Problem with last scalar fields (S8/S9) looking corrupted in ParaView output

[Pau Fradera]

Hello all,

I’m running a Nek5000 case with 9 passive scalars (so ldimt=10). The simulation itself seems fine:

• In the logfile all scalars have similar and reasonable min/max values (≈ 0 … 1×10⁻²). Moreover, all 9 scalars converge to the same residuals with similar amount of iterations.
• However, when I open the regular Nek output in Paraview, most of the scalars look totally fine but the last scalars (S8,S9) in the list look corrupted, with absurd ranges (+- 1e32), different than the ones in the logfile, and they look non-sense.
• I am observing the same problem in the files outputted by avg_all()
• I did some tests adding dummy scalars and the weird behavior of the scalars kept changing, sometimes it also reached the S7, but I have not been able to identify any pattern. So it looks like the “bad slot” shifts depending on how many scalars I have, even though the solver and per-scalar dumps seem fine.
• It may be relevant to mention that I am using neknek coupling with a half-turbulent channel (velocity only, no scalars there) to generate the inflow for the simulation.

Has anyone seen this issue where the last passive scalars look corrupted in ParaView, even though solver logs look fine? Could this be a known limitation/bug in the Nek5000 field writer or the ParaView reader when too many scalars are present? Any suggested fixes would be greatly appreciated.

Thanks in advance!

Pau

[YuHsiang Lan]

Hi Pau,

You should be able to visualize it if you choose the Nek5000 visit reader, the older one.

This is a known paraview issue that the "new" (5.12+) Nek5000 paraview reader doesn't support double precision.

and avg_all, by default, produces double precision checkpoint files.

  https://gitlab.kitware.com/paraview/paraview/-/issues/22747

The "new" reader also doesn't update the mesh, so it will not update correctly for ALE, or moving mesh in general.

  https://gitlab.kitware.com/paraview/paraview/-/issues/22890

Hope this helps,

Yu-Hsiang

--

[Pau Fradera]

Hi Yu-Hsiang, 

Thanks a lot for your reply. However, I am using Paraview 5.10.1, so not a 5.12+ version. I also checked the data that I am able to open in Paraveiw and it is indeed in double precision. So it seems the problem is something else. In that case, what would be your suggestion?

Thanks a lot again!

Pau

El dia dimarts, 26 d’agost del 2025 a les 16:18:12 UTC-7, YuHsiang Lan va escriure:

[Dillon Shaver]

Pau, 

How large is the file? I've encountered a hard upper limit of around ~2GB with paraview. Above that, the visualization is corrupted.

-Dillon

[Pau Fradera]

Hi Dillon,

Thank you for your reply. 

You were absolutely right, I splitted the output in smaller files and they are fine. Thanks for the help!

Pau

El dia dimecres, 27 d’agost del 2025 a les 6:20:22 UTC-7, Dillon Shaver va escriure: