Hi,
I’m working on CROCO on NVIDIA GPU and encountered strange problem.
When I turn on both OPENACC and GLS_MIXING in cppdefs.h, it can be compiled with nvfortran, but when it runs with regional simulation with forcing either with bulk forcing or non bulk forcing, the model tke, gls, AKv, AKt goes very large in the entire water column, which produce enormous mixing, reducing surface temperature as a result. But this doesn’t happen with both MPI and GLS_MIXING without OPENACC. Also, if I activate both OPENACC and LMD_MIXING, instead it can compute with no strange mixing. Is Dr. Lars Umlaf’s GLS mixing scheme not ready for OPENACC?