Making the most of your MATLAB Models to Improve Verification

Conference: Verification Futures 2016 (click here to see full programme)
Speaker: Graham Reith, Industry Manager: Communication, Electornics & Semiconductors
Organisation: Mathworks
Presentation Title: Making the most of your MATLAB Models to Improve Verification
Abstract: MATLAB is a popular environment for modelling algorithms before implementation in silicon. In this talk we will discuss how these models can be reused efficiently in the downstream verification activities, eg through integration with UVM verification environments, and also look at the opportunities to left-shift verification by refining the MATLAB model to include more implementation detail and improving the level of testing applied to the algorithm at the design stage. The objective is to reduce the number of issues found during verification, and thereby reducing the amount of rework, and the overall time to market.

In this session we describe:

  • Converting MATLAB functions to SystemVerilog modules using DPI-C
  • Improving the quality of models to reflect real-world implementation architectures and fixed-point effects, identifying design issues earlier
  • Improving the testbenches applied at the algorithm development stage, with coverage analysis and test vector generation
  • Interfacing MATLAB directly with RTL simulators to reuse algorithmic testbenches
Speaker Bio: Graham Reith joined MathWorks in 2002, and currently manages the Communications, Electronics and Semiconductors business in EMEA. His technical background includes both hardware and software development workflows, and he has worked on wireless communications, mixed-signal design, audio, image processing and computer vision applications.

In addition to representing the state-of-the-art of how MathWorks tools can be applied in EDA workflows today, Graham also seeks to identify areas where MathWorks should invest further to improve the tools for the future, working with the product development teams to define requirements for future releases

Before joining MathWorks, Graham worked in applied research, designing and implementing high-performance signal processing algorithms on DSPs and FPGAs for communications applications. Graham has a Master’s degree in Electronic Engineering from the University of York, UK.