From: Roy Stogner <roystgnr@ic...>  20070417 03:55:27

On Tue, 17 Apr 2007, Shengli Xu wrote: > Hello, Libmesh Users, > I have a question about the structure of the result vector of libmesh > system. > The following codes are from the GMVIO class : > > 00379 for (unsigned int n=0; n<mesh.n_nodes(); n++) > 00380 out << std::setprecision(10) << (*v)[n*n_vars + c] << " "; > My question is : If the field variables of system don't use the equal > order shape function, such as Q84 for stokes flow. The answer is simple: v is NOT the result vector of your solution. Before an IO class's write_nodal_data function is called, a serial vector v of nodal function values is calculated from the parallel solution vector. This means that if you have mixed biquadratic/bilinear data on a quad element, for example, all variables will be plotted as four bilinear quads. Unfortunately, this also means that if you have biquintic data on a quad element, it will still be plotted as four bilinear quads  currently the only format we have that supports the "raw" field variables is .xda/.xdr files.  Roy 