Benchmarks wont run

Hello all,
while comparing some stuff today I tried to run the MULTIPHASE Benchmarks (fresh checkout) using the current trunk version. For some reason only the KueperProblem-PcPnw works out of the box (on the machines I tested).
Maybe you guys have an idea what is going on:

- The BuckleyLeverett Benchmark fails due to the *.gli file which is missing the name of a point used in the *.bc file.
     When this is corrected the benchmark fails in function
         void LegacyVtkInterface::WriteVTKDataArrays(fstream &vtk_file) const
     in line
         if (_pointArrayNames[k].find("_X") != string::npos && _pointArrayNames[k + 1].find("_Y") != string::npos)
     for k=5 since the vector _pointArraynames has just 6 entries.

- The KueperProblem-PcPnw Benchmark works if the #ID tag in the *.gli is changed to #NAME

- The McWortherProblem tries to use a linear solver which is not supported (805)
The latter two problems are rather easy fix if I do not miss anything.
Oh, I tested this on windows & linux using the latest trunk version of both the benchmarks as well as the source code.

Best regards,
Tilmann

···

--
Wolf Tilmann Pfeiffer
Geohydromodelling
Institute of Geoscience
Christian-Albrechts-University of Kiel
Ludewig-Meyn-Strasse 10, 24118 Kiel, Germany
phone: 0049 (0)431 880-4805
mail: wtp@gpi.uni-kiel.de

Hoi,
regarding the problems with the gli-files, please talk to Thomas
Kallbacher. I would guess this is related to some changes he introduced
into the trunk a short while ago.
Greetings, Georg

···

On 03/25/2014 04:48 PM, Wolf Tilmann Pfeiffer wrote:

Hello all,
while comparing some stuff today I tried to run the MULTIPHASE
Benchmarks (fresh checkout) using the current trunk version. For some
reason only the KueperProblem-PcPnw works out of the box (on the
machines I tested).
Maybe you guys have an idea what is going on:

- The BuckleyLeverett Benchmark fails due to the *.gli file which is
missing the name of a point used in the *.bc file.
    When this is corrected the benchmark fails in function
        void LegacyVtkInterface::WriteVTKDataArrays(fstream &vtk_file)
const
    in line
        if (_pointArrayNames[k].find("_X") != string::npos &&
_pointArrayNames[k + 1].find("_Y") != string::npos)
    for k=5 since the vector _pointArraynames has just 6 entries.

- The KueperProblem-PcPnw Benchmark works if the #ID tag in the
*.gli is changed to #NAME

- The McWortherProblem tries to use a linear solver which is not
supported (805)
The latter two problems are rather easy fix if I do not miss anything.
Oh, I tested this on windows & linux using the latest trunk version of
both the benchmarks as well as the source code.

Best regards,
Tilmann

--
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/

Sorry for the quick repost.
I forgot to mention that the point names in the McWhorter benchmarks *.gli file cause the run to fail as well.
Tilmann

···

Am 25.03.2014 16:48, schrieb Wolf Tilmann Pfeiffer:

Hello all,
while comparing some stuff today I tried to run the MULTIPHASE Benchmarks (fresh checkout) using the current trunk version. For some reason only the KueperProblem-PcPnw works out of the box (on the machines I tested).
Maybe you guys have an idea what is going on:

- The BuckleyLeverett Benchmark fails due to the *.gli file which is missing the name of a point used in the *.bc file.
    When this is corrected the benchmark fails in function
        void LegacyVtkInterface::WriteVTKDataArrays(fstream &vtk_file) const
    in line
        if (_pointArrayNames[k].find("_X") != string::npos && _pointArrayNames[k + 1].find("_Y") != string::npos)
    for k=5 since the vector _pointArraynames has just 6 entries.

- The KueperProblem-PcPnw Benchmark works if the #ID tag in the *.gli is changed to #NAME

- The McWortherProblem tries to use a linear solver which is not supported (805)
The latter two problems are rather easy fix if I do not miss anything.
Oh, I tested this on windows & linux using the latest trunk version of both the benchmarks as well as the source code.

Best regards,
Tilmann

--
Wolf Tilmann Pfeiffer
Geohydromodelling
Institute of Geoscience
Christian-Albrechts-University of Kiel
Ludewig-Meyn-Strasse 10, 24118 Kiel, Germany
phone: 0049 (0)431 880-4805
mail: wtp@gpi.uni-kiel.de

Hi Georg and Tilmann,

 it seems not to be related to reading the gli-surface names, which

I modified.

