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
|
// -*- C++ -*-
// Copyright (C) 2007, 2008, 2009 Free Software Foundation, Inc.
//
// This file is part of the GNU ISO C++ Library. This library is free
// software; you can redistribute it and/or modify it under the terms
// of the GNU General Public License as published by the Free Software
// Foundation; either version 3, or (at your option) any later
// version.
// This library is distributed in the hope that it will be useful, but
// WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// General Public License for more details.
// Under Section 7 of GPL version 3, you are granted additional
// permissions described in the GCC Runtime Library Exception, version
// 3.1, as published by the Free Software Foundation.
// You should have received a copy of the GNU General Public License and
// a copy of the GCC Runtime Library Exception along with this program;
// see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
// <http://www.gnu.org/licenses/>.
/** @file parallel/partial_sum.h
* @brief Parallel implementation of std::partial_sum(), i.e. prefix
* sums.
* This file is a GNU parallel extension to the Standard C++ Library.
*/
// Written by Johannes Singler.
#ifndef _GLIBCXX_PARALLEL_PARTIAL_SUM_H
#define _GLIBCXX_PARALLEL_PARTIAL_SUM_H 1
#include <omp.h>
#include <new>
#include <bits/stl_algobase.h>
#include <parallel/parallel.h>
#include <parallel/numericfwd.h>
namespace __gnu_parallel
{
// Problem: there is no 0-element given.
/** @brief Base case prefix sum routine.
* @param __begin Begin iterator of input sequence.
* @param __end End iterator of input sequence.
* @param __result Begin iterator of output sequence.
* @param __bin_op Associative binary function.
* @param __value Start value. Must be passed since the neutral
* element is unknown in general.
* @return End iterator of output sequence. */
template<typename _IIter,
typename _OutputIterator,
typename _BinaryOperation>
_OutputIterator
__parallel_partial_sum_basecase(
_IIter __begin, _IIter __end, _OutputIterator __result,
_BinaryOperation __bin_op,
typename std::iterator_traits <_IIter>::value_type __value)
{
if (__begin == __end)
return __result;
while (__begin != __end)
{
__value = __bin_op(__value, *__begin);
*__result = __value;
++__result;
++__begin;
}
return __result;
}
/** @brief Parallel partial sum implementation, two-phase approach,
no recursion.
* @param __begin Begin iterator of input sequence.
* @param __end End iterator of input sequence.
* @param __result Begin iterator of output sequence.
* @param __bin_op Associative binary function.
* @param __n Length of sequence.
* @param __num_threads Number of threads to use.
* @return End iterator of output sequence.
*/
template<typename _IIter,
typename _OutputIterator,
typename _BinaryOperation>
_OutputIterator
__parallel_partial_sum_linear(
_IIter __begin, _IIter __end, _OutputIterator __result,
_BinaryOperation __bin_op,
typename std::iterator_traits<_IIter>::difference_type __n)
{
typedef std::iterator_traits<_IIter> _TraitsType;
typedef typename _TraitsType::value_type _ValueType;
typedef typename _TraitsType::difference_type _DifferenceType;
if (__begin == __end)
return __result;
_ThreadIndex __num_threads =
std::min<_DifferenceType>(__get_max_threads(), __n - 1);
if (__num_threads < 2)
{
*__result = *__begin;
return __parallel_partial_sum_basecase(
__begin + 1, __end, __result + 1, __bin_op, *__begin);
}
_DifferenceType* __borders;
_ValueType* __sums;
const _Settings& __s = _Settings::get();
# pragma omp parallel num_threads(__num_threads)
{
# pragma omp single
{
__num_threads = omp_get_num_threads();
__borders = new _DifferenceType[__num_threads + 2];
if (__s.partial_sum_dilation == 1.0f)
equally_split(__n, __num_threads + 1, __borders);
else
{
_DifferenceType __chunk_length =
((double)__n
/ ((double)__num_threads + __s.partial_sum_dilation)),
__borderstart = __n - __num_threads * __chunk_length;
__borders[0] = 0;
for (int __i = 1; __i < (__num_threads + 1); ++__i)
{
__borders[__i] = __borderstart;
__borderstart += __chunk_length;
}
__borders[__num_threads + 1] = __n;
}
__sums = static_cast<_ValueType*>(::operator new(sizeof(_ValueType)
* __num_threads));
_OutputIterator __target_end;
} //single
_ThreadIndex __iam = omp_get_thread_num();
if (__iam == 0)
{
*__result = *__begin;
__parallel_partial_sum_basecase(
__begin + 1, __begin + __borders[1], __result + 1,
__bin_op, *__begin);
::new(&(__sums[__iam])) _ValueType(*(__result + __borders[1] - 1));
}
else
{
::new(&(__sums[__iam]))
_ValueType(std::accumulate(__begin + __borders[__iam] + 1,
__begin + __borders[__iam + 1],
*(__begin + __borders[__iam]),
__bin_op,
__gnu_parallel::sequential_tag()));
}
# pragma omp barrier
# pragma omp single
__parallel_partial_sum_basecase(__sums + 1, __sums + __num_threads,
__sums + 1, __bin_op, __sums[0]);
# pragma omp barrier
// Still same team.
__parallel_partial_sum_basecase(
__begin + __borders[__iam + 1],
__begin + __borders[__iam + 2],
__result + __borders[__iam + 1],
__bin_op, __sums[__iam]);
} //parallel
::operator delete(__sums);
delete[] __borders;
return __result + __n;
}
/** @brief Parallel partial sum front-__end.
* @param __begin Begin iterator of input sequence.
* @param __end End iterator of input sequence.
* @param __result Begin iterator of output sequence.
* @param __bin_op Associative binary function.
* @return End iterator of output sequence. */
template<typename _IIter,
typename _OutputIterator,
typename _BinaryOperation>
_OutputIterator
__parallel_partial_sum(_IIter __begin, _IIter __end,
_OutputIterator __result, _BinaryOperation __bin_op)
{
_GLIBCXX_CALL(__begin - __end)
typedef std::iterator_traits<_IIter> _TraitsType;
typedef typename _TraitsType::value_type _ValueType;
typedef typename _TraitsType::difference_type _DifferenceType;
_DifferenceType __n = __end - __begin;
switch (_Settings::get().partial_sum_algorithm)
{
case LINEAR:
// Need an initial offset.
return __parallel_partial_sum_linear(
__begin, __end, __result, __bin_op, __n);
default:
// Partial_sum algorithm not implemented.
_GLIBCXX_PARALLEL_ASSERT(0);
return __result + __n;
}
}
}
#endif /* _GLIBCXX_PARALLEL_PARTIAL_SUM_H */
|
