{"title":"一个面向对象的线程包的性能","authors":"Jon Faust, H. Levy","doi":"10.1145/97945.97979","DOIUrl":null,"url":null,"abstract":"Presto is an object-oriented threads package for writing parallel programs on a shared-memory multiprocessor. The system adds thread objects and synchronization objects to C++ to allow programmers to create and control parallelism. Presto's object-oriented structure, along with its user-level thread implementation, simplifies customization of thread management primitives to meet application-specific needs.\nThe performance of thread primitives is crucial for parallel programs with fine-grained structure; therefore, the principal objective of this effort was to substantially improve Presto's performance under heavy loads without sacrificing the benefits of its object-oriented interface. We discuss design and implementation issues for shared-memory multiprocessors, and the performance impact of various designs is shown through measurements on a 20-processor Sequent Symmetry multiprocessor.","PeriodicalId":135062,"journal":{"name":"OOPSLA/ECOOP '90","volume":"81 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1990-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"33","resultStr":"{\"title\":\"The performance of an object-oriented threads package\",\"authors\":\"Jon Faust, H. Levy\",\"doi\":\"10.1145/97945.97979\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Presto is an object-oriented threads package for writing parallel programs on a shared-memory multiprocessor. The system adds thread objects and synchronization objects to C++ to allow programmers to create and control parallelism. Presto's object-oriented structure, along with its user-level thread implementation, simplifies customization of thread management primitives to meet application-specific needs.\\nThe performance of thread primitives is crucial for parallel programs with fine-grained structure; therefore, the principal objective of this effort was to substantially improve Presto's performance under heavy loads without sacrificing the benefits of its object-oriented interface. We discuss design and implementation issues for shared-memory multiprocessors, and the performance impact of various designs is shown through measurements on a 20-processor Sequent Symmetry multiprocessor.\",\"PeriodicalId\":135062,\"journal\":{\"name\":\"OOPSLA/ECOOP '90\",\"volume\":\"81 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"1990-09-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"33\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"OOPSLA/ECOOP '90\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1145/97945.97979\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"OOPSLA/ECOOP '90","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/97945.97979","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
The performance of an object-oriented threads package
Presto is an object-oriented threads package for writing parallel programs on a shared-memory multiprocessor. The system adds thread objects and synchronization objects to C++ to allow programmers to create and control parallelism. Presto's object-oriented structure, along with its user-level thread implementation, simplifies customization of thread management primitives to meet application-specific needs.
The performance of thread primitives is crucial for parallel programs with fine-grained structure; therefore, the principal objective of this effort was to substantially improve Presto's performance under heavy loads without sacrificing the benefits of its object-oriented interface. We discuss design and implementation issues for shared-memory multiprocessors, and the performance impact of various designs is shown through measurements on a 20-processor Sequent Symmetry multiprocessor.