// -*- C++ -*-
|
// -*- C++ -*-
|
|
|
// Copyright (C) 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
|
// Copyright (C) 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
|
//
|
//
|
// This file is part of the GNU ISO C++ Library. This library is free
|
// 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
|
// software; you can redistribute it and/or modify it under the terms
|
// of the GNU General Public License as published by the Free Software
|
// of the GNU General Public License as published by the Free Software
|
// Foundation; either version 3, or (at your option) any later
|
// Foundation; either version 3, or (at your option) any later
|
// version.
|
// version.
|
|
|
// This library is distributed in the hope that it will be useful, but
|
// This library is distributed in the hope that it will be useful, but
|
// WITHOUT ANY WARRANTY; without even the implied warranty of
|
// WITHOUT ANY WARRANTY; without even the implied warranty of
|
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
// General Public License for more details.
|
// General Public License for more details.
|
|
|
// Under Section 7 of GPL version 3, you are granted additional
|
// Under Section 7 of GPL version 3, you are granted additional
|
// permissions described in the GCC Runtime Library Exception, version
|
// permissions described in the GCC Runtime Library Exception, version
|
// 3.1, as published by the Free Software Foundation.
|
// 3.1, as published by the Free Software Foundation.
|
|
|
// You should have received a copy of the GNU General Public License and
|
// 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;
|
// a copy of the GCC Runtime Library Exception along with this program;
|
// see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
// see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
// <http://www.gnu.org/licenses/>.
|
// <http://www.gnu.org/licenses/>.
|
|
|
/** @file parallel/balanced_quicksort.h
|
/** @file parallel/balanced_quicksort.h
|
* @brief Implementation of a dynamically load-balanced parallel quicksort.
|
* @brief Implementation of a dynamically load-balanced parallel quicksort.
|
*
|
*
|
* It works in-place and needs only logarithmic extra memory.
|
* It works in-place and needs only logarithmic extra memory.
|
* The algorithm is similar to the one proposed in
|
* The algorithm is similar to the one proposed in
|
*
|
*
|
* P. Tsigas and Y. Zhang.
|
* P. Tsigas and Y. Zhang.
|
* A simple, fast parallel implementation of quicksort and
|
* A simple, fast parallel implementation of quicksort and
|
* its performance evaluation on SUN enterprise 10000.
|
* its performance evaluation on SUN enterprise 10000.
|
* In 11th Euromicro Conference on Parallel, Distributed and
|
* In 11th Euromicro Conference on Parallel, Distributed and
|
* Network-Based Processing, page 372, 2003.
|
* Network-Based Processing, page 372, 2003.
|
*
|
*
|
* This file is a GNU parallel extension to the Standard C++ Library.
|
* This file is a GNU parallel extension to the Standard C++ Library.
|
*/
|
*/
|
|
|
// Written by Johannes Singler.
|
// Written by Johannes Singler.
|
|
|
#ifndef _GLIBCXX_PARALLEL_BALANCED_QUICKSORT_H
|
#ifndef _GLIBCXX_PARALLEL_BALANCED_QUICKSORT_H
|
#define _GLIBCXX_PARALLEL_BALANCED_QUICKSORT_H 1
|
#define _GLIBCXX_PARALLEL_BALANCED_QUICKSORT_H 1
|
|
|
#include <parallel/basic_iterator.h>
|
#include <parallel/basic_iterator.h>
|
#include <bits/stl_algo.h>
|
#include <bits/stl_algo.h>
|
#include <bits/stl_function.h>
|
#include <bits/stl_function.h>
|
|
|
#include <parallel/settings.h>
|
#include <parallel/settings.h>
|
#include <parallel/partition.h>
|
#include <parallel/partition.h>
|
#include <parallel/random_number.h>
|
#include <parallel/random_number.h>
|
#include <parallel/queue.h>
|
#include <parallel/queue.h>
|
|
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
#include <parallel/checkers.h>
|
