[sldev] Wind Sound CPU usage.
Aimee Trescothick
aimee-sldev at ama-zing.co.uk
Wed Mar 5 05:58:20 PST 2008
Removing the damping causes audibly abrupt changes, almost clicks,
when starting and stopping movement. Try hovering in place and then
flying up or down a short distance.
Looking at the frequency spectrum of the final wind noise, most of
what's generated is being filtered out, so a lot of work is being
done for nothing. Just tried a quick experiment generating it at
14.7k and then interpolating groups of 3 samples, which seems to work
OK, giving a roughly similar spectrum.
I've attached the rough patch of what I was playing with, it's a bit
ugly and not quite right as there's a little crackle creeping in, but
shows what I'm on about :)
Aimee.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: light-wind.patch
Type: application/octet-stream
Size: 1995 bytes
Desc: not available
Url : http://lists.secondlife.com/pipermail/sldev/attachments/20080305/1df90c5c/light-wind-0001.obj
-------------- next part --------------
On Mar 5, 2008, at 09:17, Jason Giglio wrote:
>
> https://jira.secondlife.com/browse/VWR-5308
>
> I've written a optimization/cleanup patch for the wind noise
> generator. Got it down from 5.5% to 2.25%.
>
> Callum, I was able to move most of the filter code out of the inner
> loop by getting rid of the gain/frequency damping. I can't tell a
> difference with the damping gone, I don't think it was
> accomplishing much because it was damping very quickly to the
> target values.
More information about the SLDev
mailing list