device_partition.hpp 30.3 KB
Newer Older
zhuwenwen's avatar
zhuwenwen committed
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
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
// Copyright (c) 2017-2022 Advanced Micro Devices, Inc. All rights reserved.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.

#ifndef ROCPRIM_DEVICE_DEVICE_PARTITION_HPP_
#define ROCPRIM_DEVICE_DEVICE_PARTITION_HPP_

#include <algorithm>
#include <type_traits>
#include <iterator>

#include "../config.hpp"
#include "../functional.hpp"
#include "../types.hpp"
#include "../type_traits.hpp"
#include "../detail/various.hpp"

#include "device_select_config.hpp"
#include "detail/device_scan_common.hpp"
#include "detail/device_partition.hpp"
#include "device_transform.hpp"

BEGIN_ROCPRIM_NAMESPACE

/// \addtogroup devicemodule
/// @{

namespace detail
{

template<
    select_method SelectMethod,
    bool OnlySelected,
    class Config,
    class KeyIterator,
    class ValueIterator,
    class FlagIterator,
    class OutputKeyIterator,
    class OutputValueIterator,
    class InequalityOp,
    class OffsetLookbackScanState,
    class... UnaryPredicates
>
ROCPRIM_KERNEL
__launch_bounds__(Config::block_size)
void partition_kernel(KeyIterator keys_input,
                      ValueIterator values_input,
                      FlagIterator flags,
                      OutputKeyIterator keys_output,
                      OutputValueIterator values_output,
                      size_t* selected_count,
                      size_t* prev_selected_count,
                      const size_t size,
                      InequalityOp inequality_op,
                      OffsetLookbackScanState offset_scan_state,
                      const unsigned int number_of_blocks,
                      ordered_block_id<unsigned int> ordered_bid,
                      UnaryPredicates... predicates)
{
    partition_kernel_impl<SelectMethod, OnlySelected, Config>(
        keys_input, values_input, flags, keys_output, values_output, selected_count, prev_selected_count, 
        size, inequality_op, offset_scan_state, number_of_blocks, ordered_bid, predicates...
    );
}

#define ROCPRIM_DETAIL_HIP_SYNC(name, size, start) \
    if(debug_synchronous) \
    { \
        std::cout << name << "(" << size << ")"; \
        auto error = cudaStreamSynchronize(stream); \
        if(error != cudaSuccess) return error; \
        auto end = std::chrono::high_resolution_clock::now(); \
        auto d = std::chrono::duration_cast<std::chrono::duration<double>>(end - start); \
        std::cout << " " << d.count() * 1000 << " ms" << '\n'; \
    }

#define ROCPRIM_DETAIL_HIP_SYNC_AND_RETURN_ON_ERROR(name, size, start) \
    { \
        auto _error = cudaGetLastError(); \
        if(_error != cudaSuccess) return _error; \
        if(debug_synchronous) \
        { \
            std::cout << name << "(" << size << ")"; \
            auto __error = cudaStreamSynchronize(stream); \
            if(__error != cudaSuccess) return __error; \
            auto _end = std::chrono::high_resolution_clock::now(); \
            auto _d = std::chrono::duration_cast<std::chrono::duration<double>>(_end - start); \
            std::cout << " " << _d.count() * 1000 << " ms" << '\n'; \
        } \
    }

template<
    // Method of selection: flag, predicate, unique
    select_method SelectMethod,
     // if true, it doesn't copy rejected values to output
    bool OnlySelected,
    class Config,
    class OffsetT,
    class KeyIterator,
    class ValueIterator, // can be rocprim::empty_type* for key only
    class FlagIterator,
    class OutputKeyIterator,
    class OutputValueIterator, // can be rocprim::empty_type* for key only
    class InequalityOp,
    class SelectedCountOutputIterator,
    class... UnaryPredicates
>
inline
cudaError_t partition_impl(void * temporary_storage,
                          size_t& storage_size,
                          KeyIterator keys_input,
                          ValueIterator values_input,
                          FlagIterator flags,
                          OutputKeyIterator keys_output,
                          OutputValueIterator values_output,
                          SelectedCountOutputIterator selected_count_output,
                          const size_t size,
                          InequalityOp inequality_op,
                          const cudaStream_t stream,
                          bool debug_synchronous,
                          UnaryPredicates... predicates)
{
    using offset_type = OffsetT;
    using key_type = typename std::iterator_traits<KeyIterator>::value_type;
    using value_type = typename std::iterator_traits<ValueIterator>::value_type;

