This paper emphasizes on the linear assembly optimization of some time-consuming key modules such as half pixel interpolate, sum of absolute difference (SAD) and so on.

• 这些方法提高了程序代码的并行性和存储器的访问效率,其中重点介绍了算法中比较耗时的半像素插值、绝对误差和(SAD)等关键模块的线性汇编优化。