MATLAB: Is CTRL+C unpredictable or ineffective when trying to terminate a script running in MATLAB 6.0 (R12) on a SGI machine running the Common Desktop Environment (CDE)

breakccdecontrolcontrol-cctrlctrl-cMATLABstopterminateunix

When MATLAB R12 is launched from a SGI machine using the Common Desktop Environment (CDE) icon, CTRL+C does not successfully terminate running scripts as it does when MATLAB is launched from a shell command line.

Best Answer

If MATLAB 6.x (R12.x and R13) is launched from a CDE icon on a SGI machine, a minor conflict exists between the CDE and the MATLAB desktop. This causes CTRL+C to be ineffective in terminating a running MATLAB script.

There are two ways to work around this issue:

1. Open a shell and launch MATLAB from the command line.

2. Change the CDE icon so that it launches "matlab -nodesktop" rather than simply "matlab".

You can initiate the desktop from the MATLAB command line by typing:

desktop

Loading the desktop after MATLAB has been launched seems to avoid the conflict and allows CTRL+C to be used properly.

Related Solutions

MATLAB: Is MATLAB officially supported on Mac OS X machines with Intel based processors

As of R2007a, MATLAB is supported on all Intel Macs, including the MacBook Pro. Intel based Mac OS X machines are not supported on prior releases. Refer to the following link for our system requirements for Mac OS X:

https://www.mathworks.com/support/requirements/matlab-system-requirements.html

MATLAB: Are there memory limitations for MATLAB 6.0 (R12) on SGI64 machines

MATLAB 5.3 (R11) was available in a 64-bit build, however MATLAB 6.0 (R12) is only available as a 32-bit application. The reason for this difference is that the MATLAB 6.0 (R12) IDE desktop requires a Java Virtual Machine (JVM) to run, whereas MATLAB R11 did not. SGI does not produce a 64-bit JVM at this point, so it seemed there was no advantage to supporting a 64-bit version of MATLAB 6.0 (R12).

Because MATLAB 6.0 (R12) is a 32-bit application, it cannot take full advantage of the available memory on 64-bit SGI machines. Here are the main issues:

1. When running under a 64-bit SGI machine, in MATLAB R11 you can run either the SGI build (32-bit) or the SGI64 build. With the 64-bit build you can have a heap size of up to 2^40. Space is not an issue. However, there are some 32-bit limitations. The number of pointers in a given matrix cannot exceed 2^32 elements, and the precision is still out to 15 digits.

2. With MATLAB 6.0 (R12) we have ONLY the 32-bit build and in theory the heap size of up to 2^31 is allowed. Actually, the 2^31 includes all text, stack, shared libraries, and the heap.

As an example, let's look at some typical memory limits set by a user:

 cputime         unlimited
 filesize        unlimited
 datasize        unlimited
 stacksize       524288 kbytes
 coredumpsize    unlimited
 memoryuse       524288 kbytes
 vmemoryuse      unlimited
 descriptors     200
 threads         1024

As you can see, this user has a very large stack size, namely 524288 KB. That is 25% of the available space. We believe that when the program is mapped this space is not available for the heap.

In order to alleviate the problem, you can try decreasing the stack space to say 65536 KBi.e. 65 MB. MATLAB does not need such a huge stack space. It uses the heap for most of its storage.

You can set this in $MATLAB/bin/.matlab6rc.sh (where $MATLAB is the root MATLAB directory) under the SGI section to be:

       limit stacksize 65536k

You can try it first from the command line to see if it helps.

Best Answer

Related Solutions

MATLAB: Is MATLAB officially supported on Mac OS X machines with Intel based processors

MATLAB: Are there memory limitations for MATLAB 6.0 (R12) on SGI64 machines

Related Question