High-level synthesis consists of many interdependent tasks such as scheduling, allocation and binding. To make efficient use of time and area, functional unit allocation must be performed using a library of modules which contains a variety of module types with identical functionality, but different area and delay characteristics. The synthesis technique presented in the paper simultaneously performs scheduling, allocation and module selection, using problem-space genetic algorithm (PSGA) to produce area and performance optimised designs. The PSGA-based system uses an intelligent design-space exploration technique by combining a genetic algorithm with a simple and fast problem-specific heuristic to search a large design space effectively and efficiently. The efficient exploration of design-space is essential to design cost-effective architectures for problems of VLSI/ULSI complexity. The PSGA method offers several advantages such as the versatility, simplicity, objective independence and the computational advantages for problems of large size over other existing techniques. The proposed synthesis system handles multicycle functional units, chaining, conditional constructs, loops and structural pipelining. Experiments on benchmarks show very promising results.
ASJC Scopus subject areas
- Theoretical Computer Science
- Hardware and Architecture
- Computational Theory and Mathematics