Re: F90 to C interface
- From: Sebastian Hanigk <hanigk@xxxxxxxxx>
- Date: Sun, 27 May 2007 14:29:48 +0200
"Tim Springer" <Tim.Springer@xxxxxxxxxx> writes:
In case the answer to this first question in "no" can anybody give me
some hints and tips on how a F90 interface to a C-library should
look!? I am an experienced fortran programmer (many years with F77 and
the last 3 years with F90) but I have little to no knowledge on C. So
any hints/tips/pointers/examples are welcome!
I had a while ago a similar problem (interfacing Fortran code with C
libraries); theoretically you would be able to use the Fortran2003
C-interop stuff (especially C_F_POINTER), but the compiler support at
the moment is ... sparse; g95 and gfortran have some capabilities, Sun's
f95 has a usable ISO_C_BINDING module which I had used (but had to ditch
the Sun compiler because MVAPICH/VAPI wasn't compilable with it).
In the end I had written a simple wrapper routine in C for every used
library function which has a signature like
void foo_ (<type>* a, <type>* b, ..., <type>* status) { ... }
and calls the library function "foo" with pointers dereferenced where
necessary. A Fortran module was also written containing interface
definitions for the wrapper routines.
Regarding arrays, you have to be careful: onedimensional continuous
arrays ("vectors") can be passed as an assumed-size array (like C does
it), but for arrays with rank 2 or more, you either have to use the
C-interop capabilities where available, try your luck with the CHASM
library or resort to Cray pointers (what I was forced to do).
On a few occasions it was helpful to look at the source of any MPI
library for some common problems and solutions.
Sebastian
.
- References:
- F90 to C interface
- From: Tim Springer
- F90 to C interface
- Prev by Date: F90 to C interface
- Next by Date: Re: Speed versus memory
- Previous by thread: F90 to C interface
- Index(es):
Relevant Pages
|
