GROUNDWATER_FLOW vs. LIQUID_FLOW

Hi all,
I encountered some problems with setting up a "simple" advective-diffusive Transport problem with LIQUID_FLOW. Maybe someone can help me in finding out what I missed.

Everything is based on a 2D system with a triangular mesh and three domains with different material properties (hydraulic conductivity, tortuosity, porosity).

Starting point is a working configuration that utilizes GROUNDWATER_FLOW. In its simplest configuration I would like to have a diffusive problem only, therefore I set only a constant head boundary on top. Independent form the initial condition, a stationary head distribution will be calculated for the first time step. As expected, darcy fluxes in the domain are very small (around 1e-17 m/s) which gives for practical purposes zero advective transport (compared to diffusion in the domain). Even if I calculate a transient problem the flow solution is not changing....

Now I try to "convert" this problem to LIQUID_FLOW. I replaced all occurrences of GROUNDWATER_FLOW with LIQUID_FLOW and all HEAD with PRESSURE1.
Then I converted hydraulic conductivities to permeabilities, and change the initial and boundary conditions to pressure values instead of head.

Run the stuff -> equation solver does not converge

Is there something else I have to change? Do I have to change the numerical settings (and why)?

Thanks and many greetings, Georg

···

--
PD Dr. Georg Kosakowski * Laboratory for Waste Management
Paul Scherrer Institut * 5232 Villigen/PSI * Switzerland
Phone: +41-56-3104743 * Fax: +41-56-3102821
Email: georg.kosakowski@psi.ch h323:+41563104743
homepage: http://people.web.psi.ch/kosakowski/

Dear Georg,

Can you pack your input files as zip and send it also to us?

Best,
Haibing

···

On 2/4/2013 11:25 AM, Georg Kosakowski wrote:

Hi all,
I encountered some problems with setting up a "simple"
advective-diffusive Transport problem with LIQUID_FLOW. Maybe someone
can help me in finding out what I missed.

Everything is based on a 2D system with a triangular mesh and three
domains with different material properties (hydraulic conductivity,
tortuosity, porosity).

Starting point is a working configuration that utilizes
GROUNDWATER_FLOW. In its simplest configuration I would like to have a
diffusive problem only, therefore I set only a constant head boundary on
top. Independent form the initial condition, a stationary head
distribution will be calculated for the first time step. As expected,
darcy fluxes in the domain are very small (around 1e-17 m/s) which gives
for practical purposes zero advective transport (compared to diffusion
in the domain). Even if I calculate a transient problem the flow
solution is not changing....

Now I try to "convert" this problem to LIQUID_FLOW. I replaced all
occurrences of GROUNDWATER_FLOW with LIQUID_FLOW and all HEAD with
PRESSURE1.
Then I converted hydraulic conductivities to permeabilities, and change
the initial and boundary conditions to pressure values instead of head.

Run the stuff -> equation solver does not converge

Is there something else I have to change? Do I have to change the
numerical settings (and why)?

Thanks and many greetings, Georg

--
-----------------------------------------------
Dr. Haibing Shao
Department Environmental Informatics
UFZ-Helmholtz Centre for Environmental Reserach
Permoserstr. 15, D-04103 Leipzig, Germany
Tel: +49 341 235 1884 Mobile: +49 170 681 5930
Email: haibing.shao@ufz.de
-----------------------------------------------

Check the permeability of the first material. Its conductivity is 4.1e-8, bit the converted permeability is 1.0e-9.

···

On 04.02.2013 11:25, Georg Kosakowski wrote:

Hi all,
I encountered some problems with setting up a "simple" advective-diffusive Transport problem with LIQUID_FLOW. Maybe someone can help me in finding out what I missed.

Everything is based on a 2D system with a triangular mesh and three domains with different material properties (hydraulic conductivity, tortuosity, porosity).

Starting point is a working configuration that utilizes GROUNDWATER_FLOW. In its simplest configuration I would like to have a diffusive problem only, therefore I set only a constant head boundary on top. Independent form the initial condition, a stationary head distribution will be calculated for the first time step. As expected, darcy fluxes in the domain are very small (around 1e-17 m/s) which gives for practical purposes zero advective transport (compared to diffusion in the domain). Even if I calculate a transient problem the flow solution is not changing....

Now I try to "convert" this problem to LIQUID_FLOW. I replaced all occurrences of GROUNDWATER_FLOW with LIQUID_FLOW and all HEAD with PRESSURE1.
Then I converted hydraulic conductivities to permeabilities, and change the initial and boundary conditions to pressure values instead of head.

