Re: [ublas] Loop optimization problem

From: Norman Casagrande (nova77_at_[hidden])
Date: 2007-03-16 14:04:00

Use ALTAS or any other optimized library (intel or amd) for the solver
and the matrix operations. You'll get an impressive speedup.

Cheers,
Norman

On 3/15/07, Sourabh <sourabh_at_[hidden]> wrote:
> Hi all,
>
> I want to optimize the following set of lines of code. These lines form
> the part of internal loop which will be executed 2000000000 times.
>
> inplace_solve_lower<true>(slopelembda, slopeomegas);
> axpy_prod(delaylembda, slopeomegas, delayomegas, false);
> double pathdelay = 0;
> for (unsigned gate = 0; gate < delayomegas.size (); ++gate) {
>         pathdelay += delayomegas(gate);
> }
>
>
> In the above question,
> slopelembda: A strictly lower triangular matrix
> delaylembda: A strictly lower triangular matrix
> slopeomegas: A dense vector of size 100
> delayomegas: A dense vector of size 100
>
>
> Can I optimize it further ? If I am able to reduce its time by some
> margin, I will gain a lot of time because it is executed 2e9 times.
>
> Please suggest.
> --
> -- Sourabh
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
>

-- 
Bye,
Norman