[Mitgcm-support] floating point exception in the overlap

[mitgcm-support at dev.mitgcm.org]

Just found a minor bug in ini_spherical_polar_grid.F:
if you specify cosPower = .5 (or some other non-integer value, thereby
implying taking a root) -- the default is 0. -- then 
          cosFacV(J,bi,bj)=COS((yC(1,J,bi,bj)-0.5*delY(jG))*deg2rad)
     &                    **cosPower
gives a NaN for J=-2 in the global_ocean setup, because J=-2 corresponds
to yC-.5*delY=-80-3*4=-92degrees. The Sun WorkShop 6 update 2 FORTRAN 77
5.3 compiler complains at runtime and raises a floating point exception
flag. The result does not seem to be affected. I guess the halo of the
grid is not used or remasked lateron (?)

Martin

-- 
Martin Losch // mailto:mlosch at mit.edu
Department of Earth, Atmospheric, and Planetary Sciences
MIT, Bldg. 54-1523, 77 Massachusetts Ave., Cambridge, MA 02139
phone/fax: ++1 (617) 253-5938/4464