[Webkit-unassigned] [Bug 75528] Optimize the multiply-add in Biquad.cpp::process

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=75528





--- Comment #19 from xingnan.wang at intel.com  2012-02-12 23:26:19 PST ---
(From update of attachment 126465)
View in context: https://bugs.webkit.org/attachment.cgi?id=126465&action=review

>> Source/WebCore/platform/audio/Biquad.cpp:121
>> +        "addsd    %%xmm4,  %%xmm5\n\t" // x*b0 + (x2*b2+x*b0)
> 
> Doesn't xmm4 contain a1*y1 here so the result is a1*y1 + (x2*b2+x*b0)?

Sorry for that, it should be a1*y1 + (x2*b2+x*b0).

>> Source/WebCore/platform/audio/Biquad.cpp:137
>> +        :"eax", "edx", "ecx"
> 
> Is it necessary to add all of the xmm registers to this clobber list?

OK, it`s better to add all the xmm registers to the clobber-list, through there is little possibility of conflict of xmm registers here.

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