dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

On 9 March 2012 13:58, Marco Morandini <morandini@xxxxxxxxxxxxxx> wrote:
> This commit
>
> http://bazaar.launchpad.net/~dolfin-core/dolfin/1.0.x/revision/6520
>
> to the 1.0.x branch has not been applied to mainline.
>

I've cherry picked this change from 1.0 and push it to lp:dolfin.

> That said, even with this fix applied repeatedly solving a nonlinear problem
> (without re-creating the problem or the solver, but only changing the
> boundary conditions) the Jacobian matrix connectivity is
> computed each time solve() is called.
> This is because both the NonlinearDiscreteProblem and the NewtonSolver
> are re-created whenever solve() is called.
> And this, at least for my time-stepping application means
> a lot of wasted time if I run without adaptivity.
>
> Could someone review the patch applied to the branch below?
>

Sure - could you update it to lp:dolfin following the changes I just
pushed and then make a merge request on Launchpad?

Garth


> https://code.launchpad.net/~morandini/dolfin/resetjacobian
>
> It seems to work for me, but I'm not at all sure
> that it will not break things.
>
> Thanks in advance,
>
> Marco Morandini
>
> _______________________________________________
> Mailing list: https://launchpad.net/~dolfin
> Post to     : dolfin@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~dolfin
> More help   : https://help.launchpad.net/ListHelp