1 // SPDX-License-Identifier: Apache-2.0
2 //
3 // Copyright 2008-2016 Conrad Sanderson (http://conradsanderson.id.au)
4 // Copyright 2008-2016 National ICT Australia (NICTA)
5 //
6 // Licensed under the Apache License, Version 2.0 (the "License");
7 // you may not use this file except in compliance with the License.
8 // You may obtain a copy of the License at
9 // http://www.apache.org/licenses/LICENSE-2.0
10 //
11 // Unless required by applicable law or agreed to in writing, software
12 // distributed under the License is distributed on an "AS IS" BASIS,
13 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 // See the License for the specific language governing permissions and
15 // limitations under the License.
16 // ------------------------------------------------------------------------
17
18
19 //! \addtogroup glue_atan2
20 //! @{
21
22
23
24 template<typename T1, typename T2>
25 inline
26 void
apply(Mat<typename T1::elem_type> & out,const Glue<T1,T2,glue_atan2> & expr)27 glue_atan2::apply(Mat<typename T1::elem_type>& out, const Glue<T1, T2, glue_atan2>& expr)
28 {
29 arma_extra_debug_sigprint();
30
31 typedef typename T1::elem_type eT;
32
33 const Proxy<T1> P1(expr.A);
34 const Proxy<T2> P2(expr.B);
35
36 arma_assert_same_size(P1, P2, "atan2()");
37
38 const bool bad_alias = ( (Proxy<T1>::has_subview && P1.is_alias(out)) || (Proxy<T2>::has_subview && P2.is_alias(out)) );
39
40 if(bad_alias == false)
41 {
42 glue_atan2::apply_noalias(out, P1, P2);
43 }
44 else
45 {
46 Mat<eT> tmp;
47
48 glue_atan2::apply_noalias(tmp, P1, P2);
49
50 out.steal_mem(tmp);
51 }
52 }
53
54
55
56 template<typename T1, typename T2>
57 inline
58 void
apply_noalias(Mat<typename T1::elem_type> & out,const Proxy<T1> & P1,const Proxy<T2> & P2)59 glue_atan2::apply_noalias(Mat<typename T1::elem_type>& out, const Proxy<T1>& P1, const Proxy<T2>& P2)
60 {
61 arma_extra_debug_sigprint();
62
63 typedef typename T1::elem_type eT;
64
65 const uword n_rows = P1.get_n_rows();
66 const uword n_cols = P1.get_n_cols();
67 const uword n_elem = P1.get_n_elem();
68
69 out.set_size(n_rows, n_cols);
70
71 eT* out_mem = out.memptr();
72
73 const bool use_mp = arma_config::openmp && mp_gate<eT, (Proxy<T1>::use_mp || Proxy<T2>::use_mp)>::eval(n_elem);
74 const bool use_at = Proxy<T1>::use_at || Proxy<T2>::use_at;
75
76 if(use_at == false)
77 {
78 typename Proxy<T1>::ea_type eaP1 = P1.get_ea();
79 typename Proxy<T2>::ea_type eaP2 = P2.get_ea();
80
81 if(use_mp)
82 {
83 #if defined(ARMA_USE_OPENMP)
84 {
85 const int n_threads = mp_thread_limit::get();
86 #pragma omp parallel for schedule(static) num_threads(n_threads)
87 for(uword i=0; i<n_elem; ++i)
88 {
89 out_mem[i] = std::atan2( eaP1[i], eaP2[i] );
90 }
91 }
92 #endif
93 }
94 else
95 {
96 for(uword i=0; i<n_elem; ++i)
97 {
98 out_mem[i] = std::atan2( eaP1[i], eaP2[i] );
99 }
100 }
101 }
102 else
103 {
104 if(use_mp)
105 {
106 const unwrap<typename Proxy<T1>::stored_type> U1(P1.Q);
107 const unwrap<typename Proxy<T2>::stored_type> U2(P2.Q);
