I have a block tri-diagonal matrix and I would like to try the solver associated to a BandedBlockBanded matrix (see here).
I can easily construct the sparse matrix
Nx = 10 Ny = 10 T = spdiagm(0=> rand(Nx*Ny), -1=> rand(Nx*Ny-1), 1=> rand(Nx*Ny-1), Nx => rand(Nx*Ny-Nx), -Nx => rand(Nx*Ny-Nx))
However, the conversion to BandedBlockBanded is quite cumbersome. Is there a simple way rather than looping over the blocks, etc?