What happens on Ctrl + C?

I have a problem that my code becomes slower with each iteration (I used a profiler, it is none of the usual culprits: unclosed figures or non-preallocated arrays). If I click Ctrl+C and continue where I started, it runs quickly again. Can someone please tell me what exactly happens on Ctrl+C? I tried writing my variables into a file, clearing all and re-starting, but it didn't help. I would like to know which function is called on this command so I can reproduce this behavior.

[Stephen23]

"..clearing..."

You should be writing functions, not scripts.

[Tiko]

I am writing functions. I have a script calling these functions, though. This "clearing" was a workaround I trired hoping it would solve the issue.

[Looky]

As I understand it, a Ctrl-C raises an interrupt flag and makes matlab stop the running code the next time it's able to check for said flag.(this differs for unix and windows) When the interrupt is detected, it stops the execution of your code, unstacks the stack (recording the line of execution it was in each function) and returns to the workspace.

Edit: I can't tell you which function is being called. I don't think this is accessible code.

About your programm I can only guess what's wrong without the actual code. From my experience, if a programm becomes slower and slower with each iteration, the most likely problem is an increasing need for memory or a recursive call that requires more and more memory for the stack.

When you use clear in a function, all you do is clearing the "workspace" of the current function. This is not the same as clearing all memory that is beeing using by your programm. So I think it would be best to have a look at your memory use. Bring up the task manager and observe the memory or better do it programmatically and try to find out if there is a memory leak or not. But again, this is only me guessing. Without the actual code there is not much we can help.

[Tiko]

I will try to look at the memory in more detail. What I have done so far was look at the profiler. Unfortunately, it doesn't give memory consumption (is there a profiler for that?). The profiler points to some Matlab-internal functions (see images I uploaded). The "return" does suggest that there is a memory leak. Do you have an idea how to clear the problematic memory?

[Tiko]

I also observed the memory in the task manager, as well as with the "memory" command. Interestingly enough, the task manager shows a memory increase. The "memory" command, placed every few iterations, shows a slight decrease. I wrote a script that calls the same function twice. Thus, in the second call it should not be using more memory, but the task manager begs to differ and the second simulation is slower as well. In either case, the task manager claims I am using about 15% of the total memory.

[Philip Borghesani]

Some additional information would help here: What MATLAB version are you using and are you using any custom MEX files?

Answer 1

[Philip Borghesani]

Two things happen when MATLAB returns back to the command line:

1. Memory leaked by any mex files is cleaned up (mxMalloc... or MATLAB arrays). The best practice is to only rely on automatic clean up in the event of an error.
2. Any pending display/graphics messages are processed by MATLAB. This can be forced by adding a drawnow to your code. In general this should not cause an issue because there should not be a huge number of events pending but there have been bugs in the distant past.