#include <parallel/checkers.h>
|
#endif
|
#endif
|
|
|
namespace __gnu_parallel
|
namespace __gnu_parallel
|
{
|
{
|
/** @brief Information local to one thread in the parallel quicksort run. */
|
/** @brief Information local to one thread in the parallel quicksort run. */
|
template<typename _RAIter>
|
template<typename _RAIter>
|
struct _QSBThreadLocal
|
struct _QSBThreadLocal
|
{
|
{
|
typedef std::iterator_traits<_RAIter> _TraitsType;
|
typedef std::iterator_traits<_RAIter> _TraitsType;
|
typedef typename _TraitsType::difference_type _DifferenceType;
|
typedef typename _TraitsType::difference_type _DifferenceType;
|
|
|
/** @brief Continuous part of the sequence, described by an
|
/** @brief Continuous part of the sequence, described by an
|
iterator pair. */
|
iterator pair. */
|
typedef std::pair<_RAIter, _RAIter> _Piece;
|
typedef std::pair<_RAIter, _RAIter> _Piece;
|
|
|
/** @brief Initial piece to work on. */
|
/** @brief Initial piece to work on. */
|
_Piece _M_initial;
|
_Piece _M_initial;
|
|
|
/** @brief Work-stealing queue. */
|
/** @brief Work-stealing queue. */
|
_RestrictedBoundedConcurrentQueue<_Piece> _M_leftover_parts;
|
_RestrictedBoundedConcurrentQueue<_Piece> _M_leftover_parts;
|
|
|
/** @brief Number of threads involved in this algorithm. */
|
/** @brief Number of threads involved in this algorithm. */
|
_ThreadIndex _M_num_threads;
|
_ThreadIndex _M_num_threads;
|
|
|
/** @brief Pointer to a counter of elements left over to sort. */
|
/** @brief Pointer to a counter of elements left over to sort. */
|
volatile _DifferenceType* _M_elements_leftover;
|
volatile _DifferenceType* _M_elements_leftover;
|
|
|
/** @brief The complete sequence to sort. */
|
/** @brief The complete sequence to sort. */
|
_Piece _M_global;
|
_Piece _M_global;
|
|
|
/** @brief Constructor.
|
/** @brief Constructor.
|
* @param __queue_size size of the work-stealing queue. */
|
* @param __queue_size size of the work-stealing queue. */
|
_QSBThreadLocal(int __queue_size) : _M_leftover_parts(__queue_size) { }
|
_QSBThreadLocal(int __queue_size) : _M_leftover_parts(__queue_size) { }
|
};
|
};
|
|
|
/** @brief Balanced quicksort divide step.
|
/** @brief Balanced quicksort divide step.
|
* @param __begin Begin iterator of subsequence.
|
* @param __begin Begin iterator of subsequence.
|
* @param __end End iterator of subsequence.
|
* @param __end End iterator of subsequence.
|
* @param __comp Comparator.
|
* @param __comp Comparator.
|
* @param __num_threads Number of threads that are allowed to work on
|
* @param __num_threads Number of threads that are allowed to work on
|
* this part.
|
* this part.
|
* @pre @c (__end-__begin)>=1 */
|
* @pre @c (__end-__begin)>=1 */
|
template<typename _RAIter, typename _Compare>
|
template<typename _RAIter, typename _Compare>
|
typename std::iterator_traits<_RAIter>::difference_type
|
typename std::iterator_traits<_RAIter>::difference_type
|
__qsb_divide(_RAIter __begin, _RAIter __end,
|
__qsb_divide(_RAIter __begin, _RAIter __end,
|
_Compare __comp, _ThreadIndex __num_threads)
|
_Compare __comp, _ThreadIndex __num_threads)
|
{
|
{
|
_GLIBCXX_PARALLEL_ASSERT(__num_threads > 0);
|
_GLIBCXX_PARALLEL_ASSERT(__num_threads > 0);
|
|
|
typedef std::iterator_traits<_RAIter> _TraitsType;
|
typedef std::iterator_traits<_RAIter> _TraitsType;
|
typedef typename _TraitsType::value_type _ValueType;
|
typedef typename _TraitsType::value_type _ValueType;
|
typedef typename _TraitsType::difference_type _DifferenceType;
|
typedef typename _TraitsType::difference_type _DifferenceType;
|
|
|
_RAIter __pivot_pos =
|
_RAIter __pivot_pos =
|
__median_of_three_iterators(__begin, __begin + (__end - __begin) / 2,
|
__median_of_three_iterators(__begin, __begin + (__end - __begin) / 2,
|
__end - 1, __comp);
|
__end - 1, __comp);
|
|
|
#if defined(_GLIBCXX_ASSERTIONS)
|
#if defined(_GLIBCXX_ASSERTIONS)
|
// Must be in between somewhere.
|
// Must be in between somewhere.
