Hello, I am trying to use PARFOR in a loop which runs a strcmpi() function. The code is:
parfor i = 1:size(data_1,1)
idx = find(strcmpi(data_1{i,1},data_2(:,1)));
dd = data_2(idx,:);
for j = 1:size(dd,1)
fprintf(file,'%s\t%s\n',dd{j,1},dd{j,2}); % print to file
end
end
I am not able to slice the "data_2" which is a 31mn x 2 cell array due to which I am not able to run parfor. Can anyone suggest how I can modify this program for parfor? Moreover, will this effect the way I am printing the data in the file?
Thank you.

1 commentaire

find(strcmpi(data_1{i,1},data_2(:,1)))
... you are expecting multiple matches?
Is either data_1(:,1) or data_2(:,1) a set of unique strings?

Connectez-vous pour commenter.

 Réponse acceptée

Edric Ellis
Edric Ellis le 28 Nov 2017
Modifié(e) : Edric Ellis le 28 Nov 2017

0 votes

The variable data_2 is not being sliced because every iteration of the parfor loop needs access to the whole array. This is not an error though - the code analyzer warning is simply telling you that the code is incurring extra communication because the whole of data_2 is being sent to each worker.

4 commentaires

Anurag Passi
Anurag Passi le 28 Nov 2017
So, why does the parpool sits idle and shuts off? Why am I not able to run it on parallel? What can be changed in this? Below is the actual error that I get.
Walter Roberson
Walter Roberson le 28 Nov 2017
You are running out of memory trying to send the data to all of the workers. The data is converted to a byte stream and then the byte stream is sent to the workers and the workers reconstruct the objects; this requires a temporary copy of the data that could have you running out of memory.
Anurag Passi
Anurag Passi le 28 Nov 2017
So I think maybe I should try some other way of working with this data. Thank you for your replies.
Walter Roberson
Walter Roberson le 28 Nov 2017
If either data_1(:,1) or data_2(:,1) are sets of unique strings then ismember() can be used before-hand to find the indexing.

Connectez-vous pour commenter.

Plus de réponses (0)

Catégories

En savoir plus sur Large Files and Big Data dans Centre d'aide et File Exchange

Community Treasure Hunt

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

Start Hunting!

Translated by