Research Initiation Award: Highly Optimizing Compilation for Constraint Logic Programming

9308631约束逻辑编程语言是一类基于基于规则的约束编程范例的声明性编程语言。该项目涉及高度优化CLP语言编译的实用技术的开发。由于这些语言的高级性，为了获得良好的性能，编译器必须基于程序的全局属性来执行优化。这项研究将集中于编写CLP(R)语言，尽管许多技术将适用于其他CLP语言。CLP语言的程序分析已经得到了广泛的研究，研究者开发了一个抽象机来高效地执行CLP(R)，并给出了一个合适的优化编译器。这个项目将解决以下问题：分析器提供的哪些信息对于获得现实程序的有效代码实际上是有用的，在给定这些信息的情况下，需要什么编译技术来获得代码，以及如何使这种编译具有可伸缩性和可靠性。这项研究将以调查员就中电实施事宜所作的工作为基础。它将大大增加对使用全球程序分析作为高度优化的CLP(R)编译器的基础的理解，并通过实施这样的编译器首次证明其可行性。它将提供一个急需的工具，便于研究人员在程序分析中对他们的工作进行经验性评估，并将有助于识别CLP语言高性能编译中的突出研究问题。***

9308631 Michaylov The Constraint Logic Programming (CLP) languages are a class declarative programming languages based on the paradigm of rule based constraint programming. This project concerns the development of practical techniques for highly-optimizing compilation of CLP languages. To obtain good performance, it is necessary for a compiler to perform optimizations based on global properties of programs, because of the high-level nature of these languages. This study will concentrate on compiling the language CLP(R), although many of the techniques will be applicable to other CLP languages. Program analysis for CLP languages has been studied extensively, and the investigator has developed an abstract machine for executing CLP(R) efficiently, given a suitable optimizing compiler. This project will address the questions of which information provided by analyzers is actually useful for obtaining efficient code for realistic programs, what compilation techniques are needed for obtaining the code given this information, and how such compilation can be made scalable and reliable. The study will build on the investigator's work on CLP implementation issues. It will lead to a greatly increased understanding of the use of global program analysis as the basis for a highly optimizing CLP(R) compiler, and demonstrate its feasibility for the first time through the implementation of such a compiler. It will provide a much needed tool that can easily be used by researchers in program analysis to evaluate their work empirically, and will help to identify outstanding research issues in high-performance compilation of CLP languages. ***