|
_DifferenceType __n = __end - __begin;
|
_DifferenceType __n = __end - __begin;
|
|
|
_GLIBCXX_PARALLEL_ASSERT((!__comp(*__pivot_pos, *__begin)
|
_GLIBCXX_PARALLEL_ASSERT((!__comp(*__pivot_pos, *__begin)
|
&& !__comp(*(__begin + __n / 2),
|
&& !__comp(*(__begin + __n / 2),
|
*__pivot_pos))
|
*__pivot_pos))
|
|| (!__comp(*__pivot_pos, *__begin)
|
|| (!__comp(*__pivot_pos, *__begin)
|
&& !__comp(*(__end - 1), *__pivot_pos))
|
&& !__comp(*(__end - 1), *__pivot_pos))
|
|| (!__comp(*__pivot_pos, *(__begin + __n / 2))
|
|| (!__comp(*__pivot_pos, *(__begin + __n / 2))
|
&& !__comp(*__begin, *__pivot_pos))
|
&& !__comp(*__begin, *__pivot_pos))
|
|| (!__comp(*__pivot_pos, *(__begin + __n / 2))
|
|| (!__comp(*__pivot_pos, *(__begin + __n / 2))
|
&& !__comp(*(__end - 1), *__pivot_pos))
|
&& !__comp(*(__end - 1), *__pivot_pos))
|
|| (!__comp(*__pivot_pos, *(__end - 1))
|
|| (!__comp(*__pivot_pos, *(__end - 1))
|
&& !__comp(*__begin, *__pivot_pos))
|
&& !__comp(*__begin, *__pivot_pos))
|
|| (!__comp(*__pivot_pos, *(__end - 1))
|
|| (!__comp(*__pivot_pos, *(__end - 1))
|
&& !__comp(*(__begin + __n / 2),
|
&& !__comp(*(__begin + __n / 2),
|
*__pivot_pos)));
|
*__pivot_pos)));
|
#endif
|
#endif
|
|
|
// Swap pivot value to end.
|
// Swap pivot value to end.
|
if (__pivot_pos != (__end - 1))
|
if (__pivot_pos != (__end - 1))
|
std::swap(*__pivot_pos, *(__end - 1));
|
std::swap(*__pivot_pos, *(__end - 1));
|
__pivot_pos = __end - 1;
|
__pivot_pos = __end - 1;
|
|
|
__gnu_parallel::__binder2nd<_Compare, _ValueType, _ValueType, bool>
|
__gnu_parallel::__binder2nd<_Compare, _ValueType, _ValueType, bool>
|
__pred(__comp, *__pivot_pos);
|
__pred(__comp, *__pivot_pos);
|
|
|
// Divide, returning __end - __begin - 1 in the worst case.
|
// Divide, returning __end - __begin - 1 in the worst case.
|
_DifferenceType __split_pos = __parallel_partition(__begin, __end - 1,
|
_DifferenceType __split_pos = __parallel_partition(__begin, __end - 1,
|
__pred,
|
__pred,
|
__num_threads);
|
__num_threads);
|
|
|
// Swap back pivot to middle.
|
// Swap back pivot to middle.
|
std::swap(*(__begin + __split_pos), *__pivot_pos);
|
std::swap(*(__begin + __split_pos), *__pivot_pos);
|
__pivot_pos = __begin + __split_pos;
|
__pivot_pos = __begin + __split_pos;
|
|
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
_RAIter __r;
|
_RAIter __r;
|
for (__r = __begin; __r != __pivot_pos; ++__r)
|
for (__r = __begin; __r != __pivot_pos; ++__r)
|
_GLIBCXX_PARALLEL_ASSERT(__comp(*__r, *__pivot_pos));
|
_GLIBCXX_PARALLEL_ASSERT(__comp(*__r, *__pivot_pos));
|
for (; __r != __end; ++__r)
|
for (; __r != __end; ++__r)
|
_GLIBCXX_PARALLEL_ASSERT(!__comp(*__r, *__pivot_pos));
|
_GLIBCXX_PARALLEL_ASSERT(!__comp(*__r, *__pivot_pos));
|
#endif
|
#endif
|
|
|
return __split_pos;
|
return __split_pos;
|
}
|
}
|
|
|
/** @brief Quicksort conquer step.
|
/** @brief Quicksort conquer step.
|
* @param __tls Array of thread-local storages.
|
* @param __tls Array of thread-local storages.
|
* @param __begin Begin iterator of subsequence.
|
* @param __begin Begin iterator of subsequence.
|
* @param __end End iterator of subsequence.
|
* @param __end End iterator of subsequence.
|
* @param __comp Comparator.
|
* @param __comp Comparator.
|
* @param __iam Number of the thread processing this function.
|
* @param __iam Number of the thread processing this function.
