Glad you ran the tests with interesting results.  Just guessing, but probably the most likely cause of the difference is in "seeking".  There are patches to solve some seeking problems that occurred in specific user files early on (for example patch3).  Also, additional pixel formats were added (patchZ1) and a fix to add YUV420P as valid for one of the graphics hardware acceleration option (I think it was a gopro video but do not really remember).

On Fri, Jun 2, 2023 at 10:51 AM Andrea paz via Cin <cin@lists.cinelerra-gg.org> wrote:
Taking advantage of the fact that I compiled a version that uses
system (external) ffmpeg, I did some comparative testing with CinGG
appimage which uses an internal ffmpeg instead.
I report the results:

Source: blender movie "Tears of steel" 1080p, webm
(Used an interval of the first 1440 frames via In/Out Points)

mp4; preset h264.mp4
ffmpeg internal --> render at 117.4 fps (25.4MB)
ffmpeg external --> render at 175.3 fps (22.8MB)

mxf; preset DNxHR 444
ffmpeg internal --> render at 92.9 fps (1.8GB)
ffmpeg external --> render at 120.5 fps (1.8GB)

I had already noticed that ffmpeg from the command line (external) was
faster than ffmpeg inside CinGG; but now we see that even external
ffmpeg called up inside CinGG is still faster than internal ffmpeg.
What could be the cause of this difference?
--
Cin mailing list
Cin@lists.cinelerra-gg.org
https://lists.cinelerra-gg.org/mailman/listinfo/cin