next up previous contents index
Next: Edge list Up: Objects Previous: Run Time Type Identification   Contents   Index

Node and edge weights

By default, the graph partitioning algorithm used in Graphcode weights each node and link equally. However, it is possible to perform load balancing by specifying a computational weight function on each node, and a communication weight function for each edge. For example:

class foo: public object
{
  ...
  virtual idxtype weight() {return size()*size();}
  vitural idxtype edgeweight(const objref& x) {return (*x)->size();}
};



Russell Standish 2016-09-02