Nonlinear optimization using parallel input/output

I have a system that accepts a vector and returns a function value. The goal is to change the elements of the vector such that the function value is minimized using a derivative-free solver, eg. using NOMAD. The structure of the problem allows to pass n vectors to the function at once and receive n function values in return within about the same computation time.

The solvers, which I have found so far, will only pass 1 vector to the function at a time and wait for the function value to be returned.

Is there a solver, which allows "parallel processing" on a single-cpu? Any idea is appreciated.

• It sounds like you just want multiple instances of the solver. However, most solvers already take advantage of multiple CPU cores. Have you proven that you have unused CPU cores with your current solver? – Brannon Nov 26 '13 at 4:44