thanks On Sun, May 22, 2016 at 2:41 PM, Tony Kelman <t...@kelman.net> wrote:
> Shared memory or distributed? For the latter I would look at Elemental. > For the former, you can use a multithreaded BLAS in the LAPACK eigenvalue > calls, though from reports it appears MKL is often better optimized on > eigenvalue computation that openblas.