•  
• EDN Europe MAGAZINE
•  
• Free SUBSCRIPTION
• Subscription Renewal
• View Media Kit

• EDN
• EDN Australia
• EDN Asia
• EDN China
• EDN Japan

EDN.COMMENT: Whatever happened to multicore?

The Embedded World Show 2008 that took place in Nuremberg in February was notable for a number of reasons. For one thing, it was busy. Trade shows in recent years have tended to be sparsely attended, compared to events of years gone by. There could be many reasons for that, but one that seems likely is that as project deadlines get ever tighter and time-to-market pressures grow, it gets progressively more difficult for engineers to take time out to browse the offerings at a big show.

Not for this one: as far as I could tell, there were crowds of engineers engaging exhibitors with genuine technology questions, throughout each of the days of the event. Which I take to be a good sign: embedded design doesn’t seem to be suffering from any great malaise of the general economy; and, if offshoring of design is a problem, it doesn’t appear to be manifesting itself in this community.

On a high-level overview of the products and technologies that exhibitors were highlighting at the event, I found myself asking, “Whatever happened to multicore processing?”. If that question makes you reach for your keyboard to dash off an email telling me that your company has a range of tools and solutions for implementing numerous different multicore strategies in the embedded space, let me ask you to read on before you do so. What I had in mind was simply the following: at Embedded World 2007, you could not walk ten paces down any of the aisles in the show without someone saying, “Can I interest you in my multicore silicon/board/OS/toolset/development environment?”: a level of emphasis that had all but disappeared a year later.

I’m not implying that the multicore phenomenon has gone away: far from it. But unless you went looking for it at Embedded World 2008 you would not have identified it as a key—or even a major—part of the products and services on display. Why should the topic have gone from the status of “seriously hot” to—apparently—“just another technology” in such a short time? Was it just a “techno-fad”?

Of course the answer to that last question is: no. There are products at all levels to support you if you should wish to explore the multicore phenomenon. Part of the problem—if there is a problem—may be that no sooner had multicore become a “hot topic” that it had also come to mean many different things. For example, marking a transfer from the PC domain into the embedded space, an announcement from Intel notes that the company is extending its support for the embedded sector with multicore processors in 45-nm technology with seven-year supply commitments, which will let you get dual-, quad-, or octal-CPU configurations within reasonable boardspace and -power constraints. Which you can file under the heading “serious processing power”. AMD, likewise, has a programme of proliferating multicore parts from the PC space into the embedded. For the deeply embedded space there’s an example— to cite just one—of a dual-core microcontroller from Renesas (more details in next month’s edition), which that company is aiming at multimedia systems right now, but at the industrial sector in 2009. The notion of “multicore” spans everything from a handful of threads running on two processors, up to full-symmetric-multiprocessing running over arrays of CPUs. It also extends into dimensions other than processing power: the embeddedsystem designer, if he or she is inclined to venture into multicore processing at all, may do so for reasons other than achieving more computing power—a more likely motivation is to spread the activity across more than one processor and cut the clock rate—and hence power demand—on all of them.

Here’s my guess—so far, unchecked— as to what might be going on. I suspect that the embedded-design community— excepting the fraction that really needs all the processing power it can get its hands on—looked at the full spectrum of the hype about multicore, starting with adding a co-processor or companion core to a design, and took a very cautious approach. Most designers who have tried it—including those whose involvement predates recent fashion—have likely opted for the more modest end of the complexity scale, breaking their code into sections explicitly running on one core or another, and—for now—has left aside the fancier options. From the vendor side of things, I’d guess that the uptake of the more sophisticated solutions has been a bit on the slow side, hence the reduced emphasis in 2008’s exhibition promotions. Multicore may turn out to be a “slow burn”; but don’t underestimate the influence of those Intel and AMD offerings— by making the term ubiquitous in the PC space, it is bound to work its way into design thinking in the wider world, and make progress for that reason alone.