With help from the providers of development tools, the personal computer world has proven the performance of dual-core devices. Putting two identical cores on a chip roughly doubles computing power. But perhaps an even greater benefit is that chips can run at slower clock speeds, consuming less power so there is less heat generation.

    The embedded computing world is now moving to adopt multicore processors as those in the server markets make the transition to quad cores and larger devices. Automakers are gearing up to make the transition to multicore chips, which are expected to make single-core devices obsolete over the next several years.

Multicore devices, such as this Freescale
MPC5510, are being phased in by automakers for better electronics performance.

     “Specifications call for our next-generation systems to be multicore,” said Michael Grimes, powertrain technical specialist at General Motors. “That’s for the 2013 model year, which may come out in early 2012. Single-core controllers will probably live beyond 2012, but that architecture is near its end.”

     Automakers and Tier 1 suppliers have used some dual-core products before, but suppliers note that the rationale for their application is changing.

     “We have done dual-core chips in the past for redundancy. Now we’re doing it for performance,” said Paul Grimme, General Manager of Freescale’s Transportation and Standard Products Group. Grimme noted that Freescale is moving forward with a triple-core device for GM Powertrain and is working with the group on implementation of a quad-core device.

     Panelists at Freescale Semiconductor’s recent Technology Forum noted that there are many areas that will soon require the performance of dual-core products. “Fuel saving and emissions reduction are the leaders, along with safety management diagnostics and the migration of functions from hardware to software,” said Claus Baumgartner, Director of Engine Controller Development at Siemens VDO.

     The transition from single- to dualcore devices will be simpler than the switch from 16 to 32 bits, many observers note.

     “The biggest change in our software architecture is that we have to assign a task to one core or the other,” Grimes said. “We don’t expect major rewrites, but we will have to pay attention and get things that we want to be together running on the same core.”

     However, panelists also noted that designers cannot expect that moving to next-generation CPUs will bring dramatic performance boosts. “One plus one is not necessarily two; it may only be 1.2,” said Alan Rooke, Manager of FEV’s North American Automotive group.

     However, that is not unusual. Grimes noted that conventional controllers with doubled clock speeds often do not deliver the big increases that might be expected. That is particularly true in powertrains because there are no long execution paths. Parameters continuously change to respond to road and driving conditions.

     System designers will have to do some tweaking in software to take advantage of multiple cores. “Dual cores are more complex,” Baumgartner said. “Tools need to deal with that.”

     Development tool suppliers are already gearing up for the changeover. “Dualcore tools let you optimize caching, and compilers also let you further optimize caching,” said Dan Mender, Green Hills Software’s Director of Business Development.

                                                                                                                                         Terry Costlow

                                                                                                                         Edit by : Selçuk KAYABAŞI