Can't find a good block size

[lionelalexandre]

Dear All,

I've got an issue in running Conquest in parallel with diagonalisation ; I get "Can't find a good block size...".
I've looked quickly to this part of the code and docs and I think this is related to the fact that the H matrix size
is a prime number (unlucky !). Is there a way of dealing with that ?

Thank you for your precious help.
Lionel

[davidbowler]

The only solution at the moment is to set the block size to one using Diag.BlockSizeR and Diag.BlockSizeC. This may make the diagonalisation less efficient.

We do have a possible fix which involves padding the Hamiltonian to make it a good size and then discarding the resulting eigenstates, but that is not available yet (though soon!).