Skip to Main Content
This paper presents a novel Genetic Algorithm based exploration approach for integrated scheduling, allocation and binding in High Level Synthesis for Digital Media applications. The contributions of the proposed approach in this paper are: (a) Exploration of performance-hardware area tradeoffs of DSP digital media applications (b) Novel multi structure topology for chromosome encoding (c) Introduction of a novel cost function based on data pipelined performance-hardware area constraints (d) Novel load factor heuristic for chromosome decoding (e) Novel seeding process of the initial population based on serial and parallel implementation logic (f) Novel merit score (M-score) technique to assess the efficiency of the proposed approach (g) Novel cost value (C-value) metric that assesses the quality of final integrated solution. The proposed approach obtained an improvement of ≈2% in quality of final solution compared to a current technique as well as achieved an efficiency of 58.33% compared to same current approach, when applied on the digital media applications.
Date of Conference: 8-11 May 2011