unable to classify the variable 'F' in the body of the parfor-loop

I was trying to reduce the time taken by parfor and I have tried to calculate F (scatteredInterpolant() ) before the parfor loop

F = scatteredInterpolant(x,y,z,data2(1:len),'linear','linear');

Then, I tried to use interpolant object F inside the parfor loop

   parfor i=1:size(data,2)
       F.Values=data2(1:len)
   end
   

but the following error was generated

unable to classify the variable 'F' in the body of the parfor-loop. For more information ,see parallel for loops in MATLAB "SOLVE VARIABLE Classification issues in parfor-loops

Answer 1

[Edric Ellis]

This use of F creates an order depdency between loop iterations. (You're modifying F on each loop iteration). I'm not entirely sure what you're trying to do here, but perhaps you want to do something more like this?

rng('default')
x = -2.5 + 5*rand([50 1]);
y = -2.5 + 5*rand([50 1]);
v = x.*exp(-x.^2-y.^2);
F = scatteredInterpolant(x,y,v);
parfor i = 1:10
    % Make a copy of F
    Ftmp = F;
    % Modify the copy - this is fine
    vnew = x.^2 + y.^2 + i;
    Ftmp.Values = vnew;
    out(i) = F(i, i);
end

Starting parallel pool (parpool) using the 'local' profile ...
Connected to the parallel pool (number of workers: 2).

disp(out)

    0.1631   -0.0129   -0.0522   -0.0989   -0.1457   -0.1924   -0.2392   -0.2859   -0.3327   -0.3794

Answer 2

[zein]

Thanks for your help, I really appreciate it @Edric Ellis

I have already tried to modify the lines according to your recommendation in your comment, but I got the same error message.

I have also attached the M file and raw file data, The matlab code use to running the code

https://drive.google.com/drive/folders/12tiQ1pbr6mFW8EP9y3IV8yc515xCwIsv?usp=sharing

line 57 where the scatterinterpolation is done

F = scatteredInterpolant(x,y,z,data2(1:len),'linear','linear');

Then, the parfor loop

The F object is called in line 94-95

parpool(8);
%FF=zeros
%parfor (i=1:size(data,2),4)
   parfor i=1:size(data,2)
    %for i=1:size(data,2)
    filename=[surfacename,num2str(sub),var,num2str(file_no(i)),'.raw'];
    fullname=fullfile(Dir,filename);
    fid = fopen(fullname,'rb'); % rb = read binary
    NN=N*10;
    data2 = fread(fid,NN,'single');
    fclose(fid);
    start=8;
    data2=data2(start:length(data2));
    len=length(data2)/no_variables;
    if opt==1 ||opt==6;
    data(:,i)=data2(1:len);
    elseif opt==2;
    data(:,i)=data2(len+1:2*len);
    elseif opt==3;
    data(:,i)=data2(2*len+1:3*len);
    elseif opt==4;
    data(:,i)=data2(3*len+1:4*len);
    elseif opt==5;
    data(:,i)=data2(4*len+1:5*len);
    elseif opt==7;
    %F.Values=data2(1:len)
    %F = scatteredInterpolant(x,y,z,data2(1:len),'linear','linear');
    Ftmp=F;
    Ftmp.Values=data2(1:len)
    data_int=Ftmp(xq,yq,zq);   
    data_int_2= reshape(data_int,[],1);
    end
   end
   

[Edric Ellis]

In the attached code, you still have lines inside the parfor loop that assign to F.Values - lines 95; 100; 105; 109; 114. Even if these lines don't execute at runtime, the parfor analysis must be satisfied that nothing order-dependent could ever happen using static analysis and ignoring the value of opt. It might help to assign Ftmp = F at the start of the loop, and then use that everywhere inside the loop.

[zein]

That solves the problem, thanks a lot for you help @Edric Ellis