Thyra_ProductVector.hpp

// @HEADER
// ***********************************************************************
// 
//    Thyra: Interfaces and Support for Abstract Numerical Algorithms
//                 Copyright (2004) Sandia Corporation
// 
// Under terms of Contract DE-AC04-94AL85000, there is a non-exclusive
// license for use of this work by or on behalf of the U.S. Government.
// 
// This library is free software; you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as
// published by the Free Software Foundation; either version 2.1 of the
// License, or (at your option) any later version.
//  
// This library 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
// Lesser General Public License for more details.
//  
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
// USA
// Questions? Contact Michael A. Heroux (maherou@sandia.gov) 
// 
// ***********************************************************************
// @HEADER

#ifndef THYRA_PRODUCT_VECTOR_HPP
#define THYRA_PRODUCT_VECTOR_HPP

#include "Thyra_ProductVectorDecl.hpp"
#include "Thyra_ProductVectorSpace.hpp"
#include "Teuchos_Workspace.hpp"

namespace Thyra {

// Constructors/initializers/accessors

template <class Scalar>
ProductVector<Scalar>::ProductVector()
{
  uninitialize();
}

template <class Scalar>
ProductVector<Scalar>::ProductVector(
  const Teuchos::RefCountPtr<const ProductVectorSpace<Scalar> >  &productSpace
  ,const Teuchos::RefCountPtr<VectorBase<Scalar> >               vecs[]
  )
{
  initialize(productSpace,vecs);
}

template <class Scalar>
void ProductVector<Scalar>::initialize(
  const Teuchos::RefCountPtr<const ProductVectorSpace<Scalar> >  &productSpace
  ,const Teuchos::RefCountPtr<VectorBase<Scalar> >               vecs[]
  )
{
  // ToDo: Validate input!
  numBlocks_ = productSpace->numBlocks();
  productSpace_ = productSpace;
  vecs_.resize(numBlocks_);
  if(vecs) {
    std::copy( vecs, vecs + numBlocks_, &vecs_[0] );
  }
  else {
    for( int k = 0; k < numBlocks_; ++k )
      vecs_[k] = createMember(productSpace->getBlock(k));
  }
}

template <class Scalar>
void ProductVector<Scalar>::uninitialize(
  Teuchos::RefCountPtr<const ProductVectorSpace<Scalar> >  *productSpace
  ,Teuchos::RefCountPtr<VectorBase<Scalar> >               vecs[]
  )
{
  if(productSpace) *productSpace = productSpace_;
  if(vecs) std::copy( &vecs_[0], &vecs_[0]+numBlocks_, vecs );
  productSpace_ = Teuchos::null;
  vecs_.resize(0);
  numBlocks_ = 0;
}

// Overridden from ProductVectorBase

template <class Scalar>
Teuchos::RefCountPtr<const ProductVectorSpaceBase<Scalar> >
ProductVector<Scalar>::productSpace() const
{
  return productSpace_;
}

template <class Scalar>
Teuchos::RefCountPtr<VectorBase<Scalar> >
ProductVector<Scalar>::getBlock(const int k)
{
  TEST_FOR_EXCEPT( k < 0 || numBlocks_-1 < k);
  return vecs_[k];
}

template <class Scalar>
Teuchos::RefCountPtr<const VectorBase<Scalar> >
ProductVector<Scalar>::getBlock(const int k) const
{
  TEST_FOR_EXCEPT( k < 0 || numBlocks_-1 < k);
  return vecs_[k];
}

// Overridden from VectorBase

template <class Scalar>
Teuchos::RefCountPtr< const VectorSpaceBase<Scalar> >
ProductVector<Scalar>::space() const
{
  return productSpace_;
}

