Optimization Of VP8Decoder On Cortex-A8Platform

With the rapid development of mobile platform, demand of video consumption frommobile intelligent device users is becoming bigger and bigger. Hardware decoding canmeet the requirements of the smooth play in performance, but because of limitation ofcoding mode and package format, it can not meet the needs of the user’s diversification.So using of mobile intelligent device’s hardware characteristics to optimize the softwaredecoding for users to get better video play experience is required.VP8is a video format that is proposed by Google against the next generation networkstandard. It has good network compatibility, extensive equipment adaptability. Lowperformance equipment can also playback well, so it can meet mobile intelligent deviceusers’ needs better.After a few generations of renewal and development, the ARM processor has becomethe most mobile equipment processors’ choice. Among them, the Cortex-the A8is themost widely used ARM processor architecture.Cortex-A8with NEON coprocessor is based on the ARMv7instruction set. It canaccelerate multimedia applications effectively. Based on the Cortex-the A8platform, tooptimize VP8video decoding using NEON has a wide prospect of market.After analyzing the Cortex-A8processor architecture, including instruction set,assembly line characteristics and some new technologies, NEON coprocessor architectureand technical features is researched specially. At the same time, researching the VP8codecstandards, and studies the algorithm of module that can use SIMD technique to accelerateconcluding inverse quantization and inverse transform module, motion compensationmodule, loop filter module. Then the NEON optimization is implemented with hand-writtenassembly. Finally, the optimized decoder is tested by open source codecs FFmpeg oniPhone4. Test results indicate that, NEON optimization can make the whole performanceof VP8decoding increase40%at least, more than70%at most.