How to increase the memory that MATLAB uses?

I have 19 GB of memory available for running Matlab. However, Matlab is only using 1 GB of memory. This is adversely affecting my ability to run my code. How do I increase the amount of memory that Matlab is using?

Answer 1

[OCDER]

This was answered here:

https://www.mathworks.com/matlabcentral/answers/32332-increase-memory-used-by-matlab

Matlab will use more memory if you need more memory. It's not capped at 1 GB. To test this, try:

>> memory
>> a = rand(20000);
>> memory

You'll notice that the memory usage by Matlab increases

Is there an error like "Out of Memory"? If yes, then it's indeed a memory issue that needs to be fixed by changing the way your code loads large data. If not, then it's not a memory issue but slow algorithm issue. These articles can help:

https://www.mathworks.com/company/newsletters/articles/programming-patterns-maximizing-code-performance-by-optimizing-memory-access.html

https://www.mathworks.com/help/matlab/matlab_prog/resolving-out-of-memory-errors.html

https://www.mathworks.com/videos/optimizing-and-accelerating-your-matlab-code-107711.html

To help you determine the code bottleneck, use the profile: https://www.mathworks.com/help/matlab/ref/profile.html

[Babak]

@Walter Roberson My plots are very laggy.

There are 8 figures with 50 curves each and when I go to "Plot Browser" mode and uncheck "Edit Plot" the zoom in/out/drag and restore view become very laggy. How can I improve memory usage or plot algorithm?

Answer 2

[April Bryan]

Hi,

Thank you for your response. I did not have an out-of-memory issue. Also, I tried to use as much good programming practice as I can.

When I used the code you suggested, I received an increase in memory. However, when I close Matlab, the increase disappears and it is necessary for me to re-run the code again. Is there a way to obtain a permanent increase in memory?

Thanks

[Steven Lord]

When you say "permanent increase in memory" do you mean an increase in the amount of memory MATLAB is using (which is easy) or do you mean an increase in the amount of memory available to MATLAB (which may not be so easy.)

Which release of MATLAB are you using?

Are you using a 32-bit or 64-bit version of MATLAB?

[Walter Roberson]

If you are not getting "out of memory", then what adverse impact are you seeing on your code?

Are you running on a system that is being used for other memory-intensive purposes as well, and you want to somehow "reserve" memory for MATLAB so that it does not have to dynamically compete with the other programs that are asking for a lot of memory?

[Hoang-Le Tran]

If I am running on a system that is being used for other memory-intensive purposes, is there a way to reserve memory specifically for Matlab such that there is no competition?

[Walter Roberson]

@Hoang-Le Tran

You just might be able to do that with RTLinux (Real Time Linux), but as far as I know it cannot be done with regular Linux.

For modern versions of MacOS, the most you can do is increase the process priority using the renice utility. Sufficiently old versions of MacOS had other ways available as well.

For Windows, as far as I know the most you can do is increase hte process priority; https://www.auslogics.com/en/articles/change-priority-of-win10-process-with-command-line/

Answer 3

[April Bryan]

I meant an increase in the amount of memory available to Matlab. The memory >> rand >> memory series of commands, allows me to gain a temporary increase in the amount of memory that Matlab is using. However, I have to reset it every time I open Matlab for the first time.

Prior to using the memory >> rand >> memory series of commands, Matlab would only use 34 % of my available memory. When I opened other programs, then my memory usage increased. I am not sure what was capping the availability of memory. The series of memory commands allowed for the usage of up to 85 % of available memory by Matlab. However, as mentioned, I have to re-run the commands, every time I open Matlab.

The cap in the availability of memory was causing my program to run slowly. When the cap was released, I saw a 10 time improvement in performance.

[OCDER]

The memory >> rand >> memory command is only there to show you that Matlab changes it's memory usage dynamically, depending how much memory you need. And how are you removing the memory cap? Is doing a = rand(20000) making your code faster? That is indeed odd, as a = rand(20000) decreases your available memory...

Sometimes the initialization of Matlab is slow, and running a script the first time is slow. Is it still slow when you rerun the same script? Maybe this Q&A is relevant:

https://www.mathworks.com/matlabcentral/answers/353965-why-my-matlab-2017a-is-so-slow-for-even-simple-commands

Have you tried the profile function to see what is causing your program to run slowly?

%Try this example, directly from "help profile" command
profile on       %Begin tracking code performance
plot(magic(35))  %Replace this line with your code summon that's slow
profile viewer   %View time to run each line of code. Which one is slow?

Try the profiler first and locate the lines of code that are taking a long time. Without this info and code, it's hard to debug.

Last, use the Comments section instead of the Answer section to respond.

Answer 4

[April Bryan]

Thanks. That worked great! I was able to find the bottlenecks in the code with the profile command.