Re: [PATCH 07/28] powerpc: trunc/truncf refactor

["Gabriel F. T. Gomes" <gabriel@inconstante.eti.br>]

On Fri, Mar 29 2019, Adhemerval Zanella wrote:
> 
> 	* sysdeps/powerpc/fpu/trunc_to_integer.h (set_truncing_mode): Add
> 	 TRUNC handling.

Similar to floor and round, remember to add a note here about the
definition of TRUNC.

> --- a/sysdeps/powerpc/fpu/round_to_integer.h
> +++ b/sysdeps/powerpc/fpu/round_to_integer.h
> @@ -38,6 +38,7 @@ set_fenv_mode (enum round_mode mode)
>    {
>    case CEIL:  rmode = FE_UPWARD; break;
>    case FLOOR: rmode = FE_DOWNWARD; break;
> +  case TRUNC:
>    case ROUND: rmode = FE_TOWARDZERO; break;
>    default:    rmode = FE_TONEAREST; break;
>    }

Likewise, remember to add the definition of TRUNC.

The patch looks good to me with these changes.

Reviewed-by: Gabriel F. T. Gomes <gabriel@inconstante.eti.br>

> +double
> +__trunc (double x)
> +{
> +#ifdef _ARCH_PWR5X
> +  return __builtin_trunc (x);
> +#else
> +  return round_to_integer_double (TRUNC, x);
> +#endif
> +}

OK.  Arch check looks correct.

> -/* double [fp1] trunc (double x [fp1])
> -   IEEE 1003.1 trunc function.  IEEE specifies "trunc to the integer
> -   value, in floating format, nearest to but no larger in magnitude
> -   then the argument."
> -   We set "round toward Zero" mode and trunc by adding +-2**52 then
> -   subtracting +-2**52.  */

This comment got lost.  Would you deem it appropriate to reinstate it
somewhere in round_to_integer_double (you would need to make it generic
so that it fits the four rounding directions (so far), as well as
nearbyint from the subsequent patch... perhaps something along the lines
of the paragraph below)?

  IEEE 1003.1 nearest integer functions.  IEEE specifies several
  functions that approximate a floating-point to a nearby integer value.
  For each of them we set the appropriate rounding mode, then add and
  subtract +-2**52.