phynexis-deps/armadillo-code
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
de7ac59cd1e7a678ae4a4eaeb46bb2afb15ab3dc
armadillo-code/include/armadillo_bits/glue_affmul_meat.hpp
T

489 lines
13 KiB
C++
Raw Blame History

// 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_affmul
//! @{
template<typename T1, typename T2>
inline
void
glue_affmul::apply(Mat<typename T1::elem_type>& out, const Glue<T1,T2,glue_affmul>& X)
{
arma_extra_debug_sigprint();
typedef typename T1::elem_type eT;
const quasi_unwrap<T1> U1(X.A);
const quasi_unwrap<T2> U2(X.B);
const bool is_alias = (U1.is_alias(out) || U2.is_alias(out));
if(is_alias == false)
{
glue_affmul::apply_noalias(out, U1.M, U2.M);
}
else
{
Mat<eT> tmp;
glue_affmul::apply_noalias(tmp, U1.M, U2.M);
out.steal_mem(tmp);
}
}
template<typename T1, typename T2>
inline
void
glue_affmul::apply_noalias(Mat<typename T1::elem_type>& out, const T1& A, const T2& B)
{
arma_extra_debug_sigprint();
const uword A_n_cols = A.n_cols;
const uword A_n_rows = A.n_rows;
const uword B_n_rows = B.n_rows;
arma_debug_check( (A_n_cols != B_n_rows+1), "affmul(): size mismatch" );
if(A_n_rows == A_n_cols)
{
glue_affmul::apply_noalias_square(out, A, B);
}
else
if(A_n_rows == B_n_rows)
{
glue_affmul::apply_noalias_rectangle(out, A, B);
}
else
{
glue_affmul::apply_noalias_generic(out, A, B);
}
}
template<typename T1, typename T2>
inline
void
glue_affmul::apply_noalias_square(Mat<typename T1::elem_type>& out, const T1& A, const T2& B)
{
arma_extra_debug_sigprint();
typedef typename T1::elem_type eT;
// assuming that A is square sized, and A.n_cols = B.n_rows+1
const uword N = A.n_rows;
const uword B_n_cols = B.n_cols;
out.set_size(N, B_n_cols);
if(out.n_elem == 0) { return; }
const eT* A_mem = A.memptr();
switch(N)
{
case 0:
break;
case 1: // A is 1x1
out.fill(A_mem[0]);
break;
case 2: // A is 2x2
{
if(B_n_cols == 1)
{
const eT* B_mem = B.memptr();
eT* out_mem = out.memptr();
const eT x = B_mem[0];
out_mem[0] = A_mem[0]*x + A_mem[2];
out_mem[1] = A_mem[1]*x + A_mem[3];
}
else
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* out_mem = out.colptr(col);
const eT x = B_mem[0];
out_mem[0] = A_mem[0]*x + A_mem[2];
out_mem[1] = A_mem[1]*x + A_mem[3];
}
}
break;
case 3: // A is 3x3
{
if(B_n_cols == 1)
{
const eT* B_mem = B.memptr();
eT* out_mem = out.memptr();
const eT x = B_mem[0];
const eT y = B_mem[1];
out_mem[0] = A_mem[0]*x + A_mem[3]*y + A_mem[6];
out_mem[1] = A_mem[1]*x + A_mem[4]*y + A_mem[7];
out_mem[2] = A_mem[2]*x + A_mem[5]*y + A_mem[8];
}
else
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* out_mem = out.colptr(col);
const eT x = B_mem[0];
const eT y = B_mem[1];
out_mem[0] = A_mem[0]*x + A_mem[3]*y + A_mem[6];
out_mem[1] = A_mem[1]*x + A_mem[4]*y + A_mem[7];
out_mem[2] = A_mem[2]*x + A_mem[5]*y + A_mem[8];
}
}
break;
case 4: // A is 4x4
{
if(B_n_cols == 1)
{
const eT* B_mem = B.memptr();
eT* out_mem = out.memptr();
const eT x = B_mem[0];
const eT y = B_mem[1];
const eT z = B_mem[2];
out_mem[0] = A_mem[ 0]*x + A_mem[ 4]*y + A_mem[ 8]*z + A_mem[12];
out_mem[1] = A_mem[ 1]*x + A_mem[ 5]*y + A_mem[ 9]*z + A_mem[13];
out_mem[2] = A_mem[ 2]*x + A_mem[ 6]*y + A_mem[10]*z + A_mem[14];
