| 基于贝叶斯网络的验证激励向量生成研究 |
| |
| 引用本文： | 王 润,蒋剑飞,王 琴.基于贝叶斯网络的验证激励向量生成研究[J].教育技术导刊,2020,19(7):1-4. |
| |
| 作者姓名： | 王 润 蒋剑飞 王 琴 |
| |
| 作者单位： | 上海交通大学 微电子学院，上海 200240 |
| |
| 基金项目： | 国家自然科学基金项目（61176037） |
| |
| 摘 要： | 为了解决传统数字芯片验证环节中基于仿真的验证（或动态验证）功能覆盖率收敛速度慢的缺点，提出一种新的以功能覆盖为导向的测试用例生成方法，该方法基于贝叶斯网络和机器学习技术，可实现从覆盖模型到测试用例生成器反馈回路的自动关闭，在 DUT 的验证过程中，使用该方法为所测试的设计生成新的激励。实验结果表明，基于贝叶斯网络的 CDG 技术测试用例使用较少，覆盖率收敛更快，与传统基于仿真的验证技术相比，测试用例数量减少了 43%。基于贝叶斯网络的
比于传统动态验证技术而言其芯片功能验证更完善。
|
| 关 键 词： | 功能验证 贝叶斯网络 机器学习 CDG 覆盖率 |
| 收稿时间： | 2019-01-20 |
Research on Generation of Verification Incentive Vector Based on Bayesian Network |
| |
| WANG Run,JIANG Jian-fei,WANG Qin.Research on Generation of Verification Incentive Vector Based on Bayesian Network[J].Introduction of Educational Technology,2020,19(7):1-4. |
| |
| Authors: | WANG Run JIANG Jian-fei WANG Qin |
| |
| Institution: | Microelectronics Institute，Shanghai Jiao Tong University，Shanghai 200240，China |
| |
| Abstract: | Deep learning is the foundation that enables many artificial intelligence applications. Cloud service is the mainstream provisioning mode of computing power today. Providing deep learning capabili-ties by using cloud services has received widespread attention from major cloud vendors. The key to constructing such services is the adaptation of deep learning workload to cloud service models，
which involves the compatibility issues and adaptive developments of job scheduling，storage access，and resource management. OMAI deep learning platform performs a practice of adapting deep learning as a service on the cloud. It effectively solves the cloud service adaptation problem of deep learning workload through mechanisms such as the middle-layer abstraction of job scheduling，the mount of
heterogeneous storage inside containers and the matching mechanism of resource expressions. The experience of designing OMAI provides a general guidance for the development of artificial intelligence services on clouds. |
| |
| Keywords: | functional verification Bayesian network machine learning CDG coverage |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
