[Webkit-unassigned] [Bug 98131] Add ARM-NEON support to VectorMath in WebAudio

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=98131





--- Comment #10 from Gabor Rapcsanyi <rgabor at webkit.org>  2012-10-04 04:03:34 PST ---
(In reply to comment #8)
> (In reply to comment #6)
> > (In reply to comment #4)
> > > (From update of attachment 166684 [details] [details] [details])
> > > View in context: https://bugs.webkit.org/attachment.cgi?id=166684&action=review
> > > 
> > > Looks good.  Do you have any measurements to show what kind of improvements these instrinsics make?  I'm just curious.
> > > 
> > 
> > Yes I have made independent tests on the functions.
> > The vector size was 1000 with float numbers and I ran it 100000 times.
> > The average results:
> > 
> > vsma:
> > REF  elasped time is 3747 ms
> > NEON elasped time is 3642 ms
> 
> Thank you very much for the measurements.
> 
> There wasn't much difference here.  Do you have any ideas on why there was very little improvement?
> 

Yes that was strange for me as well so I played with it a little and than I realized that I forgot to use -O2 when I compiled it.
So the results with -O2 on the same tests:

vsma:
REF  elasped time is 1106 ms
NEON elasped time is 458 ms

vsmul:
REF  elasped time is 803 ms
NEON elasped time is 573 ms

vadd:
REF  elasped time is 712 ms
NEON elasped time is 361 ms

vmaxmgv:
REF  elasped time is 803 ms
NEON elasped time is 217 ms

vsvesq:
REF  elasped time is 501 ms
NEON elasped time is 250 ms

vmul:
REF  elasped time is 805 ms
NEON elasped time is 363 ms

zvmul:
REF  elasped time is 1657 ms
NEON elasped time is 953 ms


Sorry for the confusion.

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