Hello!

I'm not really new to pd since I use it now and then since a year now, but I'm still on the learning curve (is there an end to what you can learn with pd?!). As my patches get bigger and more complex, I would like to optimize ressource usage.
I hear that a lot of you experienced patchers use pd-vanilla instead of extended, and I was wondering why. Is it more ressource economic? I understand the more libraries and externals you load, the longer it takes to start pd, and the more memory it will use, but is there any difference between vanilla and extended if the same libraries are loaded? I mean besides gui stuff...
Apart from that, is there an advantage to compile the externals by yourself? Will they be more CPU/memory efficient? Or is it to get the latest version? I have the same question about Pd itself, is there any reason to compile it yourself?

the main thing to optimize resource usage is just to cut down on unnecessary cpu use in your patches. it doesn't matter too much which version of pd you run. you just have to concentrate on making your patches more efficient.

for example: if you have several synths that use [noise~] , then just make one [noise]~ object, and [s~ noise] / [r~ noise] to distribute the signal.

also, of course, use [switch~] to turn off any subpatches or abstractions when they are not generating sound.

learn the difference between [line~] and [vline~] and use them appropriately. if you really need sample accurate timing, then use [vline~], but otherwise use [line~] as it uses much less cpu.

there are lots of things you can do to greatly reduce your cpu footprint. as long as you always keep it in mind while patching, you will naturally become more efficient.

Thank you for your advices! These are some good habits I will try to make.

So the choice between vanilla and extended is only a question of personal preference? I know that pd runs two processes, one for the pd engine and another for the gui. Is the engine process the same in vanilla and in extended?

why is noise so cpu intensive?
a normal rng (don't know about noise~ generation) consists of 1 * % and +
where even a cos (osc~) should be more complex if it doesn't use a table lookup __w.o. interpolation__

one noise~ object is not particularly cpu intensive, but if you have 20 or 30 of them in a patch (quite easy to do if you make an analogue style drum kit, for example), then of course a whole bunch of [r~] objects are going to be more efficient.

as far as i know, the pd engine and gui are pretty much the same in both vanilla and extended. as for the difference between the two, i can only speak from personal experience, but here are the reasons i stopped using extended and went back to vanilla: (1) there are annoying objects in extended like [spigot~] which will not work on certain operating systems. (2) some objects in extended have weird behaviour (bugs). for example, the matrix objects looked very useful, but for some reason, they wouldn't load for me unless i opened the help patch first. (3) the most important reason i stopped using extended was because i did a commercial project using pd as the sound engine. the licence for pd vanilla allows for this, but the externals in pd extended are all covered by different licences.

also, i don't think there is much runtime difference between vanilla and extended, but as extended has a larger default search path, it does take a fraction longer to create objects. nothing worth worrying about usually.

pd-vanilla is usually a later, newer version, than the pd core pd-extended uses, which is obvious, since pd-vanilla as an app needs to be incorporated into the pd-extended releases, once it has been finished.

yet, cpu wise, pd itself has always been very gentle and all bother over the years with footprint came purely from dirty patching.

cpu wise the worst is drawing onto the gui like moving objects or redrawing graphs

totally.