Add modulation matrix

[jpcima]

Modulation matrix update

This model is operational and it can handle CC.
Support

• Amplitude
• Pan
• Width
• Position
• Pitch
• Volume

Not much tested yet, also I will clean old source which is made obsolete by this.
This will also make obsolete #326.
The CC list will be obtainable by inspecting matrix edges of CC source. Also it permits to have access to the target, and determine a useful name for the midnam item.

[jpcima]

There is some EventEnvelopesT code #if 0'd in the removal commit above. I'm not sure yet what to replace it with.
I removed all old things regarding modifiers.

[jpcima]

The commit 51b7322 ensures to always run the modulation source, even if no voice uses it.
This ensures the CC will continue to smooth in the back, even with no voice playing.

[jpcima]

@paulfd this faces a problem with the smooth filter shortcut..

For instance I'm smoothing pitch CC at center 600, value 1200.
The filter basically stops rising at around 599,121033, and it's too far off of the virtuallyZero` difference threshold such that shortcutting would be possible.

What to do in this case? consider relative difference?

[paulfd]

There is some EventEnvelopesT code #if 0'd in the removal commit above. I'm not sure yet what to replace it with.
I removed all old things regarding modifiers.

You can remove it I think, these tests were not great and we can make better ones for the FlexEGs

[paulfd]

@paulfd this faces a problem with the smooth filter shortcut..

For instance I'm smoothing pitch CC at center 600, value 1200.
The filter basically stops rising at around 599,121033, and it's too far off of the virtuallyZero` difference threshold such that shortcutting would be possible.

What to do in this case? consider relative difference?

Hmm I'd say yes, but is it OK on both ends ? (e.g. for small values too)

[jpcima]

but is it OK on both ends ? (e.g. for small values too)

hmm, as a relative distance, how about:
(abs(x-y)) / (abs(y)+virtuallyZero)

[paulfd]

Yes that seems reasonable.

[jpcima]

The only remaining feature that I can see missing, it's a modulation affecting the depth of another modulation.
eg. lfoN_amplitude_onccX
This will be a feature for another PR.

[paulfd]

Could shortcut here if depth == 0.0f

[jpcima]

Yes not only you're correct, but there is a greater problem.
The source is supposed to be buffered per region or cycle, not have its generation output used directly.
This makes the source generate more than once if in use by multiple targets.

ie. make use of Source buffer/bufferReady

[jpcima]

See a97f506.
Regarding the shortcut: a future consideration is that depth can be modulated.
When this is implemented, sourceDepth=0 is an insufficient condition, one has also to check that depth is free of any modulations.

I'm sure I to have this idea in mind previously, but postponed it until making the depth modulation.
There is not an immediate benefit anyway, the source will have to implement a fast path in generateDiscarded.

As CC as concerned, I believe the fast path is implementable in case of not smoothing, or smoothing near completed.