Synthesis of application-specific multiprocessor systems

The current VLSI/ULSI technology enables the implementation of a complicated system in a single chip at a low cost. Thus it has become cost effective to design special-purpose multiprocessor architectures for computationally-intensive applications in signal processing, control of power systems and robotics. In this paper, we introduce a novel design methodology based on the problem-space genetic algorithms for the synthesis of application-specific multiprocessor systems to meet the various cost and performance constraints, not merely the mapping of tasks onto a given system. The design methodology for the synthesis of a custom heterogeneous multiprocessor system selects the number and type of processor, finds an interconnection pattern between processors, maps the subtasks onto the system and provides a static schedule for the subtasks execution for a given application specified in terms of a task flow graph. The proposed problem-space genetic algorithms (PSGA) based design methodology combines the power of genetic algorithms, a global search method, with a problem-specific efficient heuristic to search a large design space in an intelligent way in order to find a global optimal solution within acceptable cpu times. This paper applies a novel PSGA-based design methodology and achieves a considerable improvements in the cpu times over the previous works. Comparisons are made between our and the previous works to show the strength of our methodology for the applications of large sizes. For the same designs, our technique gives optimal results in seconds, while the previous approach gives same results in hours