Memory leak due to source term???
Simulating a rising bubble I've implemented a moving reference frame similar to the approach outlined in Henrik Rusche's PhD thesis.
To take the moving reference frame into account, I've added a "source" term to the UEqn:
rhoaF = rho*aF; // density*acceleration (vector)
+ rhoaF //---------------------------acceleration term!
+ fvm::div(rhoPhi, U)
- fvm::laplacian(eta, U)
- (fvc::grad(U) & fvc::grad(eta))
rhoaF is a volVectorField. I've also tried adding rho*aF directly to the UEqn, which seems to make no difference.
At the begining of a simulation a bool is read from the solution dictionary to decide whether the reference frame has to be moved. If the reference frame is static, aF is set to (0 0 0) so UEqn remains unchanged.
Now if I simulate with a static frame, everything is fine. When I change to the moving reference frame mode, the results are reasonable, but working memory rises constantly. I've run my code with valgrind but couldn't find any errors. Moreover, there are no additional variables declared when I run the code with moving reference frame.
At the moment I'm pretty perplexed... Any advice would be welcome!
OpenFOAM Memory Leak?
I'm simulating a LES case using pisoFoam with a source term added. The memory requirement increased constantly from 2G to 8G after I ran 4E4 steps. I don't know what is the reason. May I have some advice from you?
Appreciate your help in advance.
|All times are GMT -4. The time now is 19:50.|