Execute run phase of Simulink model in parallel

I want to get the output using the function model(t,x,u,'outputs'). I have a large set of inputs, which I normally pass to the model in a for loop. I have already compiled the model before the loop using model([],[],[],'compile'). When I replaced the for with a parfor loop, it does not see the model as 'compiled'. I get the error: Error, 'model' must be compiled before it can be accessed programmatically.

I don't want to compile the model inside the loop as it would make it very slow and destroy the whole purpose of using parallelization. How can I solve this?

Answer 1

[Edric Ellis]

You can use parfevalOnAll to run the 'compile' and 'term' phases, like so:

parfevalOnAll(@vdp, 0, [], [], [], 'compile');
parfor t = 1:10
    out{t} = vdp(t, [0; 0], [], 'outputs');
end
parfevalOnAll(@vdp, 0, [], [], [], 'term');

[Osama Maqbool]

This worked the first time but did not when i ran it again. I am using this code to generate the parpool:

numCores = feature('numcores');
p = parpool(numCores);

Could it have something to do with this? The error is the same as before.

[Edric Ellis]

Strange. Did it fail the second time in a MATLAB session, or when you restarted MATLAB?

[Osama Maqbool]

In the same session. I can't seem to find a pattern on what gives the error. Many a times, I have made it work by deleting the contents of a folder generated automatically, called 'slprj', specifically the subfolder '_jitprj'.