权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Study on High-Level Programming Environments for Supercomputing (1991)

超级计算高级编程环境研究（1991）

基本信息

批准号：
02650271
负责人：
SHIMASAKI Masaaki
金额：
$ 1.86万
依托单位：
Kyushu University
依托单位国家：
日本
项目类别：
Grant-in-Aid for General Scientific Research (C)
财政年份：
1990
资助国家：
日本
起止时间：
1990 至 1991
项目状态：
已结题

项目摘要

Problems of programming environments for supercomputing were investigated from the view points : (1) programming languages and processors, (2) mathematical libraries, (3) visualization and computer graphics. Parallel processing is considered important for supercomputers in the next generation. At present, the state of art for programming environment of parallel processors is very primitive. Processor scheduling plays an important role in parallel processing to attain high performance. In order to attain high performance for a single program, it is necessary that a compiler selects an adequate scheduling algorithm. We proposed new scheduling algorithms, called 1 )the static/dynamic merged algorithm, and 2 )the modified GSS which is extended to cope with IF statements in the loop for parallel execution. We have developed an experimental parallel C compiler for 4 RISC CPU's workstation with the new scheduling algorithm. Besides memory shared multiprocessors, it is considered that distributed memory machines also play an important role in the next generation of supercomputing. We proposed an extension to LINDA language in this context. As far as visualization and computer graphics are concerned, we investigated fractals in quaternions and its application to computer graphics, It is shown that self contraction mapping functions in quaternions can generate useful tree patterns in the 3 dimensional space.

从(1)编程语言和处理器、(2)数学库、(3)可视化和计算机图形学的角度研究了超级计算的编程环境问题。并行处理被认为是下一代超级计算机的重要组成部分。目前，并行处理器的编程环境还很原始。处理器调度在并行处理获得高性能的过程中起着重要作用。为了获得单个程序的高性能，编译器必须选择适当的调度算法。我们提出了新的调度算法，称为1)静态/动态合并算法和2)改进的GSS算法，该算法被扩展以处理循环中的IF语句以实现并行执行。采用新的调度算法，我们在4个RISC CPU的工作站上开发了一个实验性的并行C编译器。除了共享内存的多处理器外，分布式存储机器也被认为在下一代超级计算中扮演着重要的角色。在此背景下，我们提出了对Linda语言的扩展。在可视化和计算机图形学方面，我们研究了四元数中的分形及其在计算机图形学中的应用，证明了四元数中的自压缩映射函数可以在三维空间中生成有用的树模式。