To support the class functionality of SparseLib++, Sparse BLAS routines have
been developed in C, and are callable through the interface proposed in [1].
The current implementation supports matrix-matrix multiply and triangular
system solves for the matrix storage formats currently in SparseLib++. The same
functionality for the symmetric versions of these formats is also available.
Permutation and integrity checking functionality will soon be added.
*** NOTE: ***
The current version of SparseLib++ (v1.5) includes only a subset of these BLAS.
The full version will be included in the next release of SparseLib++.
Official Caveat:
Neither the Authors nor the Institutions (National Institute of Standards
and Technology, University of Notre Dame, and University of Tennessee)
make any representations about the suitability of this software for any
purpose. This software is provided "as is" without expressed or implied
warranty.
Changes from v. 1.0:
Version 1.1 represents an optimization of v. 1.0 for varying vector offsets,
with the additional functionality of matvecs for symmetric matrices.
Changes from v. 1.1:
Version 1.1b includes an important bug fix to the matrix-matrix multiply
routines for coordinate sparse matrices. The distribution now includes the
type-specific sources, rather than the generic sources, since some versions
of "make" were not able to handle the source file generation.