Sam, I’d suggest you to use a shared pointer as member variable:
std::shared_ptr<FEFieldFunction<…> > f_ptr; Then the first time you know how to construct the function, you do this as f_ptr = std::shared_ptr<FEFieldFunction<…> > (new FEFieldFunction<…>(here your construction objects); you access this as a standard pointer: f_ptr->member_function(); It will be automatically destroyed when your class is destroyed. Notice that FEFieldFunction is already threadsafe. You should be able to use this at will in Workstream. L. > On 20 May 2016, at 11:32, Sam Cox <[email protected]> wrote: > > Hi all, > > Perhaps I can pick your brains. My setup is that I have two parallel > distributed meshes, tria2 is a refined-and-coarsened version of tria1 in some > way. Crucially, it may be a coarsening in at least some cells. > > I want to project a solution1 calculated on tria1 into the FE space on tria2. > Both FE spaces are DG. If I were just doing refinement then I could just > interpolate, but because I'm coarsening I need to do things properly and > create the load vector for the projection from solution1, then solve against > the mass matrix on tria2, to get the projection onto tria2. > > I don't see any way to do this other than using Functions::FEFieldFunction. > If there are other ways, please let me know! > > With FEFieldFunction, I have an implementation that works, but building the > load vector is incredibly slow so I'm looking to improve it by generating the > load vector in a WorkStream. I can't work out a way to pass the > FEFieldFunction into the local_assembly function. The below are my steps and > assumptions, so please feel free to correct any incorrect assumptions I've > made. > > 1. FEFieldFunction doesn't have a reinit function, so I can't make it a > member variable of my main class: I have to build it as and when I need it. > So I can't have access inside the local_assembly function without explicitly > passing it in in some way. > > 2. FEFieldFunction doesn't seem to have a copy constructor either, so I'm not > sure how to pass it into the scratch object. My implementation 1 is my > attempt at this. I've tried various combinations of passing in references, > pointers, the object itself, and none work. > > 3. Instead, implementation 2 tries passing the FEFieldFunction in as an > argument to local_assembly. > > Both implementations give the same error, which is also attached, but I don't > know anything near enough about threading to know what it really means, or > what the underlying problem is that they don't really point to. > > Implementation1 and Implementation2 contain a MWE that demonstrates what I > need. This doesn't include the parallel::distributed::triangulation part, nor > the DG, but is hopefully sufficient. > > > Any ideas on how to get this to work, or how I might go about this in a > better way, are much appreciated! > Running El Capitan, Xcode 7.3.1, deal.II 8.4.0 from .dmg. > > Sam > > -- > The deal.II project is located at http://www.dealii.org/ > For mailing list/forum options, see > https://groups.google.com/d/forum/dealii?hl=en > --- > You received this message because you are subscribed to the Google Groups > "deal.II User Group" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/d/optout. > <implementation1><implementation2><error_output.txt> -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
