/*! Precomputation of the 316 interactions by evaluation of the matrix kernel on the uniform grid and transformation into Fourier space.
PB: Compute() does not belong to the M2LHandler like it does in the Chebyshev kernel. This allows much nicer specialization of the M2LHandler class with respect to the homogeneity of the kernel of interaction like in the ChebyshevSym kernel.*/
/*! Precomputation of the 316 interactions by evaluation of the matrix kernel on the uniform grid and transformation into Fourier space. These interactions can be tensorial (of size ncmp) and are computed blockwise