ITR - Compiler analysis for portability across memory and consistency models

ITR - 跨内存和一致性模型的可移植性编译器分析

• 批准号: 0313033
• 负责人: Samuel Midkiff
• 金额: $ 30万
• 依托单位: Purdue University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2003
• 资助国家: 美国
• 起止时间: 2003-08-01 至 2007-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0313033&HistoricalAwards=false
• 关键词: ITR; Compiler; analysis; portability; across

The Java programming language made writing multi-threaded programs a mainstream activity. Subtle issues arise in how to compile these multi-threaded programs so that common compiler optimizations are allowed while simultaneously ensuring that actions in one thread are viewed by other threads as occurring in a legal order, i.e. one that follows the programming language consistency model. Because previous languages either did not define a memory model or were not widely used, and because multi-threaded programming was done by a small number of experts, the trade-offs between memory model ease-of-use and the ease of compilation were not extensively investigated. In this project, we are building a compiler that will accept as input a programming language memory model definition, a hardware consistency model definition, and an application program. The compiler will then optimize the application program while abiding by the constraints of both the memory and consistency model, allowing the trade-offs between ease of use and compilation to be studied. It will also allow new memory models to be investigated, which will lead to the design of easier to use languages that can be effectively compiled, and used to produce more robust, safer software.

Java编程语言使编写多线程程序成为主流活动。 在如何编译这些多线程程序以使得允许常见的编译器优化的同时确保一个线程中的动作被其他线程视为以法律的顺序（即，遵循编程语言一致性模型的顺序）发生时出现了微妙的问题。 因为以前的语言要么没有定义内存模型，要么没有被广泛使用，而且因为多线程编程是由少数专家完成的，所以内存模型易用性和编译易用性之间的权衡没有得到广泛的研究。 在这个项目中，我们正在建立一个编译器，将接受作为输入的编程语言内存模型定义，硬件一致性模型定义，和一个应用程序。 然后，编译器将优化应用程序，同时遵守内存和一致性模型的约束，允许研究易用性和编译之间的权衡。 它还将允许研究新的内存模型，这将导致设计更容易使用的语言，可以有效地编译，并用于生产更强大，更安全的软件。