search

NOAA-EMC / WW3

WAVEWATCH III
Other
254 stars 501 forks source link

Memory Leak #579

Open MatthewMasarik-NOAA opened 2 years ago

MatthewMasarik-NOAA commented 2 years ago

Describe the bug A memory leak in WW3 was detected while running as component of UFS. This was reported in UFS issue #779.

To Reproduce

Expected behavior Memory use will increase during the simulation.

Screenshots Reference plots in UFS issue #779.

Additional context N/A.

MatthewMasarik-NOAA commented 2 years ago

@DeniseWorthen I'm currently looking at memory use for running WW3 standalone, vs. coupled with the current wmesmfmd cap. This has me curious how the vmsize behaves with your new mesh cap. Have you by any chance looked at the memory profile for WW3 w/ the new cap? I believe you said there are no RT's that support the new cap currently.. if there is a canned case I could run I'd be happy to do that and look at the memory use.

DeniseWorthen commented 2 years ago

@MatthewMasarik-NOAA No, I have not done any profiling with the mesh cap. There are no standalone RTs using the mesh cap. It is only being used in the coupled model and there are RTs there (eg cpld_control_p8).

Setting the mesh cap to run a standalone RT will be needed I assume to merge into develop, so perhaps that should also be added as an issue under the project now that I think of it.

MatthewMasarik-NOAA commented 2 years ago

@DeniseWorthen Oh, my mistake.. I misunderstood. I thought you meant it did not run under any coupled model RT's yet. That's great. I'll profile it with the coupled RT's. Thanks! 👍

mickaelaccensi commented 3 months ago

@MatthewMasarik-NOAA is this issue solved ?

MatthewMasarik-NOAA commented 3 months ago

@mickaelaccensi, this has not been completely resolved and should stay open. It's thought that if a leak does exist it's relatively small and not of significant concern. The priority has been lowered on this task accordingly, but it's still open, and I hope to return to it after some of our operational deadlines are met.