[PD] IEMlib and compilation optimised for P4s [denormals once again!]

[derek holzer]

aha! revenge of the denormals! i don't know if it has been dealt with in 
PD proper entirely. i just threw out a bunch of hip~ objects [which is 
"proper pd", and not an external, afaik] tonight because they were 
denormalizing like crazy every time they got zero volume.

as hans recommends, check through the list, there are some posts on 
denormals. one suggestion was to find an Intel C compiler for windows 
and compile the objects yourself. a quick fix is to run a small amount 
of noise through all the problem objects.

i think delays and filters in combination, and in that exact order 
[always a favorite for my live sets, and always bugging my machine 
out!], are about the highest risk of denormals ever.

d.

-=lowercase is beautiful -- e.e. cummings=-

Hans-Christoph Steiner wrote:

>Sounds like the denormalize bug in P4s that causes the CPU to spike when a
>calculation is approaching zero.  AFAIK, this problem has been addressed
>in Pd proper and some externals, so there should be example code out
>there.  Check the archives of this list and pd-dev for more info.
>
>.hc
>
>On Mon, 3 Nov 2003, David NG McCallum wrote:
>
>  
>
>>Dear All,
>>
>>I had one of those CPU pig problems that I think I tracked down to the 
>>IEM filters not being optimised for P4s (just a guess).
>>
>>Does anyone have IEMlib modified and compiled for P4 optimisation for 
>>Windows or...
>>
>>Can anyone tell me that I'm wrong and that the IEM filters are an 
>>unlikely source of massive CPU consumption?
>>
>>Cheers,
>>David
>>-- 
>>.
>>. David McCallum
>>.  Music wants to be free
>>. http://mentalfloss.ca/sintheta
>>.
>>
>>
>>
>>_______________________________________________
>>PD-list mailing list
>>PD-list at iem.at
>>http://iem.at/cgi-bin/mailman/listinfo/pd-list
>>
>>    
>>
>
>	zen
>	   \
>	    \
>	     \
>
>
>_______________________________________________
>PD-list mailing list
>PD-list at iem.at
>http://iem.at/cgi-bin/mailman/listinfo/pd-list
>
>  
>