vilbjorg

excuse me if this has been a topic before, but already for a while i was searching information on how max6 is dealing with floats :: how many decimal-digits or digits can one work with.

hi
excuse me if this has been a topic before, but already for a while i was searching information on how max6 is dealing with floats :: how many decimal-digits or digits can one work with.
thank you
vilbjørg


floats-decimal-digits-max6

it's not really a matter of decimal digits, the problem is subtler. you should have a look here: 

http://en.wikipedia.org/wiki/IEEE_754-2008

hi.
it's not really a matter of decimal digits, the problem is subtler. you should have a look here: http://en.wikipedia.org/wiki/IEEE_754-2008
(Max's floats are 32-bit)

thank you, i remember reading about this some years back in PD documentation of Miller Puckkette, but was not sure where, and it could have been changed for Max6 , but looks like its not, Just in case anyone passes by here who knows if the constants in gen~ like TWOPI are of the same precision , then drop a note. -- Or actually i understand that max6 is doing some 64bit sound processing, then perhaps the 'float-values' that one can circle around (iterate on) within gen~ are 64bit??

hi
thank you, i remember reading about this some years back in PD documentation of Miller Puckkette, but was not sure where, and it could have been changed for Max6 , but looks like its not, Just in case anyone passes by here who knows if the constants in gen~ like TWOPI are of the same precision , then drop a note. -- Or actually i understand that max6 is doing some 64bit sound processing, then perhaps the 'float-values' that one can circle around (iterate on)  within gen~ are 64bit??
thank you so much
v


that will work, using signals will give you a higher precision. the same is true for expression objects (which fo course wont let you implement loops)

Gen uses 64-bit doubles throughout. as the documentation says. although you can resample these for various tasks easily enough (especially in jitter gen).

the signal connections between msp objects are now 64-bit in max6 - this only has (excellent) implications for accuracy and timing, as the application as a whole is still 32-bit. however all max/msp/jitter objects have always used 64-bit doubles internally inside the objects for calculations.

andrea was rightly talking about integers and floats as they are in max-land, and as they are in terms of display. that is also what i took your initial question to be about.

signal and control rate are different beasts with regards precision.

edit - only just saw roman's post, sorry.

thank you for the detailed answer, thats excellent. It makes me think that perhaps for some algorithms , like various stuff from fractals and chaos, one could do a parallel research/check in something like matlab to see how much the data will deviate. Though running algorithms as signals I many times will have to feed them floats along the way... I am really greatly impressed with the sound of gen~ though, hoping for a soon working implementation of if-else, -to go all the way off the topic.

hi
thank you for the detailed answer, thats excellent. It makes me think that perhaps for some algorithms , like various stuff from fractals and chaos, one could do a parallel research/check in something like matlab to see how much the data will deviate. Though running algorithms as signals I many times will have to feed them floats along the way... I am really greatly impressed with the sound of gen~ though, hoping for a soon working implementation of if-else, -to go all the way off the topic.
v


floats & decimal digits max6