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

[bugzilla-daemon at webkit.org]

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





--- Comment #15 from Chris Rogers <crogers at google.com>  2012-01-10 18:10:16 PST ---
(In reply to comment #14)
> (In reply to comment #13)
> > (In reply to comment #12)
> > > Created an attachment (id=121794)
 --> (https://bugs.webkit.org/attachment.cgi?id=121794&action=review) [details] [details] [details]
> > > Patch
> > > 
> > > Update the patch as the convolution layout test updated, new result:
> > > Tests ConvolverNode processing a mono channel with mono impulse response.
> > > 
> > > On success, you will see a series of "PASS" messages, followed by "TEST COMPLETE".
> > > 
> > > PASS Initial latency of convolver is silent.
> > > PASS Triangular portion of convolution is correct.
> > > FAIL First part of tail of convolution is not sufficiently small: 0.00244140625 dB
> > 
> > This is way too large.  This corresponds to a max tail value of about 44100.  (0 dB = 1, relative to max of the triangle.)  The expected value should be less than -146 dB.  Something is messed up.  Perhaps the test is wrong?
> 
> I think the "tail1Max" of the updated patch is exactly the same as last patch, with the threshold value 0.002 in the old layout test which should be -53.98 dB ( 20*log10(0.002) ) in dB format. But in your updated layout test the value is down to -146.7 dB,  any reason to change the threshold value?
> 
> Also, the error output in your updated test script :
> testFailed("First part of tail of convolution is not sufficiently small: " + tail1Max + " dB");
> should it be
> testFailed("First part of tail of convolution is not sufficiently small: " + tail1MaxDecibels+ " dB");
> ?

Yes, it looks like the layout test itself needs to be fixed as you recommend (to report a failure with tail1MaxDecibels).  Once you fix that, then you can find the true error in dB.  It's not -53.98 dB as you mention because it's the dB difference computed from the refMax if you see how tail1MaxDecibels is computed.

I recommend you fix the layout test and also adjust the threshold so that your code passes (assuming it's better than -100 dB or so) and upload a new patch with that fix included.

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