A Multi-Dimensional Micro-benchmark for Performance Study and Prediction
In this paper, we describe a multi-dimensional synthetic benchmark called Apex-MAP. One distinguishing feature of this benchmark is that its input parameters can be varied independent of each other between extreme values, allowing it to generate continuous performance surfaces for performance and architecture study. Moreover, Apex-MAP can be used as performance proxy for applications to provide realistic and reliable performance indicator on various target platforms.
Hongzhang Shan Erich Strohmaier
Future Technology Group, Lawrence Berkeley National Laboratory Berkeley, CA 94720
国际会议
黄山
英文
156-160
2010-05-28(万方平台首次上网日期,不代表论文的发表时间)