Problem with parfor classified variable

Question

Stefano Maffei 2019-7-30

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/474107-problem-with-parfor-classified-variable

评论： Stefano Maffei 2019-8-5

I am trying to parallelize a for loop. Here is a minimal working example of what I am trying to do at the moment.

nx = 96;
ny = nx;
Tk_all = zeros(nx,ny);
Tkp_all = zeros(2*kmax+1,kmax+1,nx,ny);
dim2 = 31;
dim3 = 31;
kmax = min(dim2,dim3);
c = parcluster('local');
c.NumWorkers = 24;
parpool(c, c.NumWorkers);
for kxi=-kmax:kmax
  parfor kyi=0:kmax
     Tk_entry = kxi*kyi*ones(nx,ny); % should be something more meaningful
     Tk_all(nx/2+(kxi-1)+1,ny/2+(kyi-1)+1) = sum(reshape(Tk_entry,nx*ny,1));
     Tkp_all(kmax+1+(kxi-1)+1,1+(kyi-1)+1,:,:) = Tk_entry;
  end
  Tk_all(nx/2-(kxi-1)+1,ny/2-([0:kmax]-1)+1) = Tk_all(nx/2+(kxi-1)+1,ny/2+([0:kmax]-1)+1);
end
delete(gcp('nocreate'))

and I receive the error

Error: The variable Tk_all in a parfor cannot be classified.

I though I was satisfying the requirements from, for example here:

https://uk.mathworks.com/help/parallel-computing/sliced-variable.html

but clearly I am missing something. Any help?

I apologize if this question seems to be a duplicate of other existing questions, but I cannot seem to figure out my specific problem.

Thanks!

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Abhilash Padma 2019-8-2

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/474107-problem-with-parfor-classified-variable#answer_385944

在 MATLAB Online 中打开

The variable Tk_alland Tkp_all are not meeting the requirements of a sliced variable. For a sliced variable, exactly one indexing expression must be of the form i, i+k, i-k, or k+i. The index i is the loop variable and k is a scalar integer constant or a simple (non-indexed) broadcast variable. Every other indexing expression is a positive integer constant, a simple (non-indexed) broadcast variable, a nested for-loop index variable, colon, or end. The following code will work in your case:

nx = 96;
ny = nx;
Tk_all = zeros(nx,ny);
dim2 = 31;
dim3 = 31;
kmax = min(dim2,dim3);
Tkp_all = zeros(2*kmax+1,kmax+1,nx,ny);
c = parcluster('local');
c.NumWorkers = 24;
parpool(c, c.NumWorkers);
for kxi=-kmax:kmax
    k1=kxi+nx/2;
    k2=ny/2;
    k3=kmax+1+kxi;
  parfor kyi=0:kmax
     Tk_entry = kxi*kyi*ones(nx,ny); % should be something more meaningful
     Tk_all(k1,k2+kyi) = 1;
     Tkp_all(k3,1+kyi,:,:) = Tk_entry;
  End
  Tk_all(nx/2-(kxi-1)+1,ny/2-((0:kmax)-1)+1) = Tk_all(nx/2+(kxi-1)+1,ny/2+((0:kmax)-1)+1);
end
delete(gcp('nocreate'));

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Stefano Maffei 2019-8-5

Hi, Thank you for the very fast answer, which seems to be working. However, as I mentioned in my original question. I have read the matlab guidelines to sliced variables, so I was familiar with what you said. My problem was in the interpretation of it, or in its application.

So for my benefit and for that of others reading the post, let me try and clarify what I was doing wrong.

If I am correct, my mistake was in using the index kxi inside the parfor loop? kxi is what you refer to as an indexed broadcast variable, while the k1, k2 and k3 variables are non-indexed and can be passed to the parfor loop. Correct?

请先登录，再进行评论。