A MULTICONDUCTOR MODELING ENGINE EMPLOYING A MONTE CARLO ALGORITHM

TRENDS IN COMPUTING TODAY RELY HEAVILY ON PARALLEL PROCESSING IN SOLVING COMPLEX FIELD MODELING PROBLEMS. BECAUSE OF INTERPROCESSOR COMMUNICATION PROBLEMS, HOWEVER, THE ULTIMATE SPEED AT WHICH PROBLEMS CAN BE SOLVED IS LIMITED. MULTIV AVOIDS THIS DILEMMA BY EMPLOYING AN OLD YETPOTENTIALLY MASSIVELY PARALLEL TECHNIQUE, NAMELY THE RANDOM WALK METHOD OF MONTE CARLO. USING THIS METHOD, A LARGE NUMBER OF LOW COST PROCESSORS OPERATING INDEPENDENTLY CAN SOLVE PROBLEMS OF ANY COMPLEX GEOMETRY EXTREMELY EFFECTIVELY. THIS PROPOSAL WILL STUDY AND COMPARE THE MONTECARLO TECHNIQUE AGAINST STANDARD TECHNIQUES, SUCH AS FINITE DIFFERENCE AND FINITE ELEMENT METHODS TO DETERMINE THE POSSIBLE LONG RANGE MERIT IN INVESTING IN PROCESSOR HARDWAREFOR THE MODELER. A PROTOTYPE MODELER BUILT USING LOW COST TRANSPUTERS WILL BE DEVELOPED. IF THE STUDY VERIFIES THE MERIT IN THE MONTE CARLO METHOD, IT WOULD HAVE NEW IMPLICATIONS FOR PROBLEM SOLVING USING FUTURE HARDWARE: AN INVESTMENT IN MASSIVE INEXPENSIVE PROCESSORS, RATHER THAN EXPENSIVE INTERCONNECTED MACHINES, COULD BE OF ECONOMIC ADVANTAGE IN SOLVING A LARGE CLASS OF PROBLEMS IN FIELD THEORY, WEATHER, NUCLEAR SCIENCE, THERMODYNAMICS AND OTHER AREAS.