561 lines
15 KiB
C++
561 lines
15 KiB
C++
// SPDX-License-Identifier: Apache-2.0
|
|
//
|
|
// Copyright 2008-2016 Conrad Sanderson (http://conradsanderson.id.au)
|
|
// Copyright 2008-2016 National ICT Australia (NICTA)
|
|
//
|
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
|
// you may not use this file except in compliance with the License.
|
|
// You may obtain a copy of the License at
|
|
// http://www.apache.org/licenses/LICENSE-2.0
|
|
//
|
|
// Unless required by applicable law or agreed to in writing, software
|
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
// See the License for the specific language governing permissions and
|
|
// limitations under the License.
|
|
// ------------------------------------------------------------------------
|
|
|
|
|
|
//! \addtogroup glue_mixed
|
|
//! @{
|
|
|
|
|
|
|
|
//! matrix multiplication with different element types
|
|
template<typename T1, typename T2>
|
|
inline
|
|
void
|
|
glue_mixed_times::apply(Mat<typename eT_promoter<T1,T2>::eT>& out, const mtGlue<typename eT_promoter<T1,T2>::eT, T1, T2, glue_mixed_times>& X)
|
|
{
|
|
arma_extra_debug_sigprint();
|
|
|
|
typedef typename T1::elem_type in_eT1;
|
|
typedef typename T2::elem_type in_eT2;
|
|
|
|
typedef typename eT_promoter<T1,T2>::eT out_eT;
|
|
|
|
const partial_unwrap<T1> tmp1(X.A);
|
|
const partial_unwrap<T2> tmp2(X.B);
|
|
|
|
const typename partial_unwrap<T1>::stored_type& A = tmp1.M;
|
|
const typename partial_unwrap<T2>::stored_type& B = tmp2.M;
|
|
|
|
const bool use_alpha = partial_unwrap<T1>::do_times || partial_unwrap<T2>::do_times;
|
|
const out_eT alpha = use_alpha ? (upgrade_val<in_eT1,in_eT2>::apply(tmp1.get_val()) * upgrade_val<in_eT1,in_eT2>::apply(tmp2.get_val())) : out_eT(0);
|
|
|
|
const bool do_trans_A = partial_unwrap<T1>::do_trans;
|
|
const bool do_trans_B = partial_unwrap<T2>::do_trans;
|
|
|
|
arma_debug_assert_trans_mul_size<do_trans_A, do_trans_B>(A.n_rows, A.n_cols, B.n_rows, B.n_cols, "matrix multiplication");
|
|
|
|
const uword out_n_rows = (do_trans_A == false) ? A.n_rows : A.n_cols;
|
|
const uword out_n_cols = (do_trans_B == false) ? B.n_cols : B.n_rows;
|
|
|
|
const bool alias = tmp1.is_alias(out) || tmp2.is_alias(out);
|
|
|
|
if(alias == false)
|
|
{
|
|
out.set_size(out_n_rows, out_n_cols);
|
|
|
|
gemm_mixed<do_trans_A, do_trans_B, use_alpha, false>::apply(out, A, B, alpha);
|
|
}
|
|
else
|
|
{
|
|
Mat<out_eT> tmp(out_n_rows, out_n_cols, arma_nozeros_indicator());
|
|
|
|
gemm_mixed<do_trans_A, do_trans_B, use_alpha, false>::apply(tmp, A, B, alpha);
|
|
|
|
out.steal_mem(tmp);
|
|
}
|
|
}
|
|
|
|
|
|
|
|
//! matrix addition with different element types
|
|
template<typename T1, typename T2>
|
|
inline
|
|
void
|
|
glue_mixed_plus::apply(Mat<typename eT_promoter<T1,T2>::eT>& out, const mtGlue<typename eT_promoter<T1,T2>::eT, T1, T2, glue_mixed_plus>& X)
|
|
{
|
|
arma_extra_debug_sigprint();
|
|
|
|
typedef typename T1::elem_type eT1;
|
|
typedef typename T2::elem_type eT2;
|
|
|
|
