open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear All,
I have been simulating hydro-morphodynamic changes (using TELEMAC2D v7p2v0 parallel) of a 16 km river reach of 300,000 mesh elements for 3 years. 216 processors for 01 hours can approx. simulate 72 hours (3 days). With the same rate, I will need 02 weeks for simulating 03 years. It looks huge time since I am calibrating my model

I don't know whether it is normal simulation time or someone has a different experience? Anyone can please share his/her experience?


I am thanking you in advance for your support!
Ateeq

Hi
It's hard to say if it's huge or not has we don't know the model...
You use a personal fortran file, are you sure there is no time loss in these code?
To reduce time, you could change some parameter but is it interesting for your calibration?
For example, k-epsilon for turbulence has a cost...

As far as I know, 300 000 elements means around 150 000 nodes and my experience would say that 216 processors could be too much. For me the optimum would be 1000 nodes per processor so around 150 procs... Maybe you could try to check this point also but this will not reduce the global computation time...

Hope this helps

Dear Coulet,
I am using TYPE OF ADVECTION=1;5 in my simulations. 1: method of characteristics for the velocities while 5: PSI distributive scheme for depth. I am also using SUPG OPTION=2;2. 2: Upwind scheme with the modified SUPG method used to make the solution smoother. I am also using a variable time step.

Actually, I don't know the theoretical background of stability criteria of these schemes. With this configuration of schemes, may I exceed courant number more than 1.0?

Thanks in advance!
Ateeq

Hi
Characteristics scheme should allow you to have courant number greater than 1...
Take care about the variable time step as this could be driven by a local effect not necessary located in the main area of interest...

Regards