Finding Inefficiencies in OpenMP Applications Automatically with Periscope

  • Karl Fürlinger
  • Michael Gerndt
Part of the Lecture Notes in Computer Science book series (LNCS, volume 3992)


Performance optimization of parallel programs can be a time-consuming and difficult task. Therefore, tools are desirable that help application developers by automatically locating inefficiencies. We present Periscope, a system for automated performance analysis based on the notion of performance properties.

We present the overall architecture of Periscope, which consists of a set of analysis agents and show how properties of OpenMP applications are detected. We describe the set of OpenMP properties we have defined so far and the data model used in the specification of these properties. Practical tests on the feasibility of our approach are performed with a number of OpenMP applications.


Performance Property Target Application Virtual Topology Analysis Agent Intermediate Agent 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


  1. 1.
    Fahringer, T., Gerndt, M., Mohr, B., Wolf, F., Riley, G., Träff, J.L.: Knowledge specification for automatic performance analysis. APART technical report, revised edition. Technical Report FZJ-ZAM-IB-2001-08, Forschungszentrum Jülich (2001)Google Scholar
  2. 2.
    Fürlinger, K., Gerndt, M.: Distributed application monitoring for clustered SMP architectures. In: Kosch, H., Böszörményi, L., Hellwagner, H. (eds.) Proceedings of the 9th International Euro-Par Conference on Parallel Processing, August 2003, pp. 127–134. Springer, Heidelberg (2003)Google Scholar
  3. 3.
    Fürlinger, K., Gerndt, M.: Performance analysis of shared-memory parallel applications using performance properties. In: Proceedings of the 2005 International Conference on High Performance Computing and Communications (HPCC-2005), Sorrento, Italy, September 2005, pp. 595–604 (2005) (Accepted for publication)Google Scholar
  4. 4.
    Miller, B.P., Callaghan, M.D., Cargille, J.M., Hollingsworth, J.K., Irvin, R.B., Karavanic, K.L., Kunchithapadam, K., Newhall, T.: The Paradyn parallel performance measurement tool. IEEE Computer 28(11), 37–46 (1995)Google Scholar
  5. 5.
    Mohr, B., Malony, A.D., Hoppe, H.-C., Schlimbach, F., Haab, G., Hoeflinger, J., Shah, S.: A performance monitoring interface for OpenMP. In: Proceedings of the Fourth Workshop on OpenMP (EWOMP 2002) (September 2002)Google Scholar
  6. 6.
    Mohr, B., Malony, A.D., Shende, S.S., Wolf, F.: Towards a performance tool interface for OpenMP: An approach based on directive rewriting. In: Proceedings of the Third Workshop on OpenMP (EWOMP 2001) (September 2001)Google Scholar
  7. 7.
    Roth, P.C., Arnold, D.C., Miller, B.P.: MRNet: A software-based multicast reduction network for scalable tools. In: Proceedings of the 2003 Conference on Supercomputing (SC 2003) (November 2003)Google Scholar
  8. 8.
    Roth, P.C., Miller, B.P.: The distributed performance consultant and the sub-graph folding algorithm: On-line automated performance diagnosis on thousands of processes. In: Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming (PPoPP 2006) (March 2005) (Accepted for Publication)Google Scholar
  9. 9.
    Wolf, F., Mohr, B.: Automatic Performance Analysis of Hybrid MPI/OpenMP Applications. In: Proceedings of the 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing (PDP 2003), February 2003, pp. 13–22. IEEE Computer Society Press, Los Alamitos (2003)CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2006

Authors and Affiliations

  • Karl Fürlinger
    • 1
  • Michael Gerndt
    • 1
  1. 1.Institut für InformatikLehrstuhl für Rechnertechnik und Rechnerorganisation Technische Universität MünchenGermany

Personalised recommendations