Run the stuff -> equation solver does not converge

Is there something else I have to change? Do I have to change the numerical settings (and why)?

Thanks and many greetings, Georg

wenqing is right...
I changed permeability of the first material to 1e-18 (which was probably, what you meant). and get this: attached pic

···

On 04.02.2013 16:36, Wenqing Wang wrote:

Check the permeability of the first material. Its conductivity is 4.1e-8, bit the converted permeability is 1.0e-9.
On 04.02.2013 11:25, Georg Kosakowski wrote:

Hi all,
I encountered some problems with setting up a "simple" advective-diffusive Transport problem with LIQUID_FLOW. Maybe someone can help me in finding out what I missed.

Everything is based on a 2D system with a triangular mesh and three domains with different material properties (hydraulic conductivity, tortuosity, porosity).

Starting point is a working configuration that utilizes GROUNDWATER_FLOW. In its simplest configuration I would like to have a diffusive problem only, therefore I set only a constant head boundary on top. Independent form the initial condition, a stationary head distribution will be calculated for the first time step. As expected, darcy fluxes in the domain are very small (around 1e-17 m/s) which gives for practical purposes zero advective transport (compared to diffusion in the domain). Even if I calculate a transient problem the flow solution is not changing....

Now I try to "convert" this problem to LIQUID_FLOW. I replaced all occurrences of GROUNDWATER_FLOW with LIQUID_FLOW and all HEAD with PRESSURE1.
Then I converted hydraulic conductivities to permeabilities, and change the initial and boundary conditions to pressure values instead of head.

Run the stuff -> equation solver does not converge

Is there something else I have to change? Do I have to change the numerical settings (and why)?

Thanks and many greetings, Georg

--
- - - - - - - - - - - - - - - - - -

Dipl.-Hydrol. Marc Walther
Wissenschaftlicher Mitarbeiter

Technische Universität Dresden
Fakultät Umweltwissenschaften
Fachrichtung Hydrowissenschaften
Institut für Grundwasserwirtschaft
01062 Dresden

Besucheradresse
Bergstr. 66, 01069 Dresden
Neubau Chemie 2. Bauabschnitt, CHE-BA2, Raum E59

Tel +49 351 463 425 56
Fax +49 351 463 425 52
Mobil +49 178 334 18 19

- - - - - - - - - - - - - - - - - -

Hoi Wenqing,
you are right. The hydraulic conductivity in the Groundwater example is wrong. It should be something like 0.1 and then the groundwater example does not work too.
Does that mean that setups that contain two materials with a permeability contrast of 8 orders of magnitude do not work at all? Explicit simulation of fractures as porous media with porosity = 1.0 would then not possible. Correct?
Greetings, Georg

···

On 02/04/2013 04:36 PM, Wenqing Wang wrote:

Check the permeability of the first material. Its conductivity is 4.1e-8, bit the converted permeability is 1.0e-9.
On 04.02.2013 11:25, Georg Kosakowski wrote:

Hi all,
I encountered some problems with setting up a "simple" advective-diffusive Transport problem with LIQUID_FLOW. Maybe someone can help me in finding out what I missed.

Everything is based on a 2D system with a triangular mesh and three domains with different material properties (hydraulic conductivity, tortuosity, porosity).

Starting point is a working configuration that utilizes GROUNDWATER_FLOW. In its simplest configuration I would like to have a diffusive problem only, therefore I set only a constant head boundary on top. Independent form the initial condition, a stationary head distribution will be calculated for the first time step. As expected, darcy fluxes in the domain are very small (around 1e-17 m/s) which gives for practical purposes zero advective transport (compared to diffusion in the domain). Even if I calculate a transient problem the flow solution is not changing....

Now I try to "convert" this problem to LIQUID_FLOW. I replaced all occurrences of GROUNDWATER_FLOW with LIQUID_FLOW and all HEAD with PRESSURE1.
Then I converted hydraulic conductivities to permeabilities, and change the initial and boundary conditions to pressure values instead of head.

Run the stuff -> equation solver does not converge

Is there something else I have to change? Do I have to change the numerical settings (and why)?

