File: get_started.cpp

package info (click to toggle)
cppad 2025.00.00.2-1
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid, trixie
  • size: 11,552 kB
  • sloc: cpp: 112,594; sh: 5,972; ansic: 179; python: 71; sed: 12; makefile: 10
file content (286 lines) | stat: -rw-r--r-- 8,357 bytes parent folder | download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
// SPDX-License-Identifier: EPL-2.0 OR GPL-2.0-or-later
// SPDX-FileCopyrightText: Bradley M. Bell <bradbell@seanet.com>
// SPDX-FileContributor: 2003-24 Bradley M. Bell
// ----------------------------------------------------------------------------
/*
{xrst_begin pthread_get_started.cpp}
{xrst_spell
  posix
}

{xrst_template ,
   example/multi_thread/template/get_started.xrst

   title: Getting Started Using @Name@ Threads With CppAD
   start source code after: // <space> BEGIN_C++
   end source code before:  // <space> END_C++

   @Name@    , Posix
   @####@    , #####
   @DEFAULT@ , USE_DEFAULT_ADFUN_CONSTRUCTOR
}

{xrst_end pthread_get_started.cpp}
------------------------------------------------------------------------------
*/
// BEGIN_C++
# include <cppad/cppad.hpp>
# include <pthread.h>

# define USE_DEFAULT_ADFUN_CONSTRUCTOR 1

namespace {
   //
   // d_vector, ad_vector, fun_vector
   typedef CPPAD_TESTVECTOR(double)                 d_vector;
   typedef CPPAD_TESTVECTOR( CppAD::AD<double> )    ad_vector;
   typedef CPPAD_TESTVECTOR( CppAD::ADFun<double> ) fun_vector;
   typedef CPPAD_TESTVECTOR( pthread_t )            pthread_vector;
   //
   // thread_info, thread_info_vector
   typedef struct {
      size_t                thread_num;
      CppAD::ADFun<double>* f_ptr;
      size_t                j_begin;
      size_t                j_end;
      const d_vector*       x_ptr;
      d_vector*             Jac_ptr;
      bool                  ok;
   } thread_info;
   typedef CPPAD_TESTVECTOR( thread_info ) thread_info_vector;
   //
   // thread_specific_key_
   pthread_key_t thread_specific_key_;
   //
   // thread_specific_destructor
   void thread_specific_destructor(void* thread_num_vptr)
   {  return; }
   //
   // sequential_execution_
   bool sequential_execution_ = true;
   //
   // in_parallel
   bool in_parallel(void)
   {  return ! sequential_execution_; }
   //
   // thread_number
   size_t thread_number(void)
   {  // get thread specific information
      void*   thread_num_vptr = pthread_getspecific(thread_specific_key_);
      size_t* thread_num_ptr  = static_cast<size_t*>(thread_num_vptr);
      size_t  thread_num      = *thread_num_ptr;
      return thread_num;
   }
   //
   // partial
   double partial(
      CppAD::ADFun<double>& f, size_t j, const d_vector& x
   )
   {  size_t nx = x.size();
      d_vector dx(nx), dy(1);
      for(size_t k = 0; k < nx; ++k)
         dx[k] = 0.0;
      dx[j] = 1.0;
      f.Forward(0, x);
      dy = f.Forward(1, dx);
      return dy[0];
   }
   //
   // run_one_thread
   void* run_one_thread(void* thread_info_vptr)
   {  //
      // thread_num, f_ptr, j_begin, j_end, x_ptr, Jac_ptr
      thread_info* thread_info_ptr =
         static_cast<thread_info*>(thread_info_vptr);
      size_t                thread_num = thread_info_ptr->thread_num;
      CppAD::ADFun<double>* f_ptr      = thread_info_ptr->f_ptr;
      size_t                j_begin    = thread_info_ptr->j_begin;
      size_t                j_end      = thread_info_ptr->j_end;
      const d_vector*       x_ptr      = thread_info_ptr->x_ptr;
      d_vector*             Jac_ptr    = thread_info_ptr->Jac_ptr;
      //
      // f, Jac, ok
      CppAD::ADFun<double>& f   = *f_ptr;
      d_vector&             Jac = *Jac_ptr;
      bool&                 ok  = thread_info_ptr->ok;
      //
      // rc
      int rc;
      //
      // phtread_setspecific, ok
      // This sets up the thread_number function for this thread.
      if( thread_num != 0 )
      {  rc  = pthread_setspecific(thread_specific_key_, &thread_num);
         ok &= rc == 0;
         ok &= thread_number() == thread_num;
      }
      //
      // f
      // This will cause an assert if Taylor coefficients were allocated
      // by a different thread.
      f.capacity_order(0);
      //
      // Jac
      for(size_t j = j_begin; j < j_end; ++j)
         Jac[j] = partial(f, j, *x_ptr);
      //
      return nullptr;
   }
}
bool get_started(void)
{  // ok
   bool ok = true;
   //
   // eps99
   double eps99 = 99.0 * std::numeric_limits<double>::epsilon();
   //
   // nx, ax
   size_t nx = 10;
   ad_vector ax(nx);
   for(size_t j = 0; j < nx; ++j)
      ax[j] = 1.0;
   CppAD::Independent(ax);
   //
   // fun
   ad_vector  ay(1);
   ay[0] = ax[0];
   for(size_t j = 1; j < nx; ++j)
      ay[0] *= ax[j];
# if USE_DEFAULT_ADFUN_CONSTRUCTOR
   CppAD::ADFun<double> fun;
   fun.Dependent(ax, ay);
# else
   // This allocates memory for first order Taylor coefficients using thread 0.
   // An assert will occur at f.capacity_order(0) in run_one_thread when
   // it is called by a different thread.
   CppAD::ADFun<double> fun(ax, ay);
# endif
   //
   // num_threads, f_thread
   size_t num_threads = 4;
   fun_vector f_thread(num_threads);
   for(size_t i = 0; i < num_threads; ++i)
      f_thread[i] = fun;
   //
   // x
   d_vector x(nx);
   for(size_t j = 0; j < nx; ++j)
      x[j] = 1.0 + 1.0 / double(j+1);
   //
   // Jac
   d_vector Jac(nx);
   //
   // n_per_thread, n_extra
   size_t n_per_thread = nx / num_threads;
   size_t n_extra      = nx % num_threads;
   //
   // thread_info_vec
   thread_info_vector thread_info_vec(num_threads);
   size_t j_begin = 0;
   size_t j_end;
   for(size_t thread_num = 0; thread_num < num_threads; ++thread_num)
   {  j_end = j_begin + n_per_thread;
      if( thread_num < n_extra )
         ++j_end;
      //
      thread_info_vec[thread_num].thread_num = thread_num;
      thread_info_vec[thread_num].f_ptr      = &f_thread[thread_num];
      thread_info_vec[thread_num].j_begin    = j_begin;
      thread_info_vec[thread_num].j_end      = j_end;
      thread_info_vec[thread_num].x_ptr      = &x;
      thread_info_vec[thread_num].Jac_ptr    = &Jac;
      thread_info_vec[thread_num].ok         = true;
      //
      j_begin = j_end;
   }
   ok &= j_end == nx;
   //
   // rc
   int rc;
   //
   // thread_specific_key_, ok
   rc  = pthread_key_create(&thread_specific_key_, thread_specific_destructor);
   ok &= rc == 0;

