Custom Query (986 matches)
Results (535 - 537 of 986)
Ticket | Owner | Reporter | Resolution | Summary |
---|---|---|---|---|
#649 | Fixed | MPDATA tiling bug | ||
Description |
TS_MPDATA developed a tiling bug in the mpdata_adiff.F routine when all the ranges were recoded to be more general. There were only a few places, but it was critical. I found the offending lines and provided a routine (attached) with correct ranges. Also, i added a new flag TS_MPDATA_LIMIT. When this option is activated, the upwind corrector fluxes are further limited to be a factor of fac = 0.25 times the computed corrector flux. We have found that in some limited applications with rapidly varying topography and strong tracer gradients, that the upwind flux was creating some oscillations. This limits that behavior. Of course, if you set the fac = 1.0, then the method is purely upwind. |
|||
#650 | Fixed | Corrected bugs in inp_par.F | ||
Description |
Several reported bugs were corrected:
The other changes are for readability. |
|||
#651 | Fixed | Corrected minor bug in mod_stepping.F | ||
Description |
Corrected a minor bug in mod_stepping.F that was triggered when ADJUST_BOUNDARY was not activated in the 4D-Var algorithms. We need to have instead: #if defined ADJUST_BOUNDARY || defined ADJUST_STFLUX || \ defined ADJUST_WSTRESS allocate ( Lfinp(Ngrids) ) Lfinp(1:Ngrids)=1 allocate ( Lfout(Ngrids) ) Lfout(1:Ngrids)=1 #endif We were initializing Lbout instead Lfout above. A compilation error resulted if ADJUST_BOUNDARY is not activated. Many thanks to Julia Levin for bringing this to my attention. I also corrected several comments that include ' or " in a line. Some compilers issue warnings. This is actually a bug in the compiler. It should not check commented statement like: ! H Q(k) = Q(k) T(k) + Gamma(k) q(k+1) e'(k) ! The above comment was changed to: ! H Q(k) = Q(k) T(k) + Gamma(k) q(k+1) transpose[e(k)] ! |