The logarithmic rescaling to handle the "overflow times underflow" problem of the Hermite functions is making it really expensive to compute. However, it's mostly the exponential causing the problem by underflowing already for oder n=0 while the polynomial will rise to overflow more slowly.
Thus, it could be beneficial to not rescale for every order, but maybe every 10-th or 20-th order which would significantly reduce the computational burden.
The logarithmic rescaling to handle the "overflow times underflow" problem of the Hermite functions is making it really expensive to compute. However, it's mostly the exponential causing the problem by underflowing already for oder
n=0while the polynomial will rise to overflow more slowly. Thus, it could be beneficial to not rescale for every order, but maybe every 10-th or 20-th order which would significantly reduce the computational burden.