Minimize MDCS Data Transfer

I am using MDCS on a Linux cluster. I repeatedly for a parfor loop in the following spirit:

while 
    parfor i=1:n 
        x(i) = a(i,:)*b; 
    end 
end

Here the vector b is used on all cores, every time I call parfor. However, I believe each time I call parfor, b is being sent to all the cores, which slows down the code. How can I repeatedly use parfor with constant data such that I do not have to keep resending the same data each time I call a parfor loop? Thank you.

Answer 1

[Edric Ellis]

You could use the Worker Object Wrapper which is designed for this sort of situation.

[Philip]

Thank you! This is exactly what I needed- it really sped up my code.