DeepChip Best of 2020: Xcelium ML

Recently, I wrote about #2a on Cooley's Best of 2020 list, which was Cadence's vManager.  See my post DeepChip Best of 2020: vManager. Number #2b on John's list is Xcelium ML. As I said in the earlier post, Paul Cunningham likes to talk about the importance of the combination of the best engines together with the best logistics in verification. For example, see my post about his recent DVCon keynote Paul Cunningham's DVCon Keynote: Verification Throughput = Engines × Logistics. So #2a being vManager is logistics, and #2b being Xcelium ML is one of the engines (the others being JasperGold for formal verification, Palladium for emulation, and Protium for FPGA prototyping). Combine them together to really increase throughput.

I wrote about Xcelium ML when it was announced. See my posts Xcelium ML: Black-Belt Verification Engineer in a Tool and Under the Hood of Xcelium ML.

John has a long piece on his  DeepChip website CDNS Xcelium ML gets 3x faster regressions is Best of 2020 #2b. As usual, there are many quotes from users of the tool. They are always anonymous, so I have no idea who they are, but here are a couple:

Xcelium ML helped us generate a 3X smaller regression set while retaining 99+% coverage.

Xcelium ML improved our regression runtimes by 2.5X vs. Xcelium.

Xcelium ML was only announced in August, so there hasn't really been enough time for a lot of design groups to have truly used the tool throughout the verification of a full project. Despite that, it was voted #2b on the best of 2020 list. Here is one user's experience:

   Cadence Xcelium has been our primary simulator for years and continues 
    to deliver great performance and efficiency.

    We've now begun working with Xcelium ML--which uses machine learning.  
    So far, we've gotten strong results from running Xcelium ML on three
    SoC System Component IPs.  

    Xcelium ML:
 
       - Improves our regression runtimes by 2.5X vs. Xcelium.

       - Requires fewer licenses for the same coverage than Xcelium
         requires.

       - Has consistently high coverage of 99+%.

You can read all about Xcelium ML in my two earlier posts linked to above. But here it is described by a user:

       - Xcelium ML's machine learning functionality learns from our 
         original Xcelium regressions, and generates an equivalent
         machine learning model.

       - Xcelium ML then uses this machine learning model to generate 
         new, smaller regression test suites. The new test suites 
         reduce the overall regression runtimes while still delivering
         coverage results comparable to our original Xcelium tests. 
 
       - In addition, Xcelium ML generates analytic reports showing 
         random control knobs and coverage bins.  

You should read John's piece linked to above. Most of it contains information from the two users quoted above, who go into a lot of detail about how they use Xcelium ML and their experience and results. The sections are subtitled:

• How it works
• How we use it
• Soak and smoke testing, simulation ranking, directed tests
• Constrained random issues
• How Xcelium ML works
• Target design for the evaluation
• Eval: Xcelium ML vs constrained random
• Eval: Xcelium ML vs regression ranking
• Cadence's umbrella switches
• Xcelium ML: 3X faster using fewer resources
• Cadence Xcelium

I talked to John. First, he pointed out that the graph I showed above is original content based on data users sent him. His view on Xcelium ML?

In my opinion, this is not so much machine learning magic, but a really good picker that chooses sensible regression vectors. It is intelligent in the sense of it not being just monkey random stuff.

A final quote from someone who can't wait to try it:

We'd love to try Cadence new machine learning enabled Xcelium ML. We've seen quite a bit of improvement in back-end tools with ML, so it will be very interesting to see how it works on front-end.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.