Grid generating failure with parallel running

[dhtjtks]

Hello ROMS users,

I have a problem with parallel running. If I generate evenly spaced grids, it works good. But if I generate un-evenly spaced grids, it doesn't work for parallel running with MPI. What I figured out my problem is this. If I have 10 by 10 grids, and the grids are evenly spaced in y direction and grids in x direction have different sizes. And if I set 'NtileI == 2' in '**.in' file to use 2 nodes, 'x_psi' shows like this: 0 1000 2000 3000 4000 5000 500 1500 2500 ... It means that the 'i' notation restart from 0 when it reach to half of grid number. That's why my grid is not working properly.

I am using 'ana_grid.h' file to change or create my grids. So what I am thinking is that if I can compile grid generating part with single process and compile other parts with multi-processing, it may works. Usually I am using 'make -j 4', 4 nodes, when I compile. Does anybody can help me?

Any help will be appreciated.

Jung

Thank you.

[kate]

Maybe if you show us what you are doing in ana_grid, we can tell you how it should be. Follow the examples given and see how they work in parallel (which they should).

[dhtjtks]

Maybe if you show us what you are doing in ana_grid, we can tell you how it should be. Follow the examples given and see how they work in parallel (which they should).

Thanks Kate. I attach a pdf file. So you can check my problem. I will wait for your comment.

Jung