Mercurial > octave-libgccjit
comparison liboctave/sparse-base-lu.h @ 8969:3ecbc236e2e0
Have sparse LU return permutation matrices rather than sparse matrices.
This could well impact user code. It'd be interesting to see if there
is any actual fall-out... Quite often, the permutation matrices are
applied to *dense* vectors. Returning permutation matrices rather
than sparse matrices is a slight performance enhancement, but likely
lost in the noise.
| author | Jason Riedy <jason@acm.org> |
|---|---|
| date | Tue, 10 Mar 2009 21:54:49 -0400 |
| parents | eb63fbe60fab |
| children | 4c0cdbe0acca |
comparison
equal
deleted
inserted
replaced
| 8968:91d53dc37f79 | 8969:3ecbc236e2e0 |
|---|---|
| 68 | 68 |
| 69 ColumnVector Pc_vec (void) const; | 69 ColumnVector Pc_vec (void) const; |
| 70 | 70 |
| 71 ColumnVector Pr_vec (void) const; | 71 ColumnVector Pr_vec (void) const; |
| 72 | 72 |
| 73 PermMatrix Pc_mat (void) const; | |
| 74 | |
| 75 PermMatrix Pr_mat (void) const; | |
| 76 | |
| 73 const octave_idx_type * row_perm (void) const { return P.fortran_vec (); } | 77 const octave_idx_type * row_perm (void) const { return P.fortran_vec (); } |
| 74 | 78 |
| 75 const octave_idx_type * col_perm (void) const { return Q.fortran_vec (); } | 79 const octave_idx_type * col_perm (void) const { return Q.fortran_vec (); } |
| 76 | 80 |
| 77 double rcond (void) const { return cond; } | 81 double rcond (void) const { return cond; } |