It is interesting to see how the once widespread concern over design-for-manufacturability (DFM) has begun to recede. Yes, progressive process nodes will present further fabrication challenges. However, users now believe that the EDA world is well-placed to address them.
EDA vendors have added more depth and breadth to their offerings. Foundries have provided necessary data for tools and, through dedicated flows, played a key role in integrating the available software. All this has given users confidence that they can introduce DFM into design flows with relatively little pain, thus promoting sub-90nm geometries to the industry as a whole.
Such comprehensive options speed progress along the learning curve. OK, so how about we try and do the same thing for ESL?
System-level concepts have been around for much longer than deep sub-micron DFM. Yet the market still appears to be made up of a plethora of point tools, many of them excellent, but which still place a considerable burden on the user to knit them into a methodology.
So much is evident on the conference circuit. At this year’s DATE and DAC, it was predominantly the IDMs that presented on the ‘how to’ of ESL, where they had convinced themselves of the ‘why’. In design tiers below Mount Olympus, many companies cannot see the ‘why’, or if they can, fear the pain and expense of broad deployment.
DFM may have had two advantages, although one is debatable. Let’s tackle that one first. It is said that the lag in ESL adoption can partially be attributed to the fact that the concept has gained greatest traction among users in Japan and Europe, while tool development is driven by US companies more wedded to their local market. By contrast, US users have led the way on DFM, hence more rapid progress. The problem with this argument is that vendors long ago realized that they are players in a global market – or not players at all.
The second reason has more force. This comes down to the foundries’ roles as tool integrators. They took on this mantle because of an economic imperative. The foundries always knew that they had to find ways to fill their lines at 90nm and below by making the technology widely accessible.
If ESL is to achieve wide acceptance, it needs someone to fill that integrator role, be it a vendor or even a design consultancy. Certainly, its increasing use among IDMs – and the benefits they report in terms of efficiency, NREs and time-tomarket – puts ESL at a tipping point. Then again, it feels like it has been at this point for a long time.