    // Get default config if Config is default_config
    using config = default_or_custom_config<
        Config,
        default_select_config<ROCPRIM_TARGET_ARCH, key_type, value_type>
    >;

    using offset_scan_state_type = detail::lookback_scan_state<offset_type>;
    using offset_scan_state_with_sleep_type = detail::lookback_scan_state<offset_type, true>;
    using ordered_block_id_type = detail::ordered_block_id<unsigned int>;


    static constexpr unsigned int block_size = config::block_size;
    static constexpr unsigned int items_per_thread = config::items_per_thread;
    static constexpr auto items_per_block = block_size * items_per_thread;

    static constexpr bool is_three_way = sizeof...(UnaryPredicates) == 2;

    static constexpr size_t size_limit = config::size_limit;
    static constexpr size_t aligned_size_limit = ::rocprim::max<size_t>(size_limit - (size_limit % items_per_block), items_per_block);
    const size_t limited_size = std::min<size_t>(size, aligned_size_limit);
    const bool use_limited_size = limited_size == aligned_size_limit;

    const unsigned int number_of_blocks = 
        static_cast<unsigned int>(::rocprim::detail::ceiling_div(limited_size, items_per_block));

    // Calculate required temporary storage
    size_t offset_scan_state_bytes = ::rocprim::detail::align_size(
        // This is valid even with offset_scan_state_with_sleep_type
        offset_scan_state_type::get_storage_size(number_of_blocks)
    );
    size_t ordered_block_id_bytes = ::rocprim::detail::align_size(
        ordered_block_id_type::get_storage_size(),
        alignof(size_t)
    );

    if(temporary_storage == nullptr)
    {
        // storage_size is never zero
        storage_size = offset_scan_state_bytes + ordered_block_id_bytes + (sizeof(size_t) * 2 * (is_three_way ? 2 : 1));

        return cudaSuccess;
    }

    // Start point for time measurements
    std::chrono::high_resolution_clock::time_point start;

    // Create and initialize lookback_scan_state obj
    auto offset_scan_state = offset_scan_state_type::create(
        temporary_storage, number_of_blocks
    );
    auto offset_scan_state_with_sleep = offset_scan_state_with_sleep_type::create(
        temporary_storage, number_of_blocks
    );
    // Create ad initialize ordered_block_id obj
    auto ptr = reinterpret_cast<char*>(temporary_storage);
    auto ordered_bid = ordered_block_id_type::create(
        reinterpret_cast<ordered_block_id_type::id_type*>(ptr + offset_scan_state_bytes)
    );

    size_t* selected_count = reinterpret_cast<size_t*>(ptr + offset_scan_state_bytes
                                                       + ordered_block_id_bytes);
    size_t* prev_selected_count
        = reinterpret_cast<size_t*>(ptr + offset_scan_state_bytes + ordered_block_id_bytes
                                    + (is_three_way ? 2 : 1) * sizeof(size_t));

    cudaError_t error;

