[Bug tree-optimization/17247] New: [3.5 Regression] TER is not always doing its job

["pinskia at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

I finnally was able to analyzed why gcc.dg/rs6000-power2-2.c was failing and noticed that TER was 
not doing its job in combining back T.1 = *(a+-1*8); into t = T.1;  why I don't know.
Here is the reduced testcase:
register double t1 __asm__("fr1");
void t(double *a, double *b)
{
        t1 = a[-1];
        b[-1] = t1;
}
Compile with "gcc -O3 -S -o - t.c -w" to see the problem.

Can also see the problem with the following source on any target if you look into the *.var file:
double t1;
void t(double *a, double *b)
{
        t1 = a[-1];
        b[-1] = t1;
}

-- 
           Summary: [3.5 Regression] TER is not always doing its job
           Product: gcc
           Version: 3.5.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: critical
          Priority: P2
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: pinskia at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
GCC target triplet: powerpc-*-*


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17247