To anyone whom is interested, I have
"tuned" MOM4 for the NEC SX-6 architecture.
I will gladly share this "tuned" code, to anyone
that requests it.
The code as released from GFDL runs (asis), 1550 seconds - 1day (46500
seconds/month). After hand optimized, 103 seconds - 1day (3090 seconds/month)
verification (relative bit error 1.0D-11)
relative bit error : the relative difference between values stored in the
resulting neutral_physics matrix (comparing ASIS .vs. Tuned) after
neutral diffusion is computed for each tracer, based up local
flow properties. Currently there are 3 tracers temperature,
salinity and a third passive tracer. For more details please see
Danabasoglu and McWilliams (1995) Sensitivity of the global
ocean circulation to parameterizations of mesoscale tracer transports.
http://web.mit.edu/globalchange/www/rpt11.html
Journal of Climate, 8, 2967-2987. and Griffies (2002),
Fundamentals z-coordinate ocean climate models.
http://www.gfdl.gov/~lat/webpages/om/om_webpage.html
Optimization:
+50 man hours. +1033 lines of code. +35 revisions.
--
John Snyder
NEC Solutions (America), Inc.
4200 Research Forest Dr., Suite 400
The Woodlands, Tx 77381
phone: 281-465-1622
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
FMS-mom4 mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/fms-mom4
