I can't believe it. It was six weeks before the Design Automation Conference and Rajeev Madhavan, the chief executive officer of Magma Design Automation Inc., sweet-talked me into sitting through one of his Web seminars. "This is the end of synthesis as you know it, John," he said. "You've got to see our new stuff! It's your duty as the ESNUG guy to do this."

So I was stuck sitting through a Magma infomercial. Blah, blah, blah. Madhavan's people barked at me about BlastCreate, their spiffy new RTL synthesis tool stuffed inside the Magma tool set.

Howard Landman gave a talk on the predictability of the new Magma RTL-to-GDSII flow. He had studied two designs by repeatedly running them through the Magma flow with minor changes in the input constraints. For example, he ran a 5k-cell microprocessor design at 2.1 nanoseconds, then 2.2 ns, then 2.3 ns, constraints, etc.

Landman then plotted the area vs. delay. Yup, the Magma flow produced the expected "banana curve." Actually, it was more of a fuzzy "hard-elbow curve," but that didn't matter much. The important thing was that when your Magma input constraints were incrementally tightened, the area of your design grew incrementally higher. That is, the Magma flow wasn't drunk and unpredictable, but acted pretty much the way a user would want.

I stopped and thought a bit. Then it hit me. Landman had performed the exact same experiments six years ago using Synopsys' Design Compiler, which he presented at SNUG'97. Same fuzzy banana curves and everything. After the Web seminar, bored, I told Rajeev this. He replied, "The big difference is that Magma is doing it now and we've added GDSII. Magma's the next generation, John." And he said that with a straight face, too. Whatever.

If you want to see both of these presentations, go to the downloads section of www.DeepChip.com and they'll be there waiting for you. The more things change, the more they stay the same. Yawn.

http://www.eet.com