We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
The improved rounding model can be rewritten as
rnd_b(x) = p_2(x, b) * eps(b)
where p_2(x, b) = p_2(x) if |x| = 0 or |x| is in the normal range and p_2(x, b) = min_pos_float(b) / eps(b) when |x| is subnormal (and not 0).
p_2(x, b) = p_2(x)
|x| = 0
|x|
p_2(x, b) = min_pos_float(b) / eps(b)
0
Activity