[petsc-users] petsc4py, cython

[Jaroslaw Piwonski (CAU)]

> Am 12.01.2018 um 16:35 schrieb Matthew Knepley <knepley at gmail.com <mailto:knepley at gmail.com>>:
> 
> On Fri, Jan 12, 2018 at 9:38 AM, Jaroslaw Piwonski <jpi at informatik.uni-kiel.de <mailto:jpi at informatik.uni-kiel.de>> wrote:
> Dear PETSc users,
> 
> I would like to use petsc4py to create (say) PETSc vectors and
> then pass them to a routine from a module written in Cython,
> which in turn passes them to a routine written in C:
> 
> main.py ========================================================
> 
> from petsc4py import PETSc
> import m3d4py
> 
> if __name__ == "__main__":
> 
>     v = PETSc.Vec()
>     v.create()
>     v.setType("standard")
>     v.setSizes(3)
>     v.set(0.3)
>     v.assemble()
> 
>     m3d4py.sub_pyx(v)
> 
> m3d4py.pyx ========================================================
> 
> cimport petsc4py.PETSc as PETSc
> 
> cdef extern from "sub.c":
>     int sub_c(Vec)
> 
> I think the declaration above is wrong. 'Vec' looks like it is seeing the Python definition, and
> probably you want 'PetscVec‘.

I tried a lot in this regard. Unfortunately nothing worked:

$> cython -f m3d4py.pyx -I /Users/jpicau/miniconda3/lib/python3.6/site-packages/petsc4py/include/

Error compiling Cython file:
------------------------------------------------------------
...
cdef extern from "sub.c":
    int sub_c(PetscVec)

def sub_pyx(PETSc.Vec vec):
    vec.view()
    sub_c(vec.vec)
            ^
------------------------------------------------------------

m3d4py.pyx:8:13: Cannot convert 'PetscVec' to Python object
make: *** [all] Error 1

This seems not to be the point … the compiler does not complain about this definition:

$> cython -f m3d4py.pyx -I /Users/jpicau/miniconda3/lib/python3.6/site-packages/petsc4py/include/

Error compiling Cython file:
------------------------------------------------------------
...
cdef extern from "sub.c":
    int sub_c(i_can_define_here_what_i_want_it_wont_matter)

def sub_pyx(PETSc.Vec vec):
    vec.view()
    sub_c(vec.vec)
            ^
------------------------------------------------------------

m3d4py.pyx:8:13: Cannot convert 'PetscVec' to Python object
make: *** [all] Error 1

Thanks,
Jaroslaw

@Matt:
Sorry, the first answer went to the wrong address.

> 
>   Thanks,
> 
>     Matt
>  
> 
> def sub_pyx(PETSc.Vec vec):
>     vec.view()
>     sub_c(vec.vec)
> 
> sub.c ========================================================
> 
> #include "petsc.h"
> 
> extern PetscErrorCode sub_c(Vec);
> 
> PetscErrorCode sub_c(Vec vec) {
>     PetscFunctionBegin;
>     VecView(vec, PETSC_VIEWER_STDOUT_WORLD);
>     PetscFunctionReturn(0);
> }
> 
> ========================================================
> 
> I used (ana/mini)conda to install petsc4py and cython. When I try to
> compile the cython file I get the following error:
> 
> $> cython -f m3d4py.pyx -I /Users/jpicau/miniconda3/lib/python3.6/site-packages/petsc4py/include/
> 
> Error compiling Cython file:
> ------------------------------------------------------------
> ...
> cdef extern from "sub.c":
>     int sub_c(Vec)
> 
> def sub_pyx(PETSc.Vec vec):
>     vec.view()
>     sub_c(vec.vec)
>             ^
> ------------------------------------------------------------
> 
> m3d4py.pyx:8:13: Cannot convert 'PetscVec' to Python object
> make: *** [all] Error 1
> 
> ========================================================
> 
> I somehow realize that this is not correct. But comparing this to the petsc4py implementation ...
> 
> petscvec.pxi:
> 	int VecView(PetscVec,PetscViewer)
> 
> Vec.pyx:
>     def view(self, Viewer viewer=None):
>         cdef PetscViewer vwr = NULL
>         if viewer is not None: vwr = viewer.vwr
>         CHKERR( VecView(self.vec, vwr) )
> 
> … I do not understand why.
> 
> Can some explain this to me or just give me a hint where I can dig further, please?
> 
> And just out of pure curiosity, why isn’t the return value of VecView() in petscvec.pxi defined as PetscErrorCode?
> The type seems to be defined:
> 
> libpetsc4py.pyx:
>     ctypedef int PetscErrorCode
> 
> Thanks in advance,
> Jaroslaw
> 
> —
> 
> Dr. Jaroslaw Piwonski
> 
> Christian-Albrechts-Universität zu Kiel
> Institut für Informatik
> Algorithmische Optimale Steuerung — CO2-Aufnahme des Meeres
> Christian-Albrechts-Platz 4 <https://maps.google.com/?q=Christian-Albrechts-Platz+4+24118+Kiel&entry=gmail&source=g>
> 24118 Kiel <https://maps.google.com/?q=Christian-Albrechts-Platz+4+24118+Kiel&entry=gmail&source=g>
> 
> Telefon: +49 431 880-7452 <tel:+49%20431%208807452>
> Telefax: +49 431 880-7618 <tel:+49%20431%208807618>
> 
> 
> 
> -- 
> 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/>
—

Dr. Jaroslaw Piwonski

Christian-Albrechts-Universität zu Kiel
Institut für Informatik
Algorithmische Optimale Steuerung — CO2-Aufnahme des Meeres
Christian-Albrechts-Platz 4
24118 Kiel

Telefon: +49 431 880-7452
Telefax: +49 431 880-7618

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20180112/98946573/attachment-0001.html>