[Bug tree-optimization/63569] New: [5.0 Regression] Wrong code with volatile and ICF

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63569

            Bug ID: 63569
           Summary: [5.0 Regression] Wrong code with volatile and ICF
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org

Take:
static int f(int t, int *a) __attribute__((noinline));

static int g(int t, volatile int *a) __attribute__((noinline));
static int g(int t, volatile int *a)
{
  int i;
  int tt = 0;
  for(i=0;i<t;i++)
    tt += *a;
  return tt;
}
static int f(int t, int *a)
{
  int i;
  int tt = 0;
  for(i=0;i<t;i++)
    tt += *a;
  return tt;
}


int h(int t, int *a)
{
  return f(t, a) + g(t, a);
}
--- CUT ---
The order of f and g are important.

The .icf dump has:
Semantic equality hit:f->g
Assembler symbol names:f->g

Which is incorrect as they are not semantic equivalent as one you can load from
a only once while the other you need to load a each time through the loop.