http://swrc.ontoware.org/ontology#Article
Efficiency of Parallel Computation on the Binary-Tree Machine CORAL'83
en
Department of Information Science Faculty of Engineering University of Tokushima
Shikoku Kakouki Co. Ltd.
Fuchu Works Toshiba Corporations.
Department of Information Science Faculty of Engineering University of Tokushima
Department of Information
Yoshizo Takahashi
Yoshitaka Yamane
Kazuyoshi Nishiyama
Fuminori Yoshitani
Katuhiro Inoue
In order to study the effectiveness of binary-tree architecture to parallel computing for general application problems a small-scale binary-tree machine CORAL'83 equipped with a software environment has been developed. The CORAL'83 consists of a host computer and a processor tree. The latter is composed of 15 processor elements each of which has an 8085 as a cpu 8kB ROM 17kB RAM and one parallel port for each of three directions. The software environment includes a C compiler for writing parallel programs an initial program loader interprocessor and host-root communication routines and several distributed operating systems. Using this facility a number of parallel computing programs for various problems were written and were executed to measure the efficiency of parallel computation. From them a parallel SOR computation of Laplace equation a tree sort the N queens problem a computation of prime numbers with the sieve of Eratosthenes and the FFT have been selected. The exploitation of parallel algorithms adequate to the binary-tree machine and the measured efficiencies of parallel computation for these programs are presented. This study proves that the binary-tree machine has modest efficiencies for both structured and unstructured problems.
In order to study the effectiveness of binary-tree architecture to parallel computing for general application problems, a small-scale binary-tree machine CORAL'83 equipped with a software environment has been developed. The CORAL'83 consists of a host computer and a processor tree. The latter is composed of 15 processor elements each of which has an 8085 as a cpu, 8kB ROM, 17kB RAM, and one parallel port for each of three directions. The software environment includes a C compiler for writing parallel programs, an initial program loader, interprocessor and host-root communication routines, and several distributed operating systems. Using this facility, a number of parallel computing programs for various problems were written and were executed to measure the efficiency of parallel computation. From them a parallel SOR computation of Laplace equation, a tree sort, the N queens problem, a computation of prime numbers with the sieve of Eratosthenes, and the FFT have been selected. The exploitation of parallel algorithms adequate to the binary-tree machine and the measured efficiencies of parallel computation for these programs are presented. This study proves that the binary-tree machine has modest efficiencies for both structured and unstructured problems.
AA00700121
Journal of Information Processing
8
4
288-299
1986-03-15
1882-6652