open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello Juli,

Actually, LIMI3D subroutine is only called by TELEMAC3D_INIT which is called once by TELEMAC3D subroutine. Its goal is initialisation as written in the headline.

In user_bord3d_work1.f there is at least one issue: ITRAC is not initialised in the subroutine (write ITRAC = 1 if it is the 1st tracer).

With user_bord3d_work2.f do you use the modified LIMI3D sent 8 days ago?
The main issue is the loop to do for surface nodes. With this implementation, you only sweep lateral boundary nodes note every node of 2D plane corresponding to the surface plane.

Hope this helps,

Chi-Tuan

Hello again,

The boundary condition used for THERMIC model is of type Neumann:
dT/dn = A * T + B
See the WAQTEL theory guide and equation 6.9.

So B corresponds to BTABOS array in the source code (1st B for B, TA for tracer, BO for boundary and S for surface).

Hope this helps,

Chi-Tuan

Hello
Yes, I was trying to mix the subroutines and see what came up.
With lateral boundary conditions is possible, I finally got it. I will check everything that you told me because I was seeing the waqtel subroutines and understanding how is management the exchange, especially if they used terms such as LITABS and TABORS, but does not.
I am still trying but in user_tr3d.f

the exchange with atmosphere

Finally with USER_TR3 is working for small domains with LITABS and TABORS with big domains is not working in parallel.

Hello Julie,

Can you send your USER_TR3 subroutine + give the number of 2D elements of the geometry file + number of horizontal planes + number of cores you use when running in parallel please?

Chi-Tuan

Hello Dear Chi-Tuan

Here is the FORTRAN file:

Number of elements 2D: 404959
Number of planes:23 mesh type 2.
Number of cores: For testing 32 and if everything is working 324.

Thanks a lot.

Can you explain in more details what you mean by "not working in parallel"?

This is the error:

I tried with 1, 2 and 3 more than that is not running.
I decreased the time step, and the number of cores

In such cases, if decreasing the time step is not enough, you have to investigate by writing results just before crashing (stop the computation the last iteration before crashing) and look at the results. It may help you to understand which process can lead to your error.

You can also try to deactivate some processes just to check that simple ones are OK (or standard TELEMAC is OK), e.g. hydrodynamics without tracers or Fortran files, hydrodynamics with tracers but without Fortran files. What about the computation if not including your modified USER_TR3?

Hope this helps,

Chi-Tuan