|
|
Ublas : |
From: Jeffrey Brent McBeth (mcbeth_at_[hidden])
Date: 2008-01-11 12:21:30
On Fri, Jan 11, 2008 at 05:12:29PM -0000, John Maddock wrote:
> Luis M. de la Cruz Salas wrote:
> >> Are you using -NDEBUG in your C++ compilation?.
>
> In addition to that, Matlab and other programs like it are highly optimised
> for vector/matrix operations, they'll use a processor-specific library as a
> backend to speed up these operations using your processor's native
> vectorised-math operations. I believe there are some bindings somewhere (in
> the sandbox maybe?) that allow you to use Ublas with an optimised BLAS
> backend, but if your compiler's optimiser supports it, you could also try
> turning on automatic optimisation of vectorisable operations (Intel C++
> supports this for example).
I would actually be disappointed if a program called "Matrix Laboratory"
wasn't fast at matrix operations :) There are definitely bindings in the
sandbox. We use the LAPACK bindings with Intel's implementation in our work.
Jeff
-- ---------------------------------------------------------------------------- "The man who does not read good books has no advantage over the man who cannot read them." -- Mark Twain ----------------------------------------------------------------------------