The ranges of float-to-integer conversions appear to be inconsistent with all approximation schemes except RoundToNegInf. For example, consider RoundToZero, which rounds
255.5 to 255
-255.5 to -255
Yet, despite this, -256.5 does not round to -256, but instead causes NegOverflow.
The ranges of float-to-integer conversions appear to be inconsistent with all approximation schemes except
RoundToNegInf
. For example, considerRoundToZero
, which rounds255.5
to255
-255.5
to-255
Yet, despite this,
-256.5
does not round to-256
, but instead causesNegOverflow
.