[Clam-devel] Pitch-Shifting steps (harmonizer related)

[Xavier Amatriain]

Pau Arumi wrote:
>>
>> Question 1: I'm using 'pow' function... do you think is worth to
>> replace it with a fixed look-up table with 12 values? (but in this
>> case user would only have 12 possible values, for example: 1.00, 2.00,
>> ..., 11.00 but nothing else, now, with 'pow' he has the option of for
>> example '3.15' if he needs it). I'm talking/asking about real-time
>> costs of that operation (i'm think not much but i'm asking anyway)
>
> I'm not sure about that. But I'm more inclined to do optimizations
> only when a profiling tool shows what the bottleneck is (as someone
> said: early optimizations are evil)
I agree with Pau. Also don't forget that we have very optimized versions 
of these functions [1]. I already came accross
these issues in related places and yes pow and sqrt are a real pain 
performance-wise.

[1] 
http://clam.iua.upf.edu/doc/CLAM-devel-doxygen/CLAM__Math_8hxx-source.html#l00266