DC Field | Value | Language |
---|---|---|
dc.contributor.author | Choi, BS | ko |
dc.contributor.author | Lee, JA | ko |
dc.contributor.author | Har, Dongsoo | ko |
dc.date.accessioned | 2019-04-15T16:32:30Z | - |
dc.date.available | 2019-04-15T16:32:30Z | - |
dc.date.created | 2013-05-08 | - |
dc.date.issued | 2004 | - |
dc.identifier.citation | ADVANCES IN COMPUTER SYSTEMS ARCHITECTURE, PROCEEDINGS Book Series: LECTURE NOTES IN COMPUTER SCIENCE, v.3189, pp.170 - 184 | - |
dc.identifier.issn | 0302-9743 | - |
dc.identifier.uri | http://hdl.handle.net/10203/255904 | - |
dc.description.abstract | Value predictor predicting result of instruction before real execution to exceed the data flow limit, redundant operation table removing redundant computation dynamically, and asynchronous bus avoiding clock synchronization problem have been proposed as high performance microprocessor design methods. However, these methods increase area cost and power consumption problems because of the larger table for value predictor and redundant operation table, and the higher switching activity in asynchronous bus. To resolve the problems of data tables for value predictor and redundant operation table, we have investigated partial tag and narrow-width operand methods, which have been recently proposed separately and present an efficient update method for value predictor and a table organization method for redundant operation table, respectively. To reduce excessive switching activity of asynchronous bus, we also propose a bus encoding method using frequent value cache, which reduces the same data transmissions. The proposed three methods - an efficient update method for value predictor, a table organization method for redundant operation table, and a frequent value cache for asynchronous bus - exploit information locality such as instruction and data locality as well as data redundancy. Analysis with a conventional microprocessor model show that the proposed three methods reduce total area cost and power consumption by about 18.2% and 26.5%, respectively, with negligible performance variance. | - |
dc.language | English | - |
dc.publisher | SPRINGER-VERLAG BERLIN | - |
dc.title | High performance microprocessor design methods exploiting information locality and data redundancy for lower area cost and power consumption | - |
dc.type | Article | - |
dc.identifier.wosid | 000224102200015 | - |
dc.identifier.scopusid | 2-s2.0-35048850912 | - |
dc.type.rims | ART | - |
dc.citation.volume | 3189 | - |
dc.citation.beginningpage | 170 | - |
dc.citation.endingpage | 184 | - |
dc.citation.publicationname | ADVANCES IN COMPUTER SYSTEMS ARCHITECTURE, PROCEEDINGS Book Series: LECTURE NOTES IN COMPUTER SCIENCE | - |
dc.contributor.localauthor | Har, Dongsoo | - |
dc.contributor.nonIdAuthor | Choi, BS | - |
dc.contributor.nonIdAuthor | Lee, JA | - |
dc.type.journalArticle | Article; Proceedings Paper | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.