The documents distributed by this server have been provided by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a noncommercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.
Publications of SPCL
|Performance Modeling for the Masses|
(Presentation - presented in Seattle, WA, USA, Nov. 2011)
AbstractPerformance modeling of parallel application has long been used only by computer scientists. We introduce intuitive motivations and techniques for performance modeling that are targeted at developers and users of scientific codes. We promote simple application models that can be mapped to algorithms and source-code constructs and guide the tuning and optimization process. Furthermore, we advocate a model-supported software development strategy where performance models are used from the very beginning throughout the whole process to monitor and tune performance of code blocks, kernels, and whole applications. We demonstrate a first model that was developed for the Blue Waters project.