Static allocation of parallel program modules onto a message passing multiprocessor system

摘要

Studies the problem of statically allocating parallel program modules onto an MPS (message passing multiprocessor system) so as to minimize the total execution time. Each node in the interconnection network of the MPS has buffers at its input ports to store messages waiting for the transmission. An algorithm for finding a route which minimizes the communication delay between processors is first given. Next, a heuristic algorithm for allocating program modules onto the MPS is presented.