dolfin team mailing list archive

[Johan Hake <hake@xxxxxxxxx>]

On Tuesday 08 September 2009 11:30:16 Anders Logg wrote:
> On Tue, Sep 08, 2009 at 11:13:50AM +0200, Garth N. Wells wrote:
> > Anders Logg wrote:
> > > On Tue, Sep 08, 2009 at 08:41:37AM +0200, Garth N. Wells wrote:
> > >> DOLFIN wrote:
> > >>> One or more new changesets pushed to the primary dolfin repository.
> > >>> A short summary of the last three changesets is included below.
> > >>>
> > >>> changeset:   6970:5d7fc35d3e597db508a005826efacb8dea6a00d9
> > >>> tag:         tip
> > >>> user:        Anders Logg <logg@xxxxxxxxx>
> > >>> date:        Tue Sep 08 00:50:08 2009 +0200
> > >>> files:       dolfin/parameter/GlobalParameters.h dolfin/plot/plot.cpp
> > >>> description:
> > >>> Work on parallel plotting from C++. Not yet working. Getting errors
> > >>> like [[10904,1],1] routed:binomial: Connection to lifeline
> > >>> [[10904,0],0] lost
> > >>
> > >> I don't see working in this direction on plotting as being very
> > >> profitable. We've spent effort on getting VTK output working in
> > >> parallel, so why not just tell Viper to load the .pvtk file?
> > >>
> > >> Also, plotting works from Python in parallel in the sense that things
> > >> pop up on the screen, but it doesn't really work because one only gets
> > >> a plot for each process with it's part of the domain. Working through
> > >> the .pvtk file would solve this too.
> > >>
> > >> Garth
> > >
> > > I think the builtin plot should do just what it does now, which is to
> > > pop up a separate window for each process. It's very nice for
> > > debugging.
> >
> > We already have what's needed for this - just let Viper load the .vtu
> > file for each process. My problems with the present plotting are that it
> > duplicates what we already have in place and that it's fragile.
> > Frequently it's been broken for me and it's not easy to fix because the
> > messages are not helpful since it's unclear whether a problem lies with
> > DOLFIN or Viper.
> 
> I don't know how much work would be involved to do this in Viper, or
> even if it's desirable. Ola can comment on that.
> 
> But even if we write .vtu files, the problem remains to call Viper on
> the command-line from C++ which currently breaks.
> 
> I think Hake is also looking into adding a C++ interface to Viper
> which may or may not be simple to add. If simple to add, it might
> solve this issue (the plot crashing when running in parallel). I'll
> hold off any fixes until I know the status of that effort.

The thing I might take a look at is to embed the viper module in C++, and call 
viper directly from within DOLFIN. This _should_ work, however it will take 
some testing and prototyping before anything is up and running, especially if 
it works, it will take some time to get up to par with the features we now 
have in dolfin::plot.

Johan

> --
> Anders
> 
> > Garth
> >
> > > And if you are running on a big cluster, you probably don't want to
> > > plot anyway using Viper but will save to VTK and visualize in
> > > Paraview.
> > >
> > >>> changeset:   6969:6f8c8a4229dd2a4901a91e8cc16972e198d7879d
> > >>> user:        "Johan Hake <hake@xxxxxxxxx>"
> > >>> date:        Mon Sep 07 21:32:34 2009 +0200
> > >>> files:       dolfin/swig/typemaps.i
> > >>> description:
> > >>> Removed outcommented code
> > >>>
> > >>>
> > >>> changeset:   6968:c9e4a19d28149a2e909e39abed118977afd3706c
> > >>> user:        "Johan Hake <hake@xxxxxxxxx>"
> > >>> date:        Mon Sep 07 21:31:43 2009 +0200
> > >>> files:       demo/la/block-matrix/cpp/main.cpp
> > >>> demo/la/block-matrix/python/demo.py demo/pde/elasticity/cpp/main.cpp
> > >>> demo/pde/stokes/taylor-hood/python/demo.py dolfin/la/EpetraMatrix.h
> > >>> dolfin/la/EpetraVector.h dolfin/la/GenericMatrix.h
> > >>> dolfin/la/GenericVector.h dolfin/la/MTL4Matrix.h
> > >>> dolfin/la/MTL4Vector.h dolfin/la/Matrix.h dolfin/la/PETScVector.cpp
> > >>> dolfin/la/PETScVector.h dolfin/la/STLMatrix.h dolfin/la/Vector.h
> > >>> dolfin/la/uBLASMatrix.h dolfin/la/uBLASVector.h description:
> > >>> Remove default argument to Foo.norm()
> > >>>   - Making PETScMatrix and Vector available in PyDOLFIN again
> > >>>
> > >>> ---------------------------------------------------------------------
> > >>>- For more details, visit http://www.fenics.org/hg/dolfin
> > >>> _______________________________________________
> > >>> DOLFIN-dev mailing list
> > >>> DOLFIN-dev@xxxxxxxxxx
> > >>> http://www.fenics.org/mailman/listinfo/dolfin-dev
> > >>
> > >> _______________________________________________
> > >> DOLFIN-dev mailing list
> > >> DOLFIN-dev@xxxxxxxxxx
> > >> http://www.fenics.org/mailman/listinfo/dolfin-dev
> > >>
> > >> ----------------------------------------------------------------------
> > >>--
> > >>
> > >> _______________________________________________
> > >> DOLFIN-dev mailing list
> > >> DOLFIN-dev@xxxxxxxxxx
> > >> http://www.fenics.org/mailman/listinfo/dolfin-dev
> >
> > _______________________________________________
> > DOLFIN-dev mailing list
> > DOLFIN-dev@xxxxxxxxxx
> > http://www.fenics.org/mailman/listinfo/dolfin-dev
>