IP quality: design's tough nut

But many design managers level withering criticism at this same IP. "We assess IP quality by surveying the users, not talking to the vendors," Naveed Sherwani, president and CEO of Open Silicon, said at a recent conference. "In this light, all IP is defective, almost by definition."

What statistical data there is on the causes of design spins appears to support Sherwani's pessimism. Respins are often, though certainly not always, caused by problems that lie in third-party IP.

Statistical data can indicate the presence of a problem, but it can't affix blame. And without knowing the mechanism of failure, it is hard to avoid it. Often, anecdotal evidence says IP has problems because it has not been evaluated in the exact functional role in which it is being used, or with the exact design flow or in the current version of the ever-changing semiconductor process.

This leads to a massive network of finger-pointing:

"This IP stinks."

"No it doesn't, no idiot would use it the way you are trying to."

"It violates design rules."

"Not the ones we have."

"No wonder — you haven't done a shuttle run with us since 1987."

And on and on.

Clearly, a solution will take cooperation. But just who will lead and who will give how much are, shall we say, not areas of initial agreement.

Taiwan Semiconductor Manufacturing Co. has added urgency to this discussion with a dual proposal (see at right). On the one hand, TSMC's Kenneth Weng and Kurt Wolf argue that IP quality is so important that there should be a chief IP officer in each user organization. On the other hand, TSMC itself rigorously tests and inspects IP listed on its Web site and now offers its own branded libraries — picking winners and losers, some would say.

TSMC has clearly taken steps toward a solution. But some are concerned that such a position gives foundries far too much power in the consolidating IP industry, and leaves the users holding far too much responsibility. Artisan Components' Mark Templeton, for instance, argues for a mutual approach to cooperation (see below), leaving the originator of each piece of the technology responsible for its quality. It is a vital question, but not one that's likely to see a quick resolution.

By Ron Wilson (rwilson@cmp.com), semiconductors editor for EE Times