search

ufs-community / UFS_UTILS

Utilities for the NCEP models.
Other
22 stars 108 forks source link

cpld_gridgen utility too slow for high resolution ATM cases #760

Open DeniseWorthen opened 1 year ago

DeniseWorthen commented 1 year ago

Currently the cpld_gridgen utility runs serially, which is adequate for the current resolutions (highest is C384, 1/4 deg ocean/ice).

When creating the weights used to map the 1/4 degree ocean mask to high resolution tiles (> C384), the execution is too slow. I'm currently working on a solution which will run on multiple PEs for the weight generation for the high resolution ATM cases.

DeniseWorthen commented 1 year ago

@GeorgeGayno-NOAA Please assign this issue to me. Thanks.

GeorgeGayno-NOAA commented 6 months ago

@DeniseWorthen - This issue has been open a long time. What is the status?

DeniseWorthen commented 5 months ago

@GeorgeGayno-NOAA I have worked on this a bit, but I ran into two issues---one was in creating a mesh from the C3072 tile files and I was waiting for a later ESMF before attempting again. That prevents even the current code from running (given enough wall clock). The second issue is even w/ the mesh created, I will need to rewrite that section of code, since I'm fairly certain at this point that the call to ESMFRegridWeightGen always runs serially 'under the covers'.

I can work on it in the background, if you would like to clean up old issues. I'm not sure whether any ATM resolution > 1152 will be needed for the coupled cases in any case.

GeorgeGayno-NOAA commented 5 months ago

@GeorgeGayno-NOAA I have worked on this a bit, but I ran into two issues---one was in creating a mesh from the C3072 tile files and I was waiting for a later ESMF before attempting again. That prevents even the current code from running (given enough wall clock). The second issue is even w/ the mesh created, I will need to rewrite that section of code, since I'm fairly certain at this point that the call to ESMFRegridWeightGen always runs serially 'under the covers'.

I can work on it in the background, if you would like to clean up old issues. I'm not sure whether any ATM resolution > 1152 will be needed for the coupled cases in any case.

Thanks for the update. I don't have a deadline for cleaning up old issues. That is up to the developer. I just wanted to know if I should keep this issue open (which I will).