Print the final solution

[soumyadipghosh]

A way to visualize the final solution in the domain would be helpful. My suggestion would be to write the final solution to a file aand then use matplotlib in a python script to do the plotting as a heatmap.

[pratikvn]

bench_ras.cpp has the overall final solution in explicit_laplacian_solution. It is a gko::Dense<> object.

It should be easy to print this out to a file.

For each subdomain, the local_solution object in the SolverBase<>::run should contain what you want. Again, you can write a function to basically take a gko::Dense<> object or a ValueType * (you can get this from gko::Dense by using ->get_values ) and print them out to ofstream.

[soumyadipghosh]

I tried doing fp << explicit_laplacian_solution->get_values() but I guess that just outputs the pointer variable. How can I get the size of a gko::Dense<> object to iterate over it ?

[pratikvn]

To get the size of an object of type gko::Dense<>, you just use ->get_size() member function. This returns an object if type gko::dim, which contains the size of the matrix. For example, to get the number of rows, you would just do explicit_laplacian_solution->get_size()[0] and number of columns would be explicit_laplacian_solution->get_size()[1]

[soumyadipghosh]

From the definition of explicit_laplacian_solution, it appears to be a 1-D vector. How do I reorder that back to a matrix representing the 2-D domain?

Edit: I can do it the sloppy way of printing the vector like a matrix but is there a library function to convert it to a matrix ?

[pratikvn]

You can technically, use strides to view it as a matrix. But in any case with matplotlib/matlab, it should be just as easy to plot the 1D vector in a 2D space as it would be for a 2D vector. So, I think printing it as a 1D vector is the best way.