@inproceedings{5360776f5e7d4f8588aebe05b17d75af,
title = "Tuning OpenCL applications with the periscope tuning framework",
abstract = "Due to the complexity and diversity of new parallel architectures automatic tuning of parallel applications has become increasingly important for achieving acceptable performance levels as well as performance portability. The European AutoTune project developed a tuning framework which closely integrates and automates performance analysis and performance tuning. The Periscope Tuning Framework relies on a flexible plugin mechanism providing tuning plugins for different tuning aspects. This paper presents plugins for tuning the execution time of OpenCL kernels on three different architectures, namely standard multicore CPUs, Xeon Phi coprocessors, and GPUs. We present OpenCL tuning via the flags used during offline kernel compilation as well as through the selection of the most appropriate NDRange configuration, which defines the organization of parallel threads used for kernel execution. Both tuning plugins show significant performance impact and a clear dependence on the target architecture and thus improve performance portability via automatic tuning.",
author = "Enes Bajrovic and Robert Mijakovic and Jiri Dokulil and Siegfried Benkner and Michael Gerndt",
note = "Publisher Copyright: {\textcopyright} 2016 IEEE.; 49th Annual Hawaii International Conference on System Sciences, HICSS 2016 ; Conference date: 05-01-2016 Through 08-01-2016",
year = "2016",
month = mar,
day = "7",
doi = "10.1109/HICSS.2016.711",
language = "English",
series = "Proceedings of the Annual Hawaii International Conference on System Sciences",
publisher = "IEEE Computer Society",
pages = "5752--5761",
editor = "Sprague, {Ralph H.} and Bui, {Tung X.}",
booktitle = "Proceedings of the 49th Annual Hawaii International Conference on System Sciences, HICSS 2016",
}