gdbsupport/parallel-for.h

   1 /* Parallel for loops
   2
   3    Copyright (C) 2019-2022 Free Software Foundation, Inc.
   4
   5    This file is part of GDB.
   6
   7    This program is free software; you can redistribute it and/or modify
   8    it under the terms of the GNU General Public License as published by
   9    the Free Software Foundation; either version 3 of the License, or
  10    (at your option) any later version.
  11
  12    This program is distributed in the hope that it will be useful,
  13    but WITHOUT ANY WARRANTY; without even the implied warranty of
  14    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15    GNU General Public License for more details.
  16
  17    You should have received a copy of the GNU General Public License
  18    along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
  19
  20 #ifndef GDBSUPPORT_PARALLEL_FOR_H
  21 #define GDBSUPPORT_PARALLEL_FOR_H
  22
  23 #include <algorithm>
  24 #include <type_traits>
  25 #include "gdbsupport/thread-pool.h"
  26
  27 namespace gdb
  28 {
  29
  30 namespace detail
  31 {
  32
  33 /* This is a helper class that is used to accumulate results for
  34    parallel_for.  There is a specialization for 'void', below.  */
  35 template<typename T>
  36 struct par_for_accumulator
  37 {
  38 public:
  39
  40   explicit par_for_accumulator (size_t n_threads)
  41     : m_futures (n_threads)
  42   {
  43   }
  44
  45   /* The result type that is accumulated.  */
  46   typedef std::vector<T> result_type;
  47
  48   /* Post the Ith task to a background thread, and store a future for
  49      later.  */
  50   void post (size_t i, std::function<T ()> task)
  51   {
  52     m_futures[i]
  53       = gdb::thread_pool::g_thread_pool->post_task (std::move (task));
  54   }
  55
  56   /* Invoke TASK in the current thread, then compute all the results
  57      from all background tasks and put them into a result vector,
  58      which is returned.  */
  59   result_type finish (gdb::function_view<T ()> task)
  60   {
  61     result_type result (m_futures.size () + 1);
  62
  63     result.back () = task ();
  64
  65     for (size_t i = 0; i < m_futures.size (); ++i)
  66       result[i] = m_futures[i].get ();
  67
  68     return result;
  69   }
  70
  71 private:
  72
  73   /* A vector of futures coming from the tasks run in the
  74      background.  */
  75   std::vector<std::future<T>> m_futures;
  76 };
  77
  78 /* See the generic template.  */
  79 template<>
  80 struct par_for_accumulator<void>
  81 {
  82 public:
  83
  84   explicit par_for_accumulator (size_t n_threads)
  85     : m_futures (n_threads)
  86   {
  87   }
  88
  89   /* This specialization does not compute results.  */
  90   typedef void result_type;
  91
  92   void post (size_t i, std::function<void ()> task)
  93   {
  94     m_futures[i]
  95       = gdb::thread_pool::g_thread_pool->post_task (std::move (task));
  96   }
  97
  98   result_type finish (gdb::function_view<void ()> task)
  99   {
 100     task ();
 101
 102     for (auto &future : m_futures)
 103       {
 104         /* Use 'get' and not 'wait', to propagate any exception.  */
 105         future.get ();
 106       }
 107   }
 108
 109 private:
 110
 111   std::vector<std::future<void>> m_futures;
 112 };
 113
 114 }
 115
 116 /* A very simple "parallel for".  This splits the range of iterators
 117    into subranges, and then passes each subrange to the callback.  The
 118    work may or may not be done in separate threads.
 119
 120    This approach was chosen over having the callback work on single
 121    items because it makes it simple for the caller to do
 122    once-per-subrange initialization and destruction.
 123
 124    The parameter N says how batching ought to be done -- there will be
 125    at least N elements processed per thread.  Setting N to 0 is not
 126    allowed.
 127
 128    If the function returns a non-void type, then a vector of the
 129    results is returned.  The size of the resulting vector depends on
 130    the number of threads that were used.  */
 131
 132 template<class RandomIt, class RangeFunction>
 133 typename gdb::detail::par_for_accumulator<
 134     std::result_of_t<RangeFunction (RandomIt, RandomIt)>
 135   >::result_type
 136 parallel_for_each (unsigned n, RandomIt first, RandomIt last,
 137                    RangeFunction callback)
 138 {
 139   typedef typename std::result_of_t<RangeFunction (RandomIt, RandomIt)>
 140     result_type;
 141
 142   size_t n_threads = thread_pool::g_thread_pool->thread_count ();
 143   size_t n_elements = last - first;
 144   size_t elts_per_thread = 0;
 145   if (n_threads > 1)
 146     {
 147       /* Require that there should be at least N elements in a
 148          thread.  */
 149       gdb_assert (n > 0);
 150       if (n_elements / n_threads < n)
 151         n_threads = std::max (n_elements / n, (size_t) 1);
 152       elts_per_thread = n_elements / n_threads;
 153     }
 154
 155   size_t count = n_threads == 0 ? 0 : n_threads - 1;
 156   gdb::detail::par_for_accumulator<result_type> results (count);
 157
 158   for (int i = 0; i < count; ++i)
 159     {
 160       RandomIt end = first + elts_per_thread;
 161       results.post (i, [=] ()
 162         {
 163           return callback (first, end);
 164         });
 165       first = end;
 166     }
 167
 168   /* Process all the remaining elements in the main thread.  */
 169   return results.finish ([=] ()
 170     {
 171       return callback (first, last);
 172     });
 173 }
 174
 175 }
 176
 177 #endif /* GDBSUPPORT_PARALLEL_FOR_H */