------- Comment #10 from amacleod at redhat dot com 2008-09-04 16:51 ------- As long as it removes any dead PHIs, it would be sufficient. Other types of dead statements don't have 'unexpected' side effects across basic block boundaries, and should be handled fine. Other than being a waste of effort of course :-)
A little further thinking leads me to think dead PHI cycles could not produce this problem. The PHI results would get conflicts since they are live out their basic blocks. It the lack of any use in this case that is the problem. But removing a PHI with no uses can easily have ripple effects in that it may have been the only use of some other PHI, etc, etc. So we're back to DCE type work anyway. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37102