|
* @param __num_threads
|
* @param __num_threads
|
* Number of threads that are allowed to work on this part. */
|
* Number of threads that are allowed to work on this part. */
|
template<typename _RAIter, typename _Compare>
|
template<typename _RAIter, typename _Compare>
|
void
|
void
|
__qsb_conquer(_QSBThreadLocal<_RAIter>** __tls,
|
__qsb_conquer(_QSBThreadLocal<_RAIter>** __tls,
|
_RAIter __begin, _RAIter __end,
|
_RAIter __begin, _RAIter __end,
|
_Compare __comp,
|
_Compare __comp,
|
_ThreadIndex __iam, _ThreadIndex __num_threads,
|
_ThreadIndex __iam, _ThreadIndex __num_threads,
|
bool __parent_wait)
|
bool __parent_wait)
|
{
|
{
|
typedef std::iterator_traits<_RAIter> _TraitsType;
|
typedef std::iterator_traits<_RAIter> _TraitsType;
|
typedef typename _TraitsType::value_type _ValueType;
|
typedef typename _TraitsType::value_type _ValueType;
|
typedef typename _TraitsType::difference_type _DifferenceType;
|
typedef typename _TraitsType::difference_type _DifferenceType;
|
|
|
_DifferenceType __n = __end - __begin;
|
_DifferenceType __n = __end - __begin;
|
|
|
if (__num_threads <= 1 || __n <= 1)
|
if (__num_threads <= 1 || __n <= 1)
|
{
|
{
|
__tls[__iam]->_M_initial.first = __begin;
|
__tls[__iam]->_M_initial.first = __begin;
|
__tls[__iam]->_M_initial.second = __end;
|
__tls[__iam]->_M_initial.second = __end;
|
|
|
__qsb_local_sort_with_helping(__tls, __comp, __iam, __parent_wait);
|
__qsb_local_sort_with_helping(__tls, __comp, __iam, __parent_wait);
|
|
|
return;
|
return;
|
}
|
}
|
|
|
// Divide step.
|
// Divide step.
|
_DifferenceType __split_pos =
|
_DifferenceType __split_pos =
|
__qsb_divide(__begin, __end, __comp, __num_threads);
|
__qsb_divide(__begin, __end, __comp, __num_threads);
|
|
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
_GLIBCXX_PARALLEL_ASSERT(0 <= __split_pos &&
|
_GLIBCXX_PARALLEL_ASSERT(0 <= __split_pos &&
|
__split_pos < (__end - __begin));
|
__split_pos < (__end - __begin));
|
#endif
|
#endif
|
|
|
_ThreadIndex
|
_ThreadIndex
|
__num_threads_leftside = std::max<_ThreadIndex>
|
__num_threads_leftside = std::max<_ThreadIndex>
|
(1, std::min<_ThreadIndex>(__num_threads - 1, __split_pos
|
(1, std::min<_ThreadIndex>(__num_threads - 1, __split_pos
|
* __num_threads / __n));
|
* __num_threads / __n));
|
|
|
# pragma omp atomic
|
# pragma omp atomic
|
*__tls[__iam]->_M_elements_leftover -= (_DifferenceType)1;
|
*__tls[__iam]->_M_elements_leftover -= (_DifferenceType)1;
|
|
|
// Conquer step.
|
// Conquer step.
|
# pragma omp parallel num_threads(2)
|
# pragma omp parallel num_threads(2)
|
{
|
{
|
bool __wait;
|
bool __wait;
|
if(omp_get_num_threads() < 2)
|
if(omp_get_num_threads() < 2)
|
__wait = false;
|
__wait = false;
|
else
|
else
|
__wait = __parent_wait;
|
__wait = __parent_wait;
|
|
|
# pragma omp sections
|
# pragma omp sections
|
{
|
{
|
# pragma omp section
|
# pragma omp section
|
{
|
{
|
__qsb_conquer(__tls, __begin, __begin + __split_pos, __comp,
|
__qsb_conquer(__tls, __begin, __begin + __split_pos, __comp,
|
__iam, __num_threads_leftside, __wait);
|
__iam, __num_threads_leftside, __wait);
|
__wait = __parent_wait;
|
__wait = __parent_wait;
|
}
|
}
|
// The pivot_pos is left in place, to ensure termination.
|
// The pivot_pos is left in place, to ensure termination.
|
# pragma omp section
|
# pragma omp section
|
{
|
{
|
__qsb_conquer(__tls, __begin + __split_pos + 1, __end, __comp,
|
__qsb_conquer(__tls, __begin + __split_pos + 1, __end, __comp,
|
__iam + __num_threads_leftside,
|
__iam + __num_threads_leftside,
|
__num_threads - __num_threads_leftside, __wait);
|
__num_threads - __num_threads_leftside, __wait);
|
__wait = __parent_wait;
|
__wait = __parent_wait;
|
}
|
}
|
}
|
}
|
}
|
}
|
}
|
}
|
|
|
/**
|
/**
|
* @brief Quicksort step doing load-balanced local sort.
