通用融合浮点点积单元(UFDP)

H. Saleh, B. Mohammad
{"title":"通用融合浮点点积单元(UFDP)","authors":"H. Saleh, B. Mohammad","doi":"10.1109/IDT.2013.6727146","DOIUrl":null,"url":null,"abstract":"A universal floating-point fused dot-product (UFDP) unit is presented that is capable of performing floating-point multiplication and addition or subtraction operations on two pairs of data, floating-point multiply add operation on three data items, floating-point multiplication of two data items and floating-point addition or subtraction of two data items. The proposed UFDP unit could be used as the only floating-point primitive a processor needs, where it easily can replace a floating-point adder, multiplier, fused multiply-add and a fused dot-product unit. Due to the dominance of multiply-add and dot-product operations in audio and video processing algorithms this unit can lead to substantial performance enhancement. The synthesis results using 32nm industry standard-cell library shows that the proposed architecture occupies an area of approximately twice a basic floating-point multiplier, can perform all operations in less 160% of a basic floating-point operation and consumes 80% more than a basic floating-point multiplier.","PeriodicalId":446826,"journal":{"name":"2013 8th IEEE Design and Test Symposium","volume":"49 4 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2013-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"Universal fused floating-point dot-product unit (UFDP)\",\"authors\":\"H. Saleh, B. Mohammad\",\"doi\":\"10.1109/IDT.2013.6727146\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"A universal floating-point fused dot-product (UFDP) unit is presented that is capable of performing floating-point multiplication and addition or subtraction operations on two pairs of data, floating-point multiply add operation on three data items, floating-point multiplication of two data items and floating-point addition or subtraction of two data items. The proposed UFDP unit could be used as the only floating-point primitive a processor needs, where it easily can replace a floating-point adder, multiplier, fused multiply-add and a fused dot-product unit. Due to the dominance of multiply-add and dot-product operations in audio and video processing algorithms this unit can lead to substantial performance enhancement. The synthesis results using 32nm industry standard-cell library shows that the proposed architecture occupies an area of approximately twice a basic floating-point multiplier, can perform all operations in less 160% of a basic floating-point operation and consumes 80% more than a basic floating-point multiplier.\",\"PeriodicalId\":446826,\"journal\":{\"name\":\"2013 8th IEEE Design and Test Symposium\",\"volume\":\"49 4 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2013-12-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2013 8th IEEE Design and Test Symposium\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/IDT.2013.6727146\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2013 8th IEEE Design and Test Symposium","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/IDT.2013.6727146","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1

摘要

提出了一种通用浮点融合点积(UFDP)单元,能够对两对数据进行浮点乘法和加减运算,对三项数据进行浮点乘法和相加运算,对两项数据进行浮点乘法和对两项数据进行浮点加减运算。所提出的UFDP单元可以用作处理器所需的唯一浮点原语,它可以很容易地取代浮点加法器、乘法器、融合乘加和融合点积单元。由于乘加和点积运算在音频和视频处理算法中占主导地位,该单元可以大大提高性能。使用32nm工业标准单元库的合成结果表明,该架构占用的面积约为基本浮点乘法器的两倍,可以在不到基本浮点运算的160%的时间内完成所有运算,并且比基本浮点乘法器多消耗80%。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Universal fused floating-point dot-product unit (UFDP)
A universal floating-point fused dot-product (UFDP) unit is presented that is capable of performing floating-point multiplication and addition or subtraction operations on two pairs of data, floating-point multiply add operation on three data items, floating-point multiplication of two data items and floating-point addition or subtraction of two data items. The proposed UFDP unit could be used as the only floating-point primitive a processor needs, where it easily can replace a floating-point adder, multiplier, fused multiply-add and a fused dot-product unit. Due to the dominance of multiply-add and dot-product operations in audio and video processing algorithms this unit can lead to substantial performance enhancement. The synthesis results using 32nm industry standard-cell library shows that the proposed architecture occupies an area of approximately twice a basic floating-point multiplier, can perform all operations in less 160% of a basic floating-point operation and consumes 80% more than a basic floating-point multiplier.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信