CODEWITHSUNDEEP
matlabmemory
HByrm
HByrm

Reputation: 41

Growing memory usage in MATLAB

I use MATLAB for programming some meta-heuristics. Recently, I have been working on an algorithm for solving an industrial engineering problem. My problem with MATLAB is getting "out of memory" errors. Now I'm trying some suggestions from Mathworks and Stackoverflow (Hope they will work). However, there is one thing I did not understand.

During the run of the algorithm in MATLAB (it takes 4000-5000 cpu sec for a medium sized problem), even though I preallocate variables, code does not demand dynamic array resizing and does not add new variables, I observe that the memory usage of the algorithm grows continuously. The main function calls some other functions written by me. What could be the reason of increase of the memory usage?

The computer I use for the running of the algorithm has 8GBs of memory and win8 64bit installed.

Upvotes: 0

Views: 681

Answers (2)

HByrm
HByrm

Reputation: 41

The reason of memory usage growth is CPlex. I tried many alternatives but I couldn't find any other useful solution than increasing virtual memory to several hundred GBs. If you don't have special reasons to insist on CPlex (commercial usage, licensing etc.), I would suggest anyone, who encounter this problem, to use GUROBI. It is free and unlimited for academic usage, totally integrable with MATLAB. That's the solution I have found for my problem with Cplex. I hope this solution works for everybody.

Upvotes: 0

Dennis Jaheruddin
Dennis Jaheruddin

Reputation: 21563

The only way to figure this out is to see where the memory is going. I think you may accidentally store results that you don't need, or that you underestimate the size of your output/intermediate variables.

Here is how I would proceed:

  1. Turn on dbstop if error
  2. Run the code till you get the out of memory error
  3. See how much memory is being used (make sure to check all work spaces)

Probably you now know where the extra memory is going. If you don't find much memory being used, continue with this:

  1. Check the memory command to see how much memory is still available
  2. Carefully look at the line being executed, perhaps you actually need a huge amount of memory for it

If all else fails share your findings here and others can help you look for it.

Upvotes: 0

Related Questions