for loops nested inside parfor loops.

3 views (last 30 days)
Hallo everyone, I have a bit of a problem with some parfor coding. The concept is this:
parfor j = 1:1:10;
for i = 1:1:10;
Q(i) = i^2;
end
R = Q(j);
end
Of course this code is non-sensicle, but it illustrates the problem. I cannot create an array of values inside a parfor loop that I can then use throughout the parfor loop. In a practical application the expression for Q(i) will depend on j, so taking the sub-loop outside the parfor in not an option. Therefore, for now I wish to know why such a simple example can't be parallelised. As far as the algorithm is concerned it has to individually loop through i and then compute R in parallel with no obvious interdependency between the loops. Many thanks.

Accepted Answer

Titus Edelhofer
Titus Edelhofer on 16 Apr 2015
Hi,
if you add the following line before the inner loop, it should work:
Q = zeros(10, 1);
Titus

More Answers (1)

Alexantrou Serb
Alexantrou Serb on 16 Apr 2015
It did indeed. Thanks!
Does this happen because the parfor needs to allocate space for Q for each independent thread it generates?
  2 Comments
Edric Ellis
Edric Ellis on 16 Apr 2015
No, it's because parfor cannot immediately deduce that you're not re-using values from Q in subsequent iterations of the loop. By assigning to the whole of Q, then it can correctly deduce that Q is a temporary variable. More details about parfor variable classification here.
BHUSHAN MUTHIYAN
BHUSHAN MUTHIYAN on 5 Sep 2017
Edited: BHUSHAN MUTHIYAN on 5 Sep 2017
Hello Edric,
If I have to change the global variable which is declared outside the parfor loop, I am getting an error. Any reasoning for this ?
My code looks something like this :
out = zeros(dimfox, dimfoy, fis(4), ins(4), 'like', sumModel);
H = vision.Convolver('OutputSize', 'Valid', 'CustomProductDataType',numerictype([],32,frac), 'CustomOutputDataType', numerictype([],32,frac));
parfor im = 1:ins(4)
carve = zeros(dimfox, dimfoy, 'like', sumModel);
for f = 1:fis(4)
for ch = 1:fis(3)
carve(:) = carve + step(H,input(:,:,ch,im),filter(:,:,ch,f));
end
carve(:) = carve + bias(1, f);
out(:,:,f,im) = carve;
end
end
I am getting error as :
The variable out in a parfor cannot be classified.

Sign in to comment.

Categories

Find more on Data Distribution Plots in Help Center and File Exchange

Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!