Dina M. Ellaithy, M. El-Moursy, G. Hamdy, Amal Zaki, A. Zekry
{"title":"用于GPU应用的精确分段均匀逼近对数/反对数转换器","authors":"Dina M. Ellaithy, M. El-Moursy, G. Hamdy, Amal Zaki, A. Zekry","doi":"10.1109/ICM.2017.8268821","DOIUrl":null,"url":null,"abstract":"Logarithmic number system (LNS) is increasingly used for low power arithmetic calculations in graphical processing unit (GPU). LNS requires less hardware. Based on uniform subdivisions and linear-Lagrange interpolation, a novel 8-subdivisions logarithmic/antilogarithmic converters are proposed. Compared with different methodologies, the proposed logarithmic/antilogarithmic converters achieve high accuracy. Also, a significant decrease in area is presented in the comparison between the proposed converters and the prior ones. Hardware implementation using shift-and-add approach is adopted to further simplify the hardware. The novel logarithmic/antilogarithmic converters are implemented and synthesized using 90 nm CMOS technology. Up to 23% and 11% decrease in relative error and area are achieved.","PeriodicalId":115975,"journal":{"name":"2017 29th International Conference on Microelectronics (ICM)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2017-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"3","resultStr":"{\"title\":\"Accurate piecewise uniform approximation logarithmic/antilogarithmic converters for GPU applications\",\"authors\":\"Dina M. Ellaithy, M. El-Moursy, G. Hamdy, Amal Zaki, A. Zekry\",\"doi\":\"10.1109/ICM.2017.8268821\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Logarithmic number system (LNS) is increasingly used for low power arithmetic calculations in graphical processing unit (GPU). LNS requires less hardware. Based on uniform subdivisions and linear-Lagrange interpolation, a novel 8-subdivisions logarithmic/antilogarithmic converters are proposed. Compared with different methodologies, the proposed logarithmic/antilogarithmic converters achieve high accuracy. Also, a significant decrease in area is presented in the comparison between the proposed converters and the prior ones. Hardware implementation using shift-and-add approach is adopted to further simplify the hardware. The novel logarithmic/antilogarithmic converters are implemented and synthesized using 90 nm CMOS technology. Up to 23% and 11% decrease in relative error and area are achieved.\",\"PeriodicalId\":115975,\"journal\":{\"name\":\"2017 29th International Conference on Microelectronics (ICM)\",\"volume\":\"29 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2017-12-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"3\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2017 29th International Conference on Microelectronics (ICM)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICM.2017.8268821\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2017 29th International Conference on Microelectronics (ICM)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICM.2017.8268821","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 3
摘要
对数系统(Logarithmic number system, LNS)越来越多地用于GPU (graphical processing unit)的低功耗运算。LNS对硬件的要求更少。基于均匀细分和线性拉格朗日插值,提出了一种新型的8细分对数/反对数变换器。与不同的方法相比,所提出的对数/反对数转换器具有较高的精度。此外,在所提出的转换器与先前的转换器之间的比较中,面积显着减少。为了进一步简化硬件,采用了shift-and-add的硬件实现方法。采用90纳米CMOS技术实现并合成了新型对数/反对数变换器。相对误差和面积分别减小了23%和11%。
Accurate piecewise uniform approximation logarithmic/antilogarithmic converters for GPU applications
Logarithmic number system (LNS) is increasingly used for low power arithmetic calculations in graphical processing unit (GPU). LNS requires less hardware. Based on uniform subdivisions and linear-Lagrange interpolation, a novel 8-subdivisions logarithmic/antilogarithmic converters are proposed. Compared with different methodologies, the proposed logarithmic/antilogarithmic converters achieve high accuracy. Also, a significant decrease in area is presented in the comparison between the proposed converters and the prior ones. Hardware implementation using shift-and-add approach is adopted to further simplify the hardware. The novel logarithmic/antilogarithmic converters are implemented and synthesized using 90 nm CMOS technology. Up to 23% and 11% decrease in relative error and area are achieved.