Le sunnuntaina 7. heinäkuuta 2024, 14.44.47 EEST Rémi Denis-Courmont a écrit : > The function is exactly identical to the C reference, only with the > constant propagated manually. It does not optimise anything.
There are presently no (other) implementations for `H264DSPContext.h264_chroma_dc_dequant_idct`, which is used by H.264 and SVQ3 decoders. I don't know if we should keep the pointers anyway or just hard-code the calls to the C implementation. I am not sure if it would be possible to marginally optimise the functions with scalar product or matrix extensions... at least those extensions that do support widening 16x16 multiplications. (This is unfortunately not a given. With the AI/ML hype,16-bit floats, 8-bit and/or 4-bit integers are the more commonly supported types.) -- Rémi Denis-Courmont http://www.remlab.net/ _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
