GG says to try CFLAG="-O2 g" make -j8 rebuild_all instead. If that does not work, if you have the window output, email that.
I attach the file with the messages from the terminal. I tried to do this test just to see if it improved the slowdown. But it's not important, I would say not to waste time on it. With my normal installation I have the following results: https://www.dropbox.com/s/h8p4tyvx2snwiyi/transition.mp4?dl=0 I summarize the results: mp4 --> wipe --> png --> dissolve --> 3gp/cut/3gp --> slide --> mp4 30 29 29 21 (png?) 30 28 30 29 30 Fps I can tell I don't have any particular problems. Repeating the test by activating "Play every frame" and "autocolor assests" to stress the system more, the result is similar. Maybe 1 fps less, but there is no noticeable difference. Repeating the test using the compile with modified CFLAGS (-O2) and then starting from: # /path/bin/./cin instead of by: # /path/cinelerra/gdb ./ci (which as you can see from the attachment does not work for me) I still get similar results: I rarely go down to 29 fps. NOTE: in the edit "neve.3gp" I cut a part of the shot to shorten it. The result is to have a cut in the middle of that shot. When the cursor passed over this cut I had a slowdown similar to those in the presence of transitions. (little, as I have already said, about 1 fps, you see in the video at 15s). So I wonder if the slowdown is due to the edges of the edits instead of the transitions. In fact, the transitions are always on the edges of the edits, of course.