| Abstract | There is much discussion of two challenges in the design of tomorrow's
electronics: the difficult "small physics" of nanoscale transistors,
and the silicon/software complexity of "big systems". But those of us
who want to build beautiful algorithms have an additional hurdle: "tall
tool-chains". If it takes 50 tool-steps to build an
industrial-strength design flow, and each tool is based on 1-2 "big
algorithms", does this mean that each new algorithm idea is worth, at
best, 1-2% of the success of a design? This seems to me a bad way of
accounting for the tremendous value that EDA brings to the world of
design. How can we have a big impact in this important technology area?
In this talk, I will offer several pieces of advice for how not to get
buried by the tall-tool-chain problem. I will discuss how to identify
design problems that can have large impact, how to embrace the strange
physics of tomorrow's silicon technologies in the service of building
beautiful algorithms, and how to get fresh (and unique) insights on
problems by spending time working with a real design team. I will use
design examples ranging from lithography, to computational finance, to
silicon-based speech recognition, to illustrate the point that this is
an exciting time to be working on tomorrow's tool and design challenges. |