[Pw_forum] parallelization issue

Giacomo Saielli giacomo.saielli at unipd.it
Fri Oct 15 10:50:40 CEST 2004


Dear developers,
I wonder if the comments of paragraph 7.4 of espresso manual v 2.1, 
parallelization issue of pw.x, apply also for the cp.x module.
This is because I found the md simulations of my system with cp.x to be 
not as fast as I need it.
I have a system made of  242 atoms, 137 of which are protons, the rest 
C, N, O. This is the relevant input section (I use Vanderbilt PPs):

&system
 ibrav = 14,

    celldm(1) = 27.5945666, 
    celldm(2) = 1.0, 
    celldm(3) = 1.0, 
    celldm(4) = 0.0, 
    celldm(5) = 0.0, 
    celldm(6) = 0.0, 
    nat = 242,
    ntyp = 4,
    nbnd = 316,
    nelec = 632,
    ecutwfc = 30.0,
    ecutrho = 180.0,
    nr1b= 10, nr2b = 10, nr3b = 10,
    xc_type = 'PBE',
 /
 &electrons
    emass = 400.d0,
    emass_cutoff = 2.5d0,
    orthogonalization = 'ortho',
    ortho_eps = 5.d-8,
    ortho_max = 15,
    electron_dynamics = 'verlet',
!    electron_damping = 0.4,
!    electron_velocities = 'zero',
    electron_temperature = 'not_controlled',
 /
 &ions
    ion_dynamics = 'verlet',
!    ion_velocities = 'zero',
    ion_temperature = 'nose',
    fnosep = 60.0,
    tempw = 300.0,
 /

A test CPMD run with cp.x takes about 50 hours of total cpu (16 procs on IBM SP4 at CINECA, all PE of the same node) for 100 steps, that is about 15 fs (dt is 6 a.u. about 0.145 fs). This makes inpossible to run the simulation for some tens of picoseconds as I would like, since it would take something of the order of 10^5 cpu hours.

The question is if this numbers sounds correct to you, for this system size, so that the only chance is to build a smaller system, or if there is space for improvement playing with the choice of kpoints grid and other grids as mentioned in the Espresso manual,concerning pw.x. 

Here I attach the beginning of the output file

thank you very much
best wishes
Giacomo

************************************************************************
****                                                                ****
****  CPV: variable-cell Car-Parrinello molecular dynamics          ****
****  using ultrasoft Vanderbilt pseudopotentials - v.2.1           ****
****                                                                ****
************************************************************************

     Parallel version (MPI)
     Number of processors in use:     16
Warning: card  &CELL ignored
Warning: card     CELL_DYNAMICS = 'NONE', ignored
Warning: card     CELL_VELOCITIES = 'ZERO', ignored
Warning: card     PRESS = 0.0D0, ignored
Warning: card     WMASS = 70000.0D0 ignored
Warning: card  / ignored
reading ppot for species #  1 from file /scratch/padpdb42/saielli/PWSCF/TMA/CP/O.pbe-van_bm.UPF
reading ppot for species #  2 from file /scratch/padpdb42/saielli/PWSCF/TMA/CP/N.pbe-van_bm.UPF
reading ppot for species #  3 from file /scratch/padpdb42/saielli/PWSCF/TMA/CP/C.pbe-van_bm.UPF
reading ppot for species #  4 from file /scratch/padpdb42/saielli/PWSCF/TMA/CP/H.pbe-van_bm.UPF
   wmass (calculated) =       233945.31



 nbeg=  1 nomore=    100    iprint=   2
 reads from 51 writes on 51
 time step =    6.0000

 parameters for electron dynamics:
 emass=     400.00  emaec=       2.50ry
 orthog. with lagrange multipliers: eps=  0.50E-07 max= 15
 verlet algorithm for electron dynamics
 with friction frice =  0.0000 , grease =  1.0000


