2008年是嵌入式系统会议(ESC)二十年庆典,ESC已经成为工业界的“风向标”。现在的问题是,所有的设计都转向嵌入式设计,那么未来20年内ESC的角色将会出现怎样的变化。

嵌入式系统的边界已经变得愈加模糊。实际上消费者所依赖的每一个部件背后都隐藏着嵌入式系统。众所周知台式机是基于微处理器运行的,然而大量的嵌入式器件已经成为当前应用的灵魂,并且控制这些器件也更加困难。

数位ESC主题演讲者达成了这样一个共识:如果20年前处理器的2%是用于台式机应用,98%是用于嵌入式设计,那么从今往后20年,这一比例会变为1:99。“在未来20年,我们希望获得更多的处理能力,”Gilder Publishing的技术分析师Nick Tredennick这样说:“唯一的不同将是每一个嵌入式器件都将基于T比特的带宽,而不是现在的吉比特。”

Tredennick长期的工业界经验颇具导向意义。在Motorola时,他设计的微处理器成为最初Apple Macintosh笔记本的核心引擎。他还曾是Altera的首席科学家,并由于他在微处理器设计方面的贡献被任命为IEEE资深会员。“我们在2008年将会看到一个一万亿美元的半导体市场,从实际应用角度来看,则都是嵌入式系统,”Tredennick在ESC上这样介绍。“所以我并不确信‘嵌入式’这个词是否要继续保留。”

这与Dinkumwave Ltd.的总裁P.J. Plauger的观点一致,他通过视频对主题演讲作了点评。Plauger的公司主要向其客户授权使用标准-正统的C和C++库以及他开发的在线文档。“在过去几年, C和C++已经成为标准的开发语言,”他说。“没有它们,我们就不会有嵌入式产业。”

Plauger和下一位ESC主题演讲人,发明第一个商用实时OS的Jim Ready都认为,当今及未来五年内嵌入式设计人员面临的最大挑战是多核处理器的软件开发难题。

“1999年我们建立了MontaVista Software,编排了嵌入式Linux商业应用的目录,但当时没人相信Linux操作系统会应用到嵌入式系统市场,”Ready这样介绍。他预计,通向多核处理器的道路还会在类似的疑问中铺就,嵌入式设计人员最终会消除这种疑问。

在视频主题演讲中,C++的设计者和最初制定人Bjarne Stroustrup向C++的标准化工作表达了敬意。C++已经成为公认的开发语言。“我的目标是设计出可以有效并优美编程的语言,”Stroustrup说。“很多语言都逼迫你只能进行二选一。”

目前Stoustrup在Texas A&M University执教,并致力于并行和分布式编程的研究。他对C++的多核应用略带遗憾:“我们正在寻找程序转换的方法,可以支持分布式计算、优化和嵌入式系统编程。”

在另一个视频演讲中,来自未来学会,但处在休假期的Paul Saffo这样预测:“我们在驶向一个‘smartifacts’的世界；也就是说我们所有的行动都受到隐藏的嵌入式处理器的引导,这将影响我们生活的每一个方面。”与其他主题演讲者意见一致,Saffo也认为“将要启动的多核技术将会对开发智能嵌入式系统产生最大的影响。”

他建议ESC的听众,在寻找多核并行编程方案过程中遇到无法解决的难题时,要留心过去的经验。“历史不会重演,”他引用马克吐温的名言,但“历史确有其规律。”

翻页查看英文原文:

ESC keynoters: Soon, all design will be embedded design

With 2008 marking the 20th anniversary of the Embedded Systems Conference, ESC has become an industry "essential." The question is how its role might shift in the next 20 years as virtually all design becomes embedded design.

Merely identifying embedded systems has become harder. There's an embedded system hidden under the hood of almost every gadget on which consumers rely. While desktop machines are based on well-known microprocessors, it's tougher to get a handle on the myriad embedded devices that form the soul of today's applications.

Several ESC keynote speakers arrived at a common conclusion: If 20 years ago 2 percent of processors were dedicated to desktop applications and 98 percent to embedded designs, 20 years from now that ratio will be 1:99. "We'll want more and more of the processing power in 20 years," said Nick Tredennick, technology analyst for Gilder Publishing. "The only difference will be that everything embedded will be terabyte-based, instead of today's gigabyte world."

Tredennick's long industry experience is a bellwether. While at Motorola, he designed the microprocessor that became the central engine for the original Apple Macintosh. He was once chief scientist at Altera, and he was named an IEEE fellow for his contributions to microprocessor design. "We'll see a $1 trillion semiconductor market in 2028, and for all practical purposes it will all be embedded," Tredennick told his ESC audience. "So I'm not sure if the 'embedded' nomenclature will survive."

That would suit Dinkumware Ltd. president P.J. Plauger, who delivered keynote remarks via video. Plauger's company licenses standards-conforming C and C++ libraries and online documentation that he developed. "Over the years, C and C++ have become the standard development languages," he said. "Without them, we would not have an embedded industry."

Plauger and fellow ESC keynoter Jim Ready, creator of the first commercial real-time OS, said software development hurdles for multicore processors are the biggest challenge embedded designers face today and in the next five years.

"When we invented the category of embedded Linux commercialization in 1999 by founding MontaVista Software, nobody believed the Linux operating system could be applied to the embedded systems market," said Ready. He predicted that the road to multicore processing is paved with similar skepticism that can and will be disspelled by embedded desigers.

In a video keynote, Bjarne Stroustrup, designer and original implementer of C++, hailed the standardization of C++ as a recognized development language. "My aim was to design a language in which I could write programs that were both efficient and elegant," Stroustrup said. "Many languages force you to choose between those two alternatives."

With regard to multicore applications for C++, Stroustrup, who teaches at Texas A&M University and is engaged in research into parallel and distributed programming, said, "We're looking at ways of using program transformation to support distributed computing, optimization and embedded systems programming."

In another video presentation, Paul Saffo, who is on a research sabbatical from the Institute for the Future, offered this prognostication: "We are moving toward a world of 'smartifacts'; that is, all our actions will be guided by embedded processors hidden from us but influencing every aspect of our lives." Like other keynoters, Saffo said he believes that "the arrival of multicore technologies will have the biggest impact on developing smart embedded systems."

He advised his ESC audience to heed the lessons of the past when looking for solutions to seemingly insurmountable hurdles such as multicore parallel processing. "History doesn't repeat itself," he said, citing an aphorism commonly attributed to Mark Twain, but "it sometimes rhymes."