LoneMonad aka don malone

anyone have a good cpu efficient reverb for ub [intel dual core}

4-6-reverb

On Aug 14, 2006, at 8:32 PM, Ali Momeni wrote:

> olaf matthes' freeverb~ and gigaverb~ are compiled for mac-intel 

> but i experienced strange CPU climbs with them; they start out 

> around 3-4% of my macbook pro and after a while they go up up up to 

Almost assuredly this is due to denormalized floating point numbers. 

If he's using the z_dsp.h macros, the version in the preliminary UB 

SDK is incorrect. Below is some correspondence on the subject from 

> I noticed that DENORM_WANT_FIX returns 0 under OSX on intel. From 

> what I've read I understand that on intel macs in will be necessary 

> to flush denormals to zero manually in code (as for windows) where 

> they might cause problems, as it cannot be done automatically (as 

> If this is the case, is it possible to have the SDK changed so that 

> it will test for an intel/ppc build and define this accordingly, so 

> that universal binaries can be easily maintained properly from one 

Thanks. Yes this was already done before the MaxMSP 4.6 public beta 

release, but has not yet made it into the preview SDK (which has an 

earlier version of z_dsp.h). It will be in a forthcoming version of 

the SDK, but for now, you can simply change z_dsp.h to the following:

#if defined(WIN_VERSION) || (defined(MAC_VERSION) 

> olaf matthes' freeverb~ and gigaverb~ are compiled for mac-intel  
> and quite efficient
> but i experienced strange CPU climbs with them; they start out  
> around 3-4% of my macbook pro and after a while they go up up up to  
> 50-60-70 and then things stop.

Almost assuredly this is due to denormalized floating point numbers.  
If he's using the z_dsp.h macros, the version in the preliminary UB  
SDK is incorrect. Below is some correspondence on the subject from  
the beta list.

> I noticed that DENORM_WANT_FIX returns 0 under OSX on intel. From  
> what I've read I understand that on intel macs in will be necessary  
> to flush denormals to zero manually in code (as for windows) where  
> they might cause problems, as it cannot be done automatically (as  
> on PPC).
>
> If this is the case, is it possible to have the SDK changed so that  
> it will test for an intel/ppc build and define this accordingly, so  
> that universal binaries can be easily maintained properly from one  
> source file?
>

Thanks. Yes this was already done before the MaxMSP 4.6 public beta  
release, but has not yet made it into the preview SDK (which has an  
earlier version of z_dsp.h). It will be in a forthcoming  version of  
the SDK, but for now, you can simply change z_dsp.h to the following:

#if defined(WIN_VERSION) || (defined(MAC_VERSION) 
&&TARGET_RT_LITTLE_ENDIAN)
#define DENORM_WANT_FIX		1
#endif

olaf matthes' freeverb~ and gigaverb~ are compiled for mac-intel and 

but i experienced strange CPU climbs with them; they start out around 

3-4% of my macbook pro and after a while they go up up up to 50-60-70 

i'd seen this in another of olaf's external's when running on a 

pentium iv and he ascribed it to floating points very close to zero 

but he fixed that for the xp external....

On Aug 14, 2006, at 3:56 PM, don malone wrote:

> anyone have a good cpu efficient reverb for ub [intel dual core}

http://faculty.roosevelt.edu/malone/default.htm

olaf matthes' freeverb~ and gigaverb~ are compiled for mac-intel and  
quite efficient
but i experienced strange CPU climbs with them; they start out around  
3-4% of my macbook pro and after a while they go up up up to 50-60-70  
and then things stop.
i'd seen this in another of olaf's external's when running on a  
pentium iv and he ascribed it to floating points very close to zero  
and what they can do to pentium's.
but he fixed that for the xp external....
perhaps it's the same story here.

i.e. here:
http://www.akustische-kunst.org/maxmsp/

>
> anyone have a good cpu efficient reverb for ub [intel dual core}
> --
> it takes all of us
> LoneMonad
> http://faculty.roosevelt.edu/malone/default.htm


many thanks for fixing this. I have played around with freeverb~ and 

gigaverb~ for a while last night and it seemd ok now.

On Aug/19/2006/, at 13/23, Olaf Matthes wrote:

> Okay, I put that in and uploaded the new versions to the usual 

> There's now also a UniversalBinary version of analogue~.

> However, I don't have an Intel Mac so it's NOT tested!

many thanks for fixing this. I have played around with freeverb~ and  
gigaverb~ for a while last night and it seemd ok now.

> Okay, I put that in and uploaded the new versions to the usual  
> http://www.akustische-kunst.org/maxmsp/.
>
> There's now also a UniversalBinary version of analogue~.
>
> However, I don't have an Intel Mac so it's NOT tested!
>
> Olaf

---
Bernhard Loibner
sound & media artist
http://loibner.cc


4.6 reverb