The Rising Stake In Software Tools

By Ed Sperling
The growing importance of software and off-the-shelf IP in semiconductor design is beginning to change the dynamics of the entire EDA tools business, setting off a string of acquisitions as the largest players realign themselves to take advantage of this shift.

The most recent example: Mentor Graphics’ acquisition this week of CodeSourcery, a GNU-based Linux toolchain and services company. The public face on the deal is that while no one can make money directly selling open source, there is plenty of money to be made in the tools and support around it.

But there’s more to this acquisition than meets the eye. Over the next couple of years some very significant changes will materialize in the SoC design world. First, more and more fabless companies will be required to develop at least some software with their hardware. Many chip companies don’t necessarily see this as a big opportunity, however. The general consensus is that no one really makes enough money selling software to warrant the investment, so open-source drivers, firmware and libraries will be very much in demand.

Second, as the industry moves to 3D stacked die, beginning in late 2012, it’s highly likely that chips won’t just have one embedded processor on them—they will have multiple processors, possibly from multiple vendors. And all of this will need to be controlled through software, most likely open-source. While it’s cheaper to use open-source software, there’s also a requirement that the tools and support be of the same caliber as commercial tools.

“We’ve seen an explosion in the software stack,” said Nick Lethaby, manager for software partner networks at Texas Instruments. “There is more connectivity and there are more sophisticated tools.”

He said that while some customers still want a slimmer real-time operating system, the bulk of the market is heading to Linux and Android (which is based on a modified Linux kernel). But he noted that customers are wary of software that is not fully supported. “From TI’s perspective, an ecosystem with commercial support is really important.”

Mentor is not alone in seeing this shift. All three of the large EDA players are now heavily invested in software tools. (Even emulators are increasingly a software verification tool, despite the fact that most of them are bought by the hardware engineering groups.)

Cadence’s entire EDA 360 push is focused on getting the EDA industry to shift direction so that application software drives everything below it. Linux and Android are key pieces of that market, and developing the tools to integrate the hardware and the software and commercial, off-the-shelf IP marks a fundamental change in how chips are architected.

What’s interesting is just how quickly this shift is occurring. Pre-downturn, companies developed much of the IP themselves and the device manufacturers developed much of the software. Post-downturn, as much as 90% of all content in a chip is re-usable IP and a greater part of the stack is now in the hands of the chip developer. That helps explain Cadence’s purchase earlier this year of Denali, which makes tools to model IP.

This shift toward commercially available IP poses huge issues, as well. The IP has to be power-aware, it has to work in complex environments for which it may not have been designed, and the various power states and voltages must be managed through software. This is a potential bonanza for EDA companies, which have been dealing with a shrinking number of design starts and relatively flat revenues for at least the last several years.

Synopsys, the third player in this EDA triumvirate, has pushed heavily into the IP itself with the development of standard I/O IP and the purchase of Virage Logic last summer. Synopsys is particularly well suited to take advantage of this shift to commercially available IP, with its expertise in tools and its understanding of proximity and power effects. The company is one of the very few that can provide users with detailed information about how its IP will perform under a variety of conditions, such as voltage or process variations and in noisy conditions.

The company also has invested heavily in software prototyping so that chipmakers can begin developing software earlier in the design cycle, thereby shrinking the time it takes to get a functioning chip out the door.

So what does this mean for EDA tools? If the large bets placed by the Big Three EDA vendors are correct, it will mean significantly greater sales to new customers rather than more tools to a shrinking number of customers. The challenge will be getting customers to pay for the perceived value of those tools, which the software industry—and increasingly the FPGA prototyping industry—have been reluctant to do in the past.

The question now is whether they can educate the software industry about the economics of automated development and change the thinking among the software engineering management. So far, no one knows the answer. But given the complex issues that need to be solved in future designs, as well as the recent spate of acquisitions, at least everyone is in agreement that an opportunity exists to tap a new customer base.