CSR-AES: Language, Compiler and Run-time Support for Irregular Applications in Emerging High-productivity Languages

该项目正在与Cray，Inc.的Chapel编程语言团队合作，开发在高性能高度并行平台上编程“不规则”并行应用程序的新技术。该项目解决了此类应用程序的几个核心问题。首先，这项工作是通过添加通用语言特性来指定对特殊情况的优化处理来扩展Chapel现有的对用户定义的不规则数据分布的支持，这通常是足够性能所必需的。其次，该项目正在设计和评估软件事务性内存（STM）算法，该算法可扩展到非常大规模的系统（例如，分布式内存集群），与之前专注于较小规模（缓存一致）系统的STM工作形成对比。第三，整个研究项目的一个（单独资助的）组成部分是开发基于指针的数据并行计算的确定性语言机制，使用类型和效果系统来表达并行性和证明不干扰。如果成功，这项研究将提高编程的便利性，并为广泛的应用程序提供高性能，这些应用程序对程序员来说尤其具有挑战性，即“不规则”应用程序。此外，这项工作旨在实现两种直接的工业技术转移：为Cray的Chapel语言和编译器工作做出贡献，以及为DARPA HPCS项目的联合语言设计做出贡献。主要的研究特征，包括语言规范和编译器原型，将作为免费的开源软件提供给研究社区使用。高性能并行程序的动态编译器基础结构应该是该领域进一步研究的一个特别有价值的工具。

This project is developing new techniques for programming "irregular"parallel applications on high-performance highly parallel platforms,in collaboration with the Chapel programming language team at Cray,Inc. The project addresses several core issues for such applications.First, the work is extending Chapel's existing support foruser-defined irregular data distributions by adding general languagefeatures for specifying optimized handling of special cases, which areusually essential for adequate performance. Second, the project isdesigning and evaluating software transactional memory (STM)algorithms that scale to very large scale systems (e.g.,distributed-memory clusters), in contrast to previous STM work thathas focused on smaller-scale (cache coherent) systems. Third, a(separately funded) component of the overall research project isdeveloping deterministic language mechanisms for parallel computationson pointer-based data using a type and effect system for expressingparallelism and proving noninterference.If successful, this research will improve ease of programming anddeliver high performance for a wide range of applications that areespecially challenging for programmers today, namely, "irregular"applications. Further, the work aims to achieve two kinds of directtechnology transfer to industry: contributions to the Chapel languageand compiler effort at Cray, and contributions to the joint languagedesign effort in the DARPA HPCS program. The major researchartifacts, including the language specification and compilerprototype, will be made available as free, open source software forthe research community to use. The dynamic compiler infrastructurefor high-performance parallel programs should be a particularlyvaluable tool for further research in this area.