Our research group is focused on examining end-to-end solutions to the problem of turning massive quantities of VLSI resources into useful computational artifacts. We're interested in the following sorts of questions:

• What will be the "killer apps" for emerging parallel architectures like multi-core and tiled?
  
  
• How do we modify the architecture to make parallelism easier to use?
  
  
• What programming constructs and compiler techniques will enable users to more easily program parallel hardware?
  
  
• How can we leverage specialization in processors to attack the recent "power wall"
  which has stagnating processor frequencies?