arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2108.00257cs.LGcs.CE
收藏

BoA-PTA,一个贝叶斯优化加速无错SPICE求解器

BoA-PTA, A Bayesian Optimization Accelerated Error-Free SPICE Solver

Beihang University(北京航空航天大学);University of Petroleum-Beijing(中国石油大学(北京));Southeast university(东南大学)

Wei W. Xing, Xiang Jin, Yi Liu, Dan Niu, Weishen Zhao, Zhou Jin

摘要:

IC设计中最大的挑战之一是重复执行计算昂贵的SPICE模拟,特别是当涉及高度复杂的芯片测试/验证时。近年来,伪瞬态分析(PTA)已被证明是最有前途的连续SPICE求解器之一。然而,PTA效率受到插入的伪参数的高度影响。在这项工作中,我们提出了BoA-PTA,贝叶斯优化加速PTA,可以大大加快模拟和提高收敛性能,而不会引入额外的错误。此外,我们的方法不需要任何预先计算数据或离线训练。加速框架可以被实现以立即加速正在进行的重复仿真,或者改进完全不同电路的新仿真。BoA-PTA配备了尖端的机器学习技术,例如,深度学习、高斯过程、贝叶斯优化、非平稳单调变换和通过参数化的变分推理。我们评估BoA-PTA在43个基准电路对其他SOTA SPICE求解器,并证明了平均2.3倍(最大3.5倍)的速度比原来的CEPTA。

英文摘要:

One of the greatest challenges in IC design is the repeated executions of computationally expensive SPICE simulations, particularly when highly complex chip testing/verification is involved. Recently, pseudo transient analysis (PTA) has shown to be one of the most promising continuation SPICE solver. However, the PTA efficiency is highly influenced by the inserted pseudo-parameters. In this work, we proposed BoA-PTA, a Bayesian optimization accelerated PTA that can substantially accelerate simulations and improve convergence performance without introducing extra errors. Furthermore, our method does not require any pre-computation data or offline training. The acceleration framework can either be implemented to speed up ongoing repeated simulations immediately or to improve new simulations of completely different circuits. BoA-PTA is equipped with cutting-edge machine learning techniques, e.g., deep learning, Gaussian process, Bayesian optimization, non-stationary monotonic transformation, and variational inference via parameterization. We assess BoA-PTA in 43 benchmark circuits against other SOTA SPICE solvers and demonstrate an average 2.3x (maximum 3.5x) speed-up over the original CEPTA.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