Parallel gradients with Fmincon

One simple question: why does fmincon offer parallel gradient computation but does not offer the option to vectorize the evaluation of objectives and constraint functions?

In my project I've to spawn jobs through more than one external simulator to get my objectives and constraints evaluated. This is done outside Matlab since there are checks on the machines and licences available to run these simulations.

In such case it seems much easier to me (I've already done it for the genetic algorithm and pattern search) to ask the optimization to vectorize the call and just pass the X vectors to my external application managing the parallelization. Rather than having to manage parallelism both in Matlab and in my external code.

Can someone give me an explanation?

Best regards, Felipe

[Steve Grikschat]

Not an answer but, to my knowledge, this is the first time that a feature like this has been requested.

I just wanted to let you know that your suggestion has been noticed.

Regards,

+Steve