I'm experimenting with some basic ice stuff, using point position on the x axis to create a spread of values instead of painting a weightmap (so that I can use that attribute to drive a shape instead of modifying shape by weightmap).
I was wondering how expensive it is comparatively. Get point pos, vector to scalar, rescale, fcurve, set attribute, execute, then another branch to use that attribute to modify a shape.
Would it really slow down playback speeds of a rig? Would it get evaluated per frame, or just once? (assuming envelope comes after in the stack)
The benefits I guess are less cluster shapes stored in a model overall, ability to change the spread interactively, and reuse on any model without any laborious painting. However, speed of playback is still king in animation terms and if setting up too may attributes will slow a rig down to a crawl then its not worth setting it all up...
cheers