issues in saturne/saturne coupling with v8.0

[lb_yuan]

Hi developing team,

I recently encountered issues with saturne/saturne coupling in version 8. I noticed there are some changes to the related source codes compared with version 7, such as in csccel.f90, csc2ts.f90, etc. The changes look fine to me, and I am not sure if the issues are caused by these changes, or may be caused by lower-level ple functions. I created simple cases for testing (with both face and cell coupling used), and I have attached them to this post. Basically, the face coupling looks fine, but the cell coupling seems to not work properly. The test cases include both v7 and v8 versions with the same setups. You will see v7 works perfectly, but v8 does not. Please take a look at them at your earliest convenience. Many thanks.

Best regards,
Bo

[Yvan Fournier]

Hello,

Thanks for the report and test case. I'll try to look.at this in temhe next few dans.

Best regards

Yvan

[Yvan Fournier]

Hello,

I actually also obtain divergence at the second time step even in the v7 version. on my machine (starting with divergence in the iterative scalar gradients).

Which release of version 7.0 are you using ? Looking at this through a debugger, i see an issue in the PLE exchange itself, which may be related to a bug fix between v7.0.4 and 7.0.5.

If this is the cause of the issue, I need to look at this in more detail.

Info on that number may help, in any case.

Also, the issue is related to handling on unlocated cells. The previous behavior would cause a shift in values found.
Now, it assumes the array passed matches the total requested number of cells, but for cell couplings, the value allocated matches the filtered (located) cells, so in all cases the issue regards handling of unlocated cells, which is not well defined here, due to multiple indirections (first indirection regarding selected cells, second regarding located cells).

So if you use selection criteria such the all cells are located, you should be able to work around the issue. pending a better fix.

Regards,

Yvan

[Yvan Fournier]

Update : this has been fixed in the master, v8.0, and v7.0 branches.

The Git repository mirror has been.synced on GitHub.

Best regards,

Yvan

[lb_yuan]

Hi Yvan,

Many thanks for your reply and prompt action on this.

The version that I tested without a problem is v7.0.4, and I did not test other v7.0 versions beyond this. So, I guess, you are right, the issue may be related to the bug fix between v7.0.4 and v7.0.5.

With your fix to the bug in the github branches, I can confirm that the cell coupling runs ok now for both v7.0 and v8.0.

Best regards,
Bo