Effect of auto-tuning with user's knowledge for numerical software

ABSTRACT

This paper evaluates the effect of an auto-tuning facility with the user's knowledge for numerical software. We proposed a new software architecture framework, named FIBER, to generalize auto-tuning facilities and obtain highly accurate estimated parameters. The FIBER framework also provides a loop-unrolling function and an algorithm selection function to support code development by library developers needing code generation and parameter registration processes. FIBER offers three kinds of parameter optimization layers---install-time, before execute-time, and run-time. The user's knowledge is needed in the before execute-time optimization layer. In this paper, eigensolver parameters that apply the FIBER framework are described and evaluated in three kinds of parallel computers: the HITACHI SR8000/MPP, Fujitsu VPP800/63, and Pentium4 PC cluster. Our evaluation of the application of the before execute-time layer indicated a maximum speed increase of 3.4 times for eigensolver parameters, and a maximum increase of 17.1 times for the algorithm selection of orthogonalization in the computation kernel of the eigensolver.

REFERENCES

• 1: ATLAS project; http://www.netlib.org/atlas/index.html.
• 2: Jeff Bilmes , Krste Asanovic , Chee-Whye Chin , Jim Demmel, Optimizing matrix multiply using PHiPAC: a portable, high-performance, ANSI C coding methodology, Proceedings of the 11th international conference on Supercomputing, p.340-347, July 07-11, 1997, Vienna, Austria [doi>10.1145/263580.263662]
• 3: J. Dongarra and V. Eijkhout. Self-adapting numerical software for next generation applications. The International Journal of High Performance Computing and Applications, 17(2):125--131, 2003.
• 4: Matteo Frigo, A fast Fourier transform compiler, Proceedings of the ACM SIGPLAN 1999 conference on Programming language design and implementation, p.169-180, May 01-04, 1999, Atlanta, Georgia, United States
• 5: T. Imamura and K. Naono. An evaluation towards an automatic tuning eigensolver with performance stability. In Proceedings of Symposium on Advanced Computing Systems and Infrastructures (SACSIS)2003, pages 145--152, 2003.
• 6: T. Katagiri, H. Kuroda, K. Ohsawa, M. Kudoh, and Y. Kanada. Impact of auto-tuning facilities for parallel numerical library. IPSJ Transaction on High Performance Computing Systems, 42(SIG 12 (HPS 4)):60--76, 2001.
• 7: Hisayasu Kuroda , Takahiro Katagiri , Yasumasa Kanada, Knowledge Discovery in Auto-tuning Parallel Numerical Library, Progress in Discovery Science, Final Report of the Japanese Discovery Science Project, p.628-639, January 2002
• 8: K. Naono and Y. Yamamoto. A framework for development of the library for massively parallel processors with auto-tuning function and with the single memory interface. IPSJ SIG Notes, (2001-HPC-87):25--30, 2001.
• 9: Randy L. Ribler , Huseyin Simitci , Daniel A. Reed, The Autopilot performance-directed adaptive control system, Future Generation Computer Systems, v.18 n.1, p.175-187, September 2001 [doi>10.1016/S0167-739X(01)00051-6]
• 10: K. Takahiro, K. Kise, H. Honda, and T. Yuba. FIBER: A general framework for auto-tuning software. Proceedings of The Fifth International Symposium on High Performance Computing, Springer Lecture Notes in Computer Science(2858):146--159, 2003.
• 11: K. Takahiro, K. Kise, H. Honda, and T. Yuba. FIBER: A framework of installation, before execution-invocation, and run-time optimization layers for auto-tuning software. IS Technical Report, Graduate School of Information Systems, The University of Electro-Communications, UEC-IS-2003-3, May 2003.
• 13: R. Whaley, A. Petitet, and J. J. Dongarra. Automated empirical optimizations of software and the ATLAS project. Parallel Computing, 27:3--35, 2001.

INDEX TERMS

Primary Classification:
  G. Mathematics of Computing
  G.4 MATHEMATICAL SOFTWARE

Additional Classification:
  D. Software
  D.2 SOFTWARE ENGINEERING
      D.2.2 Design Tools and Techniques
          Subjects: Software libraries

General Terms:
Performance

Keywords:
auto-tuning, eigensolver, numerical library, parameter optimization, performance modeling