A comprehensive, production-ready template for creating custom S-functions that interface C++ code with Simulink. This enables seamless integration of third-party libraries and custom algorithms into Simulink models and supports automated code generation with Simulink Coder.

S-functions (System-functions) provide a powerful mechanism to extend Simulink with custom C++ implementations. This repository demonstrates best practices for:

• Integrating external libraries (e.g., OpenCV) into Simulink
• Managing complex C++ objects within Simulink blocks
• Supporting both simulation and code generation workflows
• Handling memory safely with multiple memory management strategies
• Creating maintainable, well-documented embedded code

• Professional code structure with comprehensive Doxygen documentation
• Dual memory management modes for flexibility
• Cross-platform support (Windows, Linux, macOS)
• Example implementation with optical flow velocity estimation
• CMake-based build system with automatic MATLAB detection (FindMatlab.cmake)
• Code generation support with 4 auto-generated configuration files:
  • Include directories with full paths
  • Library directories for linker search paths
  • Library files with platform-specific extensions
  • Source files for compilation

├── CMakeLists.txt                     # Build configuration with FindMatlab.cmake
├── cmake/
│   └── FindMatlab.cmake               # MATLAB auto-detection module
├── include/
│   └── optical_flow_velocity.hpp  # Custom class headers (Doxygen documented)
├── src/
│   ├── s_function.cpp                 # S-function implementation (fully documented)
│   └── optical_flow_velocity.cpp      # Custom class implementation
├── test_s_function.slx                # Example Simulink model
├── include_directories.txt            # Auto-generated (include paths)
├── link_directories.txt               # Auto-generated (library search paths)
├── link_libraries.txt                 # Auto-generated (library files with extensions)
├── source_files.txt                   # Auto-generated (source files)
├── LICENSE
└── README.md

• MATLAB/Simulink with Simulink Coder (R2018b or newer recommended)
• CMake 3.10 or higher
• C++ compiler compatible with your MATLAB version:
  • Windows: MSVC (Visual Studio 2017 or newer)
  • Linux: GCC 6.3 or newer
  • macOS: Clang (Xcode 9.0 or newer)
• Third-party libraries: OpenCV (or other libraries as needed)

# Clone the repository
git clone https://github.com/amaldevh/s-function.git
cd s-function

# Create build directory
mkdir build && cd build

# Configure and build (MATLAB auto-detected via FindMatlab.cmake)
cmake ..

The build process will:

• Auto-detect your MATLAB installation (or you can specify it manually)
• Compile the MEX S-function
• Generate 4 configuration files for Simulink Coder:
  • include_directories.txt
  • link_directories.txt
  • link_libraries.txt
  • source_files.txt

Or specify MATLAB location explicitly:

On Windows:

cmake .. -DMATLAB_ROOT="C:/Program Files/MATLAB/R2023b"

On Linux/macOS:

cmake .. -DMATLAB_ROOT=/usr/local/MATLAB/R2023b

1. Copy the generated MEX file from build/ to your Simulink model directory
2. Open your Simulink model
3. Add an S-Function block from the Simulink library
4. Set the S-function name to s_function
5. Configure the S-function parameters:
   • P1: Camera focal length (meters)
   • P2: Sensor height (meters)
   • P3: Sensor width (meters)
   • P4: Instance ID (unique integer)
   • P5: Image height (pixels)
   • P6: Image width (pixels)
6. Connect inputs and run the simulation

See test_s_function.slx for a complete example.

