Efficient system design space exploration using machine learning techniques

Abstract

Computer manufacturers spend a huge amount of time, resources, and money in designing new systems and newer configurations, and their ability to reduce costs, charge competitive prices and gain market share depends on how good these systems perform. In this work, we develop predictive models for estimating the performance of systems by using performance numbers from only a small fraction of the overall design space. Specifically, we first develop three models, two based on artificial neural networks and another based on linear regression. Using these models, we analyze the published Standard Performance Evaluation Corporation (SPEC) benchmark results and show that by using the performance numbers of only 2% and 5% of the machines in the design space, we can estimate the performance of all the systems within 9.1% and 4.6% on average, respectively. Then, we show that the performance of future systems can be estimated with less than 2.2% error rate on average by using the data of systems from a previous year. We believe that these tools can accelerate the design space exploration significantly and aid in reducing the corresponding research/development cost and timeto-market.

On learning-based methods for design-space exploration with high-level synthesis

“…Learning-based methods were proposed to guide the DSE process by predicting solution qualities before running actual simulation/synthesis [1,8,10,11,17]. Compared with local-search techniques, learning-based methods can yield better solution quality as well as require shorter simulation/synthesis runtime.…”

“…Within this framework, two very recent papers [8,17] independently reported that Gaussian Process [17], a.k.a. Kriging [8], was the most promising learning model, superior to Artificial Neural Network [10,11] and other simple models. These results were obtained from DSE with processor simulators [8] or IP generators [17].…”

Boosting Design Space Explorations with Existing or Automatically Learned Knowledge

“…The link between the parameters of ANNs and understandable facts is typically hard to establish, hence they are not useful as interface for engineers to specify facts. Predictive models in a more general matter are also used by Ozisikyilmaz et al [20] to accelerate DSEs.…”

Towards Scalability and Reliability of Autonomic Systems on Chip

“…However, since the required processing power exceeds that available in the field, the generation and update of rules occurs offline. Other projects exist that use run time learning techniques such as supervised learning which require additional training information [10], or in which learning is used solely to optimize the system during design time [11]. The goal of ASoC is to provide both adaptivity and learning at run time, and to do so within the constraints of a typical system on chip design.…”