[DFTB-Plus-User] Running dftb+ 1.1 on Ibm sp6
Giovanni La Penna
glapenna at iccom.cnr.it
Tue Apr 27 11:06:27 CEST 2010
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
============================================================
More information about the DFTB-Plus-User
mailing list