optimization/6007: cfg cleanup tremendous performance hog with -O1

[Jan Hubicka]

> On Mon, Mar 25, 2002 at 09:41:09AM +0100, Jan Hubicka wrote:
> > The graph is already n^2 in size and I need to find matching outgoing edges
> > running tiny loop as n^3 that probably makes difference.
> 
> We already look at the complexity of the cfg to decide whether or
> not to run gcse.  No reason we can't do the same for crossjump.

Another posibility is to give up the search after constant amount of edges, so
we will still crossjump in trivial parts of function and ignore the expensive
bits.

I will try whether this works well.

Honza
> 
> 
> r~