Cooley ran the panel by asking questions submitted by users. Prompted by a comment in the recent Design Automation Conference trip report, he opened by asking Pravin Madhani, CEO of physical synthesis startup Sierra Design Automation, whether he was going to hire any non-Indian employees or sell his company's software to non-Indian customers.

Madhani was prepared. "I had to go cube by cube through my employee base to see who was what," he quipped. Madhani then displayed slides showing that Sierra's employee base is 50 percent white, 24 percent Indian, and 18 percent Asian, and that the customer base is 91 percent non-Indian.

Madhani also put up a slide showing that none of the company's employees are located in India, causing Cooley to ask if he was lacking an India strategy. "I can't win here, can I?" Madhani asked.

Cooley next questioned Gartner Dataquest's market statistics, which led to a discussion about ESL. Gary Smith, chief EDA analyst at Gartner Dataquest, said he sees the most ESL growth in the "non-hardware engineering" area.

"Most of the systems engineers are software guys," Smith said. "The serious systems issues are primarily software, and where we're in really deep trouble is in the software area," he said.

Cooley then asked Rajeev Madhavan, Magma Design Automation CEO, whether Magma made a mistake in sending a patent assertion letter to Synopsys and thereby triggering the Synopsys patent infringement lawsuit against Synopsys. Madhavan responded by saying Magma has to protect its intellectual property, and he then said that "this industry has to turn down the number of fights it has."

Antun Domic, general manager of Synopsys' implementation group, said that Synopsys has had only one "serious piece of litigation" other than the Magma suit in the past five years — and that was Synopsys' lawsuit against Nassda. Madhavan shot back, stating that Synopsys sued Ambit Design Systems the night before Ambit introduced its first product.

"To say there is not a large amount of litigation in EDA is sweeping everything under carpets," Madhavan said. Jacob Jacobsson, CEO of Forte Design Systems, then said he would "side with Rajeev. This industry has an image of being very lawsuit happy."

Cooley asked Jacobsson why Synopsys abandoned SystemC synthesis. Jacobsson responded that Synopsys' Behavioral Compiler didn't work because it came out too early and lacked customer input. Forte's SystemC-based Cynthesizer, he said, had considerable user input, and has made verification an important part of behavioral design.

Domic noted that Synopsys supports SystemC for simulation and high-level verification. "To add one more language for RTL synthesis, we did not think that was of much value," he said.

Robert Hum, general manager of Mentor Graphics' verification and test group, noted that Mentor's ModelSim simulator supports SystemC. Mentor's Catapult synthesis tool is based on ANSI C, he said, because that works well for the "flow-based, pipelined" designs the tool targets, and because "dataflow does not need all of SystemC."

Ted Vucurevich, Cadence Design Systems CTO, was asked why Cadence won't talk about its pending acquisition of Verisity Inc. "People are reading too much into our reticence to talk than there actually is," he said. "It's very pragmatic. It's an opportunity to integrate great people and great technology."

When Cooley asked if Cadence plans to kill Verisity's "e" language, Vucurevich didn't answer, but he did say that Verisity's Axis products are complementary to Cadence's Palladium products because Axis focuses on acceleration and Palladium focuses on in-circuit emulation.

Cooley asked Domic how many customers Synopsys has lost to Cadence's Get2Chip synthesis and Magma's Blast Chip. "Extremely few" was the answer. Domic acknowledged that there's competition, but he said the last two years have produced "dramatic improvements" to Synopsys' Design Compiler.

Madhavan argued that the whole notion of "front-end" synthesis is outdated, because much of the optimization needs to be done during placement and routing. "We are at a change," he said. "There is not going to be a front end and a back end. If I do it all in the same system I get a greater value."

"We need to do more optimization at the back end, but to say front-end optimization is useless is wrong," replied Domic.

Asked why Cadence bought Get2Chip after buying Ambit synthesis, Vucurevich said that both offered excellent technology. Get2Chip, he said, had "very bright technologists who observed the problem from higher than RTL on down," and were able to bring global optimization into the picture.

Perhaps the most contentious issue concerned 90 nm tapeouts. Madhavan said that Magma had around 44 percent of tapeouts at 0.15 microns and below in 2003, and has over 50 percent of all 90 nm tapeouts. Domic then said that Synopsys has around 60 percent of all 90 nm tapeouts, including placement and routing as well as synthesis.

This prompted Cooley to ask how Cadence can have "minus 10 percent" of the 90 nm tapeouts. Vucurevich said Cadence has had over 100 90 nm tapeouts. Cooley turned to Dataquest's Smith, who said Magma, Synopsys and Cadence each have about one-third market share for 90 nm design.

But dollars and actual tapeouts may be a different story. "Rajeev's sales force has been very disciplined in selling tools for what they're worth," Smith said. "I'm not sure the other two have followed that path. Some tools have been given away very cheaply."

In another provocative question, Cooley asked Vucurevich what Cadence plans to do about its lost market share in timing analysis and physical verification. "We will be back," Vucurevich said. He then noted that the timing analysis market is about to change as statistical timing comes on board. "You'll then see a more competitive position from Cadence," he said.

Madhavan noted that Magma's upcoming "Cobra" release includes statistical timing. Domic said Synopsys is working with IDMs to make sure it has good information for statistical timing analysis. "Providing statistical timing with information that's usable is going to be a key problem," he said.