Power and programming
Multicore processors – they’ve been here for some time now, and it appears as though they’re here to stay. So perhaps a better question than “what’s after multicore?” is “what comes next for multicore?”
Traditionally, multicore processors have been a collection of homogeneous CPUs arranged in either cache-coherent Non-Uniform Memory Architecture (ccNUMA) or Symmetric Multi-Processor (SMP) groups. Today, however, as designers are increasingly concerned with saving board real estate while improving the price/performance ratio of their offerings, conventional concepts of the multicore architecture have begun to change.
The latter was a topic of discussion at the Real-Time & Embedded Computing Conference (RTECC) in Phoenix, AZ this past December, where representatives from AMD and multicore programming startup Texas Multicore Technologies turned the conversation to heterogeneous processor architectures. Unlike traditional multicore, heterogeneous architectures look to unlock the power of diverse processors so that a pool of programmable, high-performance graphics processing elements can be integrated and applied to non-graphics data (this architecture is often in reference to combining a CPU and GPUs, but can be extended to a range of embedded processors).
Eventually the goal of this type of design is to combine the memory of the various regions of the processor into a unified model so that the communications overhead involved with copying data between CPU and GPU memory is removed. For now, however, this presents basic challenges for programmers, as a framework for simultaneously programming multiple processor resources with shared memory has not existed. In the Digital Signage section of this issue, Dave Jessel of AMD goes into further detail about heterogeneous processing, and Neil Trevett, President of the Khronos Group and VP of Mobile Content at NVIDIA, explains how the OpenCL framework can help ease heterogeneous programming pains. More information on heterogeneous processing can also be found through the Heterogeneous System Architecture (HSA) Foundation at, or the Khronos Group website at .
The power of knowledge
Something that can’t be stressed enough in the world of small form factors is power, particularly in more commercialized applications that often measure consumption on the scale of hundreds or thousands of units. Speaking on smart energy for the ZigBee feature in The Big Yet Small Picture section, Mark Grazier, ZigBee Alliance member and Program Manager, Wireless Sensor Networks and Worldwide Third Party Developer Network, Texas Instruments, emphasized the importance of educating consumers on the availability of power-saving technologies and the importance of energy consumption.
As he described to me a refrigerator magnet technology being developed that can indicate how much energy is being consumed by a household at a particular moment, Grazier stressed that it will take time for consumers to become aware of even simple, cost-effective technologies, and that some entity (in this case the utility provider) must assume the responsibility of spreading the word about useful new products. The point being that in the world of technology things move pretty fast, often too fast for end users to keep track of currently available products that can help them day in and day out.
Although the technologies we cover aren’t typically geared directly at the mass consumer market, PC/104 and Small Form Factors is in the business of making sure that the latest small form factor embedded tools and solutions are exposed to the engineers, product managers, and executives that can use them. Some of these are available in the 2013 Resource Guide beginning on page 26, including COM, SBC, I/O, Packaging, and Application Specific products.
In a final word on power, our summer issue will focus on low power solutions and medical systems. If you or your company are interested in contributing expertise on either topic, please feel free to send along a short abstract.
Brandon Lewis Associate Editor email@example.com