一种固定轮廓布图规划的迭代合并布局算法
An iterative merging placement algorithm for the fixed-outline floorplanning
摘要：
给定一组面积固定、维数可变的矩形模块和一个固定的矩形电路。固定轮廓软模块布图规划（FOFSM）的布局旨在确定电路上每个模块的尺寸和位置。提出了一种两阶段迭代合并布局算法。第一阶段迭代地将两个面积最小的模块合并为一个复合模块，得到最终的复合模块，并自底向上构建切片树。第二阶段以自顶向下的层次结构递归地确定切片树中兄弟模块的相对关系（左右或上下），并且自动确定每个叶模块的尺寸和位置。与零死空间（ZDS）算法相比，该算法在一定条件下保证了可行布局，证明了该算法在更宽松的条件下也能构造出可行布局。此外，IMP在考虑线长或不考虑零死区约束的情况下处理FOFSM时具有更好的可扩展性。
英文摘要：
Given a set of rectangular modules with fixed area and variable dimensions, and a fixed rectangular circuit. The placement of Fixed-Outline Floorplanning with Soft Modules (FOFSM) aims to determine the dimensions and position of each module on the circuit. We present a two-stage Iterative Merging Placement (IMP) algorithm for the FOFSM with zero deadspace constraint. The first stage iteratively merges two modules with the least area into a composite module to achieve a final composite module, and builds up a slicing tree in a bottom-up hierarchy. The second stage recursively determines the relative relationship (left-right or top-bottom) of the sibling modules in the slicing tree in a top-down hierarchy, and the dimensions and position of each leaf module are determined automatically. Compared with zero-dead-space (ZDS) algorithm, the only algorithm guarantees a feasible layout under some condition, we prove that the proposed IMP could construct a feasible layout under a more relaxed condition. Besides, IMP is more scalable in handling FOFSM considering the wirelength or without the zero deadspace constraint.
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。
