Re: saving parallel vectors

[Matteo Semplice <semplice@xxxxxxxxxxxx>]

Thansk for all the replies.

I always thought that binary format would be best, but I couldn't find any 
description of the binary format. Could you point me to a relevant piece 
of the docs? Can I load the PETSC binary files into MatLab or DataExplorer 
for further processing and visualization?

Matteo

On Thu, 23 Nov 2006, Barry Smith wrote:

>  NEVER use ascii for large data sets. Use the binary viewer to save
> them. See PetscViewerBinaryOpen().
>
>   Barry
>
>
> On Thu, 23 Nov 2006, Matteo Semplice wrote:
>
> > This is really a newbie question but I am struggling to solve out of memory
> > troubles on my first processor.
> >
> > I have a parallel global vector, say v, that I obtain with a call to
> > DAGetGlobalVector. I save it to disk opening an ASCII standard viewer on
> > PETSC_COMM_WORLD and calling VecView. As I understand from the manual, this
> > causes all processors to send their data to the first one and this one writes
> > them to disk. This works fine for small computational grids, but hangs my
> > program if the grid is bigger.
> >
> > I am thinking to get each processor to write its own file and reassemble the
> > data at the end of the program. I tried to create the viewer with the
> > communicator PETSC_COMM_SELF and issue VecView. I get as many files as
> > processors, but the first one contains all the data and the others are empty!
> >
> > Why? And, more importantly, what's the best way to achieve my goal?
> >
> > Thanks a lot.
> >
> >      Matteo

--
--
Matteo Semplice
Dip. di Matematica - Universita` degli Studi di Milano
Via Saldini, 50 - 20133 MILANO.   Tel: +39 02 50316170