By Jack Ganssle
Five Technologies You Need to Know About
Published 5/15/2006
An article on TechWeb (http://www.techweb.com/wire/showArticle.jhtml?articleID=159500002) tries to prognosticate into the near future about technologies that will affect "every technophile." The list:
Sweet. Yet none of those will impact the embedded systems world anytime soon. Intel did make a big push for multi-core CPUs for embedded apps at the latest Embedded Systems Conference, but I just don't see that happening in any significant way for a long time to come. Their low-end Core Duo CPU gobbles 15 watts at about 1 volt; the others suck 31 watts (see http://developer.intel.com/design/intarch/coreduo/index.htm?iid=ipp_embed+proc_coreduo&). That's almost 30 amps, which will drain an AA in a jiffy. Applications connected to the limitless power from a wall socket will have a significant cooling problem. Here in Maryland electric rates will go up 72% in July so consumers may soon start making buying decision at least partly based on power consumption.
Ajax, holographic storage and AMD's Virtualization technology had exactly zero presence at the embedded show. Certainly flash technology is important to us, but not, as yet, as a hard disk replacement for any but the most extreme embedded apps. Flash won't be a suitable disk replacement till costs tumble, yet this week Samsung and Hynix both raised their prices for NAND flash (http://eetimes.com/news/semi/showArticle.jhtml?articleID=187201352).
About 100 CPUs go into embedded systems for every desktop processor shipped. The computer business is, for all intents and purposes, the embedded market. Yet it's a stealth business; prognosticators only see the web and the insatiable craving for more PC performance.
It is fun to project ahead a few years; here are some technologies and practices that I think will be important to embedded developers.
FPGA hardware and software. Programmable devices continue to offer more capacity per dollar. This plethora of cheap transistors means manufacturers are adding both hard and soft CPU cores to their offerings. Third party vendors will sell huge amounts of IP. New tools will result in a new kind of embedded design that unifies hardware and software engineering - to some extent, at least. Compilers will translate C to hardware and hardware to software to best optimize each application's performance. Altera already has this technology.
The open source movement will continue to grow, of course, but with the release of GPL 3.0 a business-driven backlash will spawn new licensing arrangements. GPL 3.0, though still in draft form, is creating considerable controversy due to its strong stand on DRM and patents. Without getting into the debate about software patents, companies will continue to view these as important assets. The recording and movie industries will fill politicians' purses with campaign contributions to require hardware DRM enforcement in home entertainment devices. Licenses more generous than GPL 3.0 or even 2.0 will proliferate to satisfy business needs to use OSS while protecting proprietary code. An extant example is that for eCos (http://ecos.sourceware.org/license-overview.html).
Demand will continue to increase for lower-power processors and systems. More exotic sleep and current reduction modes will appear. Intel and AMD will continue to push power-hungry chips into fewer and fewer niche apps while most developers manage nanowatts. New analysis tools will help identify code chunks that can be optimized to save watt-hours. Many small systems will scavenge power from the environment.
Mesh networking will finally take off, enabled by these scavenging systems. Wires for connectivity and power will seem quaint as users scatter systems like high-tech Johnny Appleseeds.
That's four predictions. What's your fifth?