Efficient real time spectrum graphing techniques?

[Fender19]

4-6% CPU usage that some are stating here is not “light” IMO. Try to use a dozen plugins like that in a typical mixing scenario and see what happens!

Perhaps, but I wouldn't keep a dozen plugins open all displaying spectral graphs at the same time and, presumably, your plugin will only perform these calculations while the interface is open.

Excellent point - also good info for plugin USERs who like to keep lots of windows open! Thank you.

[quikquak]

But you're saying "FFTs are NOT light on resources" - have you tried Intel's IPP ?

[Fender19]

Something that should probably be mentioned is that a typical text-book radix-2 can easily be a factor of 100 (or more) slower than a properly optimised FFT library.

Yes - this would explain why my observation that FFT was slow.

Do you have a source for good, efficient audio FFTs you could point me to?

[Fender19]

But you're saying "FFTs are NOT light on resources" - have you tried Intel's IPP ?

Not yet but since you mentioned it I will!

[quikquak]

What I've done in the past is to call the FFT in the render thread using with whatever data is available at the time of rendering. Meanwhile the audio thread is filling up a short loop of data that I can read from in a write position going backwards fashion, guarranteeing there is no read/write clashes.

[vortico]

Yes, don't write your own FFT implementation for production, unless of course you want to learn how it works, or if you have a very specific use case and know what you're doing.
In very rough order of performance, some of the best FFT libraries for CPUs are: Intel IPP (proprietary freeware binary-only), Apple vDSP (Mac only), FFTW (GPLv2+), Julien Pommier's pffft (MIT-ish license). With that said, all perform within a factor of 2 compared to each other.