Running the following code in R 4.1.1 gives different results between platforms.
set.seed(1)
x <- rnorm(3)[3]
print(x, 22)
From Intel Windows:
# 1234567890123456 (digits)
# -0.83562861241004716
From M1 Mac:
# 1234567890123456 (digits)
# -0.8356286124100471557341
I know the size of difference is below .Machine$double.eps
and the extra digits do not carry meaningful information.
I am not happy with the fact that extra digits exist. How can I ensure exactly consistent results? Is there an RNG library that achieves this?
round(x, 15)
does not work:
print(round(x, 15), 22)
# -0.83562861241004704 (Intel Windows)
# -0.8356286124100470447118 (M1 Mac)
Aucun commentaire:
Enregistrer un commentaire