EDP 2007

• In electronic design there are the large conventions, such as DAC (Design Automation Conference) and DATE (Design Automation and Test in Europe.) There are the medium size shows: IC CAD, ASP-DAC, etc. Finally, there are the small discipline specific IEEE conferences: ISPD, VLSI Test, EDP, and others. These small shows are the working forums for future tools, technologies and methodologies. The shows may be small, but they are extremely influential. EDP (Electronic Design Processes Workshop) is the conference for design methodology and there have been instances of large semiconductor companies changing their research and, in some cases, methodology based on what was said at an EDP conference.
  
  
  EDP 2007, as usual, showcased the most pressing problems in electronic design today:
  
  • Power
  • DFM
  • Multi-core programmability
  
  
  Leon Stok of IBM stressed simplifying the DFM flow for the design engineers. Patrick Groeneveld of Magma Design stressed designing tools for their position in the design flow rather than looking at them as a standalone tool. RDR (Restrictive Design Rules) were brought up again and again, even from the User Community, as a must for future silicon nodes.
  
  
  On the software side the impact of Amdahl’s Law was a major topic for more than a few of the presenters. JoAnn Paul of Virginia Tech argued that Amdahl was wrong. At least, Amdahl’s law was being applied incorrectly in heterogeneous design environments. Optimization techniques used in a homogeneous, Von Neumann computing environment, when used the same way in a heterogeneous computing environment, are likely to decrease system performance not increase it. Patrick Madden of SUNY talked about the last 40 years of Amdahl’s Law and all of the failed attempts to circumvent it. Something that I hadn’t considered was not only does the sequential software programming infrastructure need replacing, but that a sequential mindset has driven us to develop sequential algorithms. Patrick argued that until we concentrate on developing concurrent algorithms our efforts in developing a concurrent software programming environment will be wasted.
  
  
  These small shows are not for marketing, but are a must attend kind of show if you are a R&D engineer or you are responsible for your company’s CAD environment.
  
  
  Gary Smith