There are lots of edge cases where the sign of zero affects computation. In floating-point, zero values represent not only the exact value zero but the result o

There are lots of edge cases where the sign of zero affects computation. In floating-point, zero values represent not only the exact value zero but the result of an underflowed computation too small to represent as a non-zero floating point value, so retaining the sign is often important to at least keep such computations on the correct side of a branch cut or limit. For example: x/+0 = +infinity x/-0 = -infinity sqrt(+0) = 0 sqrt(-0) = nan log(+0) = -infinity log(-0) = nan