LAUREL, Md. Panelists at the Military Applications of Programmable Logic Devices conference took a look at tomorrow's market for field-programmable gate arrays (FPGAs) and decided that the parts have a future in system-on-a-chip designs and possibly even in consumer electronics.
Panelists agreed that the combination of ASIC cores and programmability features will be commonplace in future devices. But tool vendors stressed verification problems with system-on-chip designs. "Already in the ASIC world you have a one-to-one correspondence of verification engineers to design engineers," said Ray Salemi, senior applications engineer for Exemplar Logic Inc. (Framingham, Mass.).
Programmable logic could actually ease those problems if added between ASIC cores, "partly to act as glue and partly to act as band-aids," Salemi said.
"The FPGA design is going to become more ASIC-like," said Grason Curtis, technical account manager for Synplicity Inc. (Durham, N.C.). "We see a greater use of IP [intellectual property] in the overall flow, greater use of integration in the overall design. Things like floor planning become critical." But timing closure remains a problem, Curtis said.
"The question of how to validate systems-on-a-chip is not solved right now," Salemi said. "We'll see something come out in the next three to five years," Salemi said. "It's got to be there, or we just stop making big chips."
"You'll see these two- and three-million-gate devices," said Brian Cronquist, technology director for Actel Corp. (Sunnyvale, Calif.).
But some panelists questioned whether such large chips could be produced by the industry with reliability, particularly for military applications.
David Hepner, mechanical engineer with the Army Research Lab, noted that the military tends to use older technology, in part because it's proven to be reliable through years of field use. "Reliability comes with time. If you've planned obsolescence into your chip, you might have killed the perfect chip," Hepner said.
When it came to future applications for programmability, panelists turned their attention away from the military and towards high-volume consumer applications.
"I would guess that small satellites are not that killer app," Cronquist said. "It will come in the form of something that's consumer-oriented."
The trend toward customizable, personalized technology will call for programmability, panelists agreed. "Anywhere that that's possible maybe the elderly [market] is where you're going to make the big bucks," Hepner said.
The result will be like a "personalized mood ring," said Donald Bouldin, a professor with the University of Tennessee's Microelectronic Systems Group (Knoxville, Tenn.). "This kind of ubiquitous, IP-addressable Palm Pilot, or whatever, that has some kind of programmability, makes a lot of sense."
Programmable logic's future in such devices is debatable, however. "If there's anything close to a killer app, somebody's going to roll a custom chip for it," said Scott Hauck, assistant professor with the University of Washington. "It might be an ASIC with some programmability, but if it's a killer app, somebody's going to go to silicon. What you're really looking for [as a win for programmable logic] is killer behavior," he said.
Salemi disagreed. "What you want to do is take your killer chip and tweak it and tweak it," he said.
Both consumer and military applications are going to require lower power consumption than today's FPGAs. Conference co-organizer Richard Katz, of NASA Goddard Space Flight Center, pointed out that FPGAs distribute clock signals throughout the chip, resulting in unwanted power consumption. This trait drives some military designers to use asynchronous circuits, he said.
Actel's Cronquist dropped hints that his company's SRAM-based technology, due to be announced later this year, will reflect the tendencies to smaller size and lower power. "You'll see when the [Actel] reprogrammable technology is introduced that we don't have all the legacy," Cronquist said.
Asked about the growing interest in open-source licensing, panelists agreed that the open-source movement won't be sweeping FPGAs any time soon.
"There is no greater rivalry, not even Ford-Chevy, than Altera-Xilinx," said Synplicity's Curtis, a former Altera employee. "That is the heart and soul of the technology. That discussion has been around for years." With open-source, "you're taking away what differentiates that company," he said.
"I don't see the FPGA vendors themselves doing it, but I do see users doing what they've done in the Linux space and the GNU space, and sharing code," Salemi said.