高级检索
朱柏成, 储著飞, 潘鸿洋, 王伦耀, 夏银水. 基于XMG的乘法器电路等价性验证算法[J]. 计算机辅助设计与图形学学报, 2024, 36(3): 443-451. DOI: 10.3724/SP.J.1089.2024.19949
引用本文: 朱柏成, 储著飞, 潘鸿洋, 王伦耀, 夏银水. 基于XMG的乘法器电路等价性验证算法[J]. 计算机辅助设计与图形学学报, 2024, 36(3): 443-451. DOI: 10.3724/SP.J.1089.2024.19949
Zhu Baicheng, Chu Zhufei, Pan Hongyang, Wang Lunyao, Xia Yinshui. An XMG Based Equivalence Checking Algorithm for Large Bit-Width Multiplier Circuits[J]. Journal of Computer-Aided Design & Computer Graphics, 2024, 36(3): 443-451. DOI: 10.3724/SP.J.1089.2024.19949
Citation: Zhu Baicheng, Chu Zhufei, Pan Hongyang, Wang Lunyao, Xia Yinshui. An XMG Based Equivalence Checking Algorithm for Large Bit-Width Multiplier Circuits[J]. Journal of Computer-Aided Design & Computer Graphics, 2024, 36(3): 443-451. DOI: 10.3724/SP.J.1089.2024.19949

基于XMG的乘法器电路等价性验证算法

An XMG Based Equivalence Checking Algorithm for Large Bit-Width Multiplier Circuits

  • 摘要: 组合电路等价性验证是数字集成电路设计自动化(EDA)中的重要部分,随着算术电路在现代计算机系统中的占比逐渐增大,传统的等价性验证算法在验证多比特算术电路,尤其是乘法器电路时面临挑战.对此,提出一种基于XOR-Majority Graph (XMG)逻辑表示的组合电路等价性验证算法.首先将2个待验证电路构建成的联接(Miter)电路进行XMG逻辑重写;然后在等价性一致的前提下对XMG的节点个数和逻辑深度进行逻辑重写优化;最后调用布尔可满足性(SAT)求解器和仿真器进行验证,得到最终等价性验证结果.实验结果表明,与ABC,Lingeling等工具相比,所提算法在验证时间上实现了平均489倍、最高1 472倍的加速.

     

    Abstract: Combinational equivalence checking is an essential part of electronic design automation (EDA). Modern computers have an increasing proportion of arithmetic logic circuits, which poses challenges for traditional equivalence checking algorithms, especially when checking multi-bit logic circuits. An equivalence checking method based on XOR-majority graph (XMG) is proposed to address this issue. Firstly, a miter circuit is constructed using an XMG representation of the referenced design and the design to be verified. Secondly, a functional rewrite is performed on the XMG of the miter, in which the logic level and number of nodes are optimized. Finally, the Boolean satisfiability (SAT) solver and simulator are used to verify the final equivalence verification result. As compared to tools such as ABC and Lingeling, the algorithm performs 489 times CPU time speedup on average, and a maximum acceleration of 1472 times in validation time.

     

/

返回文章
返回