SAN DIEGO, Calif. The EDA industry is starting to show some good growth, but needs to help solve the growing software development crisis, according to analyst Gary Smith. Further, Smith said, EDA tools need to adopt parallel programming, and many need to be completely rewritten.
Smith and other analysts from Gary Smith EDA spoke at a Sunday night presentation on the eve of the Design Automation Conference (DAC) here June 3. At past conferences, Smith has spoken as a Dataquest analyst. Dataquest closed its CAD group last fall, however, and Smith subsequently launched Gary Smith EDA.
As he's traditionally done on the night before DAC, Smith offered an EDA market forecast. This year's forecast is optimistic. It predicts 2007 EDA revenues of $4.795 billion, up from $4.423 billion in 2006. Smith predicted $6.585 billion revenues in 2011. "We're off dead center and up and moving," he said.
But the design world is changing, Smith said, and EDA is "alive and well but software challenged." He said that semiconductor vendors want EDA vendors to provide tools for all the design challenges they're facing, and their number one design problem is software.
"They're now looking for EDA vendors to provide a total hardware and software solution," Smith said. "If you want to stay providing the stuff you've always provided, you probably won't be around in five years," he said. "If you want to take advantage of the opportunity, it's a nice market."
Smith showed a second chart in which embedded software development tools are added to the EDA market. In this chart, the combined market totals over $8 billion in 2011. The problem: "Only a couple of EDA vendors really understand embedded software."
Smith struck a similar theme at last year's Design Automation Conference, where he said that the biggest problem with system-on-chip (SoC) design is embedded software development.
Smith's other point this year was that EDA vendors have a considerable amount of software work of their own to do they need to parallelize their algorithms to handle designs at 100 million gates and beyond. While some existing vendors have done "surface rewrites" of tools for parallelism, Smith said, "most tools will need to be completely replaced. This is a big deal. The ideal would be concurrent algorithms."
Support for multi-processing and multicore architectures is in fact a major theme at this year's DAC, where several startups are introducing EDA products written from the ground up to take advantage of parallel processing.
Daya Nadamuni, analyst at Gary Smith EDA, spoke about the "crisis in concurrent programming" for multicore architectures. "How do we take parallel programming, which is not well understood in the mainstream, and proliferate it out to applications developers?" she asked. "Do we have ten years to train everybody in concurrent programming before the crisis hits? Multicore is here today."
EDA vendors are not sure how to deal with software, she noted; that's always been somebody else's problem. Embedded tool and real-time operating system (RTOS) vendors don't know how to deal with multicore. Electronic system level (ESL) providers aren't sure where they stand. The key to success, Nadamuni suggested, is good programming models.
Tom Starnes, analyst at Objective Analysis, discussed hardware and software design considerations for multicore systems-on-chip. He noted challenges including task partitioning, memory hierarchy, bus structure, power domains, process technology, verification, simulation, and power dynamics.
"You've got to have a way to simulate the entire system and certainly all the processors and all the software in a timely fashion," he said. "One of the most difficult things to simulate is the dynamics of power."
"Multiple processors are here. Deal with it. You're going to have to work with the entire system design team," Starnes said.