We’re happy to give you a special Christmas (or whatever gift-giving holiday you may celebrate this time of year) present: the December issue of Verification Horizons! In addition to a fun look at how the Harry Potter books and movies can be compared to verification, we have a great article on the verification of Deep Learning designs.
After a great analysis of classical von Neuman architectures vs. Convolutional Neural Networks for implementing Deep Learning designs, my Mentor colleague Jean-Marie Brunet and noted verification expert Lauro Rizzatti explain “Why Hardware Emulation Is Necessary to Verify Deep Learning Designs.” As you’ll see, it comes down to three main requirements to verify these CNN-centric designs:
- Scalability – supporting designs from several million up to 10 billion gates.
- Virtualization – moving the emulator from the lab to the data center making it an enterprise resource.
- Determinism – ensuring the same result from one compile to the next, especially when debug is factored in.
The issue has several other valuable articles as well:
- Deadlock Prevention Made Easy with Formal Verification – explores the intricacies of safety and liveness properties to identify deadlocks in formal verification and shows how to automate deadlock detection.
- Exercising State Machines with Command Sequences – how Questa® inFact can generate command sequences that exhaustively exercise a complex state machine.
- Designing a Portable Stimulus Reuse Strategy – how to apply the new Accellera Portable Test and Stimulus Standard to maximize reusability when sharing test intent and IP between teams and projects.
- Don’t Forget the Protocol! A CDC Protocol Methodology to Avoid Bugs in Silicon – proposes a methodology to better utilize formal model checking and simulation-based technologies to overcome common CDC verification challenges.
Please take a look at the latest issue of Verification Horizons (downloadable PDF here). Feel free to leave a comment and let us know what you think. We also invite you to tell us what topics you might want to see in future issues. Enjoy!