Merge branch 'rmills/add-aijmkl'

* rmills/add-aijmkl:
Adds AIJMKL matrix class to facilitate use of MKL sparse BLAS routines
BitBucket Pull Request 737
This is the initial version of the AIJMKL matr

Merge branch 'rmills/add-aijmkl'

* rmills/add-aijmkl:
Adds AIJMKL matrix class to facilitate use of MKL sparse BLAS routines
BitBucket Pull Request 737
This is the initial version of the AIJMKL matrix class,
a subclass of AIJ that has the goal of using MKL sparse BLAS routines whenever practicable.
If the MKL version is new enough, the sparse inspector-executor
("SpMV2" -- terrible name, but that's what the MKL team called it) routines are used by default.
Currently only MatMult, MatMultAdd, MatMultTranspose, and MatMultTransposeAdd operations are supported,
but more will be added once this initial attempt is at a point
that it provides an acceptable framework for adding additional further functionality.

show more ...

MKL: Fixed compilation problems for older MKL versions without mkl_sparse_optimize().

Fixes to handle null matrices (no explicit entries) correctly in _SpMV2 routines.

Removed obsolete comment.

Edits to the MatCreateSeqAIJMKL manual page.

Added MatDiagonalScale, MatDiagonalSet, and MatAXPY methods for SeqAIJMKL; fixed ops->scale not being reset in conversion from SeqAIJMKL to SeqAIJ.

The _SpMV2 versions of the SeqAIJMKL operations now check to see if mkl_sparse_optimize() has been called, and call it if needed.

Ensured that MKL sparse inspector-executor routines are not called for matrices with no nonzero entries.

Minor edit: Fixed spacing for a function call.

More fixes for transposed versions of MatMult/MatMultAdd for SeqAIJMKL; enabled those versions.

Added MatScale_SeqAIJMKL(); fixed many errors found when testing with
rectangular matrices.

Created a function that creates MKL SpMV2 matrix handles and performs optimization to eliminate code duplication.

Removed obsolete __FUNCT__ definitions.

Prevent MKL sparse inspector-executor routines from being called on
matrices with zero rows -- the routines don't handle this case.

Fixed nrows being passed to mkl_sparse_x_create_csr() for both nrows and
ncols.

Added error messages for failures in creation/copy of matrix handles or
mkl_sparse_optimize.

Fixes following the outline of Barry's fixes to AIJPERM.

Removed DEBUG statements.

Fixes to AIJMKL MatDuplicate and MatAssemblyEnd.

Added another #ifdef to allow use of MKL that does not have
mkl_sparse_optimize().

Code cleanup to eliminate some compiler warnings.

Make sure that mkl_sparse_destroy() is only called if
mkl_sparse_optimize() has been called; more work to allow AIJMKL code to
work with older versions of MKL.

Added code to handle case in AIJMKL in which the MKL library version is
older and does not have the sparse inspector-executor routines.

Use of SpMV2 inspector-executor model can now be turned off via command
line options.

Added support for the MKL SpMV2 inspector-executor routines for
matrix-vector multiplication.