    // Memset selected_count and prev_selected_count at once
    error = cudaMemsetAsync(selected_count,
                           0,
                           sizeof(*selected_count) * 2 * (is_three_way ? 2 : 1),
                           stream);
    if (error != cudaSuccess) return error;

    cudaDeviceProp prop;
    int deviceId;
    static_cast<void>(cudaGetDevice(&deviceId));
    static_cast<void>(cudaGetDeviceProperties(&prop, deviceId));


    int asicRevision = 0;


    const size_t number_of_launches = ::rocprim::detail::ceiling_div(size, aligned_size_limit);

    if(debug_synchronous)
    {
        std::cout << "use_limited_size " << use_limited_size << '\n';
        std::cout << "aligned_size_limit " << aligned_size_limit << '\n';
        std::cout << "number_of_launches " << number_of_launches << '\n';
        std::cout << "size " << size << '\n';
        std::cout << "block_size " << block_size << '\n';
        std::cout << "number of blocks " << number_of_blocks << '\n';
        std::cout << "items_per_block " << items_per_block << '\n';
    }

    for (size_t i = 0, offset = 0; i < number_of_launches; i++, offset+=limited_size)
    {
        const unsigned int current_size = static_cast<unsigned int>(std::min<size_t>(size - offset, limited_size));

        const unsigned int current_number_of_blocks = ::rocprim::detail::ceiling_div(current_size, items_per_block);

        auto grid_size = ::rocprim::detail::ceiling_div(number_of_blocks, block_size);

        if(debug_synchronous)
        {
            std::cout << "current size " << current_size << '\n';
            std::cout << "current number of blocks " << current_number_of_blocks << '\n';

            start = std::chrono::high_resolution_clock::now();
        }


        
            init_lookback_scan_state_kernel<offset_scan_state_type>
                <<<dim3(grid_size), dim3(block_size), 0, stream>>>(
                offset_scan_state, current_number_of_blocks, ordered_bid
            );
        

        ROCPRIM_DETAIL_HIP_SYNC_AND_RETURN_ON_ERROR("init_offset_scan_state_kernel", current_number_of_blocks, start)

        if(debug_synchronous) start = std::chrono::high_resolution_clock::now();

        grid_size = current_number_of_blocks;
        

        
            partition_kernel<
                    SelectMethod, OnlySelected, config
                >
                <<<dim3(grid_size), dim3(block_size), 0, stream>>>(
                keys_input + offset, values_input + offset, flags + offset, keys_output, values_output, selected_count, prev_selected_count,
                current_size, inequality_op, offset_scan_state, current_number_of_blocks, ordered_bid, predicates...
            );
        

        ROCPRIM_DETAIL_HIP_SYNC_AND_RETURN_ON_ERROR("partition_kernel", size, start)

        std::swap(selected_count, prev_selected_count);
    }

    error = ::rocprim::transform(
        prev_selected_count, selected_count_output, (is_three_way ? 2 : 1), 
        ::rocprim::identity<>{},
        stream, debug_synchronous
    );
    if (error != cudaSuccess) return error;

