Can search algorithms save large-scale automatic performance tuning?

被引:20
作者
Balaprakash, Prasanna [1 ]
Wild, Stefan M. [1 ]
Hovland, Paul D. [1 ]
机构
[1] Argonne Natl Lab, Div Math & Comp Sci, Argonne, IL 60439 USA
来源
PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON COMPUTATIONAL SCIENCE (ICCS) | 2011年 / 4卷
关键词
autotuning; empirical tuning; optimization; performance-tuning; OPTIMIZATION;
D O I
10.1016/j.procs.2011.04.234
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
Empirical performance optimization of computer codes using autotuners has received significant attention in recent years. Given the increased complexity of computer architectures and scientific codes, evaluating all possible code variants is prohibitively expensive for all but the simplest kernels. One way for autotuners to overcome this hurdle is through use of a search algorithm that finds high-performing code variants while examining relatively few variants. In this paper we examine the search problem in autotuning from a mathematical optimization perspective. As an illustration of the power and limitations of this optimization, we conduct an experimental study of several optimization algorithms on a number of linear algebra kernel codes. We find that the algorithms considered obtain performance gains similar to the optimal ones found by complete enumeration or by large random searches but in a tiny fraction of the computation time.
引用
收藏
页码:2136 / 2145
页数:10
相关论文
共 22 条
  • [1] [Anonymous], 1998, SC 98, DOI [10.5555/509058.509096, DOI 10.1109/SC.1998.10004]
  • [2] [Anonymous], 2010, SOFTWARE AUTOMATIC T
  • [3] [Anonymous], P 23 IEEE INT PAR DI
  • [4] [Anonymous], [No title captured]
  • [5] Finding optimal algorithmic parameters using derivative-free optimization
    Audet, Charles
    Orban, Dominique
    [J]. SIAM JOURNAL ON OPTIMIZATION, 2006, 17 (03) : 642 - 664
  • [6] Conn A. R., 2009, MPS-SIAM Series on Optimization
  • [7] Optimization and Performance Modeling of Stencil Computations on Modern Microprocessors
    Datta, Kaushik
    Kamil, Shoaib
    Williams, Samuel
    Oliker, Leonid
    Shalf, John
    Yelick, Katherine
    [J]. SIAM REVIEW, 2009, 51 (01) : 129 - 159
  • [8] GENETIC ALGORITHMS FOR ESTIMATION PROBLEMS WITH MULTIPLE OPTIMA, NONDIFFERENTIABILITY, AND OTHER IRREGULAR FEATURES
    DORSEY, RE
    MAYER, WJ
    [J]. JOURNAL OF BUSINESS & ECONOMIC STATISTICS, 1995, 13 (01) : 53 - 66
  • [9] Fursin G.G., 2002, LANGUAGES COMPILERS, P305
  • [10] Gordy M., 1996, GAM MATLAB ROUTINE F