Transplantation And Optimization Of H.264 Encoder On The DSP Platform

H.264 is the latest video coding standard developed by the cooperation of ITU and MPEG. Based on H.263, it makes use of several critical technique, such as advanced intra/inter prediction modes, 4 by 4 integer DCT transfer, motion estimation at 1/4 pixel level, and content-adaptive binary arithmetic coding. As a result, H.264 has a better performance than any other video coding standards, and will have ultimate applications in many fields. However, the cost of the excellent performance is the great increase of computation complexity.Nowadays, along with the development of semiconductor and needs of the market, almost every manufactor has produced high-performance digital signal processors that aim at video appilication, and the C64 series of TI are leaders of these products. The main point of this paper is to transplant and optimize the H.264 encoder described by C language from PC to the DSP platform whose core part is the TMS320C6416 DSP of TI.First, we reviewed the history of video coding standards, including H.26x, MPEG and Windows Media series; also we reviewed the development of DSPs, and made a comparation of 3 representational DSPs.Moreover, we discussed H.264 and the DAM6416 video processing platform in detail. For H.264, we analyzed its characteristics and performance and its several critical techniques. For DAM6416 platform, we introduced its core——TMS320C6416 DSPs, including its hardware architecture, characteristics and performance and emphasizing its high parallelity. Then, we discussed other resourses on the platform and the software integrate development environment CCS.The second half part of this paper concertrated on the transplantation and optimization of the H.264 encoder from PC to DSP platform. The work of transplantation is conducted mainly by the DSP/BIOS, including thread system, the hardware timer, memory configuration and APIs provided by DSP/BIOS. We also discussed the differences between compilers of PC and DSP platform and their effect during the transplantation. Finally, several tests proved that the work of transplantation is successful.And for the optimization, we first removed the redundence of the code and analyzed the result. Then we discussed the optimizing functions and results of the compiler of CCS in detail. Moreover, we considered the optimizing means for cache and memory, whose purpose is to improve the hit rate of L2 cache and is proved by tests. Finally, we introduced the optimizing ways based on the assembly instructions. In the last chapter, we gave a discussion to the work of future.