Statistical Models for Automatic Performance Tuning
- Cite this paper as:
- Vuduc R., Demmel J.W., Bilmes J. (2001) Statistical Models for Automatic Performance Tuning. In: Alexandrov V.N., Dongarra J.J., Juliano B.A., Renner R.S., Tan C.J.K. (eds) Computational Science — ICCS 2001. ICCS 2001. Lecture Notes in Computer Science, vol 2073. Springer, Berlin, Heidelberg
Achieving peak performance from library subroutines usually requires extensive, machine-dependent tuning by hand. Automatic tuning systems have emerged in response, and they typically operate, at compile-time, by (1) generating a large number of possible implementations of a subroutine, and (2) selecting a fast implementation by an exhaustive, empirical search. This paper applies statistical techniques to exploit the large amount of performance data collected during the search. First, we develop a heuristic for stopping an exhaustive compile-time search early if a near-optimal implementation is found. Second, we show how to construct run-time decision rules, based on run-time inputs, for selecting from among a subset of the best implementations. We apply our methods to actual performance data collected by the PHiPAC tuning system for matrix multiply on a variety of hardware platforms.
Unable to display preview. Download preview PDF.