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
With #13990, we introduced a second variable for storing the plain Jacobian gradients, without the inverse Jacobians applied as a transformation as in the field
It would be better to use the untransformed Jacobian gradient also for the Hessian, in order to reduce the memory consumed/accessed by these large arrays. We should aim to fix this for the next release.
The text was updated successfully, but these errors were encountered:
This is an important step for optimizing H(div) elements with the matrix-free framework, that should be tackled soon. But it is not urgent for the upcoming release, so I'm adjusting the target milestone.
With #13990, we introduced a second variable for storing the plain Jacobian gradients, without the inverse Jacobians applied as a transformation as in the field
dealii/include/deal.II/matrix_free/mapping_info_storage.h
Lines 256 to 261 in 23cfb2b
It would be better to use the untransformed Jacobian gradient also for the Hessian, in order to reduce the memory consumed/accessed by these large arrays. We should aim to fix this for the next release.
The text was updated successfully, but these errors were encountered: