[petsc-users] [petsc4py] dm/examples/tutorials/ex2 in python

[Matthew Knepley]

On Tue, Jun 2, 2015 at 4:52 AM, Francesco Caimmi <francesco.caimmi at gmail.com
> wrote:

> Dear PETSC users,
>
> first of all, many thanks to the developers for making PETSC available.
>
> I am trying to get familiar with the library using petsc4py (my C/C++
> knowledge is rudimentary to say the least), and I am now trying to
> reproduce
> teh examples in dm/examples/tutorials/ in python to get accustomed to DMs.
>
> However, while translating ex2.c, I get an error whose cause I cannot
> understand.
> I am doing (I hope it's ok to post this short code snippet):
>
> impost sys
> import petsc4py
> petsc4py.init(sys.argv)
> from petsc4py import PETSc
> stype = PETSc.DMDA.StencilType.BOX
> bx    = PETSc.DMDA.BoundaryType.NONE
> by = PETSc.DMDA.BoundaryType.NONE
> comm = PETSc.COMM_WORLD
> rank = comm.rank
> OptDB = PETSc.Options()#get PETSc option DB
> M = OptDB.getInt('M', 10)
> N = OptDB.getInt('N', 8)
> m = OptDB.getInt('m', PETSc.DECIDE)
> n = OptDB.getInt('n', PETSc.DECIDE)
> dm = PETSc.DMDA().create(dim=2, sizes = (M,N), proc_sizes=(m,n),
>                              boundary_type=(bx,by), stencil_type=stype,
>                              stencil_width = 1, dof = 1, comm = comm
>                              )
> global_vec = dm.createGlobalVector()
> start, end = global_vec.getOwnershipRange()
> with global_vec as v:
>         for i in xrange(start,end):
>             v[i] = 5.0*rank
>

The 'with' construction just uses

  VecGetArray()

which return the raw C pointer. This is indexed from 0 always, so you want

  v[i-start] = 5.0*rank

If you want to index using [start, end), you need something like

  DMDAVecGetArray()

  Thanks,

     Matt


> As far as I understand this should be the closest python translation of
> ex2.c
> up to line 48, except for the viewer part which is still to be translated.
>
> If run on a single processor everything is ok, but when I run with
> mpiexec -n 2 <file_name>  I get the following error (on the second rank)
>
> #############################################
>         v[i] = 5.0*rank
> IndexError: index 40 is out of bounds for axis 0 with size 40
> #############################################
>
> I am on linux/x64 and I get this behaviour both with petsc3.4.3+petsc4py
> 3.4.2
> (the packages available from opensuse repositories) and with
> petsc3.5.4+petsc4py3.5.1 (that I built myself).
>
> I would have expected the code to seamlessly handle  the transition from
> one
> to multiple processors, so it's me who's doing something wrong or some
> other
> kind of problem? Up to now, I have never seen somethinf like that with
> vectors/matrices created by PETSc.Vec()/PETSc.Mat().
>
> Thank you for your attention,
> --
> Francesco Caimmi
>



-- 
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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20150602/c56bcd3a/attachment.html>