Work-in-Progress: On Leveraging Approximations for Exact System-level Design Space Exploration

• Published in: 2018 International Conference on Hardware/Software Codesign and System Synthesis (CODES+ISSS) pp. 1 - 2
• Main Authors: Neubauer, Kai, Haubelt, Christian, Wanko, Philipp, Schaub, Torsten
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-09-2018
• Subjects: Approximation; Design Space Exploration; Embedded systems; Estimation; Filtering theory; Linear programming; Search problems; Space exploration; System-level design
• DOI: 10.1109/CODESISSS.2018.8525974

In order to find good design points for embedded systems, an efficient exploration of the design space is imperative. The ever-increasing complexity of embedded systems, however, results in a deterioration of the overall exploration performance. The DSE essentially consists of two parts: (1) the search for feasible solutions and (2) the evaluation of found feasible solutions. While the search has been massively improved by ASPmT-based strategies, the evaluation emerges as the main bottleneck. Tragically, evaluating bad solutions takes as much time as evaluating good ones. Hence, in this paper we study the utilization of approximations in the evaluation process integrated in an ASPmT-based DSE to identify bad solutions more quickly while still retaining the exact Pareto-front.