The most recent Embedded Systems Conference Spring, in Chicago, again demonstrated the size and depth of our industry. Two hundred vendors filled the exhibit hall at McCormick Place, showing their niftiest goodies. Many thousands of attendees crowded the show floor, picking up the latest embedded buzz as well as the numerous freebies always available at the booths.
The Embedded show is always much more than just an exhibition; at its core it's an educational program. A decade ago, the West Coast conference originated as numerous seminars augmented by a handful of tabletop vendor displays. That philosophy still reigns. In Chicago this year, nearly 100 classes covered all aspects of embedded development, including real-time UML, using an RTOS, creating predictable real-time designs, code inspections, debugging ISRs, and so on. Once again the TCP/IP seminars were very popular, as developers struggle to incorporate connectivity into their designs.
A couple of themes seemed to pervade the conference. First, not surprisingly, was Internet connectivity. Though I remain somewhat skeptical about the utility of an Internet-aware toaster, the picture painted by many vendors was one of total appliance awareness.
emWare (www.emware.com) — which sells an interesting, though non-standard, connectivity solution for even small 8051-style processors — showed dozens of small single-board computers connected to the 'Net. One corner of the booth was a very cool complete kitchen — microwave, fridge, counter, and sink. I hoped in vain for some tasty morsels prepared by a bevy of microprocessor droids, but instead got more of a brain-than-stomach fill from their engineering VP.
The refrigerator's door sports a flat panel touch sensitive display linked via emWare's software to the 'Net. Pull a package of frozen chicken out of the freezer and the appliance's bar code scanner identifies the item and sends a request to the 'Net for information about the product. In seconds, dire warnings appear: “This food item has high cholesterol and an excess of calories.” As you place the food in the microwave, it, too, looks for information online, this time coming up with cooking recommendations which automatically set the microwave's timer to a suggested value.
Useful? I dunno. Cool? You bet. More importantly, the appliances showed a glimmer of the future touted by this and numerous other vendors. Soon, the thinking goes, everything will be smart and 'Net-connected. We'll be surrounded by incessant information processing, our every action mediated by a microprocessor, almost every thought transmitted instantly to colleagues and friends.
I don't doubt that all sorts of products will indeed be foisted off on us as “must-have” gadgets. I remain somewhat skeptical of the utility of many of these ideas. Other 'Net-connected products — not at the show but clearly possible — do intrigue me. How about glasses (for those of us optically challenged) that project encyclopedic amounts of data to us? Or why not integrate a calculator with our cell phones, so we have one less thing to carry?
As our capacity for technology increases, wise people will assess the cost/benefit ratio of each new capability. Gimmick or useful tool? How will a 'Net-connected toaster improve our lives? Contrary to common belief, the Amish — famous rejecters of anything modern — are not so much Luddites as just very careful consumers of technology. According to a fascinating story in Wired (www.wired.com/wired/archive/7.01/amish.html), Amish leaders look for the impact of new products on their social and religious lives. Phones, for instance, encourage interruptions that tear at the fabric of their close family lives. Isn't it amazing how quickly we abandon our family and friends at the first ring of the phone? When I was little, my dad never allowed us to answer the phone during dinner. At the time I thought he was odd. Now I see the wisdom in not allowing this particular bit of technology to be our master.
Wired, the source of this article on the Amish, is itself an example of misplaced capabilities. All of those bright colors and words buried in pictures hurt my middle-aged eyes. I just can't read it. Ironically the online version is more graphically traditional and much easier to read.
The Wind River show
Since summer, WRS has acquired Dr. Design, a West Coast design house; TakeFive Software, authors of the SNiFF+ source code analyzer; Diab (compilers); Software Development Systems (another compiler/debugger vendor); and Integrated Systems Inc. (who brought us pSOS).
Prior to the acquisition, WRS and ISI were the two largest embedded tool companies. Now, with some $300 million in combined revenue they dwarf all others in sales, profits, and show floor space.
The WRS/ISI deal puzzled many commentators since VxWorks and pSOS traditionally compete head-to-head. Why buy a direct competitor? Some wags suggested that eliminating pSOS frees VxWorks to become the dominant OS, creating a powerful barrier to Microsoft's CE or even to embedded Linux. In fact, their latest ad's tag line states, “Roll up your Windows.” Others figure it's a way to get a large base of smart employees in a tough labor market.
Regardless, WRS now owns most of the RTOS market, many of the more popular 32-bit compilers, and SingleStep, which might be the most common debugger of all for 32-bit Motorola processors.
ISI customers may be concerned about the future of pSOS. Happily Wind River briefed the press about their product roadmap for both pSOS and VxWorks (the press release is available at www.windriver.com/press/html/roadmap.html). They intend to offer one more release of VxWorks (code named “Cirrus”) and one of pSOS (“Stratus”), both in the third quarter of 2000. In 2001 the two products will merge into a new RTOS called “Cumulus” (I see clouds on the horizon), which will maintain compatibility with the pSOS API.
At the show, Wind River announced the acquisition of Embedded Support Tools (www.estc.com), for 6.4 million shares of stock. That's about a third of a billion dollars for a $28 million dollar outfit, one whose reported profit has never exceeded $3 million. Not a bad deal — for EST. WRS's current $2 billion market capitalization, though, means they're flush with buying power.
Why EST? This company, too, specializes primarily in tools for 32-bit Motorola processors. Their BDM and JTAG hardware debuggers, as well as board support packages, are a nice hardware complement to WRS's software offerings. It will be interesting to see how Wind River deals with EST's software debugger, which competes directly with the SingleStep debugger WRS got in buying ISI/SDS/DIAB.
WRS is still shopping, so more mergers are in the offing.
