open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,
I've been working with this huge mesh and this error keeps appearing.
It took 60 days of simulation for the error appear in parallel, but now, since i increase the refine in the oceanic border, the error appears at the first time step.

my initial and boundary conditions are well defined.
What could be causing it?

---

PROPAGATION AND DIFFUSION WITH WAVE EQUATION
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 1000 RELATIVE PRECISION: NaN

---

CALCULATION OF VERTICAL VELOCITY STEP

---

ADVECTION-DIFFUSION OF TRACERS
MURD3D OPTION: 5 1 ITERATIONS
DIFFUSION OF TAN1
GRACJG (BIEF) : SOLUTION X=0 BECAUSE L2-NORM OF B VERY SMALL: 0.000000

Hello,

It can be anything. If it happens at the first time step probably a value that is not initialised in your data, or crushed elements in your mesh without treatment of tidal flats. Compiling with tracing options would give you the very line where the first NaN occurs (with Intel compiler : -debug all -traceback -ftrapuv), I keep a version like this here, but it is not computationnally efficient, just for debugging.

Another test that you can do is putting subroutine BIEF_ALLVEC in your Fortran and replace the line:

XMAX=HUGE(100.D0) by XMAX=0.D0

If it works it is a proof that it is a value not initialised (we initialise all double precision arrays allocated in BIEF_ALLVEC with nearly NaN to trigger a crash if an array is not initialiesd, this is Quality Insurance...)

With best regards,

Jean-Michel Hervouet

the changes in bief_allvec didnt solve the problem.

Although, with your comment we were able to determine that the problem was with the initial condition of the tracer (my case temperature).

Now i'm looking for new sources of data.


Thanks Jean-Michel.