typedef typename promote_type<eT1,eT2>::result out_eT;
|
|
|
|
promote_type<eT1,eT2>::check();
|
|
|
|
const Proxy<T1> A(X.A);
|
|
const Proxy<T2> B(X.B);
|
|
|
|
arma_debug_assert_same_size(A, B, "addition");
|
|
|
|
const uword n_rows = A.get_n_rows();
|
|
const uword n_cols = A.get_n_cols();
|
|
|
|
out.set_size(n_rows, n_cols);
|
|
|
|
out_eT* out_mem = out.memptr();
|
|
const uword n_elem = out.n_elem;
|
|
|
|
const bool use_at = (Proxy<T1>::use_at || Proxy<T2>::use_at);
|
|
|
|
if(use_at == false)
|
|
{
|
|
typename Proxy<T1>::ea_type AA = A.get_ea();
|
|
typename Proxy<T2>::ea_type BB = B.get_ea();
|
|
|
|
if(memory::is_aligned(out_mem))
|
|
{
|
|
memory::mark_as_aligned(out_mem);
|
|
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) + upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) + upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword col=0; col < n_cols; ++col)
|
|
for(uword row=0; row < n_rows; ++row)
|
|
{
|
|
(*out_mem) = upgrade_val<eT1,eT2>::apply(A.at(row,col)) + upgrade_val<eT1,eT2>::apply(B.at(row,col));
|
|
out_mem++;
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
|
|
//! matrix subtraction with different element types
|
|
template<typename T1, typename T2>
|
|
inline
|
|
void
|
|
glue_mixed_minus::apply(Mat<typename eT_promoter<T1,T2>::eT>& out, const mtGlue<typename eT_promoter<T1,T2>::eT, T1, T2, glue_mixed_minus>& X)
|
|
{
|
|
arma_extra_debug_sigprint();
|
|
|
|
typedef typename T1::elem_type eT1;
|
|
typedef typename T2::elem_type eT2;
|
|
|
|
typedef typename promote_type<eT1,eT2>::result out_eT;
|
|
|
|
promote_type<eT1,eT2>::check();
|
|
|
|
const Proxy<T1> A(X.A);
|
|
const Proxy<T2> B(X.B);
|
|
|
|
arma_debug_assert_same_size(A, B, "subtraction");
|
|
|
|
const uword n_rows = A.get_n_rows();
|
|
const uword n_cols = A.get_n_cols();
|
|
|
|
out.set_size(n_rows, n_cols);
|
|
|
|
out_eT* out_mem = out.memptr();
|
|
const uword n_elem = out.n_elem;
|
|
|
|
const bool use_at = (Proxy<T1>::use_at || Proxy<T2>::use_at);
|
|
|
|
if(use_at == false)
|
|
{
|
|
typename Proxy<T1>::ea_type AA = A.get_ea();
|
|
typename Proxy<T2>::ea_type BB = B.get_ea();
|
|
|
|
if(memory::is_aligned(out_mem))
|
|
{
|
|
memory::mark_as_aligned(out_mem);
|
|
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) - upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) - upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword col=0; col < n_cols; ++col)
|
|
for(uword row=0; row < n_rows; ++row)
|
|
{
|
|
(*out_mem) = upgrade_val<eT1,eT2>::apply(A.at(row,col)) - upgrade_val<eT1,eT2>::apply(B.at(row,col));
|
|
out_mem++;
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
|
|
//! element-wise matrix division with different element types
|
|
template<typename T1, typename T2>
|
|
inline
|
|
void
|
|
glue_mixed_div::apply(Mat<typename eT_promoter<T1,T2>::eT>& out, const mtGlue<typename eT_promoter<T1,T2>::eT, T1, T2, glue_mixed_div>& X)
|
|
{
|
|
arma_extra_debug_sigprint();
|
|
|
|
typedef typename T1::elem_type eT1;
|
|
