When prefetching improves/degrades performance

ABSTRACT

We formulate a new method for evaluating any prefetching algorithm (real or hypothetical). This method allows researchers to analyze the potential improvements prefetching can bring to an application independent of any known prefetching algorithm. We characterize prefetching with the metrics: timeliness, coverage, and accuracy. We demonstrate the usefulness of this method using a Markov prefetch algorithm. Under ideal conditions, prefetching can remove nearly all of the pipeline stalls associated with a cache miss. However, in today's processors, we show that nearly all of the performance benefits derived from prefetching are eroded and, in many cases, prefetching loses performance. We do quantitative analysis of these trade-offs, and show that there are linear relationships between overall performance and coverage, accuracy, and bandwidth

REFERENCES

• 1: Alan David Berenbaum, Tor E. Jeremiassem, "History-Based Prefetch Cache Including A Time Queue", U.S. Patent 5,778,435 July 1998
• 2: David Callahan , Ken Kennedy , Allan Porterfield, Software prefetching, Proceedings of the fourth international conference on Architectural support for programming languages and operating systems, p.40-52, April 08-11, 1991, Santa Clara, California, United States
• 3: Mary Jay Charney, Correlation-based hardware prefetching, Cornell University, Ithaca, NY, 1996
• 4: T.-F. Chen , J.-L. Baer, A performance study of software and hardware data prefetching schemes, Proceedings of the 21ST annual international symposium on Computer architecture, p.223-232, April 18-21, 1994, Chicago, Illinois, United States
• 5: Jamison Collins , Suleyman Sair , Brad Calder , Dean M. Tullsen, Pointer cache assisted prefetching, Proceedings of the 35th annual ACM/IEEE international symposium on Microarchitecture, November 18-22, 2002, Istanbul, Turkey
• 6: Robert Cooksey , Stephan Jourdan , Dirk Grunwald, A stateless, content-directed data prefetching mechanism, Proceedings of the 10th international conference on Architectural support for programming languages and operating systems, October 05-09, 2002, San Jose, California
• 7: P. G. Emma, Understanding some simple processor-performance limits, IBM Journal of Research and Development, v.41 n.3, p.215-232, May 1997

INDEX TERMS

Primary Classification:
  C. Computer Systems Organization
  C.4 PERFORMANCE OF SYSTEMS
      Subjects: Design studies

Additional Classification:
  C. Computer Systems Organization
  C.4 PERFORMANCE OF SYSTEMS
      Subjects: Measurement techniques; Modeling techniques; Performance attributes

General Terms:
Algorithms, Measurement, Performance

Keywords:
accuracy, cache, coverage, prefetch, prefetching algorithm, timeliness