INT_MIN and compiler diagnostic

[David Thompson]

(Ones' complement systems are rare these days, in part because
people are more interested in getting the wrong answer as fast as
possible than in getting the right answer. )

How so? I haven't heard of any case where 1sC gives a 'right' answer
that 2sC doesn't, for integers. (For floating point, the sign of zero
sometimes matters, and TTBOMK all common formats are S&M and do
preserve it; certainly the standardized one does.)

And it was my understanding that in earlier days a main advantage of
1sC was that it was faster for negate. It was only when MSI (and
higher) made carry-lookahead effectively free (in both cost and time)
that 2sC won out.

 

[Chris Torek]

How so? I haven't heard of any case where 1sC gives a 'right' answer
that 2sC doesn't, for integers.

The one usual "good case" here is when the bit pattern for -0 is
used as a trap representation, and "uninitialized" variables of
signed integral types are set to this bit pattern. But the remark
was meant more as a jab at the usual obsession with speed, often
to the exclusion of all else, including accuracy, hence the "".