MATLAB: Does compiling a model containing Embedded MATLAB blocks using the GRT system target file take a long time in Real-Time Workshop 7.1 (R2008a)

simulink coder

I have a model which contains an Embedded MATLAB block and I pass large structures to it. The compilation takes several minutes while my other models compile without any substantial delays.

Best Answer

This enhancement has been incorporated in Release 2009b (R2009b). For previous product releases, read below for any possible workarounds:

The presence of large non-tunable parameters makes Embedded MATLAB compilation extremely slow. Since the value of a non-tunable parameter is known beforehand, Embedded MATLAB spends a lot of time trying to optimize the generated code based on the value and therefore results in a compilation delay.

Since tunable structures are not supported by Embedded MATLAB, in order to work around this issue, use arrays instead of structures for parameters.

Related Solutions

MATLAB: Does the application built using Real-Time Workshop fail with Embedded Target for Infineon C166 Microcontrollers 1.1 (R14)

This bug has been fixed for Release 14 SP1 (R14SP1). For previous releases, please read below for any possible workarounds:

This has been verified as a bug in Embedded Target for Infineon C166 Microcontrollers 1.1 (R14) in the way that it handles the parameter 'Maximum number of concurrent base-rate overruns'.

Currently, to work around this issue, refer to the following information:

In version 1.1 (R14) an incorrect value is being used for the maximum allowed number of concurrent base rate overruns. The effective value for this setting is one minus the value actually entered in the dialog under Tools - -> Real-Time Workshop --> Options--> C166 Options(1). For example, if a value of 2 is entered for "Maximum number of concurrent base-rate overruns" the then the maximum number of concurrent base rate overruns is actually 1. In particular, if a value of 0 is entered the application would fail.

MATLAB: Do I get a stack overflow exception when I debug a GRT executable containing Embedded MATLAB blocks using Microsoft Visual Studio

This behavior is observed because structures are not passed by reference in the code generated for Embedded MATLAB Function blocks, thus causing the C compiler to throw exceptions when the structures are very large.

As a workaround, one of the following suggestions may be implemented:

1. Use the eml.inline('always') directive in Embedded MATLAB Functions, in order to ensure that there are no function calls and thus no large structures being passed by value, leading to stack overflow.

2. Change the type of the parameter from structure to a matrix. Embedded MATLAB passes matrices by reference, thereby reducing the stack size.

3. Use the A=foo(A) pattern to force the structure to be passed by reference - every time a structure is passed to a function, use the following format:

[otherOutputs ... , A] = function_name(otherInputs, ... A)

Whenever Embedded MATLAB sees a function that receives a parameter and then returns it, it optimizes the function so that the parameter is passed by reference. This is the desired effect and will reduce the stack size.

Note that, in this case, the error seen when running the model in Rapid Accelerator Mode was due to the stack overflow. However, a stack overflow error does not necessarily always lead to a packet error or vice-versa.

Best Answer

Related Solutions

MATLAB: Does the application built using Real-Time Workshop fail with Embedded Target for Infineon C166 Microcontrollers 1.1 (R14)

MATLAB: Do I get a stack overflow exception when I debug a GRT executable containing Embedded MATLAB blocks using Microsoft Visual Studio

Related Question