Thanks and many greetings, Georg

--
PD Dr. Georg Kosakowski * Laboratory for Waste Management
Paul Scherrer Institut * 5232 Villigen/PSI * Switzerland
Phone: +41-56-3104743 * Fax: +41-56-3102821
Email: georg.kosakowski@psi.ch h323:+41563104743
homepage: http://people.web.psi.ch/kosakowski/

Does that mean that setups that contain two materials with a permeability contrast of 8 orders of magnitude do not work at all?

You may solve the problem with such material setting with a direct solver

Explicit simulation of fractures as porous media with porosity = 1.0 would then not possible. Correct?

I do not think it is a problem.

···

On 04.02.2013 16:44, Georg Kosakowski wrote:

Hoi Wenqing,
you are right. The hydraulic conductivity in the Groundwater example is wrong. It should be something like 0.1 and then the groundwater example does not work too.

Does that mean that setups that contain two materials with a permeability contrast of 8 orders of magnitude do not work at all?

Explicit simulation of fractures as porous media with porosity = 1.0 would then not possible. Correct?
Greetings, Georg

On 02/04/2013 04:36 PM, Wenqing Wang wrote:

Check the permeability of the first material. Its conductivity is 4.1e-8, bit the converted permeability is 1.0e-9.
On 04.02.2013 11:25, Georg Kosakowski wrote:

Hi all,
I encountered some problems with setting up a "simple" advective-diffusive Transport problem with LIQUID_FLOW. Maybe someone can help me in finding out what I missed.

Everything is based on a 2D system with a triangular mesh and three domains with different material properties (hydraulic conductivity, tortuosity, porosity).

Starting point is a working configuration that utilizes GROUNDWATER_FLOW. In its simplest configuration I would like to have a diffusive problem only, therefore I set only a constant head boundary on top. Independent form the initial condition, a stationary head distribution will be calculated for the first time step. As expected, darcy fluxes in the domain are very small (around 1e-17 m/s) which gives for practical purposes zero advective transport (compared to diffusion in the domain). Even if I calculate a transient problem the flow solution is not changing....

Now I try to "convert" this problem to LIQUID_FLOW. I replaced all occurrences of GROUNDWATER_FLOW with LIQUID_FLOW and all HEAD with PRESSURE1.
Then I converted hydraulic conductivities to permeabilities, and change the initial and boundary conditions to pressure values instead of head.

Run the stuff -> equation solver does not converge

Is there something else I have to change? Do I have to change the numerical settings (and why)?

Thanks and many greetings, Georg

Seems like we (Wenqing and me) will have to share the wine :slight_smile:

···

On 4 Feb 2013 17:04, “Wenqing Wang” wenqing.wang@ufz.de wrote:

Does that mean that setups that contain two materials with a permeability contrast of 8 orders of magnitude do not work at all?
You may solve the problem with such material setting with a direct solver
Explicit simulation of fractures as porous media with porosity = 1.0 would then not possible. Correct?
I do not think it is a problem.

On 04.02.2013 16:44, Georg Kosakowski wrote:

Hoi Wenqing,

you are right. The hydraulic conductivity in the Groundwater example is wrong. It should be something like 0.1 and then the groundwater example does not work too.

Does that mean that setups that contain two materials with a permeability contrast of 8 orders of magnitude do not work at all?

Explicit simulation of fractures as porous media with porosity = 1.0 would then not possible. Correct?

Greetings, Georg

On 02/04/2013 04:36 PM, Wenqing Wang wrote:

Check the permeability of the first material. Its conductivity is 4.1e-8, bit the converted permeability is 1.0e-9.

On 04.02.2013 11:25, Georg Kosakowski wrote:

Hi all,

I encountered some problems with setting up a “simple” advective-diffusive Transport problem with LIQUID_FLOW. Maybe someone can help me in finding out what I missed.

Everything is based on a 2D system with a triangular mesh and three domains with different material properties (hydraulic conductivity, tortuosity, porosity).

Starting point is a working configuration that utilizes GROUNDWATER_FLOW. In its simplest configuration I would like to have a diffusive problem only, therefore I set only a constant head boundary on top. Independent form the initial condition, a stationary head distribution will be calculated for the first time step. As expected, darcy fluxes in the domain are very small (around 1e-17 m/s) which gives for practical purposes zero advective transport (compared to diffusion in the domain). Even if I calculate a transient problem the flow solution is not changing…

Now I try to “convert” this problem to LIQUID_FLOW. I replaced all occurrences of GROUNDWATER_FLOW with LIQUID_FLOW and all HEAD with PRESSURE1.

