How We See Embedded Processing

Exactly 10 years ago, I read a cover article on the Scientific American (June 1997) titled "The microchip that rewires itself". The authors noted, "Computers that modify their hardware circuits as they operate are opening a new era in computer design. Because they can filter data rapidly, they excel at pattern recognition, image processing and encryption."

I was learning software development at the time using 8086 assembly, BASIC/C on DOS, etc.  The prospect for a "new era in computer design" was too exciting to ignore, and I took the leap into hardware design with FPGAs. With a low cost development board from the Xilinx University Program, my friend and our professor at Anna University, India set out to create a simple Reconfigurable Architecture Kit that would help even software engineers to learn about programmable logic technology. Although we didn't succeed then, I'm glad to be developing kits today that provide engineers with a hands-on experience in FPGA-based SoC development.

The Scientific American article went on to talk about a Hybrid-Architecture Computer combining a general-purpose microprocessor and FPGAs called the Dynamic Instruction Set Computing (DISC) being developed by Mike Wirthlin at Brigham Young University. I joined that group next and started working on an early ESL tool called JHDL (Java as a Hardware Description Language), which was another great effort in bringing the hardware and software worlds together.

Then around 2001, when the first FPGAs with embedded processor cores were released, I joined the MicroBlaze processor team at Xilinx, developing system design tools to assemble embedded systems quickly. One of the things we quickly realized was that the having a fully flexible hardware platform was not sufficient. So, we started developing tools to make the handoff to software easier. My focus has specifically been in the design and debug of hardware/software interfaces.

So, here we are now. The worldview of 'processing' is in transition. The nerve center of innovation has clearly moved from processor microarchitecture to system architecture. The discussion is no longer about RISC vs CISC and processor clock frequency but rather about multicore architectures and hybrid architectures based on relevant combinations of processor and DSP, GPU or FPGAs. But one thing is for certain – the hardware platform that addresses the next generation software challenges best will be the one to watch.

I look forward to discussions in this forum about challenges on the hardware-software boundary. Please stay tuned to my first topical post—" Tower of Babel"—on some terminology challenges.

—Navanee, Xilinx