[Webkit-unassigned] [Bug 77509] Enable IPP for Biquad filter
bugzilla-daemon at webkit.org
bugzilla-daemon at webkit.org
Tue Feb 7 10:01:44 PST 2012
https://bugs.webkit.org/show_bug.cgi?id=77509
--- Comment #20 from Raymond Toy <rtoy at chromium.org> 2012-02-07 10:01:44 PST ---
(From update of attachment 125775)
View in context: https://bugs.webkit.org/attachment.cgi?id=125775&action=review
This new patch looks much simpler than the previous one. Thanks for applying this patch over the other one that added tests.
> Source/WebCore/ChangeLog:8
> + Used IIR filter in IPP and imporved ~27% performance in linux.
"Used" -> "Use"
"and imporved..." -> "which improves performance in linux by ~27%"
> Source/WebCore/platform/audio/Biquad.cpp:56
> +
nit: Remove extra blank line.
> Source/WebCore/platform/audio/Biquad.cpp:317
> +
Question: Should ipp normalize the coefficients by dividing everything by a0 like we do for non-ipp? Makes the actual coefficients consistent between implementations. Perhaps this doesn't matter.
> Source/WebCore/platform/audio/Biquad.cpp:324
> + return;
This breaks my expectations on what setNormalizedCoefficients should do and is, I think, incorrect. In setLowShelfParams, there is the call setNormalizedCoefficients(1,0,0,1,0,0). Since a0=1, nothing is done, including setting up the actual filter parameters.
I think 323-324 should be removed. Or at least be sure you still set the filter parameters in this case.
This is probably the source of some of the test failures.
> Source/WebCore/platform/audio/Biquad.h:113
> + Ipp8u* m_buffer;
nit: Is there a more descriptive name for m_buffer?
--
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.
More information about the webkit-unassigned
mailing list