New AVX2 Code Helps FFmpeg With VVC Decoding Performance

A new AVX2 code path for FFmpeg's VVC decoding "vvcdec" is helping provide significant speed-ups for CPU-based H.266 decoding.

Open-source developer Nuo Mi has landed new optical flow AVX2 code for FFmpeg's VVC decode support within libavcodec. Up to 15% better performance is reported thanks to this optional AVX2 usage.

Advanced Vector Extensions 2 as a reminder has been around since Intel Haswell and AMD Excavator, allowing any remotely recent AMD / Intel processor to benefit from this faster VVC decode.

Those curious about this new AVX2 optimization for the VVC decode in FFmpeg can see this commit for the details.