Integrate C Code Using C Caller Blocks

You can integrate new or existing C code into Simulink^® using the C Caller block. To create custom blocks in your Simulink models, the C Caller block allows you to call external C functions specified in external source code and libraries. The advantages of the C Caller block are:

Automated integration of simple C functions
Integration with Simulink Coverage™, Simulink Test™, and Simulink Design Verifier™
Integration with Simulink Coder™

The C Caller block and the C Function block allow you to bring C algorithms into Simulink. To model dynamic systems, use the S-Function Builder instead. Next steps describe the workflow to integrate C code into Simulink using the C Caller block.

Note

C99 is the standard version of C language supported for custom C code integration into Simulink.

Specify Source Code and Dependencies

Specify your external source code file that contains your C functions.

From Simulink toolstrip, open the Configuration Parameters.
In the left pane, select Simulation Target.
To enable code parsing by the C Caller block, ensure that the Import custom code box is selected.
The directories and file paths can be absolute and relative file paths to model directories or to the current working directory. See Specify Relative Paths to Your Custom Code (Stateflow).
Select Header file and enter the name of your header file with the #include tag.

Tip
After you have entered information for Source files in the next step, you can click Auto-fill from Source files to have the header file name filled in automatically, using information contained in your source files.
Under Additional build information, select Source files and enter the path and the name of the source file. If the model and the source files are in different directories, enter the directory that contains the source file before the file name.
Select Include directories, and enter the folders where additional build information, such as header files, are stored.

To verify that your custom code can be parsed and built successfully, click Validate.

Note

If your header file declares a function but your source files do not define the function, by default the function is not visible in the C Caller block dialog. You can set the Undefined function handling parameter in Configuration Parameters to specify other behaviors in this situation, including throwing an error, generating a stub function, or ignoring the condition.

Note

To use a C Caller block in a For Each subsystem or with continuous sample time, or to optimize the use of the block in conditional input branch execution, the custom code function called by the block must be deterministic, that is, always producing the same outputs for the same inputs. Identify which custom code functions are deterministic by using the Deterministic functions and Specify by function parameters in the Simulation target pane. For a conditional input branch execution example, see Use C Caller Block with Conditional Execution.

N-D Array Handling

Simulink can pass N-D array data to custom code functions in C Caller blocks, and receive data from such blocks. When you do so, you must specify the correct array layout to achieve the intended results. See Default function array layout and Exception by function. For examples of the use of array data with C Caller blocks, see Bring Custom Image Filter Algorithms as Reusable Blocks in Simulink and Call a Legacy Lookup Table Function Using C Caller block.

You can specify the order of how matrix data is handled in your C functions. Matrix data passed to and from your C functions is converted if necessary to the array layout you specify. If the array layout is not specified, the matrix data is passed through the C function in the same order of your Simulink data, and computational errors may occur due to row-column major disagreement. Ensure that you follow the same default function array layout for all Simulink data.

Column-Major — The C function handles input array data in column-major order. Suppose that you have a 3-by-3 matrix. In the C function, this matrix is accessed in this sequence: first column, second column, and third column.
Row-Major — The C function handles input array data in row-major order. Suppose that you have a 3-by-3 matrix. In the C function, this matrix is accessed in this sequence: first row, second row, and third row.
Any — The C function is indifferent to the layout of input array data. This would be the case if, for example, the function performs only element-wise operations on the data.
Not specified — The C function makes no assumption about the layout of input array data. Compared to the Any setting, you can generate code only in column-major setting. Attempting to generate code in row-major setting produces an error. See Array layout (Simulink Coder). Select this option only if needed for compatibility with older models.

To learn more about the row-major and column-major array layouts in Simulink, see Default function array layout.

Select an array layout option under Default function array layout.
If you need to apply a specific array layout to some of the functions in your code, click Exception by function to select these functions.
Click Apply to accept your changes.