Then I converted hydraulic conductivities to permeabilities, and change the initial and boundary conditions to pressure values instead of head.

Run the stuff -> equation solver does not converge

Is there something else I have to change? Do I have to change the numerical settings (and why)?

Thanks and many greetings, Georg

You received this message because you are subscribed to the Google Groups “ogs-users” group.

To unsubscribe from this group and stop receiving emails from it, send an email to ogs-users+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.

Hi,

I've tested the problem with Pardiso solver (a direct solver) in Intel MKL. See attached results (fracture permeability is 4.1e-8 m2).

Georg, the pressure profile looks good but I'm not sure if accuracy of the velocity is enough for you.

Note that OGS has already supported Pardiso but I needed to change one parameter (iparm[12] = 1) to get the accurate results.

best,
nori

···

On 02/04/2013 05:04 PM, Wenqing Wang wrote:

Does that mean that setups that contain two materials with a
permeability contrast of 8 orders of magnitude do not work at all?

You may solve the problem with such material setting with a direct solver

Explicit simulation of fractures as porous media with porosity = 1.0
would then not possible. Correct?

I do not think it is a problem.

On 04.02.2013 16:44, Georg Kosakowski wrote:

Hoi Wenqing,
you are right. The hydraulic conductivity in the Groundwater example
is wrong. It should be something like 0.1 and then the groundwater
example does not work too.

Does that mean that setups that contain two materials with a
permeability contrast of 8 orders of magnitude do not work at all?

Explicit simulation of fractures as porous media with porosity = 1.0
would then not possible. Correct?
Greetings, Georg

On 02/04/2013 04:36 PM, Wenqing Wang wrote:

Check the permeability of the first material. Its conductivity is
4.1e-8, bit the converted permeability is 1.0e-9.
On 04.02.2013 11:25, Georg Kosakowski wrote:

Hi all,
I encountered some problems with setting up a "simple"
advective-diffusive Transport problem with LIQUID_FLOW. Maybe
someone can help me in finding out what I missed.

Everything is based on a 2D system with a triangular mesh and three
domains with different material properties (hydraulic conductivity,
tortuosity, porosity).

Starting point is a working configuration that utilizes
GROUNDWATER_FLOW. In its simplest configuration I would like to have
a diffusive problem only, therefore I set only a constant head
boundary on top. Independent form the initial condition, a
stationary head distribution will be calculated for the first time
step. As expected, darcy fluxes in the domain are very small (around
1e-17 m/s) which gives for practical purposes zero advective
transport (compared to diffusion in the domain). Even if I calculate
a transient problem the flow solution is not changing....

Now I try to "convert" this problem to LIQUID_FLOW. I replaced all
occurrences of GROUNDWATER_FLOW with LIQUID_FLOW and all HEAD with
PRESSURE1.
Then I converted hydraulic conductivities to permeabilities, and
change the initial and boundary conditions to pressure values
instead of head.

Run the stuff -> equation solver does not converge

Is there something else I have to change? Do I have to change the
numerical settings (and why)?

Thanks and many greetings, Georg

--
Norihiro Watanabe
Department of Environmental Informatics (ENVINF)
Wissenschaftler

Helmholtz-Zentrum f�r Umweltforschung GmbH - UFZ
Helmholtz Centre for Environmental Research GmbH - UFZ
Permoserstra�e 15 / 04318 Leipzig / Germany

norihiro.watanabe@ufz.de / http://www.ufz.de
Telefon +49 341 235 1806

Sitz der Gesellschaft: Leipzig
Registergericht: Amtsgericht Leipzig, Handelsregister Nr. B 4703
Vorsitzender des Aufsichtsrats: MinDirig Wilfried Kraus
Wissenschaftlicher Gesch�ftsf�hrer: Prof. Dr. Georg Teutsch
Administrativer Gesch�ftsf�hrer: Dr. Heike Gra�mann

Hoi Norihiro,
I just try to compile Intel MKL which does not seem to work easily on my LINUX box as I have to access the things via our network. It seems I have to pass some PATH options to cmake (although environment variables containing such information are already set).
In any case, I will need this setup to run with the OGS_GEM version and I have to see if I can mix intel MKL and the GEM coupling.
I am currently testing if adding the fracture as line elements "solves" the problem.
Greetings, Georg

···

On 02/05/2013 10:36 AM, Norihiro Watanabe wrote:

Hi,

I've tested the problem with Pardiso solver (a direct solver) in Intel MKL. See attached results (fracture permeability is 4.1e-8 m2).

Georg, the pressure profile looks good but I'm not sure if accuracy of the velocity is enough for you.

Note that OGS has already supported Pardiso but I needed to change one parameter (iparm[12] = 1) to get the accurate results.

best,
nori

On 02/04/2013 05:04 PM, Wenqing Wang wrote:

Does that mean that setups that contain two materials with a
permeability contrast of 8 orders of magnitude do not work at all?

You may solve the problem with such material setting with a direct solver

Explicit simulation of fractures as porous media with porosity = 1.0
would then not possible. Correct?

I do not think it is a problem.

On 04.02.2013 16:44, Georg Kosakowski wrote:

Hoi Wenqing,
you are right. The hydraulic conductivity in the Groundwater example
is wrong. It should be something like 0.1 and then the groundwater
example does not work too.

Does that mean that setups that contain two materials with a
permeability contrast of 8 orders of magnitude do not work at all?

Explicit simulation of fractures as porous media with porosity = 1.0
would then not possible. Correct?
Greetings, Georg

On 02/04/2013 04:36 PM, Wenqing Wang wrote:

Check the permeability of the first material. Its conductivity is
4.1e-8, bit the converted permeability is 1.0e-9.
On 04.02.2013 11:25, Georg Kosakowski wrote:

Hi all,
I encountered some problems with setting up a "simple"
advective-diffusive Transport problem with LIQUID_FLOW. Maybe
someone can help me in finding out what I missed.

Everything is based on a 2D system with a triangular mesh and three
domains with different material properties (hydraulic conductivity,
tortuosity, porosity).

Starting point is a working configuration that utilizes
GROUNDWATER_FLOW. In its simplest configuration I would like to have
a diffusive problem only, therefore I set only a constant head
boundary on top. Independent form the initial condition, a
stationary head distribution will be calculated for the first time
step. As expected, darcy fluxes in the domain are very small (around
1e-17 m/s) which gives for practical purposes zero advective
transport (compared to diffusion in the domain). Even if I calculate
a transient problem the flow solution is not changing....

Now I try to "convert" this problem to LIQUID_FLOW. I replaced all
occurrences of GROUNDWATER_FLOW with LIQUID_FLOW and all HEAD with
PRESSURE1.
Then I converted hydraulic conductivities to permeabilities, and
change the initial and boundary conditions to pressure values
instead of head.

Run the stuff -> equation solver does not converge

Is there something else I have to change? Do I have to change the
numerical settings (and why)?

Thanks and many greetings, Georg

--
PD Dr. Georg Kosakowski * Laboratory for Waste Management
Paul Scherrer Institut * 5232 Villigen/PSI * Switzerland
Phone: +41-56-3104743 * Fax: +41-56-3102821
Email: georg.kosakowski@psi.ch h323:+41563104743
homepage: http://people.web.psi.ch/kosakowski/

1 Like

Replacing CMakeConfiguration/FindMKL.cmake with an attached one may help you to easily set path to MKL files. With the new cmake script, you just need to set MKL_DIR (a path to MKL root directory) and hopefully it will find all related setting such as include path and library path.

best
nori

FindMKL.cmake (3.76 KB)

···

On 02/05/2013 11:01 AM, Georg Kosakowski wrote:

Hoi Norihiro,
I just try to compile Intel MKL which does not seem to work easily on my
LINUX box as I have to access the things via our network. It seems I
have to pass some PATH options to cmake (although environment variables
containing such information are already set).
In any case, I will need this setup to run with the OGS_GEM version and
I have to see if I can mix intel MKL and the GEM coupling.
I am currently testing if adding the fracture as line elements "solves"
the problem.
Greetings, Georg

On 02/05/2013 10:36 AM, Norihiro Watanabe wrote:

Hi,

I've tested the problem with Pardiso solver (a direct solver) in Intel
MKL. See attached results (fracture permeability is 4.1e-8 m2).

Georg, the pressure profile looks good but I'm not sure if accuracy of
the velocity is enough for you.

Note that OGS has already supported Pardiso but I needed to change one
parameter (iparm[12] = 1) to get the accurate results.

best,
nori

On 02/04/2013 05:04 PM, Wenqing Wang wrote:

Does that mean that setups that contain two materials with a
permeability contrast of 8 orders of magnitude do not work at all?

