整数变换在VLIW DSP上的优化与仿真

支持超长指令字（Very Long Instruction Word，VLIW）DSP的应用越来越广泛，而H．264视频压缩算法在向此类DSP平台上移植时，由于程序结构设计不合理、数据结构冗余等原因，导致缓存命中率下降，DSP的利用率降低。因此研究最新的视频压缩算法H．264／AVC在此类DSP上的优化就显得越发重要。文中介绍了VLIW DSP的体系结构特点，并研究分析了基于这些特点而提高H．264中整数变换模块运行速度的优化方法，最后在VLIW DSP平台上利用这些方法对H．264的整数变换模块进行了优化与仿真。优化后的整数变换速度为优化前的2倍以上。

Optimization and Simulation of Integral Transform Based on VLIW DSP

The VLIW supported DSP is more and more widely used. However, because of inappropriate design of program structure and the use of redundant data structure, when H.264 video compression algorithm is transplanted to these kinds of DSP platforms, the cache hit ratio is low and the efficiency of DSP decreased. So the research of the optimization methods of the newest video compression algorithm H. 264/AVC on these kinds of DSP becomes more important. In this paper,the architecture of VLIW DSP is introduced, and based on the characters, the optimization methods are analyzed to accelerate the speed of integral transformation module of H. 264. At last, the H. 264 integral transformation module is optimized and simulated on the platform of VLIM DSP. The speed of integral transformation was successfully doubled after optimization.