System Level Issues for Multiparadigm Computing and SIMD and MIMD/SIMD Architectures

多范型计算以及 SIMD 和 MIMD/SIMD 架构的系统级问题

• 批准号: 9505960
• 负责人: Jose Meseguer
• 金额: $ 23.61万
• 依托单位: SRI International
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1995
• 资助国家: 美国
• 起止时间: 1995-09-01 至 1999-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9505960&HistoricalAwards=false
• 关键词: System; Level; Issues; Multiparadigm; Computing

TITLE: and MIMD/SIMD Architectures Parallel computing applications differ widely in the type of computational paradigm that is best suited to efficiently exploit their parallelism. Current software systems for parallel hardware support only single-paradigm computing. The objective of this research project is to investigate system-level support for efficient multiparadigm parallel computation. Previous work on coordination languages and multiparadigm sequential programming languages form a solid foundation for this research, although significant new ideas are required. Although current high-performance computer architectures are moving toward a combined MIMD/SIMD organization, very little software and system support exists for such hybrid computer architectures. This research will thus also focus on aspects of multiparadigm computing for massively parallel SIMD and MIMD/SIMD machines, including the Rewrite-Rule Machine as a testbed architecture. The research results will enable future systems designers to better appreciate the engineering trade-offs to efficiently support multiparadigm computing in a parallel environment. The results of this research are expected to enhance the future programming ease of high-performance massively parallel computing systems.

标题：MIMD/SIMD架构 并行计算应用程序在最适合有效利用其并行性的计算范例类型方面存在很大差异。 目前的并行硬件软件系统只支持单范例计算。 本研究的目的是探讨系统级支持高效的多范式并行计算。 以前的工作协调语言和多范式顺序编程语言形成了坚实的基础，这项研究，虽然需要重大的新想法。 虽然当前的高性能计算机体系结构正朝着MIMD/SIMD组合组织发展，但对于这种混合计算机体系结构，存在非常少的软件和系统支持。因此，这项研究也将集中在大规模并行SIMD和MIMD/SIMD机器的多范式计算方面，包括重写规则机作为一个测试平台架构。 研究结果将使未来的系统设计者能够更好地理解工程权衡，以有效地支持多范式计算在并行环境中。 这项研究的结果，预计将提高未来的高性能大规模并行计算系统的编程方便。