MPI issue with Sphere example case

[xiongxin9000]

Hi,

I have been trying to run the example Sphere case, and it works with MPI 1 core. However, when I run with 4 cores, it seems that the solver tries to read the Python code generated initial and boundary conditions, and there are some distortions happens for the variable, probably because of the ghost cell? Please help with how to fix this issue?

Thank you.

Regards,
Xin

[mathrack]

Hello,

This example has not been tested for some time. We will test it as soon as possible to evaluate the situation.

Please note that the sphere example is based on the Case-Sandbox.f90 and on the Sandbox python toolkit. The files needed to perform the simulation should not depend on the number of MPI ranks.

Cheers,
Mathrack

[xiongxin9000]

https://github.com/xcompact3d/Incompact3d/discussions/359

And YES, it only works for iibm=1

[mathrack]

As a side remark, please note that it might take a significant time for us to test and debug the case. I suspect you can find a version of the code working for the sphere example if you rewind the git history. This case was probably working in July 2024 : https://github.com/xcompact3d/Incompact3d/commits/master/examples/Sphere

[mathrack]

I think the case was initially working with iibm = 2 (provided input file). Maybe @gauravity30 can confirm.

[xiongxin9000]

Thanks for answering. I am testing this case because I want to replace the STL file with my own geometry. However, when I test for flow around a simple square, it gives me results like the figure. I changed the ifilter and c|_filter, Re number, grid points parameters, but the issue still exists.

And the animation is my test results for iibm=1, mpi 1 core and flow around a sphere
https://github.com/user-attachments/assets/5e832a35-3716-4329-8fd8-b4ed55b46eaa

[gauravity30]

I think the case was initially working with iibm = 2 (provided input file). Maybe @gauravity30 can confirm.

Hello Everyone !
Yes, when I had pushed the commit for force-calculation, it was working with ‘iibm=2’.

[xiongxin9000]

Hi xcompact3d team,

I tested on the older version; the sphere cases run properly. Thank for your time

[mathrack]

Hi xcompact3d team,

I tested on the older version; the sphere cases run properly. Thank for your time

Hello,

Thank you for the feedback. To clarify the issue, please identify both versions tested. I guess the broken version is the current master (abb010e). Please identify the working version and provide

1. the input file used
2. the diff between the source code you were using and the working version available in the git history

Cheers,
Mathrack

[xiongxin9000]

Hi,
The broken version is the current version, abb010e,
The working version is b7f3a21
The input file for the working version is input_work
The input file for the broken version is input_broken
The diff is attached.

input_work.txt
input_broken.txt

diff.txt