typedef typename T2::elem_type eT2;
|
|
|
|
typedef typename promote_type<eT1,eT2>::result out_eT;
|
|
|
|
promote_type<eT1,eT2>::check();
|
|
|
|
const Proxy<T1> A(X.A);
|
|
const Proxy<T2> B(X.B);
|
|
|
|
arma_debug_assert_same_size(A, B, "element-wise division");
|
|
|
|
const uword n_rows = A.get_n_rows();
|
|
const uword n_cols = A.get_n_cols();
|
|
|
|
out.set_size(n_rows, n_cols);
|
|
|
|
out_eT* out_mem = out.memptr();
|
|
const uword n_elem = out.n_elem;
|
|
|
|
const bool use_at = (Proxy<T1>::use_at || Proxy<T2>::use_at);
|
|
|
|
if(use_at == false)
|
|
{
|
|
typename Proxy<T1>::ea_type AA = A.get_ea();
|
|
typename Proxy<T2>::ea_type BB = B.get_ea();
|
|
|
|
if(memory::is_aligned(out_mem))
|
|
{
|
|
memory::mark_as_aligned(out_mem);
|
|
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) / upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) / upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword col=0; col < n_cols; ++col)
|
|
for(uword row=0; row < n_rows; ++row)
|
|
{
|
|
(*out_mem) = upgrade_val<eT1,eT2>::apply(A.at(row,col)) / upgrade_val<eT1,eT2>::apply(B.at(row,col));
|
|
out_mem++;
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
|
|
//! element-wise matrix multiplication with different element types
|
|
template<typename T1, typename T2>
|
|
inline
|
|
void
|
|
glue_mixed_schur::apply(Mat<typename eT_promoter<T1,T2>::eT>& out, const mtGlue<typename eT_promoter<T1,T2>::eT, T1, T2, glue_mixed_schur>& X)
|
|
{
|
|
arma_extra_debug_sigprint();
|
|
|
|
typedef typename T1::elem_type eT1;
|
|
typedef typename T2::elem_type eT2;
|
|
|
|
typedef typename promote_type<eT1,eT2>::result out_eT;
|
|
|
|
promote_type<eT1,eT2>::check();
|
|
|
|
const Proxy<T1> A(X.A);
|
|
const Proxy<T2> B(X.B);
|
|
|
|
arma_debug_assert_same_size(A, B, "element-wise multiplication");
|
|
|
|
const uword n_rows = A.get_n_rows();
|
|
const uword n_cols = A.get_n_cols();
|
|
|
|
out.set_size(n_rows, n_cols);
|
|
|
|
out_eT* out_mem = out.memptr();
|
|
const uword n_elem = out.n_elem;
|
|
|
|
const bool use_at = (Proxy<T1>::use_at || Proxy<T2>::use_at);
|
|
|
|
if(use_at == false)
|
|
{
|
|
typename Proxy<T1>::ea_type AA = A.get_ea();
|
|
typename Proxy<T2>::ea_type BB = B.get_ea();
|
|
|
|
if(memory::is_aligned(out_mem))
|
|
{
|
|
memory::mark_as_aligned(out_mem);
|
|
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) * upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) * upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword col=0; col < n_cols; ++col)
|
|
for(uword row=0; row < n_rows; ++row)
|
|
{
|
|
(*out_mem) = upgrade_val<eT1,eT2>::apply(A.at(row,col)) * upgrade_val<eT1,eT2>::apply(B.at(row,col));
|
|
out_mem++;
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
|
|
//
|
|
//
|
|
//
|
|
|
|
|
|
|
|
//! cube addition with different element types
|
|
template<typename T1, typename T2>
|
|
inline
|
|
void
|
|
glue_mixed_plus::apply(Cube<typename eT_promoter<T1,T2>::eT>& out, const mtGlueCube<typename eT_promoter<T1,T2>::eT, T1, T2, glue_mixed_plus>& X)
