Looking at the code in SummationByParts
case (0) direction if ( bb(1) == 0 ) then il = 1 + gsize else ol = offset(1) !$omp parallel workshare dvar(1+ol,:,:) = ( q(1,1) * var(1+ol,:,:) + q(2,1) * var(2+ol,:,:) ) * idel dvar(2+ol,:,:) = ( q(1,2) * var(1+ol,:,:) + q(3,2) * var(3+ol,:,:) ) * idel !$omp end parallel workshare il = 3 + ol end if
seems to indicate that for non-boundary ends of a grid component (
bb == 0) derivatives are computed only in the interior (from
il = 1+gsize) while for boundary ends some (larger) area is valid.
This does not seem to be documented though, at least I could not find it in either documentation.tex nor interface.ccl.