CAREER: Parallel Integer Programming for Architectural-LevelVLSI Design

该项目的研究部分将导致设计和开发可移植的并行软件，以解决在VLSI系统设计中出现的架构级优化问题。这些问题中的几个可以表示为超大规模整数规划，可以使用迭代松弛技术在多项式时间内求解。然而，这些算法在串口计算机上的实际运行时间可能长得令人望而却步，特别是对于交互式设计环境。本项目开发的可伸缩并行软件的执行速度将比传统的串口代码快几个数量级。因此，它将使VLSI设计者能够在设计过程的早期阶段交互地探索他们的设计空间，并在致力于任何特定实施之前有效地考虑几种架构替代方案。该项目的教育部分旨在将并行计算无缝地整合到本科生和研究生课程中。为此，耶鲁大学将开发两门新课程。第一门课程将向本科生介绍串行和并行计算机系统组织的基础知识。第二门课程将介绍超大规模集成电路系统的计算机辅助设计的算法，重点研究并行电子设计自动化。***

The research component of this project will result in the design and development of portable parallel software for solving architectural-level optimization problems that arise in the design of VLSI systems. Several of these problems can be formulated as very large scale integer programs that can be solved in polynomial time using iterative relaxation techniques. The practical running times of these algorithms on serial computers can be prohibitively long, however, especially for interactive design environments. The scaleable parallel software developed in this project will be able to execute orders of magnitude faster than conventional serial code. It will thus enable VLSI designers to interactively explore their design space at an early stage in the design process and to efficiently consider several architectural alternatives before committing to any particular implementation. The educational part of this project aims at the seamless integration of parallel computation in the undergraduate and graduate curriculum. To that effect, two new courses will be developed at Yale. The first course will introduce undergraduates to the basics of serial and parallel computer systems organization. The second course will cover algorithms for computer-aided design of VLSI systems with a research focus on parallel electronic design automation. ***