An approach to optimise the energy efficiency of iterative computation on integrated GPU–CPU systems

详细信息    查看全文

• 作者：E. M. Garzón ; J. J. Moreno ; J. A. Martínez
• 关键词：Energy efficiency ; Heterogeneous processors ; Parallel iterative algorithms ; Integrated CPU–GPU
• 刊名：The Journal of Supercomputing
• 出版年：2017
• 出版时间：January 2017
• 年：2017
• 卷：73
• 期：1
• 页码：114-125
• 全文大小：
• 刊物类别：Computer Science
• 刊物主题：Programming Languages, Compilers, Interpreters; Processor Architectures; Computer Science, general;
• 出版者：Springer US
• ISSN：1573-0484
• 卷排序：73

文摘

Currently, the energy efficiency of computational systems is of paramount relevance. In this work, an approach for improving energy efficiency is proposed in the context of the iterative computation on integrated GPU-CPU systems. The proposal, referred to as E-ADITHE, combines iterative procedures with: (1) a heuristic scheme for processing units selection according to the estimation of energy efficiency and (2) the load balancing on heterogeneous processors. There is a wide variety of iterative algorithms related to science and engineering which can take advantage of E-ADITHE. The Beltrami filter has been selected as a representative example of such procedures and its OpenCL version has been used to validate E-ADITHE. The analysis of the results shows that E-ADITHE improves automatically the energy efficiency of parallel iterative algorithm on modern heterogeneous processors.