If your C functions accept only scalar and/or vector inputs, the Default function array layout setting has no effect.

Call C Caller Block and Specify Ports

You can start your custom C code integration into Simulink by typing C Caller in the Simulink canvas. Alternatively, drag a C Caller block from the User-Defined Functions library onto the canvas. Double-click the block to open the Block Parameters dialog box to see the names of your functions and port specifications.

Click on the Refresh button to import your source code and its dependencies.
Your C functions are displayed under Function Name. If you do not see your full list of functions, click on the to reimport your source code.
To view function definitions in the source file, click the . The source code for the selected function is displayed in the MATLAB^® Editor. If the source code is not available, the function declaration in the header file is displayed.
To change source files and their dependencies, or to define and select function array layouts, click the custom code settings button to open the Simulation Target pane in Model Configuration Parameters.

Map C Function Arguments to Simulink Ports

You can map C function arguments from your source code to Simulink ports using the Port specification table in the C Caller block and by creating a FunctionPortSpecification object through the command line. In your source code, the header file includes the C function arguments to be connected to Simulink ports.

extern void mean_filter(const unsigned char* src,
                           unsigned char* dst,
                           unsigned int width, unsigned int height,
                           unsigned int filterSize);

Port specification shows the details of your arguments and how they connect to your C Caller block in Simulink.

Name — Specifies the name of input and output arguments. Name is the function argument or parameter name as defined in your C functions from source code. This column is for reference purposes only.

Scope — Specifies how C function arguments map to the Simulink scope. Your arguments have default scopes depending on the function definition and you can change the scopes depending your function definition in the source code.

Simulink Scope	Scope to Block Mapping
`Input`	Block input port
`Output`	Block output port
`InputOutput`	Block input and output port
`Global`	Global variable used by the block
`Parameter`	Block tunable parameter
`Constant`	Constant value

For an argument passed by pointer, when you have a constant qualifier definition such as const double *u, the argument can only be an input or a parameter. When there is no constant qualifier, the argument is an InputOutput by default, and you can change it to an Input, Output, or Parameter scope. In the case of an Input or Parameter scope, ensure that the C function does not modify the memory pointed by the pointer. If the argument is of an Output scope, every element pointed to by this pointer should be reassigned in every call for the function.

C Argument	Simulink Scope
Function return	`Output`
`double u`	`Input`, `Parameter`, `Constant`
`double *u` `double u[]` `double u[][2]` `double u[2][3]`	`InputOutput` (default), `Output`, `Input`, `Parameter`
`const double *u` `const double u[]` `const double u[][2]` `const double u[2][3]`	`Input` (default), `Parameter`

Use the InputOutput port to map an input passed by a pointer in your C functions. Ports created using an InputOutput port have the same name for input and output ports. InputOutput ports enables reuse of buffer for input and output ports. This may optimize the memory use depending on the signal size and the block layout.

To map C function arguments to an InputOutput port, define the variable as a pointer in your function definitions.

extern void mean_filter(unsigned char* src,
                           unsigned int width, unsigned int height,
                           unsigned int filterSize);

Then, select the port specification to the InputOutput scope in the Port Specification table, and assign the resulting function output to the input variable in the custom function.

You can use global variables in your custom code, mapping them to the appropriate Simulink scope. To enable the use of global variables in your model, select Enable global variables as function interfaces from Model Settings > Configuration Parameters > Simulation Target. You can map the global variables to an Input, Output, InputOutput or Global scope on the C Caller block. The availability of the these scopes depend on the use of the global variable in your custom code.

A Global scope enables you to transfer data between custom code and the C Caller block and lets you use the global variable during calculations on the block. Values transferred using Global scope are not visible on the block interface. This table shows example code snippets and their default and available ports.

