[PD] Reasonable rfft computation time

[Matt Barber]

Hi list,

As part of my array-abs library – https://github.com/brbrofsvl/array-abs –
I'm including fft abstractions. I've just finished the rfft one (needs some
polish, so it's not on github yet); my question is, I'm not sure what a
reasonable computation time for rfft would be using vanilla Pd only (and
not relying on the tilde objects, which requires dsp to be running).

On my macbook pro it does very roughly (1/2500)*n*log_2(n)ms. This is ca.
1ms for a 256-pt rfft, 400ms for a 2^16-pt rfft, and a full three minutes
for 2^24 points. I have a couple of tweaks that will make it moderately
faster. I could incorporate a ton of [expr] now that it's lgpl, which could
probably make it faster yet, but it would sacrifice some patch readability
(which is important to me).

Any thoughts?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20151106/24d551a6/attachment.html>