fbpx

Vision: Promise & Challenge of Heterogeneous Processing

EE_Times-logo

By Vin Ratford
Executive Director, Embedded Vision Alliance

This blog post was originally published at EE Times' SoC Design Line. It is reprinted here with the permission of EE Times.

This week, guest blogger Vin Ratford returns to highlight the opportunities and challenges for heterogeneous processors. –Jeff

Over the last several years, improvements in digital chip manufacturing have failed to deliver the kind of gains in performance, energy efficiency, and cost effectiveness that we had previously grown accustomed to. In response, semiconductor suppliers have turned to multicore processors as the way forward. For the most demanding applications, heterogeneous multicore architectures often yield the best bang for the buck (or Watt). And these days, many of the most demanding applications are those involving visual intelligence — or as we like to call it, "embedded vision."

At the Embedded Vision Alliance, several of our member companies are helping to push the limits of processor performance and efficiency by developing new heterogeneous chips and platforms for vision-enabled systems. But new processor architectures alone aren't enough. On the contrary, as new processor architectures emerge, there's an increased need for application development infrastructure to make these processors easy to use by customers and partners throughout today's complex value chain. This infrastructure includes programming languages, libraries, tools, reference designs, development environments, and related training.

Historically, suppliers of specialized processors often tried to field all of the application development infrastructure required for their processors by themselves. But with the complexity of today's processors and applications, this is futile. It also typically leads to incompatible development flows among processors, making life difficult for system designers and application developers who want to be able to move from processor to processor as their requirements evolve.

Processor suppliers need a strong set of partners to field strong development environments for their processors. And processor users need partners to help them make good choices and make effective use of new technologies. To quote Hillary Clinton, "It takes a village." Delivering on the promise of heterogeneous processors requires suppliers and users to invest in education and development infrastructure — in addition to processors.

Individual companies have stepped in to help with the transition and to help with integrating new languages and approaches (such as OpenCL) into the existing software development flows, and universities are incorporating topics into their curriculums. But real, practical examples of how to develop systems and applications that leverage these technologies are few and far between. Even more importantly, there is little discussion about the tradeoffs among different heterogeneous platforms and the languages, tools, APIs, libraries and techniques used to program them. Our conclusion after many conversations with developers is that there is an unmet need for technical programs that educate, inform and assist the development community.

As a step in filling this gap, Alliance member company BDTI will be presenting a new workshop entitled Implementing Computer Vision and Embedded Vision: A Technical Introduction on May 28, in conjunction with the Embedded Vision Summit in Santa Clara, Calif. This workshop will present processor options for vision applications, and will introduce key tools and libraries for vision application development, including OpenCL and OpenCV.

Processor technology has improved our lives immeasurably over the past few decades. It will continue to do so in the next few decades — but to achieve this potential, large numbers of software developers will need to be able to harness the capabilities that the latest processors offer. This requires a massive training/retraining program and we are only at the beginning.

I personally find this to be an exciting time to help with this transition, and I believe that the work of the Alliance is more important than ever in helping to realize the potential of today’s most promising technologies. I invite you to join me at the next Embedded Vision Summit in May and become part of the village…

Here you’ll find a wealth of practical technical insights and expert advice to help you bring AI and visual intelligence into your products without flying blind.

Contact

Address

1646 N. California Blvd.,
Suite 360
Walnut Creek, CA 94596 USA

Phone
Phone: +1 (925) 954-1411
Scroll to Top