The “relative” function
EXPRL is useful for accurately computing
ex ‑ 1 near
x = 0. Computing
ex ‑ 1 using
EXP(
X)
‑ 1 near
x = 0 is subject to large cancellation errors.
Similarly,
ALNREL can be used to accurately compute ln(
x + 1) near
x = 0. Using the routine
ALOG to compute ln(
x + 1) near
x = 0 is subject to large cancellation errors in the computation of 1 +
X.