1. Install OpenCV, for example the precompiled binaries from sourceforge at OpenCV Compiled library Sourceforge

2. Install the MinGW-64 C/C++ toolbox from Matlab file exchange at MingW-64

3. Find Mingw-64 path by executing following command in Matlab window

   %% Find the value of MW_MINGW64_LOC set by matlab
   mingw_path = getenv('MW_MINGW64_LOC')

4. Open Powershell and execute

# Extract the OpenCV installation

# Set the MW_MINGW64_LOC env var
$env:MW_MINGW64_LOC = "mingw_path"

# Clone the repository
git clone https://github.com/amaldevh/s-function.git
cd s-function

# Create build directory
mkdir build && cd build

# Configure and build (MATLAB auto-detected via FindMatlab.cmake)
cmake .. -DCMAKE_PREFIX_PATH="\path\to\opencv\build\dir"

Place your C++ source files in src/ and headers in include/:

// include/your_algorithm.hpp
/**
 * @brief Your custom algorithm class
 */
class YourAlgorithm {
public:
    YourAlgorithm(double param1, double param2);
    void processData(const std::vector<double>& input);
    std::vector<double> getResults() const;
};

Choose one of two memory management strategies in s_function.cpp:

Option A: Persistent Memory (Recommended)

• Better for complex objects
• Explicit lifetime management
• Used by default in this template

#define USE_PERSISTENT_MEMORY

Option B: Static Objects

• Simpler implementation
• Automatic cleanup
• Better for multiple instances

// Comment out or remove USE_PERSISTENT_MEMORY definition
// #define USE_PERSISTENT_MEMORY

All callbacks are documented in src/s_function.cpp. Key functions:

Configure ports and parameters:

ssSetNumSFcnParams(S, num_parameters);
ssSetNumInputPorts(S, num_inputs);
ssSetNumOutputPorts(S, num_outputs);

Initialize your objects:

YourAlgorithm* algo = new YourAlgorithm(param1, param2);
ssSetPWorkValue(S, 0, static_cast<void*>(algo));

Process data each timestep:

YourAlgorithm* algo = static_cast<YourAlgorithm*>(ssGetPWorkValue(S, 0));
InputRealPtrsType inputs = ssGetInputPortRealSignalPtrs(S, 0);
real_T* outputs = ssGetOutputPortRealSignal(S, 0);
// Process inputs and write to outputs

Clean up resources:

YourAlgorithm* algo = static_cast<YourAlgorithm*>(ssGetPWorkValue(S, 0));
delete algo;

Edit CMakeLists.txt to add your libraries:

set(CUSTOM_PACKAGES
    OpenCV
    Eigen3
    YourLibrary
)

set(SRCS
    ${CMAKE_SOURCE_DIR}/src/s_function.cpp
    ${CMAKE_SOURCE_DIR}/src/your_implementation.cpp
)

cd build
cmake ..  # FindMatlab.cmake will auto-detect MATLAB

# Or specify MATLAB location:
# cmake .. -DMATLAB_ROOT=/path/to/matlab

# View build output
# Generated files:
#   - matlab_simulink_s_function.mex* (your S-function)
#   - include_directories.txt (include paths)
#   - link_directories.txt (library search paths)
#   - link_libraries.txt (full library paths with extensions)
#   - source_files.txt (source files)

For Simulink Coder code generation:

1. Open Model Configuration Parameters in Simulink
2. Navigate to Code Generation → Custom Code
3. Add the contents of the generated files to the appropriate fields:
   • Include directories: Contents of include_directories.txt and link_directories.txt
   • Libraries: Contents of link_libraries.txt
   • Source files: Contents of source_files.txt
4. Generate code as usual

Note: The CMake build automatically generates these files with:

• Full library paths with extensions (.so, .a, .lib, .dylib)
• Platform-specific library directories extracted from package configurations
• All necessary include paths for headers

The included example demonstrates real-time optical flow tracking for UAV applications:

Features:

• OpenCV integration for computer vision algorithms
• Lucas-Kanade pyramidal optical flow tracking
• Real-world velocity estimation from pixel motion
• State management across simulation timesteps
• Dual memory management support

Inputs:

• Port 0: Grayscale or RGB image (normalized 0-1)
• Port 1: Time delta between frames (seconds)

Outputs:

• Port 0: Velocity estimates (2×1000 matrix with vx, vy pairs)
• Port 1: Computation time (seconds)
• Port 2: Number of tracked features

Algorithm:

1. Extract Shi-Tomasi corner features
2. Track features using Lucas-Kanade optical flow
3. Convert pixel velocities to real-world velocities using camera parameters
4. Output velocity estimates for each tracked feature

The CMake build process automatically generates 4 text files for Simulink Coder integration:

Contains full paths to include directories for header files.

