[Bug tree-optimization/87987] Missed optimization with ranged-for loop on a constexpr array

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87987

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
So I think LLVM's unrolling notices the load will become constant and the whole
induction variable basically goes away for sum.

Even for this C code LLVM does the optimization:

static const int table[] = { 1,10,10, 1, 1,10, 1,10 , 1,10,10, 1, 1,10, 1,10 ,
1,10,10, 1, 1,10, 1,10 , 1,10,10, 1, 1,10, 1,10 };
int check1()
  {
    int sum = 1;
    for (unsigned long i = 0; i < sizeof(table)/sizeof(table[0]);i++)
      sum += table[i];
    return sum;
  }