How to access the local part of a distributed matrix separately for each Pid

I want to make a 5 by 5 distributed matrix in such a way that each column in a ith row has a value i. In pMatlab library, I tried implementing the following code.

PARALLEL = 1;
Map = 1;

if PARALLEL
    Map = map([Np 1], {}, 0:Np-1);
end

A = zeros(5,5,Map);
Aloc  = local(A);

for Pid = 0:Np-1
    if Pid == 0
        for i = 1:size(Aloc,1)
            Aloc(i,:) = i;
        end
    end
    if Pid == 1
        j = 4;
        for i = 1:size(Aloc,1)
            Aloc(i,:) = j;
        end
        j = j+1;
    end
end

A = put_local(A,Aloc);
A

However, it seems that the for loop from 0:Np-1 isn't able to access the local part of that Pid. Please help. How can we achive the same thing using PCT?

Answer 1

[Walter Roberson]

You overwrite all of Aloc for two different pids, 0 and 1. You can only overwrite all of an array inside parallel workers if it is a local array rather than an output array.

[Ayush Porwal]

I understand but how to work only on the local part of the array (separately for each Pid i.e. 0 & 1)? Can you please share some code?

[Walter Roberson]

Only index using the parfor variable.

parfor pid=0:5
  Aloc(:, pid+1) = pid;
end

[Walter Roberson]

parfor pid=0:5
    if pid == 0
       Aloc(:, pid+1) = 0;
    elseif pid == 1
        Aloc(:,pid+1) = 4;
    end
end

[Ayush Porwal]

Hi Walter

Thanks, I understand the code provided above. Have you worked using pMatlab library by MIT lincoln labs? parfor variable isn't available in that so I am stuck.

[Walter Roberson]

Ah, no, I had never seen pMatlab before.