Parallel Computing with surrogateopt

R2026b

Enabling Parallel Computing

surrogateopt can evaluate objective function values on multiple workers simultaneously. Unlike other solvers that use parfor, surrogateopt uses parfeval (Parallel Computing Toolbox) to send evaluations to workers asynchronously.

To enable parallel computing:

options = optimoptions("surrogateopt",UseParallel="auto");

When UseParallel is "auto" or "on", surrogateopt distributes search point evaluations to available workers. The solver continues selecting new search points while waiting for evaluations to complete on workers.

You cannot set UseParallel to "auto" or "on" and also set UseVectorized to true. If you enable both, the solver ignores UseVectorized and attempts to compute in parallel using a parallel pool, if possible.

For algorithmic details, see Parallel surrogateopt Algorithm.

Limitations

• No nested parallel loops. parfeval (Parallel Computing Toolbox) does not work in parallel when called from within another parfeval loop. If your objective or constraint functions use parfeval, the solver's parallel evaluation takes precedence and your inner parfeval runs serially.

For information on factors that affect the speed and results of parallel computations, see Improving Performance with Parallel Computing.