TODO LIST -  5 February 2004

INSTALLATION/PORTING

- expand new configure to cover as many configurations as possible

- Linux Alpha ?

- Make.irix ?

COMMON

- fix shift in eigenvalues between CP/FPMD and PW
  The eigenvalues must be the same!!!

- fix shift in total energy between CP, FPMD and PW
  for charged systems. Total energies for charged systems
  must be the same!!!

- merge (standard) diagonalization routines
  use parallel diagonalization in CP and PW

- file format unification must be finished

- merge spherical harmonics, integration routines, etc
  spherical bessel: add routine directly calculating dj_l(x)/dx,
  verify limits x=>0, fix special cases x=>0 into the codes

- variable declarations : real(kind=DP), or real(dbl), everywhere

- real vs dreal, cmplx vs dcmplx, dfloat etc

- calculation of <psi|exp(iGr)|psi'>

- process numbering: 1 to nproc or 0 to nproc-1 ?

PW

- wavefunctions extrapolation fails when some new band comes in ( and 
  some other goes out ) during the ionic evolution. This imply that
  <psi(t-dt)|psi(t)> = 0 for this "new" band ( tipically one the highest 
  unoccupied band ) and the alignment of the subspaces is no longer possible.
  In this case the history of this new band should be erased and no 
  extrapolation should be performed

- extend wavefunctions extrapolation to ionic relaxation and to NEB 
  calculations. subroutine find_alpha_and_beta() should be moved from 
  dynamics.f90 to update_pot.f90 and relative informations should be written 
  on a new file

- remove iswitch, replace with existing (minus_q, lscf)
  or new logical variables

- the assignment of correct weigths to k-points must be done
  in a more general and less obscure way (needed for better
  support of noncolinear stuff)

- maximize merge of colinear (PW) and with noncolinear (PWNC) code

- replace "use pwcom" into more "use" statement

- add traceback in error (error_handler module)

- remove analytical PPs and related variables

- remove residual direct calls to MPI routines,
  use (or merge with) existing routines in mp.f90 instead

- remove all calls to setv in PH and D3; all calls to zero and SSUM in
  CPV; calls to level-1 blas copy, scal

- remove potential mixing, save and start from rho instead of V,
  at least for scf calculations. For non-scf calculations: we may
  still start from V, but the output file MUST include the scf rho,
  not a bogus rho as it may happen now

- save rho and/or V in G-space instead of R-space
  check that what is read is sensible

- remove old punch, use new punch, add possibility to save
  wavefunctions either 'distributed' or 'collected' in
  parallel execution (requires identifier in the first line
  of the file) 

- add possibility to read atomic positions from file

- Use more standard choices for crystal axis

- fix problem with j'_l(x)Y_lm(x) for x -> 0 and l=1

- ultrasoft PP: import "small boxes" from CP 

- make hpsi/spsi/newd/addusdens/CG/DIIS  faster
  - remove complex factor i**l from beta fct and q(r) (preliminary)
  - shift structure factor from beta to psi when computing becp (reduce memory)
  - use real BLAS routine instead of COMPLEX one in hpsi/spsi (at least 2 times 
    faster).
  - use only half of the G's when computing real integrals (2 times faster)
  - seek for CG and DIIS algoritms that only use (H-eS)|psi> and not
    the two vector separately ... compute it in one single call. (this way
    S|psi> is  inexpensive)

- output should be more informative and less confused

- the various *bfgs, dynamics, vcsmd should be rewritten in a more
  uniform way with respect to what they print and how the parallel
  case is treated. Always write at the end the lattice parameters.

- write explicitly and unambiguosly the units of output quantities

POSTPROCESSING

- postprocessing/pwtools in the parallel case
  and with new_punch must be verified

- stm_wfc_matching not working

- dos, projected dos, etc: input data should be more uniform

- the dynamical matrix written by the phonon code does not contain
  the information on lattice basis vectors if ibrav=0.
  As a consequence most postprocessing codes will not work 

- reconstruction of all-electron wavefunctions

PH

- Tone: ntyp in input needed for phonon GUI ?

- better algorithm for electron-phonon (Malgorzata)

DOCUMENTATION

- examples for some features are still missing
  automated testing procedure: how to deal with numerical differences?

- expand and update the manual; add a list of FAQ, or of already
  answered questions

- complete the developer's (technical) guide

FPMD:

- Documentation is nonexistent

CPV: 

- Documentation needs serious improvements

- check on input ipp and pp type. Even better: remove ipp
  (leave either UPF or old-format Vanderbilt)

- check on small box size

- replace calls to zero, ssum, csum, blas copy, scal, with f90 commands