Matt Romein

I'm working a bpatcher system using waveforms to automate the changing of values for parameters in a large scale video processing system. I have it working but since I am more of a jitter programmer than a MSP programmer I know my waveform patch is highly inefficient. I would love to visualize the shape of the waveforms and my current setup has me doubling up the waveforms and sending one set really high frequency values so I can visualize it in scope. Is there a better way to do this? Also wondering if I can turn off the waveforms I don't have selected rather than using selector~ to funnel a specific one and leaving them all on. Attached is my current build. A master qmetro in the larger patch usually drives the snapshot~ object but I automated it for forum purposes. Appreciate any thoughts.

help-optimizing-effeciency-in-msp-patch

there are a few things (4 different oscillators and their modifiers) running in parallel, which you choose from using a gate (or selector~).

maybe you want to put these 4 things, which are used only alternatively, into 4 poly patchers os that you can turn of the off when the 4th is running.

the code layout itself can hardly be optimized, except that you could use wavetables instead of modifiers in realtime. i.e. you could read from a wavetable which already holds a folded version of a cosine instead of using cycle~ and ten apply abs~ to it, and so on.

maybe you want to put these 4 things, which are used only alternatively, into 4 poly patchers  os that you can turn of the off when the 4th is running.

the code layout itself can hardly be optimized, except that you could use wavetables instead of modifiers in realtime. i.e. you could read from a wavetable which already holds a folded version of a cosine instead of using cycle~ and ten apply abs~ to it, and so on.


you could build al of these waveform generators with floating point numbers as well, and maybe even design it in a way that you dont require metro and setting a custom grain size, but just request the next parameter value from the "LFO" when the next video frame is happing.

or downsample the [poly~]s to what you need. Will make it more efficient but still easy to maintain and change.

Inspired by Roman's suggestion, I tried to follow this kind of approach:

Just a sine wave for now. I'm encountering difficulties in implementing other types of functions...I think I'm missing some maths elements...

As roman said using wavetables is wise instead of computing things. You moved to the max domain but also you use a computed sine instead of a wavetable before, [cycle~], so I'm not sure if this is going in the right direction. again, I'd opt for downsampled polys.

Help - Optimizing Effeciency in MSP patch