|
* @brief Quicksort step doing load-balanced local sort.
|
* @param __tls Array of thread-local storages.
|
* @param __tls Array of thread-local storages.
|
* @param __comp Comparator.
|
* @param __comp Comparator.
|
* @param __iam Number of the thread processing this function.
|
* @param __iam Number of the thread processing this function.
|
*/
|
*/
|
template<typename _RAIter, typename _Compare>
|
template<typename _RAIter, typename _Compare>
|
void
|
void
|
__qsb_local_sort_with_helping(_QSBThreadLocal<_RAIter>** __tls,
|
__qsb_local_sort_with_helping(_QSBThreadLocal<_RAIter>** __tls,
|
_Compare& __comp, _ThreadIndex __iam,
|
_Compare& __comp, _ThreadIndex __iam,
|
bool __wait)
|
bool __wait)
|
{
|
{
|
typedef std::iterator_traits<_RAIter> _TraitsType;
|
typedef std::iterator_traits<_RAIter> _TraitsType;
|
typedef typename _TraitsType::value_type _ValueType;
|
typedef typename _TraitsType::value_type _ValueType;
|
typedef typename _TraitsType::difference_type _DifferenceType;
|
typedef typename _TraitsType::difference_type _DifferenceType;
|
typedef std::pair<_RAIter, _RAIter> _Piece;
|
typedef std::pair<_RAIter, _RAIter> _Piece;
|
|
|
_QSBThreadLocal<_RAIter>& __tl = *__tls[__iam];
|
_QSBThreadLocal<_RAIter>& __tl = *__tls[__iam];
|
|
|
_DifferenceType
|
_DifferenceType
|
__base_case_n = _Settings::get().sort_qsb_base_case_maximal_n;
|
__base_case_n = _Settings::get().sort_qsb_base_case_maximal_n;
|
if (__base_case_n < 2)
|
if (__base_case_n < 2)
|
__base_case_n = 2;
|
__base_case_n = 2;
|
_ThreadIndex __num_threads = __tl._M_num_threads;
|
_ThreadIndex __num_threads = __tl._M_num_threads;
|
|
|
// Every thread has its own random number generator.
|
// Every thread has its own random number generator.
|
_RandomNumber __rng(__iam + 1);
|
_RandomNumber __rng(__iam + 1);
|
|
|
_Piece __current = __tl._M_initial;
|
_Piece __current = __tl._M_initial;
|
|
|
_DifferenceType __elements_done = 0;
|
_DifferenceType __elements_done = 0;
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
_DifferenceType __total_elements_done = 0;
|
_DifferenceType __total_elements_done = 0;
|
#endif
|
#endif
|
|
|
for (;;)
|
for (;;)
|
{
|
{
|
// Invariant: __current must be a valid (maybe empty) range.
|
// Invariant: __current must be a valid (maybe empty) range.
|
_RAIter __begin = __current.first, __end = __current.second;
|
_RAIter __begin = __current.first, __end = __current.second;
|
_DifferenceType __n = __end - __begin;
|
_DifferenceType __n = __end - __begin;
|
|
|
if (__n > __base_case_n)
|
if (__n > __base_case_n)
|
{
|
{
|
// Divide.
|
// Divide.
|
_RAIter __pivot_pos = __begin + __rng(__n);
|
_RAIter __pivot_pos = __begin + __rng(__n);
|
|
|
// Swap __pivot_pos value to end.
|
// Swap __pivot_pos value to end.
|
if (__pivot_pos != (__end - 1))
|
if (__pivot_pos != (__end - 1))
|
std::swap(*__pivot_pos, *(__end - 1));
|
std::swap(*__pivot_pos, *(__end - 1));
|
__pivot_pos = __end - 1;
|
__pivot_pos = __end - 1;
|
|
|
__gnu_parallel::__binder2nd
|
__gnu_parallel::__binder2nd
|
<_Compare, _ValueType, _ValueType, bool>
|
<_Compare, _ValueType, _ValueType, bool>
|
__pred(__comp, *__pivot_pos);
|
__pred(__comp, *__pivot_pos);
|
|
|
// Divide, leave pivot unchanged in last place.
|
// Divide, leave pivot unchanged in last place.
|
_RAIter __split_pos1, __split_pos2;
|
_RAIter __split_pos1, __split_pos2;
|
__split_pos1 = __gnu_sequential::partition(__begin, __end - 1,
|
__split_pos1 = __gnu_sequential::partition(__begin, __end - 1,
|
__pred);
|
__pred);
|
|
|
// Left side: < __pivot_pos; __right side: >= __pivot_pos.
|
// Left side: < __pivot_pos; __right side: >= __pivot_pos.
