H.264 Baseline Decoder Software Optimization

• Main Authors: Xian-Liang Chen, 陳憲諒
• Other Authors: Jen-Ming Wu
• Format: Others
• Language: en_US
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/84177691927811562563

碩士 === 國立清華大學 === 通訊工程研究所 === 96 === In order to execute the application programs in the embedded Digital Signal Processor (DSP) core system platform in real-time, we need to do the software optimization. In the thesis, we adopt H.264/AVC source code as our application program. H.264/AVC is a new video coding standard proposed by ITU-T, which can achieve high compression efficiency by adding new characteristic coding functions. Therefore, the H.264/AVC program will increase computation time and computation complexity. Due to the fact, the system platform can’t display H.264/AVC video in real-time. We need to optimize the source code by software analysis and optimization methods to increase the program speed and code size to meet the system platform performance and memory resource. First we use Visual C++ to profile the H.264/AVC open source code to make the statistics about the hit counts of functions and the longest execution time by a function. We integrate the intrinsic libraries and remove unnecessary functions and variables and perform software optimization methods to optimize the H.264 source code. In additional, we have to co-simulate the software development environment and hardware develop environment to make sure that the DSP core can run the H.264 binary code generated by the software toolchain correctly. After optimizing the H.264 source code, we can estimate the required real-time DSP clock rate simulated by software and compare the actual DSP core to achieve the real-time objective.