connector not running message while executing parallel code on cluster

I am running the following test code on a cluster.

clc;clear all;

x=32;

poolobj=parpool(x);

A = zeros(32, 11);

tic

parfor (i = 1:32,x)

A(i,:) = compute(A(i,:),i);

end

toc

disp(A);

The function file is a separate file

% This function now contains the body

% of the parfor-loop

function A = compute(A,i)

for j = 1:11

A(j) = i + j - 1;

end

end

Output- It displays the A matrix but it gives a message afterwards

>> >> >> message with properties:

Identifier: 'MATLAB:connector:connector:ConnectorNotRunning'

Arguments: {}

Is the for loop actually working on 32 cores ? It does give a message that a parallel pool of 32 workers has been started. Although, the time taken for lesser cores is smaller. What does this message imply ?

[Prem Ankur]

Yes, the code will run on the 32 workers in the parpool. Total time taken depends on the length of time taken by the compute function in the code and then the overhead of sending the data to the 32 workers and receiving the results. Overhead time can be greater than the benefit of parallelizing to the 32 workers. Doing something more computationally intensive will show better scaling properties.

I am not sure about the connector message. Can you share your startup.m or matlabrc.m file?

[Soumyadipta Sengupta]

Thanks for the answer. I am running it on a cluster in which matlab is preinstalled. Hence, i don't have acess to the files.

Answer 1

[Prem Ankur]

https://www.mathworks.com/matlabcentral/answers/436498-connector-not-running-message-while-executing-parallel-code-on-cluster#comment_653284