{"title":"A programmable base MDLNS MAC with self-generated lookup table","authors":"Wenjing Zhang, G. Jullien, V. Dimitrov","doi":"10.1109/IWSOC.2004.1319850","DOIUrl":null,"url":null,"abstract":"This paper presents an new architecture for a programmable second base multi-dimensional logarithmic number system (MDLNS) multiply accumulator cell (MAC) using DRAM to store the conversion lookup table (LUT). It uses a direct mapping from nonbinary exponents to binary format with a more than 50% reduction in DRAM size compared to a recently reported architecture. With a simple modification of the DRAM data loading structure, each MAC can build its own LUT specific to the nonbinary exponents of the corresponding MDLNS coefficient in a pipelined data flow. This results not only in a considerable reduction of DRAM size, but also in the elimination of adders normally required for computing non-binary exponents in a data filtering channel.","PeriodicalId":306688,"journal":{"name":"4th IEEE International Workshop on System-on-Chip for Real-Time Applications","volume":"19 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2004-07-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"4th IEEE International Workshop on System-on-Chip for Real-Time Applications","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/IWSOC.2004.1319850","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
This paper presents an new architecture for a programmable second base multi-dimensional logarithmic number system (MDLNS) multiply accumulator cell (MAC) using DRAM to store the conversion lookup table (LUT). It uses a direct mapping from nonbinary exponents to binary format with a more than 50% reduction in DRAM size compared to a recently reported architecture. With a simple modification of the DRAM data loading structure, each MAC can build its own LUT specific to the nonbinary exponents of the corresponding MDLNS coefficient in a pipelined data flow. This results not only in a considerable reduction of DRAM size, but also in the elimination of adders normally required for computing non-binary exponents in a data filtering channel.