Experimental Software Systems: Collaborative Research: Applications of Flow Types in the Efficient, Modular, and Reliable Compilation of Higher-Order Typed Languages

9806745 Kfoury, Assaf J. Wells, Joseph B.波士顿大学实验软件系统：协作研究：流类型在高阶类型语言高效、模块化和可靠编译中的应用现代编程语言（如ML、Haskell、Scheme、Java）有几个表达性的特性，这些特性使程序员可以从低级的、系统特定的细节中抽象出来，从而获得代码安全、代码重用和模块化等好处。不幸的是，这些特性很难在编译器中有效地实现。核心的困难是编译器必须决定在不同的上下文中为相同类型的不同抽象分配什么具体的表示。传统的编译器（效率低下）通过为所有数据选择统一的表示来解决这个问题。这项研究是Church Project （http://www.cs.bu.edu/groups/church）的一部分，通过以下方式解决了这一挑战：(1)在程序的中间表示中嵌入类型和控制/数据流信息，(2)在编译器的各个阶段保持这些信息的准确性，(3)使用这些信息根据上下文定制数据表示，并驱动几个编译器优化。我们的目标是通过实验来评估这些“流类型”在编译像ML这样的现代语言中的有效性。这项研究的产品将包括一个编译工作台，可以根据流信息进行调整，实验结果表明类型和流信息在编译中的有效性，以及设计、实现和评估一个用于编译程序片段的新框架。

9806745 Kfoury, Assaf J. Wells, Joseph B. Boston University Experimental Software Systems: Collaborative Research: Applications of Flow Types in the Efficient, Modular, and Reliable Compilation of Higher-Order Typed Languages Modern programming languages (e.g. ML, Haskell, Scheme, Java) have several expressive features that let programmers abstract away from low-level, system-specific details and achieve benefits like code safety, code reuse, and modularity. Unfortunately, these features are challenging to implement efficiently in a compiler. The central difficulty is that the compiler must decide what concrete representations to assign to different abstractions of the same type in different contexts. Traditional compilers (inefficiently) solve this problem by choosing a uniform representation for all data. This research, part of the Church Project (http://www.cs.bu.edu/groups/church) addresses this challenge by (1) embedding type and control/data flow information in the intermediate representation of the program, (2) maintaining the accuracy of this information through the stages of the compiler and (3) using this information to customize data representations based on context, and to drive several compiler optimizations. The goal is to experimentally evaluate the efficacy of these "flow types" to compile a modern language like ML. Products of this research will include a compiler workbench that can be tuned with respect to flow information, experimental results indicating the effectiveness of type and flow information in compilation, and the design, implementation and evaluation of a new framework for the compilation of program fragments.