Question: causes to get 'core' files
-
modzy
Question: causes to get 'core' files
I did a successful mpi parallel, barotropic model run with a setup of a contant, say 500 m, bathymetry over the entire grid. I then doubled the number of vertical layers, say from the original 20 to 40 layers, and ran the model with reduced time step. This time after about two days of simulation, the run blew up which created a lot of cores files. However, it didn't save any 'nan' data into the perfect restart file as ROMS udually does in case of a blowing up due to nan. What might be the cause of the blowing up? Thanks!
Re: Question: causes to get 'core' files
ROMS is checking for unphysically large values for rho and velocity, as well as checking for things like NaN. It will report, then kill the job if it finds a problem.
Like ROMS, the Unix/Linux kernel is checking for certain conditions. If it finds an ill-behaved program it will kill it and generate a core file (or many). Many people find these core files unhelpful and limit their size through the coredumpsize limit. Instead, you can recompile your code with the USE_DEBUG switch and potentially find out where in ROMS your job died by using a debugger on the core file. There's more than one way your program can by ill-behaved from the point of view of the kernel and I don't know which off hand.
Like ROMS, the Unix/Linux kernel is checking for certain conditions. If it finds an ill-behaved program it will kill it and generate a core file (or many). Many people find these core files unhelpful and limit their size through the coredumpsize limit. Instead, you can recompile your code with the USE_DEBUG switch and potentially find out where in ROMS your job died by using a debugger on the core file. There's more than one way your program can by ill-behaved from the point of view of the kernel and I don't know which off hand.