Exploration of Multi-objective Tradeoff during High Level Synthesis Using Bacterial Chemotaxis and Dispersal

Abstract A novel application of bacterial foraging optimization algorithm (BFOA) in the area of design space exploration (DSE) of datapath in high level synthesis (HLS) is presented in this paper. The BFOA has been transformed into an adaptive automated DSE framework that is capable to handle tradeoffs between area-execution time during HLS. To the authors belief, no such application (or transformation) of BFOA into DSE exist in the literature. The key sub-contributions of the proposed approach can be classified as follows: i) Exploration drift using a novel chemotaxis algorithm ii) Diversity introduction in resource configuration using a novel dispersal algorithm iii) Performance analysis of proposed and related approaches on metrics such as generational distance, maximum pareto-optimal front error, spacing, spreading and weighted metric. Finally, results indicated an average improvement in Quality of Results (QoR) of ∼6% and reduction in exploration runtime of > 18% compared to three recent approaches based on PSO and GA.