[Octopus-devel] cvs commit: marques/octopus/src eigen_arpack.F90 Makefile.am config.h.in eigen.F90 eigen_jdqz.F90

[cvsusers at tddft.org]

acastro     2005/06/02 16:22:03 WEST

  Modified files:
    octopus/src          Makefile.am config.h.in eigen.F90 
  Added files:
    octopus/src          eigen_arpack.F90 
  Removed files:
    octopus/src          eigen_jdqz.F90 
  Log:
  After some experimenting and reading, I found out that the Jacobi-Davidson
  scheme would in fact be slower than the Arnoldi/Lanczos iteration
  eigensolver. So I changed the dummy interface to JDQZ by an interface
  to the ARPACK library.
  For now, only the real symmetric case is implemented. Of course the idea
  is to put the complex non-Hermitian case, since for the normal case
  the conjugate gradients seems to be faster.
  
  Revision  Changes    Path
  1.151     +4 -4      marques/octopus/src/Makefile.am
  1.19      +3 -0      marques/octopus/src/config.h.in
  1.53      +11 -20    marques/octopus/src/eigen.F90