Hallo, one of our students looked at the performance of the cholesky decomposition using ublas. The result is quite surprising: $ for i in 100 200 400 800
Hallo, matrix_matrix_binary uses restict_traits to identify the type of the iterators. Thus an expression (dense - banded) gives packed_proxy_tag as storage
I wonder what is the most efficient way to compute the product J * A * Jt where J is a sparse matrix, A is diagonal and Jt is the transpost of J. In Ublas
In the constructor of matrix_range for instance, the precondition checks are commented out. Now there is no error generated when you create a matrix_range that
The following line does not work (it is from a cholesky routine submitted a long time ago on this list): UBLAS::project(U.row (i))(UBLAS::range (i + 1, size))