Hi Derek, I can't tell you how feasible a LAPACK/uBLAS interface is today (I know one is in progress, but I've never used it), but I can provide you with a Jacobi implementation written using uBLAS. It's part of a larger research project, and the code is LGPLed. No guarantees on speed, of course, but we've spent a bit of time optimizing it. It's available as part of the BTK project (http://btk.sf.net). Download the distribution, and check the linear_algebra.h/.cpp files. There's a function titled diagonalize_symmetric(). Should be pretty straigtforward. (Note that we've used typedefs for the matrix and vector types, so you might have to do a quick text substitution if you want to copy the code somewhere else.) -tim On Fri, 17 Oct 2003, Derek wrote:
Hi,
I am a novice to all this. I am looking for a C++ library that would let me interface with BLAS/LAPACK.
The immediate problem I have is the diagonalization of a large dense symmetric (possibly hermitian) matrix. (Speed will be a consideration eventually.) I would like to use the eigenproblem solvers in LAPACK.
I understand that uBLAS does not have a full interface to LAPACK yet, but it looks like it is possible to use lapack at some level. (Does anyone have an example program excerpt?)
I would be grateful for any advice about whether uBLAS is suitable for this. I am also considering other libraries such as MTL and Blitz++.
I will be using Intel C++ compiler and/or GNU g++ on a Linux PC.
Regards Derek
_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users