However I will have a  closer look at it tomorrow.

best

Thomas
···

-- ---------------------------------------------------------------------
Dr. Thomas Kalbacher
Department of Environmental Informatics
Helmholtz-Zentrum für Umweltforschung GmbH - UFZ
Helmholtz Centre for Environmental Research - UFZ
phone +49 (0) 0341 235 1093 / fax +49 (0) 0341 235 1939
/ Sitz der Gesellschaft: Leipzig /Registergericht: Amtsgericht Leipzig, Handelsregister Nr. B 4703 / Vorsitzender des Aufsichtsrats: MinDirig Wilfried Kraus / Wissenschaftlicher Geschäftsführer: Prof. Dr. Georg Teutsch / Administrativer Geschäftsführer: Dr. Heike Graßmann
---------------------------------------------------------------------
The OpenGeoSys Project: ---------------------------------------------------------------------

thomas.kalbacher@ufz.dehttp://www.ufz.dehttp://www.opengeosys.org/foruminfo@opengeosys.org

On our jenkens server (https://svn.ufz.de/jenkins/job/Benchmarks_Linux/), all benchmarks are OK.

···

On 25.03.2014 16:48, Wolf Tilmann Pfeiffer wrote:

Hello all,
while comparing some stuff today I tried to run the MULTIPHASE Benchmarks (fresh checkout) using the current trunk version. For some reason only the KueperProblem-PcPnw works out of the box (on the machines I tested).
Maybe you guys have an idea what is going on:

- The BuckleyLeverett Benchmark fails due to the *.gli file which is missing the name of a point used in the *.bc file.
    When this is corrected the benchmark fails in function
        void LegacyVtkInterface::WriteVTKDataArrays(fstream &vtk_file) const
    in line
        if (_pointArrayNames[k].find("_X") != string::npos && _pointArrayNames[k + 1].find("_Y") != string::npos)
    for k=5 since the vector _pointArraynames has just 6 entries.

- The KueperProblem-PcPnw Benchmark works if the #ID tag in the *.gli is changed to #NAME

- The McWortherProblem tries to use a linear solver which is not supported (805)
The latter two problems are rather easy fix if I do not miss anything.
Oh, I tested this on windows & linux using the latest trunk version of both the benchmarks as well as the source code.

Best regards,
Tilmann

I don't know why but it seems only KueperProblem-PcPnw is included in automatic benchmarking on Jenkins. BuckleyLeverett is not there so that nobody has tested when new versions released.

Best regards,
Nori

···

On 03/26/2014 09:33 AM, Wenqing Wang wrote:

On our jenkens server
(https://svn.ufz.de/jenkins/job/Benchmarks_Linux/), all benchmarks are OK.

On 25.03.2014 16:48, Wolf Tilmann Pfeiffer wrote:

Hello all,
while comparing some stuff today I tried to run the MULTIPHASE
Benchmarks (fresh checkout) using the current trunk version. For some
reason only the KueperProblem-PcPnw works out of the box (on the
machines I tested).
Maybe you guys have an idea what is going on:

- The BuckleyLeverett Benchmark fails due to the *.gli file which is
missing the name of a point used in the *.bc file.
    When this is corrected the benchmark fails in function
        void LegacyVtkInterface::WriteVTKDataArrays(fstream &vtk_file)
const
    in line
        if (_pointArrayNames[k].find("_X") != string::npos &&
_pointArrayNames[k + 1].find("_Y") != string::npos)
    for k=5 since the vector _pointArraynames has just 6 entries.

- The KueperProblem-PcPnw Benchmark works if the #ID tag in the
*.gli is changed to #NAME

- The McWortherProblem tries to use a linear solver which is not
supported (805)
The latter two problems are rather easy fix if I do not miss anything.
Oh, I tested this on windows & linux using the latest trunk version of
both the benchmarks as well as the source code.

Best regards,
Tilmann

--
Dr.-Ing. Norihiro Watanabe
Department of Environmental Informatics (ENVINF)
Wissenschaftler

Helmholtz-Zentrum für Umweltforschung GmbH - UFZ
Helmholtz Centre for Environmental Research GmbH - UFZ
Permoserstraße 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 Geschäftsführer: Prof. Dr. Georg Teutsch
Administrativer Geschäftsführer: Dr. Heike Graßmann

update. actually BuckleyLeverett and other benchmarks are included in automatic benchmarking. However, those benchmarks are configured for OGS_FEM_MKL (OGS with Intel MKL lib) which is currently not supported in the benchmarking.

My suggestion is either supporting OGS_FEM_MKL in the benchmarking or changing their input files to work with OGS_FEM or OGS_FEM_LIS. Supporting OGS_FEM_MKL may be difficult because we don't have a license for MKL lib.

Best regards,
Nori

···

On 03/26/2014 10:00 AM, Norihiro Watanabe wrote:

I don't know why but it seems only KueperProblem-PcPnw is included in
automatic benchmarking on Jenkins. BuckleyLeverett is not there so that
nobody has tested when new versions released.

