基于异构多核原型芯片的NCS算法并行化  

Parallelization of NCS Algorithm Based on Heterogeneous Multi-core Prototype Chip

作　　者：钱禹 潘红兵[1,2] 何书专[1] 李丽[1,2] 李伟[1,2] 韩峰[1] 

机构地区：[1]南京大学电子科学与工程学院微电子设计研究所,江苏南京210093 [2]江苏省光电信息功能材料重点实验室,江苏南京210093

出　　处：《微电子学与计算机》2014年第4期87-91,共5页Microelectronics & Computer

基　　金：国家自然科学基金(61176024;61006018);高等学校博士学科点专项科研基金项目(20120091110029)

摘　　要：NCS算法(nonlinear chirp scaling,非线性调频变标)可以处理大耦合SAR(Synthetic Aperture Radar)回波,实现精确聚焦,但串行NCS算法的成像时间很难达到实时成像要求.为了提高算法效率,采用子孔径结构的NCS改进算法,在自主设计的NoC(Network on Chip)异构多核原型芯片上并行实现了实时NCS成像算法.与串行算法相比,并行化后可以大大缩短成像时间,通过与单次子孔径的理论计算值对比,得出实际并行效率达到90.06%.NCS algorithm（nonlinear chirp scaling） can deal with the large coupling SAR echo, achieve precise focus. However, the imaging time of the serial NCS algorithm could hardly achieve the requirements of real- time imaging. To improve efficiency of the algorithm, we use the improving algorithm with the sub-aperture in this paper. Based on the NoC heterogeneous multi-core prototype chip, we achieve the parallelization of the NCS real-- time imaging algorithm. Compared with the serial algorithm, parallel algorithm can greatly shorten the imaging time. And compared with the theoretical calculation value of a single sub--aperture operation, the actual parallel efficiency can reach 90. 06%.

关 键 词：NCS算法 NOC 异构多核原型芯片 并行化 

分 类 号：TN402[电子电信—微电子学与固体电子学]