Example Code Simulink Scope

Example Code	Simulink Scope
double data; void foo(void) { int temp = data; }	Global variable data only reads the variable `data`. Available scopes are: `Input` (default) `Global`
double data; void bar(void) { data = 0; }	Data is written to a global variable. Available scopes are: `Output` (default) `Global` `InputOutput`
double data; void foo2(void) { data = data + 1; }	Data is both read and written on a global variable. Available scopes are: `Global` (default) `InputOutput` `Output`

double data;

void foo(void)
    {
        int temp = data;  
    }

Global variable data only reads the variable data. Available scopes are:

Input (default)

Global

double data;

void bar(void)
    {
        data = 0;
    }

Data is written to a global variable. Available scopes are:

Output (default)

Global

InputOutput

double data;

void foo2(void)
    {
        data = data + 1;
    }

Data is both read and written on a global variable. Available scopes are:

Global (default)

InputOutput

Output

Label — Indicates the label for the corresponding argument in a Simulink block. By default, your argument label is the same as the argument name, unless you change it.

Simulink Scope	Simulink Port Label
`input`, `output`	Port name
`inputoutput`	Port name in both input and output port
`Global`	Port name and global variable name
`parameter`	Parameter name
`constant`	Expression for the constant value. size expressions using input argument names, for example `size(in1,1)`

Type — Demonstrates the match between the Simulink data type and the C function argument data type.

C Argument Data Type	Simulink Data Type
signed char	int8
unsigned char	uint8
char	int8 or uint8, depending on the compiler
int^*	int32
unsigned int^*	uint32
short ^*	int16
long ^*	int32 or fixdt(1,64,0), depending on the operating system
float	single
double	double
int8_t^*	int8
uint8_t^*	uint8
int16_t^*	int16
uint16_t^*	uint16
int32_t^*	int32
uint32_t^*	uint32
typedef struct {…} AStruct^**	Bus: AStruct
typedef enum {..} AnEnum^**	Enum: AnEnum
* If the C Caller takes an integer type, for example, int16_t, you can modify it to a fixed-point type with matching base type, for example to fixdt(1, 16, 3). ** The C Caller sync button prompts you to import struct or enum types used by a C function as Simulink bus and enumeration types.

Size — Specifies the data dimensions in the argument.

C Argument Dimensions	Simulink Port Dimensions
`double u`	scalar (1)
`double u[]` `double u[][2]`	inherited (-1) (default) If the argument is for an output port, the size should be specified. The size of an output port cannot be inherited.
`double *u`	inherited (-1) (default) If the argument is for an `inputoutput` port, the size cannot be inherited even though the size in the `output` port can be inherited. For global variables, size is scalar (1).
`double u[2][3]`	Size is [2, 3].

Create a `FunctionPortSpecification` Object and Edit C Caller Block Properties

To change Port Specification table properties programmatically, you can create a FunctionPortSpecification object and modify its properties. To create a FunctionPortSpecification object for a selected C Caller block in a model, type in the command line:

myCCallerConfigObj = get_param(gcb, 'FunctionPortSpecification')

myCCallerConfigObj = 

  FunctionPortSpecification with properties:

        CPrototype:  'real_T add(real_T u1, real_T u2);'
    InputArguments:  [1×2 Simulink.CustomCode.FunctionArgument]
    ReturnArgument:  [1×1 Simulink.CustomCode.FunctionArgument] 
    GlobalArguments: [1×0 Simulink.CustomCode.FunctionArgument]

The CPrototype property is read-only, and shows the declaration of C function input variables. The InputArgument and ReturnArgument properties create a FunctionArgument object that you can further edit its properties according to the rules defined for Port Specification table above. You can See FunctionPortSpecification to learn more.

To modify the global arguments in a C Caller block, create a handle to the GlobalArguments object using getGlobalArg and modify its properties.

Create a Custom C Caller Library

It is recommended to create a library model to group your C Caller blocks and keep your models organized. You can also link a data dictionary to the library to hold custom types defined in your code. Using a library model is especially useful when you have multiple models or a model reference hierarchy that uses custom C code.

