I noticed that phy2logquad had a lot of overhead. Trying to reducing it
by simplifying calls to fPsi, dPsi and such.
The function for fPsi has been made so no memory is allocated and works
under the assumption that the input array has the right size (1:numNodes)