Using Simulink

The Simulink Accelerator

The Simulink Accelerator speeds up the execution of Simulink models. The Accelerator uses portions of Real-Time Workshop, a MathWorks product that automatically generates C code from Simulink models, and your C compiler to create an executable. Note that although the Simulink Accelerator takes advantage of the Real-Time Workshop technology, Real-Time Workshop is not required to run it. Also, if you do not have a C compiler installed on your Windows PC, you can use the lcc compiler provided by The MathWorks.

Note    You must have the Simulink Accelerator product installed on your system to use the accelerator. The Accelerator uses the Real-Time Workshop technology to generate the code used to accelerate the model. However, the generated code is suitable only for acceleration of the model. If you want to generate code for other purposes, you must use Real-Time Workshop.

Accelerator Limitations

The Simulink Accelerator has the following limitations.

• The Accelerator does not support models with algebraic loops. If the accelerator detects an algebraic loop in your model, it halts the simulation and displays an error message.
• The Accelerator does not support models that pass array parameters to M-File S-Functions that are not numeric, logical, or character arrays, are sparse arrays, or that have more than two dimensions.

How the Accelerator Works

The Simulink Accelerator works by creating and compiling C code that takes the place of the interpretive code that Simulink uses when in Normal mode (that is, when Simulink is not in Accelerator mode). The Accelerator generates the C code from your Simulink model and invokes the MATLAB mex function to compile and dynamically link the generated code to Simulink.

Note    The MATLAB mex function uses the lcc compiler by default to compile the code generated by the Accelerator. An optimizing compiler, such as Microsoft Visual C/C++ 7.1, can produce faster compiled code and hence accelerate simulation still further. If you have such a compiler, you can configure the mex command to use it. See the online help for mex for more information.

The Simulink Accelerator removes much of the computational overhead required by Simulink models when in Normal mode. It works by replacing blocks that are designed to handle any possible configuration in Simulink with compiled versions customized to your particular model's configuration. Through this method, the Accelerator is able to achieve substantial improvements in performance for larger Simulink models. The performance gains are tied to the size and complexity of your model. In general, as size and complexity grow, so do gains in performance. Typically, you can expect a 2X-to-6X gain in performance for models that use built-in Simulink blocks.

Note    Blocks such as the Quantizer block might exhibit slight differences in output on some systems because of slight differences in the numerical precision of the interpreted and compiled versions of the model.

Simulink Accelerator

Running the Simulink Accelerator

© 1994-2005 The MathWorks, Inc.