Limits in setting n1 and n2

[susanhen]

Hi,

I have been testing both HOS-ocean and HOS-nwt and tried to increase the resolution of the domain. It appears that there are limits of n1xn2. With gfortran, I get a compilation error, with ifort it compiles but gives a segmentation error. Do you have an explanation for that? Is it possible to run HOS-ocean with a grid 1024x1024 and if so, how do I need to proceed?

Thank you for your help!
Best regards
Susanne

[gducrozet]

Dear Susanne,

The limitation you encountered is associated to the memory allocations in fortran. Usually, fortran does not like to pass arguments of very large size (usually above 2GB). It is possible to overcome this limit by using some compiler options -mcmodel=medium or -mcmodel=large (remind that all libraries need also to be compiled with this option). But you will probably be limited by the hardware you use at a given point.
Note that the total number of grid points is also linked to the choice of HOS nonlinear order M (through the concept of dealiasing). You can possibly use 1024x1024 with a small HOS order. By the way, you also have to keep in mind that with such discretization, the computations can take quite a long time: it is then necessary to think what is the necessary discretization in your problem.

Finally, we are running larger discretization than this one by using MPI implementation of both codes. I hope to find some time to release those versions open-source in the future... Maybe in a v2.0 with some new features available!

Hope this helps!

Best regards,
Guillaume.