summaryrefslogtreecommitdiff
path: root/boost/sort/block_indirect_sort/block_indirect_sort.hpp
blob: 62abde29a50e069de94abcd8ee6ae60296f4f805 (plain)
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
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
//----------------------------------------------------------------------------
/// @file block_indirect_sort.hpp
/// @brief block indirect sort algorithm
///
/// @author Copyright (c) 2016 Francisco Jose Tapia (fjtapia@gmail.com )\n
///         Distributed under the Boost Software License, Version 1.0.\n
///         ( See accompanying file LICENSE_1_0.txt or copy at
///           http://www.boost.org/LICENSE_1_0.txt  )
/// @version 0.1
///
/// @remarks
//-----------------------------------------------------------------------------
#ifndef __BOOST_SORT_PARALLEL_DETAIL_BLOCK_INDIRECT_SORT_HPP
#define __BOOST_SORT_PARALLEL_DETAIL_BLOCK_INDIRECT_SORT_HPP

#include <atomic>
#include <boost/sort/block_indirect_sort/blk_detail/merge_blocks.hpp>
#include <boost/sort/block_indirect_sort/blk_detail/move_blocks.hpp>
#include <boost/sort/block_indirect_sort/blk_detail/parallel_sort.hpp>
#include <boost/sort/pdqsort/pdqsort.hpp>
#include <boost/sort/common/util/traits.hpp>
#include <boost/sort/common/util/algorithm.hpp>
#include <future>
#include <iterator>

// This value is the minimal number of threads for to use the
// block_indirect_sort algorithm
#define BOOST_NTHREAD_BORDER 6

namespace boost
{
namespace sort
{
namespace blk_detail
{
//---------------------------------------------------------------------------
//         USING SENTENCES
//---------------------------------------------------------------------------
namespace bs = boost::sort;
namespace bsc = bs::common;
namespace bscu = bsc::util;
using bscu::compare_iter;
using bscu::value_iter;
using bsc::range;
using bsc::destroy;
using bsc::initialize;
using bscu::nbits64;
using bs::pdqsort;
using bscu::enable_if_string;
using bscu::enable_if_not_string;
using bscu::tmsb;
//
///---------------------------------------------------------------------------
/// @struct block_indirect_sort
/// @brief This class is the entry point of the block indirect sort. The code
///        of this algorithm is divided in several classes:
///        bis/block.hpp : basic structures used in the algorithm
///        bis/backbone.hpp : data used by all the classes
///        bis/merge_blocks.hpp : merge the internal blocks
///        bis/move_blocks.hpp : move the blocks, and obtain all the elements
///                              phisicaly sorted
///        bis/parallel_sort.hpp : make the parallel sort of each part in the
///                                initial division of the data
///
//----------------------------------------------------------------------------
template<uint32_t Block_size, uint32_t Group_size, class Iter_t,
                class Compare = compare_iter<Iter_t> >
struct block_indirect_sort
{
    //------------------------------------------------------------------------
    //                  D E F I N I T I O N S
    //------------------------------------------------------------------------
    typedef typename std::iterator_traits<Iter_t>::value_type value_t;
    typedef std::atomic<uint32_t> atomic_t;
    typedef range<size_t> range_pos;
    typedef range<Iter_t> range_it;
    typedef range<value_t *> range_buf;
    typedef std::function<void(void)> function_t;

    // classes used in the internal operations of the algorithm
    typedef block_pos block_pos_t;
    typedef block<Block_size, Iter_t> block_t;
    typedef backbone<Block_size, Iter_t, Compare> backbone_t;
    typedef parallel_sort<Block_size, Iter_t, Compare> parallel_sort_t;

    typedef merge_blocks<Block_size, Group_size, Iter_t, Compare> merge_blocks_t;
    typedef move_blocks<Block_size, Group_size, Iter_t, Compare> move_blocks_t;
    typedef compare_block_pos<Block_size, Iter_t, Compare> compare_block_pos_t;
    //
    //------------------------------------------------------------------------
    //       V A R I A B L E S   A N D  C O N S T A N T S
    //------------------------------------------------------------------------
    // contains the data and the internal data structures of the algorithm for
    // to be shared between the classes which are part of the algorithm
    backbone_t bk;
    // atomic counter for to detect the end of the works created inside
    // the object
    atomic_t counter;
    // pointer to the uninitialized memory used for the thread buffers
    value_t *ptr;
    // indicate if the memory pointed by ptr is initialized
    bool construct;
    // range from extract the buffers for the threads
    range_buf rglobal_buf;
    // number of threads to use
    uint32_t nthread;
    //
    //------------------------------------------------------------------------
    //                F U N C T I O N S
    //------------------------------------------------------------------------

