Joe Landman wrote:
Ralph Castain wrote:
Ummm....not to put gasoline on the fire, but...if the data exchange is
blocking, why do you need to call a barrier op first? Just use an
appropriate blocking data exchange call (collective or whatever) and
it will "barrier" anyway.
Since I don't run these codes, I would have to defer to those that do.
This said, I am not sure if they are running the coupling as two
separate MPI codes or as one code sharing a communications handle or
whatnot else.
I do agree that letting the data exchange provide a (natural) barrier
makes a great deal of sense, though the codes may not be amenable to
this mode of operations. Gerry could likely shed light on this.
What we do is to invoke one of several barrier methods. The two codes
we are working with have very different time-steps and different CFL
limits, as well as having simplified physics in one vice the other. Gus
is more eloquent than I in his explanation, but I didn't want to dilute
his comments with mine.
Simply, we have to calculate exchange points, and we can
programmatically use a barrier call as one of several possible solutions
to creating a state where we're waiting for exchange to occur. Use of
MpiBarrier came before I started working on this project and I've not
gotten to the point that I'm willing to rewrite it (it's working).
When coupling ocean and atmosphere models one has to consider the time
constants of the media. We see a time constant of minutes to small
numbers of hours to resolve a phenomenon in the atmosphere, while for
the ocean, it's 10's of hours to days. Thus, the timesteps don't
oincide often, and when they DO coincide, it's when we want to exchange
data.
With regard to the models in question, WRF is essentially a monolith for
th is purpose, for atmosphere, land surface, and tropo/stratosphere. We
run ROMS to incorporate the fluid ocean and sea ice. The exchange is
flux data and wind forcings. We're running as a climate simulation,
meaning there are several explicit conditions that simplify (mostly) our
calculations compared to _weather_ simulations.
Use of a natural barrier would be better but that means it's harder for
us to get 100 years of simulation time (or more) accomplished when we
have to add natural barriers and restarts.
gc
--
Gerry Creager -- gerry.crea...@tamu.edu
Texas Mesonet -- AATLT, Texas A&M University
Cell: 979.229.5301 Office: 979.458.4020 FAX: 979.862.3983
Office: 1700 Research Parkway Ste 160, TAMU, College Station, TX 77843
