big FFT~, multicore processing?

Will there be any use-case examples of FFT processing in RNBO?

i read in the announcement thread, i think it was Josh Clayton, mentioned to someone asking about fft, that there will be some examps eventually(just need to be patient, i bet the rollout requires some care in working out immediate kinks here and there, first).

it’s difficult to give FFT advice without seeing the patch, and i don’t recommend sharing your RNBO commercial product here, either, but going by what i’m reading, it seems you might be reaching too hard:
-there’s rarely any need to process each bin discretely at all(if you have 2048 bins, that would be an intense amount of detail you’re trying to work over the spectrum going bin-by-bin(maybe you get an AI to do this instead, haha), plus with an FFT size of 2048, there’s 1024 bins(upper-half) that might be redundant for you depending on what you‘re doing)
-instead of going through this and running the risk of confusing you with the way i barely understand the semantics, i’d recommend you go through these tutorials if you haven’t yet:
PhaseVocoder-Part1
PhaseVocoder-Part2

-you’re not off base with that one part you mentioned(for an FFT-size of 2048, and a desired ‘overlap’ of 4, you need a ‘hop-size’ of 512), i’m just not sure how you’re doing it, there’s many ways(see the phasevocoder-part1 article i linked above), you don’t necessarily need multiple ‘ifft~’/‘ifftstream~’ objects, depending on how you’re processing the data you get from the input/analysis stage(again, see the part1 article linked above, notice they have multiple fft~, but only one fftout~… but also, for that article, you’ll need to remember it’s within a pfft~… there will be things that you’ll need to figure out on your own from there specific to fftstream~ in rnbo~, hopefully with help of incoming examps c74 might give us(?))

-for your codebox question, first off, codebox is not necessarily easier nor more-efficient(you don’t necessarily need it at all, but it’s good to learn to have more options/techniques)
with that in mind, your specific answer is one i’m wondering about myself: we don’t have a way to create visual objects in rnbo~ and then click to see the codebox equivalent(and that probably means they need to create better docs for rnbo~ codebox, otherwise, most of us will be flying in the dark as to what we can or can’t do - i don’t see how it registers ‘fftstream’ as an op at all, nor even just ‘fft’?)

but if you want to study codebox in general, try within gen~ first, try this: create a gen~ object(maybe outside of rnbo~ to start), then go into the gen~ object, create a [buffer wut what] op, and a [sample wut @interp cubic] op, then connect the [in 1] op to the input of the [sample] op and [out 1] to the output of the [sample] op, now with everything connected, click the ‘C’ icon in the right-hand toolbar, and you’ll see an equivalent code for codebox(this is how you can learn the more advanced/detailed parts of how to codebox, for example, how to codebox the “@interp cubic” part of the sample op and what order all the arguments go in, etc.).

that said, i don’t see how to get help for using ‘fftstream’ in a codebox within rnbo~-naked(my word for it), perhaps it just means codebox in rnbo~-naked will be different than in gen~(you’ll just need to use it for the specific uses documented only?)

anyways, that’s my 2cents, it’s Thanksgiving weekend, too, so if you’re still lost on your codebox question about fft in rnbo after you give some time for folks to come back from the holiday here, maybe bump this thread again(it‘s definitely an interesting/worthwhile question about using fftstream in codebox, i‘m curious too :) hope this helps.