Mark Coniglio
TroikaTronix
/ Moderator
Dec 3, 2008, 9:05 AM
Post #4 of 4
(674 views)
Shortcut
|
|
Re: [ursullivision] 1.2.9.8 "slower"
[In reply to]
|
Can't Post
|
|
Dear Eric,
OK -- I was looking at Leatherback 1.2.9c.izz on both the current version 1.2.9.8 that I have in-house (soon to be out) and 1.2.9. I was playing three DV-PAL movies with sound.
Stage size was 1152 x 288 in size (Half of a Triple Width 768 x 576 resolution). (Obviously this is on the main screen, not an external card.) Target Frame Rate in Isadora is 25fps.
This was done on my Mac Book Pro (CoreDuo) with Mac OS 10.5.5, QuickTime 7.5.5
Here's what I measured
DV PAL WITH SOUND:
1.2.9 = 25.0 fps, 160 cycles, 2 vpo
1.2.9.8 = 25.0 fps, 108 cycles, 19 vpo (standard configuration)
1.2.9.8 = 25.0 fps, 110 cycles, 6 vpo (old render method)
DV PAL WITHOUT SOUND:
1.2.9 = 25.0 fps, 160 cycles, 2 vpo
1.2.9.8 = 25.0 fps, 115 cycles, 18 vpo (standard configuration)
1.2.9.8 = 25.0 fps, 130 cycles, 6 vpo (old render method)
Now, as I mentioned on the forum, many things have changed "under the hood" in Isadora since 1.2.9. First, it uses Apple's recommended method for rendering the movies -- to a more modern form of the bitmap image. Second, switching to the "pull" mechanism, where data is pulled by the Projector actor from the actors that supply it, as opposed to pushing data from the Movie Player. And third, the entire timing mechanism that tries to keep the frame rate consistent was redone.
While it is true that 1.2.9 has a higher cycle count, to my eye, 1.2.9.8 has smoother playback. I'm seeing fewer "glitches" than are exhibited in 1.2.9.
For sure, the higher VPO seems to have something to with the new rendering method; In addition, I am measuring the VPO at different points than in 1.2.9, giving a more complete picture as to the true VPO. (In other words, I don't believe the measurement in 1.2.9 was fully accurate.) For the moment, I would not use the VPO as a meaningful measurement.
However, the cycle count does matter. And it is less in 1.2.9.8 than in 1.2.9, and that is something to be investigated. It is most likely a result of changing the way the timing mechanism works to get more accurate frame timing (i.e, the distance between frames) which was very _inaccurate_ in 1.2.9.
As an experiment, I forced Isadora to render using the "old" method (noted above as "force old render method") where the movies are rendered to a bitmap called a GWorld. (Which you're not even supposed to be using GWorlds in 10.5 anyway, as they are warning that this will be removed in the next big release of MacOS --GWorlds remains in 10.5 for compatibility but I have no choice but to move in this direction if Isadora will continue to work with Mac OS in the future.) This lowered the VPO singificantly, and the cycle count went up a bit, but not a lot. I'll look into this to see if in fact I'm doing something "wrong" in the way I'm pulling the data from the movie using Apple newer programming interface.
I think the most important thing to consider is your eyes and ears, in terms of how the result looks and sounds. Second, look at the FPS -- this is a truly meaningful measurement, because it shows when you fall below the target frame rate (which you should match to your source material -- 25 fps in the case of PAL, or 29.97 or 30 fps in the case NTSC.)
I will continue to take a look at your example case and others, as I'm trying to lock down Isadora 1.3. But at the moment the only thing I can suggest if you need for some reason to see a higher cycle count, is to increase the frequency the General Service Tasks popup in the General tab of the Preferences.
Let me know how you make out.
Best Wishes,
Mark
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Mark Coniglio / TroikaTronix
Creator of Isadora
|
MAIN
SEARCH
WHO'S 
LOG