The synthesis story began in the early 1980s at General Electric (GE) in Research Triangle Park, N.C. We were working on a technology to automate the way design engineers get a circuit from logic equations to netlist. Back then, engineers would start with a design description, as we do today, but then they had to manually enter the design into a schematic entry package to get a netlist. It was tedious, painful, and error-prone work.
To test our new, automated approach, we took an existing design at GE (somewhere around 70 gates) and entered its logic equations into a computer. As we expected, the computer gave back a netlist automatically, but with one unexpected fringe benefit: It was both 30% smaller and 30% faster than the one derived manually from the original equations. Over the next few months, we ran test after test. Sure enough, we had made a key breakthrough. Not only had we automated an onerous process, but we had improved area and timing.
As we approached designers with this new process, we would typically ask them, "What if there were a 'Technology X' that could read in your design and timing parameters, then automatically create a netlist? What if it could achieve better results than your manual design methods—and do so in a matter of hours, instead of months?" Of course, they said they would love it, but they didn't believe such a process could exist.
So we took their schematics, entered them as equations, and the computer churned out a netlist. You can imagine their surprise when we took a design that they had worked on for months, synthesized it in hours, and got a netlist for a circuit that was both smaller and faster than what their experts had designed.
At first, it was more than surprise: Designers flat-out didn't believe it. The design engineers couldn't figure out what the computer had done, or even if the netlist was the same as their original design. We ended up having to create a schematic generator to translate the netlist back into the schematic form, just so they had the necessary proof that the results were real!
This process wasn't merely the beginning of synthesis. It started the timing optimization technology that we have continued to use for 17 years. Recently, I saw a design that had used the synthesis process on almost 7 million gates. That's 100,000 times more complex than our first design, and the technology is still going strong. Synthesis is proving to be the "Energizer Bunny" of design methodologies.
People sometimes ask if I miss the excitement of the early days. Inventing a technology that revolutionized an industry was exciting, but when I see a 7-million-gate design, I still get excited. The revolution has barely even begun.