[Mitgcm-support] Float exchanges

[mitgcm-support at dev.mitgcm.org]

Alistair,

currently I'm in thge debugging phase of the new float code. I've built
in the changes regarding the distribution and exchange of the floats you
suggested.

The exchanges work in principle but from time to time some of them got
lost. Although these floats are written to the buffer correctly and send
they are seem not to be recognized on the target tile. In this context I
have a few short questions:

In EXCH_RL_SEND_PUT_X/EXCH_RL_RECV_GET_X there is a bunch of IDs for
signal completion (e.g. exchReqIdX, eastRecvAck). Now, when it comes to
receiving there is a flag in EXCH_RL_RECV_GET_X called exchUsesBarrier
that does not use all these IDs but simply calls a barrier. Even if this
flag is not set by default it seems that calling the barrier would be
more safe that the whole flag ID story. Is that true? And why is no
mpi_wait needed in that case?

And, maybe more difficult: Is there any possibility or trick to test the
exchange between tiles?

Thanks for your answers, but I got lost here...

Ciao,
	Arne

P.S. Did you had any time yet to include a new advection scheme?

-- 

Dr. Arne Biastoch
Scripps Institution of Oceanography        phone: +1-858-822-3787
Physical Oceanography Research Division    fax  : +1-858-534-9820
MS: 0230                                   email: abiastoch at ucsd.edu
8605 La Jolla Shores Dr.            
La Jolla, CA 92093-0230, U.S.A.   http://www.ecco.ucsd.edu/~biastoch