CPU issues, can not get my patch to run well

yannseznec

Hi!

I'm trying to run some granular Pure Data patches on my Bela, and I simply can not get it to work without crashing due to CPU issues.

I've made a super-stripped down version of the patches I'm trying to run here: https://github.com/yannseznec/granular_test

That test patch uses a single sound file in a granular patch with 4 grain voices, triggered very slowly. I've tried on two Bela boards (one mini, one big), and both times they basically don't run. They sometimes start and run for a few seconds but then crash.

The IDE gives this error:
error: array: no method for 'edit'
Underrun detected: 1 blocks dropped
Xenomai/cobalt: watchdog triggered
CPU time limit exceeded
make: *** [runide] Error 152
Makefile:621: recipe for target 'runide' failed
Bela stopped

I've tried changing the block size in the IDE and it makes no difference.

The patch doesn't seem to me to be super CPU intensive, but maybe I'm naive! Is it just too much for the Bela to handle? I certainly feel like I've done more intensive things on a Bela and a Raspberry Pi before.

(one caveat: I'm not sure what version of the Bela software these are on, but I think it's relatively recent. They are running Pd 0.51-4.)

So I'm not sure if this is a Bela problem, or a problem with how I've made my Pd patches, or something else entirely. Any tips would be great!

giuliomoro

yannseznec error: array: no method for 'edit'

This I don't know about. Probably you are sending an "edit' message to an array object?

yannseznec Xenomai/cobalt: watchdog triggered
CPU time limit exceeded

That's a way to say that you are trying yo use way too much CPU

yannseznec

Im confused by that “edit” error as well, I don’t think I’m sending an edit message. But yeah I understand that the cpu usage is super high, I just don’t understand why! The patch does not seem very intensive to me.

giuliomoro

yannseznec The patch does not seem very intensive to me.

Care to share it ?

yannseznec

giuliomoro this is the patch https://github.com/yannseznec/granular_test

yannseznec

Yes the GitHub link to the full patch is up on the original post!

giuliomoro

@Donnerbono thanks for the investigation. Yes, [bob~] tends to be a CPU hog. yannseznec I have a slightly inaccurate version of [bob~] that is much lighter on the CPU . See details here.

Donnerbono

Hi, I had a quick play with this. It's a nice patch!

The edit array is a message on loadbang, just search for "edit" and delete.

I'd remove all the VU / display objects unless you plan on using them as they're continually updating things you aren't using.

I got it to play by removing the bob~ objects in the sampling stereo part as must be a processor hog and maybe use a lop~ instead or something? Down around 30% now for me unless I deleted more things unknowingly! But sounds like it does running on the PC.

Cheers.

yannseznec

Whoa amazing thank you both! I will give that a try.

yannseznec

yes indeed, removing the bob~ object seems to do the trick. I wasn't even using that subpatch, it was a sampler system I stuck in there some time ago! I took the whole section out, along with some of the VU/GUI stuff and it worked much better. I've now made a new test patch that has three instances of the granular patch running at once, without issues. The GitHub is updated: https://github.com/yannseznec/granular_test

@Donnerbono if you liked the granular patch I have a more fully featured version here (which I will now be updating with these changes!) https://github.com/yannseznec/ys.granular

Thanks so much everyone.