[Bug target/81647] New: inconsistent LTGT behavior at different optimization levels on AArch64.

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81647

            Bug ID: 81647
           Summary: inconsistent LTGT behavior at different optimization
                    levels on AArch64.
           Product: gcc
           Version: 8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: amker at gcc dot gnu.org
  Target Milestone: ---

Given below test from Richard S,
#include <fenv.h>

double x[16], y[16];
int res[16];

int
main (void)
{
  for (int i = 0; i < 16; ++i)
    {
      x[i] = __builtin_nan ("");
      y[i] = i;
    }
  asm volatile ("" ::: "memory");
  feclearexcept (FE_ALL_EXCEPT);
  for (int i = 0; i < 16; ++i)
    res[i] = __builtin_islessgreater (x[i], y[i]);
  asm volatile ("" ::: "memory");
  return fetestexcept (FE_ALL_EXCEPT) != 0;
}
If we compile and run it with:
$ ./g++ -O3 test.cc -o test.O3.exe
$ ./test.O3.exe
$ echo $?
1

If we compile and run it with:

$ ./g++ -O2 test.cc -o test.O2.exe
$ ./test.O2.exe
$ echo $?
0

It is because of different implementation of LTGT w/o vectorization.