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
_GLIBCXX_PARALLEL_ASSERT(__begin <= __split_pos1
|
_GLIBCXX_PARALLEL_ASSERT(__begin <= __split_pos1
|
&& __split_pos1 < __end);
|
&& __split_pos1 < __end);
|
#endif
|
#endif
|
// Swap pivot back to middle.
|
// Swap pivot back to middle.
|
if (__split_pos1 != __pivot_pos)
|
if (__split_pos1 != __pivot_pos)
|
std::swap(*__split_pos1, *__pivot_pos);
|
std::swap(*__split_pos1, *__pivot_pos);
|
__pivot_pos = __split_pos1;
|
__pivot_pos = __split_pos1;
|
|
|
// In case all elements are equal, __split_pos1 == 0.
|
// In case all elements are equal, __split_pos1 == 0.
|
if ((__split_pos1 + 1 - __begin) < (__n >> 7)
|
if ((__split_pos1 + 1 - __begin) < (__n >> 7)
|
|| (__end - __split_pos1) < (__n >> 7))
|
|| (__end - __split_pos1) < (__n >> 7))
|
{
|
{
|
// Very unequal split, one part smaller than one 128th
|
// Very unequal split, one part smaller than one 128th
|
// elements not strictly larger than the pivot.
|
// elements not strictly larger than the pivot.
|
__gnu_parallel::__unary_negate<__gnu_parallel::__binder1st
|
__gnu_parallel::__unary_negate<__gnu_parallel::__binder1st
|
<_Compare, _ValueType, _ValueType, bool>, _ValueType>
|
<_Compare, _ValueType, _ValueType, bool>, _ValueType>
|
__pred(__gnu_parallel::__binder1st
|
__pred(__gnu_parallel::__binder1st
|
<_Compare, _ValueType, _ValueType, bool>
|
<_Compare, _ValueType, _ValueType, bool>
|
(__comp, *__pivot_pos));
|
(__comp, *__pivot_pos));
|
|
|
// Find other end of pivot-equal range.
|
// Find other end of pivot-equal range.
|
__split_pos2 = __gnu_sequential::partition(__split_pos1 + 1,
|
__split_pos2 = __gnu_sequential::partition(__split_pos1 + 1,
|
__end, __pred);
|
__end, __pred);
|
}
|
}
|
else
|
else
|
// Only skip the pivot.
|
// Only skip the pivot.
|
__split_pos2 = __split_pos1 + 1;
|
__split_pos2 = __split_pos1 + 1;
|
|
|
// Elements equal to pivot are done.
|
// Elements equal to pivot are done.
|
__elements_done += (__split_pos2 - __split_pos1);
|
__elements_done += (__split_pos2 - __split_pos1);
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
__total_elements_done += (__split_pos2 - __split_pos1);
|
__total_elements_done += (__split_pos2 - __split_pos1);
|
#endif
|
#endif
|
// Always push larger part onto stack.
|
// Always push larger part onto stack.
|
if (((__split_pos1 + 1) - __begin) < (__end - (__split_pos2)))
|
if (((__split_pos1 + 1) - __begin) < (__end - (__split_pos2)))
|
{
|
{
|
// Right side larger.
|
// Right side larger.
|
if ((__split_pos2) != __end)
|
if ((__split_pos2) != __end)
|
__tl._M_leftover_parts.push_front
|
__tl._M_leftover_parts.push_front
|
(std::make_pair(__split_pos2, __end));
|
(std::make_pair(__split_pos2, __end));
|
|
|
//__current.first = __begin; //already set anyway
|
//__current.first = __begin; //already set anyway
|
__current.second = __split_pos1;
|
__current.second = __split_pos1;
|
continue;
|
continue;
|
}
|
}
|
else
|
else
|
{
|
{
|
// Left side larger.
|
// Left side larger.
|
if (__begin != __split_pos1)
|
if (__begin != __split_pos1)
|
__tl._M_leftover_parts.push_front(std::make_pair
|
__tl._M_leftover_parts.push_front(std::make_pair
|
(__begin, __split_pos1));
|
(__begin, __split_pos1));
|
|
|
__current.first = __split_pos2;
|
__current.first = __split_pos2;
|
//__current.second = __end; //already set anyway
|
//__current.second = __end; //already set anyway
|
continue;
|
continue;
|
}
|
}
|
}
|
}
|
else
|
else
|
{
|
{
|
__gnu_sequential::sort(__begin, __end, __comp);
|
__gnu_sequential::sort(__begin, __end, __comp);
|
__elements_done += __n;
|
__elements_done += __n;
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
__total_elements_done += __n;
|
__total_elements_done += __n;
|
#endif
|
#endif
|
|
|
// Prefer own stack, small pieces.
