open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello Rafia,

Unfortunately, the issue of discrepancies between serial and parallel simulations is a well-known one in the TELEMAC system. That said, the resulting error is typically very small and therefore manageable.

From what I understand of your results, the error is less than 1E-4 over a volume of 500,000 m³. That’s actually very good, and so far, the TELEMAC team hasn’t found a way to improve this further. Additionally, you're displaying elevation values around 0.000289, which is below a millimetre – predicting to the centimetre would already be quite fortunate!

Given this, I’m not sure why this is causing significant delays in your research. These kinds of differences are not something you should be concerned about.

Hope this helps,
Sébastien.

Good morning Mr. Bourban,

Thank you very much for your reply. I completely agree with you that the discrepancies in these results remain small. However, these values correspond to the results of a TELEMAC-3D example, the Vent case, which is highly simplified. In contrast, when I run my simulation in the Strait of Gibraltar, where the spatial resolution is much higher, the discrepancies between the parallel and sequential versions become significantly larger.

For further clarification, I have attached the results of both simulations: one using the sequential version and the other using the parallel version with only 2 cores. The attachments include screenshots of the listing files and attribute tables showing the velocities at an arbitrary point on the surface.

Thank you in advance

I apologize, please disregard the attachment in the previous post, as I mistakenly uploaded the wrong figures. In this post, I am attaching the correct ones.

This could be great if you could share all the files required to run your case because in your case the differences are significant but this remain strange for Telemac.
Nevertheless, without the case, it's hard to investigate and help you
Regards

Good morning,

Thank you very much for your response. Would you mind sharing your email address so I can send you the files related to my case?

Best regards,

It depends on the file size but you could join a zip of all the file here or give a link to a shared drive.

Greetings,

Please find below a Drive link to a folder related to one of the simulations I'm working on for the Strait of Gibraltar. I’ve also attached screenshots showing the file listings and attribute tables for two simulation runs on the cluster—one without parallelization and one using parallelization with only 2 cores. I hope this helps provide an overview or offers some clues about the issue.
Thank you in advance.

Link: drive.google.com/drive/folders/1pPOVofre...Y9z6wvY2?usp=sharing