( ESNUG 555 Item 1 ) -------------------------------------------- [01/22/16]
Subject: Anirudh on CDNS creating Genus, a 5th or 6th CDNS RTL synth tool
DAC'15 Troublemakers Panel in San Francisco, CA
Cooley: Anirudh, why is Cadence doing yet another RTL synthesis
tool? I think this is the 5th or 6th version.
Anirudh: Well John, thank you for an excellent question.
Cooley: And my follow-up question is how many patents are
in that tool?
Anirudh: We have a lot of patents... So in RTL synthesis,
first of all I think there is a big opportunity. And
what I mean by that as we go to lower nodes EDA companies
have to provide a full flow, it's not just about individual
tools. And we put a lot of effort in place and route, but
you know it is not just about place and route, nor timing.
We have to provide a full flow and that has to start with
RTL synthesis. Because these days with most design, the
customers I go to they look at RTL-to-GDSII, or really it's
RTL-to-signoff.
And there's still a lot of problems in the RTL synthesis
space. One of them is correlation to implementation and
the full flow.
And the second one is this whole RTL productivity issue.
These chips are getting bigger and I think RTL productivity
can be much higher than it is right now.
Hogan: So you're saying Anton, after 30 years, hasn't got it
right yet?
[ audience laughs ]
Cooley: Can you say that a little louder in the microphone
because the SNPS lawyers will want to hear that later.
[ more laughter ]
Anirudh: I'm not saying any individual names but I think there
is a lot of scope for improvement in RTL synthesis. And
there's a lot of scope improvement in synthesis-to-
implementation-to-signoff.
Cooley: But what you're saying physical synthesis is not a
reality. You're trying to do it now for the first time.
Even though Synopsys has been pitching that idea for maybe
close to a decade?
Anirudh: You mean... OK, so what happens for example is that
physical synthesis, these are very big words with a lot of
meanings, right? Physical synthesis means yes you have some
physical information in RTL synthesis.
Cooley: Right.
Anirudh: But the question is how much information do you have?
You know so some placement information is old technology,
in synthesis. But I don't think just placement information
in RTL synthesis is enough. Because these days routing impacts,
timing impacts, extraction impacts.
So there needs to be a lot more synergy between implementation
and RTL synthesis than just the old placement -- and the
productivity of the flow -- and that's what we see with real
examples. So everything has to evolve to meet the new issues.
Join
Index
Next->Item
|
|
