open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi

I'm trying to program subroutine METEO to include a time varying wind which is constant troughout my entire domain.

I'm using formatted data file 1 in channel 26 to include wind data.

I'm attaching formatted file + princi.f

I get the following error code (scalar and parallel) at the beginning of first time step :

DIFFUSION-PROPAGATION STEP
GRACJG (BIEF) : SOLUTION X=0 BECAUSE L2-NORM OF B VERY SMALL: 0.2764187E-67

---

BALANCE OF WATER VOLUME
VOLUME IN THE DOMAIN : 0.1677428E+11 M3
FLUX BOUNDARY 1: -0.000000 M3/S ( >0 : ENTERING <0 : EXITING )
FLUX BOUNDARY 2: -0.000000 M3/S ( >0 : ENTERING <0 : EXITING )
FLUX BOUNDARY 3: -0.000000 M3/S ( >0 : ENTERING <0 : EXITING )
FLUX BOUNDARY 4: -0.000000 M3/S ( >0 : ENTERING <0 : EXITING )
FLUX BOUNDARY 5: -0.000000 M3/S ( >0 : ENTERING <0 : EXITING )
RELATIVE ERROR IN VOLUME AT T = 300.0 S : 0.000000
MAXIMUM COURANT NUMBER: 0.8471072E-71
TIME-STEP : 1.000000
_____________
runcode::main:
/media/sf_host/39566/wind_test:
|runCAS: fail to run
| /usr/bin/mpiexec -wdir /media/sf_host/39566/wind_test/cas5.txt_2013-01-15-03h04min45s -n 4 /media/sf_host/39566/wind_test/cas5.txt_2013-01-15-03h04min45s/out_princi


What does this error means??

From reading in other posts it seems that i should use binary file instead for temporal varying winds ??? Am I correct ??? Is there any other way around with the formatted file???

Thanks again

Mathieu

Hello,

This is not an error, it just means that you solve a linear system with a right-hand side which is zero, and the program just tells you it will give a solution equal to zero, without even trying to invert the matrix. This message sometimes happens at the beginning of a computation when there is no velocity and a flat free surface. If the programs stops, it is not for this reason, I would advise that you try to run your case without the specific meteo subroutine, and then if it works to insert prints in your subroutine meteo to see if it is OK, I would suspect that something crashes there, though the listing does not mention any error.

Regards,

JMH

Hi Jean-Michel,

Thanks for reply

I tried running the simulation without meteo subroutine. The program doesn't crash.

So as you said, my problem remains in the my princi.f file.

I'll try debugging with prints.

My final question
Is it possible to simulate (parallel) variable winds using the formatted file instead of binary file ???

Thanks

Hello,

You can also use the formatted file, keeping in mind that it will be copied as many times as processors (see explanations in the header of telemac/V6P2/telemac2d/tel2d_v6p2/lib/telemac2dv6p2.dico, this file is declared as PARAL, if you want a single copy change the dictionary and put SCAL instead of PARAL in the line that starts by "SUBMIT = 'T2DFO1...."). In all cases all the processors will have the complete file at their disposal, and you will have to give only the winds of the sub-domain, depending on what is in this file it may happen to be tricky.

Regards,

JMH

Great

thanks Jean-Michel for helpful comments,

I've corrected the fortran file...i needed to save a few variables before they died at the end of the subroutine.

From what i understand :
using formatted file is ok if wind is constant troughout the entire domain (but still variable in time). If wind is variable in both time and space, using the binary file is more appropriate.

Thanks again

Hi Mathieu

I was just trying to run my model including a text file of the wind that varies with time. I had some small problems in my fortran file and I found your code on the Forum, but you said that there are some variables that you need to save. I was just checking the code and it looks fine to me, could you let me know which are these variables.

Thanks

Viviana

Hello Viviana,

This previous code had two problems in it.

1. Wind speeds (Veast,Vnorth) were initialized as REAL instead of DOUBLE PRECISION explaining why i was getting the SOLUTION X = 0 remainder.

2. The modified subroutine is called every iteration. When the subroutine finishes, the initialized variables (TV,VX,VY,TV0,VX0,VY0) die. Next iteration, the code could not remember their previous values so i was getting a crash at 2nd iteration. The lines i have added at the beginning are :

LOGICAL DEJA
DATA DEJA /.FALSE./
DOUBLE PRECISION :: TV,VX,VY,TV0,VX0,VY0
SAVE DEJA, TV,VX,VY,TV0,VX0,VY0

I'm attaching the updated code here.

Take note that i do not interpolate data between timesteps.

Have fun

Mathieu

Thanks so much for sharing the code with me. I will try and see how it goes this time.

Thanks again

Viviana