You may solve the problem with such material setting with a direct
solver

Explicit simulation of fractures as porous media with porosity = 1.0
would then not possible. Correct?

I do not think it is a problem.

On 04.02.2013 16:44, Georg Kosakowski wrote:

Hoi Wenqing,
you are right. The hydraulic conductivity in the Groundwater example
is wrong. It should be something like 0.1 and then the groundwater
example does not work too.

Does that mean that setups that contain two materials with a
permeability contrast of 8 orders of magnitude do not work at all?

Explicit simulation of fractures as porous media with porosity = 1.0
would then not possible. Correct?
Greetings, Georg

On 02/04/2013 04:36 PM, Wenqing Wang wrote:

Check the permeability of the first material. Its conductivity is
4.1e-8, bit the converted permeability is 1.0e-9.
On 04.02.2013 11:25, Georg Kosakowski wrote:

Hi all,
I encountered some problems with setting up a "simple"
advective-diffusive Transport problem with LIQUID_FLOW. Maybe
someone can help me in finding out what I missed.

Everything is based on a 2D system with a triangular mesh and three
domains with different material properties (hydraulic conductivity,
tortuosity, porosity).

Starting point is a working configuration that utilizes
GROUNDWATER_FLOW. In its simplest configuration I would like to have
a diffusive problem only, therefore I set only a constant head
boundary on top. Independent form the initial condition, a
stationary head distribution will be calculated for the first time
step. As expected, darcy fluxes in the domain are very small (around
1e-17 m/s) which gives for practical purposes zero advective
transport (compared to diffusion in the domain). Even if I calculate
a transient problem the flow solution is not changing....

Now I try to "convert" this problem to LIQUID_FLOW. I replaced all
occurrences of GROUNDWATER_FLOW with LIQUID_FLOW and all HEAD with
PRESSURE1.
Then I converted hydraulic conductivities to permeabilities, and
change the initial and boundary conditions to pressure values
instead of head.

Run the stuff -> equation solver does not converge

Is there something else I have to change? Do I have to change the
numerical settings (and why)?

Thanks and many greetings, Georg

--
Norihiro Watanabe
Department of Environmental Informatics (ENVINF)
Wissenschaftler

Helmholtz-Zentrum f�r Umweltforschung GmbH - UFZ
Helmholtz Centre for Environmental Research GmbH - UFZ
Permoserstra�e 15 / 04318 Leipzig / Germany

norihiro.watanabe@ufz.de / http://www.ufz.de
Telefon +49 341 235 1806

Sitz der Gesellschaft: Leipzig
Registergericht: Amtsgericht Leipzig, Handelsregister Nr. B 4703
Vorsitzender des Aufsichtsrats: MinDirig Wilfried Kraus
Wissenschaftlicher Gesch�ftsf�hrer: Prof. Dr. Georg Teutsch
Administrativer Gesch�ftsf�hrer: Dr. Heike Gra�mann

Dear All,

v5.4 (parallel fem with PETSc) is in the trunk now. A description about it will be ready in today.

Wenqing

Dear all,

Here is the release notes. An instruction can be found on https://svn.ufz.de/ogs/wiki/PETScPage

