The dual-core processors that dominate the PC world are slowly making their way into automotive applications. However, for the technology to make significant inroads, OEMs and Tier 1 suppliers are seeking help from tool and chip providers.
In powertrains, dual-core processors are still about five years away. But suppliers are now pressing forward with more urgency in case automakers shorten their development cycles for twin CPUs, which provide more processing capability and generate less heat than single-core devices.
“Our performance requirements will hit the dual-core levels in 2013, though that could go faster if there’s greater demand from something like higher fuel prices,” said Claus Baumgartner, Engineering Director at Continental Automotive Powertrain Electronics.
Speaking at the Green Hills Software Embedded Software Summit, he noted that tool companies will play a critical role in the transition. “High-performance development tools are the most significant enabler,” Baumgartner said. “We have a lot of projects for dual-core chips, but we don’t yet have a lot of expertise with them.”
Green Hills Chief Technical Officer David Kleidermacher noted that his company and other tool providers have a number of tools for dual-core applications. But he also acknowledged that the chips, which are used in only about 10% of the embedded markets, pose bigger challenges than single-core architectures that have been used for years.
“Debugging multi cores is a lot harder,” Kleidermacher said. “Up to 80% of the development cycle for systems is now spent on debugging, and the numbers are on the rise for multi cores.”
Chip makers are also stepping up their efforts to help customers transition to multi-core processors. As engineering staffs struggle to add more complexity while shortening design cycle times, chip suppliers are providing more support, particularly in the emerging multi-core field.
“OEMs are relying on semiconductor companies for systems solutions. Integrating software and multi-core chips is one of the key challenges,” said Michel Mayer, CEO at Freescale Semiconductor.
Some observers predict that dual-core chips will see their first usage in body applications and infotainment. Baumgartner will leverage the knowledge gained by other Continental groups that address these applications, but he noted that performance as well as environmental and reliability issues are different, so the powertrain group will have to perform its own studies.
At this point, there are still questions about the best ways to use chips that have two processors. Kleidermacher noted that techniques for parallelizing code play a major role in the performance benefits gained by using dual cores. Those benefits can range from nearly doubling throughput to as little as a 20% speed-up. At Continental, analyzing this problem is an ongoing process.
“We have time components we need to deal with, usually associated with things like the crank angle,” Baumgartner said. “We could use one core to crunch algorithms and another for I/O. That hasn’t been decided yet.”
Though there are big challenges with first-generation designs, automotive engineers are looking forward to successive designs that will use four or more processors on one die. “Once we have mastered dual cores, we will move forward with more cores,” Baumgartner said.
However, these complex chips come with yet another set of issues. Sharing data from many cores requires lots of communication bandwidth.
“When you scale to more cores, the memory bottleneck is the key roadblock,” Kleidermacher said. “You can’t have just one memory bus. The semiconductor companies are all looking at solutions for this.”