Sign In
Forgot Password?
Sign In | | Create Account

Multi-core/multi-OS - terminology

Colin Walls

Colin Walls

Posted Aug 24, 2009
8 Comments

Working for Mentor Graphics is interesting, but can have some challenges. One of the facets of the company that makes us unique is embedded software. There are a couple of other companies that Mentor would regard as key competitors, but they are both totally focussed on electronic hardware design – “electronic design automation” [EDA]. A very large part of Mentor has that orientation, but we are unique in having the Embedded Systems Division. Of course, there are plenty of other embedded software companies, but they do not operate in the hardware design space.

The logic behind an EDA firm being involved in software is clear: an electronic system of any significant level of complexity delivers functionality via a combination of hardware and software. As a company, understanding both domains gives us the chance to serve our customers better – to offer more complete solutions.

Of course, we do all of this, but we run into communication problems …

Although it is now more than 13 years since Mentor acquired the company that I worked for and, hence, started its investment in embedded software, I still find myself surrounded by people who speak a different language. They talk about “synthesis” and “netlists” and think that “IP” stands for “intellectual property” [whereas I think "Internet protocol"]. An operating system is something that runs on a desktop computer – an RTOS is not familiar. And when I talk about embedding Linux in a device, confused looks are the result.

I recently ran into a new terminological challenge. This came about because there is increasing interest in multi-core systems: devices that incorporate multiple CPUs – either multiple chips or multiple processor cores on a single chip. Typically, the functionality of a device is spread across the CPUs so that one [or one group] performs the high speed, hard real time activities and another looks after functionality which in not time critical. An example might be a digital TV system, where one processor looks after the processing of the media stream and another looks after file management and the user interface.

This kind of system architecture is interesting because the operating system requirements of the two parts of the system are different. The hard real time part needs a true RTOS [like our own Nucleus OS]; the other part may be best served by Linux.

The question is: what do we call these two distinctly different parts of a multi-core system? My EDA colleagues make references to the “data plane” and the “control plane”, but these terms are not intuitive to me. Am I out of touch or is it a hardware/software thing? I would appreciate any feedback either via a comment or by email. Thanks in advance.

Multi-OS, RTOS, Multicore, Nucleus Kernel, Nucleus, Linux

More Blog Posts

About Colin Walls Follow on Twitter

Colin WallsI have over twenty-five years experience in the electronics industry, largely dedicated to embedded software. A frequent presenter at conferences and seminars and author of numerous technical articles and two books on embedded software, I am a member of the marketing team of the Mentor Graphics Embedded Systems Division, and am based in the UK. Away from work, I have a wide range of interests including photography and trying to point my two daughters in the right direction in life. Learn more about Colin, including his go-to karaoke song and the best parts of being British: http://go.mentor.com/3_acv Visit The Colin Walls Blog

More Posts by Colin Walls

Comments 8

Post a Comment
I also find the control/data plane terminology a bit too specific. After all there can be situations where there is no visible boundary segregating control from data. I am more comfortable with the generic term AMP (Asymmetric multi-processing) that encompasses such use cases.

Faheem Sheikh
6:39 AM Aug 25, 2009

Thanks for this input Faheem.

Colin Walls
8:28 AM Aug 25, 2009

In the Bluetooth world the terms "host" and "controller" are used to describe this partitioning. The upper protocol stack runs on the host processor of the PC, phone, etc. while the lower stack runs in the controller, which is the Bluetooth chip itself. I have also used the terms "application processor" and "network processor" to describe the same partitioning in a more general way.

Jason Hillyard
8:50 PM Aug 25, 2009

Good input Jason. In the context of a mobile handset [for example] "application processor" and "network processor" make sense, but in other areas the terms are not general enough. For example, a high-end laser printer: one CPU looks after UI and file management and the other deals with actual printing.

Colin Walls
10:43 AM Aug 26, 2009

Colin, how about "HRT Domain" and "SRT Domain", (standing for Hard/Soft RealTime) - while it functionally saggregates the two domains (or Planes by EDA guys' talk) - it stays generic across devices having such splits...? Just a quick thought i got.

Sumeet Gupta
8:20 AM Sep 13, 2009

Sumeet: Thanks for that. I like the use of the word "domain". Not so sure about "SRT" and "HRT" - "HRT" is used to stand for "hormone replacement therapy". :-)

Colin Walls
9:27 AM Sep 14, 2009

Collin, I guess question is tricky to understand at the first as it drives to think about distributed processing and symmetrical parallel processing. I guess you're right on money,terminology is yet to be defined. I would naturally compare this system with our body. I would put brain as master processor and organs as ASP(Application Specific Processor). After that i would say CCPP(Control Command Processing Path) and CCEP(Control Command Execution Path).I guess i sound naive..gauche :)

Hardik
9:22 AM Sep 21, 2009

Hardik: Thanks. All contributions are welcome.

Colin Walls
10:10 AM Sep 21, 2009

Add Your Comment

Please complete the following information to comment or sign in.

(Your email will not be published)

Archives

Tags

 
Online Chat