Lithodriven Detailed Routing

1
Litho-driven Detailed Routing

• Minsik Cho (thyeros_at_cerc.utexas.edu)
• Joydeep, Peng
• Dept. of ECE, University of Texas at Austin

2
Problem Definition

• Assumptions
• Global routing and Track routing are finished.
• Any wire longer than G-cell side is permanently
  embedded in DRC compatible manner (remaining
  wires are completely inside a G-cell).
• Every G-cell doesnt have litho-hotspot (they
  dont need any litho-driven optimization)
• Fast OPC and litho simulator are available.
• Goal
• Find detailed routing (final connecting) in each
  G-cell such that printability can be improved in
  DRC compatible manner.

3
Challenges in Litho-driven routing

• The sum of light intensity on a spot is not
  decomposable.
• OPC can play an important role.
• A routing after OPC can be better than estimated
  during routing.
• Litho simulation with OPC is very expensive for
  large area.
• Two major papers
• Li-Das DAC04 paper
• Joydeeps DAC05 paper

Need to break a problem into small litho
independent problems
4
Proposed Approach

• Mainly based on graph
• Each unrouted wire -gt vertex
• Between wires -gt edge
• Edge is NOT inserted
• Two wires are fairly apart(600nm?)
• Separated by long prerouted wires (from track
  routing)
• Cluster vertices such that each cluster can be
  afforded by OPC/litho simulater.
• Prerouted some wires if needed.

A
B
Preroute this
5
Proposed Approach

• For each clustered edges, enumerate possible
  routing solutions.
• Each solution will be evaluated by OPC/litho
  simulator.
• Edge sharpness will be measured.
• The best routing with OPC features will be
  picked.
• Questions to be answered
• Which wire will be prerouted if needed?
• Minimum impact on flexibility
• Providing maximum independency between vertices
• How small the cluster should be?
• Depends on OPC algorithm and litho simulator
• How to enumerate possible routing solutions
• Heuristic to pick good ones selectively