基于置信度评估的预测路径数据自适应存储策略.docVIP

面向微处理器猜测执行过程中预载入数据的 Cache污染控制方法 张骏（西安交通大学电学院， 陕西 西安710049）Zhangjun2008@ 摘 要：“存储墙”问题已经成为处理器性能提升的主要障碍，而处理器内核猜测执行预测路径上访存指令所导致Cache污染会严重影响处理器性能。本文提出一种针对数据的CSDA。首先，利用置信度评估技术从所有预测路径中分离出错误概率较大的路径。然后，根据低置信度污染型访存指令识别历史表将低置信度预测路径上的访存指令划分为预取型和污染型，为污染型的访存指令建立低优先级Load/Store队列，并采用污染数据Cache存储污染数据。仿真结果表明，在双核模式下，CSDA策略相对于baseline结构来说， L1 D-Cache缺失率降低幅度从9%-23%，平均降低了17%；L2 Cache缺失率的下降范围从1.02%-14.39%，平均为5.67%；IPC的提升幅度从0.19%-5.59%，平均为2.21%。 关键词：处理器; 存储系统; 置信度评估; Cache污染中图法分类号：TP303 文献标识码：AA Cache Pollution Control Method for Pre-Loaded Data During Microprocessor’s Speculative Execution Zhang Jun (Department of Computer Science and Engineering, Xi’an JiaoTong University, Xi ’an, Shaan’Xi, China 710049) Abstract：“Memory Wall” problem already became the chief obstacle for processor performance promotion. Cache pollution caused by the speculative execution of memory accessing instructions in predictive path may affect the performance of processor seriously. This paper proposes a cache pollution control method for predictive path during speculative execution based on confidence estimation, called CSDA. First of all, confidence estimation technique can be used to separate out the more likely mis-predictive paths from all predictive paths adaptively. Second, all memory accessing instructions in low confidence predictive path are divided into pre-fetching type and polluting type according to the low confidence polluting type memory accessing recognition history table adaptively. A special low priority Load/Store queue is created for these polluting type memory accessing instructions, and polluting data cache is used to store polluting data. In brief, CSDA may alleviate the negative effect of cache polluting data, and promotes the effective memory bandwidth. Simulation result indicates that, in dual-core configuration, CSDA could avoid most cache pollution and upgrade performance. Relative to the baseline architecture, miss rate reduction of D-Cache ranges from 9% to 23%, averagely 17%. Miss r