|
// Prefer own stack, small pieces.
|
if (__tl._M_leftover_parts.pop_front(__current))
|
if (__tl._M_leftover_parts.pop_front(__current))
|
continue;
|
continue;
|
|
|
# pragma omp atomic
|
# pragma omp atomic
|
*__tl._M_elements_leftover -= __elements_done;
|
*__tl._M_elements_leftover -= __elements_done;
|
|
|
__elements_done = 0;
|
__elements_done = 0;
|
|
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
double __search_start = omp_get_wtime();
|
double __search_start = omp_get_wtime();
|
#endif
|
#endif
|
|
|
// Look for new work.
|
// Look for new work.
|
bool __successfully_stolen = false;
|
bool __successfully_stolen = false;
|
while (__wait && *__tl._M_elements_leftover > 0
|
while (__wait && *__tl._M_elements_leftover > 0
|
&& !__successfully_stolen
|
&& !__successfully_stolen
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
// Possible dead-lock.
|
// Possible dead-lock.
|
&& (omp_get_wtime() < (__search_start + 1.0))
|
&& (omp_get_wtime() < (__search_start + 1.0))
|
#endif
|
#endif
|
)
|
)
|
{
|
{
|
_ThreadIndex __victim;
|
_ThreadIndex __victim;
|
__victim = __rng(__num_threads);
|
__victim = __rng(__num_threads);
|
|
|
// Large pieces.
|
// Large pieces.
|
__successfully_stolen = (__victim != __iam)
|
__successfully_stolen = (__victim != __iam)
|
&& __tls[__victim]->_M_leftover_parts.pop_back(__current);
|
&& __tls[__victim]->_M_leftover_parts.pop_back(__current);
|
if (!__successfully_stolen)
|
if (!__successfully_stolen)
|
__yield();
|
__yield();
|
#if !defined(__ICC) && !defined(__ECC)
|
#if !defined(__ICC) && !defined(__ECC)
|
# pragma omp flush
|
# pragma omp flush
|
#endif
|
#endif
|
}
|
}
|
|
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
if (omp_get_wtime() >= (__search_start + 1.0))
|
if (omp_get_wtime() >= (__search_start + 1.0))
|
{
|
{
|
sleep(1);
|
sleep(1);
|
_GLIBCXX_PARALLEL_ASSERT(omp_get_wtime()
|
_GLIBCXX_PARALLEL_ASSERT(omp_get_wtime()
|
< (__search_start + 1.0));
|
< (__search_start + 1.0));
|
}
|
}
|
#endif
|
#endif
|
if (!__successfully_stolen)
|
if (!__successfully_stolen)
|
{
|
{
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
_GLIBCXX_PARALLEL_ASSERT(*__tl._M_elements_leftover == 0);
|
_GLIBCXX_PARALLEL_ASSERT(*__tl._M_elements_leftover == 0);
|
#endif
|
#endif
|
return;
|
return;
|
}
|
}
|
}
|
}
|
}
|
}
|
}
|
}
|
|
|
/** @brief Top-level quicksort routine.
|
/** @brief Top-level quicksort routine.
|
* @param __begin Begin iterator of sequence.
|
* @param __begin Begin iterator of sequence.
|
* @param __end End iterator of sequence.
|
* @param __end End iterator of sequence.
|
* @param __comp Comparator.
|
* @param __comp Comparator.
|
* @param __num_threads Number of threads that are allowed to work on
|
* @param __num_threads Number of threads that are allowed to work on
|
* this part.
|
* this part.
|
*/
|
*/
|
template<typename _RAIter, typename _Compare>
|
template<typename _RAIter, typename _Compare>
|
void
|
void
|
__parallel_sort_qsb(_RAIter __begin, _RAIter __end,
|
__parallel_sort_qsb(_RAIter __begin, _RAIter __end,
|
_Compare __comp, _ThreadIndex __num_threads)
|
_Compare __comp, _ThreadIndex __num_threads)
|
{
|
{
|
_GLIBCXX_CALL(__end - __begin)
|
_GLIBCXX_CALL(__end - __begin)
|
|
|
typedef std::iterator_traits<_RAIter> _TraitsType;
|
typedef std::iterator_traits<_RAIter> _TraitsType;
|
typedef typename _TraitsType::value_type _ValueType;
|
typedef typename _TraitsType::value_type _ValueType;
|
typedef typename _TraitsType::difference_type _DifferenceType;
|
typedef typename _TraitsType::difference_type _DifferenceType;
|
typedef std::pair<_RAIter, _RAIter> _Piece;
|
typedef std::pair<_RAIter, _RAIter> _Piece;
|
|
|
typedef _QSBThreadLocal<_RAIter> _TLSType;
|
typedef _QSBThreadLocal<_RAIter> _TLSType;
|
|
|
_DifferenceType __n = __end - __begin;
|
_DifferenceType __n = __end - __begin;
|
|
|
if (__n <= 1)
|
if (__n <= 1)
|
return;
|
return;
|
|
|
// At least one element per processor.
