基于功能和数据划分的H.264并行解码算法

H.264 Parallel Decoding Algorithm Based on Function and Data Partition

针对H.264视频解码算法的并行模块选择、划分及解码速度优化等问题,面向TilePro64多核平台,提出一种可扩展的H.264并行解码算法。对该算法的内部功能模块进行整合和划分,根据核间数据的依赖关系,动态分配功能模块及优化算法并行效率。实验结果表明,该算法在解码效率、多核并行程度、解码时延等方面均有较好性能,相比传统并行解码算法,其并行加速比提高约25%。

An extensible parallel decoding algorithm for TilePro64 multicore platform is proposed to improve the choice and partition of the parallel function module,optimize the speed of the parallel decoding.The algorithm organizes the function modules of the algorithm and divides them into several function groups.The algorithm assigns the function module dynamically according to the relationship among the data of the processors to optimize the efficiency of the parallel decoder.Experimental results indicate that the algorithm improves the efficiency and the parallelization,reduces the latency time,achieves a 25% performance increase of speedup compared to the traditional parallel algorithm.