Title: Multi-objective optimization of HPC kernels for performance, power, and energy
Authors: Prasanna Balaprakash, Ananta Tiwari, Stefan Wild
Abstract: Code optimization in the high-performance computing realm has traditionally focused on reducing execution time. The problem, in mathematical terms, has been expressed as a single-objective optimization problem. The expected concerns of next-generation systems, however, demand a more detailed analysis of the interplay among execution time and other metrics. Metrics such as power, performance, energy, and resiliency may all be targeted together and traded against one another. We present a multi-objective formulation of the code optimization problem. Our proposed framework helps one explore potential tradeoffs among multiple objectives and provides a significantly richer analysis than can be achieved by treating additional metrics as hard constraints. We empirically examine a variety of metrics, architectures, and code optimization decisions and provide evidence that such tradeoffs exist in practice.
Keywords: empirical tuning; multi-objective optimization; performance-tuning; energy, time, power tradeoffs
Thanks: Support for this work was provided through the Scientific Discovery through Advanced Computing (SciDAC) program funded by the U.S. Department of Energy, Office of Science, Advanced Scientific Computing Research, under Contract No. DE-AC02-06CH11357. This research used the computational resources of the Argonne Leadership Computing Facility under a Director's discretionary allocation. We thank Paul Hovland for valuable discussions and Kazutomo Yoshii for insights on the power monitoring systems of BG/Q and Intel Xeon Phi and valuable feedback.
Status: To appear in Proceedings of the 4th International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (PMBS13), November 2013.
Links: [PDF available from ANL tech reports],
BibTeX:
@inproceedings{Multiobj13,
	author = {Prasanna Balaprakash and Ananta Tiwari and Stefan M. Wild}, 
	booktitle = {Proceedings of the 4th International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (PMBS13)},
	month = {November},
	year = {2013},
	title = {Multi-Objective Optimization of {HPC} Kernels for Performance, Power, and Energy},
        note = {To appear, available at \url{http://www.mcs.anl.gov/papers/P4069-0413.pdf}}
}
	
Back to Stefan Wild's homepage