[_ron_]

Hello,

I am using SU2 for simulations at transonic Mach numbers. I have encountered an increase in time per iteration for large problems (50M+ cells). For example, I found that at the beginning of the simulation it took 20s per iteration, but after about 1000 iterations it took 50s per iteration. I suspected that the number of iterations needed for the convergence of the linear solver increases with time, so I exported LINSOL as screen output. However, I found out that at all times the linear solver converged after 6 iterations.
It is worth mentioning that when I restart the simulation from previous iteration, the time per iteration reverts back to 20s from 50s.
Has anyone encountered this problem before?

Thank you.

[pcg]

No this seems odd. Can you share the config, version of the code, etc. Can you monitor how much memory is being used and if it increases in time?

[_ron_]

Documents.zip

Originally Posted by pcg

No this seems odd. Can you share the config, version of the code, etc. Can you monitor how much memory is being used and if it increases in time?

Hi, of course, the input file and the corresponding output are attached. You can see that the time per iteration increases dramatically. As I mentioned before, when I restart the simulation the time reduces back to normal. Lately I found out that when I run an unsteady simulation, the time reduces back at each time iteration, namely it only increases with inner iterations. The output of the unsteady simulation is attached as well. I used version 7.5.1. I did not check the memory usage during the simulation, but it stands close to 50% at most times (I run the simulation on a shared memory machine).

[bigfootedrockmidget]

Do you have the su2 mesh as well?

[_ron_]

Originally Posted by bigfootedrockmidget

Do you have the su2 mesh as well?

I am sorry, it is not available to me right now, but it happens at a few geometries if the mesh is adequately dense (say above 50M cells).