108
109 out = arma::atan2(U1.M, U2.M);
110 }
111 else
112 {
113 for(uword col=0; col < n_cols; ++col)
114 for(uword row=0; row < n_rows; ++row)
115 {
116 *out_mem = std::atan2( P1.at(row,col), P2.at(row,col) );
117 out_mem++;
118 }
119 }
120 }
121 }
122
123
124
125 template<typename T1, typename T2>
126 inline
127 void
apply(Cube<typename T1::elem_type> & out,const GlueCube<T1,T2,glue_atan2> & expr)128 glue_atan2::apply(Cube<typename T1::elem_type>& out, const GlueCube<T1, T2, glue_atan2>& expr)
129 {
130 arma_extra_debug_sigprint();
131
132 typedef typename T1::elem_type eT;
133
134 const ProxyCube<T1> P1(expr.A);
135 const ProxyCube<T2> P2(expr.B);
136
137 arma_assert_same_size(P1, P2, "atan2()");
138
139 const bool bad_alias = ( (ProxyCube<T1>::has_subview && P1.is_alias(out)) || (ProxyCube<T2>::has_subview && P2.is_alias(out)) );
140
141 if(bad_alias == false)
142 {
143 glue_atan2::apply_noalias(out, P1, P2);
144 }
145 else
146 {
147 Cube<eT> tmp;
148
149 glue_atan2::apply_noalias(tmp, P1, P2);
150
151 out.steal_mem(tmp);
152 }
153 }
154
155
156
157 template<typename T1, typename T2>
158 inline
159 void
apply_noalias(Cube<typename T1::elem_type> & out,const ProxyCube<T1> & P1,const ProxyCube<T2> & P2)160 glue_atan2::apply_noalias(Cube<typename T1::elem_type>& out, const ProxyCube<T1>& P1, const ProxyCube<T2>& P2)
161 {
162 arma_extra_debug_sigprint();
163
164 typedef typename T1::elem_type eT;
165
166 const uword n_rows = P1.get_n_rows();
167 const uword n_cols = P1.get_n_cols();
168 const uword n_slices = P1.get_n_slices();
169 const uword n_elem = P1.get_n_elem();
170
171 out.set_size(n_rows, n_cols, n_slices);
172
173 eT* out_mem = out.memptr();
174
175 const bool use_mp = arma_config::openmp && mp_gate<eT, (ProxyCube<T1>::use_mp || ProxyCube<T2>::use_mp)>::eval(n_elem);
176 const bool use_at = ProxyCube<T1>::use_at || ProxyCube<T2>::use_at;
177
178 if(use_at == false)
179 {
180 typename ProxyCube<T1>::ea_type eaP1 = P1.get_ea();
181 typename ProxyCube<T2>::ea_type eaP2 = P2.get_ea();
182
183 if(use_mp)
184 {
185 #if defined(ARMA_USE_OPENMP)
186 {
187 const int n_threads = mp_thread_limit::get();
188 #pragma omp parallel for schedule(static) num_threads(n_threads)
189 for(uword i=0; i<n_elem; ++i)
190 {
191 out_mem[i] = std::atan2( eaP1[i], eaP2[i] );
192 }
193 }
194 #endif
195 }
196 else
197 {
198 for(uword i=0; i<n_elem; ++i)
199 {
200 out_mem[i] = std::atan2( eaP1[i], eaP2[i] );
201 }
202 }
203 }
204 else
205 {
206 if(use_mp)
207 {
208 const unwrap_cube<typename ProxyCube<T1>::stored_type> U1(P1.Q);
209 const unwrap_cube<typename ProxyCube<T2>::stored_type> U2(P2.Q);
210
211 out = arma::atan2(U1.M, U2.M);
212 }
213 else
214 {
215 for(uword slice=0; slice < n_slices; ++slice)
216 for(uword col=0; col < n_cols; ++col )
217 for(uword row=0; row < n_rows; ++row )
218 {
219 *out_mem = std::atan2( P1.at(row,col,slice), P2.at(row,col,slice) );
220 out_mem++;
221 }
222 }
223 }
224 }
225
226
227
228 //! @}
229