 ions are allowed to move
 ion dynamics with fricp =  0.0000 and greasp =  1.0000
 ion dynamics with nose` temp. control:
 temperature required= 300.00000(kelvin) nose` mass =  16599.975


 internal stress tensor calculated
 cell parameters are not allowed to move


 iprsta =  1

  unit vectors of full simulation cell
 in real space:                         in reciprocal space:
   27.5946    0.0000    0.0000              1.0000    0.0000    0.0000
    0.0000   27.5946    0.0000              0.0000    1.0000    0.0000
    0.0000    0.0000   27.5946              0.0000    0.0000    1.0000
 Proc  planes cols    G   planes cols    G    columns  G
       (dense grid)     (smooth grid)   (wavefct grid)
  1      8    680  53548    7    456  29172  114   3634
  2      8    680  53548    7    456  29172  114   3634
  3      8    680  53548    7    456  29172  114   3630
  4      8    680  53548    7    456  29152  114   3630
  5      8    681  53549    7    455  29151  113   3629
  6      8    680  53548    7    456  29168  114   3630
  7      8    682  53554    7    456  29132  114   3630
  8      8    682  53554    7    456  29132  114   3630
  9      7    682  53554    7    454  29126  112   3628
 10      7    682  53554    7    454  29118  112   3628
 11      7    682  53554    7    454  29126  112   3628
 12      7    682  53550    7    456  29160  116   3632
 13      7    682  53550    7    456  29148  112   3628
 14      7    682  53550    7    456  29156  112   3628
 15      7    682  53550    6    456  29168  112   3628
 16      7    682  53550    6    456  29164  114   3630
  0    120  10901 856809  110   7289 466417 1813  58077
 
 ggen:  # of g vectors < gcut   ng=  26774
 ggen:  # of g vectors < gcuts ngs=  14586
 ggen:  # of g vectors < gcutw ngw=   1817
 ggen:  # of g shells  < gcut  ngl=   2655
 
 
 unit vectors of box grid cell
 in real space:                         in reciprocal space:
    2.2995    0.0000    0.0000              1.0000    0.0000    0.0000
    0.0000    2.2995    0.0000              0.0000    1.0000    0.0000
    0.0000    0.0000    2.2995              0.0000    0.0000    1.0000
 
 ggenb: # of gb vectors < gcutb ngb =    243
 ggenb: # of gb shells  < gcutb ngbl=     22
 initialization 

 ibrav= 14 alat= 27.595 omega=21012.1616
 gcut= 3471.84    gcuts= 2314.56 gcutw=  578.64
 k-points: nkpt= 1

 meshes:
  dense grid: nr1 ,nr2, nr3  =  120 120 120  nr1x, nr2x, nr3x =  120 120 120
 smooth grid: nr1s,nr2s,nr3s =  110 110 110  nr1sx,nr2sx,nr3sx=  110 110 110
    box grid: nr1b,nr2b,nr3b =   10  10  10  nr1bx,nr2bx,nr3bx=   10  10  10

 exchange-correlation potential:  SLA  PW   PBE  PBE 

 ecutw=   30.0 ryd    ecuts=  120.0 ryd    ecut=  180.0 ryd

 # of electrons=  632 # of states=  316


 nspin= 1
 occupation numbers:
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0

 # of atomic species    4
 is=  1
  na=  25  atomic mass= 16.00 gaussian rcmax=  0.50
 atomic coordinates:
       5.34  27.52   7.83
       0.95  14.73   1.13
       7.35  10.64  21.18
      19.98  18.54   1.18
       2.28   5.20   1.24
.....
.....
.....
.....





-- 
Giacomo Saielli
Istituto per la Tecnologia delle Membrane del CNR, Sezione di Padova;
Via Marzolo, 1 - 35131 Padova, Italy; 
Tel: +39-049-8275279; Fax: -5239;
http://www.chfi.unipd.it/home/g.saielli/




More information about the users mailing list