SIMD Optimization Season

Written for FFmpeg by Timothy Gu on 2014-02-24

FFmpeg has a large number of hand-written SIMD optimization that utilizes CPU's instruction set extensions like MMX, SSE, and AVX for x86, and NEON for ARM to create a speed boost [1]. Recently we have added various SIMD optimization that causes VP9, DCA, and FLAC decoding to speed up.

Here is a partial list of all the optimization we've added since late last year:

* VP9 decoder: new MMX, SSE2, SSSE3, and AVX optimizations
* DCA decoder: new SSE, SSE2, and SSE4 optimizations
* FLAC 32-bit LPC coder: 50% faster than the C-based code thanks to the new SSE 4 and XOP optimizations
* HEVC decoder: optimizations currently being reviewed

* All NEON optimizations written for regular ARM architecture are ported to the AArch64 64-bit platform

A huge thanks to all the developers who have contributed to the FFmpeg project!


Read all announcements