Collaborative Research: CPA-SEL: Implementation Techniques for High-level Parallel Languages

多核体系结构的快速趋势保证了计算机程序的更快执行，但由于缺乏良好的编程模型来利用这种体系结构中的并行性，因此给软件开发带来了巨大的困难。对于编程语言研究来说，这种情况是一个重要的机会，可以为编写桌面应用程序提供有效的语言和工具，同时利用多核硬件的性能。众所周知，函数式编程语言为并发和并行编程提供了良好的语义基础，但要想让这类语言取得成功，它们需要提供具有竞争力的性能。研究的重点是并行函数式语言高效实现的技术挑战。多核和多核体系结构的特点要求实现在并行构造中保留顺序语义，管理并行线程的粒度和调度，并了解数据的局部性。这项研究探索了一系列技术，这些技术结合了静态程序分析、编译器转换和动态运行时策略。对传统并行基准和小型应用的实证分析被用来评估本研究开发的技术的有效性。通过解决性能问题，这项研究将使并行函数式编程语言在广泛的应用程序中得到实际应用。

The rapid trend toward multi-core architectures promises faster execution of computer programs but poses significant difficulties for software development due to the lack of good programming models for exploiting the parallelism in such architectures. This situation is a significant opportunity for programming-language research to supply effective languages and tools for writing desktop applications while exploiting the performance of multi-core hardware. It is well known that functional-programming languages provide a good semantic base for concurrent and parallel programming, but for such languages to be successful, they need to provide competitive performance. The research focuses on the technical challenges in the efficient implementation of parallel functional languages. The characteristics of multi-core and many-core architectures demand that implementations preserve sequential semantics in parallel constructs, manage the granularity and scheduling of parallel threads, and be aware of the locality of data. The research explores a collection of techniques that combine static program analyses, compiler transformations, and dynamic runtime policies. Empirical analysis of both traditional parallel benchmarks and small applications is used to evaluate the effectiveness of the techniques developed by this research. By addressing performance concerns, the research will enable the practical use of parallel functional programming languages for a broad range of applications.