   //
   // thread_setspecific, ok
   // must be set for this thread before calling parall_setup or parallel_ad
   rc = pthread_setspecific(
         thread_specific_key_, &thread_info_vec[0].thread_num
   );
   ok &= rc == 0;
   ok &= thread_number() == 0;
   //
   // parallel_setup
   CppAD::thread_alloc::parallel_setup(
      num_threads, in_parallel, thread_number
   );
   //
   // parallel_ad
   CppAD::parallel_ad<double>();
   //
   // hold_memory
   // optional and may improve speed if you do a lot of memory allocation
   CppAD::thread_alloc::hold_memory(true);
   //
   // ptread_vec
   pthread_vector pthread_vec(num_threads - 1);
   //
   // sequential_execution_, ok
   sequential_execution_ = false;
   ok &= in_parallel();
   //
   // Jac, pthread_vec, ok
   // Launch num_threads - 1 posix threads
   for(size_t thread_num = 1; thread_num < num_threads; ++thread_num)
   {  pthread_attr_t* no_attr = nullptr;
      rc = pthread_create(
         &pthread_vec[thread_num-1],
         no_attr,
         run_one_thread,
         &thread_info_vec[thread_num]
      );
      ok &= rc == 0;
   }
   {  // run master thread's indices
      size_t thread_num = 0;
      run_one_thread(&thread_info_vec[thread_num]);
   }
   // wait for other threads to finish
   for(size_t thread_num = 1; thread_num < num_threads; ++thread_num)
   {  void* no_status = nullptr;
      rc  = pthread_join(pthread_vec[thread_num-1], &no_status);
      ok &= rc == 0;
   }
   //
   // sequential_execution_, ok
   sequential_execution_ = true;
   CppAD::thread_alloc::parallel_setup(1, nullptr, nullptr);
   ok &= ! in_parallel();
   //
   // hold_memory, ok
   // free memory for other threads before this (the master thread)
   ok &= thread_number() == 0;
   CppAD::thread_alloc::hold_memory(false);
   for(size_t thread_num = 1; thread_num < num_threads; ++thread_num)
   {  ok &= thread_info_vec[thread_num].ok;
      CppAD::thread_alloc::free_available(thread_num);
   }
   ok &= thread_info_vec[0].ok;
   CppAD::thread_alloc::free_available(0);
   //
   // ok
   for(size_t j = 0; j < nx; ++j)
   {  //
      // check
      double check = 1.0;
      for(size_t k = 0; k < nx; ++k)
         if(k != j)
            check *= x[k];
      //
      // ok
      ok &= CppAD::NearEqual(Jac[j], check, eps99, eps99);
   }
   return ok;
}
// END_C++