[FFmpeg-devel] [PATCH] lavc/aarch64: Add pix_abs16_x2 neon implementation

Hubert Mazur hum at semihalf.com
Tue Jul 12 12:15:32 EEST 2022


Thanks for the feedback. I made changes to the patch.
The performance has increased now to ~7 boost compared
to C implementation.

Changes:
- Do not use v8-v15 registers.
- Use urhadd instruction.
- Reorder the instructions to increase performance.

// Hubert



More information about the ffmpeg-devel mailing list