open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello everyone,
Does exist any tip to compute a rough estimate of the necessary CPU time for a given T2D simulation? A method which involves duration, time step, number of listings, turbulence model, mesh parameters, number of cores ant their characterstics, etc.?
Any remark or suggestion will be appreciated
Greetings

Hello

I'm not sure I've understood your question.
At the end of any computation of Telemac-2D the CPU time is written.
Do you need other specific features?

kind regards
Riadh

Thanks Riadh,
I ask for an estimate before the run. Because we have a cluster of 512 cores but we are many researchers to work with. I need to submit my simulations plan and give rough estimates of the expected CPU time for each simulation. I've seen in a paper of Moulinec that relation between core numbers and CPU times is almost linear, hoxever this is not true in my case. My question is : is there a empirical or pther formulas between CPU time and the : 1-Mesh size (nodes,elements) 2-Duration, 3-Time step, 4-number of cores and theirs characteristics 5- Turbulence model, 5- Dry areas, 6-Number of listings ?

Hi,

I can share a bit of my own experience.
I am maybe wrong but there is no such method to estimate the run time prior to running it since it depends largely on your system characteristics.
I also ran jobs on a HPC cluster (Hartree center in UK) and I found that the scaling factor was not linear after a certain number of cores.
The best method I found was to run a short test job using same set-up and same number of cores than the final run and then scale it up to the actual DURATION. I also use the same method for jobs on my own machine...

Good luck!

Best regards
PL