MATLAB: How to handle a TLC Compiler out of memory error upon generating code for a Simulink model on a newer version of MATLAB while it works fine on an older version

Embedded Coder

I have a Simulink model that builds with successful code generation in release R2009b. I have upgraded to the latest version and reuse the same model to generate code. However, I receive an 'OUT_OF_MEMORY' error reported by the TLC Compiler and the model encounters a build failure.

ERROR: ### Loading TLC function libraries
TLC Compiler encountered an OUT_OF_MEMORY condition. Attempting to clean up...

I have the following system configuration:

RAM: 4GB

Swap Virtual Memory: 2GB

I checked memory in use using Windows Task Manager utility, but the physical memory in use is 90% a the point of OUT_OF_MEMORY error. This effectively means, that 10% remaining RAM is still available for use.

Best Answer

Memory requirements of a system change between different releases of MATLAB. Mostly with additional features and toolboxes added to every release of MATLAB, the memory requirements increase. However, this doesn't directly affect the memory usage by a TLC compiler during the build phase.

In an scenario described above, the process can be investigated by studying the build process for the model to code conversion. The Simulink model prior to generation of code, undergoes several phases that include:

1. Data initialization

2. Intermediate representation of the model

3. TLC build

4. Final code generation

Each of the steps above require larger memory for a very large model. The strict requirement implicit to the process is a contiguous block of memory, however large the memory requirement is. Based on this reasoning, if the final TLC build phase does not have a contiguous memory space as required in the RAM, the TLC compiler throws the OUT_OF_MEMORY exception, further, aborting the model build process.

Some to this problem on a Windows 32 bit machine are listed below. Their detailed usage is described enlisted in the documentation page listed.

Techniques to counter TLC OUT_OF_MEMORY exception:

1. Use 3GB switch to allocate more memory for MATLAB process

2. Start MATLAB from the $MATLABROOT$\bin directory as opposed to $MATLABROOT$\bin\win32 directory

3. Start MATLAB using -nojvm and -no desktop option

4. If the model does not use any Java Virtual Machine call, set the Java Heap space to the minimum. This can be done in "Preferences->General->Java Heap Memory".

Please refer to the following link for 3GB switch:

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

Related Solutions

MATLAB: Can I access more than 2GB of memory when running a 32-bit MATLAB 7.5 (2007b) on a Mac

In theory, 32-bit applications can address up to 4GB of virtual memory.

However, a number of things affect the actual memory accessibility:

1. The Operating System reserves memory for itself, reducing the amount of memory its applications can use.

2. Other applications running at the same time can eat up memory as well.

3. MATLAB reserves memory for itself. Even when there is no code running and there are no variables in the workspace, MATLAB is still using memory.

4. The memory available to MATLAB may be fragmented when starting additional tools or toolboxes, or loading MATLAB functions into memory. This limits the size of the largest contiguous block of memory. Since MATLAB matrices need to be stored in contiguous blocks, Out of Memory errors can occur while trying to create a matrix larger than the size of the largest contiguous block, even if that memory exists in several fragmented blocks.

There are tools for memory management for Macs:

The TOP command allows you to view the RPRVT (amount of memory in use) and VSIZE(amount of virtual memory paged so far) for the applications with top CPU usage: