1. extend transpose to CCS/CRS (in-place)
2. extend spblas_dgemm to CRS
3. range checking on get/set - move part of the routine into the .h header file, and put the rest in a _source file ?
4. sorted csc,csr
5. document: spmatrix_compress
6. _csc, _csr : should these accept matrix input argument? instead of allocating matrix ?
7. example program for parallel assembly / FLG_FIXED flag documentation
8. scipy routines:
   gsl_spmatrix_diagonal - extract diagonal
   gsl_spmatrix_memcpy - convert CSR to CSC etc, csr_tocsc
   csr_tobsr ?
   csr_sort_indices
   get_csr_submatrix
9. spmatrix_max_index + tests for min/max_index
10. document FLG_FIXED and FLG_GROW ?