Heuristic genetic algorithm of computational pipelines optimization

Abstract

An analysis of computational pipelines and their optimization methods has been performed. A class of dataflow pipelines that do not use resource sharing and obtain high throughput has been considered. Pipeline optimization tasks being solved during high-level synthesis from random logic behavioral specifications in programming and hardware description languages have been considered. A heuristic genetic optimization algorithm which is capable of handling large designs has been proposed and investigated.

Keywords

• data flow
• computational pipeline
• high-level synthesis
• optimization