[tree-ssa] New regressions as of 2003-11-04

[law@redhat.com]

In message <20031105145858.GS13705@kam.mff.cuni.cz>, Jan Hubicka writes:
 >We can do it that way, but the function should be mostly NOOP now.
 >cleanup_cfg+dead code removal should've caught all cases previously.
It's definitely not a NOP.  Have you actually compared the tree dumps
with and without this pass?

Also don't forget it's now used before we build the CFG to clean up obviously 
useless junk.

 >The point is that we *want* to create temporary registers so unless
 >there is good reason to do so, we don't want single register to be set
 >two times.  That confuses RTL.
More correctly, having a register set multiple times can inhibit certain
optimizations.  Just as an example, if a register is set more than once
we're a lot less aggressive about marking it as a pointer.


jeff