Hi Spencer, thanks for the command. I was able to merge partitions with it, and convert to VTU. However, some problems show up for highly partitioned cases. I attached an example for two cases merged from p=2, and p=100 (p number of partitions). The contours are the mean spanwise vorticity, extracted with the slightly extended meanmode-module that I wrote. In the case of p=100, there are clear oscillations in some areas. I didn't observe any oscillations for velocity and pressure fields in any cases. So, I guess calculation of derivatives giving problems in highly-partitioned case. I was wondering what could be the reason for this. Thank you very much in advance for any feedback. Best, Asim On Mon, Nov 21, 2016 at 7:29 AM, Sherwin, Spencer J < s.sherwin@imperial.ac.uk> wrote:

HI Asim,

If you put all the files into a directory called meanField.chk/

The only thing you then need in this directory is a file called info.xml which tell you which partitions contain which elements. This can be generated from FieldConvert using

[image: page98image37800]

FieldConvert --nprocs 10 file2.xml file2.fld/Info.xml:info

There should also be some details about this in the user-guide in section 5.5.1 (user-guide-4.3.2.pdf).

I hope this helps.

Spencer.

On 16 Nov 2016, at 09:26, Asım Önder <asim.onder@gmail.com> wrote:

Dear all,

Due to memory limitations, often I have to partition large data files using something like

FieldConvert --part-only=100 config.xml fields/field_1.chk

, and postprocess them piece by piece:

FieldConvert -v -m meanstats config_xml/P000000"$p".xml fields/field_1.chk meanFields/meanField_1_p"$p".fld

In the case of meanmode module for example, the output is a collection of many small files associated with partitions, which can be analyzed piece by piece. I was wondering if there is a way to merge the outputs back into one piece again like one piece fld file.

Any feedback is appreciated.

Thanks, Asim _______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users

Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ

s.sherwin@imperial.ac.uk +44 (0) 20 759 45052

-- Asim

Hi Spencer, thanks for the feedback. I think vtu conversion is problematic in general for massively parallel cases. Besides the reported issue with superficial noises in the field, there are also problems sometimes with just opening the files. For example, I cannot open some merged instances in Paraview for p=100 case: vtkXMLUnstructuredGridReader (0x9414c40): Cannot read point data array "P" from PointData in piece 1851. The data array in the element may be too short. It is strange that I get this problem for some cases, while having no troubles with the rest of database. The workflow is identical except the specified input file to the meanmode module. I was kindly wondering if there is a way to convert the .fld files directly to simple raw data-files in this format [x,y,z,data1,data2,...]. This way I can skip this apparently problematic vtu-conversion step. Thanks a lot again for any feedback. Best, Asim On Sun, Nov 27, 2016 at 7:41 PM, Sherwin, Spencer J < s.sherwin@imperial.ac.uk> wrote:

Hi Asim,

I am just resending this email to the whole nektar-users list in case others have experienced this issue.

Thanks for your summary. This is certainly strange. It looks like some noise is being added in your p=100 case however I have not seen this previously.

I am cc’ing Dave & Douglas to see if they have experienced this at all?

Cheers, Spencer.

On 25 Nov 2016, at 06:07, Asım Önder <asim.onder@gmail.com> wrote:

Hi Spencer,

thanks for the command. I was able to merge partitions with it, and convert to VTU. However, some problems show up for highly partitioned cases.

I attached an example for two cases merged from p=2, and p=100 (p number of partitions). The contours are the mean spanwise vorticity, extracted with the slightly extended meanmode-module that I wrote. In the case of p=100, there are clear oscillations in some areas.

I didn't observe any oscillations for velocity and pressure fields in any cases. So, I guess calculation of derivatives giving problems in highly-partitioned case. I was wondering what could be the reason for this.

Thank you very much in advance for any feedback.

Best, Asim

On Mon, Nov 21, 2016 at 7:29 AM, Sherwin, Spencer J < s.sherwin@imperial.ac.uk>wrote:

...

HI Asim,

If you put all the files into a directory called meanField.chk/

The only thing you then need in this directory is a file called info.xml which tell you which partitions contain which elements. This can be generated from FieldConvert using

<page98image37800.png>

FieldConvert --nprocs 10 file2.xml file2.fld/Info.xml:info

There should also be some details about this in the user-guide in section 5.5.1 (user-guide-4.3.2.pdf).

I hope this helps.

Spencer.

On 16 Nov 2016, at 09:26, Asım Önder <asim.onder@gmail.com> wrote:

Dear all,

Due to memory limitations, often I have to partition large data files using something like

FieldConvert --part-only=100 config.xml fields/field_1.chk

, and postprocess them piece by piece:

FieldConvert -v -m meanstats config_xml/P000000"$p".xml fields/field_1.chk meanFields/meanField_1_p"$p".fld

In the case of meanmode module for example, the output is a collection of many small files associated with partitions, which can be analyzed piece by piece. I was wondering if there is a way to merge the outputs back into one piece again like one piece fld file.

Any feedback is appreciated.

Thanks, Asim _______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users

Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ

s.sherwin@imperial.ac.uk +44 (0) 20 759 45052

-- Asim <mean_spanwise_vorticity.png>

Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ

s.sherwin@imperial.ac.uk +44 (0) 20 759 45052

-- Asim

Hi Asim, Yep that is certainly one way and Killian has been working on various interpolation routines that may well help improve your output. You could also try the Tecplot output (which is what I use) there is both a .dat and a binary .plt output for Tecplot which can also be read by VisIt although I am not sure whether Paraview can read this. Cheers, Spencer. On 29 Nov 2016, at 10:18, Asım Önder <asim.onder@gmail.com<mailto:asim.onder@gmail.com>> wrote: Hi, I just noticed conversion to raw data already exists in the form of '.pts' format, which works nicely for me at the moment. You may ignore my query. Thanks, Asim On Tue, Nov 29, 2016 at 3:35 PM, Asım Önder <asim.onder@gmail.com<mailto:asim.onder@gmail.com>> wrote: Hi Spencer, thanks for the feedback. I think vtu conversion is problematic in general for massively parallel cases. Besides the reported issue with superficial noises in the field, there are also problems sometimes with just opening the files. For example, I cannot open some merged instances in Paraview for p=100 case: vtkXMLUnstructuredGridReader (0x9414c40): Cannot read point data array "P" from PointData in piece 1851. The data array in the element may be too short. It is strange that I get this problem for some cases, while having no troubles with the rest of database. The workflow is identical except the specified input file to the meanmode module. I was kindly wondering if there is a way to convert the .fld files directly to simple raw data-files in this format [x,y,z,data1,data2,...]. This way I can skip this apparently problematic vtu-conversion step. Thanks a lot again for any feedback. Best, Asim On Sun, Nov 27, 2016 at 7:41 PM, Sherwin, Spencer J <s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk>> wrote: Hi Asim, I am just resending this email to the whole nektar-users list in case others have experienced this issue. Thanks for your summary. This is certainly strange. It looks like some noise is being added in your p=100 case however I have not seen this previously. I am cc’ing Dave & Douglas to see if they have experienced this at all? Cheers, Spencer. On 25 Nov 2016, at 06:07, Asım Önder <asim.onder@gmail.com<mailto:asim.onder@gmail.com>> wrote: Hi Spencer, thanks for the command. I was able to merge partitions with it, and convert to VTU. However, some problems show up for highly partitioned cases. I attached an example for two cases merged from p=2, and p=100 (p number of partitions). The contours are the mean spanwise vorticity, extracted with the slightly extended meanmode-module that I wrote. In the case of p=100, there are clear oscillations in some areas. I didn't observe any oscillations for velocity and pressure fields in any cases. So, I guess calculation of derivatives giving problems in highly-partitioned case. I was wondering what could be the reason for this. Thank you very much in advance for any feedback. Best, Asim On Mon, Nov 21, 2016 at 7:29 AM, Sherwin, Spencer J <s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk>>wrote: HI Asim, If you put all the files into a directory called meanField.chk/ The only thing you then need in this directory is a file called info.xml which tell you which partitions contain which elements. This can be generated from FieldConvert using <page98image37800.png> FieldConvert --nprocs 10 file2.xml file2.fld/Info.xml:info There should also be some details about this in the user-guide in section 5.5.1 (user-guide-4.3.2.pdf). I hope this helps. Spencer. On 16 Nov 2016, at 09:26, Asım Önder <asim.onder@gmail.com<mailto:asim.onder@gmail.com>> wrote: Dear all, Due to memory limitations, often I have to partition large data files using something like FieldConvert --part-only=100 config.xml fields/field_1.chk , and postprocess them piece by piece: FieldConvert -v -m meanstats config_xml/P000000"$p".xml fields/field_1.chk meanFields/meanField_1_p"$p".fld In the case of meanmode module for example, the output is a collection of many small files associated with partitions, which can be analyzed piece by piece. I was wondering if there is a way to merge the outputs back into one piece again like one piece fld file. Any feedback is appreciated. Thanks, Asim _______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk<mailto:Nektar-users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052 -- Asim <mean_spanwise_vorticity.png> Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052 -- Asim -- Asim Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052