tracking
and/or chemistry integration using the new cpuLoad switch in the cloudProperties
or chemistryProperties dictionary files respectively and used for
multi-constraint load-balancing by the fvMeshDistributorsLoadBalancer specified
in the dynamicMeshDict file
distributor
{
type loadBalancer;
libs ("libfvMeshDistributors.so");
multiConstraint true;
redistributionInterval 10;
}
which used the distributor specified in the decomposeParDict file, e.g.
numberOfSubdomains 12;
decomposer simple;
distributor zoltan;
libs ("libzoltanDecomp.so");
simpleCoeffs
{
n (2 2 3);
}
zoltanCoeffs
{
lb_method rcb;
}
The incompressibleDenseParticleFluid/cyclone case has been updated to
demonstrate this new functionality and shows a speedup ~50% using the Zoltan RCB
multi-constraint distributor. The multicomponentFluid/counterFlowFlame2D_GRI
case has also been updated to use the new cpuLoad switch.
The fact that these names create sources in their associated transport
equations is clear in context, so the name does not need to contain
'Source'.
Having 'Source' in the name is a historic convention that dates back to
when fvModels and fvConstraints were combined in a single fvOptions
interface. In this interface, disambiguation between sources and
constraints was necessary.
The full set of name changes is as follows:
accelerationSource -> acceleration
actuationDiskSource -> actuationDisk
effectivenessHeatExchangerSource -> effectivenessHeatExchanger
explicitPorositySource -> porosityForce
radialActuationDiskSource -> radialActuationDisk
rotorDiskSource -> rotorDisk
sixDoFAccelerationSource -> sixDoFAcceleration
solidEquilibriumEnergySource -> solidThermalEquilibrium
solidificationMeltingSource -> solidificationMelting
volumeFractionSource -> volumeBlockage
interRegionExplicitPorositySource -> interRegionPorosityForce
VoFSolidificationMeltingSource -> VoFSolidificationMelting
The old names are still available for backwards compatibility.
The fact that these names refer to constraints is clear in context, so
the name does not need to contain 'Constraint'.
Having 'Constraint' in the name is a historic convention that dates back
to when fvConstraints and fvModels were combined in a single fvOptions
interface. In this interface, disambiguation between sources and
constraints was necessary.
This change has been applied to the 'fixedValue' and 'fixedTemperature'
constraints, which were formerly named 'fixedValueConstraint' and
'fixedTemperatureConstraint', respectively.
The old names are still available for backwards compatibility.
When an fvModel source introduces fluid into a simulation it should also
create a corresponding source term for all properties transported into
the domain by that injection. The source is, effectively, an alternative
form of inlet boundary, on which all transported properties need an
inlet value specified.
These values are now specified in the property field files. The
following is an example of a 0/U file in which the velocity of fluid
introduced by a fvModel source called "injection1" is set to a fixed
value of (-1 0 0):
dimensions [0 1 -1 0 0 0 0];
internalField uniform (0 0 0);
boundaryField
{
#includeEtc "caseDicts/setConstraintTypes"
wall
{
type noSlip;
}
atmosphere
{
type pressureInletOutletVelocity;
value $internalField;
}
}
// *** NEW ***
sources
{
injection1
{
type uniformFixedValue;
uniformValue (-1 0 0);
}
}
And the following entry in the 0/k file specifies the turbulent kinetic
energy introduced as a fraction of the mean flow kinetic energy:
sources
{
injection1
{
type turbulentIntensityKineticEnergy;
intensity 0.05;
}
}
The specification is directly analogous to boundary conditions. The
conditions are run-time selectable and can be concisely implemented.
They can access each other and be inter-dependent (e.g., the above,
where turbulent kinetic energy depends on velocity). The syntax keeps
field data localised and makes the source model (e.g., massSource,
volumeSource, ...) specification independent from what other models and
fields are present in the simulation. The 'fieldValues' entry previously
required by source models is now no longer required.
If source values need specifying and no source condition has been
supplied in the relevant field file then an error will be generated.
This error is similar to that generated for missing boundary conditions.
This replaces the behaviour where sources such as these would introduce
a value of zero, either silently or with a warning. This is now
considered unacceptable. Zero might be a tolerable default for certain
fields (U, k), but is wholly inappropriate for others (T, epsilon, rho).
This change additionally makes it possible to inject fluid into a
multicomponent solver with a specified temperature. Previously, it was
not possible to do this as there was no means of evaluating the energy
of fluid with the injected composition.
Lagrangian's dependency set is simpler than it used to be. There is no
longer a need to maintain a separate library for models that depend on
the momentum transport modelling.
setFormat no longer defaults to the value of graphFormat optionally set in
controlDict and must be set in the functionObject dictionary.
boundaryFoam, financialFoam and pdfPlot still require a graphFormat entry in
controlDict but this is now read directly rather than by Time.
The parcelsPerSecond control can now be specified as a time-varying
function. This provides additional control over the temporal
distribution of injected parcels, which may be advantageous if, for
example, the mass flow rate varies significantly. It also enables
variable flow rates of particulates in cases which have a fixed number
of particles per parcel.
