Autoincrement pass

[Michael Hayes]

Joern Rennecke writes:
 > > I decided that the current situation was getting rather messy and have
 > > written a separate pass that it is run before lifetime analysis.  This
 > 
 > So you probably don't know what registers are actually life at any given
 > point.  This could lead to spurious decisions about using auto-increment.

I pondered about this situation but cannot think of a case where it
would be a problem.  OK, I may generate a spurious autoincrement for
an insn that may be dead but the lifetime analysis pass will then
delete it.  If you have some testcases that I can try, please send
them.

The alternative is to run the autoincrement pass between the first and 
second passes of lifetime analysis.

Michael.