Year of Graduation
Applied Mathematics and Information Science
Program autotuning has been shown to achieve better performance in a number of domains. However, autotuners themselves are rarely universal. That is why everyone should adjust or write tuner by himself. It is a compilcated task. Good autotuner requires advanced machine learning techniques and powerful computing resources for fast optimization. This paper introduces a new universal cloud service for program autotuning. The service is based on Everest platform and OpenTuner framework for building autotuners. Service uses distributed computing on a cluster to improve performance and supports a wide-range of optimization techiniques which are provided by OpenTuner. The paper demostrates a process of development and examples of using the service.