基于申威421处理器的视频帧间预测优化研究

Research on video interframe prediction optimization based on Shenwei 421 processor

由于X265编码库目前还不适应申威平台的软件生态环境,造成移植到申威处理器上的X265视频编码库性能较低。为了解决这个问题,通过分析X265编码库中视频编码的各个编码模块,提出了基于申威平台的帧间预测模块优化方法,该方法采用4个线程对帧间预测模块进行优化,提升了X265编码库执行效率,同时解决了多线程中存在的线程数据竞争、分配不均衡和迁移耗时的问题,进一步提升了X265编码库的性能。通过对比6个测试序列优化前后的编码转换时间和PSNR,证明了本文方法的有效性与可行性。

As X265 coding library is not suitable for the software ecological environment of Shenwei platform,the performance of X265 video coding library transplanted to Shenwei processor is low.To solve this problem,by analyzing the various coding modules of the video coding in the X265 coding library,an optimization method for the inter frame prediction module based on the Shenwei platform is proposed.This method uses four threads to optimize the inter-frame prediction module to improve the X265 coding library execution efficiency.At the same time,it solves the problems of thread data competition,unbalanced distribution and time consuming migration in multithreading,and as a result further improves the performance of the X265 coding library.By comparing the code conversion time and PSNR of six test sequences before and after optimization,the effectiveness and feasibility of the proposed method are proved.