[MITgcm-support] large setup issues
Matthew Mazloff
mmazloff at MIT.EDU
Wed Oct 29 17:33:49 EDT 2008
Hello,
I'm having trouble initializing large setups on ranger. The _BARRIER
doesnt seem to be acting as a BARRIER. So when one processer is
still reading a namelist another processor is on a different list and
may overwrite the scratch file and thus crash the model. (E.g. one
proc. wants to read scratch1 thinking it is the ini_parms name list
but another processor has overwritten the packages boot namelist there.)
A fix would be to have every processor write to its own file (e.g.
WRITE(scrname1,'(3a)') 'scratch',myProcessStr(1:4),'_1'
WRITE(scrname2,'(3a)') 'scratch',myProcessStr(1:4),'_2'
OPEN(UNIT=scrUnit1,FILE=scrname1,STATUS='UNKNOWN')
OPEN(UNIT=scrUnit2,FILE=scrname2,STATUS='UNKNOWN')
) as I have done before...but shouldnt a barrier work...whats going on?
-Matt
More information about the MITgcm-support
mailing list