Abstract
Data value prediction has been widely accepted as an effective mechanism to break data hazards for high performance processor design. Several works have reported promising performance potential. However, there is hardly enough information that is presented in a clear way about performance comparison of these prediction mechanisms. This paper investigates the performance impact of four previously proposed value predictors, namely last value predictor, stride value predictor, two-level value predictor and hybrid (stride+two-level) predictor. The impact of misprediction penalty, which has been frequently ignored, is discussed in detail. Several other implementation issues, including instruction window size, issue width and branch predictor are also addressed and simulated. Simulation results indicate that data value predictors act differently under different configurations. In some cases, simpler schemes may be more beneficial than complicated ones. In some particular cases, value prediction may have negative impact on performance.
Similar content being viewed by others
References
Wu Y F, Chen D Y, Fang J. Better exploration of region-level value locality with integrated computation reuse and value prediction. In Proc. 28th Int. Symp. Computer Architecture, Göteborg, Sweden, Jul. 2001, pp.98–108.
Lipasti M H, Shen J P. Exceeding the data flow limit via value prediction. In Proc. 29th Int. Symp. Microarchitecture, Paris, France, Dec. 1996, pp.226–237.
Lipasti M H, Shen J. Exploiting value locality to exceed the data flow limit. International Journal of Parallel Programming, Aug. 1998, 28(4): 505–538.
Zhou H, Flanagan J, Conte T M. Detecting global stride locality in value streams. In Proc. 30th Int. Symp. Computer Architecture, San Diego, California, Jun. 2003, pp.324–335.
Lee S J, Yew P C. On some implementation issues for value prediction on wide-issue ILP processors. In Proc. 2000 Int. Conf. Parallel Architectures and Compilation Techniques, Philadelphia, Oct. 2000, pp.145–156.
Gonzalez J, Gonzalez A. The potential of data value speculation to boost ILP. In Proc. 12th Int. Conf. Supercomputing, Melbourne, Australia, Jul. 1998, pp.21–28.
Lipasti M H, Wilkerson C B, Shen J P. Value locality and load value prediction. In Proc. VIIth Int. Conf. Architectural Support for Programming Languages and Operating Systems, Cambridge, Massachusetts, Oct. 1996, pp.138–147.
Wang K, Franklin M. Highly accurate data value prediction using hybrid predictors. In Proc. 30th Int. Symp. Microarchitecture, Research Triangle Park, North Carolina, Dec. 1997, pp.281–290.
Rychlik B, Faitl J, Krug B, Shen J P. Efficacy and performance impact of value prediction. In Proc. 1998 Int. Conf. Parallel Architectures and Compilation Techniques, Paris, France, Oct. 1998, pp.148–157.
Sazeides Y. Modeling value speculation. In Proc. 8th Int. Symp. High Performance Computer Architecture, Boston, Massachusetts, Feb. 2002, pp.211–222.
Burger D C, Austin T M. The SimpleScalar tool set, version 2.0. Technical Report CSTR-97-1342, University of Wisconsin, Madison, Jun. 1997.
Lee S J. Data Value Predictors. http://www.simplescalar.com/
SPEC CPU2000 Benchmarks. http://www.spec.org/osg/cpu-2000/.
Gabbay F, Mendelson A. The effect of instruction fetch bandwidth on value prediction. In Proc. 25th Int. Symp. Computer Architecture, Barcelona, Spain, Jun. 1998, pp.272–281.
Lee S J, Wang Y, Yew P C. Decoupled value prediction on trace processors. In Proc. 6th Int. Symp. High Performance Computer Architecture, Toulouse, France, Jan. 2000, pp.231–240.
Calder B, Reinman G, Tullsen D. Selective value prediction. In Proc. 26th Int. Symp. Computer Architecture, Atlanta, Georgia, Jun.1999, pp.64–74.
Bhargava R, John L K. Performance and energy impact of instruction-level value predictor filtering. First Value-Prediction Workshop (VPW1) (held with ISCA'03), San Diego, California, Jun. 2003, pp.71–78.
Author information
Authors and Affiliations
Corresponding author
Additional information
Supported by the National Natural Science Foundation of China under Grant No. 90307001.
Xiao Yong is a Ph.D. candidate. His research interests include data value prediction and binary translation.
Xing-Ming Zhou is a professor and Ph.D. supervisor. He is a fellow of Chinese Academy of Sciences. His research interests include high performance computer architecture and mobile computing.
Rights and permissions
About this article
Cite this article
Xiao, Y., Zhou, XM. Performance Evaluation of Data Value Prediction Schemes. J Comput Sci Technol 20, 615–623 (2005). https://doi.org/10.1007/s11390-005-0615-y
Revised:
Issue Date:
DOI: https://doi.org/10.1007/s11390-005-0615-y