( ESNUG 579 Item 1 ) ---------------------------------------------- [12/14/17]
Subject: Prakash and Anirudh spar on Real Intent Linting vs. Jasper Formal
DAC'17 Troublemakers Panel in Austin, TX
Prakash: ... Yeah, sure they [Synopsys Atrenta] have their advantage;
but we at Real Intent have the advantage that we have much
superior functionality to offer.
There's less price competition right now than we had with
Atrenta, so those two factors are contributing to our current
success. We're doing very well.
Anirudh: I do have a comment here.
Cooley: Go ahead.
Anirudh: Linting is part of verification. So, what I believe there are
four main engines in verification:
1. Formal,
2. Simulation,
3. Emulation,
4. and Prototyping.
Linting is an important piece, but to me the right place it
belongs is with Formal because Formal is a structural analysis
of the netlist. Formal already reached that, and has much more
powerful engines that have been traditionally used in Linting.
So, from a Cadence perspective, what we are doing is to enhance
Jasper to do more of this Linting.
In the long run, there must be more Formal techniques applied,
rather than just structural analysis of the netlist. So that's
what we're trying to do in Cadence. That's why we bought Jasper,
that's where the investment is going for this area.
Prakash: We saw your announcement. You are throwing Formal at a frontend
digital problem; which is very different from a backend problem.
I almost don't belong on this panel -- which has a huge amount
of user interaction. The capabilities that you are touting is
parallelism and debugging.
Well, your Cadence Superlint, when we benchmarked against it,
was Supernoisy. [laughter]
So, your strategy is to throw more debug support instead of
deleting the noise. When you say this whole thing belongs
inside Formal, I say it's a complete misrepresentation.
Actually, Formal belongs inside Static. It's all wrong,
it's inverted, and as long as they [Cadence] invert, the
principle is going to be...
I mean your CDC announcement -- it doesn't seem like a CDC tool.
It seems like a funnel to feed in more problems into your Jasper
Formal tool. It's really not trying to solve a CDC problem.
Cooley: I love it when they fight. [laughter]
Prakash: You, Anirudh, can say that everything belongs in Formal, but
I really think Formal belongs inside (Static)
[Joe moves his chair back.]
Cooley: Dean, can you move back out of the way? Fight, fight, ...
Anirudh: Well, I stand by my statement. There is more...
Cooley: He says as he's sitting down.
Prakash: You [to Anirudh] have to have an answer, right?
Anirudh: When I look at this area, there is more fundamental technology
that is there in Formal verification -- it's a much more
powerful engine. I believe over time that is the basis of
doing all these structural changes.
Cooley: But Real Intent is making money in that space.
Anirudh: Sure, yeah...
Prakash: [to Anirudh] Why do you say that the "powerful technology" is
the best way to go? At the end of the day, you have to invent
the simplest technology that will solve the problem.
And that's what Static is. When you said Formal technology
is the way to go, without thinking about what technology is
needed to solve the problem, you don't understand the problem.
We've been doing CDC for 10 years.
Anirudh: [to Prakesh] Sure, you're entitled to your opinion and I don't
want to...
But in the end, you are reading a netlist, you are doing some
analysis. And the two engines I think are critical in this, is
either a timing engine, which is Tempus because you have to check
all these constraints, or a formal engine, which is Jasper, and
this is where we believe we're making the investment.
Now there may be other ways to solve this problem, and that's
fine. But I think over the long run, that's where I'm say
personally, from a Cadence perspective, that's where our bet is.
Prakash: There's one more point. CDC, today our customers want us to
solve 500 million gate CDC problems. And you're saying Formal
is the way to go?
Anirudh: I think what you have remember is -- what I'm trying to say is --
yes, I'm not saying you do Formal verification of 500 million
gates, because Formal doesn't do that.
But the fact is, Jasper requires the whole design to be read,
right? It requires some structural analysis. So that is
the right place to solve this problem.
It's not the same thing as equivalence checking in Formal. But
it has all the variables you need to solve the problem. So
that's where we at Cadence are investing.
Prakash: See, it will be every verification engineers dream, if every bug
found was by a Lint tool. Because that is a tool that is most
efficient tool. It is the lowest cost-per-bug found. It's just
that it cannot find every problem. So, touting Formal technology
as the way to go, is absolutely the wrong way to go.
What you need to do, is you need to invent technologies that'll
solve the problems in the most efficient way. Then you have come
at it from the customer's point of view. You have to think about
their efficiency requirements. You have to think about the
throughput requirements. You have to think about the debug
requirements. And then invent the technology, and it may not be
Formal.
But it is Static. So, I just think that Formal, in my opinion
after doing this for a long time, Formal is a small piece in the
overall puzzle -- but it is not very important. We've been
successful. We have customers who are using our CDC and Linting
tools very effectively in very advanced designs to sign off, and
Formal technology plays a very small role in this process.
Cooley: Cool.
---- ---- ---- ---- ---- ---- ----
Related Articles:
Sawicki tells what it's like for Mentor to be acquired by Siemens
Synopsys layoffs means ICC2 rewrite is unknown for 3 to 4 years out
Dean on nGDP, his next gen graph-based design data managment tool
Prakash and Anirudh spar on Real Intent Linting vs. Jasper Formal
Join
Index
Next->Item
|
|
