Parfor error: parfor variable is indexed in different ways

I am trying to convert my code to run in parallel, but I get an error when I use the parfor : "in a parfor loop, variable is indexed in different ways". I can not understand, how is this being indexed in different ways? The order that each iteration is completed does not matter, and each iteration is completely independent.

   parfor g = 1:ng

            [A] = function(g)

            B(g, :, 1) = mean( A(:,:) );
            B(g, :, 2) = std( A(:,:) );

thanks end

Answer 1

[Edric Ellis]

Your variable 'B' has two different lists of subscripts - to run in PARFOR, you need to provide precisely the same list of subscripts. You can fix this by converting your expression to be a single indexed assignment. You need something like this

numColsInB = size(B,2);
...
parfor ...
  newVals = [mean(A(:, :)), std(A(:, :))];
  B(g, :, 1:2) = reshape(newVals, 1, numColsInB, 2);
end

[Denis Menshykau]

Dear Edric,

I have a similar problem. Following your example I created the following simple test:

   nPar=10;
data=zeros(nPar,2);
parfor n=1:nPar
    v1=n^2;
    v2=n^2-1;
    newV=[v1, v2];
    data(n,1:2)=reshape(newV,1,2);
end

however matlab generates an error: "Error using testParfor Error: The variable data in a parfor cannot be classified."

What is the problem?

Regards, Denis