|
|
Ublas : |
Subject: Re: [ublas] [BLAS bindings] help appreciated on using trmv and gbmv
From: Rutger ter Borg (rutger_at_[hidden])
Date: 2012-10-10 09:26:30
On 2012-10-10 14:51, Florent Teichteil wrote:
>
> Ok, so I was not totally wrong after all...
>
Nope, thanks for finding these cases. An hour ago, I already added
support for banded matrices in bindings::trans(), the other items are
almost done, but will take a bit more testing.
But do remember that the bindings work with industry-standard libraries
(e.g., BLAS, and LAPACK), which are also provided by manufacturers, and
we are already providing an interface that checks way more than is done
in C or Fortran.
>
> Actually, I tried to be constructive and help you improve the bindings,
> since I find them very interesting. I also spent valuable time on
> searching information on the web about using the trmv and gbmv functions
> of the bindings, but could not find anything (whereas they're available
> for almost 3 years?).
Try looking for the same names, but prefix your stuff with 'd'. Try
Googling for 'dtrmv' and 'dgbmv', for instance. You should get hits on
netlib.org,
www.netlib.org/blas/dtrmv.f
http://www.netlib.org/blas/dgbmv.f
Or, look here (albeit incomplete)
http://svn.boost.org/svn/boost/sandbox/numeric_bindings/libs/numeric/bindings/doc/html/boost_numeric_bindings/reference/blas/level_2_blas/trmv.html
http://svn.boost.org/svn/boost/sandbox/numeric_bindings/libs/numeric/bindings/doc/html/boost_numeric_bindings/reference/blas/level_2_blas/gbmv.html
> I guess I am stupid or perhaps just one the first
> people who try to use these functions. I'm sorry guys, but because of
> lack of time, I have to move to a more intuitive and less error-prone
> library, for less intelligent people like me.
Actually, you are probably right to say that you are the first who tried
these functions in combination with those adaptors/views. The library
should have asserted your arguments.
If you are not interested in BLAS or LAPACK, the bindings may indeed not
be the best option for you. However, if raw speed or specialized
routines is what you're after, it is certainly worth the trouble and
maybe steep learning curve.
Cheers,
Rutger