    return cudaSuccess;
}

#undef ROCPRIM_DETAIL_HIP_SYNC_AND_RETURN_ON_ERROR
#undef ROCPRIM_DETAIL_HIP_SYNC

} // end of detail namespace

/// \brief Parallel select primitive for device level using range of flags.
///
/// Performs a device-wide partition based on input \p flags. Partition copies
/// the values from \p input to \p output in such a way that all values for which the corresponding
/// items from /p flags are \p true (or can be implicitly converted to \p true) precede
/// the elements for which the corresponding items from /p flags are \p false.
///
/// \par Overview
/// * Returns the required size of \p temporary_storage in \p storage_size
/// if \p temporary_storage in a null pointer.
/// * Ranges specified by \p input, \p flags and \p output must have at least \p size elements.
/// * Range specified by \p selected_count_output must have at least 1 element.
/// * Values of \p flag range should be implicitly convertible to `bool` type.
/// * Relative order is preserved for the elements for which the corresponding values from \p flags
/// are \p true. Other elements are copied in reverse order.
///
/// \tparam Config - [optional] configuration of the primitive. It can be \p select_config or
/// a custom class with the same members.
/// \tparam InputIterator - random-access iterator type of the input range. It can be
/// a simple pointer type.
/// \tparam FlagIterator - random-access iterator type of the flag range. It can be
/// a simple pointer type.
/// \tparam OutputIterator - random-access iterator type of the output range. It can be
/// a simple pointer type.
/// \tparam SelectedCountOutputIterator - random-access iterator type of the selected_count_output
/// value. It can be a simple pointer type.
///
/// \param [in] temporary_storage - pointer to a device-accessible temporary storage. When
/// a null pointer is passed, the required allocation size (in bytes) is written to
/// \p storage_size and function returns without performing the select operation.
/// \param [in,out] storage_size - reference to a size (in bytes) of \p temporary_storage.
/// \param [in] input - iterator to the first element in the range to select values from.
/// \param [in] flags - iterator to the selection flag corresponding to the first element from \p input range.
/// \param [out] output - iterator to the first element in the output range.
/// \param [out] selected_count_output - iterator to the total number of selected values (length of \p output).
/// \param [in] size - number of element in the input range.
/// \param [in] stream - [optional] HIP stream object. The default is \p 0 (default stream).
/// \param [in] debug_synchronous - [optional] If true, synchronization after every kernel
/// launch is forced in order to check for errors. The default value is \p false.
///
/// \par Example
/// \parblock
/// In this example a device-level partition operation is performed on an array of
/// integer values with array of <tt>char</tt>s used as flags.
///
/// \code{.cpp}
/// #include <rocprim/rocprim.hpp>
///
/// // Prepare input and output (declare pointers, allocate device memory etc.)
/// size_t input_size;     // e.g., 8
/// int * input;           // e.g., [1, 2, 3, 4, 5, 6, 7, 8]
/// char * flags;          // e.g., [0, 1, 1, 0, 0, 1, 0, 1]
/// int * output;          // empty array of 8 elements
/// size_t * output_count; // empty array of 1 element
///
/// size_t temporary_storage_size_bytes;
/// void * temporary_storage_ptr = nullptr;
/// // Get required size of the temporary storage
/// rocprim::partition(
///     temporary_storage_ptr, temporary_storage_size_bytes,
///     input, flags,
///     output, output_count,
///     input_size
/// );
///
/// // allocate temporary storage
/// cudaMalloc(&temporary_storage_ptr, temporary_storage_size_bytes);
///
/// // perform partition
/// rocprim::partition(
///     temporary_storage_ptr, temporary_storage_size_bytes,
///     input, flags,
///     output, output_count,
///     input_size
/// );
/// // output: [2, 3, 6, 8, 7, 5, 4, 1]
/// // output_count: 4
/// \endcode
/// \endparblock
template<
    class Config = default_config,
    class InputIterator,
    class FlagIterator,
    class OutputIterator,
    class SelectedCountOutputIterator
>
inline
cudaError_t partition(void * temporary_storage,
                     size_t& storage_size,
                     InputIterator input,
                     FlagIterator flags,
                     OutputIterator output,
                     SelectedCountOutputIterator selected_count_output,
                     const size_t size,
                     const cudaStream_t stream = 0,
                     const bool debug_synchronous = false)
{
    // Dummy unary predicate
    using unary_predicate_type = ::rocprim::empty_type;
    // Dummy inequality operation
    using inequality_op_type = ::rocprim::empty_type;
    using offset_type = unsigned int;
    rocprim::empty_type* const no_values = nullptr; // key only

    return detail::partition_impl<detail::select_method::flag, false, Config, offset_type>(
        temporary_storage, storage_size, input, no_values, flags, output, no_values, selected_count_output,
        size, inequality_op_type(), stream, debug_synchronous, unary_predicate_type()
    );
}

