RISC-V / GCC 10: invalid operation not set in acos and asin

[Adhemerval Zanella]

On 13/05/2020 15:49, Joseph Myers wrote:
> On Wed, 13 May 2020, Aurelien Jarno wrote:
> 
>> I have tracked the issue to the following piece of code from
>> sysdeps/ieee754/dbl-64/e_asin.c:
>>
>> |   else {
>> |     u.i[HIGH_HALF]=0x7ff00000;
>> |     v.i[HIGH_HALF]=0x7ff00000;
>> |     u.i[LOW_HALF]=0;
>> |     v.i[LOW_HALF]=0;
>> |     return u.x/v.x;
>> |   }
>>
>> GCC 10 optimizes out that code and loads a NaN value from memory instead
>> of doing the division, which explains why the invalid operation
>> exception is not set.
> 
> So that should be filed as a regression bug in GCC.  If fixed quickly we 
> can then put something on the per-release wiki pages about the fix for GCC 
> bug N (bug present in 10.1, fixed in 10.2) being needed to avoid those 
> failures.
> 

Is is an invalid optimization? Because:

double foo (double x)
{
  x = NAN;
  x = x / x;
  return x;
}

is also optimized away even with gcc-9 (and math_force_eval
or math_opt_barrier does not really help).