Optimization Of H.265/HEVC Decoder On ARMv8 Architecture

Compared to H.264/AVC,H.265/HEVC provides a more flexible, reliable and stable video codec. H.265/HEVC can improve the compression efficiency of the video greatly while the H.265/HEVC decoder needs more complex process and higher performance. ARMv8, which is the new generation of mobile CPU architecture proposed by ARM, opens a new era of mobile computing by its 64-bit architecture. The optimization of the H.265/HEVC decoder on ARMv8 can make the playback of H.265/HEVC video more smoothly and more clearly on mobile devices. It can promote the promotion and popularization of H.265/HEVC in the mobile network video and provide a great significance to mobile video industry application platform.This topic focuses on the basic process and key technologies of H.265/HEVC decoding, analyses the decoding modules to find the key parts to optimize. The key parts are IDCT transform and Motion Compensation. The idea is using the AArch64 to optimize the overall to get the final effects after using the NEON first to optimize the parallel modules of the IDCT transform and Sub pixel interpolation part. This paper aims to optimize the HEVC coder in FFmpeg and uses the NEON and AArch64 instruction set to achieve specific optimization scheme. Finally iPhone 6 plus, which is with the ARMv8 Architecture, is used to test the optimized decoder by playing many videos.The result shows that, the optimization based on NEON can improve the performance of H.265/HEVC decoding about 30% while the AArch64 can help the HEVC decoder after NEON optimization to improve about 10%.