http://swrc.ontoware.org/ontology#Article
A Binary Tree Multiprocessor: CORAL
en
Tokushima University Tokushima Japan.
Tokushima University Tokushima Japan.
Tokushima University Tokushima Japan.
Yoshizo Takahashi
Naoki Wakabayashi
Yoshihiro Nobutomo
A feasible architecture for a highly parallel processing system which consists of 100 or more processors is studied. The commonly used shared data system in which a shared store is connected directly to the multiple of processors is inapplicable when the number of the processors is very large. An alternative system is the distributed data system in which the processors are loosely connected and the data are distributed to the processors prior to the processing. In this system the time to distribute the data and the time consumed in the inter-processor communications deteriorate the speed-up ratio. After various connections of the distributed data system including star chain loop lattice and binary tree were studied it was revealed that the binary tree connection has the best performance. This binary tree multiprocessor is named CORAL. An algorithm to handle the inter-Processor communications in the CORAL is developed. As possible applications of the CORAL the parallel solutions of partial differential equations in one-dimensional heat conduction problem and the Potential problem are studied. A prototype of CORAL consisting of seven microcomputers is described.
A feasible architecture for a highly parallel processing system which consists of 100 or more processors is studied. The commonly used shared data system in which a shared store is connected directly to the multiple of processors is inapplicable when the number of the processors is very large. An alternative system is the distributed data system in which the processors are loosely connected and the data are distributed to the processors prior to the processing. In this system the time to distribute the data and the time consumed in the inter-processor communications deteriorate the speed-up ratio. After various connections of the distributed data system, including star, chain, loop, lattice, and binary tree, were studied, it was revealed that the binary tree connection has the best performance. This binary tree multiprocessor is named CORAL. An algorithm to handle the inter-Processor communications in the CORAL is developed. As possible applications of the CORAL, the parallel solutions of partial differential equations in one-dimensional heat conduction problem and the Potential problem are studied. A prototype of CORAL consisting of seven microcomputers is described.
AA00700121
Journal of Information Processing
3
4
230-237
1981-02-15
1882-6652