Combining analytical and simulation-based design space exploration for time-critical systems

In the context of the design on time-critical systems, analytical models with worst case workloads are used to identify safe solutions that guarantee hard timing constraints. However, the focus on the worst case often leads to unnecessarily pessimistic and inefficient solutions, in particular for mixed-critical systems. To overcome the situation, the paper proposes a novel design flow integrating analytical and simulation-based design space exploration (DSE). This combined approach is capable to find more efficient design solutions, without sacrificing timing guarantees. For it, a first analytical DSE phase obtains a set of solutions compliant with the critical time constraints. Search of the optimum solution is done among this set, but it is delegated to a second simulation-based search, for fine tuning and average-case optimisation. The potential of our approach is illustrated by a proof-of-concept implementation of the proposed DSE flow and an accompanying DSE example.