Generic vector valued function problem

[Gideon Simpson]

I asked a question elsewhere, but it's part of a more general problem I'm trying to work on, so I thought I'd start a new thread to highlight what I'm after.  I'm interested in accelerating the computation of vector valued functions.  For example, maybe I have a parametric curve like t - > (cos(t), sin(t)), or I have a vector field (x,y) - > (x*y, x-y).  So, in general, I have an input argument thats in R^n (or C^n), and an output variable thats in R^m (or C^m).  n and m are known beforehand, but they may not be the same.  The goal is to accelerate this, ideally with multithreading as an option.  What would people recommend?