DATE 2007

• DATE (Design, Automation and Test in Europe) was extremely successful this year. The most obvious success was the location. Both Nice and the conference facilities more than made up for the loss of Paris as a venue. But far more important was the fact that they have finally succeeded in turning DATE into a true Electronic System Design Conference.
  
  
  Europe has been always a systems-oriented design environment. During the initial efforts at designing the 3G phone, Europe became aware of the major negative impact that embedded Software was having on these new systems. That grew into the realization that it had become impossible to address electronic design without including the Embedded Software Engineers in the discussion. That was not lost on the DATE committee, for they realized that their conference would shrink, or worse become irrelevant, if they were unable to attract the entire design community, both hardware and software to their show. After years of trying, this year they succeeded.
  
  
  As I prepare for DAC each year, I call up my User clients and ask them what are the most difficult design issues that they have faced during the year. Usually I get three or four “Hot” items and then I concentrate on those topics on my What To See @ DAC list. This year I discovered five “Hot” issues: Software, Software, Software, DFM and Power.
  
  
  The first software issue, and the most pressing, was Amdahl’s Law’s limitation on using more than four microprocessors in a general purpose computing environment. The key here is “general purpose.” We’ve been using parallel processing to address Embarrassingly Parallel Problems for quite some time now. However in today’s Von Neumann-based, sequential programming environment, Amdahl’s Law pretty much spells out the end of parallel processing for general purpose computers at four homogeneous processors. That is the Homogeneous multi-processing problem.
  
  
  The second issue is the Heterogeneous Processor problem. Cell phones are usually mentioned in this context. Can we do it ... yes. Can we do it efficiently, the answer is decidedly no. Most designs today are based on Padded Cell architectures, which were borrowed from the security community. Basically each application becomes its own stand alone function, its own processor, its own memory, its own I/O, etc. There are some minor improvements on the concept but all in all these designs are pretty ugly. This is not the world of optimized resources. In the world of cell phones, the power problem demands optimization.
  
  
  The third software problem is hardware/software partitioning. The European 3G phone efforts proved the point; 80ish percent of the power optimization “must” be accomplished at the initial HW/SW partitioning. Ties in nicely with the fifth hot issue doesn’t it?
  
  
  At panel session after panel session this year, the questions from the audience were on the three software issues. This year’s attendees were a good mix of hardware and software engineers. The software crisis was on everyone’s minds. DATE has become a truly Electronic System Design Conference.
  
  
  Gary Smith