out_mem[3] = A_mem[ 3]*x + A_mem[ 7]*y + A_mem[11]*z + A_mem[15];
}
else
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* out_mem = out.colptr(col);
const eT x = B_mem[0];
const eT y = B_mem[1];
const eT z = B_mem[2];
out_mem[0] = A_mem[ 0]*x + A_mem[ 4]*y + A_mem[ 8]*z + A_mem[12];
out_mem[1] = A_mem[ 1]*x + A_mem[ 5]*y + A_mem[ 9]*z + A_mem[13];
out_mem[2] = A_mem[ 2]*x + A_mem[ 6]*y + A_mem[10]*z + A_mem[14];
out_mem[3] = A_mem[ 3]*x + A_mem[ 7]*y + A_mem[11]*z + A_mem[15];
}
}
break;
case 5: // A is 5x5
{
if(B_n_cols == 1)
{
const eT* B_mem = B.memptr();
eT* out_mem = out.memptr();
const eT x = B_mem[0];
const eT y = B_mem[1];
const eT z = B_mem[2];
const eT w = B_mem[3];
out_mem[0] = A_mem[ 0]*x + A_mem[ 5]*y + A_mem[10]*z + A_mem[15]*w + A_mem[20];
out_mem[1] = A_mem[ 1]*x + A_mem[ 6]*y + A_mem[11]*z + A_mem[16]*w + A_mem[21];
out_mem[2] = A_mem[ 2]*x + A_mem[ 7]*y + A_mem[12]*z + A_mem[17]*w + A_mem[22];
out_mem[3] = A_mem[ 3]*x + A_mem[ 8]*y + A_mem[13]*z + A_mem[18]*w + A_mem[23];
out_mem[4] = A_mem[ 4]*x + A_mem[ 9]*y + A_mem[14]*z + A_mem[19]*w + A_mem[24];
}
else
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* out_mem = out.colptr(col);
const eT x = B_mem[0];
const eT y = B_mem[1];
const eT z = B_mem[2];
const eT w = B_mem[3];
out_mem[0] = A_mem[ 0]*x + A_mem[ 5]*y + A_mem[10]*z + A_mem[15]*w + A_mem[20];
out_mem[1] = A_mem[ 1]*x + A_mem[ 6]*y + A_mem[11]*z + A_mem[16]*w + A_mem[21];
out_mem[2] = A_mem[ 2]*x + A_mem[ 7]*y + A_mem[12]*z + A_mem[17]*w + A_mem[22];
out_mem[3] = A_mem[ 3]*x + A_mem[ 8]*y + A_mem[13]*z + A_mem[18]*w + A_mem[23];
out_mem[4] = A_mem[ 4]*x + A_mem[ 9]*y + A_mem[14]*z + A_mem[19]*w + A_mem[24];
}
}
break;
default:
{
if(B_n_cols == 1)
{
Col<eT> tmp(N, arma_nozeros_indicator());
eT* tmp_mem = tmp.memptr();
arrayops::copy(tmp_mem, B.memptr(), N-1);
tmp_mem[N-1] = eT(1);
out = A * tmp;
}
else
{
Mat<eT> tmp(N, B_n_cols, arma_nozeros_indicator());
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* tmp_mem = tmp.colptr(col);
arrayops::copy(tmp_mem, B_mem, N-1);
tmp_mem[N-1] = eT(1);
}
out = A * tmp;
}
}
}
}
template<typename T1, typename T2>
inline
void
glue_affmul::apply_noalias_rectangle(Mat<typename T1::elem_type>& out, const T1& A, const T2& B)
{
arma_extra_debug_sigprint();
typedef typename T1::elem_type eT;
// assuming that A.n_rows = A.n_cols-1, and A.n_cols = B.n_rows+1
// (A and B have the same number of rows)
const uword A_n_rows = A.n_rows;
const uword B_n_cols = B.n_cols;
out.set_size(A_n_rows, B_n_cols);
if(out.n_elem == 0) { return; }
const eT* A_mem = A.memptr();
switch(A_n_rows)
{
case 0:
break;
case 1: // A is 1x2
{
if(B_n_cols == 1)
{
const eT* B_mem = B.memptr();
eT* out_mem = out.memptr();
const eT x = B_mem[0];
out_mem[0] = A_mem[0]*x + A_mem[1];
}
else
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* out_mem = out.colptr(col);
const eT x = B_mem[0];
out_mem[0] = A_mem[0]*x + A_mem[1];
}
}
break;
case 2: // A is 2x3
{
if(B_n_cols == 1)
{
const eT* B_mem = B.memptr();
eT* out_mem = out.memptr();
const eT x = B_mem[0];
const eT y = B_mem[1];
out_mem[0] = A_mem[0]*x + A_mem[2]*y + A_mem[4];
