面向视频处理的高效二维流存储系统

An Efficient 2-D Stream Memory System for Video Processing

为了解决高度并行的视频处理中存储系统的瓶颈问题,设计了一种高效的基于流的二维存储系统.该系统完成了二维逻辑空间到物理上多路并行存储器模块的映射.在二维逻辑空间中的每一个数据阵列都可以根据给出的基地址、二维偏移量、长度、数据粒度来进行灵活的流访问.数据在物理存储器上进行了交织存放以支持行阵列和列阵列的同时访问.该交织算法在之前的交织算法基础上做了面向流访问的改进.实验结果表明,在实时视频处理中,所述的二维流存储系统可以减少平均约32.0%的存储器访问率,以及25.4%的实时处理所需要的时钟周期数.

An efficient stream memory system for video processing is proposed in this paper to remove the bottleneck of memory system in parallel video processing. The system maps a 2- D logic space onto physical parallel memory modules. Each array of data elements in a 2-D space can be streaming accessed by given base address, 2-D offset, and granularity. Data are interleaved in the memory to support both horizontal and vertical array access. The interleave scheme is improved for streaming access based on the previous interleave schemes. The experimental results show that the proposed stream memory system could reduce 32.0% of the memory access rate and 25.4% of the execution cycle counts in real time video processing.