Why accessing 2d matrix in parfor so slow?

Let's say I have a large matrix A:

A = rand(10000,10000);

The following serial code took around 0.5 seconds

tic
for i=1:5
    r=9999*rand(1);
    disp(A(round(r)+1, round(r)+1))
end
toc

Whereas the following code with parfor took around 47 seconds

tic
parfor i=1:5
    r=9999*rand(1);
    disp(A(round(r)+1, round(r)+1))
end
toc

How can I speed this up?

Answer 1

[Stephen23]

"How can I speed this up?"

1. Do you have a parallel pool open? I.e. created using parpool, or automatically? If no pool is open then a parfor loop will not be run in parallel anyway.
2. Don't use parfor.

Why do you think that parfor should be faster? The only time that parfor is faster is when the time saved doing the operations in parallel is greater than the overhead required. This topic has been discussed many time on this forum, which also suggest possible improvements to your code:

https://www.mathworks.com/matlabcentral/answers/389191-parfor-is-slower-than-for#answer_310685

https://www.mathworks.com/matlabcentral/answers/21478-slow-execution-of-parfor-loops-due-to-communication-overhead-load-static-data-into-worker-workspace

https://www.mathworks.com/matlabcentral/answers/32460-parfor-slower-than-for-for-simple-tutorial-example

https://www.mathworks.com/matlabcentral/answers/179468-trying-to-use-parfor-but-it-is-very-slow

https://www.mathworks.com/matlabcentral/answers/75166-why-is-parfor-slower-than-for

This is also explained in the documentation:

https://www.mathworks.com/help/distcomp/interactively-run-a-loop-in-parallel.html

Your loop contains very fast commands anyway, mostly just round and indexing, so it is quite possible that you will not see any benefit from parfor.