[IMPORTANT] DSP / signal calculation optimization

[psykoruptor]

[ THIS IS REALLY IMPORTANT TO ME AND IF THERE IS A WAY TO GENERALLY FIX PLUGDATA ON THIS POINT I CAN PUT MONEY ON IT TO HELP !!! ]
100$ if this it "patched" (lol, and "lets be honest i'm not rich at all" for scale)

---

Hi! I use plugdata a lot (2k+ patches) since ~ 1.5 year and I just found something concerning.

I made all my patches in a certain philosophy but I just found out that it doesn't work. I thought that when you [switch~] off an abstraction or a sub patch, every connected signal sleeps, because it sounded really obvious to me, and I didn't even try to mesure this, until now.

I found that :

• when you use [*~ 0], it still uses cpu to process
• when you try to avoid it using switches, it still uses (even more) cpu to apply the sum of signals
  (and dynamic patching with connect messages is pretty unstable / laggy)

My question is : is it a fundamental pure data thing that we can't change ?
If we can change it, lets change it, it would save so much cpu to make these signal connections sleep

thank you

[timothyschoen]

• when you use [*~ 0], it still uses cpu to process

Correct, that's expected. Though a single multiplication uses very little (basically negligable) CPU time. There's also no way around this: we can't say "stop processing when the input is zero" because then any number of zeros over a connection cord would cause it to skip samples, which would break the basic assumption signal objects are built on.

• when you try to avoid it using switches, it still uses (even more) cpu to apply the sum of signals
  (and dynamic patching with connect messages is pretty unstable / laggy)

I'm not sure what your test setup is, but from what I can tell, that's not accurate. If I put an [osc~] object inside a subpatch and I use switch~ to turn the subpatch off, I can confirm that the objects inside the subpatch don't do any processing. Let me know if I'm misunderstanding what you mean.

is it a fundamental pure data thing that we can't change

I'm afraid the answer is yes.

[psykoruptor]

Hi, thank you for your answer!

I think I need to clarify two things

1. I didn't mean to to detect when cords are 0, I meant to turn off cords from A to B if A or B is off, I thought it would be possible to basically "switch on / off the cords" according to the patches they link, like each edges of a cord has little bool saying "this side it's on" and the cord skips calculation if there is one off

2. And for the second thing, it wasn't the situation I talked about ; I know that everything inside a subpatch/abstraction that is off is sleeping. But If you have boxes A and B, both in a parent patch P, and lets say both linked to a same outlet, if you turn off A, it still needs to calculate the sum, that's these kind of situations.

And a last thing I'll add : for my use of plugdata, multiplications aren't negligable at all since I have an enormous patch of 8k lines, with a lot of processing including a lot of multiplications and switches
I really care about cpu when I'm programming but I still have cpu issues... (40-50% (plugdata display)) but I also need to run visuals and that's pretty harsh

If we can't handle it I don't know what I can do, that seems to be a big limit to my patching
Anyway, thank you for your answer and your amazing work that lets people express a new kind of creativity