Editorial

Man vs. Machine: The IC Designer's Dilemma Too great a dependence on automation can leave IC designers spinning aimlessly in space, waiting for a rescue that only they can ultimately provide. by Jonah McLeod

I watched 2001: A Space Odyssey the other night and was struck not only by Arthur C. Clarke's and Stanley Kubrick's vision of the future, but by their grasp of the subtle battle between mankind and technology. The most memorable part of the film, the conflict between the computer HAL and his human antagonists, resonates even more in silicon-driven 1998 than it did in 1968.

What does the conflict between humans and HAL have to do with design automation? Plenty. It mirrors the conflict between engineers and the automation they view as taking over increasing amounts of their creativity. Fundamentally, the satisfaction engineers derive from their labor is inherently tied up in their coming to grips with a machine, a process, a function. They must understand how a product is supposed to behave, then conjure an invention to make it do so. Design automation, though, automates that problem-solving process, abstracting increasing amounts of the details.

The past twenty years of development in electronic circuit design, after all, has been an abstraction from the first analog circuits. As digital replaced analog, sine waves gave way to 1s and 0s, and transistors--with their complex formulae for transconductance and impedance--gave way to gates with Boolean predictability. Now gates are giving way to complex reusable functions. In so many ways, the designer has lost touch with silicon.

The problem began as far back as when Spice tools, created to comprehend the vagaries of transistor functions, were introduced. From that point on, tools increasingly took on the labor-intensive parts of an integrated circuit's design. In the back end, for example, draftsmen once created the IC's layout by hand-applying tape to large mylar sheets--hence the term "tape-out." The draftsmen were replaced with workstations and software tools that partially automated tape-out. Today's placement and routing tools completely automate the process, leaving the designer with hands in pockets.

In the front end, logic simulators replaced the hardware prototype and its rat's nest of wire-wrapped interconnect. Then followed the logic synthesis tool, which led to the adoption of synthesis methodology. Introduced as a means of retargeting an existing design into another design, it's a glaring example of the struggle between automation and those being automated. Though most designers willingly relinquished such retargeting tasks at first, the tool eventually replaced them in the actual creation of logic circuits.

One more apparent example of where automation may have moved too far, too fast--but isn't fully embraced, for now--is the silicon compiler, whereby the designer fills out a form and the design tool creates a circuit. Why did such an apparently nifty idea fail to attract designers? Did the tool alienate the designers from their inventions? Did it require too dramatic a redirection in design methodology? Or were the companies promoting silicon compilation such bad marketers that they couldn't sell the concept? In the end it was probably a combination of all three, which means that everyone involved shared the responsibility for the compiler's failure--and that designers need to rethink their responses to newly introduced technologies.

As for areas where automation is mainstream, we're starting to see the consequences. Today, IC designers are creating circuits that their tools have told them are functional--only to find the circuits failing in the field. Just as with HAL, IC designers have relied upon constructs to abstract themselves from their fundamental tasks, but now they're finding that the constructs aren't working. The constructs have been unable to account for fundamental forces such as crosstalk and EMF, forcing IC designers to reacquire a body of ignored information.

In 2001, the human eventually conquers the machine, disconnecting his computing functions until it becomes a helpless pile of circuits. In the real world, the monster confronting IC designers isn't so easily subdued. You may stop it from creating circuits that are unworkable or failure-prone, but then you're left to do the job by hand--a prospect equally abhorrent.

To voice an opinion on this or any Integrated System Design article, please email your message to miker@isdmag.com.

integrated system design  October 1998