Best regards,

wenqing

== OpenGeoSys 5.4 (WW) - 2013-03-26 ==
=== Enhancements ===
  * Parallel finite element scheme by using PETSc. It works for two-phase flow, Richards flow, heat-transport, liquid flow, groundwater flow. It should work for mass transport with one component. The output are limited to point wise and domain. In present release, the domain output is performed on subdmains with Tecplot with different files, and a single file output will be appeared in the next release. A detailed instruction about how to use this feature can be found [wiki:PETScPage here].
  * New mesh partition [https://svn.ufz.de/ogs/browser/trunk/partmesh, ]
=== New parallel benchmarks ===
  There are four new [https://svn.ufz.de/ogs/browser/trunk/benchmarks/PETSc, benchmarks ] to test the parallel computing with PETSc, which about two-phase flow, heat transport, liquid flow and Richards flow.
=== New keyword ===
  * '''$TECPLOT_ZONE_SHARE''' for output in order to have less verbose Tecplot file. Do not use it if keyword DEACTIVATE is sued.
=== Changes ===
Replace all '''endl''' to '''\n''' in order to avoid flushing buffer on Linux clusters or super computers. Due to this change, [[Color(background-color,red, one who will prepare a new version had better to do this change before committing in order to save conflicts)]].
There are some other changes or big fix that I could not remember in detail.
=== File added ===
mesh_lib_ext.cpp, PETscLinearSolver.cpp/h, findPETSc
=== File removed ===
intrface.cpp/h msh_gen.cpp msh_mat.cpp
=== Benchmark status ===
The following benchmarks do not work
{{{
       18 - NB_FILECOMPARE_FLUID_PROPERTIES/HT_EOS/FLUID_PROP (Failed)
       186 - BG_FILECOMPARE_ECLIPSE_DUMUX/1phase_radialflow_1phase_transport/1pf_1pt (Failed)
       188 - BG_FILECOMPARE_ECLIPSE_DUMUX/2phase_flow_2phase_tracertransport/2pf_2pt (Failed)
       190 - BG_FILECOMPARE_ECLIPSE_DUMUX/2phase_flow_radialmodel/2pf_radialmodel (Failed)
       192 - BG_FILECOMPARE_ECLIPSE_DUMUX/kinetic_CO2phase_generation_E100/CO2phase_gen_E100 (Failed)
       194 - BG_FILECOMPARE_ECLIPSE_DUMUX/kinetic_CO2phase_generation_E300/CO2phase_gen_E300 (Failed)
       196 - BG_FILECOMPARE_C/2d_Cl_transport_Clay/Nuklidtransport (Failed)
}}}
where 18 has problem before.

···

On 27.02.2013 14:47, Wenqing Wang wrote:

Dear All,

v5.4 (parallel fem with PETSc) is in the trunk now. A description about it will be ready in today.

Wenqing

Dear All,

Finally, the bugs in the native parallel FEM computation are fixed.

Just updated the trunk in the repository with bug fixes. The changes are in
1. CMakeList.txt in the root directory.
2. FEM/problem.cpp

A bookmark of this commit can be found at
https://svn.ufz.de/svn/ogs/tags/5.4/new/sources

Besides, the related benchmarks (https://svn.ufz.de/svn/ogs/trunk/benchmark_parallel) are updated to the present input syntax, and a batch script (mpi_fem_job) is add as well.

Best,

Wenqing

Hi Wenqing,

Do you have any results from Task B1? I am preparing the presentation for Korea Workshop now.

Hua

···

-----Ursprüngliche Nachricht-----
Von: ogs-users@googlegroups.com [mailto:ogs-users@googlegroups.com] Im Auftrag von Wenqing Wang
Gesendet: Mittwoch, 3. April 2013 11:44
An: ogs-users@googlegroups.com
Cc: ogs-devs@googlegroups.com
Betreff: [ogs-users] Bug fixes in v5.4

Dear All,

Finally, the bugs in the native parallel FEM computation are fixed.

Just updated the trunk in the repository with bug fixes. The changes are in
1. CMakeList.txt in the root directory.
2. FEM/problem.cpp

A bookmark of this commit can be found at
https://svn.ufz.de/svn/ogs/tags/5.4/new/sources

Besides, the related benchmarks
(https://svn.ufz.de/svn/ogs/trunk/benchmark_parallel) are updated to the
present input syntax, and a batch script (mpi_fem_job) is add as well.

Best,

Wenqing

--
You received this message because you are subscribed to the Google Groups "ogs-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ogs-users+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

Dear Hua,

I have not got time to do that yet. Hope tomorrow is free for that.

Best regards,

Wenqing

···

On 03.04.2013 12:04, Shao, Hua wrote:

Hi Wenqing,

Do you have any results from Task B1? I am preparing the presentation for Korea Workshop now.

Hua

-----Urspr�ngliche Nachricht-----
Von: ogs-users@googlegroups.com [mailto:ogs-users@googlegroups.com] Im Auftrag von Wenqing Wang
Gesendet: Mittwoch, 3. April 2013 11:44
An: ogs-users@googlegroups.com
Cc: ogs-devs@googlegroups.com
Betreff: [ogs-users] Bug fixes in v5.4

Dear All,

Finally, the bugs in the native parallel FEM computation are fixed.

Just updated the trunk in the repository with bug fixes. The changes are in
1. CMakeList.txt in the root directory.
2. FEM/problem.cpp

A bookmark of this commit can be found at
https://svn.ufz.de/svn/ogs/tags/5.4/new/sources

Besides, the related benchmarks
(https://svn.ufz.de/svn/ogs/trunk/benchmark_parallel) are updated to the
present input syntax, and a batch script (mpi_fem_job) is add as well.

Best,

Wenqing