Minutes
Date: 12th March, 2019
Attendees:
- Marshall Ward (MW) (Chair) NCI
- Aidan Heerdegen (AH) CLEX, Andrew Kiss (AK) COSIMA, ANU
- Russ Fiedler (RF), Matt Chamberlain(MC) CSIRO Hobart
Updates
MW: Submitted parallel IO FMS patch. New automake made PR more complicated. FMS now buildable by automake. If we add new files/dependency build will fail. Not very auto. Works well. Some tuning with lustre. Just need a good io_layout with large contiguous chunks.
AH: Compression? MW: IO runtimes double with compression. Testing some of the newer algos and getting better numbers. AH: How does he get other compression into netCDF? MW: Custom libraries. Will make accessible when time comes. Have been using it. Good. MW: Haven’t heard back from GFDL yet. AH: will be needed for further high res models.
MW: RF big help with all the fill value stuff. RF: You put missing values in missing tiles. There is a mppnccombine bug which stuffs up. MW: Used netCDF fill value for restarts, MOM sets it to CMOR fill value.
automake goal is to make FMS into a library that can be centrally installed.
AH: Should just link against libFMS rather than have in the repo.
RF: part of decadal project is to update FMS. Wants to use it in AR4. MW: make it loadable module? AH: dynamic linking? AH: need access to compiled module files?
MW: Issues with GNU/Intel. Dale roberts wooing on this. Also module versions are an issue. Intel 18 can’t read 19 mod files for example. Don’t think there are any includes. If there are, should be fixed. RF: There is include mpi.f. Not sure how they include netcdf.
MW: Issues with GNU/Intel. Dale roberts wooing on this. Also module versions are an issue. Intel 18 can’t read 19 mod files for example. Don’t think there are any includes. If there are, should be fixed. RF: There is include mpi.f. Not sure how they include netcdf.
RF: Advanced with getting WOMBAT into CM2 harmonised MOM5. Fixed a few issues with redundant ifdefs. Hopefully sometime this week can finish. Pretty close by the end of the day.
MC: Might have an executable with WOMBAT compiled into it? RF: Yeah, just changed a few interfaces to get rid off ifdefs, changed to optional arguments for WOMBAT arguments.
MC: If Matear is there tomorrow, can try something.
AH: Now have a suite of models. Different resolutions, now with BGC as well. Great position to be in.