[MITgcm-devel] restart testing

Jean-Michel Campin jmc at ocean.mit.edu
Fri Jan 11 12:06:27 EST 2008


Hi,

I set-up an automatic test (currently daily, on faulks) 
for the restart (test 2+2=4), which is done after the normal testreport.
It gets to the testing web page (type="restart" instead of "forward"
or "adjoint"), e.g., the one from from last night:
http://mitgcm.org/testing/results/2008_01/rs_faulks_20080111_0/summary.txt
and added recently the "log" files of experiments that fail,
in the corresponding directory (e.g., in
http://mitgcm.org/testing/results/2008_01/rs_faulks_20080111_0/ )

one thing is that it assumes that we can run each tested experiment 
for 4 more time steps.

Dimitris:
Unfortunately, it seems that the new experiment "seaice_obcs"
cannot run for more than 9 time-steps = the length of the 
current test (I think there are not enought reccords in some
obcs input files, and when I try just 10 iters, starting from
iter=0, get the same problem as in this restart test).
It would be a good thing to have more reccords in the input files
to enable to run the test for 4 more iterations, so that this
set-up could also be tested for the restart.
I tried to reduce to 5 the length of the test-run and then 
did a 2+2=4 test, but did not get the same answer: 2+2 <> 4.

Cheers,
Jean-Michel




More information about the MITgcm-devel mailing list