 |
 |
----- Original Message ----- From: "Aleksandar Donev" <[log in to unmask]> To: <[log in to unmask]> Sent: Wednesday, October 31, 2001 12:11 PM Subject: PETSc library --- C dynamic arrays and Fortran > Hello, > > I am wondering if someone has used the parallel linear algebra package > PETSc from Fortran and what the experience has been. The package does > provide a Fortran interface to its distinctly C-like routines, but this > interface is somewhat ``scary'' and not legal. > > For example, a lot of the routines there return a C pointer to an array > allocated internally by the library. To use these in Fortran, they > imploy a dirty trick of passing an array of 1 element and returning the > integer offset of the allocated array from this array, so one can index > the array in Fortran (but passing beyond the array bounds): > > Here is a sligtly tweaked example from their page: > > INTEGER(C_pointer) Mapping > integer Array(1) > INTEGER offset, error_status > > call ISGetIndices(Mapping, Array, offset, error_status) > > The allocated array is then Array(offset: offset+array_size-1) in > Fortran, though this is illegal (maybe the wrong word, but you get the > point) since the array is only of size 1. > If I understand you correctly, this is the old pre-77 trick, which, if it works, is considered to be the same as integer Array(*) which means no bounds checking on many compilers. Since there is no standard yet on interfacing C to Fortran, you could hardly call it "legal," or expect any guarantee about portability. It appears to assume call by reference, a fairly common but not universal type of implementation.