[springcold]

I have tried to do tutorial 2 on the SU2 website, using v3.0 release. There wasn't any compilation problem but the results was wrong when invoke in parallel.



I have tried using:
- master branch and 3.0 release
- using another machine

But still produce incorrect result. Wonder anyone had similar issue?

[awendorff]

What system are you running on? Can you provide more detail into what incorrect results you are encountering? Do you have any pictures that show the errors you are experiencing?

Andrew

[springcold]

It was done on intel-based system, compiled with icpc and mpicpc, metis 4.0.3.

The problem is there's no errors appearing when I compile and run the instance. However, when I output surface_flow.vtk on Paraview, it is different from serial, as shown in the image above.

The console output file here, if you need it for understanding.
parallel.txt

Thanks!

[awendorff]

Can you attach the plots. If it is working in serial, but not parallel, you might want to configure it again in parallel making sure you have everything you need.

[springcold]

@awendorff: Not sure if you meant the mesh file when you said plots. I used the mesh file from the Testcase provided by Stanford. ONERAM6 mesh

In addition, the configuration I used is
./configure CC=icc CXX=icpc --with-MPI=/path/to/mpi/mpiicpc --prefix=/path/to/install/ --with-Metis-lib=/path/to/metis-4.0.3/ --with-Metis-include=/path/to/metis-4.0.3/Lib/

I have tried using gnu compiler, openmpi and mpich. There was no compilation error (Iteration error included). In all the instance, the same output was seen in Paraview..

[economon]

Hi guys,

Just a note that the slight difference you see between the serial and parallel in ParaView is simply due to some additional information being in the parallel version of the file. In the parallel version, we happen to store the x, y, & z coordinates of the mesh nodes as well, and by default in ParaView (for a reason that I am not familiar), it defaults to viewing the x coordinates when you load the file.

However, note that all of the same fields should still be there under the pull-down menu for the different available fields, and you should find the same types of data such as Mach, Pressure Coefficient, etc. In the attached image, I have changed over to the pressure coefficient for a parallel file. Lastly, remember that you might not see exactly the same numbers for the values due to the slight differences between serial and parallel execution (but all of the same fields should appear).

Hope this helps,
T

[springcold]

That solves the problem. Thanks!

[economon]

No problem! As a quick follow-up: we have adjusted the output so that this visualization discrepancy shouldn't occur in future releases.