A problem has been found (with thanks to Sebastian Scher, MISU, Stockholm University), where if the intel compiler (all versions) is used and the stochastic physics is enabled in the model, the model will crash with a floating point exception, divide by zero.

The issue is related to the optimization the intel compiler makes regarding if-then-else blocks.

The solution is to modify the intel compiler options as shown below. This fix will be included in future releases.

Modify make/cfg/intel-*.cfg files

For each intel-*.cfg file used, edit the file and replace:

$OIFS_FFLAGS{?} = -g -m64 -openmp -O1 -fp-model precise -convert big_endian -traceback

with

$OIFS_FFLAGS{?} = -g -m64 -openmp -O2 -fp-model precise -fp-speculation=safe -convert big_endian -traceback

adding the extra option: -fp-speculation=safe.  The optimization level can also safely be raised to -O2.