/// \brief Parallel select primitive for device level using selection predicate.
///
/// Performs a device-wide partition using selection predicate. Partition copies
/// the values from \p input to \p output  in such a way that all values for which
/// the \p predicate returns \p true precede the elements for which it returns \p false.
///
/// \par Overview
/// * Returns the required size of \p temporary_storage in \p storage_size
/// if \p temporary_storage in a null pointer.
/// * Ranges specified by \p input, \p flags and \p output must have at least \p size elements.
/// * Range specified by \p selected_count_output must have at least 1 element.
/// * Relative order is preserved for the elements for which the \p predicate returns \p true. Other
/// elements are copied in reverse order.
///
/// \tparam Config - [optional] configuration of the primitive. It can be \p select_config or
/// a custom class with the same members.
/// \tparam InputIterator - random-access iterator type of the input range. It can be
/// a simple pointer type.
/// \tparam OutputIterator - random-access iterator type of the output range. It can be
/// a simple pointer type.
/// \tparam SelectedCountOutputIterator - random-access iterator type of the selected_count_output
/// value. It can be a simple pointer type.
/// \tparam UnaryPredicate - type of a unary selection predicate.
///
/// \param [in] temporary_storage - pointer to a device-accessible temporary storage. When
/// a null pointer is passed, the required allocation size (in bytes) is written to
/// \p storage_size and function returns without performing the select operation.
/// \param [in,out] storage_size - reference to a size (in bytes) of \p temporary_storage.
/// \param [in] input - iterator to the first element in the range to select values from.
/// \param [out] output - iterator to the first element in the output range.
/// \param [out] selected_count_output - iterator to the total number of selected values (length of \p output).
/// \param [in] size - number of element in the input range.
/// \param [in] predicate - unary function object which returns /p true if the element should be
/// ordered before other elements.
/// The signature of the function should be equivalent to the following:
/// <tt>bool f(const T &a);</tt>. The signature does not need to have
/// <tt>const &</tt>, but function object must not modify the object passed to it.
/// \param [in] stream - [optional] HIP stream object. The default is \p 0 (default stream).
/// \param [in] debug_synchronous - [optional] If true, synchronization after every kernel
/// launch is forced in order to check for errors. The default value is \p false.
///
/// \par Example
/// \parblock
/// In this example a device-level partition operation is performed on an array of
/// integer values, even values are copied before odd values.
///
/// \code{.cpp}
/// #include <rocprim/rocprim.hpp>///
///
/// auto predicate =
///     [] __device__ (int a) -> bool
///     {
///         return (a%2) == 0;
///     };
///
/// // Prepare input and output (declare pointers, allocate device memory etc.)
/// size_t input_size;     // e.g., 8
/// int * input;           // e.g., [1, 2, 3, 4, 5, 6, 7, 8]
/// int * output;          // empty array of 8 elements
/// size_t * output_count; // empty array of 1 element
///
/// size_t temporary_storage_size_bytes;
/// void * temporary_storage_ptr = nullptr;
/// // Get required size of the temporary storage
/// rocprim::partition(
///     temporary_storage_ptr, temporary_storage_size_bytes,
///     input,
///     output, output_count,
///     input_size,
///     predicate
/// );
///
/// // allocate temporary storage
/// cudaMalloc(&temporary_storage_ptr, temporary_storage_size_bytes);
///
/// // perform partition
/// rocprim::partition(
///     temporary_storage_ptr, temporary_storage_size_bytes,
///     input,
///     output, output_count,
///     input_size,
///     predicate
/// );
/// // output: [2, 4, 6, 8, 7, 5, 3, 1]
/// // output_count: 4
/// \endcode
/// \endparblock
template<
    class Config = default_config,
    class InputIterator,
    class OutputIterator,
    class SelectedCountOutputIterator,
    class UnaryPredicate
>
inline
cudaError_t partition(void * temporary_storage,
                     size_t& storage_size,
                     InputIterator input,
                     OutputIterator output,
                     SelectedCountOutputIterator selected_count_output,
                     const size_t size,
                     UnaryPredicate predicate,
                     const cudaStream_t stream = 0,
                     const bool debug_synchronous = false)
{
    // Dummy flag type
    using flag_type = ::rocprim::empty_type;
    flag_type * flags = nullptr;
    // Dummy inequality operation
    using inequality_op_type = ::rocprim::empty_type;
    using offset_type = unsigned int;
    rocprim::empty_type* const no_values = nullptr; // key only

    return detail::partition_impl<detail::select_method::predicate, false, Config, offset_type>(
        temporary_storage, storage_size, input, no_values, flags, output, no_values, selected_count_output,
        size, inequality_op_type(), stream, debug_synchronous, predicate
    );
}

