Libraries Description
The
libraries contain routines which are independent of the rest of
the package, e.g. they do not use the include linking the main
programs. They thus have a special status and are kept together
in the $MCTDH_DIR/source/lib directory.
When changing a library routine, try to avoid changing an
interface. This could lead to unforeseen problems in a
program.
The following libraries are employed by the MCTDH program:
lapack: Contains all the LAPACK routines
that MCTDH uses. This is the reference implementation from
NETLIB.
blas: Contains all the BLAS routines that
MCTDH uses. Again, this is the reference implementation from
NETLIB.
la: Contains all other linear algebra
calculations, such as matrix multiplication or determination of
eigenvalues.
ode: Contains all integrators, e.g.
Bulirsch-Stoer and Lanczos.
num: Contains all numerical operations
(e.g. FFT, special functions, ...).
sys: Contains routines concerning the
machine and the operating system and the input of data.
The libraries are automatically compiled as
required when a program is compiled. It is also possible to
compile the libraries separately using the general compile
script. Typing
will compile all the library archives. A single archive can
also be compiled, e.g.
will compile the NUM library.
One useful feature of the compilation procedure is that it is
possible to generate debug code for a program with optimised
libraries, e.g.
compiles a debug version (mctdh<ver>d) of the MCTDH
program with optimised libraries. This is useful as the library
routines need a large percentage of the computer time, and this
speeds up the process of debugging. To compile a program with
debug information also for the library, use the -D option.
For more details on the $MCTDH_DIR/bin/compile script, see Compiling the
programs.
This is not the complete LAPACK library, but only those
selected routines from NETLIB's reference implementation which are needed by
MCTDH and auxiliary programs.
Like above, this is not the complete
BLAS library, but only those selected routines from NETLIB's reference
implementation which are needed by MCTDH and auxiliary
programs.
The library la contains routines for linear
algebra calculations. The code is contained in various files:
- eigchess.f:
Routine to obtain eigenvalues and -vectors of a complex upper
Hessenberg matrix.
- eigcs.f:
Routine to obtain eigenvalues and -vectors of a complex
symmetric matrix.
- lanczos.f:
Lanczos diagonalisation routine.
- svd.f: Routines
for performing a singular value decomposition.
- svdz.f:
Routines for performing a singular value decomposition of
complex symmetric matrices.
- lineq.f:
Routines for solving linear equations, inverting matrices,
etc.
- mmlib.f:
Matrix*Matrix routines.
- mtlib.f:
Matrix*Tensor routines.
- op1lib.f:
Routines operating on one object.
- op2lib.f:
Routines operating on two objects of the same type.
- rmlib.f:
Routines to form "density" matrices.
- schmidtortho.f:
Routines for orthogonalising orbitals.
- ttlib.f:
Routines involving multiplication of two tensors of third
order.
- xvlib.f:
Routines involving scalars and vectors.
- d1tens.f:
Additional subroutines for matrices and tensors.
- d2tens.f:
Additional subroutines for matrices and tensors.
- lalib.f:
Other routines.
The library ode contains various
integrators for ordinary differential equations. The following
files and routines exist:
- abmlib.f:
Adams-Bashforth-Moulton predictor-corrector integrator.
- bslib.f:
Bulirsch-Stoer extrapolation method.
- sodlib.f:
Second order differencing scheme.
- sillib.f: Short
iterative Lanczos algorithm.
- csillib.f:
Lanczos-Arnoldi (complex Lanczos) integrator.
- odedp5.f:
Runge-Kutta integrator of order 5.
- oded85.f:
Runge-Kutta integrator of order 8.
- odeuti.f: Various
utility routines for the Runge-Kutta integrators.
The library num contains routines for
various numerical operations. The following files and routines
exist:
- cmdvr.f:
Calculates Hamiltonian matrix in Colbert/Miller DVR basis. (see
J. Chem. Phys. 96, 1982 (1992) for more details) Kinetic energy
elements are that of Eqs. A6 of Appendix A.
- fft.f:
Calculates the FFT-transform of a complex tensor of third order
in the middle index. No normalisation is done.
- func.f:
Calculates various functions.
- legendre.f:
Legendre and associated Legendre functions, spherical
harmonics.
- numeric.f: Various
numerical routines.
- splinenm.f:
Routines to compute cubic-splines for an array of data and to
evaluate the spline.
The library sys contains routines
concerning the machine and the operating system. The following
files and routines exist:
- einglib.f: Contains
routines to control the input of data.
- keyunits.f:
Contains routines for converting units.
- myflush.f: Contains
a self-written routine for enforcing a file buffer to be
written to disk.
- myflush_lib.f:
Calls the flush routine provided by some
compilers.
- myflush_no.f:
Contains a routine that ignores the flush-command.
- showtime.c:
Contains routines to determine the system time, host name,
etc.
- strutil.f: Contains
routines for processing character strings.
- syslib.f:
Contains various routines, e. g. for determining the machine
precision.
- dvrweights.f:
Routine for computing the DVR weights.
- platform.F:
Contains various wrappers for platform-dependent routines.
- fileutil.f:
Contains utility routines for common file operations.