论文部分内容阅读
本文通过把时延约束条件转化为一个布尔矩阵，利用这个布尔矩阵构造能量函数的时延约束项，从而解决了用神经网络处理时延约束这类不等式约束的难题，并据此提出了一个性能驱动的ＶＬＳＩ系统划分的均场退火算法．算法不仅考虑了模块间的连接关系，还考虑了版图的物理结构，是一种逻辑与版图相结合的划分方法．实验表明，该算法具有较强的寻优能力
In this paper, we transform the delay constraint into a Boolean matrix, and use this Boolean matrix to construct the delay constraint of energy function, which solves the problem of using neural networks to deal with inequality constraints such as delay constraints, and proposes a performance Homogenization Annealing Algorithms for Driving VLSI Systems. The algorithm not only considers the connection between modules, but also considers the physical structure of the layout. It is a combination of logic and layout. Experiments show that this algorithm has a strong ability to search