dolfin team mailing list archive
-
dolfin team
-
Mailing list archive
-
Message #08237
Re: Saving data in time dependent simulations
In the following address are the code and some testing/screenshots of the RAW
format in Xd3d.
http://ptmat.ptmat.fc.ul.pt/%7Endl/other/dolfin.dir/README.html#sec-26
On Monday 16 June 2008, Anders Logg wrote:
> On Mon, Jun 16, 2008 at 07:44:22PM +0200, Kent-Andre Mardal wrote:
> > man, 16.06.2008 kl. 15.26 +0100, skrev Nuno David Lopes:
> > > What post-processor do you want to use?
> > > For Xd3d i have the solution.
> > > I don't know if there are other Post-processor that reads the files
> > > like xd3d.
> > >
> > > I've simplified/modified the vtk format and made a raw format that i'm
> > > testing.
> > > It's possible to save only the solutions.(much smaller files)
> > > The idea is exactly what you are thinking, time-dependent problems with
> > > the same mesh.
> > >
> > > Example of usage:
> > > File g_file("g.raw");
> > > g_file<<g;
> > >
> > > I've made some tests and it seems to work fine.
> >
> > I'm using vtk and I'd very much like to see your code :)
> >
> > Kent
>
> I think we need to implement a special purpose format for fast access
> to solutions, in particular time series. For ODE solutions, Benjamin
> Kehlet is working on a binary format with two levels of cache (memory
> and disk) which makes it possible to save a time series of ODE
> solutions to file and quickly access the solution at any given time t,
> including interpolation.
>
> If the time series fits in memory, it will already be in
> memory. Otherwise it will be fetched from disk.
>
> This is currently being implemented in the class ODESolution, but
> maybe we can reuse the code for PDE solutions as well. If we limit
> ourselves to solutions on a fixed constant mesh, then we just need to
> store a time series of vectors.
>
> How about a class TimeSeries that does this? It can then be reused
> by two classes ODESolution and PDESolution which take care of the
> rest (for example the mesh which can stay in memory).
>
> If this sounds good, I can ask Benjamin to move the relevant code from
> ODESolution to TimeSeries.
--
Nuno David Lopes
References