Open a new library model. On the Simulation tab, select New > Library.
On the Modeling tab, under Design, click Simulation Custom Code.
Select C or C++ in the Language option, depending on your code, and ensure the Import custom code box is selected.
Follow the instructions in Specify Source Code and Dependencies to add your source files and their dependencies.
Create C Caller blocks to call C functions.
To insert a block from your library model to a Simulink model, simply drag the block into your model.

You can also create a library of C Caller blocks from your custom code using the Simulink Code Importer. See Import Custom C/C++ Code Using the Simulink Code Importer.

Generate Debug Symbols for Custom Code

To attach an external debugger to the MATLAB process, and debug external C code, make sure that the Model Configuration Parameter Simulate custom code in a separate process is cleared, then generate the debug symbols using:

clear mex
Simulink.CustomCode.debugSymbols('on')

After you turn this setting on and update your model, your debug symbols are generated and you can attach an external debugger to your MATLAB process by attaching the process to MATLAB.exe. Set breakpoints in your external C code using the external debugger, and when you simulate your model it should stop at the breakpoints.

Turn this setting off using:

Simulink.CustomCode.debugSymbols('off')

Generate Code from Model

The C Caller supports code generation. In the code generated from your model, each execution of a C Caller block corresponds to a call to the external C function associated with the block. In order to build the generated code, the Code Generation > Custom Code pane of the Model Configuration Parameters must be populated with the correct information regarding the custom code. See Model Configuration Parameters: Code Generation Custom Code (Simulink Coder).

Simulate Custom Code in a Separate Process

Open Model

When simulating a model containing custom C or C++ code, you have the option to run the custom code in a separate process outside of MATLAB. This option may be useful when debugging your custom code. By running in a separate process, problems with the custom code do not cause MATLAB to crash, and you can more easily debug and resolve such problems. Problems could arise due to unexpected exceptions in the custom code or errors in the interface between Simulink and the custom code.

To enable this option, in the model configuration parameters, in the Simulation Target pane, select Simulate custom code in a separate process. The option applies to custom C/C++ code integrated into your model using any of these blocks:

C Caller
C Function
MATLAB Function
MATLAB System
Stateflow® chart

For example, this model contains a C Caller block that calls the function adder(), which accesses an object called adderObj. Before calling the function, the object must be created, which can be done by calling initAdder() from the Initialize function of the Simulation Target pane of the model configuration parameters.

If initAdder() is not called before adder(), then adder() attempts to access an uninitialized pointer, which causes a run-time exception. If the Simulate custom code in a separate process parameter is not selected, this exception could cause MATLAB to crash when you simulate the model. However, if the parameter is selected, simulating the model produces an error message within Simulink.

You can then click Open to launch your external debugger and resolve the issue that caused the error.

After the debugger launches, it will restart the simulation and stop at the breakpoints of custom function entries automatically.

Once your custom code is fully debugged, you can clear Simulate custom code in a separate process for faster simulation time.

Limitations

Global Variables — Global variables as function input outputs do not support multidimensional arrays.
Initialization/Termination of Custom Code Settings — If you need to allocate and deallocate memory for your custom code, insert allocate and deallocate in the Initialize function and Terminate function fields of custom code settings, or use a C Function block.
Complex Data Support — The C Caller block does not support complex data types in Simulink.
Variable Arguments — Variable arguments in C are not supported, for example, int sprintf(char *str, const char *format, ...).
C++ Syntax — The C Caller block does not support native C++ syntax directly. You need to write a C function wrapper to interface with C++ code.

To test models that include C Caller blocks, see Test Integrated C Code (Simulink Test).

Note

If a model has custom code, after the model is updated or run, the slprj folder may be locked due to the loaded custom code simulation executable file. The folder cannot be deleted when it is locked. To unload the executable file and unlock the slprj folder, use the clear mex command. See clear.