An efficient joint analytical and simulation-based design space exploration flow for predictable multi-core systems
2015 (English)In: ACM International Conference Proceeding Series, ACM Digital Library, 2015Conference paper (Refereed)
Recent work has proposed two-phase joint analytical and simulation-based design space exploration (JAS-DSE) approaches. In such approaches, a first analytical phase relies on static performance estimation and either on exhaustive or heuristic search, to perform a very fast filtering of the design space. Then, a second phase obtains the Pareto solutions after an exhaustive simulation of the solutions found as compliant by the analytical phase. However, the capability of such approaches to find solutions close to the actual Pareto set at a reasonable time cost is compromised by current system complexities. This limitation is due to the fact that such approaches do not support an heuristic exploration on the simulation-based phase. It is not straightforward because in the second phase the heuristic is constrained to consider only the custom set of solutions found in the first phase. This set is in general unconnected and irregularly distributed, which prevents the application of existing heuristics. This paper provides as a solution a novel search heuristic called ARS (Adaptive Random Sampling). The ARS strategy enables the application of heuristic search in the two phases of the JAS-DSE flow, by enabling the application of heuristic in the second phase, regardless the type of performance estimation done at each phase. Moreover, it enables the definition of N-phase DSE flows. The paper shows on an experiment focused on predictable multi-core systems how this enhanced JAS-DSE is capable to find more efficient solutions and to tune the trade-off between exploration time and accuracy in finding actual Pareto solutions.
Place, publisher, year, edition, pages
ACM Digital Library, 2015.
Design space exploration, Electronic system-level design, Predictable systems, Design, Economic and social effects, Heuristic algorithms, Modular robots, System theory, Electronic system level design, Exhaustive simulation, Multi-core systems, Performance estimation, Search heuristics, Simulation-based designs, Static performance estimation, Systems analysis
Electrical Engineering, Electronic Engineering, Information Engineering
IdentifiersURN: urn:nbn:se:kth:diva-194641DOI: 10.1145/2693433.2693435ScopusID: 2-s2.0-84984972726OAI: oai:DiVA.org:kth-194641DiVA: diva2:1050405
2015 Workshop on Rapid Simulation and Performance Evaluation: Methods and Tools, RAPIDO 2015, 19 January 2015 through 21 January 2015
Funding Details: FP7 611146, EC, European Commission
QC 201611292016-11-292016-10-312016-11-29Bibliographically approved