From: Guillaume ANCIAUX <anciaux@la...>  20051107 17:08:14

why do you make the difference with the solution vector ? there is a=20 privileged communication process for it ?=20 nevertheless for that kind of operations (even for solution vector) do you= =20 need to send the complete vector to your neighbors ? this would be pretty=20 much non scalable (my opinion). Interpolation is not a key operation in the FE programs ? (i'm kind of newb= ie=20 in that). I need to compute a deformation gradient on quadrature points=20 (\sum_I \dphi_I(X) u_I) to compute a non linear behavior at each explicit=20 step. i know i can compute dphi and phi shapes functions locally as long as they= =20 are precomputed from reference element projection... but i really need to g= et=20 the nodes values (u_I) ... if it implies sending the complete vector ... :( maybe i didn't get something... thanks for your advices. Le Lundi 7 Novembre 2005 15:56, vous avez =E9crit=A0: > Hi, > > If I understand correctly, you want to communicate a field variable > (other than the current solution) involving only the minimum amount > of data required (i.e. values at the ghost nodes). I think you will > obtain the desired effect for any vector which is added to the system > using the "System::add_vector()" routine. However, I believe this > involves communicating all the values in the vector, not just those > for the ghost nodes... > > J > > Guillaume ANCIAUX writes: > > I wonder if there is a way to get a field (described by a shared > > NumericVector) interpolated in a finest way that localizing the shared > > vector. I mean by doing only the necessary communication in order to g= et > > the ghosts values .... > > > > This method should be on a system fonctionality but i can't find it ..= =2E. > > > > Anyone could give me help ? > > > > Thx... 