( ESNUG 563 Item 1 ) -------------------------------------------- [11/01/16]
Subject: Sawicki on how MENT Calibre DRC/LVS "breaks at almost every node"
DAC'16 Troublemakers Panel in Austin, TX
Cooley: Joe.
Sawicki: Yes sir.
Cooley: You have this tool called Calibre DRC/LVS.
Sawicki: I've heard of it.
Cooley: Ok, the more I've learned about that market and fabs, I've
learned, well two things: first, things are getting small...
Sawicki: Itty bitty.
Cooley: So when is Calibre going to break? 5nm? 3nm?
Sawicki: Umm... It broke ages ago.
Cooley: Huh?
Sawicki: Calibre broke ages ago. And we had to rewrite a portion of
it. And it broke again the next node, we had to rewrite a
large portion. It breaks all the time.
Cooley: But you're starting to get to quantum physics type weird
stuff.
Sawicki: Yeah, the thing about it is if I look at right now the code
base of Calibre -- vs. what it was 5 years ago -- we probably
have rewritten at least half the Calibre code at this point
in time.
The fundamental algorithm that does distribution... In our
last... since we did our first distribution, well let's go
back to SMP. We did our first SMP right around the year
2000 time frame.
Cooley: SMP?
Sawicki: Sorry Symmetric Multi-Processing, single CPU distribution on
multiple cores in Calibre.
Cooley: Alright.
Sawicki: We have probably rewritten our parallelization engine inside
Calibre... Hmmmm... At least 6 times.
Cooley: Wow.
Sawicki: To be able to get the DRC/LVS scalability and performance you
need out of this stuff...
Cooley: That's speed up stuff. I get speed up stuff. I'm talking
about the physics starting to get weird.
Sawicki: Well, what physics are you doing? If you're in the DRC space,
that's fundamentally a specification-based system. The physics
will result in rules around what your metal enclosures need to
look like, what sort of density equation you need to have.
Most of the stuff we hit on in physics happens in the OPC
space -- where the things we have to do is model the photo
process through the scanner, down through the resist, down
through etch formation -- has all completely changed.
Calibre changes every single node because of that weirdness.
I mean we now have to model things like the light, we call
light - I wouldn't have this EuV stuff shine in my eyes - as
it goes down into the resist, then it bounces off the under-
layer deposition layer for the p-well region, then bounces up,
and then affects how you're doing imaging on your FinFET.
And the modeling of all this is just a bitch.
But yes, our Calibre breaks at almost every node in some big
fundamental way and we rewrite it every node in some big
fundamental way.
Cooley: Ok.
---- ---- ---- ---- ---- ---- ----
Related Articles
Sawicki on how MENT Calibre DRC/LVS "breaks at almost every node"
Anirudh defends JasperGold's rep and the CDNS-RocketSim purchase
Dean on his new peer-to-peer chip design workspace accelerator
Amit on 55nm non-Gaussian variation starts, plus IoT low VDD issues
Raik on it's only OneSpin vs. Jasper in formal; not SNPS nor MENT
Join
Index
Next->Item
|
|
