[DFTB-Plus-User] Running dftb+ 1.1 on Ibm sp6

[Giovanni La Penna]

Dear DFTB+ users and developers,

  In the following makefile there are directives for
compiling dftb+ 1.1 for 32-cores Ibm-sp6 nodes:

### Start makefile
# -*- makefile -*-
############################################################################
# System dependent Makefile options for
# AIX, IBM xlf compiler (version 3.1)
############################################################################

# Fortran 90 compiler
# cf. http://www.nersc.gov/nusers/resources/software/ibm/xlf.php
# Hints from Jump (fz-juelich.de) documentation on PowerPC6
FC90 = xlf95_r -qsuffix=f=f90 # -qsuppress=cmpmsg

# Optimization flags for the Fortran 90 compiler
FC90OPT = -O3 -qsmp=omp -qstrict -qarch=pwr6 -qtune=pwr6

# Preprocessor (leave empty, if the compiler has a built in preprocessor)
CPP = cpp

# Options for preprocessing
CPPOPT = -WF,-DDEBUG=$(DEBUG)

# Postprocessing of the preprocessor output (add-on pipe)
CPPPOST = $(ROOT)/utils/fpp/fpp.sh noln2

# Linker
LN = $(FC90)

# Linker options
LNOPT =

# Override options for different DEBUG modes
ifeq ($(DEBUG),1)
     FC90OPT = -g
endif
ifeq ($(DEBUG),2)
     FC90OPT = -g
endif
ifeq ($(DEBUG),3)
     FC90OPT = -g -C
endif

# Library options in general
# module load lapack
LIBOPT = -L${LAPACK_LIB}

# How to link specific libraries

# Standard Lapack
#LIB_BLAS   =  -lblas
#LIB_LAPACK = -llapack -lxlsmp

#lapack/esslsmp
#LIB_BLAS   =  -lblas
LIB_LAPACK = -llapack -lesslsmp -lxlsmp

### End makefile

The Ibm esslsmp library apparently works. This library
is the Ibm equivalent of Atlas or Mkl within Intel/Amd
based architectures.
The following input

### Start input
# Zn2+ 2[Abeta(11-16)] in a water box
# Second dimer
# System: 1204 atoms, 223 solute + 327 water molecules
# super-cell: 21.976 x 27.947 x 22.881 Ang^3
# This run: minimization

Geometry = GenFormat {
  <<< "min.gen"
}

Driver = ConjugateGradient {
   MovedAtoms = 1:-1
   MaxForceComponent = 1.e-03
   MaxSteps = 100
   OutputPrefix = "min2"
   AppendGeometries = Yes
}

Hamiltonian = DFTB {
   SCC = Yes
   SCCTolerance = 1.e-06
   MaxSCCIterations = 200
   Mixer = Broyden {
   }
   SlaterKosterFiles = Type2FileNames {
     Prefix = "your_path_to_znorg-0-1_skffiles"
     Separator = "-"
     Suffix = ".skf"
   }
   MaxAngularMomentum = {
     N = "p"
     O = "p"
     C = "p"
     H = "s"
     Zn = "d"
   }
   Charge = 0.
   ReadInitialCharges = No
   SpinPolarisation = {}
   Filling = Fermi {
     Temperature [Kelvin] = 0.0
   }
   EwaldParameter = 0.
   KPointsAndWeights = SupercellFolding {
    1 0 0
    0 1 0
    0 0 1
    0.0 0.0 0.0
   }
}

Options = {}

ParserOptions = {
   ParserVersion = 4
}
### End input

produces the following timing
(output of time unix command):

Using 16 threads
real    111m6.618s
user    349m17.847s
sys     11m46.250s

So it makes 100 scc cicles (each of about 10 steps)
in about 100 minutes, with about 1 minute per cicle.
I hope the input corresponds to a periodic system
(in the input gen-file there is S and the cell at the end),
but with K=0 (so-called gamma-point in usual PW-DFT
calculations).

Every suggestion is welcome for increasing performance
or correcting the input (for comparisons with gamma-point
PW-DFT calculations (quantum-espresso.org)).

Best regards,
                   Giovanni La Penna

============================================================
Giovanni La Penna - National research council (Cnr)
Institute for chemistry of organo-metallic compounds (Iccom)
via Madonna del Piano 10,
I-50019 Sesto Fiorentino, Firenze, Italy
tel.: +39 055 522-5264, fax: +39 055 522-5203
e-mail: glapenna at iccom.cnr.it - http://www.iccom.cnr.it/lapenna
skype: giovannilapenna
============================================================