template <class Scalar>
void ProductVector<Scalar>::applyOp(
  const RTOpPack::RTOpT<Scalar>    &op
  ,const int                       num_vecs
  ,const VectorBase<Scalar>*       vecs[]
  ,const int                       num_targ_vecs
  ,VectorBase<Scalar>*             targ_vecs[]
  ,RTOpPack::ReductTarget          *reduct_obj
  ,const Index                     first_ele_in
  ,const Index                     sub_dim_in
  ,const Index                     global_offset_in
  ) const
{
  using Teuchos::Workspace;
  Teuchos::WorkspaceStore* wss = Teuchos::get_default_workspace_store().get();
  //
  const Index n = productSpace_->dim();
  // Validate the compatibility of the vectors!
#ifdef _DEBUG
  TEST_FOR_EXCEPTION(
    !(1 <= first_ele_in && first_ele_in <= n), std::out_of_range
    ,"ProductVector::applyOp(...): Error, "
    "first_ele_in = " << first_ele_in << " is not in range [1,"<<n<<"]" );
  TEST_FOR_EXCEPTION(
    sub_dim_in < 0, std::invalid_argument
    ,"ProductVector::applyOp(...): Error, "
    "sub_dim_in = " << sub_dim_in << " is not acceptable" );
  TEST_FOR_EXCEPTION(
    global_offset_in < 0, std::invalid_argument
    ,"ProductVector::applyOp(...): Error, "
    "global_offset_in = " << global_offset_in << " is not acceptable" );
  TEST_FOR_EXCEPTION(
    sub_dim_in > 0 && sub_dim_in - (first_ele_in - 1) > n, std::length_error
    ,"ProductVector::applyOp(...): Error, "
    "global_offset_in = " << global_offset_in << ", sub_dim_in = " << sub_dim_in
    << "first_ele_in = " << first_ele_in << " and n = " << n
    << " are not compatible" );
  bool test_failed;
  for(int k = 0; k < num_vecs; ++k) {
    test_failed = !this->space()->isCompatible(*vecs[k]->space());
    TEST_FOR_EXCEPTION(
      test_failed, Exceptions::IncompatibleVectorSpaces
      ,"ProductVector::applyOp(...): Error vecs["<<k<<"]->space() "
      <<"of type \'"<<typeid(*vecs[k]->space()).name()<<"\' is not compatible with this "
      <<"\'VectorSpaceBlocked\' vector space!"
      );
  }
  for(int k = 0; k < num_targ_vecs; ++k) {
    test_failed = !this->space()->isCompatible(*targ_vecs[k]->space());
    TEST_FOR_EXCEPTION(
      test_failed, Exceptions::IncompatibleVectorSpaces
      ,"ProductVector::applyOp(...): Error targ_vecs["<<k<<"]->space() "
      <<"of type \'"<<typeid(*vecs[k]->space()).name()<<"\' is not compatible with this "
      <<"\'VectorSpaceBlocked\' vector space!"
      );
  }
#endif
  // Get the index of an incore-only input vector and input/output vector?
  const bool this_isInCore = productSpace_->isInCore();
  int incore_vec_k = -1, incore_targ_vec_k = -1;
  // Dynamic cast the pointers for the vector arguments
  Workspace<const ProductVector<Scalar>*>
    vecs_args(wss,num_vecs,false);
  for(int k = 0; k < num_vecs; ++k) {
    vecs_args[k] = dynamic_cast<const ProductVector<Scalar>*>(vecs[k]);
    if( vecs_args[k] == NULL ) {
      const bool isInCore_k = vecs[k]->space()->isInCore();
      if( this_isInCore && isInCore_k ) {
        incore_vec_k = k;
        break;
      }
      TEST_FOR_EXCEPTION(
        !this_isInCore || (this_isInCore && !isInCore_k), Exceptions::IncompatibleVectorSpaces
        ,"ProductVector::applyOp(...): Error vecs["<<k<<"] "
        <<"of type \'"<<typeid(*vecs[k]).name()<<"\' does not support the "
        <<"\'ProductVector<Scalar>\' interface and is not an incore vector or this is not an incore vector!"
        );
    }
  }
  Workspace<ProductVector<Scalar>*>
    targ_vecs_args(wss,num_targ_vecs,false);
  for(int k = 0; k < num_targ_vecs; ++k) {
    targ_vecs_args[k] = dynamic_cast<ProductVector<Scalar>*>(targ_vecs[k]);
    if( targ_vecs_args[k] == NULL ) {
      const bool isInCore_k = targ_vecs[k]->space()->isInCore();
      if( this_isInCore && isInCore_k ) {
        incore_targ_vec_k = k;
        break;
      }
      TEST_FOR_EXCEPTION(
        !this_isInCore || (this_isInCore && !isInCore_k), Exceptions::IncompatibleVectorSpaces
        ,"ProductVector::applyOp(...): Error targ_vecs["<<k<<"] "
        <<"of type \'"<<typeid(*targ_vecs[k]).name()<<"\' does not support the "
        <<"\'ProductVector<Scalar>\' interface and is not an incore vector or this is not an incore vector!"
        );
    }
  }
  // Let a incore-only vector handle this through explicit vector access
  // if all else fails?
  if( incore_vec_k >= 0 ) {
    vecs[incore_vec_k]->applyOp(
      op,num_vecs,vecs,num_targ_vecs,targ_vecs,reduct_obj
      ,first_ele_in,sub_dim_in,global_offset_in
      );
    return;
  }
  else if ( incore_targ_vec_k >= 0 ) {
    targ_vecs[incore_targ_vec_k]->applyOp(
      op,num_vecs,vecs,num_targ_vecs,targ_vecs,reduct_obj
      ,first_ele_in,sub_dim_in,global_offset_in
      );
    return;
  }
  // Perform the reduction on each vector segment at a time.
  const Index this_dim = n;
  const Index sub_dim  = ( sub_dim_in == 0
                           ? this_dim - (first_ele_in - 1)
                           : sub_dim_in );
  Index num_elements_remaining = sub_dim;
  const int  numBlocks = productSpace_->numBlocks();
  Workspace<const VectorBase<Scalar>*>
    sub_vecs(wss,num_vecs,false);
  Workspace<VectorBase<Scalar>*>
    sub_targ_vecs(wss,num_targ_vecs,false);
  Index g_off = -first_ele_in + 1;
  for(int k = 0; k < numBlocks; ++k) {
    const Index local_dim = productSpace_->getBlock(k)->dim();
    if( g_off < 0 && -g_off > local_dim ) {
      g_off += local_dim;
      continue;
    }
    const Index
      local_sub_dim = ( g_off >= 0
                        ? std::min( local_dim, num_elements_remaining )
                        : std::min( local_dim + g_off, num_elements_remaining ) );
    if( local_sub_dim <= 0 )
      break;
    for( int i = 0; i < num_vecs; ++i )       // Fill constituent vectors for block k
      sub_vecs[i] = &*vecs_args[i]->vecs_[k];
    for( int j = 0; j < num_targ_vecs; ++j )  // Fill constituent target vectors for block k
      sub_targ_vecs[j] = &*targ_vecs_args[j]->vecs_[k];
    Thyra::applyOp( 
      op
      ,num_vecs,num_vecs?&sub_vecs[0]:NULL
      ,num_targ_vecs,num_targ_vecs?&sub_targ_vecs[0]:NULL
      ,reduct_obj
      ,g_off < 0 ? -g_off + 1 : 1                                // first_ele
      ,local_sub_dim                                             // sub_dim
      ,g_off < 0 ? global_offset_in : global_offset_in + g_off   // global_offset
      );
    g_off += local_dim;
    num_elements_remaining -= local_sub_dim;
  }
  TEST_FOR_EXCEPT(!(num_elements_remaining==0));
}

