Tpetra_Experimental_BlockView.hpp

Go to the documentation of this file.

// @HEADER
// ***********************************************************************
//
//          Tpetra: Templated Linear Algebra Services Package
//                 Copyright (2008) Sandia Corporation
//
// Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation,
// the U.S. Government retains certain rights in this software.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// 1. Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
//
// 2. Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the distribution.
//
// 3. Neither the name of the Corporation nor the names of the
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY SANDIA CORPORATION "AS IS" AND ANY
// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SANDIA CORPORATION OR THE
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Questions? Contact Michael A. Heroux (maherou@sandia.gov)
//
// ************************************************************************
// @HEADER

#ifndef TPETRA_EXPERIMENTAL_BLOCKVIEW_HPP
#define TPETRA_EXPERIMENTAL_BLOCKVIEW_HPP


#include <Teuchos_ScalarTraits.hpp>
#include <Teuchos_LAPACK.hpp>


//#include "Teuchos_LAPACK_wrappers.hpp"
//extern "C" {int DGETRF_F77(const int *, const int *, double *, const int*, int *, int*);}

namespace Tpetra {

namespace Experimental {

template<class Scalar, class LO>
class LittleBlock {
private:
  typedef Teuchos::ScalarTraits<Scalar> STS;

public:
  LittleBlock (Scalar* const A,
               const LO blockSize,
               const LO strideX,
               const LO strideY) :
    A_ (A), blockSize_ (blockSize), strideX_ (strideX), strideY_ (strideY)
  {
  }

  LO getBlockSize () const {
    return blockSize_;
  }

  Scalar* getRawPtr () const {
    return A_;
  }

  Scalar& operator() (const LO i, const LO j) const {
    return A_[i * strideX_ + j * strideY_];
  }

  template<class LittleBlockType>
  void update (const Scalar& alpha, const LittleBlockType& X) const {
    for (LO j = 0; j < blockSize_; ++j) {
      for (LO i = 0; i < blockSize_; ++i) {
        (*this)(i,j) += alpha * X(i,j);
      }
    }
  }

  template<class LittleBlockType>
  void assign (const LittleBlockType& X) const {
    for (LO j = 0; j < blockSize_; ++j) {
      for (LO i = 0; i < blockSize_; ++i) {
        (*this)(i,j) = X(i,j);
      }
    }
  }

  void scale (const Scalar& alpha) const {
    for (LO j = 0; j < blockSize_; ++j) {
      for (LO i = 0; i < blockSize_; ++i) {
        (*this)(i,j) *= alpha;
      }
    }
  }

  void fill (const Scalar& alpha) const {
    for (LO j = 0; j < blockSize_; ++j) {
      for (LO i = 0; i < blockSize_; ++i) {
        (*this)(i,j) = alpha;
      }
    }
  }

  template<class LittleBlockType>
  void absmax (const LittleBlockType& X) const {
    for (LO j = 0; j < blockSize_; ++j) {
      for (LO i = 0; i < blockSize_; ++i) {
        Scalar& Y_ij = (*this)(i,j);
        const Scalar X_ij = X(i,j);
        Y_ij = std::max (STS::magnitude (Y_ij), STS::magnitude (X_ij));
      }
    }
  }



  void factorize (int* ipiv, int & info)
  {
    Teuchos::LAPACK<int, Scalar> lapack;
    lapack.GETRF(blockSize_, blockSize_, A_, blockSize_, ipiv, &info);
  }

  template<class LittleVectorType>
  void solve (LittleVectorType & X, const int* ipiv) const
  {
    int info;
    Teuchos::LAPACK<int, Scalar> lapack;
    char trans = 'T';
    lapack.GETRS(trans, blockSize_, 1, A_, blockSize_, ipiv, X.getRawPtr(), blockSize_, &info);
  }

private:
  Scalar* const A_;
  const LO blockSize_;
  const LO strideX_;
  const LO strideY_;
  std::vector<int> ipiv_;
};


template<class Scalar, class LO>
class LittleVector {
private:
  typedef Teuchos::ScalarTraits<Scalar> STS;

public:
  LittleVector (Scalar* const A, const LO blockSize, const LO stride) :
    A_ (A), blockSize_ (blockSize), strideX_ (stride)
  {}

  Scalar* getRawPtr () const {
    return A_;
  }

  LO getBlockSize () const {
    return blockSize_;
  }

  LO getStride () const {
    return strideX_;
  }

  Scalar& operator() (const LO i) const {
    return A_[i * strideX_];
  }

  template<class LittleVectorType>
  void update (const Scalar& alpha, const LittleVectorType& X) const {
    for (LO i = 0; i < blockSize_; ++i) {
      (*this)(i) += alpha * X(i);
    }
  }

  template<class LittleVectorType>
  void assign (const LittleVectorType& X) const {
    for (LO i = 0; i < blockSize_; ++i) {
      (*this)(i) = X(i);
    }
  }

  void scale (const Scalar& alpha) const {
    for (LO i = 0; i < blockSize_; ++i) {
      (*this)(i) *= alpha;
    }
  }

  void fill (const Scalar& alpha) const {
    for (LO i = 0; i < blockSize_; ++i) {
      (*this)(i) = alpha;
    }
  }

  template<class LittleVectorType>
  void absmax (const LittleVectorType& X) const {
    for (LO i = 0; i < blockSize_; ++i) {
      Scalar& Y_i = (*this)(i);
      Y_i = std::max (STS::magnitude (Y_i), STS::magnitude (X (i)));
    }
  }

  template<class LittleVectorType>
  bool equal (const LittleVectorType& X) const {
    if (getBlockSize () != X.getBlockSize ()) {
      return false;
    }
    for (LO i = 0; i < blockSize_; ++i) {
      if ((*this)(i) != X(i)) {
        return false;
      }
    }
    return true;
  }

  template<class LittleBlockType, class LittleVectorType>
  void
  matvecUpdate (const Scalar& alpha,
                const LittleBlockType& A,
                const LittleVectorType& X) const
  {
    // FIXME (mfh 07 May 2014) This is suitable for column major, not
    // for row major.  Of course, we'll have to change other loops
    // above as well to make row major faster.
    for (LO j = 0; j < blockSize_; ++j) {
      for (LO i = 0; i < blockSize_; ++i) {
        (*this)(i) += alpha * A(i,j) * X(j);
      }
    }
  }

private:
  Scalar* const A_;
  const LO blockSize_;
  const LO strideX_;
};

} // namespace Experimental
} // namespace Tpetra

#endif // TPETRA_EXPERIMENTAL_BLOCKVIEW_HPP