prev FPGA 2002 done!
Conclusions

  1. created new analytic techniques to design switch blocks

  2. checkerboards increase diversity

  3. commutative design simplifies analysis

  4. create diverse switch blocks

  5. diversity admirable goal (Wilton works well)
    • maximizing diversity does not improve routability/area
    • difficult for existing CAD algorithms
    • new CAD algorithms needed?

  6. can save 3-4% area with different switch blocks
    • by reducing flexibility (# of connections on endpoint)
    • by adding diversity
    • by other topological changes, eg using checkerboard
    • is this the limit?