[petsc-users] DMPlexGetCone in Fortran

[Satish Balay]

I see you are using petsc-3.7. PETSc 3.8 has replaced [most of] the
include file usage with modules - so there is better typechecking.

Check 'Fortran' section of http://www.mcs.anl.gov/petsc/documentation/changes/38.html

So suggest upgrading to using latest petsc release [currently 3.9]

Unfortunately fortran doesn't enforce prototype requirements - so some
checks [that C/C++ does wrt includes] is not possible with fortran.

Satish


On Mon, 9 Apr 2018, Bernardo Rocha wrote:

> OK, just found my mistake.
> 
> I was using other include files such as "petscksp.h" and etc that I needed
> from previous codes.
> 
> I had to include ONLY
> "#include <petsc/finclude/petsc.h90>"
> to fix the problem.
> 
> After that the GetCone, GetTransitiveClosure and other routines returning
> arrays stop
> to result in Segmentation Fault.
> 
> It would be nice to give some warning when compiling to Fortran 90 and
> including
> the wrong header files.
> 
> Anyway, thanks.
> Best regards,
> Bernardo
> 
> 
> 
> On Sat, Apr 7, 2018 at 4:13 PM, Bernardo Rocha <
> bernardomartinsrocha at gmail.com> wrote:
> 
> > OK, here is the entire code.
> > I created a simple mesh to test DMPlex.
> > As you can see, the GetCone is not working.
> > As I said, I followed one of the examples for DMPlex using Fortran, I'm
> > not an exper in F90.
> >
> > Best regards,
> > Bernardo M. Rocha
> >
> >
> > On Fri, Apr 6, 2018 at 11:48 PM, Matthew Knepley <knepley at gmail.com>
> > wrote:
> >
> >> If you send the whole code, we can try it here. I am assuming that the
> >> examples run for you.
> >>
> >>   Matt
> >>
> >> On Fri, Apr 6, 2018 at 5:47 PM, Bernardo Rocha <
> >> bernardomartinsrocha at gmail.com> wrote:
> >>
> >>> Hello,
> >>>
> >>> I can't get a a simple code to work for DMPlexGetCone.
> >>> I have tried different ways of handling the array/pointer which is the
> >>> third argument of DMPlexGetCone,
> >>> but neither of them worked for me.
> >>>
> >>> The last version that I got compiling is this one, which is very much
> >>> based on the
> >>> following example: .../petsc-3.7.6/src/dm/impls/p
> >>> lex/examples/tutorials/ex1f90.F
> >>>
> >>>     PetscInt, target, dimension(6) :: icone
> >>>     PetscInt, pointer :: pIcone(:)
> >>>
> >>>     call DMPlexCreateFromCellList(PETSC_COMM_WORLD,
> >>>      &                              ndim,numel,numnp,nen,PETSC_TRUE,
> >>>      &                              ielement,ndim,xnodes,dm,ierr)
> >>>       ...
> >>>       call DMPlexGetCone(dm,0,pIcone,ierr)
> >>>       write(*,*) icone(1),icone(2),icone(3),icone(4),icone(5),icone(6)
> >>>       call DMPlexRestoreCone(dm,0,pIcone,ierr)
> >>>
> >>> I always get a segmentation fault.
> >>>
> >>> I've created a simple cuboid mesh with 2x2x2 hexaedral elements.
> >>> Many other routines are working and give reasonable results.
> >>>
> >>> Output of DMView
> >>>
> >>> DM Object: 1 MPI processes
> >>>   type: plex
> >>> DM_0x84000000_0 in 3 dimensions:
> >>>   0-cells: 27
> >>>   1-cells: 66
> >>>   2-cells: 44
> >>>   3-cells: 8
> >>> Labels:
> >>>   depth: 4 strata of sizes (27, 66, 44, 8)
> >>>
> >>>
> >>>
> >>>
> >>> On Thu, Apr 5, 2018 at 10:57 PM, Bernardo Rocha <
> >>> bernardomartinsrocha at gmail.com> wrote:
> >>>
> >>>> OK, thanks a lot.
> >>>> I had to rename the files and add -ffixed-form as argument to the
> >>>> compiler.
> >>>> Best regards,
> >>>> Bernardo
> >>>>
> >>>> On Thu, Apr 5, 2018 at 8:21 PM, Matthew Knepley <knepley at gmail.com>
> >>>> wrote:
> >>>>
> >>>>> On Thu, Apr 5, 2018 at 7:16 PM, Bernardo Rocha <
> >>>>> bernardomartinsrocha at gmail.com> wrote:
> >>>>>
> >>>>>> Hello everyone,
> >>>>>>
> >>>>>> I've been trying to use DMPlex in a Fortran 77 application
> >>>>>> (unfortunately this is legacy code and I can't move to F90).
> >>>>>>
> >>>>>> Is there a way to use DMPlexGetCone on it?
> >>>>>> Although the documentation online says it is only available for F90,
> >>>>>> I'm wondering if there is a trick like the ones in the users manual
> >>>>>> for routines
> >>>>>> that return an array. I tried, but I'm not sure it it is right.
> >>>>>>
> >>>>>
> >>>>> We have standardized our array handling with F90, and I don't think we
> >>>>> will go back
> >>>>> to supporting new things for F77-style arrays.
> >>>>>
> >>>>> That being said, I think all you have to do is compile that source
> >>>>> with an F90 compiler. You
> >>>>> should not have to change anything. If you want the F90 array to look
> >>>>> like an F77 array,
> >>>>> just pass it as a function argument (I think this works).
> >>>>>
> >>>>>   Thanks,
> >>>>>
> >>>>>      Matt
> >>>>>
> >>>>>
> >>>>>> Best regards,
> >>>>>> Bernardo M. Rocha
> >>>>>>
> >>>>> --
> >>>>> What most experimenters take for granted before they begin their
> >>>>> experiments is infinitely more interesting than any results to which their
> >>>>> experiments lead.
> >>>>> -- Norbert Wiener
> >>>>>
> >>>>> https://www.cse.buffalo.edu/~knepley/
> >>>>> <http://www.caam.rice.edu/~mk51/>
> >>>>>
> >>>>
> >>>>
> >>>
> >>
> >>
> >> --
> >> What most experimenters take for granted before they begin their
> >> experiments is infinitely more interesting than any results to which their
> >> experiments lead.
> >> -- Norbert Wiener
> >>
> >> https://www.cse.buffalo.edu/~knepley/ <http://www.caam.rice.edu/~mk51/>
> >>
> >
> >
>