|
// At least one element per processor.
|
if (__num_threads > __n)
|
if (__num_threads > __n)
|
__num_threads = static_cast<_ThreadIndex>(__n);
|
__num_threads = static_cast<_ThreadIndex>(__n);
|
|
|
// Initialize thread local storage
|
// Initialize thread local storage
|
_TLSType** __tls = new _TLSType*[__num_threads];
|
_TLSType** __tls = new _TLSType*[__num_threads];
|
_DifferenceType __queue_size = (__num_threads
|
_DifferenceType __queue_size = (__num_threads
|
* (_ThreadIndex)(__rd_log2(__n) + 1));
|
* (_ThreadIndex)(__rd_log2(__n) + 1));
|
for (_ThreadIndex __t = 0; __t < __num_threads; ++__t)
|
for (_ThreadIndex __t = 0; __t < __num_threads; ++__t)
|
__tls[__t] = new _QSBThreadLocal<_RAIter>(__queue_size);
|
__tls[__t] = new _QSBThreadLocal<_RAIter>(__queue_size);
|
|
|
// There can never be more than ceil(__rd_log2(__n)) ranges on the
|
// There can never be more than ceil(__rd_log2(__n)) ranges on the
|
// stack, because
|
// stack, because
|
// 1. Only one processor pushes onto the stack
|
// 1. Only one processor pushes onto the stack
|
// 2. The largest range has at most length __n
|
// 2. The largest range has at most length __n
|
// 3. Each range is larger than half of the range remaining
|
// 3. Each range is larger than half of the range remaining
|
volatile _DifferenceType __elements_leftover = __n;
|
volatile _DifferenceType __elements_leftover = __n;
|
for (_ThreadIndex __i = 0; __i < __num_threads; ++__i)
|
for (_ThreadIndex __i = 0; __i < __num_threads; ++__i)
|
{
|
{
|
__tls[__i]->_M_elements_leftover = &__elements_leftover;
|
__tls[__i]->_M_elements_leftover = &__elements_leftover;
|
__tls[__i]->_M_num_threads = __num_threads;
|
__tls[__i]->_M_num_threads = __num_threads;
|
__tls[__i]->_M_global = std::make_pair(__begin, __end);
|
__tls[__i]->_M_global = std::make_pair(__begin, __end);
|
|
|
// Just in case nothing is left to assign.
|
// Just in case nothing is left to assign.
|
__tls[__i]->_M_initial = std::make_pair(__end, __end);
|
__tls[__i]->_M_initial = std::make_pair(__end, __end);
|
}
|
}
|
|
|
// Main recursion call.
|
// Main recursion call.
|
__qsb_conquer(__tls, __begin, __begin + __n, __comp, 0,
|
__qsb_conquer(__tls, __begin, __begin + __n, __comp, 0,
|
__num_threads, true);
|
__num_threads, true);
|
|
|
#if _GLIBCXX_ASSERTIONS
|
#if _GLIBCXX_ASSERTIONS
|
// All stack must be empty.
|
// All stack must be empty.
|
_Piece __dummy;
|
_Piece __dummy;
|
for (_ThreadIndex __i = 1; __i < __num_threads; ++__i)
|
for (_ThreadIndex __i = 1; __i < __num_threads; ++__i)
|
_GLIBCXX_PARALLEL_ASSERT(
|
_GLIBCXX_PARALLEL_ASSERT(
|
!__tls[__i]->_M_leftover_parts.pop_back(__dummy));
|
!__tls[__i]->_M_leftover_parts.pop_back(__dummy));
|
#endif
|
#endif
|
|
|
for (_ThreadIndex __i = 0; __i < __num_threads; ++__i)
|
for (_ThreadIndex __i = 0; __i < __num_threads; ++__i)
|
delete __tls[__i];
|
delete __tls[__i];
|
delete[] __tls;
|
delete[] __tls;
|
}
|
}
|
} // namespace __gnu_parallel
|
} // namespace __gnu_parallel
|
|
|
#endif /* _GLIBCXX_PARALLEL_BALANCED_QUICKSORT_H */
|
#endif /* _GLIBCXX_PARALLEL_BALANCED_QUICKSORT_H */
|
|
|