|
|
{
|
|
arma_extra_debug_sigprint();
|
|
|
|
typedef typename T1::elem_type eT1;
|
|
typedef typename T2::elem_type eT2;
|
|
|
|
typedef typename promote_type<eT1,eT2>::result out_eT;
|
|
|
|
promote_type<eT1,eT2>::check();
|
|
|
|
const ProxyCube<T1> A(X.A);
|
|
const ProxyCube<T2> B(X.B);
|
|
|
|
arma_debug_assert_same_size(A, B, "addition");
|
|
|
|
const uword n_rows = A.get_n_rows();
|
|
const uword n_cols = A.get_n_cols();
|
|
const uword n_slices = A.get_n_slices();
|
|
|
|
out.set_size(n_rows, n_cols, n_slices);
|
|
|
|
out_eT* out_mem = out.memptr();
|
|
const uword n_elem = out.n_elem;
|
|
|
|
const bool use_at = (ProxyCube<T1>::use_at || ProxyCube<T2>::use_at);
|
|
|
|
if(use_at == false)
|
|
{
|
|
typename ProxyCube<T1>::ea_type AA = A.get_ea();
|
|
typename ProxyCube<T2>::ea_type BB = B.get_ea();
|
|
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) + upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword slice = 0; slice < n_slices; ++slice)
|
|
for(uword col = 0; col < n_cols; ++col )
|
|
for(uword row = 0; row < n_rows; ++row )
|
|
{
|
|
(*out_mem) = upgrade_val<eT1,eT2>::apply(A.at(row,col,slice)) + upgrade_val<eT1,eT2>::apply(B.at(row,col,slice));
|
|
out_mem++;
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
|
|
//! cube subtraction with different element types
|
|
template<typename T1, typename T2>
|
|
inline
|
|
void
|
|
glue_mixed_minus::apply(Cube<typename eT_promoter<T1,T2>::eT>& out, const mtGlueCube<typename eT_promoter<T1,T2>::eT, T1, T2, glue_mixed_minus>& X)
|
|
{
|
|
arma_extra_debug_sigprint();
|
|
|
|
typedef typename T1::elem_type eT1;
|
|
typedef typename T2::elem_type eT2;
|
|
|
|
typedef typename promote_type<eT1,eT2>::result out_eT;
|
|
|
|
promote_type<eT1,eT2>::check();
|
|
|
|
const ProxyCube<T1> A(X.A);
|
|
const ProxyCube<T2> B(X.B);
|
|
|
|
arma_debug_assert_same_size(A, B, "subtraction");
|
|
|
|
const uword n_rows = A.get_n_rows();
|
|
const uword n_cols = A.get_n_cols();
|
|
const uword n_slices = A.get_n_slices();
|
|
|
|
out.set_size(n_rows, n_cols, n_slices);
|
|
|
|
out_eT* out_mem = out.memptr();
|
|
const uword n_elem = out.n_elem;
|
|
|
|
const bool use_at = (ProxyCube<T1>::use_at || ProxyCube<T2>::use_at);
|
|
|
|
if(use_at == false)
|
|
{
|
|
typename ProxyCube<T1>::ea_type AA = A.get_ea();
|
|
typename ProxyCube<T2>::ea_type BB = B.get_ea();
|
|
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) - upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword slice = 0; slice < n_slices; ++slice)
|
|
for(uword col = 0; col < n_cols; ++col )
|
|
for(uword row = 0; row < n_rows; ++row )
|
|
{
|
|
(*out_mem) = upgrade_val<eT1,eT2>::apply(A.at(row,col,slice)) - upgrade_val<eT1,eT2>::apply(B.at(row,col,slice));
|
|
out_mem++;
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
|
|
//! element-wise cube division with different element types
|
|
template<typename T1, typename T2>
|
|
inline
|
|
void
|
|
