Creating a big Variable with splitting into numbered .mat-Files

3 views (last 30 days)
Hello Guys, i need to create a "big" Variable but my Memory(4GB) is not big enough so i want now to split it into some numbered .mat-Files. With this code i would have the Variable Buff with 6.765.201 rows and 4 columns:
Buff = zeros(numel(1350:2:1450)*numel(-150:2:-50)*numel(825:2:925)*numel(1225:2:1325),4);
Ctr = 1;
for step1 = 1350:2:1450
for step2=-150:2:-50
for step3=825:2:925
for step4=1225:2:1325
Buff(Ctr, 1) = step1;
Buff(Ctr, 2) = step2;
Buff(Ctr, 3) = step3;
Buff(Ctr, 4) = step4;
Ctr = Ctr+1;
end
end
end
end
But i want that matlab create Buff with 1 to 500.000 rows and 4 columns into saving_Buff_001.mat then the next 500.000 to 1.000.000 rows and 4 columns into saving_Buff_002.mat and 1.000.000 to 1.500.000 to saving_Buff_003 and so on...
My idea was something like this, but matlab creates first the big variable and then starts splitting
idx = mod(Ctr-1,500000)+1;
if idx == 500000
index = Ctr/500000;
save(sprintf('saving_Buff_%03d.mat',index),'Buff');
end
Hope somebody can help my, big thanks
  2 Comments
Matt J
Matt J on 12 Jan 2013
With this code i would have the Variable Buff with 6.765.201 rows and 4 columns.
A matrix of that size is only 206 MB. Should be no problem for a system with 4 GB RAM.

Sign in to comment.

Answers (2)

Matt J
Matt J on 12 Jan 2013
In addition to memmapfile, there is also matfile, and you could put the whole thing in one file. No need to split into multiple files, as far as I can see, though I would recommend writing column-wise instead of row-wise
matObj=matfile('myfile','Writable',true);
for i=1:4
matObj.Buff(:,i)=...;
end

Trier87
Trier87 on 12 Jan 2013
Thx for the informations. This was only a example, my Real Variable has 215.000.0000 rows and 4 columns and thats a very large Variable so i need a pre initialization of the Variable an the Splitting. So any ideas how to solve this Problem. Thanks
  2 Comments
José-Luis
José-Luis on 12 Jan 2013
Do you really need so large a variable in memory? Could you not just load what you need? Also, in the future, please don't place comments as answers as it might get confusing for those trying to join the discussion.

Sign in to comment.

Community Treasure Hunt

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

Start Hunting!