Repelling Circles performance

as already mentioned in riot, the repelling circles demo is awesome but has a hefty performance hog, which i’m trying to understand.

when i add 300 circles, the renderer window can’t be moved anymore or becomes very unresponsive. The cpu and gpu looks fine, there is still performance left. how would i track down what’s killing the thing. in comparison, Natans vvvv beta example https://vvvv.org/contribution/boubles-proximity-dynamic does 512 circles with ease.

there seems a similar issue with the PackingCharacters demo, when you run it a few seconds, the renderer window freezes and vvvv gamma is on the edge of crashing.

@u7angel I haven’t looked at it yet but I guess it’s not doing anything clever to optimise spatially. It’s a n*n problem. I’m pretty sure Natan’s boubles does spatial binning of some kind to circumvent this

@mrboni, sure. it is not the best idea to have badly optimized demos in the list. especially the packing characters patch freezes the window without user interaction. the demos are the first contact for fresh users…

Re: circles - the problem is things get much more complicated when doing spatial binning, which could put people off from exploring these kind of techniques.

I’d suggest (to the devvvvs) the demo include a mention of safe count limits, a quick explanation why these limits exist, and a link to another patch with a super sexy spatially binned interaction example ;)

caveat - I still haven’t looked at the patch in question!

quick note, this is a current bug in the mainloop, where the mainloop has more priority than the user interaction. so it gets unresponsive for expensive patches. will be fixed soon.

1 Like

@mrboni I don’t remember boubles having binning, not sure, not in front of a pc, but I recall it being brute force…

From the description:

This module is usefull when you have a large number of interacting agents: it’s a grid based proximity detection that gives, for each agent-particle, only the indices of neighbour agents (this means optimization in interaction evaluation).

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.