How to synchronize parallel code execution while also sharing data?

myCluster = parcluster('local'); myCluster.NumWorkers=I; saveProfile(myCluster);
parpool(I)
spmd
    i=labindex;
    for j=0:100
    %each worker in matlab is assigned a node and each worker indexes into
    %the matrices accordingly
    c_t=c(i,:); x_t=x(:,i); A_t=A{i}; a_t=a(i,:); X_t=X(:,i);
    mem_num_t=mem_num(1,i); mem_den_t=mem_den(1,i);
    [x_hat,v_hat,mem_num_t,mem_den_t] = node(x_t,c_t,beta,a_t,A_t,b,v,j,mem_num_t,mem_den_t,X_t);
    v_hat
    v(:,i) = v_hat;
    labBarrier;
    end
end

In the above code I have I workers (in my example let I be 5). So, each worker executes the function node (I am implementing a distributed algorithm). The variables c, x, A, a and X are all variables declared before the spmd section. I am indexing them according to the worker id. However, in the code I noticed that I always end up indexing the first row and columns of the matrices. The same happens in the end where I want to store the column vector into the matrix v (which was also declared before spmd) but the line

v(:,i) = v_hat;

only indexes the first column of my v matrix and I cannot understand how I can index the matrix v according to worker id. I have searched online and am unable to solve this query. Hopefully, someone experienced here can answer me. Thank you!

[Vineet Ahirkar]

Try to print the value of "i" just before using it for indexing to verify its value.

I tried to create a similar script and was able to get the correct results on my end.

Try running this script and check the results -

I = 5;

if isempty(gcp('nocreate'))
    parpool(I)
end

v = rand(3,5);
disp('value of v: ');
disp(v);

spmd
    disp(labindex);
    
    for j=0:4
        disp(j)
        v(:,labindex) = rand(3, 1);
        disp('value of v: ');
        disp(v);
    end
    
    labBarrier;
end

Also, have a look at composite objects, they might help you in this scenario.

https://www.mathworks.com/help/distcomp/composite.composite.html