linpart.h

// Copyright 2009-2017 Sandia Corporation. Under the terms
// of Contract DE-AC04-94AL85000 with Sandia Corporation, the U.S.
// Government retains certain rights in this software.
//
// Copyright (c) 2009-2017, Sandia Corporation
// All rights reserved.
//
// This file is part of the SST software package. For license
// information, see the LICENSE file in the top level directory of the
// distribution.


#ifndef SST_CORE_PART_LINEAR
#define SST_CORE_PART_LINEAR

#include <sst/core/part/sstpart.h>
#include <sst/core/elementinfo.h>

using namespace SST;
using namespace SST::Partition;

namespace SST {

class Output;

namespace Partition {


    
/**
Performs a linear partition scheme of an SST simulation configuration. In this
scheme a list of components (supplied as a graph) are grouped by slicing the list
into approximately equal parts. A "part" is generated for each MPI rank performing
the simulation. This means Components with sequential ids will be placed close together.
In general this scheme provides a very effective partition for most parallel
simulations which generate man similar components of interest close together in the input
Python configuration. It is also very fast to compute a linear partiton scheme. For
more aggressive partition schemes users should try either a simple or Zoltan-based
partitioner.
*/
class SSTLinearPartition : public SST::Partition::SSTPartitioner {

protected:
    /** Number of ranks in the simulation */
    RankInfo rankcount;
    /** Output object to print partitioning information */
    Output* partOutput;
    
public:
    /**
       Creates a new linear partition scheme.
       \param mpiRankCount Number of MPI ranks in the simulation
       \param verbosity The level of information to output
    */
    SSTLinearPartition(RankInfo rankCount, RankInfo my_rank, int verbosity);
    
    /**
       Performs a partition of an SST simulation configuration
       \param graph The simulation configuration to partition
    */
    void performPartition(PartitionGraph* graph) override;
    
    bool requiresConfigGraph() override { return false; }
    bool spawnOnAllRanks() override { return false; }
    
    // static SSTPartitioner* allocate(RankInfo total_ranks, RankInfo my_rank, int verbosity) {
    //     return new SSTLinearPartition(total_ranks, my_rank, verbosity);
    // }
    
    SST_ELI_REGISTER_PARTITIONER(SSTLinearPartition,"sst","linear","Partitions components by dividing Component ID space into roughly equal portions.  Components with sequential IDs will be placed close together.")

    SST_ELI_DOCUMENT_VERSION(1,0,0)
};

}
}

#endif