Performance of SystemObject vs Class

Lately, I was introduced to SystemObjects and some of its benefits, for example that one can use them in Matlab as well as in Simulink which is important to me. So I started playing around with SystemObjects and quickly ended up in a scenario where I have a main SystemObject in which I create instances of other SystemObjects which themselves create instances of other SystemObjects. So to say: nested SystemObjects.

Now, for my understanding: If I want to use the main SystemObject only in both Matlab and Simulink then I could replace the other SystemObjects with normal classes. And here comes the question: Would that increase the performance of the generated code?

The reason why I am wondering is the fact that in languages like C++ inheritance from classes with virtual methods cause runtime costs (at least that's what I understood). I guess in the case of SystemObjects which inherit from the matlab.System class something similar will happen?

Answer 1

[Sanju]

Hi Friedrich,

Replacing nested SystemObjects with normal classes may not necessarily lead to a performance increase in the generated code. While languages like C++ incur runtime costs due to inheritance from classes with virtual methods, MATLAB's SystemObjects, which inherit from "matlab.System class", handle this differently.

In MATLAB, the overhead associated with inheritance is generally minimal, especially in the case of SystemObjects. By replacing the nested SystemObjects with normal classes, you can potentially reduce this overhead and improve the performance of the generated code. Normal classes in MATLAB do not have the same runtime costs associated with virtual methods as in languages like C++. However, it's important to note that the performance improvement will depend on the specific implementation and usage of the SystemObjects.

To determine the actual impact on performance, it is recommended to benchmark and compare the performance of the code with nested SystemObjects and the code with normal classes. This will help you make an informed decision based on your specific requirements and constraints.

To benchmark the performance of your MATLAB code, you can use the built-in profiling tools provided by MATLAB. These tools allow you to measure the execution time of different parts of your code and identify potential bottlenecks.

One commonly used tool is the MATLAB Profiler, which provides a detailed report of the time spent in each function or line of code. To use the profiler, you can follow these steps:

1. Open the MATLAB Profiler by clicking on the "Profile" button in the MATLAB toolbar.
2. In the Profiler window, click on the "Start Profiling" button to start recording the execution time.
3. Run your MATLAB code or execute specific functions that you want to benchmark.
4. Once your code has finished running, click on the "Stop Profiling" button in the Profiler window.
5. The Profiler will display a report showing the time spent in each function or line of code, as well as other useful information such as the number of calls and memory usage.

By analyzing the profiler report, you can identify the parts of your code that are taking the most time and optimize them if necessary.

Another useful tool for benchmarking is the tic and toc functions. You can use these functions to measure the execution time of specific sections of your code. Here's an example,

tic
% Code to be benchmarked
toc

The output of toc will give you the elapsed time in seconds.

Note: Remember to run your code multiple times and take the average execution time to get a more accurate measurement of performance.

You can also refer to the following links for further information,

https://www.mathworks.com/help/matlab/matlab_prog/techniques-for-improving-performance.html

https://www.mathworks.com/matlabcentral/answers/579270-nested-functions-vs-object-oriented-programming-k-d-tree-example-and-relative-performance

https://www.mathworks.com/help/matlab/matlab_prog/profiling-for-improving-performance.html

https://www.mathworks.com/help/matlab/ref/profile.html

https://www.mathworks.com/help/matlab/ref/tic.html

Hope this Helps!

[Friedrich]

Thanks @Sanju for your reply! From what I can see in Matlab, replacing the nested SystemObjects with regular classes reduces the runtime costs in Matlab significantly in my case. If this also has an impact on the generated code is yet to be checked.