From: David K. <dkn...@se...> - 2013-01-22 04:19:58
|
On 01/21/2013 11:06 PM, Roy Stogner wrote: > > On Mon, 21 Jan 2013, David Knezevic wrote: > >> You have to make sure you call >> heterogenously_constrain_element_matrix_and_vector (instead of >> constrain_element_matrix_and_vector) as in introduction_ex4. > > Well, wait - this is actually a little more subtle than that. Some > systems may still need to be constrained homogeneously (e.g. if you're > solving for (du := u_{n+1} - u_{n}) rather than directly for u_{n+1} > in a time step or a newton step. It depends on how you do your > solver. Yes, certainly, it depends what he's doing in the problem... I thought not calling heterogenously_constrain_element_matrix_and_vector might be the issue in K's case though, based on what he described, but I shouldn't have implied that heterogenously_constrain_element_matrix_and_vector is always relevant! David |