Boosting Design Space Explorations with Existing or Automatically Learned Knowledge

Abstract

Abstract. During development, processor architectures can be tuned and configured by many different parameters. For benchmarking, automatic design space explorations (DSEs) with heuristic algorithms are a helpful approach to find the best settings for these parameters according to multiple objectives, e.g. performance, energy consumption, or real-time constraints. But if the setup is slightly changed and a new DSE has to be performed, it will start from scratch, resulting in very long evaluation times.To reduce the evaluation times we extend the NSGA-II algorithm in this article, such that automatic DSEs can be supported with a set of transformation rules defined in a highly readable format, the fuzzy control language (FCL). Rules can be specified by an engineer, thereby representing existing knowledge. Beyond this, a decision tree classifying high-quality configurations can be constructed automatically and translated into transformation rules. These can also be seen as very valuable result of a DSE because they allow drawing conclusions on the influence of parameters and describe regions of the design space with high density of good configurations.Our evaluations show that automatically generated decision trees can classify near optimal configurations for the hardware parameters of the Grid ALU Processor (GAP) and M-Sim 2. Further evaluations show that automatically constructed transformation rules can reduce the number of evaluations required to reach the same quality of results as without rules by 43%, leading to a significant saving of time of about 25%. In the demonstrated example using rules also leads to better results.

A Comparison of Multi-objective Algorithms for the Automatic Design Space Exploration of a Superscalar System

“…L2 cache must be larger than L1 cache) to more complex ones. More details about FADSE can be found in [21] [22].…”

Multi-objective optimisations for a superscalar architecture with selective value prediction

“…We can conclude that imposing a high probability of the rules will reduce diversity, especially with a small number of rules. In our previous work more rules were used and the membership functions had many intervals (associated linguistic terms) [28]. In this situation the runs with Gaussian probability provided better results.…”

DeSpErate++: An Enhanced Design Space Exploration Framework Using Predictive Simulation Scheduling

“…In the context of the DSE phase, researchers proposed to better tackle the problem with the usage of approximate system models. These models are analytic approximations of system performance that are learned after an initial training phase [36] and used to determine where to focus the costly evaluations [5], [6], [37]. Widely used methods are linear regressions [38], radial basis functions [5], [39], neural networks [40], regression trees [28], [41], and statistical techniques such as Gaussian processes [42] and Kriging interpolation [11], [43].…”

“…In this paper, we demonstrate that state-of-the-art analytic performance prediction techniques such as [5], [6], and [11]- [13] do not represent the best DSE approach when a parallel computing system (e.g., a multicore processor or a computer cluster) is exploited to run concurrently different simulations. To clarify this idea, Fig.…”

“…To reduce the simulation time, a solution is to use analytic models to predict the simulation results [5]. Once analytic models are trained to fit the behavior of some observed simulations, they can be used to prune the design space by focusing the exploration on the most promising design points [6].…”