Mercurial > hg > octave-terminal
changeset 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 | 91d53dc37f79 |
children | b37a6c27c23f |
files | liboctave/ChangeLog liboctave/sparse-base-lu.cc liboctave/sparse-base-lu.h src/ChangeLog src/DLD-FUNCTIONS/lu.cc |
diffstat | 5 files changed, 41 insertions(+), 8 deletions(-) [+] |
line wrap: on
line diff
--- a/liboctave/ChangeLog +++ b/liboctave/ChangeLog @@ -1,3 +1,13 @@ +2009-03-10 Jason Riedy <jason@acm.org> + + * sparse-base-lu.cc (Pr_mat): New member function. Return the row + permutation as a PermMatrix. + (Pc_mat): New member function. Return the col permutation as a + PermMatrix. + + * sparse-base-lu.h (sparse_base_lu): Declare Pc_mat and Pr_mat + member functions. + 2009-03-09 Jason Riedy <jason@acm.org> * Sparse-diag-op-defs.h (octave_impl::inner_do_add_sm_dm): New
--- a/liboctave/sparse-base-lu.cc +++ b/liboctave/sparse-base-lu.cc @@ -98,6 +98,13 @@ } template <class lu_type, class lu_elt_type, class p_type, class p_elt_type> +PermMatrix +sparse_base_lu <lu_type, lu_elt_type, p_type, p_elt_type> :: Pr_mat (void) const +{ + return PermMatrix (P, false); +} + +template <class lu_type, class lu_elt_type, class p_type, class p_elt_type> p_type sparse_base_lu <lu_type, lu_elt_type, p_type, p_elt_type> :: Pc (void) const { @@ -131,6 +138,13 @@ return Pout; } +template <class lu_type, class lu_elt_type, class p_type, class p_elt_type> +PermMatrix +sparse_base_lu <lu_type, lu_elt_type, p_type, p_elt_type> :: Pc_mat (void) const +{ + return PermMatrix (Q, true); +} + /* ;;; Local Variables: *** ;;; mode: C++ ***
--- a/liboctave/sparse-base-lu.h +++ b/liboctave/sparse-base-lu.h @@ -70,6 +70,10 @@ ColumnVector Pr_vec (void) const; + PermMatrix Pc_mat (void) const; + + PermMatrix Pr_mat (void) const; + const octave_idx_type * row_perm (void) const { return P.fortran_vec (); } const octave_idx_type * col_perm (void) const { return Q.fortran_vec (); }
--- a/src/ChangeLog +++ b/src/ChangeLog @@ -1,3 +1,8 @@ +2009-03-10 Jason Riedy <jason@acm.org> + + * DLD-FUNCTIONS/lu.cc (lu): Call fact.Pr_mat () and fact.Pc_mat () + to return permutation matrices in the sparse case. + 2009-03-12 John W. Eaton <jwe@octave.org> * ls-mat-ascii.cc (get_mat_data_input_line): If we are looking at
--- a/src/DLD-FUNCTIONS/lu.cc +++ b/src/DLD-FUNCTIONS/lu.cc @@ -226,7 +226,7 @@ retval (0) = fact.Y (); else { - SparseMatrix P = fact.Pr (); + PermMatrix P = fact.Pr_mat (); SparseMatrix L = P.transpose () * fact.L (); retval(1) = octave_value (fact.U (), MatrixType (MatrixType::Upper)); @@ -245,7 +245,7 @@ if (vecout) retval (2) = fact.Pr_vec (); else - retval(2) = fact.Pr (); + retval(2) = fact.Pr_mat (); retval(1) = octave_value (fact.U (), MatrixType (MatrixType::Upper)); @@ -269,8 +269,8 @@ } else { - retval(3) = fact.Pc (); - retval(2) = fact.Pr (); + retval(3) = fact.Pc_mat (); + retval(2) = fact.Pr_mat (); } retval(1) = octave_value (fact.U (), MatrixType (MatrixType::Upper)); @@ -296,7 +296,7 @@ retval (0) = fact.Y (); else { - SparseMatrix P = fact.Pr (); + PermMatrix P = fact.Pr_mat (); SparseComplexMatrix L = P.transpose () * fact.L (); retval(1) = octave_value (fact.U (), MatrixType (MatrixType::Upper)); @@ -315,7 +315,7 @@ if (vecout) retval (2) = fact.Pr_vec (); else - retval(2) = fact.Pr (); + retval(2) = fact.Pr_mat (); retval(1) = octave_value (fact.U (), MatrixType (MatrixType::Upper)); @@ -339,8 +339,8 @@ } else { - retval(3) = fact.Pc (); - retval(2) = fact.Pr (); + retval(3) = fact.Pc_mat (); + retval(2) = fact.Pr_mat (); } retval(1) = octave_value (fact.U (), MatrixType (MatrixType::Upper));