Automatic Communication Performance Debugging in PGAS Languages

  • Jimmy Su
  • Katherine Yelick
Part of the Lecture Notes in Computer Science book series (LNCS, volume 5234)

Abstract

Recent studies have shown that programming in a Partition Global Address Space (PGAS) language can be more productive than programming in a message passing model. One reason for this is the ability to access remote memory implicitly through shared memory reads and writes. But this benefit does not come without a cost. It is very difficult to spot communication by looking at the program text, since remote reads and writes look exactly the same as local reads and writes. This makes manual communication performance debugging an arduous task. In this paper, we describe a tool called ti-trend-prof that can do automatic performance debugging using only program traces from small processor configurations and small input sizes in Titanium [13], a PGAS language. ti-trend-prof presents trends to the programmer to help spot possible communication performance bugs even for processor configurations and input sizes that have not been run. We used ti-trend-prof on two of the largest Titanium applications and found bugs that would have taken days in under an hour.

Keywords

PGAS languages automatic performance debugging 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Alexandrov, A., Ionescu, M.F., Schauser, K.E., Scheiman, C.: LogGP: Incorporating long messages into the LogP model. Journal of Parallel and Distributed Computing 44(1), 71–79 (1997)CrossRefGoogle Scholar
  2. 2.
    Bonachea, D.: GASNet specifications (2003)Google Scholar
  3. 3.
    Brewer, E.A.: High-level optimization via automated statistical modeling. In: PPOPP 1995: Proceedings of the 5th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, pp. 80–91. ACM Press, New York (1995)CrossRefGoogle Scholar
  4. 4.
    Coarfa, C., Mellor-Crummey, J., Froyd, N., Dotsenko, Y.: Scalability Analysis of SPMD Codes Using Expectations. In: PPOPP (2007)Google Scholar
  5. 5.
    Culler, D.E., Karp, R.M., Patterson, D.A., Sahay, A., Schauser, K.E., Santos, E., Subramonian, R., von Eicken, T.: LogP: Towards a realistic model of parallel computation. In: Proceedings 4th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, pp. 1–12 (1993)Google Scholar
  6. 6.
    Givelberg, E., Yelick, K.: Distributed Immersed Boundary Simulation in Titanium (2004)Google Scholar
  7. 7.
    Goldsmith, S., Aiken, A., Wilkerson, D.: Measuring Empirical Computational Complexity. Foundations of Software Engineering (2007)Google Scholar
  8. 8.
    Graham, S.L., Kessler, P.B., Mckusick, M.K.: Gprof: A call graph execution profiler. In: SIGPLAN 1982: Proceedings of the 1982 SIGPLAN Symposium on Compiler Construction, pp. 120–126. ACM Press, New York (1982)CrossRefGoogle Scholar
  9. 9.
    Kluge, M., Knüpfer, A., Nagel, W.E.: Knowledge based automatic scalability analysis and extrapolation for MPI programs. In: Cunha, J.C., Medeiros, P.D. (eds.) Euro-Par 2005. LNCS, vol. 3648. Springer, Heidelberg (2005)Google Scholar
  10. 10.
    Rugina, R., Schauser, K.: Predicting the running times of parallel programs by simulation. In: Proceedings of the 12th International Parallel Processing Symposium and 9th Symposium on Parallel and Distributed Processing (1998)Google Scholar
  11. 11.
    Vetter, J., Worley, P.: Asserting performance expectations. In: SC (2002)Google Scholar
  12. 12.
    Wen, T., Colella, P.: Adaptive Mesh Refinement in Titanium. In: IPDPS (2005)Google Scholar
  13. 13.
    Yelick, K., Semenzato, L., Pike, G., Miyamoto, C., Liblit, B., Krishnamurthy, A., Hilfinger, P., Graham, S., Gay, D., Colella, P., Aiken, A.: Titanium: A high-performance Java dialect. In: Workshop on Java for High-Performance Network Computing (1998)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2008

Authors and Affiliations

  • Jimmy Su
    • 1
  • Katherine Yelick
    • 1
    • 2
  1. 1.Computer Science DivisionUniversity of California at Berkeley 
  2. 2.Lawrence Berkeley National Laboratory 

Personalised recommendations