You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
* (*)Fixed dimensional inconsistency in P3M_functions
Corrected dimensionally inconsistent expressions in P3M_functions.F90,
notably in P3M_limiter and monotonize_cubic and a complete rewrite and
simplification of is_cubic_monotonic. Also added comments documenting the
units of all real variables in this module, and changed the code to use logical
variables in place of integer "booleans", including in the return value from
is_cubic_monotonic. These changes will change (fix) the answers when remapping
variables with small numerical values, but no answers change in the
MOM6-examples test cases.
* +Added REMAPPING_2018 runtime option
Added a new runtime option, REMAPPING_2018, which if set to false triggers the
use of new, more accurate expressions in various parts of the ALE remapping
code. By default, the older expressions are used, and all answers are bitwise
identical, but there are new optional arguments to various routines related to
remapping to trigger the use of new mathematically equivalent expressions. By
default all answers are bitwise identical, but there are new and reordered
entries in the MOM6_parameter_doc files.
* Corrected the formatting of a doxygen comment
* Added conversion factors to forcing diagnostics
Added conversion factors to 4 mass-flux diagnostics and comments to 4 others
on why no conversion factors are needed. All answers are bitwise identical.
* Added correct scaling factors to chksum calls
Added scale arguments to 5 chksum calls and grouped another two chksum calls
while also adding the right scaling argument. All answers are bitwise identical.
* +Unscales area before taking global sum
Undoes the dimensional scaling of the cell areas before taking their global
sum, so that the reproducing sum does not overflow when there is dimensional
rescaling. All answers are bitwise identical when there is no rescaling, but
this eliminates a source of inadvertent overflows or underflows in the global
sums, and there is a new optional argument to compute_global_grid_integrals.
* (*)Correct dimensionally inconsistent advective CFL
Corrects the dimensionally inconsistent expressions for the CFL number in
the tracer advection code, in which a negligible thickness had been added to
the cell volume to avoid division by zero. This change does not alter the
solutions in the MOM6-examples test cases, but now it permits dimensional
rescaling of lengths over a much larger range, and it could change answers if
the minimum layer thicknesses are small enough.
* Unscale sea level before averaging
Unscale interface heights before taking a global average via a reproducing sum
in non-Boussinesq mode global diagnostics to permit dimensional consistency
testing over a larger range. All answers are bitwise identical.
* +Added an optional tmp_scale arg to global_i_mean
Added an optional tmp_scale argument to global_i_mean and global_j_mean to
specify an internal rescaling of variables being averaged before the reproducing
sum. All answers are bitwise identical, but there are new optional arguments
to two public interfaces.
* Expand consistency testing with i-mean sponges
Use tmp_scale when taking the i-mean interface heights for i-mean sponges, to
give a greatly expanded range of dimensional consistency testing. All answers
are bitwise identical.
0 commit comments