    block_indirect_sort(Iter_t first, Iter_t last, Compare cmp, uint32_t nthr);

    block_indirect_sort(Iter_t first, Iter_t last) :
                        block_indirect_sort(first, last, Compare(),
                        std::thread::hardware_concurrency()) { }


    block_indirect_sort(Iter_t first, Iter_t last, Compare cmp) :
                        block_indirect_sort(first, last, cmp,
                        std::thread::hardware_concurrency()) { }


    block_indirect_sort(Iter_t first, Iter_t last, uint32_t nthread) :
                        block_indirect_sort(first, last, Compare(), nthread){}


    //
    //------------------------------------------------------------------------
    //  function :destroy_all
    /// @brief destructor all the data structures of the class (if the memory
    ///        is constructed, is destroyed) and  return the uninitialized
    ///        memory
    //------------------------------------------------------------------------
    void destroy_all(void)
    {
        if (ptr != nullptr)
        {
            if (construct)
            {
                destroy(rglobal_buf);
                construct = false;
            };
            std::return_temporary_buffer(ptr);
            ptr = nullptr;
        };
    }
    //
    //------------------------------------------------------------------------
    //  function :~block_indirect_sort
    /// @brief destructor of the class (if the memory is constructed, is
    ///        destroyed) and  return the uninitialized memory
    //------------------------------------------------------------------------
    ~block_indirect_sort(void)
    {
        destroy_all();
    }

    void split_range(size_t pos_index1, size_t pos_index2,
                    uint32_t level_thread);

    void start_function(void);

//-------------------------------------------------------------------------
}; // End class block_indirect_sort
//----------------------------------------------------------------------------
//
//############################################################################
//                                                                          ##
//                                                                          ##
//            N O N     I N L I N E      F U N C T I O N S                  ##
//                                                                          ##
//                                                                          ##
//############################################################################
//
//-------------------------------------------------------------------------
//  function : block_indirect_sort
/// @brief begin with the execution of the functions stored in works
/// @param first : iterator to the first element of the range to sort
/// @param last : iterator after the last element to the range to sort
/// @param comp : object for to compare two elements pointed by Iter_t
///               iterators
/// @param nthr : Number of threads to use in the process.When this value
///               is lower than 2, the sorting is done with 1 thread
//-------------------------------------------------------------------------
template<uint32_t Block_size, uint32_t Group_size, class Iter_t, class Compare>
block_indirect_sort<Block_size, Group_size, Iter_t, Compare>
::block_indirect_sort(Iter_t first, Iter_t last, Compare cmp, uint32_t nthr)
: bk(first, last, cmp), counter(0), ptr(nullptr), construct(false),
  nthread(nthr)
{
    try
    {
        assert((last - first) >= 0);
        size_t nelem = size_t(last - first);
        if (nelem == 0) return;

        //------------------- check if sort -----------------------------------
        bool sorted = true;
        for (Iter_t it1 = first, it2 = first + 1; it2 != last and (sorted =
                        not bk.cmp(*it2, *it1)); it1 = it2++);
        if (sorted) return;

        //------------------- check if reverse sort ---------------------------
        sorted = true;
        for (Iter_t it1 = first, it2 = first + 1; it2 != last and (sorted =
                        not bk.cmp(*it1, *it2)); it1 = it2++);

        if (sorted)
        {
            size_t nelem2 = nelem >> 1;
            Iter_t it1 = first, it2 = last - 1;
            for (size_t i = 0; i < nelem2; ++i)
            {
                std::swap(*(it1++), *(it2--));
            };
            return;
        };

        //---------------- check if only single thread -----------------------
        size_t nthreadmax = nelem / (Block_size * Group_size) + 1;
        if (nthread > nthreadmax) nthread = (uint32_t) nthreadmax;

        uint32_t nbits_size = (nbits64(sizeof(value_t)) >> 1);
        if (nbits_size > 5) nbits_size = 5;
        size_t max_per_thread = 1 << (18 - nbits_size);

        if (nelem < (max_per_thread) or nthread < 2)
        {
            //intro_sort (first, last, bk.cmp);
            pdqsort(first, last, bk.cmp);
            return;
        };

        //----------- creation of the temporary buffer --------------------
        ptr = std::get_temporary_buffer<value_t>(Block_size * nthread).first;
        if (ptr == nullptr)
        {
            bk.error = true;
            throw std::bad_alloc();
        };

        rglobal_buf = range_buf(ptr, ptr + (Block_size * nthread));
        initialize(rglobal_buf, *first);
        construct = true;

