The Efficient Use of Parallel Computers

本研究项目旨在开发和分析算法 用于支持高级并行编程模型。 图 嵌入已经成功地用于比较不同的 并行架构，并将并行计算映射到并行 网络. 该项目的一个组成部分是发展 图嵌入的新技术。 第二个组成部分旨在 开发和分析有效的负载平衡算法，以支持 运行时的控制级并行性。 第三种成分是定向的 来评估不同编程模型的表达能力。 最后一个部分是Fluent机器的设计， 支持高级指令集的并行系统 有效地研究重点是算法和嵌入 可以被严格证明有效的技术。

This research project is aimed at developing and analyzing algorithms for supporting high-level parallel programming models. Graph embeddings have been successfully employed to compare different parallel architectures and to map parallel computations into parallel networks. One component of the project is directed towards developing new techniques for graph embeddings. A second component aims to develop and analyze efficient load-balancing algorithms for supporting control-level parallelism at run time. A third component is directed toward evaluating the expressive power of different programming models. The final component is the design of the Fluent machine, a massively parallel system which supports a very high-level instruction set efficiently. The research emphasizes algorithms and embedding techniques that can be rigorously proved efficient.