Big vs. small
The panel held a wide ranging debate on the value of open source to the embedded community. Under repeated and focused questions from the audience, each of the panel's vendors made it clear, though, that open source is a vehicle to build companies by charging developers for their products and services. Their very clear message: don't confuse “open source” with “free.”
For years we've seen the big vendors retreating from 8- and 16-bit operating systems and tools. The Linux craze, too, is a 32-bit venue, one that's creating opportunities and headaches for these high-end companies. Yet 8- and 16-bit processors and their toolchains continue to thrive. For example, CMX (www.cmx.com) showed their beta MicroNet TCP/IP stack with web server that runs on most processors, burning a meager 12K on an 8051. While not 100% RFC-compliant, the product handles most embedded networking chores adequately. CMX's primary business is real-time OSes for deeply embedded systems. I was delighted to find them handing out a complete price list with their datasheets, a refreshing change from more common complex pricing models. In fact, in a recent discussion about build vs. buy issues with RTOSes, one developer told me they elected to build their own when salesmen from two competing vendors left the developers baffled about costs.
The 8-bit world grew with the addition of two new Z80 derivatives. Zilog (www.zilog.com) showed their eZ80 (a very clever name) while Rabbit Semiconductor (www.rabbitsemiconductor.com) touted the Rabbit 2000. The eZ80 offers binary compatibility with the Z80, yet extends the address space to 16Mbits and cranks the clock to 80MHz. The on-board MAC offers DSP-like performance for some applications.
The Rabbit is more an extension of the Z180 architecture than of the Z80, though it does not offer binary or even true source compatibility. A few Z80 instructions were dropped to pack most instructions into a single byte, while new C-friendly opcodes bring the CPU into the modern era.Perhaps the best feature is the addition of interrupt levels, which allows easy implementation of critical handlers inside of ISRs and operating systems. The Z80's primitive interrupt structure required an awful lot of work to create true reentrant code.
Both Zilog and Rabbit offer their own proprietary tools, a move which seems a throwback to the early embedded days. Perhaps their rationale parallels that of the open source dream: getting cheap tools into developers' hands. At $99 for Zilog's compiler/debugger and $139 for Rabbit's development board with the Dynamic C environment it's pretty hard to complain about pricing. Yet I wouldn't be surprised to see these semiconductor providers contracting with third-party tool companies to give potential customers more options.
The Chicago show brought the decline of in-circuit emulators into focus for the first time. I saw this in the vendors who were not there, some of whom failed in the last year (Orion and Pentica). It was apparent in how chip vendors, including Rabbit, showed ICE-less debuggers. I think WRI's valuation of EST is a warning to all hardware tool vendors, as EST primarily sells BDM/JTAG hardware. It appears emulators for high-end CPUs will wither in favor of the much cheaper BDMs. Perhaps low-end ICEs will survive, especially as many 8-bit CPUs just don't support serial debug ports. But even there we're seeing change, as in Triscend's 8031 E5 core, which contains a complete serial debugging environment.
Eight- and 16-bit compiler and debugger sales, though, are stronger than ever. In this 32-bit world it ain't easy getting much respect in eight bits, yet Keil (www.keil.com) told me growth in their 8051 and 167 tool sales is phenomenal. IAR (www.iar.com), which provides compilers and debuggers across a wide spectrum of small and large processors, showed their very cool visualSTATE tool that creates code for state machines that you design graphically, on screen.
Virtual hardware design is nothing new; for a decade or more large programmable logic devices let designers create “hardware” designs using tools that closely parallel our firmware-creating compilers. Each year, however, finds more convergence as hardware design requires ever less soldering and more programming. Last year ARC Cores (www.arccores.com) introduced their synthesizable CPU, encouraging customers to modify the instruction set to suit particular needs. Scary stuff. Then Triscend (www.triscend.com) demonstrated that a CPU core surrounded by an FPGA gives engineers customizable virtual peripherals. This show brought Tensilica (www.tensilica.com) to the forefront with their Xtensa 32-bit synthesizable RISC processor.
Tensilica's only product is the intellectual property behind the core and the supporting tools. They don't make chips. At 24,000 gates the CPU is pretty tiny, eating up only 0.7 square millimeters in .18-micron geometry. If you'd like to build an ASIC with the processor you'll pay them a six-figure licensing fee plus royalties. Alternatively there's a $64,000 “binary only” deal in case you'd like to dump the core into a very large FPGA. Though the fees keep these processors out of the reach of low-volume products, it's clearly an indicator of things to come.
Another parallel with software, in this case the open source movement, was not at the show. The OpenRISC 1000, soon to be available at www.opencores.org, will be a freely downloadable 32-bit RISC processor. Created by students in Slovenia, it's awfully hard to see what the ultimate success of this effort will be. But pundits said the same thing about Linux not too many years ago.
Finally, I was struck by the number of outsourcing companies with booths at this year's show. The shortage of engineers, more complex products, and shorter time-to-market pressures have created a land-office business for these hired guns. The startup WebPRN (www.webprn.com) is so virtual they provide neither products nor product development; instead their mission is to electronically connect companies looking for developers with these outsourcing businesses.
Stellcom (www.stellcom.com) told me they've started a new division whose charter is simply to help prospective customers convert raw product ideas into complete business plans and specifications. Apparently the dot-com mania spawns many businesses with little more than an idea and capital, with no idea how to build a business around the concept.
On another note, I recently obtained a Write-Only Memory chip. Signetics created this “product” 25 years ago to celebrate April Fool's Day. Since I don't expect to be creating products that need write once, read never capability, I'm holding a contest to give away the part and its hilarious datasheet. See www.ganssle.com.
Jack G. Ganssle is a lecturer and consultant on embedded development issues. He conducts seminars on embedded systems and helps companies with their embedded challenges. He founded two companies specializing in embedded systems. Contact him at .