| Recently, with the increment of the need of high-definition video, the video decoding technique has been developed rapidly. The compression rate and the definition of the videos are both becoming much higher, but meanwhile the mathematical and logical computation for decoding is more and more large. If the video player of the computer directly use CPU to decode the video, the CPU utilization rate would rise sharply which will lead to the fact that the computer cannot affluently play high-definition video, or even don’t work at all. To solve such a problem, part of the decoding job of CPU is offloaded to GPU so that the CPU utilization rate can be reduced. At the same time the system performance would be improved and the energy cost will be decreased.This paper is about the decoding technique for videos of high-rate definition on the domestic processor of FT. The FT CPU has the high ability on multi- thread processing, but meanwhile it don’t performance well with single-core, and it don’t support the instruction of multi- media. The traditional software video decoding technique can’t work well on the platform of FT CPU. Thus, implementing a scheme of playing high-definition video on the platform of FT CPU is contributing to the spreading and the improving of video decoding technique on the domestic processor of FT.First, targeting at the hardware decoding for videos of high-rate definition, combing with the multi- core feature of FT1000 A CPU, we use multi- thread technology to optimize the high-definition video decoding procedure, and convert the video decoding procedure from serial to parallel. According to the different type, it can be sorted into function parallel and data parallel, and have implemented the multi- thread decoding technique scheme on the platform of domestic FT processor.Then, combining the UVD video hardware decoding technique of universal engine technology with the VDPAU hardware acceleration interface, we propose a GPU high-definition video decoding technique scheme for the platform of domestic FT processor. The scheme achieve the goal of supporting multi- format high-definition video decoding acceleration. Based on the framework above, and by solving the problem of size endian convertion and page table length alignment, we achieve the goal of hardware decoding playing the high-definition videos with low cost and power consumption upon the platform of FT CPU.Last, we test the function and performance of the high-definition video decoding scheme proposed in this paper under the FT1000 A system. The experimental results demonstrated that the multi-thread decoding technique on FT CPU can improve the performance more than twenty percent comparing with the software decoding method on single core. But the CPU utilization rate still upon fifty percent when decoding 23 fps and 1080 P videos, the optimizing of performance is not satisfying. When using the GPU hardware decoding scheme upon VDPAU to play 1080 P high-definition videos, the CPU utilization rate can be controlled under thirty percent. |