摘要
Turbo码因具有卓越的纠错能力和接近香农理论极限的性能而受到广泛关注。为了满足译码的实时性需求,利用Intel公司的众核处理器的多核并行处理和迭代运算能力,并结合OpenMP在编译层面可以自动将程序并行化的能力,对Turbo译码的CPU程序进行众核移植和并行化。在保证译码性能的同时,使用折线逼近的Log-MAP算法,并在代码层面进行调整和优化。针对实际卫星信号,采用基于数据的MIC多线程并行处理模式,相比利用CPU进行处理,使用Intel Xeon Phi Coprocessor 7120众核处理器在计算速度上有将近60倍的提升,并且可以实现8路突发信号的实时处理。
Turbo-code has received wide attention due to excellent error correction ability and performance near to the Shannon limit. In order to satisfy the real-time requirement of decoding, the multi-core parallel processing and iterative computing capabilities of Intel′s many integrated core(MIC) processor and OpenMP′s ability to automatically parallelize programs are used to decode Turbo-code. Then MIC migration and parallelization are used to the CPU program. While ensuring the decoding performance, the Log-MAP algorithm with the polyline approximation is used and adjusted and simplified at the code level. For the actual satellite signal, using the data-based MIC multi-thread parallel processing mode. Compared to the CPU processing, the calculation speed of the Intel Xeon Phi Coprocessor 7120 Many Integrated Core processor has nearly 60 times improvement. What is more, it can achieve 8 Real-time processing of the burst signal.
作者
赵梦伟
陈永锋
刘凯
孙超群
Zhao Mengwei;Chen Yongfeng;Sun Chaoqun(Key Laboratory of Specialty Fiber Optics and Optical Access Networks, Shanghai University,Shanghai 200444,China;Shanghai Branch of South-West Electron and Telecom Technology Institute,Shanghai 200434,China)
出处
《电子测量技术》
2019年第1期126-131,共6页
Electronic Measurement Technology
