[Gfs-users] convergence of the poisson solver. Help installing hypre

[Jose M. López-Herrera Sánchez]

Hi all,

I have noted that the poisson solver has sometimes severe difficulties to converge for electric problems specially if neumann bc conditions are used. I wonder how would perform the hypre solver. So I want to install hypre and perform some test.  I have already downloaded hypre software but I am not able to install it properly.

Can somebody teach me how to install it properly in order to have hypre available for gerris?

Cheers

Jose

PD: Please, take into account that I am a total dummy in these issues.

[Daniel Fuster]

Hi Jose

Which system are you using? It should be as simple as downloading the
hypre libraries and reconfigure gerris. I do not remember any
difficulty to have the libraries installed

In ubuntu you can open the synaptic package manager and look for
libhypre-dev. After that you will have to reconfigure gerris

make clean
./configure ... (or sh autogen ...)
make && make install

good luck!
Daniel

2011/9/12 Jose M. López-Herrera Sánchez <jose.lopez...@gmail.com>:

> ------------------------------------------------------------------------------
> Doing More with Less: The Next Generation Virtual Desktop
> What are the key obstacles that have prevented many mid-market businesses
> from deploying virtual desktops?   How do next-generation virtual desktops
> provide companies an easier-to-deploy, easier-to-manage and more affordable
> virtual desktop model.http://www.accelacomm.com/jaw/sfnl/114/51426474/
> _______________________________________________
> Gfs-users mailing list
> Gfs-...@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/gfs-users
>
>

--
www.danielfuster.com

------------------------------------------------------------------------------
Doing More with Less: The Next Generation Virtual Desktop
What are the key obstacles that have prevented many mid-market businesses
from deploying virtual desktops? How do next-generation virtual desktops
provide companies an easier-to-deploy, easier-to-manage and more affordable
virtual desktop model.http://www.accelacomm.com/jaw/sfnl/114/51426474/
_______________________________________________
Gfs-users mailing list
Gfs-...@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/gfs-users

[Jose M. López-Herrera Sánchez]

Hi Daniel

Thanks a lot!!

The problem I had were due from the fact I was installing from the hypre sources (the tar file). Installing using synaptic has no problem as you pointed out except that together with libhypre-dev it installs a crappy version of mpi that causes in turn a crappy Gerris installation. To avoid problems I have found it is better to follow this order:

1.-  sudo apt-get install openmpi-bin openmpi-dev
2.- using synaptic install libhypre-dev

and then as you suggest

make clean
./configure ... (or sh autogen ...)
make && make install

However, to have the Module hypre working together with the Module electrohydro is not such as straightfrorward  as I though. I have to think a little about it.

cheers

Jose

[Vladimir Kolobov]

Dear Jose,

Do you observe poor convergence of the Poisson solver with Dirichlet
bc's on solids ?
As far as we see, you do not use Neumann bc's in the EHD Module ?

We have already had a discussion with you regarding poor convergence of
the Poisson solver with
Dirichlet bc's on solids. We observed that convergence got worse after
you changed the scheme
(switched to cell-center based potential) in the Poisson solver with
Dirichlet bc's.
We observed that the older scheme performed much better, especially in 3D.

It seems that hypre module does support Dirichlet bc's on solids in the
Poisson solver.
Pls let us know if it works better for you.

Thank you,
Vladimir

On 9/12/2011 8:08 AM, Jose M. López-Herrera Sánchez wrote:
>
> I have noted that the poisson solver has sometimes severe difficulties
> to converge for electric problems specially if neumann bc conditions
> are used. I wonder how would perform the hypre solver.

------------------------------------------------------------------------------
BlackBerry&reg; DevCon Americas, Oct. 18-20, San Francisco, CA
Learn about the latest advances in developing for the
BlackBerry&reg; mobile platform with sessions, labs & more.
See new tools and technologies. Register for BlackBerry&reg; DevCon today!
http://p.sf.net/sfu/rim-devcon-copy1

[Jose M. López-Herrera Sánchez]

Dear Vladimir,

First, thanks for your feedback.

 

Do you observe poor convergence of the Poisson solver with Dirichlet
bc's on solids ?

 The worst scenenario I observe is: 

solid with dirichlet bc + box boundaries neumann 0 

 

As far as we see, you do not use Neumann bc's in the EHD Module ?

Actually, the poisson solver supports Neumann conditions different from zero in the box boundaries.

It is not the case with solids, it assumes always neumann zero.

It seems that hypre module does support Dirichlet bc's on solids in the
Poisson solver.

I think so. Fortunately Stephane did that work for us, 

http://gerris.dalembert.upmc.fr/gerris/tests/tests/poisson.html#toc2

Pls let us know if it  works better  for you.

Of course!!

 cheers

Jose

[Vladimir Kolobov]

Dear Jose,

How is your experience with hypre ?

We observe a significant spike in the electric field Ex on the MPI boundary when running your
EHD Module with the Gerris Poisson Solver. This spike is not there for serial runs.

To reproduce the problem, please run the attached file with

mpirun -n 4 gerris2D -m -s 2 electro.gfs

and plot Ex.

Do you observe this spike using hypre ?

Thanks,

Vladimir

[Jose M. López-Herrera Sánchez]

Dear Vladimir,

How is your experience with hypre ?

For the moment not very good but I have to say that rather to explore thoroughly the issue I have just made a few trials. This issue is in my pile of "work to do".  It is too early to be categorical.

We observe a significant spike in the electric field Ex on the MPI boundary when running your
EHD Module with the Gerris Poisson Solver. This spike is not there for serial runs.

To reproduce the problem, please run the attached file with

mpirun -n 4 gerris2D -m -s 2 electro.gfs

and plot Ex.

The only spike I see is a negative  one (Ex= -53...)  that appears in the mixed cell which is adjacent to the left boundary.  I am afraid that this negative spike is due to non-compatible boundary conditions on the electric potential. Note that the electric potential has to be continuous and in that cell you imposing on the left face a voltage -Ex*x and in the solid the voltage is V=0. There is an unphysical discontinuity that the solver manage creating a negative spike.

I hope it helps

Cheers

Jose

[Vladimir Kolobov]

Dear Jose,

The spike does not exist when running Gerris in serial:

gerris2D -m electro.gfs

So it may not be an issue of the non-compatible boundary condition, but an MPI issue related to handling of Dirichlet bc's and/or stencil.

Thank you,

Vladimir.

On 10/14/2011 8:36 AM, Jose M. López-Herrera Sánchez wrote:

We observe a significant spike in the electric field Ex on the MPI boundary when running your
EHD Module with the Gerris Poisson Solver. This spike is not there for serial runs.

To reproduce the problem, please run the attached file with

mpirun -n 4 gerris2D -m -s 2 electro.gfs

and plot Ex.

The only spike I see is a negative  one (Ex= -53...)  that appears in the mixed cell which is adjacent to the left boundary.  I am afraid that this negative spike is due to non-compatible boundary conditions on the electric potential. Note that the electric potential has to be continuous and in that cell you imposing on the left face a voltage -Ex*x and in the solid the voltage is V=0. There is an unphysical discontinuity that the solver manage creating a negative spike.

I hope it helps

Cheers

Jose

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2d-oct