Hi,
I experienced a segmentation fault in my LES simulation using the NBA option with sfs_opt = 1 or 2. When I turn off the option (sfs_opt = 0), the code runs to the end. Is there any known issue about NBA model ?
I didn't investigate so deeply the error since I may run without NBA model but here are some features of the bug :
- I am using WRF v4.5
- The code crashes almost 27 minutes after the beginning of LES (LES starts at 8AM on 2020-05-18)
- There is no clear error message in rsl.error.0000. The last lines are below :
Timing for main: time 2020-05-18_08:26:56 on domain 3: 2.41545 elapsed seconds
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
Backtrace for this error:
- I output the data every 5 minutes and noticed nothing wrong at 8:20AM but at 8:25AM there are NaNs in the whole domain except at the boundaries for all quantities (velocities, temperature, pressure, ...). In a previous test, NaNs appeared just before the output timestep and only a part of the domain was covered of NaNs.
- I output a time serie (tslist) in the center of my domains and all ts outputs stop at time 48.308643 which corresponds to 8:18AM (since my simulation start date is 8AM on 2020-05-16)
- When sfs_opt is disabled, the simulation runs to the end (11AM on 2020-05-18)
My simulation :
- is composed of 5 domains : 3 RANS and 2 LES, the crash occur on d04 (the first LES domains), RANS domains run normally during 2 days
- the comparison between with and without NBA is made with a restart at 7AM and only this option is changed
- NBA is used only on LES domains
- I use a timestep a bit higher than advised. I tried using a smaller time step in a former test and experienced the same bug
- My namelist.input is attached
I experienced a segmentation fault in my LES simulation using the NBA option with sfs_opt = 1 or 2. When I turn off the option (sfs_opt = 0), the code runs to the end. Is there any known issue about NBA model ?
I didn't investigate so deeply the error since I may run without NBA model but here are some features of the bug :
- I am using WRF v4.5
- The code crashes almost 27 minutes after the beginning of LES (LES starts at 8AM on 2020-05-18)
- There is no clear error message in rsl.error.0000. The last lines are below :
Timing for main: time 2020-05-18_08:26:56 on domain 3: 2.41545 elapsed seconds
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
Backtrace for this error:
- I output the data every 5 minutes and noticed nothing wrong at 8:20AM but at 8:25AM there are NaNs in the whole domain except at the boundaries for all quantities (velocities, temperature, pressure, ...). In a previous test, NaNs appeared just before the output timestep and only a part of the domain was covered of NaNs.
- I output a time serie (tslist) in the center of my domains and all ts outputs stop at time 48.308643 which corresponds to 8:18AM (since my simulation start date is 8AM on 2020-05-16)
- When sfs_opt is disabled, the simulation runs to the end (11AM on 2020-05-18)
My simulation :
- is composed of 5 domains : 3 RANS and 2 LES, the crash occur on d04 (the first LES domains), RANS domains run normally during 2 days
- the comparison between with and without NBA is made with a restart at 7AM and only this option is changed
- NBA is used only on LES domains
- I use a timestep a bit higher than advised. I tried using a smaller time step in a former test and experienced the same bug
- My namelist.input is attached