@article {982, title = {Experiences in Autotuning Matrix Multiplication for Energy Minimization on GPUs}, journal = {Concurrency and Computation: Practice and Experience}, volume = {27}, year = {2015}, month = {12-Oct}, pages = {5096 - 5113}, abstract = {In this paper, we report extensive results and analysis of autotuning the computationally intensive graphics processing units kernel for dense matrix{\textendash}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.}, keywords = {Autotuning, energy efficiency, hardware accelerators, matrix multiplication, power}, doi = {10.1002/cpe.3516}, url = {http://doi.wiley.com/10.1002/cpe.3516https://api.wiley.com/onlinelibrary/tdm/v1/articles/10.1002\%2Fcpe.3516}, author = {Hartwig Anzt and Blake Haugen and Jakub Kurzak and Piotr Luszczek and Jack Dongarra} }