On 2015-08-21 13:27:22 -0400, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> > On 2015-08-21 13:03:42 -0400, Tom Lane wrote:
> >> The behavior is well-defined, at least as long as we don't shift far
> >> enough to have integer overflow
>
> > Unfortunately not:
> > 5.8.2: The value of E1 << E2 is E1 left-shifted E2 bit positions;
> > vacated bits are zero-filled. If E1 has an unsigned type, the value of
> > the result is E1 � 2 E2 , reduced modulo one more than the maximum value
> > representable in the result type. Otherwise, if E1 has a signed type and
> > non-negative value, and E1 � 2 E2 is representable in the result type,
> > then that is the resulting value; >>otherwise, the behavior is undefined<<.
>
> [ rolls eyes... ] There isn't any machine in the world where the behavior
> isn't well-defined short of overflow.
> Why do the C spec authors persist in pretending otherwise?
Yea, it's way past time that C is redefined being based on 2-s
completement. And why this is declared undefined rather than
implementation defined is completely beyond me.
FWIW, icc apparently has been observed to recognize that a negative
value cannot be shifted and thus optimized based on the assumption that
the number is positive...
