Sherlock: A Multi-Objective Design Space Exploration Framework

Design space exploration (DSE) provides intelligent methods to tune the large number of optimization parameters present in modern FPGA high-level synthesis tools. High-level parameter tuning is a time-consuming process due lengthy hardware compilation times—synthesizing an design can take tens hours. DSE helps find optimal solution faster than brute-force without relying on designer intuition achieve high-quality results. Sherlock framework that handle multiple conflicting objectives and aggressively focuses finding Pareto-optimal solutions. integrates model selection choose regression reach faster. designs strategy based around multi-armed bandit problem, opting balance exploitation learned expected decrease importance models do not provide correct estimates, reaching capable tailoring its choice problem at hand, leading best reflects application space. We have tested dataset problems found converges toward set similar frameworks.