Graph coloring for register allocation?

[Michael Hayes]

Richard Henderson writes:
 > On Thu, Jan 25, 2001 at 11:53:29AM +1300, Michael Hayes wrote:
 > >  > What's
 > >  > left is making sure that copies from hard argument register to
 > >  > pseudo don't move from the start of the function.
 > > 
 > > Does this also handle the case where incoming args are in hard
 > > registers?
 > 
 > Er, no.  As I said, that exactly the case that's left to do.

That's what I thought but I've noticed that the interaction between
the scheduler and register allocator does not seem as bad as it used
to be for these incoming arguments.

Michael.