Correction - Not creepy!
Published by André Michelle October 21st, 2008 in newsAfter the last post some of you may think that we are complaining that everything is bad about the new Sound API. Well, this is certainly not true. The new Sound API works - but unfortunately not as expected. Our reaction may be seen as a bit overreacting cause most of you guys won’t run into the same issues we had when the final Flash Player launched last week.
Again the story, better explained
We ask for dynamic sound and we finally got it. Since Adobe is running a public beta phase, we could adopt very early the new Sound API. We implemented it in the Hobnox AudioTool and we were pleased how good it worked. Then we had some new challenges to solve. There are very CPU intensive tasks every audio application needs to do (creating new plugins, redrawing large areas on the screen). Since we do not have Threads we figured out another way to avoid playback gaps while intensive computation are made by the player. First, you need to pass a bunch of samples when the Sound API asks for them. Speaking for the ‘old’ version before the release, we just passed the minimum number of samples, getting a very good latency time of 46ms on a MAC. When we expect a critical process (CPU intensive), we passed the maximum number of samples. The sound card buffer had more data to play and the player more time to perform the intensive computation. The perfect solution actually.
This however changed
With the new changes it is not possible to go back to the minimum number of samples - or in short - it won’t go back to the minimum latency time. It will stay at 830ms on a MAC. So you can see, that our workaround has become useless. Building workarounds on an API running in Beta might be a bad idea. However we were sure that changes would be passed the beta group. Furthermore passing the minimal number of samples to get the lowest latency is no longer stable enough. So we were forced to switch up to 160ms AND getting playback glitches/gaps when intensive computations occur.
Work is not just work
When spending so much time on a proper audio environment in Flash - it should be comprehensive that we felt really wronged when seeing the new changes in the Sound API, done in the last minute before launch.
Peace! Adobe - but next time please tell us earlier. If there is any chance to revert that behavior in a minor update, we would all appreciate this!
I’m trying to get get some more information so we can help avoid issues like this in the future. My understanding is that the prerelease list was notified of this change before the final release. Is that not the case?
Thanks,
Lee
Hi Lee,
Unfortunately not, the final release has been shipped and modified without any notification to the prerelease board.
best,
Thibault
I’m a bit offtopic here, but I want to share an idea that I’ve seen at the latest flex camp : you can use pixelbender for sound computation (and other CPU intensive tasks). you can see what I mean here : http://miti.pricope.com/2008/11/10/playing-with-pixel-bender/
It should be a good performance boost, if you didn’t integrate this already