glue_mixed_div::apply(Cube<typename eT_promoter<T1,T2>::eT>& out, const mtGlueCube<typename eT_promoter<T1,T2>::eT, T1, T2, glue_mixed_div>& X)
|
|
{
|
|
arma_extra_debug_sigprint();
|
|
|
|
typedef typename T1::elem_type eT1;
|
|
typedef typename T2::elem_type eT2;
|
|
|
|
typedef typename promote_type<eT1,eT2>::result out_eT;
|
|
|
|
promote_type<eT1,eT2>::check();
|
|
|
|
const ProxyCube<T1> A(X.A);
|
|
const ProxyCube<T2> B(X.B);
|
|
|
|
arma_debug_assert_same_size(A, B, "element-wise division");
|
|
|
|
const uword n_rows = A.get_n_rows();
|
|
const uword n_cols = A.get_n_cols();
|
|
const uword n_slices = A.get_n_slices();
|
|
|
|
out.set_size(n_rows, n_cols, n_slices);
|
|
|
|
out_eT* out_mem = out.memptr();
|
|
const uword n_elem = out.n_elem;
|
|
|
|
const bool use_at = (ProxyCube<T1>::use_at || ProxyCube<T2>::use_at);
|
|
|
|
if(use_at == false)
|
|
{
|
|
typename ProxyCube<T1>::ea_type AA = A.get_ea();
|
|
typename ProxyCube<T2>::ea_type BB = B.get_ea();
|
|
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) / upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword slice = 0; slice < n_slices; ++slice)
|
|
for(uword col = 0; col < n_cols; ++col )
|
|
for(uword row = 0; row < n_rows; ++row )
|
|
{
|
|
(*out_mem) = upgrade_val<eT1,eT2>::apply(A.at(row,col,slice)) / upgrade_val<eT1,eT2>::apply(B.at(row,col,slice));
|
|
out_mem++;
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
|
|
//! element-wise cube multiplication with different element types
|
|
template<typename T1, typename T2>
|
|
inline
|
|
void
|
|
glue_mixed_schur::apply(Cube<typename eT_promoter<T1,T2>::eT>& out, const mtGlueCube<typename eT_promoter<T1,T2>::eT, T1, T2, glue_mixed_schur>& X)
|
|
{
|
|
arma_extra_debug_sigprint();
|
|
|
|
typedef typename T1::elem_type eT1;
|
|
typedef typename T2::elem_type eT2;
|
|
|
|
typedef typename promote_type<eT1,eT2>::result out_eT;
|
|
|
|
promote_type<eT1,eT2>::check();
|
|
|
|
const ProxyCube<T1> A(X.A);
|
|
const ProxyCube<T2> B(X.B);
|
|
|
|
arma_debug_assert_same_size(A, B, "element-wise multiplication");
|
|
|
|
const uword n_rows = A.get_n_rows();
|
|
const uword n_cols = A.get_n_cols();
|
|
const uword n_slices = A.get_n_slices();
|
|
|
|
out.set_size(n_rows, n_cols, n_slices);
|
|
|
|
out_eT* out_mem = out.memptr();
|
|
const uword n_elem = out.n_elem;
|
|
|
|
const bool use_at = (ProxyCube<T1>::use_at || ProxyCube<T2>::use_at);
|
|
|
|
if(use_at == false)
|
|
{
|
|
typename ProxyCube<T1>::ea_type AA = A.get_ea();
|
|
typename ProxyCube<T2>::ea_type BB = B.get_ea();
|
|
|
|
for(uword i=0; i<n_elem; ++i)
|
|
{
|
|
out_mem[i] = upgrade_val<eT1,eT2>::apply(AA[i]) * upgrade_val<eT1,eT2>::apply(BB[i]);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for(uword slice = 0; slice < n_slices; ++slice)
|
|
for(uword col = 0; col < n_cols; ++col )
|
|
for(uword row = 0; row < n_rows; ++row )
|
|
{
|
|
(*out_mem) = upgrade_val<eT1,eT2>::apply(A.at(row,col,slice)) * upgrade_val<eT1,eT2>::apply(B.at(row,col,slice));
|
|
out_mem++;
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
|
|
//! @}
|