/// \brief Parallel select primitive for device level using two selection predicates.
///
/// Performs a device-wide three-way partition using two selection predicates. Partition copies
/// the values from \p input to either \p output_first_part or \p output_second_part or
/// \p output_unselected according to the following criteria:
/// The value is copied to \p output_first_part if the predicate \p select_first_part_op invoked
/// with the value returns \p true. It is copied to \p output_second_part if \p select_first_part_op
/// returns \p false and \p select_second_part_op returns \p true, and it is copied to
/// \p output_unselected otherwise.
///
/// \par Overview
/// * Returns the required size of \p temporary_storage in \p storage_size
/// if \p temporary_storage is a null pointer.
/// * Range specified by \p selected_count_output must have at least 2 elements.
/// * Relative order is preserved for the elements.
/// * The number of elements written to \p output_first_part is equal to the number of elements
/// in the input for which \p select_first_part_op returned \p true.
/// * The number of elements written to \p output_second_part is equal to the number of elements
/// in the input for which \p select_first_part_op returned \p false and \p select_second_part_op
/// returned \p true.
/// * The number of elements written to \p output_unselected is equal to the number of input elements
/// minus the number of elements written to \p output_first_part minus the number of elements written
/// to \p output_second_part.
///
/// \tparam Config - [optional] configuration of the primitive. It can be \p select_config or
/// a custom class with the same members.
/// \tparam InputIterator - random-access iterator type of the input range. It can be
/// a simple pointer type.
/// \tparam FirstOutputIterator - random-access iterator type of the first output range. It can be
/// a simple pointer type.
/// \tparam SecondOutputIterator - random-access iterator type of the second output range. It can be
/// a simple pointer type.
/// \tparam UnselectedOutputIterator - random-access iterator type of the unselected output range.
/// It can be a simple pointer type.
/// \tparam SelectedCountOutputIterator - random-access iterator type of the selected_count_output
/// value. It can be a simple pointer type.
/// \tparam FirstUnaryPredicate - type of the first unary selection predicate.
/// \tparam SecondUnaryPredicate - type of the second unary selection predicate.
///
/// \param [in] temporary_storage - pointer to a device-accessible temporary storage. When
/// a null pointer is passed, the required allocation size (in bytes) is written to
/// \p storage_size and function returns without performing the select operation.
/// \param [in,out] storage_size - reference to a size (in bytes) of \p temporary_storage.
/// \param [in] input - iterator to the first element in the range to select values from.
/// \param [out] output_first_part - iterator to the first element in the first output range.
/// \param [out] output_second_part - iterator to the first element in the second output range.
/// \param [out] output_unselected - iterator to the first element in the unselected output range.
/// \param [out] selected_count_output - iterator to the total number of selected values in
/// \p output_first_part and \p output_second_part respectively.
/// \param [in] size - number of element in the input range.
/// \param [in] select_first_part_op - unary function object which returns \p true if the element
/// should be in \p output_first_part range
/// The signature of the function should be equivalent to the following:
/// <tt>bool f(const T &a);</tt>. The signature does not need to have
/// <tt>const &</tt>, but function object must not modify the object passed to it.
/// \param [in] select_second_part_op - unary function object which returns \p true if the element
/// should be in \p output_second_part range (given that \p select_first_part_op returned \p false)
/// The signature of the function should be equivalent to the following:
/// <tt>bool f(const T &a);</tt>. The signature does not need to have
/// <tt>const &</tt>, but function object must not modify the object passed to it.
/// \param [in] stream - [optional] HIP stream object. The default is \p 0 (default stream).
/// \param [in] debug_synchronous - [optional] If true, synchronization after every kernel
/// launch is forced in order to check for errors. The default value is \p false.