Example content:

/usr/include/opencv4
/home/user/project/include

Simulink field: Code Generation → Custom Code → Include directories

Contains library search paths where linker should look for libraries.

Example content:

/usr/lib/x86_64-linux-gnu
/usr/local/lib

Simulink field: Code Generation → Custom Code → Library search path

Contains full library file paths with platform-specific extensions.

Example content (Linux):

/usr/lib/x86_64-linux-gnu/libopencv_core.so
/usr/lib/x86_64-linux-gnu/libopencv_imgproc.so
/usr/lib/x86_64-linux-gnu/libopencv_video.so

Example content (Windows):

C:/opencv/lib/opencv_core454.lib
C:/opencv/lib/opencv_imgproc454.lib

Simulink field: Code Generation → Custom Code → Libraries

Contains paths to all source files that need to be compiled.

Example content:

/home/user/project/src/s_function.cpp
/home/user/project/src/optical_flow_velocity.cpp

Simulink field: Code Generation → Custom Code → Source files

This template supports two memory management strategies:

• Comprehensive documentation: All functions documented with Doxygen
• Memory safety: No memory leaks, proper RAII where applicable
• Error handling: Try-catch blocks for OpenCV exceptions
• Type safety: Explicit type conversions, const-correctness
• Cross-platform: Tested on Windows, Linux, and macOS

Problem: MATLAB_BIN_DIR not set

Solution: Specify MATLAB path explicitly:
  cmake .. -DMATLAB_BIN_DIR=/path/to/matlab/bin

Problem: Could not find OpenCV

Solution: Install OpenCV or specify OpenCV_DIR:
  cmake .. -DOpenCV_DIR=/path/to/opencv/build

Problem: Compiler not compatible with MATLAB

Solution: Check MATLAB documentation for supported compilers:
  mex -setup C++

Problem: Optical flow tracker not initialized

Solution: Ensure mdlStart() completes successfully. Check MATLAB console for errors.

Problem: Incorrect output dimensions

Solution: Verify output port dimensions match your algorithm's output size.

Problem: Memory leaks in simulation

Solution: Ensure mdlTerminate() properly deletes all allocated objects.

Problem: S-function not found in Simulink

Solution:
  1. Copy MEX file to model directory
  2. Add build directory to MATLAB path: addpath('build')

Problem: Parameter mismatch error

Solution: Verify number of S-function parameters matches ssSetNumSFcnParams() call.

1. Always initialize all member variables in constructors
2. Use const-correctness for parameters that shouldn't be modified
3. Handle exceptions from third-party libraries (e.g., OpenCV)
4. Document all parameters with units and ranges
5. Zero-initialize output buffers to avoid undefined behavior
6. Test with multiple instances if using static memory mode
7. Validate input dimensions before processing

• Use -O3 optimization in MEX compilation for production code
• Pre-allocate buffers in mdlStart() to avoid repeated allocations
• Profile using MATLAB Profiler to identify bottlenecks
• Consider using MEX function caching for frequently called operations
• Use appropriate data types (prefer float over double when precision allows)

If you find this repository helpful, please ⭐ star it!

• MATLAB S-Function Documentation
• Writing C MEX S-Functions
• Simulink Coder User's Guide

• Custom Code Integration
• S-Function Code Generation

• MATLAB Central File Exchange
• Simulink Answers

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Follow the existing code style and documentation standards
4. Test your changes thoroughly
5. Submit a pull request with a clear description

See the LICENSE file for details.