        // creation of the buffers for the threads
        std::vector<value_t *> vbuf(nthread);
        for (uint32_t i = 0; i < nthread; ++i)
        {
            vbuf[i] = ptr + (i * Block_size);
        };

        // Insert the first work in the stack
        bscu::atomic_write(counter, 1);
        function_t f1 = [&]( )
        {
            start_function ( );
            bscu::atomic_sub (counter, 1);
        };
        bk.works.emplace_back(f1);

        //---------------------------------------------------------------------
        //                    PROCESS
        //---------------------------------------------------------------------
        std::vector<std::future<void> > vfuture(nthread);

        // The function launched with the futures is "execute the functions of
        // the stack until this->counter is zero
        // vbuf[i] is the memory from the main thread for to configure the
        // thread local buffer
        for (uint32_t i = 0; i < nthread; ++i)
        {
            auto f1 = [=, &vbuf]( )
            {   bk.exec (vbuf[i], this->counter);};
            vfuture[i] = std::async(std::launch::async, f1);
        };
        for (uint32_t i = 0; i < nthread; ++i)
            vfuture[i].get();
        if (bk.error) throw std::bad_alloc();
    }
    catch (std::bad_alloc &)
    {
        destroy_all();
        throw;
    }
};
//
//-----------------------------------------------------------------------------
//  function : split_rage
/// @brief this function splits a range of positions in the index, and
///        depending of the size, sort directly or make to a recursive call
///        to split_range
/// @param pos_index1 : first position in the index
/// @param pos_index2 : position after the last in the index
/// @param level_thread : depth of the call. When 0 sort the blocks
//-----------------------------------------------------------------------------
template<uint32_t Block_size, uint32_t Group_size, class Iter_t, class Compare>
void block_indirect_sort<Block_size, Group_size, Iter_t, Compare>
::split_range(size_t pos_index1, size_t pos_index2, uint32_t level_thread)
{
    size_t nblock = pos_index2 - pos_index1;

    //-------------------------------------------------------------------------
    // In the blocks not sorted, the physical position is the logical position
    //-------------------------------------------------------------------------
    Iter_t first = bk.get_block(pos_index1).first;
    Iter_t last = bk.get_range(pos_index2 - 1).last;

    if (nblock < Group_size)
    {
        pdqsort(first, last, bk.cmp);
        return;
    };

    size_t pos_index_mid = pos_index1 + (nblock >> 1);
    atomic_t son_counter(1);

    //-------------------------------------------------------------------------
    // Insert in the stack the work for the second part, and the actual thread,
    // execute the first part
    //-------------------------------------------------------------------------
    if (level_thread != 0)
    {
        auto f1 = [=, &son_counter]( )
        {
            split_range (pos_index_mid, pos_index2, level_thread - 1);
            bscu::atomic_sub (son_counter, 1);
        };
        bk.works.emplace_back(f1);
        if (bk.error) return;
        split_range(pos_index1, pos_index_mid, level_thread - 1);
    }
    else
    {
        Iter_t mid = first + ((nblock >> 1) * Block_size);
        auto f1 = [=, &son_counter]( )
        {
            parallel_sort_t (bk, mid, last);
            bscu::atomic_sub (son_counter, 1);
        };
        bk.works.emplace_back(f1);
        if (bk.error) return;
        parallel_sort_t(bk, first, mid);
    };
    bk.exec(son_counter);
    if (bk.error) return;
    merge_blocks_t(bk, pos_index1, pos_index_mid, pos_index2);
};

//
//-----------------------------------------------------------------------------
//  function : start_function
/// @brief this function init the process. When the number of threads is lower
///        than a predefined value, sort the elements with a parallel pdqsort.
//-----------------------------------------------------------------------------
template<uint32_t Block_size, uint32_t Group_size, class Iter_t, class Compare>
void block_indirect_sort<Block_size, Group_size, Iter_t, Compare>
::start_function(void)
{
    if (nthread < BOOST_NTHREAD_BORDER)
    {
        parallel_sort_t(bk, bk.global_range.first, bk.global_range.last);
    }
    else
    {
        size_t level_thread = nbits64(nthread - 1) - 1;
        split_range(0, bk.nblock, level_thread - 1);
        if (bk.error) return;
        move_blocks_t k(bk);
    };
};

