It appears that you have two questions.
Q1. How many cores/processors can fmincon use at a time?
If you set the UseParallel option to always then fmincon will use up to 8 local workers (MATLAB labs) to perform the optimization. The number of local workers that fmincon uses is determined by the number of cores/processors available on your system. With Parallel Computing Toolbox the maximum number of workers that you can use is 8. However, if you have MDCS, the number of workers used by fmincon is limited by the number of MDCS licenses.
Q2. How can I execute two fmincon calls at the same time in parallel, with each fmincon call using 4 workers?
On a local machine with 8 cores it is currently not possible to execute two calls to fmincon with each call using 4 workers at the same time (in parallel). However, it is possible to execute two calls to fmincon, at the same time (in parallel) with each fmincon using 3 workers. To execute these two separate evaluations of fmincon in parallel you will need to use the batch job API. For example, let's say that you have the following function:
function xx = myfmincon(x,c)
% UseParallel option is set to 'always'
xx = fmincon(@(x) myfun(x,c));
end
Then to evaluate two instances of this function in parallel you will need to use the following series of commands:
sched = findResource('scheduler','type','local');
job1 = batch(sched, @myfmincon, 1, {x1,c1}, 'Matlabpool',3);
job2 = batch(sched, @myfmincon, 1, {x2,c2}, 'Matlabpool',3);
wait(job1);
wait(job2);
xx1 = getAllOutputArguments(job1);
xx2 = getAllOutputArguments(job2);
For more information of the batch command please refer to the following documentation page:
