CPPLapack Code

*Some warning messages are modified.

*dgsmatrix::pardiso was added. It is a parallel sparse direct solver.
*uint32_t was substituted by long.

*Some for loops are a bit speeded up.
*Some bug in zhsmatrix are fixed.

VERBOSE is substituted by CPPL_VERBOSE.

small fixes in sparse-io functions

*The default file format of write and read function was changed.
The write functions append a '#' at the very begining of the file.
For the backward compatibility, read functions can read files without the '#'.

*The style of "[ERROR]" messages was slightly changed.

minor modifications to suppress warnings about implicit casts

A lot of minor changes has been done.
Probably there are no need to modify your project codes using CPPLapack.

Main modifications are the followings.
1)All the private menmer objects (e.g. M, N, Array, etc.) were removed and put in public.
This modification made CPPLapack unsafe a bit and extendable a lot.
We highly recomend CPPLapack users to use valgrind with building your codes in debug mode.
2)Some implicit cast operators were removed.
Use to_XXXmatrix() functions instead.

specs of dgsmatrix and dssmatrix were changed a little.
They save uint32_t data for matrix index data instead of long integer.

The data storage for dgsmatrix and dssmatrix was modified.