On Thu, 28 Apr 2005, Virginie Quequet wrote: > In previous version of PWSCF (1.2.0), it was possible to choose > diagonalization = david_nooverlap. I wonder why it is no longer possible, > do this method cause particular problem?
no, but it didn't have any advantage on davidson diagonalization with overlap, that works also for ultrasoft pseudopotentials, so finally we retained only one version (the difference is whether you generate a subspace with orthonormalized correction vectors and the current estimated wavefunctions, and diagonalize with conventional diagonalization; or whether you use unnormalized correction vectors instead and diagonalize with an overlap matrix) Paolo
