AudioProcessor float precision woes (not just in processBlock)


#1

Some heated discussion about floating point precision in previous posts. Personally I think JUCE should not impose 32 bit precision on developers for things such as audio samples and parameter values. Surely it should be the choice of the developer, or at least choose the higher precision at the framework level and let the developer do internal processing at lower precision if so desired.

Please can we have a typdeffed float for audio processor parameter values or a templated AudioProcessorParameter? I would like my parameters to be double precision, because single precision is causing some annoying rounding errors with a DSP framework I want to combine with JUCE that prefers doubles.

thanks!

 

 


#2

+1

 


#4

do you mean 64bit where you wrote 32bit? 

 


#6

thanks for your reply, but I am really keen to use the parameter and syncronization stuff that Jules is adding to JUCE rather than having another system. In fact i already ported IPlug's rather nice IParam class to JUCE which I can use instead of AudioProcessorParameter, but want to do it the JUCE way, with as few dependencies as possible.

VST3 parameters "ParamValue" are double precision, so it would be nice to be able to write JUCE plugins that support that resolution.