timestep is HUGE

[Antonella Succar]

Dear Basilisk users,
 I am currently using basilisk to run an 3D test case. My code is running well with high performance computing until recently, when I installed basilisk on a new cluster, after running for several timesteps, a timestep dt=HUGE occurs, ending the simulation:
Timestep i=1096 at time t=1024.946470; dt=1.147455
...
Jet tip xMin= -22.153132, velo_x_f=-0.000165, out_val=-0.000165, at iteration 1098 and time 1027.286303 on process 56
Timestep i=1098 at time t=1027.286303; dt=1.227044
Jet tip xMin= -22.153323, velo_x_f=-0.000165, out_val=-0.000165, at iteration 1099 and time 1028.513346 on process 56
Timestep i=1099 at time t=1028.513346; dt=1.270474
Jet tip xMin= -22.153521, velo_x_f=-0.000165, out_val=-0.000165, at iteration 1100 and time 1029.783821 on process 56
Saving movie at timestep 1100
Jet tip xMin= -22.153521, velo_x_f=-0.000165, out_val=-0.000165, at iteration 1101 and time 1000000000000000019884624838656.000000 on process 56

I am including the following headers:
#include "grid/octree.h"
#include "navier-stokes/centered.h"
#include "two-phase.h"
#include "tension.h"
#include "output_vtu_foreach_0.h"
#include "tag.h"

By placing several printfs in different events, and not seeing the corresponding output at i=1100, including for the stability event in navier-stokes/centered.h, it seems that the stability event is not running at i=1100 (among other events), which may imply that the default DT value is used (HUGE). This behavior appears very odd to me, especially that the same code works perfectly well on several other clusters.

Has this happened to someone before?  

Thank you!

Antonella

[Haochen Huang]

Hi Antonella,

You can use CFLAG=events to print the execute order of events. The error time step is always related to events named "stability" not only in centered.h but also to those in tension.h, vof.h etc. 

The very last time similar problem (wrong time step) occurred to me is when I use 'restore', the execute order of events is misplacing due to different name of events that initiates the computation ('init' in centered.h but 'initial' in my file). Thus 'stability' has to cope with NULL data and cause error timestep. But if I just run the same program in one shot without using 'restore', the output is perfectly correct.

I hope you find this helpful.

Best regards,

Haochen