One thing I didn't find at the recent Design Automation Conference (DAC) was much activity in system-level design. We've heard for years that the move beyond register transfer level (RTL) is coming-but we just can't seem to get there from here.
Seems to me I've been writing about system-level design ever since I started covering the electronic design automation industry in 1985. I still remember a company called Endot that had a system-level language in those days. It didn't get far. And then there was Silc, a company that had a behavioral logic synthesis tool with a Lisp-like language. It was about 20 years ahead of its time, and it soon vanished.
Moving into the 1990s, Redwood Design Automation came forth with what it claimed to be a new, post-RTL design paradigm shift. A few years later, Cadence Design Systems Inc. bought what remained of that company. The Behavioral Compiler from Synopsys Inc. was touted as a revolutionary breakthrough when it was rolled out, but it's turned into a niche tool used by a small group of adherents.
And then, in the mid-1990s, something called ESDA was all the rage. Can you remember what ESDA stood for? If not, see the end of this column.
A couple of years ago, there was considerable publicity and debate about C-language design. Synopsys and CoWare Inc. started the Open SystemC Initiative, and there were a number of panels at which supporters talked about C-language design as the next great thing, while chip designers talked about how much they hated it.
DAC was quiet this year on the system-level front. CoWare announced some upgrades, and startup Novilit Inc. offered its AnyWare product for communications protocols, but the fanfare just wasn't there. The C-language design panels seem to be a thing of the past.
So why can't we ever seem to get beyond RTL? Perhaps we've been looking in the wrong place. I think the move beyond RTL will happen not in design, but in verification. We have already seen how languages like Vera and "e" have raised the abstraction level, and now it appears that assertion-based design will become a significant and widely used technology.
The best chance for easing beyond RTL is the new SystemVerilog 3.0 standard, with its support for assertions, C/C++ data types and a high-level "interface" construct.
Oh, yes. ESDA stood for Electronic System Design Automation. Now, how many ESDA companies can you name, and where are they now?