Is there a way to run many anonymous functions simultaneously?

Hi All,

I have a problem where I need to evaluate 18 anonymous functions before combining them into three variables. This is incredibly slow and painful at points so I was wondering whether there is a way to run these simultaneously (on an 8 core desktop and cluster)?

The reason we have so many functions is we are evaluating our test problem exactly through integration. We plan to create some numerical quadratures in the near(ish) future but need to be sure our method works well first for our test cases. An example of my code is below:

Here all the f_i's are integrals in space and time for an arbitary function.

if conditions == 1
    Tf1_1 = f1_1(arg1, arg2, arg3);
    Tf1_2 = f1_2(arg1, arg2, arg3);
    Tf1_3 = f1_3(arg1, arg2, arg3);
    Tf1_4 = f1_4(arg1, arg2, arg3);
    Tf1_5 = f1_5(arg1, arg2, arg3);
    Tf1_6 = f1_6(arg1, arg2, arg3);
    
    Tf2_1 = f2_1(arg1, arg2, arg3);
    Tf2_2 = f2_2(arg1, arg2, arg3);
    Tf2_3 = f2_3(arg1, arg2, arg3);
    Tf2_4 = f2_4(arg1, arg2, arg3);
    Tf2_5 = f2_5(arg1, arg2, arg3);
    Tf2_6 = f2_6(arg1, arg2, arg3);
    
    Tf3_1 = f3_1(arg1, arg2, arg3);
    Tf3_2 = f3_2(arg1, arg2, arg3);
    Tf3_3 = f3_3(arg1, arg2, arg3);
    Tf3_4 = f3_4(arg1, arg2, arg3);
    Tf3_5 = f3_5(arg1, arg2, arg3);
    Tf3_6 = f3_6(arg1, arg2, arg3);
end
%Combining the integrals 
loc_f1 = Tf1_1 + Tf1_2 + Tf1_3 + Tf1_4 + Tf1_5 + Tf1_6;

loc_f2 = Tf2_1 + Tf2_2 + Tf2_3 + Tf2_4 + Tf2_5 + Tf2_6;

loc_f3 = Tf3_1 + Tf3_2 + Tf3_3 + Tf3_4 + Tf3_5 + Tf3_6;
            

Any tips or ideas would be appreciated!

Answer 1

[Matt J]

If you have the Parallel Computing Toolbox, you can use parFeval,

https://www.mathworks.com/help/parallel-computing/parallel.pool.parfeval.html#mw_ccb5ab01-9b07-4593-896a-9b05ea63dfb9

[KieranSQ]

Perfect, this worked well! Many thanks.