Skip to content
Permalink
Browse files

Update Fortran documentation

- hint on reverse index order for multidimensional arrays
- mention that macros in libint2_params.h may be needed in Fortran
  • Loading branch information...
pseewald committed May 10, 2019
1 parent c049c31 commit 8c8039615cd230e6aa84bb290a926326a083e5e5
Showing with 9 additions and 2 deletions.
  1. +9 −2 doc/progman/progman.tex
@@ -653,6 +653,10 @@ \subsection{Modern Fortran}
The definition of the Fortran derived type {\tt Libint\_t} is shown in Listing \ref{lst:libinttf}.
The only notable difference to the corresponding C struct (Listing \ref{lst:libintt}) is that variable names starting with an underscore in C are prefixed with a letter {\tt f} in Fortran.
Refer to Listing \ref{lst:apif} for Fortran prototypes of procedures and data.
Note that the index order of multidimensional arrays is reverse in fortran
compared to C due to row-major vs. column-major order such that e.g. {\tt
libint2\_build\_eri[am1][am2][am3][am4]} in C corresponds to {\tt
libint2\_build\_eri(am4, am3, am2, am1)} in Fortran.

\begin{lstlisting}[label=lst:libinttf,caption=Fortran definition of the \LIBINT\ integral evaluator type.]{}
type, bind(c) :: Libint_t
@@ -661,7 +665,7 @@ \subsection{Modern Fortran}
real(LIBINT2_REALTYPE), dimension(VECLEN) :: f_aB_s___0___ElecPot_s___0___Ab__up_2
real(LIBINT2_REALTYPE), dimension(VECLEN) :: f_aB_s___0___ElecPot_s___0___Ab__up_3
real(LIBINT2_REALTYPE), dimension(VECLEN) :: f_aB_s___0___ElecPot_s___0___Ab__up_4
! and so on until 4 * LIBINT2_MAX_AM_ERI
! and so on until 4 * LIBINT2_MAX_AM_eri

real(LIBINT2_REALTYPE), dimension(VECLEN) :: WP_x, WP_y, WP_z
real(LIBINT2_REALTYPE), dimension(VECLEN) :: WQ_x, WQ_y, WQ_z
@@ -721,7 +725,9 @@ \subsection{Modern Fortran}
Fortran pointers are not directly interoperable with C pointers, therefore special datatypes for C pointers ({\tt c\_ptr} and {\tt c\_funptr}) and methods
from {\tt iso\_c\_binding} to convert to Fortran pointers ({\tt c\_f\_pointer} and {\tt c\_f\_procpointer}) are needed.
If your code should be compatible with different configurations of libint compiler, you need to use conditional compilation based on the macros provided
in {\tt libint2/config.h} (such as macro {\tt LIBINT\_CONTRACTED\_INTS}, see listing \ref{lst:usefortran}).
in {\tt libint2/config.h} and {\tt libint2/libint2\_params.h} such as the
macros {\tt LIBINT2\_MAX\_AM\_eri} and {\tt LIBINT\_CONTRACTED\_INTS} (for
their use see listings \ref{lst:libinttf} and \ref{lst:usefortran}).

An example subroutine that demonstrates how to actually calculate integrals using \LIBINT\ is provided in Listing \ref{lst:usefortran}.
This example is a translation of Listing \ref{lst:usecpp} to Fortran and evaluates four-center two-body Coulomb integrals over primitive Gaussians.
@@ -730,6 +736,7 @@ \subsection{Modern Fortran}
\begin{lstlisting}[label=lst:usefortran, caption=Using \LIBINT\ from Fortran.]{}
! include libint configuration-specific macros (may not be needed)
#include <libint2/config.h>
#include <libint2/libint2_params.h>

module libint_f_example

0 comments on commit 8c80396

Please sign in to comment.
You can’t perform that action at this time.