Musings on the IEEE HPEC Conference

Maybe it’s a degree of cynicism that comes with age, but it frequently occurs to me that there is little that is truly new in our particular niche of the technology world.

For example, there has been a lot of writing in the press over the past couple of years about this new-fangled “HPEC” stuff and how to define it. Even within GE, I have had discussions with people who are convinced that we invented something when we formed our HPEC Center Of Excellence.

To set this into its proper perspective, you need look no further than the HPEC Conference that was held this month. The last three events have been held underHPEC conference the auspices of the Institute of Electrical and Electronics Engineers, and the “E” of HPEC has been recast as “Extreme” rather than its predecessor, “Embedded,” opening it up to other endeavors of hard problems like exploitation of Big Data.

However, its history goes back to at least 1996, which was probably my first encounter with the term High Performance Embedded Computing. I don’t remember what we called it before that. “Computing,” probably. MIT Lincoln Laboratory recognized the growing importance of this field and inaugurated a yearly workshop/conference to bring together people from industry and academia to discuss developments and needs.

I attended the conference, as I have done for many years. As always, I found it entertaining and instructive. However, afterwards I found it interesting to review the agendas of some previous years’ events to see how far we have come.

Ten years ago, we were discussing how to use FPGAs and maybe even these new-fangled GPU things to accelerate radar processing. Today—well, pretty much the same. OK, the performance of the devices has moved on significantly (but nothing that Mr. Moore didn’t predict through process technology evolution).

And the tools have moved on. Then, we had Matlab to FPGA; now, we have C to FPGA. Then, we had OpenGL which meant we had to turn a signal processing problem on its head to fit a render pipeline; now, we have CUDA. Nice, but they do little more than make life a little easier.

Nine years ago we were enthralled by how the Cell processor was going to change our world. Today—well, amazingly enough there is still discussion of the vagaries of programming the Cell even though it is dead and buried as a viable technology for HPEC and has been for several years.

HPEC in the fieldWe talk of novel memory designs that accelerate a specific application. I remember programming a CD&A system (remember them, anyone?) for a contract I had with a Japanese manufacturer to do CT X-ray examination of engine blocks to detect non-visible cracks and flaws. That system had a memory subsystem specifically designed to extract arbitrary 2D planes from a 3D volume of data at huge speed. I’d call that a novel memory architecture.

I guess what I’m asking is, where is the “holy-crap-that’s-amazing” quantum leap in capability that would push us forward in a truly meaningful manner? Following the Intel / Freescale / ATI / NVIDIA roadmap ain’t it. Not that those advances aren’t nice to have—but they follow the curve, not a step function.

I have to say I was humbled by the challenges faced by a speaker from NASA who was working on autonomous navigation for remote exploration vehicles. Next time I find myself thinking how hard it is to shoehorn that GMTI processor into the space available on a small unmanned vehicle, I will have to ground myself by remembering how they have to provide enough compute horsepower to do autonomous nav. With a power budget of 10W. And it has to run all the time, every time—no field service on Mars. And it has to survive all kinds of radiation that would fry standard electronics in seconds. Plus of course, no GPS up there. Kinda puts it into perspective, doesn’t it?

I manage an over-50s soccer team, and each week I do a tongue-in-cheek post-match report. I have taken to awarding “Curmudgeon of the Week” to the member of our team who displayed the most unsportsmanlike behavior on the field. My daughter approves, as she once described our games as a bunch of old guys running around a field swearing at each other. Somehow, it feels like I should get the HPEC curmudgeon award this week. I wonder if that comes with a cold beer like the soccer games do?

IEEE image credit

Peter Thompson

Peter Thompson is senior business development manager for High Performance Embedded Computing. He first started working on High Performance Embedded Computing systems when a 1 MFLOP machine was enough to give him a hernia while carrying it from the parking lot to a customer’s lab. He is now very happy to have 27,000 times more compute power in his phone, which weighs considerably less.

More Posts