Calling multiple outputs from a function in a parfor loop

Hi,

I am trying to call the multiple outputs of different functions in a 'parfor' loop, using multiple workers by 'matlabpool' command. To the best of my knowledge, I've got to use a 'nested Cell Array', as I have copied below.

i) The called function (1)

    function [ M1,T1,G1 ] = Tr1(Nul)

    syms m1 t1 

    M1=sym('M1',[5 1]);

    T1=sym('T1',[5 5]);

    G1=ones(2,2)*300;

    for i=1:5

        M1(i)=m1^i;

        T1(i)=t1*i; 

    end

    end

ii) The called function (2)

function [ M2,T2,G2 ] = Tr2(Nul)

syms m2 t2 

M2=sym('M2',[5 1]);

T2=sym('T2',[5 5]);

G2=ones(2,2)*600;

for i=1:5

      M2(i)=m2^i;

      T2(i)=t2*i; 

end

end

iii) And the main program with the error message

matlabpool 2

Funlist={@Tr1, @Tr2};

parfor i=1:2

[Data{i^2:3*i}]=Funlist{i}(0)

end

And the error

Error: The variable Data in a parfor cannot be classified.
See Parallel for Loops in MATLAB, "Overview".

I even tried to simplify the expressions and then pass them into the nested Cell Array "Data" and again the same error, copied below

>> parfor i=1:2

t=i^2;

j=3*i;

[Data{t,j}]=Funlist{i}(0)

end

Error: The variable Data in a parfor cannot be classified.
See Parallel for Loops in MATLAB, "Overview".

I am expecting to get Data as a nested cell array Data={[1*3],[1*3]}, where Data{1}=[5*1],[5*5],[2,2]

This is a simple example I am tying to work with to figure out how to handle a much more time consuming and complicated program, where I need to call the Multiple Outputs of 4 different functions into a main program and to save the time, I have to do it in parallel. (matlabpool, parfor). I know that this structure works with "for loops", but not of much help to me. Could you please recommend a solution to handle such issues? (MATLAB Version: 8.0.0.783 (R2012b))

Thanks a lot!

Answer 1

[Matt J]

Something like the following perhaps? If Data isn't in the precise shape you want you can reshape() it after the loop.

     Funlist={@Tr1, @Tr2}; %sliced variable

     Data=cell(1,2); %sliced variable

     parfor i=1:2

          out={}; %temporary variable

          numout=3*i+1-i^2; %temporary variable

       [out{1:numout}]=Funlist{i}(0);

       Data{i}=out; 

     end

[Matt J]

It's also good practice to plan how you want the parfor loop to classify your variables using the classification rules described at,

http://www.mathworks.com/help/releases/R2014b/distcomp/classification-of-variables-in-parfor-loops.html

and then label them accordingly as I have done (sliced, temporary, reduction, etc...) using code comments.

[Parmida Boroumand]

Thanx :)