out_mem[1] = A_mem[1]*x + A_mem[3]*y + A_mem[5];
}
else
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* out_mem = out.colptr(col);
const eT x = B_mem[0];
const eT y = B_mem[1];
out_mem[0] = A_mem[0]*x + A_mem[2]*y + A_mem[4];
out_mem[1] = A_mem[1]*x + A_mem[3]*y + A_mem[5];
}
}
break;
case 3: // A is 3x4
{
if(B_n_cols == 1)
{
const eT* B_mem = B.memptr();
eT* out_mem = out.memptr();
const eT x = B_mem[0];
const eT y = B_mem[1];
const eT z = B_mem[2];
out_mem[0] = A_mem[ 0]*x + A_mem[ 3]*y + A_mem[ 6]*z + A_mem[ 9];
out_mem[1] = A_mem[ 1]*x + A_mem[ 4]*y + A_mem[ 7]*z + A_mem[10];
out_mem[2] = A_mem[ 2]*x + A_mem[ 5]*y + A_mem[ 8]*z + A_mem[11];
}
else
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* out_mem = out.colptr(col);
const eT x = B_mem[0];
const eT y = B_mem[1];
const eT z = B_mem[2];
out_mem[0] = A_mem[ 0]*x + A_mem[ 3]*y + A_mem[ 6]*z + A_mem[ 9];
out_mem[1] = A_mem[ 1]*x + A_mem[ 4]*y + A_mem[ 7]*z + A_mem[10];
out_mem[2] = A_mem[ 2]*x + A_mem[ 5]*y + A_mem[ 8]*z + A_mem[11];
}
}
break;
case 4: // A is 4x5
{
if(B_n_cols == 1)
{
const eT* B_mem = B.memptr();
eT* out_mem = out.memptr();
const eT x = B_mem[0];
const eT y = B_mem[1];
const eT z = B_mem[2];
const eT w = B_mem[3];
out_mem[0] = A_mem[ 0]*x + A_mem[ 4]*y + A_mem[ 8]*z + A_mem[12]*w + A_mem[16];
out_mem[1] = A_mem[ 1]*x + A_mem[ 5]*y + A_mem[ 9]*z + A_mem[13]*w + A_mem[17];
out_mem[2] = A_mem[ 2]*x + A_mem[ 6]*y + A_mem[10]*z + A_mem[14]*w + A_mem[18];
out_mem[3] = A_mem[ 3]*x + A_mem[ 7]*y + A_mem[11]*z + A_mem[15]*w + A_mem[19];
}
else
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* out_mem = out.colptr(col);
const eT x = B_mem[0];
const eT y = B_mem[1];
const eT z = B_mem[2];
const eT w = B_mem[3];
out_mem[0] = A_mem[ 0]*x + A_mem[ 4]*y + A_mem[ 8]*z + A_mem[12]*w + A_mem[16];
out_mem[1] = A_mem[ 1]*x + A_mem[ 5]*y + A_mem[ 9]*z + A_mem[13]*w + A_mem[17];
out_mem[2] = A_mem[ 2]*x + A_mem[ 6]*y + A_mem[10]*z + A_mem[14]*w + A_mem[18];
out_mem[3] = A_mem[ 3]*x + A_mem[ 7]*y + A_mem[11]*z + A_mem[15]*w + A_mem[19];
}
}
break;
default:
{
const uword A_n_cols = A.n_cols;
if(B_n_cols == 1)
{
Col<eT> tmp(A_n_cols, arma_nozeros_indicator());
eT* tmp_mem = tmp.memptr();
arrayops::copy(tmp_mem, B.memptr(), A_n_cols-1);
tmp_mem[A_n_cols-1] = eT(1);
out = A * tmp;
}
else
{
Mat<eT> tmp(A_n_cols, B_n_cols, arma_nozeros_indicator());
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* tmp_mem = tmp.colptr(col);
arrayops::copy(tmp_mem, B_mem, A_n_cols-1);
tmp_mem[A_n_cols-1] = eT(1);
}
out = A * tmp;
}
}
}
}
template<typename T1, typename T2>
inline
void
glue_affmul::apply_noalias_generic(Mat<typename T1::elem_type>& out, const T1& A, const T2& B)
{
arma_extra_debug_sigprint();
typedef typename T1::elem_type eT;
// assuming that A.n_cols = B.n_rows+1
const uword B_n_rows = B.n_rows;
const uword B_n_cols = B.n_cols;
Mat<eT> tmp(B_n_rows+1, B_n_cols, arma_nozeros_indicator());
for(uword col=0; col < B_n_cols; ++col)
{
const eT* B_mem = B.colptr(col);
eT* tmp_mem = tmp.colptr(col);
arrayops::copy(tmp_mem, B_mem, B_n_rows);
tmp_mem[B_n_rows] = eT(1);
}
out = A * tmp;
}
//! @}
Reference in New Issue View Git Blame Copy Permalink