MATLAB Answers

0

Help: restructuring parfor

Asked by Pramit Biswas on 6 Apr 2018
Latest activity Edited by Bhavanithya Thiraviaraja on 17 Apr 2018
totpV=5;
vars = rand(totpV,2);
for loop1=1:100
for loop2=1:totpV % Make this loop2 parallel using parfor
parameter=vars(loop2,:,loop1);
result(loop2,loop1)=feval('sum',parameter); % sample function
if min(result(loop2,:))<=result(loop2,loop1)
best=result(loop2,loop1);
end
vars(loop2,:,(loop1+1))= vars(loop2,:,loop1) + rand(1,2).*best;
end
end
I want to process loop2 parallelly, however, not getting how. From This page, probably this is due to Sliced Variables. Any idea to restructure?
Errors: The PARFOR loop cannot run due to the way variable 'vars', 'result' is used.

  1 Comment

Adam
on 6 Apr 2018
What are the warnings you get when you try?

Sign in to comment.

Tags

Products

1 Answer


There seems to be a dependency between the iterations in the line

min(result(loop2,:))<=result(loop2,loop1)

In parfor, the iterations are performed on independent MATLAB workers and hence it requires 'No Dependency'. In this case, the min(result(loop2,:)) seems to use the previously calculated values also, hence making it dependent.

Refer https://de.mathworks.com/help/distcomp/ensure-that-parfor-loop-iterations-are-independent.html for examples!

The dependency has to be broken without affecting the algorithm to use parfor.

  0 Comments

Sign in to comment.