[sldev] Wind Sound CPU usage.

[Aimee Trescothick]

I still hear some clicks, sorry  :S :)

I'm on a Mac, so maybe that's the difference. I also have some  
background in sound engineering, so I'm probably more critical than  
most. I'll record a couple of samples in a minute so we can see if  
we're comparing like with like.

The pinking filter as it's implemented is overkill, I've cut it down  
some with no noticeable difference using the "economy" method from  
http://www.firstpr.com.au/dsp/pink-noise/

My sub-sampling seems to be fine at 22K, it's only when I go down to  
14.7K it becomes noisy, I'm probably doing something stupid  
somewhere. The filter is cutting over 60dB at 7Khz (audio), so a 14K  
sample rate should be sufficient according to Mr. Nyquist and co.

I've incorporated some of the clean up from yours and tidied things  
up a bit, would you mind profiling the attached patch and seeing if  
I'm having any impact?

Aimee.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: LightWind2.patch
Type: application/octet-stream
Size: 5314 bytes
Desc: not available
Url : http://lists.secondlife.com/pipermail/sldev/attachments/20080305/ac31d868/LightWind2-0001.obj
-------------- next part --------------


On Mar 5, 2008, at 22:02, Jason Giglio wrote:

> Jason Giglio wrote:
>
> Sorry Aimee,
>
> You were right.
>
> Upon listening again to mine at full volume I did detect some  
> subtle clicking on rapid velocity change.  I added interpolation of  
> the gain only back in.
>
> Interpolation of the frequency and pan-gain does not seem to be  
> necessary (I sat on a fast spinning physical cube to make sure), so  
> my optimization still works.
>
> A new patch is attached to this message that interpolates gain only  
> to fix this subtle clicking and still only uses about 2% CPU.