Skip to main content

Experiences in autotuning matrix multiplication for energy minimization on GPUs

by Hartwig Anzt, Blake Haugen, Jakub Kurzak, P Luszczek, Jack J Dongarra
Publication Type
Journal Name
Concurrency and Computation: Practice and Experience
Publication Date
Page Numbers
5096 to 5113

In this paper, we report extensive results and analysis of autotuning the computationally intensive graphics processing units kernel for dense matrix-matrix multiplication in double precision. In contrast to traditional autotuning and/or optimization for runtime performance only, we also take the energy efficiency into account. For kernels achieving equal performance, we show significant differences in their energy balance. We also identify the memory throughput as the most influential metric that trades off performance and energy efficiency. As a result, the performance optimal case ends up not being the most efficient kernel in overall resource use. Copyright (c) 2015John Wiley & Sons, Ltd.