dolfin team mailing list archive

[Johan Hake <johan.hake@xxxxxxxxx>]

On Thursday January 13 2011 15:06:10 Garth N. Wells wrote:
> On 13/01/11 22:10, Johan Hake wrote:
> > On Thursday January 13 2011 13:35:05 Garth N. Wells wrote:
> >> On 13/01/11 21:30, Johan Hake wrote:
> >>> Do you have plans to create a shared_ptr version of all methods
> >>> receiving and returning a GenericFoo?
> >> 
> >> Not just yet. I'm starting with BlockFoo because it's more explicit that
> >> BlockFoo usually represent views.
> >> 
> >> I had a go at fixing thins in SWIG, but the uBLAS matrix objects are
> >> giving me trouble. Could you take a look?
> > 
> > The point is that none of the matrices or vectors in DOLFIN is stored
> > using shared_ptr. Do we want that? Then we need to add shared_ptr
> > versions of all methods?
> 
> Do we have many? I don't see a presssing need to change other linear
> algebra objects.

Ok, but I couldn't get all the stuff in directors working for shared_ptr 
stored LA structures.

> > I have tried to do this but unfortunately there are places we need to
> > update to get this going.
> 
> BlockFoo is different from the other matrix and vector objects because
> it's really just a holder of other objects. I can't see how we could
> have a robust BlockFoo implementation without using shared pointers.

I see this, and I think the change is logical, but to be able to access a 
shared_ptr<GenericMatrix> argument in Python we need to store _all_ our 
GenericTensors as shared_ptr in Python. 

When we do not provide an interface for the rest of the C++ library things can 
happen. It mostly goes well as SWIG provides a dereference typemap for most 
arguments, but again, I stumbled over a problem with directors.

Johan

> Garth
> 
> > It looked like the director implementation of for
> > example NonLinearProblem broke.
> > 
> > I think we shoudl wait with this...
> > 
> > Should
> > 
> >> Garth
> >> 
> >>> Johan
> >>> 
> >>> On Thursday January 13 2011 13:22:28 noreply@xxxxxxxxxxxxx wrote:
> >>>> ------------------------------------------------------------
> >>>> revno: 5498
> >>>> committer: Garth N. Wells <gnw20@xxxxxxxxx>
> >>>> branch nick: dolfin-all
> >>>> timestamp: Thu 2011-01-13 21:19:51 +0000
> >>>> 
> >>>> message:
> >>>>   More work on shared pointers in linear algebra.
> >>>> 
> >>>> modified:
> >>>>   demo/undocumented/block-matrix/cpp/main.cpp
> >>>>   dolfin/la/BlockMatrix.cpp
> >>>>   dolfin/la/BlockMatrix.h
> >>>>   dolfin/la/BlockVector.cpp
> >>>>   dolfin/la/BlockVector.h
> >>>>   dolfin/swig/la_post.i
> >>>>   dolfin/swig/la_pre.i
> >>>> 
> >>>> --
> >>>> lp:dolfin
> >>>> https://code.launchpad.net/~dolfin-core/dolfin/main
> >>>> 
> >>>> Your team DOLFIN Core Team is subscribed to branch lp:dolfin.
> >>>> To unsubscribe from this branch go to
> >>>> https://code.launchpad.net/~dolfin-core/dolfin/main/+edit-subscription
> >>> 
> >>> _______________________________________________
> >>> Mailing list: https://launchpad.net/~dolfin
> >>> Post to     : dolfin@xxxxxxxxxxxxxxxxxxx
> >>> Unsubscribe : https://launchpad.net/~dolfin
> >>> More help   : https://help.launchpad.net/ListHelp
> >> 
> >> _______________________________________________
> >> Mailing list: https://launchpad.net/~dolfin
> >> Post to     : dolfin@xxxxxxxxxxxxxxxxxxx
> >> Unsubscribe : https://launchpad.net/~dolfin
> >> More help   : https://help.launchpad.net/ListHelp