[Q] liveness analysis of executable code

Hi,

I'm doing some research on interprocedural liveness analysis of executable
code right now. As far as I can tell the literature on such analyses
mainly seems to originate from the University of Arizona and from
Digital/Compaq/Intel (the OM and Spike groups).

Does anyone know of any other literature concerning liveness analysis of
executable code? More specifically, I'd like to know how many registers on
average are typically live/dead in a program. Statically or dynamically.

Thanks,

Bjorn De Sutter
  Postdoctoral Fellow of the Fund for Scientific Research -- Flanders
  affiliated with Ghent University http://www.elis.ugent.be/~brdsutte/