Energy saving of value prediction by expanding branch target buffer

被引:0
作者
Shimomura, Yoshio [1 ]
Kobayashi, Ryotaro [1 ]
机构
[1] Toyohashi University of Technology, Toyohashi 441-8580, 1-1, Hibarigaoka, Tenpaku-cho
关键词
BTB; Reference reduction; Stride; Value prediction;
D O I
10.1541/ieejeiss.132.1706
中图分类号
学科分类号
摘要
Recent advances in computer design allow microprocessors to enable parallelism in programs in order to achieve high performance. True dependency is one of the many factors that restrict parallelism in a computer program. To overcome the restrictions of true dependency, a number of schemes for value prediction have been proposed. The value predictor predicts an instruction result on the basis of the value history obtained from Value History Table (VHT). However, it consumes considerable energy because VHT is large and is referenced very frequently. In this paper, we propose a more efficient mechanism for a value predictor that extends the use of an existing branch target buffer (BTB) to reduce the number of invalid VHT references. We introduced a predictability bit (p bit) to identify an instruction that has a predictable value, and we added a field to BTB to ensure that the p-bits corresponded to successive instructions. The proposed mechanism controls VHT references on the basis of p bits. The evaluation results show that the proposed mechanism reduces invalid references by 27.6% and 42.4% and energy by 16.6% and 25.7% with 0.1% performance loss on average in the 8 and 32 p bits length, respectively. © 2012 The Institute of Electrical Engineers of Japan.
引用
收藏
页码:1706 / 1718+21
相关论文
共 21 条
[1]  
Akkary H., Driscoll M.A., A dynamic multithreading processor, Proc. MICRO-31, pp. 226-236, (1998)
[2]  
Burger D., Austin T.M., The simplescalar tool set version 2.0, ACM SIGARCH Computer Architecture News, 25, 3, (1997)
[3]  
Burtscher M., Zorn B.G., Hybridizing and Coalescing Load Value Predictor, (2000)
[4]  
Chen T.F., Baer J.L., Effective hardware-based data prefetching for high performance processors, IEEE T COM-PUT, 44, 5, pp. 609-623, (1995)
[5]  
Gabbay F., Mendelson A., Speculative Execution Based on Value Prediction, (1996)
[6]  
Gonzalez Jose, Gonzalez Antonio, Control-flow speculation through value prediction for superscalar processors, Parallel Architectures and Compilation Techniques - Conference Proceedings, PACT, pp. 57-65, (1999)
[7]  
Ketterlin A., Clauss Ph., Prediction and trace compression of data access addresses through nested loop recognition, Proceedings of the 2008 CGO - Sixth International Symposium on Code Generation and Optimization, pp. 94-103, (2008)
[8]  
Koushiro T., Sato T., Arita I., A trace-level value predictor for contrail processors, ACM SIGARCH Computer Architecture News, 31, 3, pp. 42-47, (2003)
[9]  
Lipasti M.H., Wilkerson C.B., Shen J.P., Value locality and load value prediction, Proc. ASPLOS-7, pp. 138-147, (1996)
[10]  
Lipasti M.H., Shen P.J., Exceeding the dataflow limit via value prediction, Proc. MICRO-29, pp. 226-237, (1996)