Are You Ready For a Change?
UVM Drivers and Monitors

Getting the Least Out of Your Verification Effort

It's quite common to read blog posts or industry newsletters giving you tips on how to succeed with your verification efforts. But all of these people are probably trying to sell you something. Don't let them do it! Follow these easy steps and you'll be amazed at how much less stressful your life becomes:

#1 - Wait until you're finished with your testbench before starting random regressions.

We all know that EDA and server vendors are always trying to steal your money by suggesting that you run more random regressions earlier in your verification cycle. Stick it to the man by waiting until your testbench is completely finished before wasting extra CPU cycles on those worthless random regressions. Remember - always hardcode your simulation seed in your run scripts to 1, and bonus points if you simply leave out random constraints so that anyone who foolishly tries to run a regression still sees all passing results. Ha!

#2 - Don't waste time writing funcional coverage.

Line, conditional, and toggle coverage were good enough for grandpa, so they're good enough for me! Because we all know that functional coverage was added just so simulator vendors could sell licenses for fancy-pants tools claiming to help you map your functional coverage back to a specification. Jokes on them - we don't need no stinkin' specs either!

#3 - Real engineers don't need block diagrams. 

Every so often, you'll hear some highfalutin consultant or new team member ask where they can find a copy of the latest top-level block diagram. They'll try to say they need this to create a testplan. But in reality, all this shows is that they aren't really team players. And they probably just want you to buy a copy of Visio or OmniGraffle. Losers... 

#4 - Outsource all local decision making to your CAD team.

Some people think they know everything. But how could they if they're not a member of the internal CAD team? CAD engineers go through special training where they purposely never work on real projects. That way their minds aren't polluted with reality - they can see what should be instead of what is! Don't even think about rotating these engineers through product teams. You've spent all this time training them to create special tools for you. If you let them work on a real project they might start having different ideas about what works best. And then all that training goes down the toilet. 

#5 90% of 90%

Mathematically, 0.9 x 0.9 is greater than, say, 1.0 x 0.5. So if you have the opportunity to get 90% of 90% of the tasks done, take it. Because spending time actually completing any one task is just not an effective use of your time. For bonus points, see how many different tasks you can get to 90% completion each week. Remember, you can always go back and finish that extra 10% later. And why not, it will only take 10% of the time as the rest of the task - it can wait!

Remember, just follow these easy steps and you'll be a verification superstar in no time!

-----

Want to discuss ideas for a UVM-based Birds of a Feather at DAC?  Check out Cool Verification on Facebook

Comments