基于GPU的干涉测量VDIF格式数据编帧方法*

A new data encoding method for VDIF data based on GPU

摘要:

为提高深空测控网 VLBI 基带转换器灵活性和多种数据格式适应能力，设计一种基于 GPU 的 VDIF 格式数据编帧方法。基于 CUDA 并行运算架构开发符合 VDIF 规范的帧头计算模块、多通道单线程编帧模块和多通道多线程编帧模块。为了实现编帧过程效率的优化，设计基于纹理缓存查找表的数据量化方法和基于流式架构的多帧信号异步编帧方法。试验验证表明，设计的编帧方法结果正确，高效可靠，有望为我国深空测控干涉测量数据记录系统参与国际 VLBI 联合观测提供有效的支持。

Abstract:

In order to improve the flexibility of the deep space station VLBI baseband converter and the adaptability of multiple data formats, a data encoding method which conforms to VDIF specification is developed based-on GPU, including frame header computing module, multi-channel single-datathread encoding module and multi-channel multi-datathread encoding module. To optimize the efficiency of encoding process, a data quantization method based on lookup table using texture memory and a multi-frame asynchronous encoding method based on streaming architecture are designed. Experimental verification shows that the encoding method designed in this paper is accurate, efficient and reliable, which is expected to provide effective support for VLBI data recording system of China's deep space station to participate in international VLBI joint observation.