///---------------------------------------------------------------------------
//  function block_indirect_sort_call
/// @brief This class is select the block size in the block_indirect_sort
///        algorithm depending of the type and size of the data to sort
///
//----------------------------------------------------------------------------
template <class Iter_t, class Compare,
         enable_if_string<value_iter<Iter_t>> * = nullptr>
inline void block_indirect_sort_call(Iter_t first, Iter_t last, Compare cmp,
                uint32_t nthr)
{
    block_indirect_sort<128, 128, Iter_t, Compare>(first, last, cmp, nthr);
};

template<size_t Size>
struct block_size
{
    static constexpr const uint32_t BitsSize =
                    (Size == 0) ? 0 : (Size > 256) ? 9 : tmsb[Size - 1];
    static constexpr const uint32_t sz[10] =
    { 4096, 4096, 4096, 4096, 2048, 1024, 768, 512, 256, 128 };
    static constexpr const uint32_t data = sz[BitsSize];
};
//
///---------------------------------------------------------------------------
/// @struct block_indirect_sort_call
/// @brief This class is select the block size in the block_indirect_sort
///        algorithm depending of the type and size of the data to sort
///
//----------------------------------------------------------------------------
template <class Iter_t, class Compare,
          enable_if_not_string<value_iter<Iter_t>> * = nullptr>
inline void block_indirect_sort_call (Iter_t first, Iter_t last, Compare cmp,
                                      uint32_t nthr)
{
    block_indirect_sort<block_size<sizeof (value_iter<Iter_t> )>::data, 64,
                        Iter_t, Compare> (first, last, cmp, nthr);
};

//
//****************************************************************************
}; //    End namespace blk_detail
//****************************************************************************
//
namespace bscu = boost::sort::common::util;
//
//############################################################################
//                                                                          ##
//                                                                          ##
//               B L O C K _ I N D I R E C T _ S O R T                      ##
//                                                                          ##
//                                                                          ##
//############################################################################
//
//-----------------------------------------------------------------------------
//  function : block_indirect_sort
/// @brief parallel sample sort  algorithm (stable sort)
///
/// @param first : iterator to the first element of the range to sort
/// @param last : iterator after the last element to the range to sort
//-----------------------------------------------------------------------------
template<class Iter_t>
void block_indirect_sort(Iter_t first, Iter_t last)
{
    typedef bscu::compare_iter<Iter_t> Compare;
    blk_detail::block_indirect_sort_call (first, last, Compare(),
                                          std::thread::hardware_concurrency());
}

//
//-----------------------------------------------------------------------------
//  function : block_indirect_sort
/// @brief parallel sample sort  algorithm (stable sort)
///
/// @param first : iterator to the first element of the range to sort
/// @param last : iterator after the last element to the range to sort
/// @param nthread : Number of threads to use in the process. When this value
///                  is lower than 2, the sorting is done with 1 thread
//-----------------------------------------------------------------------------
template<class Iter_t>
void block_indirect_sort(Iter_t first, Iter_t last, uint32_t nthread)
{
    typedef bscu::compare_iter<Iter_t> Compare;
    blk_detail::block_indirect_sort_call(first, last, Compare(), nthread);
}
//
//-----------------------------------------------------------------------------
//  function : block_indirect_sort
/// @brief parallel sample sort  algorithm (stable sort)
///
/// @param first : iterator to the first element of the range to sort
/// @param last : iterator after the last element to the range to sort
/// @param comp : object for to compare two elements pointed by Iter_t
///               iterators
//-----------------------------------------------------------------------------
template <class Iter_t, class Compare,
          bscu::enable_if_not_integral<Compare> * = nullptr>
void block_indirect_sort(Iter_t first, Iter_t last, Compare comp)
{
    blk_detail::block_indirect_sort_call (first, last, comp,
                                      std::thread::hardware_concurrency());
}

//
//-----------------------------------------------------------------------------
//  function : block_indirect_sort
/// @brief parallel sample sort  algorithm (stable sort)
///
/// @param first : iterator to the first element of the range to sort
/// @param last : iterator after the last element to the range to sort
/// @param comp : object for to compare two elements pointed by Iter_t
///               iterators
/// @param nthread : Number of threads to use in the process. When this value
///                  is lower than 2, the sorting is done with 1 thread
//-----------------------------------------------------------------------------
template<class Iter_t, class Compare>
void block_indirect_sort (Iter_t first, Iter_t last, Compare comp,
                          uint32_t nthread)
{
    blk_detail::block_indirect_sort_call(first, last, comp, nthread);
}
//
//****************************************************************************
}; //    End namespace sort
}; //    End namespace boost
//****************************************************************************
//
#endif