IV-OPTs causes huge regression

[David Edelsohn]

	Zdenek has submitted a patch to fix an oversight in the original
ivopts patch that considers "important" candidates:

http://gcc.gnu.org/ml/gcc-patches/2004-09/msg02937.html

Some colleagues observed related performance problems.  The patch might
help the SPEC95 regressions.

	The patch thread is awaiting a response from Zdenek regarding the
logical bitmap operation used to prune the candidates:

http://gcc.gnu.org/ml/gcc-patches/2004-09/msg03085.html

David