System for parallel implementation of combinatorial optimization in a multiprocessor network for generating search graphs for solving enumerative problems

A method of operating a plurality of digital processors wherein a root processor interfaces with a user and a PICO system operates in interemdiacy between said root processor and a network of multiprocessors to solve enumerative problems and the like using a shadow node system generated under a speculative node expansion mode, in said PICO system, to use idle processor capacity within said network of processors and said shadow node to be prospectively integrated into a successor node by the user and convergence between said shadow nodes and said successor nodes is confirmed, comprising the steps of:

• storing a root node on a list of nodes on an expansion queue in a node generation subsystem of the PICO system:
• allocating said expanded nodes to the network of multiprocessors;
• hashing said expanded nodes;
• generating offsprings forming shadow nodes to occupy idle capacity of the network of multiprocessors;
• interacting with a message subsystem, initialized by the user, to refine said shadow nodes comprising the steps of:
• routing shadow node offsprings via a message channel subsystem in the PICO system;
• removing nodes to curtail offspring generation:
• broadcasting bounds, values and upper and lower limits of said offspring to thereby compare with said root node and determine convergence;
• and
• routing said broadcast to deploy message to the network of processors via a message channel subsystem in the PICO system;
• connecting the PICO system to said root processor via a state vector;
• and
• managing memory and shutdown operation of the PICO system by an auxiliary function means.