Best regards,
Nori

On 03/26/2014 09:33 AM, Wenqing Wang wrote:

On our jenkens server
(https://svn.ufz.de/jenkins/job/Benchmarks_Linux/), all benchmarks are
OK.

On 25.03.2014 16:48, Wolf Tilmann Pfeiffer wrote:

Hello all,
while comparing some stuff today I tried to run the MULTIPHASE
Benchmarks (fresh checkout) using the current trunk version. For some
reason only the KueperProblem-PcPnw works out of the box (on the
machines I tested).
Maybe you guys have an idea what is going on:

- The BuckleyLeverett Benchmark fails due to the *.gli file which is
missing the name of a point used in the *.bc file.
    When this is corrected the benchmark fails in function
        void LegacyVtkInterface::WriteVTKDataArrays(fstream &vtk_file)
const
    in line
        if (_pointArrayNames[k].find("_X") != string::npos &&
_pointArrayNames[k + 1].find("_Y") != string::npos)
    for k=5 since the vector _pointArraynames has just 6 entries.

- The KueperProblem-PcPnw Benchmark works if the #ID tag in the
*.gli is changed to #NAME

- The McWortherProblem tries to use a linear solver which is not
supported (805)
The latter two problems are rather easy fix if I do not miss anything.
Oh, I tested this on windows & linux using the latest trunk version of
both the benchmarks as well as the source code.

Best regards,
Tilmann

--
Dr.-Ing. Norihiro Watanabe
Department of Environmental Informatics (ENVINF)
Wissenschaftler

Helmholtz-Zentrum für Umweltforschung GmbH - UFZ
Helmholtz Centre for Environmental Research GmbH - UFZ
Permoserstraße 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 Geschäftsführer: Prof. Dr. Georg Teutsch
Administrativer Geschäftsführer: Dr. Heike Graßmann

Thank you all for your quick replies. I agree that the benchmarks in question are actually not tested on the server. Are all of the MULTIPHASE/ benchmarks configured for OGS_FEM_MKL?
I am just asking since I get most of them to run on my local machine if I correct the *.gli files...

@Chan-Hee:
Thank you for your comprehensive reply! I am working on some changes to the OGS-ECLIPSE interface (concerning the multiphase flow process) and those changes are supposed to be availabe in a future trunk version.
Obviously this means I have to check how my changes are affecting the benchmarks, which is why I ran into the problems in the first place.

Best regards,
Tilmann

···

Am 26.03.2014 10:30, schrieb Norihiro Watanabe:

update. actually BuckleyLeverett and other benchmarks are included in automatic benchmarking. However, those benchmarks are configured for OGS_FEM_MKL (OGS with Intel MKL lib) which is currently not supported in the benchmarking.

My suggestion is either supporting OGS_FEM_MKL in the benchmarking or changing their input files to work with OGS_FEM or OGS_FEM_LIS. Supporting OGS_FEM_MKL may be difficult because we don't have a license for MKL lib.

Best regards,
Nori

On 03/26/2014 10:00 AM, Norihiro Watanabe wrote:

I don't know why but it seems only KueperProblem-PcPnw is included in
automatic benchmarking on Jenkins. BuckleyLeverett is not there so that
nobody has tested when new versions released.

Best regards,
Nori

On 03/26/2014 09:33 AM, Wenqing Wang wrote:

On our jenkens server
(https://svn.ufz.de/jenkins/job/Benchmarks_Linux/), all benchmarks are
OK.

On 25.03.2014 16:48, Wolf Tilmann Pfeiffer wrote:

Hello all,
while comparing some stuff today I tried to run the MULTIPHASE
Benchmarks (fresh checkout) using the current trunk version. For some
reason only the KueperProblem-PcPnw works out of the box (on the
machines I tested).
Maybe you guys have an idea what is going on:

- The BuckleyLeverett Benchmark fails due to the *.gli file which is
missing the name of a point used in the *.bc file.
    When this is corrected the benchmark fails in function
        void LegacyVtkInterface::WriteVTKDataArrays(fstream &vtk_file)
const
    in line
        if (_pointArrayNames[k].find("_X") != string::npos &&
_pointArrayNames[k + 1].find("_Y") != string::npos)
    for k=5 since the vector _pointArraynames has just 6 entries.

- The KueperProblem-PcPnw Benchmark works if the #ID tag in the
*.gli is changed to #NAME

- The McWortherProblem tries to use a linear solver which is not
supported (805)
The latter two problems are rather easy fix if I do not miss anything.
Oh, I tested this on windows & linux using the latest trunk version of
both the benchmarks as well as the source code.

Best regards,
Tilmann

--
Wolf Tilmann Pfeiffer
Geohydromodelling
Institute of Geoscience
Christian-Albrechts-University of Kiel
Ludewig-Meyn-Strasse 10, 24118 Kiel, Germany
phone: 0049 (0)431 880-4805
mail: wtp@gpi.uni-kiel.de