Information Technology Reference
In-Depth Information
TI and Lockheed we'll find that. The other thing I'm doing is reading the history
of how television developed, and what a small team did to make the most crucial
developments. And I've read about how radar was developed and how the first
computers were developed. I'm trying to take historical lessons, and see where
a small team could make the critical steps. That's my approach. I'm trying to
look for those little openings and I'm trying to learn from history a little bit.
Wa lus :
So what have you learned? [Laughter.]
Wo lkow:
Well, I have learned that it
is
possible for a small team to take the
most critical first steps. I'm motivated by that. I think it's still possible. You
don't have to be a thousand-man team to get going.
Back to identifying the killer app: I'm hoping to make some ultra-low power
circuitry that, no matter how expensive it is to make, will be just a little
more ecient so it won't cause the clock in a satellite to be just a little bit
inaccurate (which makes GPS positioning a little less accurate than it could
otherwise be). That is one really narrow thing that could be a very good first
objective. And everything just seems right, because when you talk about plan-
ning a new satellite, you're actually talking ten years in the future because
everything takes a long time. So it doesn't matter how much it costs, we have
a lot of time, they only need a few, and I've got a lot of people who can
advise me. It seems like one example of just the right kind of thing for us.
So it's not the
killer
app, it's the first app.
Porod:
I think you're absolutely right. And, as I mentioned at the beginning,
you need something reliable, like a building block, that you can engineer with
and that works. And even if you have that, you still need to know what you're
going to build with it, right? This is the application you're asking for.
QCA is so different than CMOS. I think the main difference comes in the
communication between the devices. In CMOS, you can have wires and can com-
municate over long distances at very high speeds. In QCA you pay a very heavy
price for communication that carries a heavy penalty. So I think that the chal-
lenge to computer architects and for people who think about these applications
is this: What kind of applications rely heavily on local computation, so you don't
have to communicate very much but you compute at every step. This is one of
the reasons why we're looking into systolic architectures, where you push data
through and at each step you do some computation. There are different ways of
thinking about applications, but I think the key is minimizing communication
and maximizing computation. I think that is what QCA needs. And I think
you're absolutely right, just taking the layout of a microprocessor and then just
redoing everything the QCA way has very little chance in computing.
Vemuru
(Srinivasa): Are systolic arrays the best solution?
Porod:
I don't know, I don't have the answer. But I think that is a critical
problem: What applications are there, what opportunities are there, what killer
apps are there where you do heavy computation at every step, where you do
mostly computation and very little communication. It seems to me that is key.
Search WWH ::
Custom Search