From: Karl Meerbergen (Karl.Meerbergen_at_[hidden])
Date: 2007-08-27 07:06:25
The BLAS are for dense matrices and vectors only. In theory you should
be able to apply the blas to the value_array of the sparse matrix types
if this would be useful.
Blas and lapack assume column major matrices, since they are defined as
fortran subprograms. A row major matrix is most likely interpreted as
its transpose, but without swapping size1 and size2! So, be careful.
Michele De Stefano wrote:
> Hi Kresmir.
> First of all, thank you very much.
> The problem was I had the ATLAS compiled library installed but not the
> corresponding headers
> (I had done the installation through "Yum Extender" on my Fedora 7,
> but I had not added the
> development SW ... now I have done it and I was able to compile the
> I have other questions for you.
> Are your bindings able to treat sparse matrices too ?
> I ask this because I've tried to use your "gemv" binding ( blas.hpp)
> and I was able to make the
> program work only with a dense matrix, column_major storage. Is this
> behavior correct or am
> I doing something wrong ?
> Saying it in other terms, is it possible to use row_major storage with
> your "gemv" too ?
> (I ask this only for information, because I already know that the same
> operation can be done using
> the "axpy_prod" function of the official Boost).
> On 8/24/07, * Kresimir Fresl* <fresl_at_[hidden] <mailto:fresl_at_[hidden]>>
> Hi Michele,
> > The first error I receive is that the compiler (gcc 4.1.2)
> cannot find
> > atlas/clapack.h or atlas/cblas.h.
> If I understand correctly, these are ATLAS headers. Their location
> depends on your installation (of ATLAS). You should edit
> bindings/atlas/cblas_inc.hpp and clapack_inc.hpp to provide your
> > [...]
> > Has someone an idea ?
> Unfortunately, I cannot help for other errors. Although it probably
> sounds strange, I currently don't have a working version of ublas and
> ublas mailing list
> ublas_at_[hidden] <mailto:ublas_at_[hidden]>
>ublas mailing list