Error running fv 2x2.5 high freq history

[boyle5@llnl_gov]

I am running the CAM3.1p2, fv2x2.5 on Thunder at LLNL.
NHTFRQ(1) determines the frequency of the default history file.
If I set NHTFRQ(1) to anything but 0 (the default), I get an error involving memory allocation as the model tries to write out the history file.
If I use 0 all goes OK.
I have run the model in the T42 configuration, on the same platform and this error does not occur.

Has anyone else observed this behaviour?

 

[eaton]

Our regression testing uses several different output frequencies for the primary history file. In fact we don't write any monthly avg output during regression testing.

 

[boyle5@llnl_gov]

Brian -
Thanks for the reply.
I have downloaded the latest distribution from NCAR CAM3.1p2.
I have built and run the model on the NERSC machine Bassi which is an IBM p575 Power5 ( like BlueVista, BlueIce, I think) to try and emulate the NCAR environment(AIX, loadleveler, poe).
I get the same problem as on Thunder, if the default history (1) is not 0 the model hangs.
As far as I can tell I am running the same build, run as in the distribution.

Could you point me to the whereabouts of the build/run scripts for the regression tests? Maybe examining these will reveal my error. I do have access to NCAR.

--Jim