Speed comparison S-Function - Simulink blocks

Hi!

I have simulink Level 2 M File S-Function block and I have also a subsystem with simulink blocks written to do exactly the same thing, they solve the same differential equations. The subsytem runs much faster thant the S-function block.

How should the performance of a S-Function block compare to a fully simulink based model? In this case I am talking about a Level 2 M file S Function.

Thank you for your answers.

Answer 1

[Guy Rouleau]

MATLAB-file s-functions are interpreted by the MATLAB engine. Core blocks are compiled.

As said on wikipedia:

"The main disadvantage of interpreting is a much slower speed of program execution"

http://en.wikipedia.org/wiki/Interpreted_language#Disadvantages_of_interpreted_languages

If you want to write MATLAB code that runs fast, you can use the MATLAB Function Block (previously called Embedded MATLAB Function block). Or you also have the option to write a C-Mex s-function.

I wrote posts about including MATLAB code in a simulation:

http://blogs.mathworks.com/seth/2010/07/18/including-matlab-code-in-a-simulation/

and about speeding up a simulation:

http://blogs.mathworks.com/seth/2010/10/28/tips-for-simulation-performance/

Answer 2

[Thiago Marinho]

Thank you! That was of great help.

Although Level-2 S-functions are slower I just cannot use Matalab function block for my purposes.

Is there a way to compile the m-file s-function block or do I have to write a C-Mex s-function?

How will the C-Mex s-function compare to a model made only out of simulink core blocks?

[Kaustubha Govind]

Thiago: Unfortunately the only way to get performance comparable with core blocks with MATLAB code is to use the MATLAB Function block. The only alternative currently is to write a C-MEX S-function, whose performance should be comparable to that of core blocks.