VUnit - The Best Value for Initial Effort - Part 3#
This article was originally posted on LinkedIn where you may find some comments on its contents.
After spending one minute on installing VUnit and one minute on creating a run script for incremental compilation it is time to go full automation. Five lines of code or roughly 30 seconds of work for each testbench is what it takes to get the following added values:
A single command to verify your project designs or, which I will show in the next blog, part of the designs.
Support for distributing the simulations over many CPU cores. If you have a quad core CPU you can have a 4x speed-up. This requires more simulator licenses but if you use a free version of a commercial simulator or the open source GHDL that is no problem.
Continuous integration with the help of tools like Jenkins and Travis. Did you notice the build status badge in the example project repository? Whenever someone pushes new code to the repository Travis will use VUnit to run all the tests for that code and present the result. Click on the badge or here and you’ll see.
This video clip will show the details for this step and the resulting code is in my repository.
All in all I’ve spent less than ten minutes to convert this project but I haven’t really made any changes to how the testbenches are designed. If you look at the code you will find several thousands of assert statements which can be enhanced with VUnit functionality. However, replacing all of them is not a quick fix so VUnit must be able to handle legacy asserts in order to be the low-effort solution we’re striving for. A better approach is to continuously improve on the code you have, for example when developing tests for new functionality, making manual testbenches self-checking, or when debugging designs.
This was all for now. In the next blog I will talk about test cases in VUnit and how they bring clarity and extra speed to your testbenches. As always, comments and questions are welcomed here or in our user community chat room.