view liboctave/fCmplxQRP.cc @ 8987:542015fada9e

Eliminate the workspace in sparse transpose. The output's cidx (column start offset array) can serve as the workspace, so the routines operate in the space of their output.
author Jason Riedy <jason@acm.org>
date Mon, 16 Mar 2009 17:03:07 -0400
parents eb63fbe60fab
children 7918eb15040c
line wrap: on
line source

/*

Copyright (C) 1994, 1995, 1996, 1997, 2002, 2003, 2004, 2005, 2007,
              2008, 2009 John W. Eaton

This file is part of Octave.

Octave is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 3 of the License, or (at your
option) any later version.

Octave is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
for more details.

You should have received a copy of the GNU General Public License
along with Octave; see the file COPYING.  If not, see
<http://www.gnu.org/licenses/>.

*/

#ifdef HAVE_CONFIG_H
#include <config.h>
#endif

#include <cassert>

#include "fCmplxQRP.h"
#include "f77-fcn.h"
#include "lo-error.h"
#include "oct-locbuf.h"

extern "C"
{
  F77_RET_T
  F77_FUNC (cgeqp3, CGEQP3) (const octave_idx_type&, const octave_idx_type&, FloatComplex*,
			     const octave_idx_type&, octave_idx_type*, FloatComplex*, FloatComplex*,
			     const octave_idx_type&, float*, octave_idx_type&);
}

// It would be best to share some of this code with FloatComplexQR class...

FloatComplexQRP::FloatComplexQRP (const FloatComplexMatrix& a, QR::type qr_type)
  : FloatComplexQR (), p ()
{
  init (a, qr_type);
}

void
FloatComplexQRP::init (const FloatComplexMatrix& a, QR::type qr_type)
{
  assert (qr_type != QR::raw);

  octave_idx_type m = a.rows ();
  octave_idx_type n = a.cols ();

  octave_idx_type min_mn = m < n ? m : n;
  OCTAVE_LOCAL_BUFFER (FloatComplex, tau, min_mn);

  octave_idx_type info = 0;

  FloatComplexMatrix afact = a;
  if (m > n && qr_type == QR::std)
    afact.resize (m, m);

  MArray<octave_idx_type> jpvt (n, 0);

  if (m > 0)
    {
      OCTAVE_LOCAL_BUFFER (float, rwork, 2*n);

      // workspace query.
      FloatComplex clwork;
      F77_XFCN (cgeqp3, CGEQP3, (m, n, afact.fortran_vec (), m, jpvt.fortran_vec (),
                                 tau, &clwork, -1, rwork, info));

      // allocate buffer and do the job.
      octave_idx_type lwork = clwork.real ();
      lwork = std::max (lwork, static_cast<octave_idx_type> (1));
      OCTAVE_LOCAL_BUFFER (FloatComplex, work, lwork);
      F77_XFCN (cgeqp3, CGEQP3, (m, n, afact.fortran_vec (), m, jpvt.fortran_vec (),
                                 tau, work, lwork, rwork, info));
    }
  else
    for (octave_idx_type i = 0; i < n; i++) jpvt(i) = i+1;

  // Form Permutation matrix (if economy is requested, return the
  // indices only!)

  jpvt -= static_cast<octave_idx_type> (1);
  p = PermMatrix (jpvt, true);


  form (n, afact, tau, qr_type);
}

FloatColumnVector
FloatComplexQRP::Pvec (void) const
{
  Array<float> pa (p.pvec ());
  FloatColumnVector pv (MArray<float> (pa) + 1.0f);
  return pv;
}

/*
;;; Local Variables: ***
;;; mode: C++ ***
;;; End: ***
*/