Re: [ublas] Bad prod efficiency of sparse/symmetry matrix ?

From: Epic John (epicjohn_at_[hidden])
Date: 2006-12-23 15:54:59

sparse_prod is from operation_sparse.h

In my experience it is much faster than axpy_prod. I haven't investigated
why, but if anyone can recommend what the most efficient product functions
(ideally it should be seamless) are for various matrix types (namely
sparse/dense) I would appreciate it.

Thanks,

John

On 12/9/06, Seabook <cipoke_at_[hidden]> wrote:
>
>
> Hello Jihn,
>
> I never used  "sparse_prod".
> I need to include which file if I want to use it ?
> Thx~
>
>
> Epic John wrote:
> >
> > What about sparse_prod? Should that ever be used instead of
> axpy_prod?When
> > I
> > tried it, it seemed to run faster for sparse=sparse*sparse .
> > John
> >
> >
> >
> >
> >> My classmate ever demo Altas library for me.
> >> It's really fast when your matrix is bigger.
> >>
> >> any way, thanks a lot again
> >>
> >> If I am free ,I will try to test all kind  efficiency  of prod .
> >>
> >> Kuo-Shih
> >>
> >> Gunter Winkler wrote:
> >> >
> >> > On Friday 08 December 2006 15:59, Seabook wrote:
> >> >> I will try that but I still don't know how to use axpy_prod.
> >> >
> >> > 1) modify the line in operation.hpp marked with +/i in my diff
> >> > 2) use axpy_prod(A,X,Y,true) for Y = prod(A,X)
> >> >    and axpy_prod(A,X,Y,false) for Y += prod(A,X)
> >> >
> >> >> Could you paste your source code with axpy_prod operation?
> >> >
> >> > I should have atached it already :(
> >> >
> >> > BTW. ublas' product of dense matrices is a lot slower than any
> >> optimized
> >> > blas3
> >> > routine. However, you can use the atlas-bindings to use blas instead
> of
> >> > prod.
> >> >
> >> > mfg
> >> > Gunter
> >> >
> >> > _______________________________________________
> >> > ublas mailing list
> >> > ublas_at_[hidden]
> >> > http://lists.boost.org/mailman/listinfo.cgi/ublas
> >> >
> >> >
> >>
> >> --
> >> View this message in context:
> >>
> http://www.nabble.com/Bad-prod-efficiency-of-sparse-symmetry-matrix---tf2766020.html#a7759673
> >> Sent from the Boost - uBLAS mailing list archive at Nabble.com.
> >>
> >> _______________________________________________
> >> ublas mailing list
> >> ublas_at_[hidden]
> >> http://lists.boost.org/mailman/listinfo.cgi/ublas
> >>
> >
> > _______________________________________________
> > ublas mailing list
> > ublas_at_[hidden]
> > http://lists.boost.org/mailman/listinfo.cgi/ublas
> >
> >
>
> --
> View this message in context:
> http://www.nabble.com/Bad-prod-efficiency-of-sparse-symmetry-matrix---tf2766020.html#a7776525
> Sent from the Boost - uBLAS mailing list archive at Nabble.com.
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
>