March 2-5, 2020

DoubleTree Hotel, San Jose, CA

Industry Leaders Panel: Did We Create the Verification Gap? - Part 5

Industry Leaders Panel: Did We Create the Verification Gap? - Part 5


Download the MP3


Part 5
(Don Smith): (Don Smith) with (Cypress).  Jim's last comment was a good lead into my question.  I wanted to ask Jim and Bill when the ROI starts to accentuate the verification gap, how have you challenged your teams, or what effective solutions have you found to close that ROI gap?
Bill Grundmann: We have many challenges.  First off, my definition of verification is more than the logic domain, which was stolen by the (Synthesis) crowd back in the 1980s.  Verification, to me, is electrical, the timing, all that stuff is part of the verification.  Logic is just one component.  Actually, in some developments where it's analog, it's the small component.  So the challenge is, is how to modernize.  As mentioned earlier, you know, you have some people who've done RTL.  You know, to ask them to write some C-code and then go through a high level synthesis is just asking too much.  They weren't trained.  That's not the way their mindset is.  But you have to keep moving.
The onset of bringing in gate libraries – okay.  A full custom designer, for those who came from the full custom crowd where every transistor was sacred, I can size every individual transistor.  It was foreign to go to a library base.  In fact, there was quite a few custom designers that said, "Over my dead body am I going to do this stuff."
Male #2: And they're dead now.
Bill Grundmann: So it's constantly training, it's constantly elevating people's experience and modernizing how they do things.  And it's a constant retraining of people.  And that's really the only way to handle it overall.  You can bring in tools, but if people don't understand why the tool is there, it's not going to go anywhere.  You just had an expense.
Jim Caravella: Yes.  I'd just build on that.  I think the teams need to understand the why, first of all.
Second thing, I would also agree with Bill on is it's not just going to happen.  All right?  What it really takes – it may sound real simple, but what it really takes is a lot of adult supervision.  You know?  You've got to get in there, you've got to work with the teams hands-on.  The team has got to see the value in it.  Right?  Just even sending them to training isn't going to really do anything.  You going to send them to training, get them the training.  But then really work with the teams in the trenches to where they – you know, eventually then a light bulb goes on and says, wow, if I don't size every one of my inverters, I actually get a lot more output which, by the way, then I get a bigger check at the end of the day.  Right?  So it's not something you just give them a handbook or send them to training and they come back and life is good.  It's a culture change.  It's tough.  It really takes engineering leadership in the teams hands-on, at least from my experience.
Bill Grundmann: Even the concept of IP, there's (sounds like:  design) who said, "I can design that better."  And that's probably true.  But the project would be dead because the project schedule would slip out so much that they had to design it.  So it's the idea of, you know, reusability, the libraries, all these things kind of add over time.
Jim Caravella: Yes.  It takes the team seeing it, oh, wow, okay, this actually does work.
Bill Grundmann: There's an ROI on that.
Jim Caravella: Yes.
Bill Grundmann: It's the ROI they have to – you have to educate them on the ROI.
(Don Smith): And how much efficiency would you say you're able to gain by, you know, putting those techniques in place.
Bill Grundmann: I can't give you any metrics – numerical metrics compared to what we were doing with the individual transistors.  Or we used to have a metric that how many transistors you can design per week.  Okay?
(Don Smith): And then it went to day.
Bill Grundmann: Today, you know, it's how many gates.  Now it's how many IPs can you place in the day.  You know, you could build a whole chip in a day if you have a good wizard and construction mechanisms.  That was unheard of some time ago, not too long ago.
(Don Smith): All right, thank you.