Genetic algorithm driven hardware-software partitioning for dynamically reconfigurable embedded systems

The need for inexpensive, compact and adaptive systems prompted considerabl e interest in the hardware-software co-design of embedded systems. In parti cular, Dynamically Reconfigurable Embedded Systems, which exploit the advan ces in Field Programmable Gate Array (FPGA) technology, facilitate customis ation of their hardware resources during runtime to meet the demands of exe cuting, applications. The ability to estimate the resultant acceleration ob tained is highly desirable, as time to market deadlines are being ever shor tened. The performance of such systems is fundamentally dependent on the ha rdware-software partition. In this paper, a genetic algorithm-based (GA) ha rdware-software partitioning method is presented. Demonstrative application s are used to illustrate the effectiveness of the GA approach at exploiting the inherent reconfigurable nature of such systems to obtain optimal or ne ar optimal performance speedup relative to a conventional software implemen tation. (C) 2001 Elsevier Science B.V. All rights reserved.