Hardware design lags behind software design by about 10 years. Software used to be written by one or two solo programmers. Usually, little code was reused. Everything was written from scratch.

At that time, programmers complained that compiled code ran slower than hand-assembled code and that high-level languages didn't produce the tight code density of assembly language. They hated commercial operating systems because rolling your own kernel was cheaper. But none of that mattered. Programmers discovered that productivity was more important than writing everything from scratch.

Today, the same thing is happening in hardware design.

It's impossible to bit-twiddle a hardware design for an ASIC or system-on-chip and get anything done in reasonable time. To produce significant value (i.e., a million-gate chip), you can't obsess over individual transistors, gates or even macro functions. At that level, even a microprocessor is just a function block. Encryption engines, DSPs and memory arrays become just so many blocks to arrange in the overall design.

Engineers must overcome the not-invented-here factor. Yes, you probably could design a slightly better Ethernet controller than the one you could license, but why? Would your company-or your customers, or your product-benefit from the months of design, testing and integration that such a task would require?

Licensing all the intellectual property (IP) you possibly can means 80 percent of your project is already done. The real engineering comes in finishing the last 20 percent-the unique value of the product.

To do this, engineers must overcome the urge to design each individual component and even abandon skills that were honed over a career.

Since most functions and components exist as reusable, synthesizable RTL designs, it makes sense to concentrate on their organization-the chip architecture-not on the components themselves.

Tomorrow's chip designers need to think in more high-level architectural terms and less in terms of detailed design. Architecture is what most designers aspire to, and now it's within almost everyone's grasp. Leave the grubby details to the IP construction companies.

Chip design will value "renaissance people," those who can think holistically and at a high level of abstraction. Speaking the languages of both hardware and software will be key. Switch-hitters will be rewarded. A working knowledge of gate-level interactions will be an historical anomaly, a skill that is no longer valuable to all but a few specialists.

So don't sweat the details, even if that's what you know best. Trust your tools. Think in terms of hardware and software working together. And architect products at a higher, more abstract level. In other words, save your talents for the cool stuff.

Jim Turley is Vice President of Marketing at Arc Cores Ltd. (Elstree, England).