Quick performance evaluation for iterative optimization
Weiheng Zhu · Journal of Jinan University · 2011
Iterative optimization is an effective approach to achieve high performance on modern architectures.Various approaches have been developed to accelerate this time-consuming iterative optimization process.Most of them focus on how to quickly identify good points within the optimization space.An alternative approach to solve this problem is proposed,by testing multiple points in each execution,so that more points can be evaluated in a fixed amount of time.Experimental results are presented to demonstrate the feasibility of this statistic-based approach.It shows that only a few samples are needed to accurately evaluate the performance of a given point.