{"title":"高性能、功率感知处理器的检查点替代方案","authors":"A. Moshovos","doi":"10.1109/LPE.2003.1231886","DOIUrl":null,"url":null,"abstract":"High performance processors use checkpointing to rapidly recover from branch mispredictions and possibly other exceptions. We demonstrate that conventional checkpointing becomes unattractive in terms of resource and power requirements for future generation processors. We propose out-of-order checkpoint release and checkpoint prediction, two alternatives that require significantly less resources and power while maintaining high-performance. We demonstrate their utility at the register alias table (RAT). Our methods reduce the number of RAT checkpoints to 1/3 (from 48 down to 16) for an aggressive, 8-way superscalar processor with a 256-entry instruction window. Using a 0.18 /spl mu/m process model we estimate that RAT power is reduced by 24%.","PeriodicalId":355883,"journal":{"name":"Proceedings of the 2003 International Symposium on Low Power Electronics and Design, 2003. ISLPED '03.","volume":"10 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"18","resultStr":"{\"title\":\"Checkpointing alternatives for high-performance, power-aware processors\",\"authors\":\"A. Moshovos\",\"doi\":\"10.1109/LPE.2003.1231886\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"High performance processors use checkpointing to rapidly recover from branch mispredictions and possibly other exceptions. We demonstrate that conventional checkpointing becomes unattractive in terms of resource and power requirements for future generation processors. We propose out-of-order checkpoint release and checkpoint prediction, two alternatives that require significantly less resources and power while maintaining high-performance. We demonstrate their utility at the register alias table (RAT). Our methods reduce the number of RAT checkpoints to 1/3 (from 48 down to 16) for an aggressive, 8-way superscalar processor with a 256-entry instruction window. Using a 0.18 /spl mu/m process model we estimate that RAT power is reduced by 24%.\",\"PeriodicalId\":355883,\"journal\":{\"name\":\"Proceedings of the 2003 International Symposium on Low Power Electronics and Design, 2003. ISLPED '03.\",\"volume\":\"10 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"1900-01-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"18\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Proceedings of the 2003 International Symposium on Low Power Electronics and Design, 2003. ISLPED '03.\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/LPE.2003.1231886\",\"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 of the 2003 International Symposium on Low Power Electronics and Design, 2003. ISLPED '03.","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/LPE.2003.1231886","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Checkpointing alternatives for high-performance, power-aware processors
High performance processors use checkpointing to rapidly recover from branch mispredictions and possibly other exceptions. We demonstrate that conventional checkpointing becomes unattractive in terms of resource and power requirements for future generation processors. We propose out-of-order checkpoint release and checkpoint prediction, two alternatives that require significantly less resources and power while maintaining high-performance. We demonstrate their utility at the register alias table (RAT). Our methods reduce the number of RAT checkpoints to 1/3 (from 48 down to 16) for an aggressive, 8-way superscalar processor with a 256-entry instruction window. Using a 0.18 /spl mu/m process model we estimate that RAT power is reduced by 24%.