DSP Robotics Support

(héhé this post has been deleted but with all the spam that was surrounding it i understand why !)

Here's a fast blue array memin reader.
It take the memin and write it in a normal mem in a modified way.

Every SSE will be write like this : sample0,sample-1,sample-2,sample-3

So we could only read one time to get an interpolated result. (Here it's only a linear interpolation)
It's not bandlimited, but if you provide sin like array with frequency below nyquist the aliasing is low.
(I see it more like a table lookup.)

In white i provided an example but it's very bad and glitchy..
Fast attack will read an unloaded array..
Each time a new note is done, it might or not reload the array..
> That's where it seams impossible to manage even is we fix the fast attack...

For curiosity.. (because this might not be a really optimized way to do, and it do aliasing)

Here's a white fast sin array reader.
It only use an array of 32 memory, which make the aliasing.
The array is load in one cpu cycle each time a new note is done.
(for a 32 memory seams not so hard)

Maybe another interpolation could absorb some aliasing.
But it will take more cpu..