{"title":"在消息传递并行机器上进行静态数组分区的两种技术","authors":"Eric Hung-Yu Tseng, J. Gaudiot","doi":"10.1109/PACT.1997.644018","DOIUrl":null,"url":null,"abstract":"We present two techniques for partitioning arrays in parallel DoAll loops for message-passing parallel machines. (1) Communication-free array partitioning: a general solution of communication-free partitioning is derived for arrays in a DoAll loop. The derivation is based on the Smith normal form decomposition of the matrix which characterizes the array references in a DoAll loop. (2) One block-communication partitioning: when communication-free partitioning is not possible, we derive the partitioning equations which allocate all remote data to a unique processor. Thus, at most one block-communication is required for each processor to obtain the remote data it needs during computation.","PeriodicalId":177411,"journal":{"name":"Proceedings 1997 International Conference on Parallel Architectures and Compilation Techniques","volume":"58 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1997-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"Two techniques for static array partitioning on message-passing parallel machines\",\"authors\":\"Eric Hung-Yu Tseng, J. Gaudiot\",\"doi\":\"10.1109/PACT.1997.644018\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"We present two techniques for partitioning arrays in parallel DoAll loops for message-passing parallel machines. (1) Communication-free array partitioning: a general solution of communication-free partitioning is derived for arrays in a DoAll loop. The derivation is based on the Smith normal form decomposition of the matrix which characterizes the array references in a DoAll loop. (2) One block-communication partitioning: when communication-free partitioning is not possible, we derive the partitioning equations which allocate all remote data to a unique processor. Thus, at most one block-communication is required for each processor to obtain the remote data it needs during computation.\",\"PeriodicalId\":177411,\"journal\":{\"name\":\"Proceedings 1997 International Conference on Parallel Architectures and Compilation Techniques\",\"volume\":\"58 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"1997-11-11\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Proceedings 1997 International Conference on Parallel Architectures and Compilation Techniques\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/PACT.1997.644018\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings 1997 International Conference on Parallel Architectures and Compilation Techniques","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/PACT.1997.644018","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Two techniques for static array partitioning on message-passing parallel machines
We present two techniques for partitioning arrays in parallel DoAll loops for message-passing parallel machines. (1) Communication-free array partitioning: a general solution of communication-free partitioning is derived for arrays in a DoAll loop. The derivation is based on the Smith normal form decomposition of the matrix which characterizes the array references in a DoAll loop. (2) One block-communication partitioning: when communication-free partitioning is not possible, we derive the partitioning equations which allocate all remote data to a unique processor. Thus, at most one block-communication is required for each processor to obtain the remote data it needs during computation.