Heard at DAC: Is workflow automation the next frontier for EDA?

Three exhibitors on the show floor at DAC this year stood out from the crowd-not for their huge booths or the novelty of their new products, but simply by being so different from the mainstream of EDA thinking. The three were Dassault Systemes ENOVIA, IBM Software Group, and Methodics. Each is involved, in one way or another, with Product Lifecycle Management. PLM plays a growing role in enterprise operations generally, and as the scale of IC design efforts increases, like it or not PLM probably lies in the future for many chip design teams.

But what is it? PLM is one of those grand terms so capacious that it risks conveying nothing at all. But to the IC design profession it has some specific meanings. One-perhaps the first to be automated in the EDA world-is simply version control. A second aspect is workflow management: tracking data sets, tasks, and outcomes through an organization as the project progresses. Combine these and season with message-tracking, conferencing, and perhaps a Wiki, and you get an inter-team collaboration platform.

But wait-there’s more. Also crucial to PLM is the notion of traceability. Every object in the design, from a line of C++ to a signal to the OPC decorations on a polygon to a field-service call should be traceable back to the specific design requirements that justified it. And then there is documentation: of every step in design, verification, production, and test. As you can imagine we are starting to talk really big data bases here-this is not a job for one person with Excel.

So far all these are familiar tasks. Chip design teams have been finding ways to do version control or build control, and have taken semi-automated stabs at workflow for years. Aerospace designers have lived with traceability and documentation mandates, such as DO-178B, at least in the software world. And some of the vendors are familiar as well. A major component of Dassault’s product traces its heritage to Synchronicity, which was trying to sell collaboration tools to chip designers years ago. Similarly, the Methodics project IP control tool is known and used in parts of the industry.

Two points are less familiar. First, the many point-tools are showing signs of polymerizing into a comprehensive platform. Such a platform would support all aspects of the product life cycle and offer unprecedented opportunities for data mining. And second, separate from this trend but enormously enriching to it, design teams are beginning to attach quantitative figures of merit to the outcomes of design and verification tasks.

Indavong Vongsavady, projects director at STMicroelectronics, said that for chip designers the importance of PLM tools may lie in their ability to methodically collect design and verification metrics to support decisions. “Most tools produce metrics,” Vongsavady said, “but not all engineers interpret them the same way. There are different ways of measuring quality of results.”

Vongsavady said there are some obvious measures of design goodness-such as the distribution of timing slack in a block or power-consumption estimates. There are also less obvious metrics: the degree of manipulation or the number of buffers a netlist requires in order to close timing, for example. Experience tells a design manager how a certain structure should behave, allowing the manager to normalize such metrics and estimate whether a block is doing fine or whether it requires intervention. PLM tools can play a key role in capturing, organizing, and preserving that experience, and in mining for nuggets that might not be obvious on the surface. A similar case can be made for verification metrics.

The sheer scale of next-generation designs, comprising as they will huge gate counts, IP from a wide array of sources, and massive software efforts, will require some sort of project-management automation. Existing PLM tools, although they may seem foreign to the EDA world, are likely the best candidates. And in applying PLM to electronic systems design we may find that the data-mining capabilities of the tools add a new dimension to productivity improvement.