Le 10/10/2024 à 22:45, Michael Niedermayer a écrit :
This makes a 16bit RGB raw sample 25% faster at a 2% loss of compression with 
rawlsb=4

Please test and comment, especially if you are an archivist caring about 
compression and speed
Id like to know if this is a direction (that is trading compression against 
speed) that
is wanted

This is something I was having in mind for a long time but never spent the time to implement it, great that this is done!
Definitely something useful as lsb at 16-bit can not really be compressed.


[...]
+//         put_symbol(c, state, f->rawlsb, 0);

I was having in mind to add the count of zeroed lsb and to not store them, 
sometimes the lsb are upscale with zeroes but we need to keep the hint that it 
is 16-bit.


[...]

In practice the 16-bit version of the compressor could be used also for 16-bit RGB content, it may be a bit faster especially on huge frames with a bit less memory usage, but this adaptation can be made in a later step.


_______________________________________________
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".

Reply via email to