|
Vol: 47(61) No: 2 / June 2002 Understanding Value Prediction through Complex Simulations Adrian Florea University"L. Blaga", Department of Computer Science, Str. E. Cioran, No. 4, Sibiu-2400, ROMANIA, phone: ++40-269-212716, e-mail: aflorea@vectra.ulbsibiu.ro Lucian N. Vintan University "L. Blaga", Department of Computer Science, Str. E. Cioran, No. 4, Sibiu-2400, ROMANIA, e-mail: vintan@jupiter.ulbsibiu.ro Dorin Sima University "L. Blaga", Department of Computer Science, Str. E. Cioran, No. 4, Sibiu-2400, ROMANIA Keywords: advanced computer architectures, value locality, value prediction, pipeline, simulation, benchmarking. Abstract Through this paper we investigated the value locality degree present in real-worlds program, and extended the value locality concept to all general-purpose registers (MIPS architecture). Also we exploited this concept through a prediction mechanism of load instruction values from two distinct perspectives: producer-centric and memory-centric. To extract the maximum degree of load value locality we performed execution driven simulations on SPEC95 benchmarks with unbounded table sizes. The encouraging obtained results facilitate implementation of much simpler prediction structures (at most 32/64 locations), reducing the hardware cost and complexity. References [1] Sodani A., Sohi G. - “Dynamic Instruction Reuse”, Proceedings of the 24th International Symposium on Computer Architecture, pp. 194-205, June 1997. [2] Wang K., Franklin M. - “Highly Accurate Data Value Prediction using Hybrid Predictors”, Proceedings of the 30th Annual ACM/IEEE International Symposium on Microarchitecture, December 1997. [3] Lepak K., Lipasti M. - “On the Value Locality of Store Instructions”, Proceedings of the 27th Annual International Symposium on Computer Architecture, Vancouver, June 2000. [4] Lipasti M., Wilkerson C.B., Shen J.P. - “Value Locality and Load Value Prediction”, Proceedings of the 7th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS-VII), pp. 138-147, October 1996. [5] Lipasti M., Shen J.P. - “Exceeding the Dataflow Limit via Value Prediction”, Proceedings of the 29th Annual ACM/IEEE International Symposium on Microarchitecture, December 1996. [6] Sazeides Y., Smith J.E. - “The Predictability of Data Values”, Proceedings of Micro-30, December 1-3, 1997 in Research Triangle Park, North Carolina. [7] Larus J. - \"SPIM S20: A MIPS R2000 Simulator\", Morgan Kaufmann Publishers, 1993. [8] Burger D., Austin T.M., Bennet S. - \"Evaluating future microprocessors: The SimpleScalar tool set\" - Tech. Rep. CS-TR-96-1308, University of Wisconsin - Madison, July 1996. [9] http://www.spec.org [10] Deswet V., Goeman B., Bosschere K. – Independent hashing as confidence mechanism for value predictors in microprocessors, Int’l Conf. EuroPar, Augsburg, Germany, 2002 [11] Sazeides Y.T. – An Analysis of Value Predictibality and its Application to a Superscalar Processor, PhD Thesis, University of Wisconsin-Madison, SUA, 1999 [12] Vintan L. - Arhitecturi de procesoare cu paralelism la nivelul instrucţiunilor, Editura Academiei Române, Bucureşti, (264 pg.), ISBN 973-27-0734-8, 2000 [13] Vintan L. - Predictia valorilor instructiunilor, NET-Report nr. 115, pg.11-17, ISSN 1582-4497, aprilie 2002. |