[Webkit-unassigned] [Bug 77509] Enable IPP for Biquad filter

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=77509





--- Comment #29 from xingnan.wang at intel.com  2012-02-13 18:14:21 PST ---
(In reply to comment #28)
> (From update of attachment 126449 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=126449&action=review
> 
> >> LayoutTests/webaudio/resources/biquad-testing.js:73
> >> +        a2 = 0;
> > 
> > nit: Can you explain why this is needed now when it wasn't needed without IPP?  When I run this with freq = 0 (and q = 1 and gain = 1) I get b0=b1=b2=0 and a1=-2 and a2=1.  Is it because there's a pole at 1?
> 
> I found you set b0=b1=b2=a1=a2=0 in setLowpassparams() in Biquad.cpp, and did not in the test script, I`ve thought they should be the same. Any reason to treat freq==0 condition differently?
> 
Ray, there is nothing special for IPP. So I won`t change the code here, if it should be different for condition freq=0.
> >> LayoutTests/webaudio/resources/biquad-testing.js:386
> >> +    var a2 = filterCoef.a2;
> > 
> > Sorry about that.  I'm surprised this worked at all before.
> 
> All right, it seems JS is smarter than C++ ~
> 
> >> Source/WebCore/platform/audio/Biquad.cpp:256
> >> +    setNormalizedCoefficients(m_b0, m_b1, m_b2, 1, m_a1, m_a2);
> > 
> > It seems odd that we first set the values directly on the member variables in this method, then call setNormalizedCoefficients().
> > 
> > Can we improve this in this method and setHighpassParams() and any similar methods to be more consistent and similar to what we do in setLowShelfParams()
> 
> That`s right, I`ll refine it.

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