{"title":"Experiments with a virtual tree machine using transputers","authors":"D. McBurney, M. Sleep","doi":"10.1109/HICSS.1989.47177","DOIUrl":null,"url":null,"abstract":"A number of experiments with a virtual process tree architecture called ZAPP (zero assignment parallel processor) are described. These experiments were performed using a network of Inmos transputers. Most of the experiments involve the parallel execution of simple process trees grown by rewrite rules that recursively decompose a large grain of work into smaller grains. Some experiments extend the model by supporting limited notions of global communication in the process tree. The basic principles of the ZAPP architecture are presented, and implementation details of a ZAPP kernel used to perform the experiments on a network of Inmos transputers are given. Results of early parallel experiments covering specific applications are described and compared with direct sequential execution on a single transputer. The results confirm earlier sequential simulation experiments, which showed that nearly 100% processor utilization can be obtained for problems that generate process trees much larger than the physical architecture. They also show that real speedups can be obtained with relatively low degrees of parallelism.<<ETX>>","PeriodicalId":300182,"journal":{"name":"[1989] Proceedings of the Twenty-Second Annual Hawaii International Conference on System Sciences. Volume 1: Architecture Track","volume":"233 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1989-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"[1989] Proceedings of the Twenty-Second Annual Hawaii International Conference on System Sciences. Volume 1: Architecture Track","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/HICSS.1989.47177","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
A number of experiments with a virtual process tree architecture called ZAPP (zero assignment parallel processor) are described. These experiments were performed using a network of Inmos transputers. Most of the experiments involve the parallel execution of simple process trees grown by rewrite rules that recursively decompose a large grain of work into smaller grains. Some experiments extend the model by supporting limited notions of global communication in the process tree. The basic principles of the ZAPP architecture are presented, and implementation details of a ZAPP kernel used to perform the experiments on a network of Inmos transputers are given. Results of early parallel experiments covering specific applications are described and compared with direct sequential execution on a single transputer. The results confirm earlier sequential simulation experiments, which showed that nearly 100% processor utilization can be obtained for problems that generate process trees much larger than the physical architecture. They also show that real speedups can be obtained with relatively low degrees of parallelism.<>