[ADMB Users] example to interface with Lapack routine using Openblas

dave fournier davef at otter-rsch.com
Sun Aug 18 09:09:39 PDT 2013

Over the years I have had various requests for ways to use some Lapack 
these are optimized linear algebra routines which can use multi-threading.

A difficulty with doing this easily is that these are based on old 
Fortran code which
assumes that matrices are a contiguous array stored by column. In ADMB
matrices are not contiguous and are stored by row.

However, I realized that matrices and be constructed in such a way that they
are contiguous although they are still stored by row.  If in addition the
matrix is symmetric the distinction between rows and columns disappears,
so this is a simple case to start with.

The attached code shows how to use the Lapack routine dgesdd to
get the singular value decomposition of a symmetric matrix.
The particular implementation of Lapack is Openblas which contains


Maybe we can extend these ideas at the workshop if there is any interest.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: stest1.cpp
Type: text/x-c++src
Size: 4364 bytes
Desc: not available
URL: <http://lists.admb-project.org/pipermail/users/attachments/20130818/0c093481/attachment.cpp>

More information about the Users mailing list