Hey,
What are the best ways to optimize a rig ? (classic stuff: skinning, blendshapes, FK/IK, constraint expressions, deformers, constraints nodes, driven keys, ...)
In particular, I'm trying to find out if it's relevant to recode some chunks of the rig using GPU parallelism (cuda, opencl): for instance some maya nodes, or isolated workflows.
In a more generic point of view:
What are the bottlenecks during the evaluation of a scene frame? Indeed, if maya consumes all the CPU&GPU only for displaying meshes, there is no interest in optimizing the rest of the evaluation tree.
Are there any tools, like code profiler, to analyze the execution process of maya? Without reverse engineering it of course.
One example: here is a link about recoding the maya blendshape node, to be executed on the gpu:
http://www.chadvernon.com/blog/resources/cvcudablendshape/
Thanks!
