parfor loop variable classification

Hi all,

I have always had trouble getting parlor loops to work for me, and again I'm having a variable classification issue. The code is for a machine vision program that we are using to detect blobs in multiple behavioral chambers in a video. We would basically like to use a parfor loop to find the blobs in each frame more rapidly. We'll track them across frames later. This way it would seem like if we built a small computer cluster, we could speed up the code a lot. Basically this is what we are trying to do with several of our variables (I've shown only one for an example).

recordAValue = NaN(blobsPerChamber, numberOfChambers, numberOfFrames);

parfor g = 1:numberOfFrames
  load(frame);
  for k = 1:numberOfChambers
    findBlobs
    for j = 1: blobsPerChamber
      recordAValue(j,k,g) = someCalculation;
    end
  end
end

The error that I would get is that recordAValue cannot be classified. I've looked at the documentation but I still can't make sense of it. The loops all seem independent to me and the matrix recordAValue should be able to be generated one loop at a time.

Any help is appreciated. Thanks in advance.

[Edric Ellis]

Hm. The following simplified example works for me:

out = NaN(6, 5, 4);
parfor idx = 1:4
    for jdx = 1:5
        for kdx = 1:6
            out(kdx, jdx, idx) = 7;
        end
    end
end

What version of MATLAB are you using?

[Adam]

What are 'someCalculation' and 'findBlobs' and where/how are they defined? They may be breaking transparency rules for parfor loops as they could theoretically be anything as it is currently stated.