[Webkit-unassigned] [Bug 75522] Enable IPP for FFTFrame

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=75522





--- Comment #6 from xingnan.wang at intel.com  2012-01-04 17:49:25 PST ---
(In reply to comment #5)
> Thanks for the patch!  From what I understand, IPP is among the fastest, so it's great to have it as an option.  I'll have a closer look in a little while, but I wanted to ask about testing...
> 
> The timing of this patch is very good, since Ray is about to add a new very important layout test which will (indirectly) test if this code is working correctly or not:
> https://bugs.webkit.org/show_bug.cgi?id=75126
> 
> Have you tested this code in any way?  For example, have you tried running a Web Audio API demo such as:
> http://chromium.googlecode.com/svn/trunk/samples/audio/convolution-effects.html

Yes, I did some test for FFTFrameIPP.
For the correctness, I tried to run http://chromium.googlecode.com/svn/trunk/samples/audio/convolution-effects.html
but it always failed to load the page (maybe network reason), so wrote a simple test case using ConvolverNode and it was OK.
Also I tried to run the drum machine (http://chromium.googlecode.com/svn/trunk/samples/audio/shiny-drum-machine.html), it works very well.
For the performance, I roughly did a function test in ConvolverNode::process(). 
It boosted about 20% if we used accurate DFT in IPP(ippAlgHintFast).
It boosted about 100% if we used fast DFT in IPP(ippAlgHintAccurate).
All tests in 
Intel(R) Core(TM) i7 CPU 870  @ 2.93GHz 
Ubuntu 11.10

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