///
/// \par Example
/// \parblock
/// In this example a device-level three-way partition operation is performed on an array of
/// integer values, even values are copied to the first partition, odd and 3-divisible values
/// are copied to the second partition, and the rest of the values are copied to the
/// unselected partition
///
/// \code{.cpp}
/// #include <rocprim/rocprim.hpp>
///
/// auto first_predicate =
///     [] __device__ (int a) -> bool
///     {
///         return (a%2) == 0;
///     };
/// auto second_predicate =
///     [] __device__ (int a) -> bool
///     {
///         return (a%3) == 0;
///     };
///
/// // Prepare input and output (declare pointers, allocate device memory etc.)
/// size_t input_size;          // e.g., 8
/// int * input;                // e.g., [1, 2, 3, 4, 5, 6, 7, 8]
/// int * output_first_part;    // array of 8 elements
/// int * output_second_part;   // array of 8 elements
/// int * output_unselected;    // array of 8 elements
/// size_t * output_count;      // array of 2 elements
///
/// size_t temporary_storage_size_bytes;
/// void * temporary_storage_ptr = nullptr;
/// // Get required size of the temporary storage
/// rocprim::partition_three_way(
///     temporary_storage_ptr, temporary_storage_size_bytes,
///     input,
///     output_first_part, output_second_part, output_unselected,
///     output_count,
///     input_size,
///     first_predicate,
///     second_predicate
/// );
///
/// // allocate temporary storage
/// cudaMalloc(&temporary_storage_ptr, temporary_storage_size_bytes);
///
/// // perform partition
/// rocprim::partition_three_way(
///     temporary_storage_ptr, temporary_storage_size_bytes,
///     input,
///     output_first_part, output_second_part, output_unselected,
///     output_count,
///     input_size,
///     first_predicate,
///     second_predicate
/// );
/// // elements denoted by '*' were not modified
/// // output_first_part:  [2, 4, 6, 8, *, *, *, *]
/// // output_second_part: [3, *, *, *, *, *, *, *]
/// // output_unselected:  [1, 5, 7, *, *, *, *, *]
/// // output_count:       [4, 1]
/// \endcode
/// \endparblock
template <
    class Config = default_config,
    typename InputIterator,
    typename FirstOutputIterator,
    typename SecondOutputIterator,
    typename UnselectedOutputIterator,
    typename SelectedCountOutputIterator,
    typename FirstUnaryPredicate,
    typename SecondUnaryPredicate>
inline
cudaError_t partition_three_way(void * temporary_storage,
                               size_t& storage_size,
                               InputIterator input,
                               FirstOutputIterator output_first_part,
                               SecondOutputIterator output_second_part,
                               UnselectedOutputIterator output_unselected,
                               SelectedCountOutputIterator selected_count_output,
                               const size_t size,
                               FirstUnaryPredicate select_first_part_op,
                               SecondUnaryPredicate select_second_part_op,
                               const cudaStream_t stream = 0,
                               const bool debug_synchronous = false)
{
    // Dummy flag type
    using flag_type = ::rocprim::empty_type;
    flag_type * flags = nullptr;
    // Dummy inequality operation
    using inequality_op_type = ::rocprim::empty_type;
    using offset_type = uint2;
    using output_key_iterator_tuple = tuple<
        FirstOutputIterator,
        SecondOutputIterator,
        UnselectedOutputIterator>;
    using output_value_iterator_tuple
        = tuple<::rocprim::empty_type*, ::rocprim::empty_type*, ::rocprim::empty_type*>;
    rocprim::empty_type* const no_input_values = nullptr; // key only
    const output_value_iterator_tuple no_output_values {nullptr, nullptr, nullptr}; // key only

    output_key_iterator_tuple output{ output_first_part, output_second_part, output_unselected };

    return detail::partition_impl<detail::select_method::predicate, false, Config, offset_type>(
        temporary_storage, storage_size, input, no_input_values, flags, output, no_output_values, selected_count_output,
        size, inequality_op_type(), stream, debug_synchronous,
        select_first_part_op, select_second_part_op
    );
}

/// @}
// end of group devicemodule

END_ROCPRIM_NAMESPACE

#endif // ROCPRIM_DEVICE_DEVICE_PARTITION_HPP_