[RFC] Implementing detection of saturation and rounding arithmetic

[Liu Hao]

在 5/12/21 5:13 PM, Tamar Christina via Gcc 写道:
> 
> int f (int a, int b)
> {
>      int res;
>      if (__builtin_add_overflow (a, b, &res))
>        {
>            if (res >= 0)
>              return INT_MAX;
>            else
>              return INT_MIN;
>        }
>      return res;
> }
> 
> Should be recognized as saturating as well.  But yeah, following the same approach we
> would rewrite the sequence to something like res = .ADD_SAT (a, b);
> 

Is this a correct saturating addition implementation?

If the addition has overflowed, you get a positive result or zero for the sum of two negative 
numbers (or a negative one for two positive numbers); and it is not straightforward to write it this 
way.


This should be

   int f (int a, int b)
   {
     int res;
     if (__builtin_add_overflow (a, b, &res))
       {
         if (a >= 0)      /* changed from `res` to `a`  */
           return INT_MAX;
         else
           return INT_MIN;
       }
     return res;
   }

which can be optimized further as

   int f (int a, int b)
   {
     int res;
     if (__builtin_add_overflow (a, b, &res))
       res = (a >> sizeof(int) * CHAR_BIT - 1) ^ INT_MAX;
     return res;
   }


-- 
Best regards,
Liu Hao

-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 840 bytes
Desc: OpenPGP digital signature
URL: <https://gcc.gnu.org/pipermail/gcc/attachments/20210512/489cafc6/attachment.sig>