[MITgcm-devel] globalFiles

Jean-Michel Campin jmc at ocean.mit.edu
Thu Nov 3 12:24:10 EST 2005


Hi,

I did some test with globalFiles=T, on cg01 & aces cluster,
2 processors MPI runs, with 6 tiles & EXCH2, and get strange
output files (sometimes even of wrong size).

I also tried useSingleCpuIO=T, and get better results, but
still have problems when using globalFiles=T. with files
written by write_local_rl/rs (i.e. called within bi,bj loops).

I added a "warning" when globalFiles is used with multiprocessors
(nPx*nPy>1) and would like to turn it to a STOP, since
this option is not safe, at least on the 2 platforms I tested.

An other solution would be to put the stop only if
globalFiles=T & useSingleCpuIO=F & nPx*nPy>1,
if we can find a way to fix the write_local_rl/rs.
I was thinking to force to use globalFile=F inside
write_local_rl/rs in the case where nSx*nSy>1.

Does this seems a good fix ? or do we stick with the (simpler)
first solution (i.e.: if globalFiles=T & nPx*nPy>1 : STOP) ?

Cheers,
Jean-Michel



More information about the MITgcm-devel mailing list