Representation Specification and Optimization of Object-Oriented Languages

高级编程语言，如Smalltalk、ML和Java，可以使编程比低级编程语言（如C和c++）更容易、更可靠和更灵活。不幸的是，这些软件工程的好处往往是以牺牲运行时效率为代价的。高级语言实现良好性能的一个重要障碍是相对低效的数据结构表示和布局方法，其中数据结构统一表示为通过指针间接引用的堆分配结构。本研究旨在降低这种高级数据模型的性能成本，同时为程序员保留其软件工程优势。将开发用于指定数据结构布局和指定数据结构布局优化的声明技术。基于程序员建议、自动静态分析和动态配置反馈的混合，将开发用于决定应用哪种布局优化的技术。将设计一个灵活的，独立于语言的编译器中间表示，以显式地维护，检查和优化表示。这些技术将在优化编译器中实现，并在各种面向对象编程语言的大型基准程序中测量其有效性。

Craig ChambersHigher-level programming languages, such as Smalltalk, ML, and Java, can makeprogramming easier, more reliable, and more flexible than lower-levelprogramming languages, such as C and C++. Unfortunately, these softwareengineering benefits often come at significant cost to run-time efficiency. Oneimportant obstacle to achieving good performance for higher-level languages isthe relatively inefficient approach to the representation and layout of datastructures, where data structures are represented uniformly as heap-allocatedstructures referenced indirectly via pointers.This research aims to reduce the performance costs of this high-level datamodelwhile retaining its software engineering benefits for programmers.Declarativetechniques will be developed for specifying the layout of data structures andfor specifying optimizations to data structure layouts. Techniques will bedeveloped for deciding which layout optimizations to apply, based on a mix ofprogrammer suggestions, automatic static analyses, and dynamic profilefeedback. A flexible, language-independent compiler intermediaterepresentationwill be designed that explicitly maintains, checks, and optimizesrepresentations. The techniques will be implemented in an optimizingcompiler,and their effectiveness measured on a range of large benchmark programs in avariety of object-oriented programming languages.