template <class Scalar>
void ProductVector<Scalar>::getSubVector(
  const Range1D& rng_in, RTOpPack::SubVectorT<Scalar>* sub_vec
  ) const
{
  const Range1D
    rng = rng_in.full_range() ? Range1D( 1, productSpace_->dim()) : rng_in;
  int    kth_vector_space  = -1;
  Index  kth_global_offset = 0;
  productSpace_->getVecSpcPoss(rng.lbound(),&kth_vector_space,&kth_global_offset);
#ifdef _DEBUG
  TEST_FOR_EXCEPT( !( 0 <= kth_vector_space && kth_vector_space <= numBlocks_ ) );
#endif
  if( rng.lbound() + rng.size() <= kth_global_offset + 1 + vecs_[kth_vector_space]->space()->dim() ) {
    // This involves only one sub-vector so just return it.
    const_cast<const VectorBase<Scalar>*>(&*vecs_[kth_vector_space])->getSubVector(
      rng - kth_global_offset, sub_vec );
    sub_vec->setGlobalOffset( sub_vec->globalOffset() + kth_global_offset );
  }
  else {
    // Just let the default implementation handle this.  ToDo: In the future
    // we could manually construct an explicit sub-vector that spanned
    // two or more constituent vectors but this would be a lot of work.
    // However, this would require the use of temporary memory but
    // so what.
    VectorBase<Scalar>::getSubVector(rng_in,sub_vec);
  }
}

