Main Content

Use Dynamic Memory Allocation for Variable-Size Arrays in a MATLAB Function Block

This example shows how to use dynamic memory allocation for variable-size arrays in a MATLAB Function block. Dynamic memory allocation allocates memory on the heap as needed at run time, instead of allocating memory statically on the stack. Dynamic memory allocation is beneficial when:

  • You do not know the upper bound of an array.

  • You do not want to allocate memory on the stack for large arrays.

For more information about how Simulink® handles dynamic arrays, see Unbounded Variable-Size Signals.

You can use dynamic memory allocation only for arrays that are local to the MATLAB Function block.

You cannot use dynamic memory allocation for:

  • Parameters. Parameters must be fixed-size.

  • Discrete state properties of System objects associated with a MATLAB System block.

Configure Model for Dynamic Memory Allocation

Make sure that you configure the model to use dynamic memory allocation for variable-size arrays in MATLAB Function blocks. In the Configuration Parameters window, in the Simulation Target > Advanced parameters category, make sure that:

  • Dynamic memory allocation in MATLAB functions is selected.

  • The Dynamic memory allocation threshold in MATLAB functions parameter has the default value 65536.

Create Model

Create this Simulink model that has a MATLAB Function block with an unbounded variable-size array.

This image shows the model you created. A MATLAB Function block uses a constant block as an input and an Outport as an output.

  1. Create a Simulink model mymodel.

  2. Add a MATLAB Function block to the model.

  3. In the MATLAB Function block, add this code:

    function s = myfcn(n)
Z = rand(1,n);
s = sum(Z);
end

  4. Add a Constant block to the left of the MATLAB Function block.

  5. Add an Outport block to the right of the MATLAB Function block.

  6. Connect the blocks.

Simulate Model Using Dynamic Memory Allocation

  1. Simulate the model.

  2. In the MATLAB Function Editor, to open the MATLAB® function report, click Function Report.

    The variables tab shows that Z is a 1-by-:? array. The colon (:) indicates that the second dimension is variable size. The question mark (?) indicates that the second dimension is unbounded.

Simulation must use dynamic memory allocation for Z because the second dimension of Z does not have an upper bound.

Use Dynamic Memory Allocation for Bounded Arrays

When an array is unbounded, the code generator must use dynamic memory allocation. If an array is bounded, the code generator uses dynamic memory allocation only if the array size, in bytes, is greater than or equal to the dynamic memory allocation threshold. The default value for this threshold is 65536.

Dynamic memory has a run-time performance cost. By controlling its use, you can improve execution speed.

If you make Z a bounded variable-size array with a size that is greater than the threshold, the code generator uses dynamic memory allocation for Z. For example:

  1. In mymodel, modify myfcn so that Z has an upper bound of 500.

    function s = myfcn(n)
assert(n < 500);
Z = rand(1,n);
s = sum(Z);
end

  2. Simulate the model.

    In the MATLAB function report, you see that Z is a 1-by-:500 array. It is variable size with an upper bound of 500.

  3. Lower the dynamic memory allocation to a value less than or equal to 4000, which is the size, in bytes, of Z. In the Configuration Parameters window, in the Simulation Target > Advanced parameters category, set the Dynamic memory allocation threshold in MATLAB functions parameter to 4000.

  4. Simulate the model.

    The code generator uses dynamic memory allocation because the size of Z is equal to the dynamic memory allocation threshold, 4000.

Generate C Code That Uses Dynamic Memory Allocation

If you have Simulink Coder™, you can generate C code for this model. Then, you can see how the code generator represents dynamically allocated arrays.

  1. Configure the model to use a fixed-step solver. In the Configuration Parameters dialog box, in the Solver pane, under Solver selection:

    • For Type, select Fixed-step.

    • For Solver, select discrete (no continuous states).

  2. Configure the model to create and use a code generation report. In the Configuration Parameters dialog box, in the Code Generation > Report pane, select Create code generation report and Open report automatically.

  3. Edit the code in the MATLAB Function block so that it looks like this code:

    function s = myfcn(n)
Z = rand(1,n);
s = sum(Z);
end

    Z is an unbounded variable-size array.

  4. Build the model.

  5. In the code generation report, open mymodel.c. You can tell that the code generator used dynamic memory allocation for Z because you see the emxArray type emxArray_real_T_mymodel_T and emxArray utility functions, such as mymodel_emxInit_real_T. The code generator uses an emxArray type for variables whose memory is dynamically allocated. The generated code uses the emxArray utility functions to manage the emxArrays.

If you have Embedded Coder®, you can customize the identifiers for emxArray types and the utility functions. See Identifier Format Control (Embedded Coder).

See Also

Related Topics