Investigating the Effects of Hardware Parameters on Power Consumptions in SPMV Algorithms on Graphics Processing Units (GPUs)

Although Sparse matrix-vector multiplication (SPMVs) algorithms are simple, they include important parts of Linear Algebra algorithms in Mathematics and Physics areas. As these algorithms can be run in parallel, Graphics Processing Units (GPUs) has been considered as one of the best candidates to run these algorithms. In the recent years, power consumption has been considered as one of the metrics that should be taken into consideration in addition to performance.  In spite of this importance, to the best of our knowledge, studies on power consumptions in SPMVs algorithms on GPUs are scarce.  In this paper, we investigate the effects of hardware parameters on power consumptions in SPMV algorithms on GPUs. For this, we leverage the possibility of setting the GPU’s parameters to investigate the effects of these parameters on power consumptions. These configurations have been applied to different formats of Sparse Matrices, and the best parameters are selected for having the best performance per power metric. Therefore, as the results of this study the settings can be applied in running different Linear Algebra algorithms on GPUs to obtain the best performance per power.