template <class Scalar>
void ProductVector<Scalar>::freeSubVector(
  RTOpPack::SubVectorT<Scalar>* sub_vec
  ) const
{
  if( sub_vec->values() == NULL ) return;
  int    kth_vector_space  = -1;
  Index  kth_global_offset = 0;
  productSpace_->getVecSpcPoss(sub_vec->globalOffset()+1,&kth_vector_space,&kth_global_offset);
#ifdef _DEBUG
  TEST_FOR_EXCEPT( !( 0 <= kth_vector_space && kth_vector_space <= numBlocks_ ) );
#endif
  if( sub_vec->globalOffset() + sub_vec->subDim() <= kth_global_offset +  vecs_[kth_vector_space]->space()->dim() ) {
    // This sub_vec was extracted from a single constituent vector
    sub_vec->setGlobalOffset( sub_vec->globalOffset() - kth_global_offset );
    vecs_[kth_vector_space]->freeSubVector(sub_vec);
  }
  else {
    // This sub_vec was created by the default implementation!
    VectorBase<Scalar>::freeSubVector(sub_vec);
  }
}

template <class Scalar>
void ProductVector<Scalar>::getSubVector(
  const Range1D& rng_in, RTOpPack::MutableSubVectorT<Scalar>* sub_vec
  )
{
  const Range1D
    rng = rng_in.full_range() ? Range1D( 1, productSpace_->dim()) : rng_in;
  int    kth_vector_space  = -1;
  Index  kth_global_offset = 0;
  productSpace_->getVecSpcPoss(rng.lbound(),&kth_vector_space,&kth_global_offset);
#ifdef _DEBUG
  TEST_FOR_EXCEPT( !( 0 <= kth_vector_space && kth_vector_space <= numBlocks_ ) );
#endif
  if( rng.lbound() + rng.size() <= kth_global_offset + 1 + vecs_[kth_vector_space]->space()->dim() ) {
    // This involves only one sub-vector so just return it.
    vecs_[kth_vector_space]->getSubVector(
      rng - kth_global_offset, sub_vec );
    sub_vec->setGlobalOffset( sub_vec->globalOffset() + kth_global_offset );
  }
  else {
    // Just let the default implementation handle this.  ToDo: In the future
    // we could manually construct an explicit sub-vector that spanned
    // two or more constituent vectors but this would be a lot of work.
    // However, this would require the use of temporary memory but
    // so what.
    VectorBase<Scalar>::getSubVector(rng_in,sub_vec);
  }
}

template <class Scalar>
void ProductVector<Scalar>::commitSubVector(
  RTOpPack::MutableSubVectorT<Scalar>* sub_vec
  )
{
  if( sub_vec->values() == NULL ) return;
  int    kth_vector_space  = -1;
  Index  kth_global_offset = 0;
  productSpace_->getVecSpcPoss(sub_vec->globalOffset()+1,&kth_vector_space,&kth_global_offset);
#ifdef _DEBUG
  TEST_FOR_EXCEPT( !( 0 <= kth_vector_space && kth_vector_space <= numBlocks_ ) );
#endif
  if( sub_vec->globalOffset() + sub_vec->subDim() <= kth_global_offset +  vecs_[kth_vector_space]->space()->dim() ) {
    // This sub_vec was extracted from a single constituent vector
    sub_vec->setGlobalOffset( sub_vec->globalOffset() - kth_global_offset );
    vecs_[kth_vector_space]->commitSubVector(sub_vec);
  }
  else {
    // This sub_vec was created by the default implementation!
    VectorBase<Scalar>::commitSubVector(sub_vec);
  }
}

template <class Scalar>
void ProductVector<Scalar>::setSubVector(
  const RTOpPack::SparseSubVectorT<Scalar>& sub_vec
  )
{
  int    kth_vector_space  = -1;
  Index  kth_global_offset = 0;
  productSpace_->getVecSpcPoss(sub_vec.globalOffset()+1,&kth_vector_space,&kth_global_offset);
#ifdef _DEBUG
  TEST_FOR_EXCEPT( !( 0 <= kth_vector_space && kth_vector_space <= numBlocks_ ) );
#endif
  if( sub_vec.globalOffset() + sub_vec.subDim() <= kth_global_offset + vecs_[kth_vector_space]->space()->dim() ) {
    // This sub-vector fits into a single constituent vector
    RTOpPack::SparseSubVectorT<Scalar> sub_vec_g = sub_vec;
    sub_vec_g.setGlobalOffset( sub_vec_g.globalOffset() - kth_global_offset );
    vecs_[kth_vector_space]->setSubVector(sub_vec_g);
  }
  else {
    // Let the default implementation take care of this.  ToDo: In the future
    // it would be possible to manually set the relevant constituent
    // vectors with no temp memory allocations.
    VectorBase<Scalar>::setSubVector(sub_vec);
  }
}

} // namespace Thyra

#endif // THYRA_PRODUCT_VECTOR_HPP

---