摘要
视频压缩算法在向DSP(digital signal processing)平台上移植时,大多存在程序结构设计不合理、数据结构冗余等问题,因而会导致缓存命中率下降、DSP的利用率降低.针对这些问题,通过对cache miss、指令跳转等降低DSP效率的不利因素的分析,在H.263的DCT、量化、预测等模块采取重新定义数据结构、改进算法、优化专用指令等一系列措施提高缓存命中率,减少跳转指令,以提高DSP的有效利用率,使得优化后的H.263编码器在保持较高清晰度的情况下,可以对通用中间格式(CIF)的视频序列进行实时处理.
Because of inappropriate design of program structurc and the use of redundant data structure, when video compression algorithm is transplanted to DSP（digital signal processing） platforms, the cache hit ratio is low and the efficiency of DSP decreased. By analyzing the causes of cache miss and frequent jump operations, the data structure was redefined, the arithmetic was improved. The PNX1300＇ s instruction set was used in DCT, Quantization, ME, Coding module of H. 263 to increase the cache hit ratio and eliminate the frequent jump operations. The improved H. 263 encoder can compress CIF-format video sequence at real-time with high definition.
