摘要
在科学研究和实时信号处理中,需要进行大量的带状矩阵乘法运算.在分析经典二维,串行心动结构矩阵乘法器的基础上,提出了基于三维的、串并行数据传输的新型结构.改进后的矩阵乘法器具有高度的并行性和流水线的特点,均衡阵列负载,提高运算速度和优化资源的利用率.
A large quantity of band-matrix multiplier calculations must be executed in scientific research and real-tlme signal processing. After analyzing the classic 2-D and serial transportation systolic array architecture, a novel structure matrix multiplier based on 3-D square and serial-parallel mixed transportation was proposed. The matrix multiplier of the new structure possesses advantages of parallel processing and flow computation. It can balance the load and increase the speed of calculation and the usage of resource.
出处
《复旦学报(自然科学版)》
CAS
CSCD
北大核心
2005年第6期989-992,共4页
Journal of Fudan University:Natural Science
关键词
半导体技术
矩阵乘法器
并行运算
心动阵列
semiconductor technology
matrix multiplier
parallel calculation
systolic array