| 处理器验证激励在线筛选技术 |
| |
| 引用本文： | 郭崎.处理器验证激励在线筛选技术[J].计算机辅助设计与图形学学报,2012,24(5):690-698. |
| |
| 作者姓名： | 郭崎 |
| |
| 作者单位： | 中国科学院计算技术研究所计算机体系结构国家重点实验室 北京100190;中国科学院研究生院 北京 100039;北京龙芯中科技术服务中心有限公司 北京 100190 |
| |
| 基金项目： | 国家“核高基”科技重大专项，国家自然科学基金，国家“八六三”高技术研究发展计划，中国科学院战略性先导科技专项 |
| |
| 摘 要： | 如何生成高质量的验证激励是功能验证中的核心问题之一.随着功能验证的不断进行,验证激励的有效性也随之降低.为了提高验证激励的质量,提出一种在线筛选技术来处理验证激励.该技术采用单分类支持向量机来在线地构建分类器,以对新生成的验证激励是否冗余进行预测,如果是冗余的,则不进入仿真阶段进行仿真.在此基础上,进一步提出指令序列核函数来衡量不同指令序列的相似程度.实验结果表明,与约束随机生成技术相比,文中技术可以减少约83%的验证激励及79%的验证时间.
|
| 关 键 词： | 功能验证 激励生成 在线学习 功能覆盖率 核函数 |
Online Filtration of Stimuli for Microprocessor Verification |
| |
| Guo Qi.Online Filtration of Stimuli for Microprocessor Verification[J].Journal of Computer-Aided Design & Computer Graphics,2012,24(5):690-698. |
| |
| Authors: | Guo Qi |
| |
| Affiliation: | Guo Qi1,2,3) 1)(State Laboratory of Computer Architecture,Institute of Computing Technology,Chinese Academy of Sciences,Beijing 100190) 2)(Graduate School of Chinese Academy of Sciences,Beijing 100049) 3)(Loongson Technologies Corporation Limited,Beijing 100190) |
| |
| Abstract: | One of the most critical issues during functional verification is to generate highly effective stimuli.As verification proceeds,the effectiveness of verification stimuli decreases.To improve the effectiveness of stimuli,an online filteration technique to process is proposed to generated stimuli.This technique employs one-class support vector machines to online construct a classifier to predict whether or not a newly generated stimulus is redundant,and the predicted redundant stimulus will not be sent for simulation.Besides,we also propose an instruction sequence kernel to measure the similarities among instruction sequences.Experimental results demonstrate that this technique can reduce about 83% stimuli and 79% verification time in comparison with conventional constrained random generation. |
| |
| Keywords: | functional verification stimuli generation online learning functional coverage kernel function |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
