Research Initiation: VLSI Floating-Point Digital Signal Processing

现代信号处理算法及其相应的基于矩阵的运算对计算要求很高，需要在专用VLSI芯片和系统中实现。由于执行现代信号处理的电路既大又复杂，因此有必要使用最先进的工具来设计芯片和系统。理解这类工具的性质是当今的一个重要研究问题。首席研究员正在研究一种高级硅编译器的结构，该编译器致力于数字信号处理电路的设计。他打算分析和评估不同的数字格式及其有限字长对重要信号处理算法的影响，并评估高效VLSI实现所需的体系结构权衡。结果正在与高级计算机辅助设计优化程序相关联。他还通过一个适当的电子设计数据库，将系统权衡与较低级别的硅汇编工具相结合。

Modern signal processing algorithms and their corresponding matrix-based operations are computationally demanding and require implementation into application-specific VLSI chips and systems. Since circuits for executing modern signal processing are large and complex, it is necessary to have state-of-the-art tools to design the chips and systems. Understanding the nature of such tools is an important research issue today. The principal investigator is investigating the structure of an advanced silicon compiler dedicated to the design of digital signal processing circuits. He intends to analyze and evaluate different number formats and their finite word-length effects on important signal processing algorithms, and to evaluate architectural tradeoffs needed for efficient VLSI realization. Results are being incorproated into high-level computer-aided-design optimization programs. He is also integrating the system tradeoffs with lower-level silicon assembler tools through an appropriate electronic design database.