[Webkit-unassigned] [Bug 77856] Have the DynamicsCompressorNode support multi-channel data

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=77856





--- Comment #7 from Chris Rogers <crogers at google.com>  2012-02-08 10:45:03 PST ---
(In reply to comment #5)
> > 
> > Hi Raymond, so far I've tested:
> > 0.5f * (abs(undelayedL) + abs(undelayedR));
> > 
> > which appears to work well on a variety of real-world source material.  Please give me a little time to test your second approach (using max).  I understand your reasoning, but want to be very careful about how the change translates to real-world source material which I've been testing with the algorithm for a long time.  I'll get back with the results...
> 
> Hi Chris
> 
> I test the 0.5f * (abs(undelayedL) + abs(undelayedR)) approaching and max approach again today.
> 
> With 3 scenario
> 
> 1: L=sin(t) R=sin(t+PI)
> 2: L=sin(t) R=sin(t+PI/2)
> 3: L=sin(t) R=0
> 
> the 0.5f * (abs(undelayedL) + abs(undelayedR)) not only works for case 1, but not works for case 2 and case 3.
> 
> While the max approach works fine for all these 3 cases.

Hi Raymond, I've had a chance to test the max approach on a variety of source material and it seems to work well.  So, this seems fine to me.  I think it should be safe to land, but I will continue listening carefully as the changes go in...

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.