open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

I'm running into an issue with TELEMAC3D (v8p1r1) when using great than 40 CPU's with a binary boundary data file. I have run the model with TPXO tidal forcing only, with up to 200 CPU's fine. However, when I add HYCOM binary boundary [eta, U, V, Temp, Salinity] files to the model, it will run fine up to 40 CPU's but fails with great than this number.

The model produces the following error: "GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 60 RELATIVE PRECISION: NaN"

Thanks

Attached (hopefully) model files

Attached HYCOM binary boundary

Hello Glen,

There is a bug in the reading of BOUNDARY DATA FILE in v8p1r1 (and previous releases), solved in the trunk version. Can you try the enclosed fix (2 files) and tell me if your issue still occurs please?

Hope this helps,

Chi-Tuan

Hi Chi-Tuan,

I have add the new file you attached and recompiled the code. However, I'm still getting the NaN error when the model uses greater than 40 CPU's.

Regards

Hi Chi-Tuan,

Update: I downloaded and compiled the trunk just to check to see whether the binary boundary issue was still present. I also increased the maximum number of iterations to 500. The code runs but still produces the "GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 500 RELATIVE PRECISION: NaN" error. I also turned off all the tide forcing (TPXO inputs), effectively only using the a binary boundary as an input and the error is still present.

Regards

Hello Glenn,

Sorry I forgot to answer to your last post 2 weeks ago.
I had a quick look at the source code and I think it is too be deeply investigated because of the definition of sizes of arrays.
This feature is quite experimental and not so validated I think. That should be the reason why it was not so documented.

Anyway, if possible, I would advise you to stop using this feature temporarily. You will have to wait for a good fix (I do not know when I could have time to look at it, a colleague or someone else in the TELEMAC community).

Sorry,

Chi-Tuan

Hello Glenn,

Not so much time to investigate your issue, but today was OK.

Can you try the enclosed fix in the archive hg_hycom_test_dbg.tar and tell me if it is OK for you?
I tried with 224 cores + TPXO global solution (too long to download Pacific Ocean at home) and I did not have any nan during the first 10 time steps (with the trunk I got some with only 56 cores as you may have had).

As there have been a lot of fixes for TELEMAC-3D since release v8p1r1, you can also try the trunk release with the enclosed bord3d.f tentative fix.

I hope this helps,

Chi-Tuan

PS: sorry in advance if you do not have an answer from me during the next weeks, I am starting my holidays. :)

Hi Chi-Tuan

I can download the bord3d.f file fine but it seems I require some sort of permission to download the attached *.tar file

Forbidden

You don't have permission to access /media/kunena/attachments/438/hg_hycom_test_dbg.tar on this server.

Thanks

Glen

Sorry Glenn,

I thought .tar was supported by this forum.

The split files, I let you reorder them.

Chi-Tuan