Here's the pertinent section of my genmake file.
"marcellus.navo.hpc.mil" is an IBM SP4 machine.
case marcellus.navo.hpc.mil:
# modified from A. Biastoch, SIO.
echo "Configuring for MARCELLUS IBM SP POWER4"
set SOURCEDIRS = ( ./ $SOURCEDIRS )
set LN = ( 'ln -s' )
set DEFINES = ( ${DEFINES} '-DTARGET_PWR3 -DTARGET_SGI
-DWORDLENGTH=4' )
set INCLUDES = ( '-I/usr/lpp/ppe.poe/include' )
set FC = ( 'mpxlf95' )c
set LINK = ( 'mpxlf95' )
set FLAGS = ( '-O3 -qarch=pwr4 -qtune=pwr4 -qcache=auto
-qmaxmem=-1' \
'-bmaxdata:0x80000000 -bloadmap:mitgcmuv.map'
)
set FFLAGS = ( '-qfixed=132 -O3 -qarch=pwr4 -qtune=pwr4
-qcache=auto -qmaxmem=-1' \
'-bmaxdata:0x80000000 ' )
set LDFLAGS = ( '-O3 -qarch=pwr4 -qtune=pwr4 -qcache=auto
-qmaxmem=-1' \
'-bmaxdata:0x80000000' )
set LIBS = ( ' -L/usr/local/apps/mass -lmass' )
set NOOPTFLAGS = ( '-O0' )
set NOOPTFILES = ( 'cost_obcs_ageos.F' )
set RMFILES = ( 'rii_files' )
breaksw
Second, are you sure that you are applying o.b. values at all the o.b.
points? Is the dense water at temperature 0? This could be a missing
o.b. value. To avoid the messiness with o.b. masks, I have no
topographic gradient at the boundary, a.k.a. the bottom slope is flat at
the boundary.
Third, I have seen waves emanating from the intersection of the o.b.'s
and the tile boundaries. Do you see anything like this at the original
spinup time? Jean-Michel has an idea for adding extra exchanges to the
o.b. routines to eliminate this.
Jake