[Pw_forum] Problems with nscf diago with many bands

Conor Hogan conor.hogan at roma2.infn.it
Tue Sep 30 11:57:08 CEST 2008


Dear forum,
I'm doing a slab calculation, and for reasons I wont explain, I need to
obtain the nscf KS states for many many unoccupied bands. There are 96
occupied bands: using the davidson I can diagonalize ok for 500 bands and
1000 bands, but if I push it up to 1500 I get problems with (3.2.3):
     from cdiaghg : error #      2684
     info =/= 0
or (4.0):
     from  pzpotrf  : error #        94
      problems computing cholesky decomposition

I fear the conjugate gradient will be very slow, so before I try that, any
suggestions for either approach with so many bands?

I'm using the system compiled pw.x on BCX.
Cheers,
Conor

 &system
      ecutwfc =12.0,
      nbnd=1500,
 /
 &electrons
    diago_thr_init=5.0e-6,
    diago_full_acc=.true.
    electron_maxstep = 70,
    diagonalization='david'
    mixing_mode = 'plain'
    mixing_beta = 0.7
    conv_thr =  1.0d-8,
 /



----

Dr. Conor Hogan                                --- -----  ---   ---
Dipartimento di Fisica e CNR-INFM             |      :   |     :
Universita' di Roma "Tor Vergata"             |--    :    ---  |---
Tel: +39 06 7259 4121 *NEW*                   |      :       | :
Fax: +39 06 2023507                            ---   :    ---  :
http://www.fisica.uniroma2.it/~cmtheo-group/  European Theoretical
                                              Spectroscopy Facility
                                              http://www.etsf.eu

Q. What do you call a fish with no eyes?
A. A fsh.   (Tommy Cooper)

The early bird gets the worm, but the second mouse gets the cheese. - S.
Wright

If you go through a lot of hammers each month, I don't think it necessarily
means you're a hard worker. It may just mean that you have a lot to learn
about proper hammer maintenance - J. Handey


More information about the Pw_forum mailing list