Re: P1788.1/M001.01 Missing roundTiesToEven/roundTiesToAway
On Wed, 7 Oct 2015 04:57:16 -0700 (PDT), Dima wrote:
> So I suggest to keep regularity and do not omit roundTiesToEven
> and roundTiesToAway from the basic standard.
I agree, and it seems to me to be even simpler than Dima suggests.
Overflow is no concern since large magnitudes are integers already.
The general assumption for the basic standard is that there is an
underlying IEEE 754-2008 implementation (otherwise why be tied to that
representation?) -- so there should be no trouble for roundTiesToEven.
For roundTiesToAway, which is optional for Binary in 754-2008, a sign
test is needed to decide whether to add or subtract 0.5, with rounding
towards zero.
Of course, if the underlying system only supports 64-bit integers...
Michel.
---Sent: 2015-10-07 13:25:16 UTC