Should write_tasks_per_group be the same as LAYOUT_X or LAYOUT_Y in predef_grid_params.yaml?

[lo-y-wni]

Hello everyone:

I have a question about IO quilting option. I know the total processors = LAYOUT_X * LAYOUT_Y + write_tasks_per_group * write_groups. Suppose write_groups is 1, I want to use total 144 processors and hope LAYOUT_X=14 and LAYOUT_Y=10, then write_tasks_per_group must be 4 at this configuration. But I don't know is the configuration illegal or not since the predefined domain are almost let write_tasks_per_group=LAYOUT_Y or write_tasks_per_group=LAYOUT_X.

Yi-Chuan

[gspetro-NOAA]

Hi Yi-Chuan,

I've reached out to @EdwardSnyder-NOAA, and he is looking into this for you. He should get back to you within a few days.

Best,
Gillian Petro, Advanced User Support (AUS) Team

[EdwardSnyder-NOAA]

Good afternoon Yi-Chuan,

The configuration you described isn't illegal. The LAYOUT_X and LAYOUT_Y variables (number of MPI tasks to use in the horizontal x and y directions) are independent from the write_tasks_per_group variable (number of MPI tasks to allocate for each write group). I'm not sure why the predefined grids have some of these variables equal to each other. But you should be able to run the model with this configuration.

Please let me know if you have any additional questions! And I'll address your other question soon!

Thanks!

Eddie Snyder | Advanced User Support (AUS) Team

[lo-y-wni]

Hi Edward:

Thanks your response! It's very happy to hear write_tasks_per_group doesn't need to be the same as LAYOUT_X or LAYOUT_Y. It lets me tune grid decomposition more flexible!

Yi-Chuan