[Cfp-interest] Three round nearest modes and FLT_ROUNDS

[Jim Thomas]

My understanding is that 754-2019 does not intend for roundTiesToZero to be a new user selectable rounding direction. C 7.31.4 already reserves the names with an FE_ prefix, so the names would be available if a future revision of 754 added a user selectable rounding direction. It wouldn’t be appropriate for C to get in front of 754 on this.

- Jim Thomas

> On Oct 23, 2018, at 8:46 AM, Fred J. Tydeman <tydeman at tybor.com> wrote:
> 
> The current draft of 754 has three different rounding modes for midway
> 
> ties to even (for binary and decimal)
> ties away from zero (for decimal; optional for binary)
> ties toward zero (for augmented operations; not required for any other operations)
> 
> WG14 paper N2124 is adding FE_TONEARESTFROMZERO to C
> 
> Should we be adding something like:
>  FE_TONEARESTTOZERO
>  FE_DEC_TONEARESTTOZERO
> to C?
> 
> Given that there could be three different meanings for Round to Nearest,
> what should FLT_ROUNDS be?  Should FLT_ROUNDS have a value 
> for each of the possible FE_* rounding macros?
> 
> C has (in 7.6.1)
> 216) Even though the rounding direction macros might expand to 
> constants corresponding to the values of FLT_ROUNDS, they
> are not required to do so.
> 
> F.2 has:
> The value of FLT_ROUNDS applies to all IEC 60559 types 
> supported by the implementation, but need not apply to non-IEC 60559 types.
> 
> F.3 has:
> The fegetround and fesetround functions in <fenv.h> provide the facility to select among
> the IEC 60559 directed rounding modes represented by the rounding direction macros in
> <fenv.h> (FE_TONEAREST, FE_UPWARD, FE_DOWNWARD, FE_TOWARDZERO)
> and the values 0, 1 , 2, and 3 of FLT_ROUNDS are the IEC 60559 directed rounding modes.
> 
> 
> ---
> Fred J. Tydeman        Tydeman Consulting
> tydeman at tybor.com      Testing, numerics, programming
> +1 (702) 608-6093      Vice-chair of PL22.11 (ANSI "C")
> Sample C99+FPCE tests: http://www.tybor.com
> Savers sleep well, investors eat well, spenders work forever.
> 
> _______________________________________________
> Cfp-interest mailing list
> Cfp-interest at oakapple.net
> http://mailman.oakapple.net/mailman/listinfo/cfp-interest